Wide Band-Gap Nanowires for Light Emitting Diodes

by Jordan Chesin

B.Sc. Materials Science and Engineering

Brown University, 2009

Submitted to the Department of Materials Science and Engineering In Partial Fulfillment of the Requirements for the Degree of

Doctor of Philosophy

at the Massachusetts Institute of Technology

June 2015

© 2015 Massachusetts Institute of Technology. All rights reserved.

Signature of Author: Department of Materials Science and Engineering

April 9th, 2015

Certified by: Silvija Gradečak

Associate Professor in Materials Science and Engineering Thesis Supervisor

Accepted by: Donald R. Sadoway

Chair, Departmental Committee on Graduate Students

1

2

Wide Band-Gap Nanowires for Light Emitting Diodes

by Jordan Chesin

Submitted to the Department of Materials Science and Engineering on May 22nd, 2015

In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy in Materials Science & Engineering

Wide band-gap nanowires composed of GaN and ZnO are promising materials for unique

designs and potential efficiency improvement of light emitting diodes (LEDs) for solid state lighting. The large surface-to-volume ratio of nanowires provides facile strain-relaxation such that nanowires can be grown on substrates with a large lattice mismatch and remain free of threading dislocations. Specifically, the growth of wide band-gap nanowires directly on Si substrates is a promising platform for the fabrication of wafer-scale nanowire array-based LEDs. While nanowire-based LEDs have been previously demonstrated, there has been no work directly comparing the different potential designs of nanowire-based LEDs addressing how material-specific properties affect the light extraction and internal quantum efficiency (IQE). Furthermore, for scalable fabrication of nanowire array-based LEDs on Si a large degree of control over the nanowire synthesis is necessary, especially with regard to the nanowire length uniformity, vertical alignment relative to the growth substrate and the nanowire areal density.

In this work we directly compare feasible designs for GaN-InGaN nanowire-based LEDs using a combination of photonic simulation and modeling. We compared the directed external quantum efficiency of III-nitride LEDs on silicon based on axial and radial nanowire heterostructures, considering m- and c-directional nanowires. The directed extraction efficiency was calculated using photonic simulations and the IQE was estimated using the A-B-C model. We found that m-directional axial heterostructures have the highest directed extraction efficiency, due to the strong polarization anisotropy of III-nitrides, and display similar IQE as c-directional axial heterostructures. By combining IQE and directed extraction, a range of expected directed external quantum efficiencies (EQEs) reveal that m-directional axial heterostructures have EQEs up to three times that of c-directional axial heterostructures, providing guidelines for the design of future III-nitride nanowire-based LEDs.

While III-nitride nanowires are promising candidates, ZnO is an alternative with a higher exciton binding energy and excellent optical properties. To create a platform for the fabrication of ZnO nanowire array-based LEDs on Si, the growth of ZnO was investigated primarily using ZnO solution-processed seed-layers in vapor transport and condensation growth at high temperatures. Due to dependency of the carbothermal reduction of ZnO powder, which acts as the precursor source in the growth, the nanowire areal density was dependent on O2 flow. At low nanowire areal density, growth proceeded in a regime in which continuous nucleation of nanowires occurred throughout the growth, resulting in nanowires with a fixed aspect ratio, but widely varying lengths. At higher nanowire areal densities, the nanowires competed for source precursors in a surface-diffusion limited regime of growth in which the growth rate was dependent upon the nanowire diameter. We observed a critical nucleation diameter for nanowires in the continuous-nucleation regime, which was higher at lower oxygen flow rates. Thus, to achieve length uniformity we developed a two-stage growth method in which nanowires are nucleated at low oxygen flow in the continuous nucleation regime to set the nanowire diameter.

3

In the second stage of growth, where conditions were shifted to the surface-diffusion limited regime, the large diameters set by the first stage of growth were designed to be in the range at which the growth rate does not vary substantially with diameter. The concept of this approach was extended to include control over the nanowire areal density, using sparse ZnO seed-layers. These ZnO nanowires retain excellent optical properties and we observed both demonstrative p-type and n-type doping, dependent on processing conditions, using individual nanowire electrical characterization. Thus, by achieving ZnO nanowire arrays with controlled nanowire areal density, excellent length uniformity and vertical alignment relative to the substrate, we have demonstrated a promising platform for the fabrication of scalable ZnO nanowire array-based LEDs.

Thesis Supervisor: Silvija Gradečak Title: Associate Professor of Materials Science & Engineering

4

Acknowledgements

The work in this thesis would not have been possible without the tremendous support of those in the Gradečak research group, the amazing faculty and staff at MIT and the unrelenting support of my family and friends.

First of all, I would like to thank Professor Silvija Gradečak or her guidance and support throughout my time at MIT. Her approach to research problems and the communication of research ideas has been an invaluable resource even as the research itself seemed daunting. It has been amazing to be a part of her group as she advanced so much through her career, gaining tenure and even welcome her son Stribor into the world. The spirit in her research group that she has cultivated is a large part of what made my time at MIT so enjoyable. I would also like to thank Professors Lionel Kimerling and Polina Anikeeva for providing excellent feedback and also for being excellent teachers.

The members of the Gradaček group have also been an invaluable resource for both research and friendship. I may not have joined the group if I had not first met with Dr. Matt Smith, who indirectly convinced me to join the group and gave the best advice upon graduation I have heard to date. I owe a great debt of gratitude to Dr. Mike Tambe, who helped me to develop the approach to nanowire electrical characterization presented in Chapter 6 and was always up for a rousing discussion about Lost in its final season. Thanks to Dr. Sung-Keun Lim and Dr. Sam Crawford for teaching me valuable maintenance skills and critical assessment of nanowire growth via CVD and MOCVD as well as providing samples for electrical characterization, with a special thanks to Sam who was willing to eat steak when the entire MIT campus lost power. The work on ZnO nanowires would not have been possible without training on the system by Dr. Megan Brewster and discussions regarding seed-layer preparation with Dr. Sehoon Chan. I first joined the group with my classmates Dr. Xiang Zhou and (soon-to-be Dr.) Eric Jones. Thanks to Xiang for the collaboration on GaN nanowire growth and the book chapter on nanowire lasers. Thanks to Eric for always being available to discuss research ideas, leak check systems and start prank-wars with snowballs from condensation on the liquid nitrogen dewars.

I would also like to thank the younger members in the group for their contributions. I extend special thanks to the best office mates anyone could ask for, Sema Ermez and John Hanson (and honorary office mate Eric Jones), who have also been great friends through occasional rough times. I appreciate the willingness to discuss research and make a fresh pot of coffee whenever productivity lulled. I also extend thanks to Jayce Cheng, Amirezza Kiani, Tina Safaei and Paul Rekemeyer for discussion regarding ZnO nanowire growth, with special thanks to Paul for conducting CL with me and being generally awesome at getting the new EBIC system to MIT, even if I didn’t get a chance to use it. To Zhibo Zhao, Olivia Hentz and Kevin Boegart: I wish you good luck in your research endeavors.

DMSE at MIT has also played a huge supporting role in time here. I thoroughly enjoyed teaching the excellent MIT undergrads mechanical behavior of materials for Professor Krystyn Van Vliet and teaching structure for Professors Jeff Grossman and Chris Schuh. I truly enjoyed writing evilly thoughtful exams with Professor Schuh and I think the students are better for it. Furthermore, Angelita Mireles and Elissa Haverty have been a great source of guidance through the process of qualifying exams, defining a minor and helping to make the department an example of excellence. I would like to thank all the members of the Graduate Materials Council that I served with as well as past and present members who help to make the culture of this department so special.

5

Other faculty members at MIT have also helped advance my research. Thanks to Professors Geoffrey Beach, Vladimir Bulović, Tomas Palacios, Gene Fitzgerald and Michael Rubner for allowing me to use resources in or from their labs. The staff at the shared facilities at MIT has also been exceptional at helping students achieve their research. This work would not have been possible without the flexibility and resiliently positive attitude of Kurt Broderick in the MTL, who was endlessly helpful. I would also like to thank Dr. Shaihn Chen, Dr, Yong Zhang and Patrick Boisvert for many trainings and the upkeep of the electron microscopy facility in CMSE. Some students whom I thank for their help in experiments or using their equipment are Patrick Brown, Joel Jean, Geoffrey Supran, Kunal Mukherjee, Daniel Franke, Uwe Bauer, John Gilbert, Ryan Iutzi and Neil Patel.

Though the environment at MIT was an excellent place for research pursuits, I would not have remained sane without my friends in Boston and Cambridge, or those who frequently visited. Moving here from Providence was greatly smoothed thanks to Jeanine Pollard and Betsy Milarcik, who make great room-mates, even if I had to commute on the green-line. Thanks to Su-Yee Lin, Dan Meltzer, Joseph P. Browne, Siamrut Patanavanich and Brian Spatacco, without whom I may not have persevered through the first semester. Thanks to my awesome room-mates, Patrick Brown and Matthew Pinson whom I may not have seen much during certain times of our research, but who were always there for commiseration and support. Thanks to Ben and Melissa Kline-Struhl for making sure I wasn’t always in lab every weekend and demonstrating that work-life balance can be achieved. I am especially grateful to the friendship of An Son Leong and Danny Farnand who were always ready for an adventure and who have been there for me when I needed them most.

This thesis also would not have been possible without the tremendous support, love and understanding from my wonderful partner, Irene Ros. Her pragmatic approach to life has helped me to make the best of the research results I had to write this thesis, rather than worry about the infinite possible experiments I could try to fit in at the last minute. Her presence, along with Allie McBeagle Ros, has been a great source of calm and clarity ever since we met and I am so lucky that she is in my life.

Finally, I must thank my amazing family for their love and support, even when their own lives are frantic and ever-changing. My sisters Darion, Rachael, Nicolle and Michaela have always been an inspiration, and I cannot wait to support them in their future endeavors as they have supported me. I must give thanks to my step-mother Patricia who never let me settle for average in my life and showed me my own potential as early as elementary school. To my father and best-friend (Guyyy), who has been as proud of me as I am of him, I could not be more grateful for all of the unconditional support and love I’ve received throughout the years.

This thesis is dedicated to my mother, Lisa R. Cummings, who taught me all of the things I

value most through her example and whom I will always hold in my heart.

This work was supported by The Center for Excitonics, an Energy Frontier Research Center

funded by the US Department of Energy, Office of Science, Office of Basic Energy Sciences under Award Number DE-SC0001088. Access to shared facilities in the Center for Materials Science and Engineering was supported in part by the MRSEC program of the NSF under award DMR-0213282.

6

Table of Contents

List of Figures ........................................................................................................................... 9

List of Tables .......................................................................................................................... 12

List of Abbreviations and Acronyms ...................................................................................... 13

Chapter 1: Introduction ........................................................................................................... 14

1.1. Solid state lighting ....................................................................................................... 14

1.2. Wide band-gap materials for LEDs ............................................................................. 16

1.3. Wide band-gap nanowires............................................................................................ 21

1.4. Design and synthesis of wide band-gap nanowire-based LEDs .................................. 24

Chapter 2: Design of Nanowire Light Emitting Diodes ......................................................... 27

2.1. Thin-films .................................................................................................................... 27

2.2. Nanowire-based LED design ....................................................................................... 27

2.2.1. Homojunctions ...................................................................................................... 29

2.2.2. Nanowire-substrate heterojunctions...................................................................... 29

2.2.3. Axial heterostructures ........................................................................................... 29

2.2.4. Radial heterostructures.......................................................................................... 30

2.3. Efficiency considerations ............................................................................................. 32

2.3.1. Light extraction ..................................................................................................... 32

2.3.2. The quantum-confined Stark effect ...................................................................... 34

2.3.3. Other design considerations .................................................................................. 35

2.4. Summary ...................................................................................................................... 36

Chapter 3: Evaluating GaN/InGaN Nanowire-Based LED Designs ...................................... 37

3.1. Finite difference time domain simulations .................................................................. 37

3.1.1. Simulation Setup ................................................................................................... 38

3.1.2. Modal properties of nanowire waveguides ........................................................... 40

7

3.1.3. Point emission sources .......................................................................................... 41

3.1.4. Plane emission sources ......................................................................................... 47

3.2. Modeling the internal quantum efficiency ................................................................... 48

3.3. External quantum efficiencies ...................................................................................... 52

3.4. Further enhancing extraction efficiency ...................................................................... 54

3.5. Summary ...................................................................................................................... 57

3.6. Device outlook ............................................................................................................. 58

Chapter 4: ZnO Nanowire Synthesis on Si ............................................................................. 60

4.1. Nanowire growth mechanisms ..................................................................................... 61

4.1.1. VTC growth with seed-particles ........................................................................... 61

4.1.2. Anisotropic vapor-solid growth ............................................................................ 63

4.2. Hydrothermal synthesis of ZnO nanowires ................................................................. 64

4.2.1. Hydrothermal synthesis of ZnO nanowires on Si ................................................. 64

4.2.2. Hydrothermal synthesis of ZnO nanowires on flexible substrates ....................... 68

4.3. VTC ZnO nanowire synthesis with Au seed-particles ................................................. 69

4.3.1. Nanowire morphology shifts with seed-particle size ............................................ 69

4.3.2. Discussion of using Au for synthesis .................................................................... 73

4.4. VTC ZnO nanowire synthesis on silicon with continuous ZnO seed-layers ............... 75

4.4.1. Density Control via O2 Flow/Seed Layer Thickness ............................................ 77

4.4.2. Continuous nucleation growth regime at low nanowire areal density .................. 81

4.4.3. Surface-diffusion limited growth regime at higher nanowire areal density ......... 86

4.4.4. Two-Step Growth Technique ................................................................................ 94

4.5. VTC ZnO nanowire synthesis on silicon with sparse ZnO seed-layers ....................... 98

4.5.1. Controlling nanowire areal density with sparse seed-layers ................................. 99

4.5.2. Controlling nanowire morphology on sparse seed-layers ................................... 101

8

4.6. Summary .................................................................................................................... 102

Chapter 5: ZnO Nanowire Optical Properties ....................................................................... 104

5.1. Optical properties of ZnO nanowires ......................................................................... 104

5.2. Photoluminescence of ZnO nanowires ...................................................................... 105

5.3. Summary .................................................................................................................... 113

Chapter 6: Towards Scalable Nanowire-Based LEDs on Si ................................................. 114

6.1. Individual nanowire electrical characterization ......................................................... 114

6.2. Electrical properties of ZnO nanowires ..................................................................... 118

6.3. Fabrication methods for nanowire array devices ....................................................... 121

Chapter 7: Conclusions ......................................................................................................... 125

7.1. Summary and implications of this work .................................................................... 125

7.1.1. The ideal GaN nanowire-based LED .................................................................. 129

7.2. Future Work ............................................................................................................... 131

7.2.1. Controllable electrical properties via extrinsic dopants ...................................... 131

7.2.2. Controlled defect luminescence for SH white nanowire-based LEDs................ 131

7.2.3. Development of QW materials and nanowire-based LED design ...................... 132

References ............................................................................................................................. 133

Appendix A: GaN Nanowires - Synthesis and Properties.................................................... 148

A.1. Growth on sapphire................................................................................................. 149

A.2. Growth on silicon.................................................................................................... 151

A.3. Electrical characterization of GaN nanowires........................................................ 153

A.4. Optical properties of GaN nanowires..................................................................... 154

9

List of Figures

Figure 1-1. Summary of LED structures in thin-films and nanowires. ........................................ 17

Figure 1-2. Defect and impurity levels in GaN and ZnO. ............................................................ 20

Figure 1-3. The wurtzite crystal structure of GaN and ZnO. ........................................................ 22

Figure 1-4. Common growth directions of GaN nanowires on various substrates. ...................... 24

Figure 2-1. Selected designs of GaN nanowire-based LEDs. ....................................................... 28

Figure 2-2. Comparison of the active area in nanowire heterostructures and thin-films. ............. 31

Figure 2-3. Schematic of polarization anisotropy in different nanowire-based LED designs. ..... 33

Figure 2-4. The quantum-confined Stark effect. ........................................................................... 34

Figure 3-1. Simulation setup used for FDTD photonic simulations. ............................................ 39

Figure 3-2. The modal properties of hexagonal and triangular nanowire waveguides. ................ 40

Figure 3-3. Directed extraction efficiency dependence on point source position. ........................ 42

Figure 3-4. Fundamental mode profile intensity cross-section for c-directional nanowires. ....... 44

Figure 3-5. Directed extraction dependence on point source position in cross-section. ............... 45

Figure 3-6. Effect of realistic nanowire cross-sections on extraction. .......................................... 47

Figure 3-7. Compared directed extraction for nanowire-based LED designs. ............................. 48

Figure 3-8. Comparing changes in coefficients of the A-B-C model for finding IQE. ................ 50

Figure 3-9. Overall estimated EQE comparison of nanowire axial heterostructures. .................. 54

Figure 3-10. Extraction from square arrays of m-directional axial heterostructures. ................... 55

Figure 3-11. Distributed Bragg reflector via diameter modulations. ............................................ 57

Figure 4-1. ZnO nanowires synthesis methods. ............................................................................ 61

Figure 4-2. VLS growth mechanism schematic. ........................................................................... 62

Figure 4-3. Hydrothermally synthesized ZnO nanowires. ............................................................ 65

Figure 4-4. Nanowire diameter distribution of HTS-grown ZnO nanowires. .............................. 66

Figure 4-5. HTS ZnO nanowire length at varying growth times. ................................................. 67

Figure 4-6. SEM of HTS ZnO nanowires on PEN. ...................................................................... 69

Figure 4-7. ZnO Nanowire diameters as a function Au film thickness on a-sapphire. ................ 71

Figure 4-8. Nanowire areal density decrease with increasing Au-film thickness. ........................ 72

Figure 4-9. Large Au particles at substrate in ZnO nanowire growth. ......................................... 73

Figure 4-10. Growth attempts of ZnO nanowires on Si with Au seeds. ....................................... 74

10

Figure 4-11. Demonstration of shifting reactor conditions. .......................................................... 76

Figure 4-12. The role of oxidation in reproducibility. .................................................................. 76

Figure 4-13. Ellingham diagram of relevant reactions in ZnO reduction. .................................... 78

Figure 4-14. Nanowire areal density dependence on seed-layer thickness and O2 flow. ............. 79

Figure 4-15. Fine-tuning nanowire areal density with O2 flow. ................................................... 80

Figure 4-16. Linear dependence of nanowire length on nanowire diameter. ............................... 82

Figure 4-17. Evolution of nanowire and seed-layer morphology with growth time. ................... 82

Figure 4-18. Pyramidal faceting observed at ZnO nanowire bases. ............................................. 83

Figure 4-19. Diameter of pyramidal c-facet and corresponding nanowire length. ....................... 84

Figure 4-20. Continuous nucleation growth regime. .................................................................... 85

Figure 4-21. Surface-diffusion limited growth model schematic. ................................................ 87

Figure 4-22. Surface-diffusion limited growth of ZnO nanowires. .............................................. 88

Figure 4-23. Temperature dependence of surface diffusion length and source flux. ................... 90

Figure 4-24. Growth regime summary for VTC ZnO nanowires on continuous seed-layers. ...... 92

Figure 4-25. Comparing low pressure and atmospheric growth. .................................................. 94

Figure 4-26. Growth parameters used for the two stage growth................................................... 95

Figure 4-27. Two-step growth process for nanowire arrays with uniform lengths. ..................... 97

Figure 4-28. Fine-tuning the nanowire areal density with sparse seed-layers. ........................... 100

Figure 4-29. Average diameter and length of nanowires on sparse seed-layers. ........................ 101

Figure 5-1. Comparison of luminescence efficiency of GaN and ZnO nanowires. .................... 106

Figure 5-2. PL spectra of hydrothermally synthesized ZnO on Si and PEN. ............................. 107

Figure 5-3. PL spectra of ZnO nanowires grown with increasing oxygen flow. ........................ 109

Figure 5-4. Annealing-induced green-luminescence in ZnO nanowires. ................................... 111

Figure 5-5. CL of an annealed ZnO nanowire demonstrating bulk defect luminescence. .......... 112

Figure 5-6. PL spectra of ZnO nanowires on sparse seed-layers. ............................................... 113

Figure 6-1. Method for individual nanowire electrical characterization. ................................... 115

Figure 6-2. Electrical properties of individual ZnO nanowires. ................................................. 120

Figure 6-3. Electrical properties of ZnO nanowires after H2 anneal. ......................................... 121

Figure 6-4. PECVD SiOx as a dielectric filler material. ............................................................. 123

Figure 7-1. Schematic of the ideal nanowire-based LED on Si. ................................................. 130

Figure A-1. (MO)CVD system setup used for the growth of GaN nanowires. ......................... 149

11

Figure A-2. m-directional GaN nanowires on r-plane sapphire................................................. 150

Figure A-3. Six-fold symmetry of GaN nanowires grown on c-plane sapphire......................... 151

Figure A-4. Growth of GaN on Si substrates with different pre-growth treatments.................. 152

Figure A-5. TEM of m-directional GaN nanowire grown on bare Si. ....................................... 152

Figure A-6. GaN nanowires on bare Si at different V/III ratios. ............................................... 153

Figure A-7. Individual nanowire electrical characterization of GaN nanowires. ...................... 154

12

List of Tables

Table 1-1. Threading dislocations in GaN and ZnO thin-films via different methods. ................ 19

Table 3-1. Coefficients of the A-B-C model used in modeling IQE. ........................................... 52

Table 5-1. Summary of nanowire morphologies compared in Figure 5-3. ................................. 109

13

List of Abbreviations and Acronyms

CB conduction band CL cathodoluminescence CRI color rendering index CVD chemical vapor deposition DH double heterojunction DI deionized EQE external quantum efficiency FDTD finite difference time domain HF hydrofluoric acid HTS hydrothermal synthesis IQE internal quantum efficiency IR infrared LED light emitting diode MBE molecular beam epitaxy MOCVD metal-organic chemical vapor deposition MQW multiple quantum-well PAMBE plasma-assisted molecular beam epitaxy PECVD plasma-enhanced chemical vapor deposition PMMA poly-methyl methacrylate PML perfectly matched layer QCSE quantum-confined Stark effect QW quantum-well RIE reactive ion etching SEBL scanning electron beam lithography SEM scanning electron microscopy SH single heterojunction STEM scanning transmission electron microscopy TD threading dislocation TEM transmission electron microscopy UV ultraviolet VB valence band VLS vapor-liquid-solid VS vapor-solid VTC vapor transport and condensation WPE wall plug efficiency

14

Chapter 1: Introduction

1.1. Solid state lighting

Solid state lighting has the potential to significantly improve the energy efficiency of

lighting, providing conservation of energy resources and helping to reduce energy consumption

to a sustainable level. In the United States, lighting comprises at least 14% of commercial and

residential electricity usage[1]. As of 2008, by globally replacing current lighting sources with

solid state lighting it would be possible to save 1.9×1020 J over 10 years, equivalent to 1.83

trillion dollars, 10.7×109 tons of CO2 emissions or 962×106 barrels of oil[2]. Solid state lighting

takes advantage of light emitting diodes (LEDs) to generate light directly from electrical current,

via electroluminescence, which can be significantly more efficient than the currently

implemented technologies, with wall-plug efficiencies (WPE) of up to 70% as compared to

incandescent (WPE ≈ 5%) and compact fluorescent lighting (WPE ≈ 20%)[3].

A common approach to achieve white light is to combine a blue LED with a yellow

phosphor, while the approach for the best color rendering index (CRI) uses three separate LEDs

to emit red, green and blue light[3]. Thus, it is critically important to develop high-efficiency

blue emitters based on wide band-gap materials for LEDs, such as gallium nitride (GaN) or zinc

oxide (ZnO). GaN is commonly used, as control over both n-type and p-type doping has been

demonstrated reliably on a manufacturing scale using methods common in industrial processing

such as chemical vapor deposition (CVD), metal-organic CVD (MOCVD), molecular beam

epitaxy (MBE) or plasma-assisted MBE (PAMBE)[4]. Furthermore, the indium gallium nitride

(InxGa1-xN) material system can be tuned to emit from the infrared (the bandgap of InN is 0.7

eV), through the visible spectrum, and into the ultraviolet (the bandgap of GaN is 3.4 eV), by

adjusting the alloy indium composition[5]. Thus, it may be possible to design white LEDs using

a single material system with different compositions using the InxGa1-xN material system.

In comparison, ZnO is relatively less developed for commercial use. Traditionally, p-type

doping in ZnO has been unreliable and difficult to consistently produce on a large scale[6];

however, recently there have been many reports of p-type doping in ZnO, especially in

nanowires[7-10]. There are only limited reports of tuning the emission wavelength of ZnO-

compatible materials systems, mostly deeper into the UV or just slightly into the blue regime of

15

the spectra, such as Al-doped ZnO, which can emit from the band-gap of ZnO (3.37 eV) up to

3.6 eV due to a Burnstein-Moss Fermi-band filling effect[11], using MgxZn1-xO for emission

from 3.37 eV up to 4.55 eV for Mg0.51Zn0.49O[12] or doping ZnO with Cu to introduce d-orbitals

and decrease the band-gap blue emission near 2.66 eV[13]. Alternatively, alloying ZnO with

SnO2 or Ag can create materials with both blue and yellow emission peaks for white-light from

individual nanowires[14, 15]. One advantage of ZnO is its relatively large excitonic binding

energy of 60 meV, compared to 25 meV in GaN, which suggests that ZnO materials may have

brighter emission[16]. Furthermore, Zn is a commodity costing only about $2/kg as compared to

Ga which costs $511/kg (USD as of 2013 year-end)[17]. Thus, while GaN is the more developed

materials system for solid-state lighting, ZnO is a promising alternative.

Current GaN-based LEDs and ZnO films grown on sapphire typically have a high threading

dislocation density (around 108 /cm2), which cause non-radiative recombination and thus

decreased efficiency[3]. It has recently been suggested that the threading dislocation density

must be reduced below 106 /cm2 to achieve 100% internal quantum efficiency[18]. Such low

threading dislocation densities have been achieved using epitaxial lateral overgrowth[19, 20];

however, this approach requires an extra patterning and growth step. One of the most limiting

aspects of current GaN-based LEDs is that they must be grown on expensive and insulating

substrates, such as sapphire (Al2O3) or silicon carbide (SiC) to achieve low dislocation

densities[19]. The growth of ZnO thin-films is less developed than GaN thin-films, but uses

many of the same approaches, such as epitaxial growth on sapphire or the use of buffer

layers[21, 22] and can even be fabricated via hydrothermal synthesis (HTS) at low

temperatures[23]; however, the largest challenge is controlling the point defect density[24].

Nanowires, one-dimensional crystals that are tens of nm thick and microns long have the

potential to increase LED efficiencies while decreasing manufacturing costs. GaN nanowires

have been shown to be dislocation free[25, 26] and other III-V nanowires have been shown to be

grown on largely mismatched substrates, such as silicon[21, 22]. ZnO nanowires often grow via

the anisotropic nature of the wurtzite crystal system and have been demonstrated on a variety of

substrates, including silicon, with vertical alignment due to highly oriented seed-layers via

solution-based[27, 28] and deposition-based synthesis[29, 30]. The capability to grow nanowires

on substrates with large lattice mismatch is generally attributed to strain-relaxation due to the

large surface-area to volume ratio and the nanoscale interface. Not only do nanowires provide a

16

route to high quality materials on low-cost substrates, but many of their properties can be

advantageous for use in optoelectronics, especially LEDs. Nanowires fabricated from bottom-up

approaches, as discussed in this work, tend to have smooth side-wall facets and cross-sections

that are smaller than or close to the wavelength of light they emit, thus they can act as excellent

waveguides for photonics applications or for directing emitted light towards extraction in

LEDs[31-33]. The arrangement of vertical nanowires on a substrate can also be used to create a

photonic crystal, limiting the propagation direction of emitted light for higher extraction

efficiency in LEDs[34]. The high aspect ratio of the nanowires can act as an excellent optical

cavity for low-threshold lasers[35]. Nanowire-based devices also have unique opportunities for

device design based on their morphologies that can enhance efficiency or brightness. By

providing dislocation-free material on inexpensive and conducting substrates such as silicon and

many unique properties advantageous for optoelectronics, nanowire-based LEDs are an exciting

prospect for both enhancing efficiency and decreasing the cost of solid state lighting.

1.2. Wide band-gap materials for LEDs

Thin-film LEDs and nanowire-LEDs both operate based on electroluminescence from a

diode. The simplest form of an LED is a p-n homojunction. Emission from homojunction devices

has been demonstrated in both thin-film[36] and nanowire homojunctions[37] (see Figure

1-1(a)); however, carriers recombine in the depletion region and within their diffusion lengths

(Ln and Lp) from the junction, forming a wide active region for emission that is not efficient for

radiative recombination[4]. A similarly simple LED is based on heterojunction diodes. Of

particular interest for simple large scale nanowire LEDs is the type II heterojunction diode, for

example n-type nanowires of high crystalline quality can be grown directly on inexpensive p-

type substrates (i.e. – n-GaN or n-ZnO nanowires on p-Si). Under forward bias this sort of

junction can operate similarly to a homojunction to emit light. Some early thin-film LEDs were

based on this type of heterojunction[38], known as a single heterojunctions (SH). Similarly,

crossed-nanowire heterostructures have been demonstrated[39] (see Figure 1-1(b)).

The most efficient and thus most common type of LED is based on a double heterojunction

(DH), in which a material of a narrow band gap is sandwiched between a p-type and n-type

regions of a wider band gap semiconductor (see Figure 1-1(c)), forming symmetrical type I

junctions (in which the narrower gap falls completely within the wider gap). The primary

17

advantage of the DH is to spatially confine both electrons and holes in the narrow band gap

material under forward bias to promote radiative recombination and to waveguide the emission.

Multi-quantum-well (MQW) DH LEDs, both in thin-film[40] and nanowire-based devices[41]

have both been demonstrated.

Figure 1-1. Summary of LED structures in thin-films and nanowires. The schematic

demonstrates the operating principle of (a) homojunction, (b) single-heterojunction and (c)

double-heterojunction LEDs by comparing (i) the equilibrium band-diagram to (ii) the forward-

bias band-diagram and their physical structures in (iii) thin-film devices and (iv) nanowires.

18

The use of nanowires as a dislocation-free material is one of the key reasons for the pursuit

of nanowire-based LEDs. The effects of extended defects are still not fully understood in GaN-

based LEDs or in ZnO films. In most compound semiconductors used for optoelectronics, such

as GaAs and GaP, threading dislocations act as non-radiative recombination centers and

drastically reduce the efficiency of LEDs[42, 43]. For example, it has been shown that a

threading dislocation (TD) density of 106 cm-2 in GaP completely quenches emission[42]. In

contrast, a TD density of 106 cm-2 is actually quite low in GaN LEDs, where TD densities above

108 cm-2 are common, due to the lattice mismatch between GaN and sapphire or SiC[44, 45] and

high-brightness GaN LEDs have been demonstrated at TD densities above 1010 cm-2[46]. Bright

electroluminescence from ZnO has been also been observed at TD densities of 1010 cm-2[47]. It

has been suggested that dislocations act as non-radiative recombination centers in GaN[48],

however dislocations in both GaN[49, 50] and ZnO[51] were also found to repulse carriers.

Interestingly, dislocations induced via plastic deformation have been shown to increase

luminescence in ZnO[51] and high TD densities have been shown to limit the carrier

mobility[47]. Regardless of the mechanism with which dislocations effect the efficiency in GaN

and ZnO LEDs, it has been shown that a high TD density attributes to GaN-LED droop[52] and

it has been suggested that a TD density of less than 106 cm-2 is necessary for 100% internal

quantum efficiency[18], therefore great strides have been made in the growth of thin-film GaN

and ZnO to reduce the TD density, as summarized in Table 1-1.

Table 1-1 exemplifies the difficulties in achieving high-quality GaN and ZnO thin-film

material. Of these examples, only epitaxial lateral overgrowth (ELOG) of GaN on sapphire

substrates can achieve TD densities of less than 106 cm-2, with all strategies for thin-film growth

on silicon resulting in high dislocation densities (above 109 cm-2). While epitaxial growth of ZnO

has not yet yielded films with TD densities less than 106 cm-2, bulk single-crystalline ZnO has

been fabricated with dislocation densities on the order of 104-105 cm-2; however, these bulk

crystals still suffer from poor control over point defects and contain many impurities[53].

Alternatively, GaN and ZnO nanowires are typically free of extended defects and can be grown

on a variety of substrates[25, 54]. Thus, GaN nanowire-based LEDs should be able to

circumvent most of the issues associated with dislocations in GaN thin-film LEDs. Point defects

on the other hand affect the properties, especially optical emission and electrical conductivity, of

both GaN and ZnO thin-films [55-57] and nanowires [58-61]. It must also be noted that surface

19

states can play a larger role in nanowires, such as the size-dependent optical properties observed

in ZnO nanowires[62].

Table 1-1. Threading dislocations in GaN and ZnO thin-films via different methods.

Material Technique Substrate Buffer Layer TD Density (cm-2

)

GaN MOCVD[20] c-sapphire - 2×109

GaN MOCVD with substrate

patterning[20] c-sapphire - 2×108

GaN MOCVD with ELOG and

substrate patterning[20] c-sapphire - 6×105

GaN MOCVD[63] c-sapphire AlN 4×107

GaN MOCVD[64] Si (111) SiNX 9×109

GaN MOCVD[65] Si (111) AlN ≈1012

ZnO HTS with ELOG[66] MgAl2O4 (111) - 2×108

ZnO PAMBE[67] c-sapphire - 3×109

ZnO MBE[68] Si (111) Lu2O3 3×1010

The control of intrinsic point defects and impurities is critical to tuning the electrical and

optical properties in GaN and ZnO films[57, 69] and nanowires[55, 59, 70, 71]. If control over

point-defects can be attained, they can be used as dopants or to tune the emission properties of

the nanowires[14]. Both as-grown GaN and ZnO tend to be n-type, likely due to intrinsic point

defects or prevalent impurities acting as shallow donors. In GaN the responsible defect for as-

grown n-type carriers has been attributed to nitrogen vacancies[72], oxygen impurities[73, 74] or

silicon impurities. Silicon is generally used as the n-type dopant in GaN[5], as it has the tendency

to form the donor SiGa. The most common p-type dopant in GaN is MgGa. The activation energy

of Mg acceptors (224 meV), is significantly higher than that of Si shallow donors (30.8

meV)[56] and initially caused p-type doping difficulties due to the formation of Mg-H

complexes[75], which passivates the Mg dopants[69] and necessitates low-energy electron-beam

irradiation or thermal annealing to break apart these complexes and produce Mg acceptors and

H2 gas from the out-diffused H+. Other point defects and impurities can also have dramatic

effects on electrical and optical properties in GaN and are summarized in Figure 1-2(a). Other

20

defects such as di-vacancies (VGaVN), gallium interstitials (Gai), and nitrogen anti-sites (NGa) are

unlikely to form due to high formation energies[56]. Another impurity which is likely to be

introduced during MOCVD growth of GaN is carbon. The role of carbon in GaN is unclear,

though it is likely to act as an acceptor (CN) and it may form levels associated with yellow defect

luminescence common in thin-films[56, 76].

Figure 1-2. Defect and impurity levels in GaN and ZnO. Approximate defect and impurity

levels are presented (to scale) in (a) GaN and (b) ZnO. Compiled from [51, 56, 74, 77, 78].

The role of point defects in ZnO is less clear than in GaN, though there has recently been a

large effort in understanding the role of these defects, especially as they relate to a broad

emission commonly seen centered around green or yellow wavelengths. Though oxygen

vacancies in ZnO have long been suggested as a candidate for shallow donors, recent work has

shown that they are more likely only compensating donors in p-type ZnO[51, 78]. It is likely that

21

hydrogen plays an important role in the n-type conductivity of as-grown ZnO, whether on its

own or as a complex with other defects or impurities[51, 79]. Group III elements are often used

as electron donors when they incorporate on the Zn site: both Al and Ga have been demonstrated

as donors[80, 81]. p-type doping in ZnO is more varied: group I elements such as K and Li

incorporating on a Zn-site[82, 83], group V elements such as N, As and Sb incorporating on an

O-site[7, 84-86] and transition metals such as Ag have been used as acceptors[87]. The energy

levels of defects and selected impurities are summarized in Figure 1-2(b). In general, it is critical

to keep the defect concentrations low in both GaN and ZnO, to allow the more shallow impurity-

dopants to control the electronic properties. Finally, while nanowire-based wide band-gap

materials eliminate the role of dislocations in LED performance, the effects of surface states

must be considered. The surface is especially critical for the electrical properties of nanowires,

where Fermi-level pinning at the nanowire side-wall facets can result in a depletion region which

can completely prevent conduction in nanowires that have small enough diameters[88, 89].

General approaches to passivating surface states are to deposit a shell of higher band-gap

material, such as AlGaN for GaN nanowires[41] or MgZnO for ZnO nanowires[90].

1.3. Wide band-gap nanowires

GaN and ZnO both have a hexagonal wurtzite crystal structure. The crystal structure plays a

critical role in determining the nanowire morphology, surface properties and the crystallographic

relationship to the substrate on which the nanowires are grown. Sapphire substrates often used in

the growth of wide band-gap materials are also hexagonal. The wurtzite crystal structure is polar

in the <0001> c-directions and non-polar in the <1010> m-directions and in the <1120> a-

directions. The polarity induces an electric field due to the stacking of layers of atoms with

different electronegativity and can cause an undesired spontaneous electric field across important

regions of LEDs. Furthermore, the polarity of the c-facet can induce unique growth structures

and the polarity contributes significantly to the high-surface energy of the c-facet, which

accounts for anisotropic growth-rates, especially observed in the case of ZnO[23]. The wurtzite

crystal structure and relevant crystallographic planes and directions in the hexagonal crystal

system are summarized in Figure 1-3.

22

Figure 1-3. The wurtzite crystal structure of GaN and ZnO. Schematics of (a) the wurtzite

crystal structure for GaN or ZnO and important crystallographic (b) directions and (c) planes in

the hexagonal crystal system.

Nanowires can be fabricated with a variety of methods using either top-down approaches

(beginning with bulk materials to create nanoscale features) or from the bottom-up (fabricating

nanomaterials from chemical precursors). Top-down approaches include nanoscale lithography

followed by reactive ion etching[91] or metal-assisted etching[92]. Bottom-up methods include

particle-mediated growth in CVD[93] and MOCVD[25] and “self-induced” GaN nanowire

growth in MBE[94] and similar anisotropic growth methods of ZnO nanowires via solution

methods[95] or vapor-transport and condensation (VTC)[96]. The morphology of nanowires

produced via these various methods depends on the conditions during growth and the underlying

substrate. Nanowires with various cross-sections (square, triangular, hexagonal), amounts of

tapering and vertical alignment relative to the substrate have been observed.

Perhaps the most important aspect of nanowire growth for nanowire-based LED fabrication

is the growth direction, or crystallographic orientation, relative to the substrate. For device

fabrication, especially in the case of more complicated high-efficiency devices where inter-

nanowire spacing can be critical for photonic applications and most of the light is guided along

the nanowire axis, it is desirable to have arrays of vertically oriented nanowires. Nanowire

orientation has been found to be dependent on many factors, including the underlying substrate

orientation, diameter, growth rate and source flows (i.e. V/III ratio for GaN)[93]. Under optimal

growth conditions, however, the growth direction of GaN nanowires has been found to depend

23

primarily on the orientation of the growth substrate[25, 97], where the relation between the

nanowire and substrate is typically epitaxial. Figure 1-4 summarizes the growth direction and

cross-sections of GaN nanowires grown on various substrates, as well as indicating the growth

direction in the wurtzite hexagonal cell. Notably, vertical growth of GaN nanowires in the c-

direction has been demonstrated on c-plane sapphire[97] and Si (111)[98], where nanowires

have a hexagonal cross-section. On a-plane GaN films, triangular cross-sectional GaN nanowires

grow at a 60° incline from the surface in the m-direction[25]. However, it seems that when

epitaxy is not achieved, due to either large mismatch or growth on amorphous surfaces, this m-

directional growth of GaN is energetically preferred[93]. In general, GaN nanowires are grown

with a seed-mediated method, which leave an Au seed-particle at the nanowire tip; however in

some instances of c-directional GaN nanowires in PAMBE, the growth is self-induced and has a

flat top facet with no foreign metal seed at the tip[99].

As opposed to GaN nanowires, ZnO nanowires almost always grow in the c-direction. The

morphology of c-directional ZnO nanowires is very similar to c-directional GaN nanowires,

composed of a hexagonal cross-section with non-polar m-plane side-facets and a flat c-plane top-

facet, typically with no metal seed-particle. Seed-mediated growth of ZnO nanowires has been

shown[100], but generally the growth proceeds via a direct vapor-solid mechanism and the

particles remain at the nanowire bases near the substrate[54]. Occasionally additional faceting is

observed at the nanowire tips with higher order facets forming a hexagonal pyramidal structure

at the nanowire tip[101, 102]. A seed-layer film is often employed or inadvertently grown when

synthesizing ZnO nanowires, which inherently textures in the c-direction[103], likely due the

polarity in the c-axis[104]. This texturing in the c-direction is what generally causes ZnO

nanowires to grow generally vertically on many substrates, including Si (100)[30], Si (111)[105],

c-plane and a-plane sapphire[106] and even amorphous indium-tin oxide[107]. This inherent

texturing of ZnO nanowires that leads to vertical growth on most substrates is an advantage of

ZnO nanowires for LEDs over GaN nanowires, which require epitaxial growth in the c-direction

for vertical alignment; however the ability to grow GaN nanowires in a non-polar direction may

be an advantage for GaN nanowire for LEDs.

24

Figure 1-4. Common growth directions of GaN nanowires on various substrates. The

highlighted growth direction of the nanowire within the hexagonal crystal system, the cross-

section of the nanowire and the alignment of the nanowire relative to the substrate are

schematically pictured for (a) c-directional GaN nanowire growth on c-GaN, c-sapphire or Si and

(b) m-directional GaN nanowire growth on r-sapphire or a-GaN.

1.4. Design and synthesis of wide band-gap nanowire-based LEDs

The interplay of the growth mechanism, electrical properties, optical properties and nanowire

morphology as discussed above are critical in designing and synthesizing nanowire-based LEDs.

To fully consider the design of possible nanowire-based LED architectures the nanowire

synthesis and the effect of nanowire morphology must be considered relative to the orientation of

the active region of the device (the light-emitting area) and the crystallographic direction across

the active region, especially considering the polarity in that direction. All of these factors play a

role in the efficiency of the carrier recombination in the active region as well as the way in which

light is emitted from that region and guided through the nanowire towards emission.

Furthermore, there are strict requirements on the morphology of nanowire arrays to allow for

reliable device fabrication. Ideally, nanowire areal density should be controlled and vertical

alignment of nanowires relative to the substrate should be achieved with length uniformity such

25

that most nanowires can be contacted via conventional fabrication methods. For example, if

nanowires are not aligned (i.e. growing in random directions) relative to the substrate, even with

nanowire length uniformity, the actual height of the nanowire tips would vary tremendously.

Generally, contact to nanowire tips can be made using a spin-on insulator to fill the gaps between

the nanowires and a transparent conducting material to connect the nanowire tips to a metallic

contact. Thus, when processing to contact nanowire tips of the most vertical nanowires relative

to the substrate, only a small percentage of the nanowire tips would actually be contacted. Note

that this fabrication strategy relies on a conductive substrate, such that contact can be made

through the substrate-nanowire interface and at the nanowire tip. Such vertical nanowire arrays

on conductive substrates are the most scalable method for nanowire-based devices, but there are

still many challenges to address.

The first challenge addressed in this work is determining the optimal nanowire design for

such large-scale nanowire-based LEDs grown directly on Si. In Chapter 2:

26

Design of Nanowire Light Emitting Diodes, the critical aspects of GaN nanowire-based

LED design are discussed. To further elucidate the properties of various nanowire-based LED

designs, photonic simulation and modeling are applied to discuss the trade-offs between possible

designs and evaluate the overall efficiencies. The results of the design evaluation are discussed in

Chapter 3: Evaluating GaN/InGaN Nanowire-Based LED Designs. Upon pursuit of

synthesizing the designs evaluated in this work, the structures could not be adequately

synthesized with the resources available for, thus the alternative design of ZnO-nanowire based

LEDs was pursued. Controlled synthesis of ZnO nanowires on Si was achieved via an in-depth

understanding of the growth-mechanisms, as reported in Chapter 4: ZnO Nanowire Synthesis

on Si. The ZnO nanowire arrays were synthesized with uniform nanowire lengths and the ability

to tune the nanowire areal density. The excellent optical properties with high-brightness emission

in the UV and demonstrated control over bright luminescence in the green make these nanowire-

arrays excellent candidates for the fabrication of nanowire-based LEDs, as shown in Chapter 5:

ZnO Nanowire Optical Properties. Additionally, methods were developed to characterize the

electrical properties of individual nanowires to evaluate control over doping and fabrication

methods to provide a direct path towards the fabrication of large-scale nanowire-based LEDs,

which are discussed in Chapter 6: Towards Scalable Nanowire-Based LEDs on Si. Finally the

contributions of this work are related to the progress in the field and the future directions for this

research are discussed in Chapter 7: Conclusions.

27

Chapter 2: Design of Nanowire Light Emitting Diodes

This chapter discusses the different designs of nanowire-based LEDs and how they may

affect the light extraction efficiency and internal quantum efficiency of the device. The design of

nanowire-based LEDs is critical for high-efficiency devices and control over the emission color

and distribution of emission angles. The interplay of the nanowire crystallography and the

electrical and optical properties of the device’s active region require careful consideration of

many aspects, such as the placement and crystallographic orientation of the QW and the size and

morphology of the nanowire. This chapter focuses on GaN nanowires with InGaN nanowires to

discuss the relevant aspects of nanowire-based LED design to overall efficiency and identifies

some key questions that are explored in Chapter 3: Evaluating GaN/InGaN Nanowire-Based

LED Designs.

2.1. Thin-films

The extraction efficiency of GaN thin-films is generally limited by total internal reflection

due to the large index of refraction contrast between GaN and air. Only light within an escape

cone (22° for GaN) is extracted for a thin-film with an unmodified planar surface. For isotropic

emission and considering emission toward the substrate lost, this would result in an extraction

efficiency of only 0.06. In contrast, nanowires confine light into waveguide modes, guiding

emission towards extraction. In the regime in which the nanowire supports waveguide modes

with effective indices (neff) greater than the superstrate refractive index, the electric-field

intensity of the light is primarily confined within the nanowire. If the cross-sectional area is too

small, the nanowire will still guide the fundamental mode; however, most of the electric-field

intensity of such a “leaky mode” will be outside of the nanowire and the extraction efficiency

will depend heavily on its surroundings, typically other nanowires.

2.2. Nanowire-based LED design

There are many potential designs for GaN nanowire-based LEDs and some of the critical

parameters include nanowire growth direction and geometry of the active quantum well (QW)

region. The cross-sectional shape and growth direction will change the nanowire waveguiding

properties and the polarization anisotropy observed in III-nitrides will affect the efficiency of

coupling to these modes[108, 109]. Furthermore, different surface facets can affect the

28

magnitude of surface recombination and the efficacy of contacts[110, 111]. The choice the QW

orientation within the nanowire relative to the growth direction determines the extent of the

quantum-confined Stark effect (QCSE): an electric-field can shift apart the electron and hole

wave functions, resulting in a reduction in spontaneous emission efficiency and a redshift of the

emission[112]. Since the electric fields are caused by the crystal polarity and piezoelectric effect,

QCSE is particularly observed in the case of the polar c-plane QWs.

Demonstration-scale nanowire-based LEDs have been fabricated, typically as c-

directional radial[113] and axial heterostructures[114, 115] or m-direction radial

heterostructures[116, 117]; however there has been no direct comparison of these designs to

guide future work towards high-efficiency nanowire-based LEDs. Here we discuss the relative

tradeoffs between different designs: axial and radial homojunctions, nanowire-substrate

heterojunctions, c-directional and m-directional radial heterostructures, and finally c-directional

axial heterostructures and m-directional axial heterostructures. These designs are described in

detail below and the m-directional axial heterostructure, c-directional radial heterostructure and

c-directional axial heterostructure are illustrated in Figure 2-1(a), (b) and (c), respectively.

Figure 2-1. Selected designs of GaN nanowire-based LEDs. Illustration of the morphology

and placement of the InGaN QW and alignment relative to crystallographic c- and m-directions

for (a) m-directional axial heterostructures, (b) c-directional radial heterostructures and (c) c-

directional axial heterostructures.

29

2.2.1. Homojunctions

Nanowire homojunction devices contain no quantum well or heterojunction, only p-type and

n-type sections, either in an axial or radial configuration. These devices are simple p-n junctions

within a nanowire. Fabrication of the axial homojunction can be achieved by switching the

dopant source part-way through the growth, while fabrication of the radial homojunction can be

achieved by first growing the nanowire with one dopant and then switching the growth

conditions to promote vapor-solid (film) growth with a different dopant type to deposit an

epitaxial shell on the nanowire sidewall facets. While these structures are simple, the radiative

recombination rates are typically low compared to using a double-heterostructure due to less

wavefunction overlap of the electrons and holes. These structures also have a larger active region

from which light can be emitted than double heterostructure, which is equivalent to the sum of

the diffusion length of the minority carriers in each side of the junction and the depletion region

width. In axial homojunctions the large active region can be an issue when trying to maximize

coupling to the waveguided modes within the nanowire waveguide.

2.2.2. Nanowire-substrate heterojunctions

Nanowire-substrate heterojunctions are formed simply by growing a doped nanowire of a

dopant type opposite to the substrate used, such as n-GaN nanowire on a p-type substrate. These

devices are actually likely the simplest nanowire-based LEDs to fabricate, as only one type of

doping needs to be achieved, and they can act as a first step towards more refined devices. With

regards to growth of nanowires on Si substrates, there are some key issues with these single-

heterojunction devices. Since Si has an indirect band gap in the IR, any emission from Si would

be useless for solid-state lighting and the efficiency would be very low. To maximize emission

from the nanowire, a highly doped Si substrate can be used, which should create a larger

depletion region within the nanowire, enhancing emission from the nanowire; however, all of

this emission would occur at the nanowire base and would have to be guided out of the array to

be extracted.

2.2.3. Axial heterostructures

Axial heterostructures are double-heterostructure devices with InGaN regions inserted along

the length of the nanowire between the p-n junction. In these devices the active region is limited

30

to the size of the InGaN region, where carriers are trapped. This allows for designing the optimal

position of the InGaN region along the length of the nanowire. Quantum well positioning to

maximize the extraction through the nanowire tip has not previously been thoroughly studied,

and we address this aspect of design in the next chapter. In GaN, the axial heterostructure can be

fabricated in either triangular cross-sectioned non-polar m-directional nanowires (Figure 2-1(a))

or hexagonal cross-section polar c-directional nanowires (Figure 2-1(c)). The effect of the

polarity can play a large role in the IQE from the device, as discussed in the next section.

2.2.4. Radial heterostructures

Radial heterostructures are double-heterostructure devices with the active InGaN region

wrapped around an n-GaN core followed by the deposition of a p-GaN shell. Radial

heterostructures can be fabricated in either c-directional nanowires (Figure 2-1(b)) or m-

directional nanowires. It has been shown that the optimal placement of the active region in c-

directional radial heterostructures is at 90% of the nanowire radius where light extraction was

maximized[118]. Using this guideline we can investigate one of the most important advantages

of nanowire radial heterostructures: the increase in active area for high-brightness applications.

The overall amount of emitting material per surface area of the device can be increased relative

to thin-films and axial heterostructures, since it is wrapped around a vertical nanowire core. This

comparison is made in Figure 2-2 by defining an active region multiplier relative to a thin-film as

a function of the nanowire areal density, assuming a 7 nm InGaN quantum well region in 1 µm

long m-directional and c-directional GaN nanowires, where m-directional nanowires are

approximated as regular triangular prisms with a side-length of 100 nm and c-directional

nanowires are approximated as regular hexagonal prisms with a side-length of 50 nm, such that

the maximum width of the two nanowires are both 100 nm. As experimentally observed in m-

directional radial heterostructures, it can be difficult to grow continuous InGaN active regions

around the entire nanowire, since two of the side-facets of m-directional are semi-polar and one

is polar[41] (see Figure 3-1(b)), thus only two facets are modelled with an active InGaN region

for m-directional radial heterostructures. It should be noted that an increase in active area will

necessitate the increase of injected carriers to maintain the same brightness level per active area,

assuming other parameters of the device remain similar. In general, the axial nanowire

heterostructures only approach an active area multiplier of one at a fill-factor of one (i.e. a thin

31

film). Radial heterostructures, however, can far surpass the active area of thin-films at fill-factors

as low as ~1 nanowire/µm2. This effect can be further expanded by using longer nanowires.

Figure 2-2. Comparison of the active area in nanowire heterostructures and thin-films. The

active area multiplier of GaN nanowire-based LEDs is plotted as a function of the nanowire areal

density for nanowires 1 µm in length with a maximum width of 100 nm and a 7 nm active area,

demonstrating that the active area of radial heterostructures can far surpass thin-films.

While the active area of radial heterostructures can be advantageous for high-brightness

applications, as opposed to axial heterostructures, a comprehensive comparison of different

nanowire-based LED designs to identify the most efficient has not been previously done. In

Chapter 3 we choose the most promising candidates for high efficiency: c-directional radial and

axial heterostructures and m-directional axial heterostructures to focus on the material factors

affecting light extraction efficiency and internal quantum efficiency (IQE) to deduce the overall

external quantum efficiency (EQE) and directly compare the efficiencies of these three designs

for nanowire-based LEDs. The factors considered include waveguiding through differently

shaped cross-sectional nanowire waveguides, the role of polarity across quantum wells, and the

active region emission’s dependence on crystallographic orientation of the quantum well and

how that can affect the waveguiding and extraction efficiency. We discuss these factors in depth

in the following section.

32

2.3. Efficiency considerations

The overall efficiency of light-emitting diodes is the EQE, which is a measure of the overall

number of photons generated per injected carrier. The EQE is the product of the light extraction

efficiency η and the IQE. The light extraction efficiency is a measure of the percentage of

photons generated within the device that escape the device towards the observer. The IQE is a

measure of the number of photons generated per injected carrier. Light extraction efficiency is a

property of the device design and is often enhanced in thin-films by inducing scattering. Light

extraction efficiency from nanowire-based devices depends on the nanowire morphology and the

overall arrangement of the nanowire array, as discussed in depth later in this work. The IQE is

more of a function of material quality and can be affected by the presence of defects which act as

non-radiative recombination centers. The electronic design of the device is also critical to high

IQEs, such as the double-heterostructure which traps carriers in the quantum well to enhance

radiative recombination and increase the IQE.

2.3.1. Light extraction

One of the greatest challenges of enhancing the overall efficiency of LEDs has been low light

extraction efficiencies. Many techniques have been developed to improve light extraction, such

as the flip-chip LED[119], surface roughening[120] and even the integration of photonic crystals

into LED designs[121]. Nanowires can offer unique solutions to increasing light extraction

efficiency, as the large index contrast between GaN and most insulators means that the nanowire

acts as a waveguide along the vertical axis of the nanowire and can act to guide emitted light

towards extraction[118]. The cross-sectional shape and growth direction influence the nanowire

waveguiding properties, whereas the polarization anisotropy observed in III-nitrides strongly

affects the efficiency of coupling to these modes[122, 123].

The polarization anisotropy can have a profound effect on the expected extraction

efficiencies from the considered nanowire LED designs, and therefore must be taken into

account in the case of highly anisotropic III-nitride nanowires. Polarization anisotropy of

emission from InGaN QWs in GaN thin-films has been reported for c-plane and m-plane

QWs[123, 124] due to an estimated energy gap of 49 meV between the polarization states at

room temperature[122]. Thus, most of the radiation has an electric-field perpendicular to the c-

33

axis, denoted as E⊥c, as opposed to parallel to the c-axis, denoted as E||c. The degree of

polarization anisotropy is described as the polarization ratio, the ratio between the difference in

intensity between E⊥c and E||c and the total intensity. The polarization ratio seen in the literature

are at values of 0.8 for m-plane QWs[122] and 0.56 for c-plane QWs[123].

Figure 2-3. Schematic of polarization anisotropy in different nanowire-based LED designs.

The polarization anisotropy from InGaN m-plane QWs results in most emission directed along

the nanowire axis for (a) m-directional axial heterostructures and perpendicular to the nanowire

axis for (b) c-directional radial heterostructures. The anisotropy is less pronounced in c-plane

quantum wells, resulting in more even emission both along the nanowire axis and perpendicular

to it for (c) c-directional axial heterostructures.

In the case of nanowires, the highest directed extraction efficiency is expected for the case

with the maximized emission coupling to the waveguide modes. For the m-directional axial

heterostructure, the orientation of the m-plane QW results in most of the emission oriented along

the vertical axis (shown schematically in Figure 2-3(a)). The opposite is expected in the case of

the c-directional radial heterostructure, in which most of the emission is oriented parallel to the

plane of extraction (Figure 2-3(b)). The same effect is seen for c-directional axial

heterostructures, but it is less pronounced (Figure 2-3(c)) due to the smaller polarization ratio for

c-plane QWs. Emission along the vertical axis is expected to couple to the waveguide modes

more efficiently than for emission parallel to the plane of extraction[118], thus the highest

extraction efficiency is expected for the m-directional axial heterostructure.

34

2.3.2. The quantum-confined Stark effect

Figure 2-4. The quantum-confined Stark effect. The electric field present across a polar QW

results in a physical separation of where the carriers are confined and a reduction in the energy

that will be emitted upon recombination (red-shift), as demonstrated in the band-diagram

schematic, where CB and VB are the conduction and valence bands, respectively. On the other

hand, in non-polar QWs these effects do not occur since there is no electric field across the

quantum well.

The QW orientation within the nanowire relative to the growth direction determines the

extent of the QCSE: an electric-field can shift apart the electron and hole wave functions,

resulting in a reduction in spontaneous emission efficiency and a redshift of the emission[125].

The band-diagram of polar and non-polar is schematically shown in Figure 2-4, demonstrating

how electrons and holes become spatially separated (which results in reduced recombination

efficiency) and how the effective emission from recombination is red-shifted due to the band-

bending within the carrier-confined region. Since the electric fields are caused by the crystal

polarity and piezoelectric effect, the QCSE is particularly important in the case of c-plane QWs.

This reduction in spontaneous emission efficiency equates to a reduced IQE in the case of polar

c-plane QWs; however m-plane QWs have no electric field across the active region and thus do

not experience the QCSE. Thus, both m-directional axial heterostructures and c-directional radial

35

heterostructures which have m-plane quantum wells have higher IQEs than c-directional axial

heterostructures which have c-plane quantum wells.

2.3.3. Other design considerations

In thin-film GaN LEDs on insulating substrates, current crowding can be an issue. A mesa

structure must be used to contact both the p-type and n-type regions; however this causes current

to crowd around the edges of the mesa closest to the bottom contact. Due to significant resistivity

in the underlying thin-film, the current tends to avoid flowing through the film[4]. Current

crowding has also been an issue in core-shell nanowire LEDs for similar reasons, where the

current flow in a core-shell nanowire LED tends to flow primarily through the more conductive

nanowire core than the shell, thus current is not flowing to the actual junction at the top of the

nanowire resulting in a much smaller active region[126]. However, with the growth of axial

heterostructure nanowires on a conductive substrate, these issues may be circumvented, by being

able to contact the nanowires through the substrate. It should be noted that ideal core-shell

nanowire LEDs would emit from the entire surface of the core, while the emission area in the

axial heterostructure is limited to the cross-section of the nanowire. Due to the need to make

electrical contact only to the outer shell in the case of core-shell nanowire LEDs, higher

nanowire densities are likely possible with axial heterostructures, thus the overall active area

may be similar.

In thin-film GaN LEDs heat extraction can be an issue, especially because these structures

are grown on sapphire, which has a thermal conductivity of approximately 35 W/m·K at room

temperature[127]. This thermal conductivity is low compared to GaN, which is 230 W/m·K at

room temperature[5], thus it acts to limit heat extraction from where it is typically generated in

GaN LEDs: the active region[128]. This heat can cause reliability issues[129]. This, along with

better light extraction, has motivated techniques to develop complicated “flip-chip” LEDs, as

well as the integration of GaN LEDs on silicon[119]. Silicon has a thermal conductivity more

than four times that of sapphire, 156 W/m·K at room temperature[130]. As nanowire-based

LEDs can be grown directly on silicon, this can greatly improve heat extraction. However, in the

case of nanowire-based LEDs, the large aspect of nanowires presents the challenge of extracting

heat from only a small interface. This challenge must be overcome with advanced designs, such

as using a thermal conductor to fill the spaces between vertical nanowires, which would allow

36

for enhanced heat extraction compared to thin-film counterparts[129]. Though the effects from

current crowding and thermal management are not directly considered in this work, they are

useful considerations to keep in mind when considering different designs for nanowire-based

LEDs.

2.4. Summary

In this chapter we discussed the different potential designs for nanowire-based LEDs and the

design considerations that must be balanced when comparing the efficiency of these designs to

determine which is best-suited for high-efficiency solid-state lighting. Considering the specific

case of GaN-InGaN double-heterostructures, we recognized the need to address a few key

unknowns. How the nanowire waveguide cross-section in m-directional and c-directional

nanowires affects the modal properties and how the quantum well position in axial

heterostructures affects coupling to these modes in the context of light-extraction from nanowire-

based LEDs must be directly compared for the designs. Furthermore, known polarization

anisotropy in emission from InGaN QWs must be included in this analysis of light extraction.

The change in IQE for polar versus non-polar oriented quantum wells, due to the QCSE, must be

considered for an overall comparison of the EQE of these different design. In the next chapter,

we assess these different nanowire-based LED designs to directly compare the EQE using a

combination of finite-difference time-domain photonic simulations and IQE modelling.

37

Chapter 3: Evaluating GaN/InGaN Nanowire-Based LED Designs

Demonstration-scale nanowire-based LEDs have been fabricated, typically as c-directional

radial[131] and axial heterostructures[132, 133] or m-directional radial heterostructures[134,

135]; however there has been no direct comparison of these designs to guide future work towards

high-efficiency nanowire-based LEDs. The modal properties of nanowires have been calculated

for a range of devices, originally for cylindrical nanowires[136], and then for more realistic

hexagonal[118] and triangular cross-sections[137]. The extraction efficiency has been reported

in studies of circular and hexagonal nanowires, generally assuming isotropic emission [118, 136,

138]. Recently, a study of c-directional radial heterostructures has also been conducted,

including polarization anisotropy[139]. However, a direct comparison of m- and c-directional

axial heterostructures requires a comparative study of the modal properties, the extraction

efficiency, and the effects of crystallographic growth directions on the IQE, which has not been

previously conducted. Here, we compare several typical nanowire-based III-nitride LED designs

on a Si substrate, focusing on c-directional and m-directional axial heterostructures in which the

QW is inserted along the nanowire length, as well as the c-directional radial heterostructure in

which the QW is wrapped around the nanowire in a core-shell configuration. The directed

extraction efficiency is considered as a function of QW position within the nanowire using finite-

difference time domain (FDTD) photonic simulations. The IQE is modeled by considering

radiative and non-radiative recombination rates, allowing for direct comparison with

experimental results in the literature. Finally, by combining as-determined directed extraction

efficiency and IQE, a range for the directed EQE is estimated and tradeoffs of the different

designs are discussed. We note that this is the first study to directly compare the common

nanowire architectures while including the important effects from polarization anisotropy,

different cross-sectional geometries and crystallographic growth directions.

3.1. Finite difference time domain simulations

FDTD breaks up Maxwell’s equations into dependencies on the electronic and magnetic

fields as a function of coordinates and time (x,y,z,t). The simulated mesh then discretizes these

fields in space to segment the problem into many parts. Starting with the initial conditions in

each segment, a new H-field is calculated in each segment over the entire volume. Time is then

38

advanced a half-step and the new E-field is solved based the previously calculated H-field. Time

is advanced another half step and the process is repeated for the amount of time designated.

FDTD is unique among photonic simulations, in that it is time-dependent, but it can be

computationally demanding. Typically a perfectly matched layer (PML) boundary condition is

applied, which is completely absorbing at the edges of the simulation with a grading applied for

continuity ensuring no reflection from boundaries back into the region of interest. To ensure the

PML is not affecting the simulation, the bounds of the simulation region must be a few times

greater than the region of interest. Periodic boundary conditions and transparent boundaries can

also be applied.

3.1.1. Simulation Setup

To find the directed extraction efficiency for the different nanowire-based LED designs, we

used three-dimensional FDTD Lumerical software with a maximum mesh size of λ/15n,

assuming a 1 µm long nanowire with a refractive index n = 2.6. The nanowire was oriented

vertically on a silicon substrate with flat facets on both ends, as depicted in Figure 3-1(a). PMLs

were used as the boundary conditions to assure that any radiation reaching the boundary does not

reflect back to interfere with the simulation in the region of interest. A power monitor was used

to observe the amount of radiation passing through a plane immediately above the nanowire tip.

The directed extraction efficiency was calculated as the ratio of power through the monitor plane

to the power emitted by the emission source (e.g. QW), and as a function of the QW position

within the nanowire, h (Figure 3-1(a)). By using a single monitor, we are effectively defining the

directed extraction efficiency as the light which is directed toward the observer, composed

mostly of emission from the waveguided modes, and effectively ignoring higher angle emission.

While much of the light which escapes the nanowire will be outside this limit, we are interested

in designing nanowire-based LEDs such that we can tune the characteristics on an individual

nanowire level, which emission from guided modes is best suited for. m-directional and c-

directional nanowires were approximated as prisms with the cross-section of an equilateral

triangle of side-length d (Figure 3-1(b)) and a regular hexagon of side-length aeq (Figure 3-1(c)),

respectively. The dimensions d and aeq were chosen such that the nanowires had the same cross-

sectional area for direct comparison, specifically d = ⋅6 aeq.

39

Figure 3-1. Simulation setup used for FDTD photonic simulations. (a) Schematic of

simulation setup in x-z plane, with emission source (QW) at a distance h from the substrate. The

nanowire cross-sections are assumed to be (b) an equilateral triangle of side-length d for m-

directional nanowires, with two semi-polar 1122 facets and one polar c-plane (0002) facet, and

(c) a regular hexagon of side-length aeq for c-directional nanowires with six non-polar m-plane

1100 facets. Black dots in (b) and (c) represent the center of the cross-section and gray dots

represent the location of additional point sources used to simulate the full QW.

To simulate incoherent spontaneous emission, as expected from a QW, the emitted electric-

field intensity was calculated as a weighted average of the electric-field intensities emitted from

three independent orthogonal dipole sources[140]. The relative weight of emission from the

different dipoles was assigned dependent upon the transition probability of each

polarization[141], as discussed in Section 2.3.1. Light extraction. Unless otherwise noted, QWs

were simulated by averaging the results of spatially equally distributed point sources within the

QW, taking advantage of symmetry to reduce the number of simulations when possible (Figure

3-1(b-c)). The results reported here assume a typical 7 nm QW of In0.13Ga0.87N, and hence a

wavelength of 405 nm.

40

3.1.2. Modal properties of nanowire waveguides

Figure 3-2. The modal properties of hexagonal and triangular nanowire waveguides.

(a) Dispersion relation of the first six modes of nanowire waveguides plotted for both m- (solid

lines) and c-directional (dashed lines) nanowires, also plotted against d (or aeq 6⋅ for hexagonal

cross-sections) at the emission wavelength of 405 nm. (b) The cross-sectional electric-field

intensity profiles of the lower order modes. Note that the symmetry is broken due to the

difference in cross-section along the x- and y-axes.

Though the modal properties of hexagonal[118] and triangular cross-sections[137] have been

previously reported, we calculated the modal dispersion of each case here for a direct

comparison. To estimate nanowire side-length regimes for which well-confined waveguiding is

sustained (i.e. neff > 1), we calculated the dispersion relation for the first six modes in GaN

nanowire waveguides (Figure 3-2(a)). The normalized propagation constant, ωd/c, where ω is

the angular frequency, is proportional to the ratio of nanowire side-length and wavelength, d/λ,

while the effective index, neff, is higher for more confined modes. At the wavelength of 405 nm,

the single-mode regime – where only the fundamental mode HE11 exists – is calculated to be

sustained for d < 160 nm and aeq < 76 nm in the case of m- and c-directional nanowires,

respectively, which is in agreement with previously reported values[136, 137]. Each mode has a

distinct cross-sectional electric-field intensity profile (Figure 3-2(b)). The overall intensity varies

41

along the length of the nanowire with the effective wavelength of the mode, but the relative

intensities within the cross-section maintain the distinct modal profile. Note that the dispersion is

remarkably similar for the m- and c- directional fundamental modes, especially in the single-

mode regime, which indicates that for nanowires of equal cross-sectional area the modal

properties should be similar. The cross-sectional intensity profiles of the fundamental mode are

also both approximately radially symmetric indicating that coupling to the fundamental mode

should have the same trend with respect to distance from the nanowire center, which is important

when considering coupling changes throughout the cross-section. The higher order modes vary

more when comparing the cross-sections and would be important to understanding coupling from

emission in multimode nanowire waveguides[142].

3.1.3. Point emission sources

For the analysis of a point source’s position-dependence on the directed extraction efficiency,

we focus on m-directional nanowires, for which the most coupling to modes is expected based on

the polarization argument discussed above, before moving on to more computationally

demanding and time-consuming plane-source simulations. We calculate the directed extraction

efficiency for m-directional nanowires in various waveguide regimes (leaky single-mode

nanowires – with neff of the fundamental mode < 1, well-confined single-mode nanowires with

neff of the fundamental mode > 1 and multimode nanowires) and compare it with a 1 µm c-plane

GaN thin-film on silicon as a function of the QW position (Figure 3-3(a)). Here, QWs were

approximated as a single, centered point source to reduce the number of simulations necessary

for this general comparison. While this approximation is an overestimate, the general trends are

the same as observed for full plane-source simulations and the actual values are only about 8%

lower than this approximation.

42

Figure 3-3. Directed extraction efficiency dependence on point source position. (a) Directed

extraction efficiency as a function of the QW placement for leaky and well-confined single-mode

and multimode m-directional nanowire waveguides compared to a thin-film. (b) Comparison of

well-confined single-mode nanowire waveguide extraction and electric-field intensity of the

fundamental mode profile.

For the thin film, the extraction efficiency is independent of QW position and limited to

about 4% by total internal reflection. For nanowires, however, the extraction efficiency is

generally higher and dependent on the QW position and the nanowire diameter. Extraction from

43

a thin single-mode nanowire (d = 60 nm, neff < 1) shows a maximum directed extraction

efficiency of 45% for a QW position near the nanowire tip and a minimum of 1.2% for a QW

position near the substrate with oscillations approximately equivalent to the free-space

wavelength, λ0 = 405 nm. The steady increase in extraction as the QW is positioned closer to the

nanowire tip is due to an increased emission from extraction cones on the nanowire side-facets as

the source is moved towards the plane of extraction. The oscillations can be attributed to the fact

that for leaky single-mode nanowires extracted light mostly propagates outside the nanowire and

interferes with itself after reflection from the surface of the silicon substrate. The oscillations in

extraction efficiency from a thicker (d = 145 nm, neff > 1) single-mode nanowire waveguide

between a maximum of 56% and a minimum of 21% (near the substrate) can be ascribed to the

same effect, where the oscillations correspond to the effective wavelength of the mode, λ0/neff.

We further explore the origin of the oscillations in directed extraction efficiency by directly

comparing it with the electric-field intensity profile of the standing wave excited by the

fundamental mode in a single-mode nanowire waveguide with neff > 1 (Figure 3-3(b)). Since the

directed extraction efficiency is primarily due to extraction of the fundamental mode, the

intensity profile and directed extraction efficiency follow the same oscillation pattern (coupling

of the dipole emission is expected to be strongest at maxima in the electric-field intensity

profile[118]). This trend is obscured in c-directional nanowires due the polarization anisotropy,

which results in less coupling to the fundamental mode. When the nanowire side-length is

increased even further such that multiple modes become active, much more complicated

dependency on QW position is observed (Figure 3-3(a)). For example, for a nanowire with d =

300 nm, up to six total modes exist within the nanowire, resulting in overall lower directed

extraction due to lower overall coupling from a discrete source, with maximum and minimum

extraction efficiencies of 48% and 13%, respectively.

The design of single nanowires can be expanded to nanowire arrays without considering the

complications from scattering and interference if the nanowires are adequately spaced apart. For

adequate spacing the average inter-nanowire distance should be large enough to avoid overlap in

the emission from neighboring nanowires. To understand the range of nanowire areal densities

under which only individual nanowire-based LED designs need to be considered, we gain insight

from considering the cross-sectional modal profile. Since single-mode nanowire-based LEDs

have the highest extract, we consider nanowires which are near the multi-mode cut-off (d = 145

44

nm or deq = 120 nm for hexagonal cross-sections). The approximate maximum density at which

hexagonal nanowire modal profiles overlap minimally is 20 nanowires/µm2 as shown in Figure

3-4(a). If the nanowire is smaller in diameter then the maximum areal density decreases (to 13

nanowires/µm2 for deq = 80 nm, where neff of the fundamental mode is ~1) since a larger portion

of the mode is outside of the nanowire as shown in Figure 3-4(b). These principles can be used to

guide the synthesis of wide-band gap nanowires for LEDs: for individual nanowire-based LED

design to scale and have a maximized extraction the nanowire should be single mode (d from

100-160 nm for m-directional nanowires and deq from 80-120 nm for c-directional nanowires)

and the maximum nanowire areal density to avoid modal overlap is about 13-20 nanowires/µm2.

Figure 3-4. Fundamental mode profile intensity cross-section for c-directional nanowires.

Cross-section of the intensity of the fundamental mode, HE11x, for c-directional nanowires with

the width of the mode (equivalent to the necessary inter-nanowire spacing to prevent modal

overlap) for (a) deq = 120 nm and (b) deq = 80 nm.

45

Figure 3-5. Directed extraction dependence on point source position in cross-section.

Extraction efficiency for a single-mode m-directional nanowire (d = 145 nm) as a function of

emission distance from the substrate (h), with points at the center, near the facets and near the

corners represented. As the source moves away from the center, the emission coupling to the

fundamental mode, and thus extraction, decreases.

To fully simulate emission from a nanowire QW, equally distributed point sources within the

nanowire cross-section were independently simulated. Analysis of the extraction efficiency

changes with respect to the position of the dipole source can provide insights into the regions

which contribute most to extraction and can be useful when emission is not perfectly distributed

across the entire plane of the quantum well (QW). For example, in the presence of a surface

depletion region, most of the current will flow through the core of the nanowire[89] therefore

limiting the cross-sectional position of the emission source to the center. For the single-mode m-

directional nanowire studied in this work, in which only the fundamental mode HE11 exists, the

electric-field intensity is highest in the center of the nanowire, thus it is expected that as the

point-source is moved away from the nanowire center the coupling into the mode will decrease

(see Figure 3-2(b)). Since most of the extracted light corresponds to this guided mode, the

46

extraction efficiency will also decrease as the point-source is moved away from the center of the

nanowire cross-section (Figure 3-5). The oscillations of extraction efficiency with the emission

distance from the base do not change significantly, where the oscillations are due to changes in

coupling into the fundamental mode. It is notable that emission at the top corner (oriented on the

y-axis) is less extracted than the comparable right corner (oriented on the x-axis). Since most of

the emission is coming from a y-oriented dipole, it is more likely to be scattered by the corner

along the y-axis, as opposed to the corner along the x-axis. We note that for the c-directional

nanowires, the above discussed trends are more difficult to observe, since the dominant dipole is

oriented along the z-axis, which does not couple well to the fundamental mode.

This work generally assumes an equilateral triangle and regular hexagon for the cross-

sections of m-and c-directional nanowires, respectively. The experimentally grown nanowires,

however, generally do not have these ideal cross-sections. Nanowires grown in the c-direction

typically have slightly faceted corners that can decrease the effective index of waveguide modes

in the nanowire. However, so long as the faceting is symmetric, the general properties of the

modes will remain similar. The cross-section of m-directional nanowires has recently been

thoroughly investigated with electron tomography in InN-InGaN nanowire heterostructures[143].

The m-directional nanowires have a less ideal cross-section and the actual cross-section deviates

slightly from an equilateral triangle, with two corners with angles of about 58° and the other with

64°. We simulated the extraction from a centered point source via FDTD for such a nanowire

(Figure 3-6). The broken symmetry results in a lifting in degeneracy of the fundamental modes,

HE11x and HE11

y. As expected, because the nanowire is smaller along the x-axis, HE11x has a

smaller effective index (and thus larger effective wavelength) as compared to HE11y. Recall that

due to the polarization anisotropy most of the emission is polarized with the electric field

perpendicular to the c-axis. In this case such emission corresponds to emission from the dipole

oriented along the y-axis. Thus, due to the polarization-anisotropy the extraction efficiency

oscillates with the effective wavelength expected from HE11y. The effective index of this

fundamental mode will be smaller as compared to the ideal structure as the cross-sectional area is

decreased and thus confinement of the mode is decreased. The dipole oriented along the z-axis

oscillates approximately with the free-space wavelength, indicating most of this emission is

back-reflected from the Si substrate. These results emphasize that for careful optimization of

nanowire-based LED design it may be important to consider realistic cross-sections.

47

Figure 3-6. Effect of realistic nanowire cross-sections on extraction. Extraction efficiency as

a function of emission distance from the substrate (h), for the three-orthogonal dipoles simulated

at the center of a realistic m-directional nanowire via FDTD. The total extraction follows the

same trend as the dipole oriented in the y-axis due to polarization-anisotropy.

3.1.4. Plane emission sources

After considering different nanowire waveguide regimes, we next compare the overall

directed extraction efficiency of well-confined single-mode m-directional axial and c-directional

axial and radial heterostructures (Figure 3-7), with d = 145 nm and aeq = 59 nm, respectively.

Unless otherwise noted, we focus on these well-confined single-mode nanowires for the

remainder of this work. The c-directional radial heterostructure was found for a QW placed at

0.9 the distance from the center to the surface, where extraction has previously been reported to

be optimal[118]. The m-directional axial heterostructure has strikingly larger extraction as

compared to the c-directional axial and radial heterostructures, but the placement of the QW is

critical to optimize the extraction efficiency, as discussed above. In contrast, the c-directional

axial heterostructure has much less variation, as most of the emission is oriented parallel to the

plane of extraction. While this result indicates that m-directional axial heterostructures have an

advantage in terms of directed light extraction, expected differences in the IQE must also be

considered, which we discuss in the next section.

48

Our results demonstrate that regardless of their side-length, nanowires exhibit enhanced

extraction efficiency over an unprocessed thin-film. It should be noted, however, that extraction

efficiencies from thin-films can be greatly enhanced via a variety of techniques, even reaching

extraction efficiencies of up to 80% in optimized devices[144]. However, well-confined single-

mode m-directional nanowires can achieve extraction efficiencies near 50% without any

additional processing steps, simply by placing the QW at the proper position along the nanowire.

We also note that nanowire arrays may be able to reach even higher extraction efficiencies, but

the calculation of extraction efficiency here is meant as a starting point to compare m- and c-

directional nanowires for LED applications.

Figure 3-7. Compared directed extraction for nanowire-based LED designs. Simulated

directed extraction efficiency as a function of the QW placement for m-directional and c-

directional axial heterostructures, with the solid line indicating the extraction efficiency for a c-

directional radial heterostructure with the QW placed at 0.9 the distance from the center to the

surface.

3.2. Modeling the internal quantum efficiency

The A-B-C model is often used to model the IQE and compare relative recombination rates

for LEDs. This simple model is intuitive and provides a physical insight into the dominant

49

recombination mechanisms in the device[145]. The current density, J, and the IQE are dependent

on the carrier density, n, in the active region of the device (in this case the QW) as shown in

Equation 3-1 and Equation 3-2, respectively:

= + + Equation 3-1

IQE = + + Equation 3-2

where q is the elementary charge, w is the active region width (approximated by the QW width)

and the coefficients A, B and C represent the relative rates of non-radiative, radiative, and Auger

recombination, respectively. The model assumes that the electron and hole densities are of the

same order in the active region at non-equilibrium, that there is no carrier leakage across the

QW, and that A, B and C are only weakly dependent on n[145]. This model has been fit to

experimental results for a c-directional axial nanowire heterostructure[146], resulting in room-

temperature values of A=7×108 s-1, B=3×10-10 cm3s-1, and C=4.5×10-29 cm6s-1. Though this

device is different from the simpler ones we are simulating, it is a useful starting point for

comparison, as the characteristics of a multi-QW system are often dominated by a single QW,

typically the one closest to the more resistive p-type layer[145].

While the values for B and C depend primarily on the LED design (QW orientation, number

of QWs, existence of electron blocking layers), the values for A are specific to size and geometry

in nanowire-based LEDs, thus we cannot simply use A from the experimental reference[146].

Non-radiative recombination can occur due to trap states or non-radiative recombination centers

such as dislocations or surface states. In nanowires it is expected that non-radiative

recombination is mostly governed by the surface recombination due to the high surface-to-

volume-ratio and the fact that dislocations are not present[147]. For a typical thin-film GaN

LED, the non-radiative recombination rate has been reported as Abulk=5.4×107 s-1, representing an

estimate for bulk non-radiative recombination rate in the regime of low threading dislocation

densities[148]. (In an ideal case with no dislocations and low surface recombination, A would be

equivalent to such a bulk value.) We used this bulk value as a comparison against the values

dominated by surface recombination, as expected in a nanowire-based device. For a c-directional

50

axial heterostructure with non-polar facets, based on the values measured experimentally[146,

149, 150], assuming a surface depletion region[151, 152] based on a doping density of 1018 cm-3

and extracting to the geometries considered here, we obtain a value of Ac-directional=2.9×108 s-1.

Similarly, for m-directional nanowires with higher surface recombination at the polar c-plane

facets[149] we estimate a higher Am-directional=5.5×108 s-1. Using the experimental values for B

and C cited above and varying A from the bulk value to the calculated values for c- and m-

directional axial heterostructures (as summarized in Table I), the general IQE is found to

decrease from the ideal bulk-like case to the c-directional axial heterostructure, with the m-

directional axial heterostructure displaying the lowest IQE (Figure 3-8(a)).

Figure 3-8. Comparing changes in coefficients of the A-B-C model for finding IQE. (a) The

effect of increasing surface recombination on expected IQE assuming constant B and C, and

varying A. (b) The range of expected IQE’s for bulk and nanowire cases, where minimal IQE

represents radiative recombination from a c-plane QW and the dashed and maximum IQE curves

represent low and high-end estimates for B from an m-plane QW. (c) The range of expected IQE

from m- and c-directional axial heterostructures using the low end estimate of B for the m-plane

QW with the range minimum and maximum determined by extent of surface recombination. See

Table 3-1.

The above consideration only takes into account the trend with surface recombination and

does not account for changes in radiative recombination between m- and c-directional axial

heterostructures, which we discuss next. The coefficient B represents the amount of radiative

recombination, thus for high IQE it is desirable for B to be large. Since the radiative

recombination probability is expressed by Fermi’s golden rule, which is proportional to the

51

matrix element of wave function overlap, |M|2, B is directly proportional to |M|2. To estimate B

for different nanowire configurations, the QCSE must be taken into account. Due to the electric-

field induced by the polarity in the c-direction of GaN and the piezoelectric effect shifting apart

the electron and hole wave functions[125], the QCSE is significant for c-plane QWs resulting in

a lower |M|2 for c-plane QWs than m-plane QWs, where the effect is absent. This difference

results in a higher spontaneous emission rate for m-plane QWs, and thus a higher B. In

simulating the QCSE in InGaN QWs using self-consistent and flat-band models, it has been

reported that for m-plane QWs |M|2 is enhanced by up to a factor of eight and at least a factor of

two compared to c-plane QWs, respectively[125]. These factors are thus multiplied by the

baseline B for c-plane QWs (3×10-10 cm3s-1) to arrive at an estimate of B for both the self-

consistent (2.4×10-9 cm3s-1) and flat-band models (6×10-10 cm3s-1). The resulting IQE for this

range of B values is shown in Figure 3-8(b), for both nanowire-like and bulk-like non-radiative

recombination. The minimum IQE curves in each case represent a c-plane QW and the

intermediate (dashed) and maximum curves represent the flat-band model (low) and self-

consistent model (high) estimates for an m-plane QW, respectively (see Table 1). Experimental

results have shown that thin-film devices grown in the m-direction have IQEs about a factor of

two higher than the devices grown in the c-direction[153], which supports using the flat-band

model estimate in which |M|2 is also a factor of two higher in m-plane QWs than in c-plane QWs.

Thus, a value of B=6×10-10 cm3s-1 is used henceforth for m-directional axial heterostructures.

Effects of the QCSE and surface recombination discussed above were then combined (Figure

3-8(c)) to calculate a range of IQEs for both c- and m-directional axial heterostructures (see

Table 1). In general, it is likely that c-directional axial heterostructures will have a higher IQE at

lower current densities due to lower surface recombination compared with m-directional axial

heterostructures. As the current density increases, m-directional axial heterostructures will likely

have higher IQEs due to enhanced radiative recombination as compared to c-directional axial

heterostructures. These results do not include potential effects of the crystallographic orientation

on electron overflow through the QW. It has been shown that non-polar QW orientations

decrease electron overflow relative to polar QW orientations without an electron-blocking layer

(EBL); however, the inclusion of an EBL results in similar output powers[154]. Note that the c-

directional radial heterostructure has been excluded from this analysis, as a direct comparison to

experimental results could not be made.

52

Table 3-1. Coefficients of the A-B-C model used in modeling IQE.

A[s-1

] B[cm3s

-1] C[cm

6s

-1]

experimental

reference[146] 7×108 3×10-10 4.5×10-29

Figure 3-8(a)

comparing

surface

recombination

bulk[148] 5.4×107

3×10-10 4.5×10-29 c-dir[149, 150] 2.9×108

m-dir[149, 150] 5.5×108

Figure 3-8(b)

comparing QW

orientation

QCSE and

surface

recombination

bulk[148] 5.4×107 c-plane QW[146] 3×10-10

4.5×10-29

nanowire[149, 150] 5.5×108 m-plane QW[125]

low 6×10-10

high 2.4×10-9

Figure 3-8(c)

comparing

overall IQEs

c-dir [149, 150] 2.9×108 3×10-10 4.5×10-29

m-dir [149, 150] 5.5×108 6×10-10

3.3. External quantum efficiencies

Finally, we estimate the EQE range (Figure 3-9) by multiplying the directed extraction

efficiency dependent on QW placement (as determined by FDTD, Figure 3-7) by the expected

IQE and plotting the range from the minimum to maximum overall EQE curve. It is assumed that

nanowires are not surface-passivated, effectively giving a conservative estimate for the EQE. For

the optimal QW placement (700 nm from the substrate), m-directional axial heterostructures

generally have a higher EQE, mostly due to the differences in extraction. For the least optimal

QW placement in m-directional axial heterostructures (950 nm from the substrate), c-directional

axial heterostructures have a higher EQE for very small current densities, but the m-directional

axial heterostructures still outperform the c-directional axial heterostructures at higher current

53

densities, despite a higher amount of surface recombination expected for the m-directional

nanowires.

Several assumptions were made to arrive at the EQE results in Figure 3-9 and here we

discuss deviations from these assumptions. First, it is assumed that the cross-sections of the

nanowires are highly symmetric. In reality, these cross-sections do not generally have perfectly

sharp corners[143]; changes in symmetry will further lift the degeneracy of the modes (as

discussed in Section 3.1.3. Point emission sources) and slightly change their effective indices,

which should not play a significant role on the results besides on the optimal QW placement.

Nanowire devices are often encapsulated by a superstrate other than air, such as

benzocyclobutane[155], which would raise the index of refraction of the superstrate and reduce

the effective indices of the waveguide modes, shifting dispersion and the optimal QW placement.

Similarly, for different QW compositions and sizes, the emission wavelength would be different,

resulting in shifting to a different point in the dispersion, and thus a different optimal QW

placement. At longer wavelengths the modes will be less confined and thus the directed

efficiency would be decreased. Whereas the nanowire diameter could then be increased to obtain

a well-confined single mode at λ0 = 550 nm (green), this would move the regime at λ0 = 405 nm

(blue) to multi-mode, making designs for multiple QW challenging to optimize. For multiple

QWs of different emission wavelengths, the trade-off between optimal QW placement and

electrical injection would also need to be considered. Furthermore, the silicon substrate assumed

here reflects much of the visible spectrum, which is taken advantage of in this design.

Alternative substrates such as sapphire are transparent, which would necessitate the development

of different designs, as most of the emission reaching the substrate in this design would be lost

for transparent substrates. Such designs would likely mimic thin-film devices, which often emit

through the sapphire and are designed with a silver-based contact to reflect emitted light back

towards the sapphire and tune emission[144, 156].

54

Figure 3-9. Overall estimated EQE comparison of nanowire axial heterostructures.

Estimated EQE ranges for m- and c-directional axial heterostructures obtained by combining IQE

modeling and directed extraction efficiency simulations from FDTD.

Besides considerations that affect optimal QW placement and the role of nanowire density,

the material quality and processing can play important roles. The effect of the surface depletion

region, while considered in the IQE modeling, was not integrated into the FDTD simulations (as

calculated widths were small, only on the order of 10 nm), however large surface depletion

regions could actually enhance the extraction efficiency, by limiting emission to the center of the

nanowire, where extraction efficiency is expected to be highest. Though nanowires grown by the

self-induced method in plasma-assisted molecular beam epitaxy often display flat surface

facets[133], seed-mediated nanowires would normally have a metallic seed-particle at the

nanowire tip, which may quench emission or scatter the waveguide modes, effectively reducing

extraction efficiency. It is possible to chemically etch such seed-particles[157]. Finally, nanowire

sidewall surface roughness would induce losses and decrease the overall extraction efficiency.

3.4. Further enhancing extraction efficiency

This work has simulated only individual nanowires, thus motivating the use of directed

extraction rather than overall extraction from the nanowire (i.e. – including free space emission

from the nanowire sidewalls). We note that in many cases, much of the light is in fact not guided

55

by the nanowire (i.e. – undirected) and thus not considered in our definition of directed

extraction efficiency, even though the light does escape the nanowire, it is not directed

substantially away from the substrate toward an observer. For cases where most of the light is

confined within the nanowire, array effects should be minimal with nanowires spaced adequately

apart – and thus the overall efficiency would be close to the directed efficiency in such cases. It

should be noted that thinner nanowires will have large portions of light intensity outside of the

nanowire, for which the density of nanowires on the substrate should more strongly affect the

extraction[158]. Furthermore, it may be possible to form a photonic crystal with an array of

nanowires to prevent propagation along the plane extraction and suppress emission not oriented

towards extraction[34].

Figure 3-10. Extraction from square arrays of m-directional axial heterostructures. The

extraction efficiency from a nanowire in the center of an square array was found as a function of

the array pitch a and the QW position along the length of the nanowire for thin (d = 60 nm)

nanowires. In general, the extraction efficiency increased as the QW position is moved toward

the tip and as the array pitch increases; however, the comparison to a single nanowire of the

same dimensions indicates the arrays must be carefully optimized to enhance extraction. A

schematic of the nanowire arrays simulated is inset.

56

As a starting point for exploring the effects of nanowire arrays on light extraction, we used

FDTD to estimate the extraction from square arrays of thin (d = 60 nm) m-directional axial

heterostructures on Si as a function of QW position along the length of the nanowire and the

pitch (a) of the square arrays. Each simulation was performed for an emitting nanowire at the

center of an array of at least 5×5 nanowires, which was limited by memory requirements for

arrays with larger pitches. The results for extraction for nanowire arrays with pitches from 70 nm

to 200 nm actually result in generally lower extraction efficiencies that just an individual

nanowire. In this case, most of the emission is not guided within the nanowire and thus the

scattering of the nearby nanowires actually results in more light being scattered toward the

substrate. Larger pitches may be able to scatter emission towards extraction; however this still

needs to be investigated. To specifically tune the arrays to form a photonic crystal may

dramatically enhance extraction as well, thus the use of ordered nanowires arrays have great

potential to increase extraction efficiency, but their designs must be well optimized.

Furthermore, in traditional optoelectronics a distributed Bragg-reflector can be used to create

a perfect reflector at certain wavelengths by periodically alternating the effective index within a

waveguide. Recently, facile diameter modulation of III-nitride nanowires via metal-organic

chemical vapor deposition has been demonstrated by our group, where the diameter was

modulated by as much as a factor of two[143, 159]. FDTD was used to evaluate the potential for

creating an effective distributed Bragg reflector of the fundamental mode via such diameter

modulations. The reflector was designed to have a smaller base side-length of ds = 145 nm with

10 periodic repetitions of larger (doubled) diameter (dl = 290 nm), where the small and large

diameter segments have a length of ts and tl, respectively, as shown in Figure 3-11(a). Figure

3-11(b) demonstrate the local maximum of R = 0.82 for the smallest optimal segment lengths,

which occurs at ts = 92 nm and tl = 26 nm. Such reflecting structures at the nanowire base are a

promising structure to encourage light extraction by growing them prior to the growth of the

light emitting portion of the nanowire. The abruptness of the diameter modulation will be critical

to high reflectivity from the pseudo-Bragg reflector, and would be the limiting factor on

producing such structures, thus our simulations of abrupt diameter modulations are likely a high-

end estimate of what could actually be fabricated via diameter modulation using a seed-particle.

57

Figure 3-11. Distributed Bragg reflector via diameter modulations. (a) Schematic structure

of diameter modulation in m-directional GaN nanowires, denoting the diameters/lengths of the

small and large diameter segments, ds/ts and dl/tl, respectively. (b) The local maximum for

reflection at the smallest tl/ts with 0.82 reflectivity achieved at ts = 92 nm and tl = 26 nm.

3.5. Summary

In this chapter, we compared designs of III-nitride nanowire-based LEDs using a

combination of FDTD photonic simulations and the A-B-C model for IQE. We showed that

individual well-confined single-mode nanowires have the highest directed extraction efficiency

and that the variation of directed extraction efficiency as a function of QW placement could be

directly related to the expected coupling to the fundamental mode in this case. Furthermore, due

to the large polarization anisotropy expected in InGaN QWs, m-directional axial heterostructures

were found to have the highest directed extraction efficiencies as compared to c-directional axial

and radial heterostructures in the well-confined single-mode regime, as more of the emission can

be coupled to the fundamental mode. The placement of the QW was also found to be more

important in the m-directional axial heterostructure due to the strong coupling to the fundamental

mode, demonstrating oscillation between a maximum directed extraction efficiency of 49% and a

minimum of 24%. The A-B-C model was used to discuss the effects of non-radiative and

58

radiative recombination within m- and c-directional axial heterostructures. The m-directional

nanowires are expected to have a slightly higher non-radiative recombination rate, due to the

polar c-plane side-facet; however, the m-directional axial heterostructure is also expected to have

a higher radiative recombination rate due to the QCSE reducing the radiative recombination rate

in c-directional axial heterostructures. Finally, by combining these results, the expected EQE

ranges were estimated for both m- and c-directional axial heterostructures, demonstrating that the

m-directional axial heterostructures generally have a higher EQE, due primarily to the large role

that polarization anisotropy plays in the directed extraction efficiency and the enhanced radiative

recombination rate by avoiding the QCSE.

Though this work only considered blue-emitters, the same approach used here can be applied

to any emission wavelength to find the optimal QW placement. By looking at the optimal

placement for green and red wavelengths it would be possible to design a white LED with QWs

emitting at each wavelength positioned optimally all within a single nanowire. High efficiency

LEDs using m-directional axial heterostructures operated at high current densities could be

further improved by coating the nanowires with a high band-gap shell, such as AlxGa1-xN, to

reduce surface-recombination and boost the IQE. Furthermore, if the nanowires were arranged

periodically in a photonic crystal arrangement, it should be possible to boost the extraction

efficiency by suppressing propagation parallel to the plane of extraction. Another option to

enhance extraction was shown to be periodic diameter modulations at the base of the nanowire to

reflect more of the waveguided light away from the substrate toward extraction. This work will

help to guide the design of such future III-nitride nanowire-based LEDs to reach higher

efficiencies which will help to progress the solid state lighting revolution.

3.6. Device outlook

In this chapter we evaluated the design of specific arrangements of GaN nanowires on Si

substrates. To fabricate such structures requires control over GaN nanowire dimensions (lengths

and effective diameters), areal nanowire densities and most importantly, vertical alignment

relative to the substrate. Furthermore, excellent optical and electrical properties must be

maintained while achieving this control over the nanowire synthesis. Practically, the largest

challenge in fabricating such devices is synthesizing vertical or even aligned m-directional

nanowire growth on Si. Such alignment likely requires an epitaxial relationship between the

59

nanowire and the substrate; however growth of m-directional GaN nanowires on Si resulted in

growth with no particular alignment to the substrate, despite growing the nanowires with a

variety of seed-particle metals at various growth conditions (see Appendix A). Furthermore, the

GaN nanowires grown via both CVD and MOCVD in our growth system demonstrated very low

luminescence; especially relative to ZnO nanowires (see Section 5.1. Optical properties of ZnO

nanowires). The poor optical properties of our GaN nanowires and their potential sources are

discussed further in Appendix A. To continue on a path towards the fabrication of nanowire-

based LEDs, ZnO nanowires were pursued as an alternative to GaN. The simulation work

performed in this chapter could then be reiterated upon potential designs for ZnO nanowires,

though this is left for future work. Thus, in the next chapter, we discuss the synthesis of ZnO

nanowires on Si towards the fabrication of nanowire-based LEDs, with an emphasis on control

over the nanowire vertical alignment relative to the substrate and the uniformity of nanowire

length: two critical properties for scalable nanowire-array device fabrication.

It should be noted here that these simulation results may also be able to be loosely extended

to ZnO nanowire-based devices with CdZnO QWs. The QCSE also has a large effect in c-plane

CdZnO QWs[160]. It has also been theoretically shown that due to the same effect of an energy

gap between different polarization states as seen in InGaN, that polarization anisotropy of the

emission should be present for m-plane CdZnO[161], however not to the same extent as InGaN,

which allows for a potential improvement in directed extraction from c-directional radial

heterostructures relative to InGaN QWs. Obviously this simulation work could easily be repeated

for the case of ZnO-CdZnO nanowire-based LED designed as CdZnO is further developed and

understood.

This chapter resulted in the publication of a conference proceeding and a journal article:

• Chesin J, Zhou X and Gradečak S 2012 Light extraction in individual GaN nanowires on

Si for LEDs Proc. of SPIE 8467 846703 1-10.

• Chesin J and Gradecak S 2014 Comparing directed efficiency of III-nitride nanowire

light-emitting diodes Journal of Nanophotonics 8 083095.

60

Chapter 4: ZnO Nanowire Synthesis on Si

This chapter demonstrates the relative ability to control ZnO nanowire vertical alignment

relative to the growth substrate, nanowire length uniformity and nanowire areal density for ZnO

nanowire arrays grown via different techniques and compares the applicability of these

techniques as a realistic device fabrication step. First, HTS was used to produce ZnO nanowires

on Si and polyethylene naphthalate, which are both platforms for scalable device fabrication;

however, this technique limits the control over important nanowire properties relevant for

nanowire array-based LEDs, such as the nanowire areal density and high optical quality. To

achieve better control over the nanowire properties, ZnO nanowires were next synthesized via

VTC using Au seed-particles on sapphire. ZnO nanowires with controlled diameter and

morphology were synthesized using this method, but because Au is known to quench

luminescence in ZnO, the presence of the foreign seed particles is a potential disadvantage of this

approach. To overcome these issues, ZnO nanowires were synthesized on Si substrates via VTC

using a solution-processed ZnO seed-layer film, without any foreign metal seed-particles. The

oxygen partial pressure was determined to be critical to tuning the nanowire areal density and the

specific growth kinetics, from a continuous nucleation regime at lower nanowire areal densities

around 1 nanowire/µm2 (producing nanowires with a fixed aspect ratio) to a surface-diffusion

limited regime at higher areal densities around 10 nanowires/µm2 (producing nanowires with

length inversely proportional to the nanowire diameter). Nanowires with uniform lengths were

achieved via a two-stage growth method by first forming nanowire bases followed by growth

under the surface-diffusion limited regime. In the surface-diffusion limited regime, the growth

rate is roughly independent of diameter at larger diameters, thus by defining large diameter bases

in the first stage of the two-stage growth method nanowires with uniform lengths were achieved.

To further control the nanowire areal density and maintain the uniform nanowire lengths, sparse

seed layers were used. With control of ZnO nanowire areal density, nanowire length uniformity

and vertical alignment relative to the Si substrates, these nanowire-arrays are ideal for pursuing

the fabrication of nanowire-based LEDs.

61

4.1. Nanowire growth mechanisms

In this work, ZnO nanowires were synthesized via HTS and VTC methods. For HTS

synthesis the growth proceeds according to an anisotropic growth method on a ZnO seed-layer

processed on Si substrates. The substrate is submerged in an aqueous solution containing Zn and

O ions to supply the nanowire growth as demonstrated by the schematic in Figure 1-1Figure

4-1(a). For ZnO nanowire synthesis using the VTC method, a hot-walled three-zone quartz-tube

furnace was used, as demonstrated by the schematic in Figure 4-1(b). VTC synthesis of ZnO

nanowires were conducted using Au seed-particles via the VLS mechanism or on ZnO seed-

layers via anisotropic growth. The growth mechanisms are discussed in further detail below.

Figure 4-1. ZnO nanowires synthesis methods. Schematics of the setup for ZnO nanowire

synthesis using (a)HTS at relatively low temperatures of 85ºC and (b) VTC at higher

temperatures from 800-1000ºC.

4.1.1. VTC growth with seed-particles

ZnO nanowires are often grown via the VTC method, which takes advantage of the

carbothermal reduction of ZnO powder at much lower temperatures (~900°C) than ZnO typically

decomposes (~1,975°C). The reaction of ZnO with C powder acts as source for Zn vapor. The

specifics of this reaction are discussed further in Section 4.4.1. Density Control via O2

Flow/Seed Layer Thickness. The vapor-liquid-solid (VLS) mechanism is a typical seed-particle

mediated method used to grow nanowires and has been used extensively to grow nanowires with

Au particles. Au is considered a noble metal resistant to corrosion and oxidation and is thus ideal

62

for use as a seed-particle in VLS growth of nanowires, used to fabricate Si[162, 163], GaAs[164,

165], GaN[25, 26] and particularly ZnO nanowires[54, 100], as well as many others.

Figure 4-2. VLS growth mechanism schematic. VLS mechanism of nanowire growth as

described by a Si-Au phase diagram with the growth temperature denoted as Tgrowth and the

respective stages of growth indicated from 1-4: solid Au seed (1), a liquid Au-Si seed formed as

Si incorporates into Au (2), the first Si monolayer nucleates at the liquid-solid interface with the

substrate once the Au-Si alloy supersaturates with Si (3), and nanowire growth continues at the

liquid-solid interface (4).

A classic example of VLS growth is the growth of silicon nanowires via CVD using a gold

seed particle, as demonstrated in Figure 4-2. Gold nanoparticles are dispersed on the surface of

the substrate. The temperature is elevated to above the Si-Au eutectic point and silane (SiH4) is

flowed as a Si precursor. At these temperatures SiH4 decomposes to H2 and Si vapors. The Au

seed particle absorbs Si, expanding in volume until it is supersaturated, at which point solid Si

nucleates preferentially at the liquid-solid interface between the substrate and the liquid seed

particle[166]. The nucleation and absorption of Si into the seed-particle reach a steady state, such

that the volume of the seed particle is constant and determines the radius of the nanowire. As the

growth progresses, nucleation continues to occur at the liquid-solid interface, which is between

63

the last monolayer of Si formed and the seed-particle, thus creating a pillar of Si with the seed-

particle on top. In this technique, the diameter is directly related to the size of the seed-particle

and the length of the nanowire is determined by the growth time[167]. In compound

semiconductors, such as ZnO and GaN, the growth can be more complicated, for example,

because N is not soluble in Au, it only reacts with the III-precursor via the triple-phase boundary

between the liquid seed-particle, solid nanowire and vapor phase[25].

In addition to acting as a seed particle during the VLS growth of ZnO nanowires Au can also

act – depending on the oxygen overpressure – as a collection site for the vapor source. In this

growth regime, substantial film growth on the substrate can also occur, drawing reactants from

the Au particles. The collection of vapor source from the seed-particles in combination with the

film-growth results in a network-like film with exposed c-plane facets. Nanowires can then grow

via an anisotropic vapor-solid (VS) mechanism, since the c-facet grows anisotropically faster

than the m-facets[54], forming nanowires with a c-plane top facet and m-plane sidewall facets. It

has been previously shown that Au can transition from acting as a seed-particle in ZnO nanowire

growth to passively forming a particle on the nanowire sidewall, presumably due to the surface

facet stability of the Au particle changing as oxygen partial pressure decreased[54]. Similarly,

this transition from Au-assisted to non-assisted anisotropic VS growth has been observed as a

function of temperature, where at higher temperatures the Au particle is more likely to be liquid

and easily absorb Zn precursors to alloy toward the eutectic composition[168].

4.1.2. Anisotropic vapor-solid growth

The anisotropic VS growth of ZnO nanowires has been observed to form c-directional

nanowires using many growth methods, implying that under most conditions the polar c-plane

facet has the highest surface energy relative to other facets, and thus grows most quickly.

Particularly, this has been observed in HTS of ZnO and ZnS crystals[23], for CVD at low-

temperatures (140°C)[101] and even for CVD-grown nanowires at temperatures as high as

1200°C[169]. Usually, growth is seeded with a textured ZnO polycrystalline film (the seed-

layer) with each small grain oriented roughly in the c-direction[30, 100, 170], though under the

proper conditions even the seed-layer is not necessary since the growth results in its own island-

like film which acts as the seed-layer[101, 171]. It has been shown that the polarity of the

exposed seed-layer facet can be critical to growth. In most vapor-source methods only Zn-

64

terminated surfaces will be chemically active and result in nanowires, though in some growth

methods, such as chemical bath deposition, both Zn and O-terminated surface resulted in

nanowires[104]. In general, the growth method is assumed to be the anisotropic growth of c-

planes which eventually expose only m-plane nanowire sidewall facets that grow much slower.

The role of higher order 1011 is less clear since they seem to only form upon nanowire

cooling if source is still available[102], indicating that they may grow faster even than c-plane

facets at high temperatures.

The specific nanowire growth kinetics are relevant to work discussed later in this chapter.

Typically in VLS-grown nanowires limited by the diffusion of source precursors, the Gibb’s-

Thompson effect can play a large role at low nanowire diameters. The curvature of the smaller

seed-particles decreases the effective chemical potential difference between the liquid and vapor

phase and thus the nanowires with smaller diameters grow more slowly[172]. Eventually, the

effect of needing more atoms per monolayer in thicker nanowires results in an inverse

dependence of growth rate on nanowire diameter, overcoming the Gibb’s Thompson effect. The

Gibb’s Thompson effect is not observed in anisotropic VS-growth, since there are no seed-

particles used to mediate the growth. Typically the growth rate follows a ~1/d relationship in the

anisotropic VS growth of ZnO nanowires[173], which indicates a surface-diffusion limited

regime in which the growth rate of nanowires with smaller effective diameters grow much faster

than thicker nanowires.

4.2. Hydrothermal synthesis of ZnO nanowires

4.2.1. Hydrothermal synthesis of ZnO nanowires on Si

HTS is commonly used to grow ZnO nanowires on a variety of substrates at low

temperatures and was thus used as a starting point for ZnO nanowire synthesis on Si. Nanowire

vertical alignment relative to the growth substrate is generally good in HTS ZnO nanowires and

the nanowire lengths tend to be fairly uniform, both of which are properties of nanowire arrays

necessary for the fabrication of nanowire array-based LEDs, which makes the HTS growth

method an excellent starting point towards the fabrication of such devices. First, 0.5×0.5 in2

squares of p-Si (100) substrates were cut by die-saw and sonicated in acetone, methanol and

deionized (DI) water immediately prior to seed-layer deposition. Zinc-acetate dihydrate,

Zn(O2CCH3)2(H2O)2, was dissolved in methanol at a concentration of approximately 10 mM.

65

The zinc-acetate solution was spun onto the cleaned substrates at 3000 revolutions per min (rpm)

for 10 s. Each coating was performed 10 times and annealed at 350°C in air for 30 min. As-

prepared substrates were affixed to a glass slide using double-sided tape on the back side of the

substrate, and single-sided tape on two edges of the substrate wrapped around the glass slide.

The slide was placed into a glass jar containing the growth solution (see Figure 4-1(a)),

composed of 100 mL of 10 mM zinc-nitrate (Zn(NO3)2) and 4 mL of ammonium hydroxide

(NH4OH). The jar was sealed with a lid and placed into a box furnace at 85°C for 2-6 h. After the

growth, substrates were rinsed in DI water and gently blown dry with compressed air. The

amount of white precipitation, indicative of high supersaturation in the growth solution in the

growth solution causing homogenous nucleation of ZnO, was qualitatively noted as growth

parameters were changed.

Figure 4-3. Hydrothermally synthesized ZnO nanowires. SEM images of ZnO nanowires on

a ZnO seed-layer on (100) Si: (a) 45° tilted view and (b) top-down plane-view.

Representative HTS-grown ZnO nanowires, imaged using scanning electron microscopy

(SEM), are shown in Figure 4-3. Though lengths of the nanowires are uniform, the top-down

view in Figure 4-3(b) demonstrates the large nanowire diameter distribution, which is further

illustrated in Figure 4-4. The diameter of HTS-grown nanowires has been shown to depend on

the seed-layer preparation conditions[27, 103, 174] and the thickness/grain-size of the seed-layer

film[27, 170], though achieving diameter uniformity with these methods has not been

demonstrated. To control the nanowire optical properties (such as the confinement as an optical

waveguide), control of the nanowire diameter uniformity is critical. To further assess the

potential of HTS ZnO nanowires on Si for nanowire array-based LEDs, we next look the

66

potential for controlling nanowire length and length uniformity in HTS ZnO nanowires. Control

over nanowire length is necessary for implementing nanowire-based LED designs, such as the

placement of a QW region or a p-n homojunction along the length of the nanowire.

Figure 4-4. Nanowire diameter distribution of HTS-grown ZnO nanowires. The histogram

demonstrates the large distribution of nanowire diameters, with most nanowires (~50%) from 60

to 120 nm in diameter. The median nanowire diameter is 100 nm, near the peak in the

distribution, and the standard deviation is 65 nm.

It is critical to control nanowire lengths for many applications. For most bottom-up methods

of nanowire synthesis, the nanowire length can be controlled by adjusting the growth time. HTS-

grown ZnO nanowires are limited in length by the depletion of precursor source as the reaction

continues. The nanowire length does not increase substantially after two-hour growth times, as

demonstrated in Figure 4-5. After this period, enough of the Zn ions in solution have either

incorporated into the nanowire or precipitated separately in solution, such that the solution is no

longer supersaturated thus reducing the nanowire growth driving force. Though nanowires of this

length (~2 µm) may be adequate for nanowire array-based LEDs, the flexibility to grow longer

nanowires allows for the possibility of nanowire lasers and four-point electrical characterization.

Attempts at growing longer nanowires by using a higher concentration (23 mM) of zinc nitrate in

the growth solution resulted in shorter nanowires and more precipitation in solution than growth

with lower concentrations (10 mM), likely because higher concentrations favor homogenous

67

nucleation of ZnO particles in solution (i.e. precipitation) which leaves less precursor for the

nanowire growth.

Figure 4-5. HTS ZnO nanowire length at varying growth times. The ZnO nanowire length is

mostly unchanged as a function of growth time. The error bars represent standard deviation and

the insets are cross-sectional SEM images of the nanowire growths.

The HTS-grown ZnO nanowires are hexagonal in cross-section (see Figure 4-3(b)) and retain

a uniform diameter along the length of the nanowire. The hexagonal cross-section indicates c-

directional nanowire growth, which is commonly observed for ZnO nanowires[95, 174]. While

control over the average nanowire diameter can be tuned with the grain size of the seed-layer

film[27, 170], achieving control over diameter uniformity, the nanowire length and other

nanowire properties along the length of the nanowire would require the development of a

continuous flow reactor (which is currently in development in our lab). Furthermore, control of

the nanowire areal density is critical for the development and fabrication of optoelectronic

devices, which is a key disadvantage of HTS. While some control over the density is possible by

tuning the seed-layer, all HTS ZnO nanowire areal densities seen in this work are greater than 30

nanowires/µm2. Other techniques have been developed to control the areal density in HTS.

Seedless growth on a 50 nm thick Au film has been used to demonstrate control over the

68

nanowire areal density[175, 176], but this approach is not ideal for many applications, due to the

presence of the Au film. Very thin (< 3 nm) seed-layer films have also been used to tune the

areal of density of HTS ZnO nanowires[177]; however, in this case the control of the aspect

ratios becomes dependent on the nanowire density. In brief, while HTS of ZnO nanowires on Si

can be used to attain uniform nanowires lengths and decent vertical alignment relative to the

growth substrate, the lack of control over diameter uniformity, nanowire lengths above 2 µm and

nanowire areal density limit their applicability for the fabrication of nanowire array-based LEDs.

4.2.2. Hydrothermal synthesis of ZnO nanowires on flexible substrates

The largest factor that differentiates hydrothermal synthesis of ZnO nanowires from other

growth methods is the ability to grow at low temperatures. Low temperature growth allows for

the use of substrates that normally would not be stable at temperatures used for other synthesis

methods, such as flexible plastics. We have successfully demonstrated the growth of ZnO

nanowires on polyethylene naphthalate (PEN) using these methods. The ZnO seed-layer was

spun-on to PEN substrates using similar methods as for Si substrates by affixing the plastic to a

rigid substrate using double-sided tape. To maintain the structural integrity of PEN, the seed-

layer anneal had to be lowered to a temperature of 200°C, below the glass-transition temperature

of PEN. Since the seed-layer occurred at lower temperatures, the annealing time was extended to

one hour (from 30 min for seed-layers on Si). The results are similar to the typical results on Si

substrates and are shown in the SEM image in Figure 4-6. The nanowires on flexible substrates

can be useful in a range of applications, such a piezoelectric nano-generators or large area

processing; however, the photoluminescence from these nanowires indicates large defect

luminescence (as discussed in Chapter 5: ZnO Nanowire Optical Properties). To develop control

over nanowire areal density and the optoelectronic properties necessary for LEDs, the use of

high temperature methods is considered for the remainder of this chapter.

69

Figure 4-6. SEM of HTS ZnO nanowires on PEN. The nanowire morphology on flexible PEN

is similar to the morphology observed on rigid Si substrates, demonstrating the ability to

synthesize nanowires directly onto flexible substrates.

4.3. VTC ZnO nanowire synthesis with Au seed-particles

In this section we evaluate how to use Au to control the nanowire areal density and the

distribution of nanowire diameters in VTC and discuss the trade-offs between using Au seed-

particles and other methods to synthesize ZnO which leave no Au particles behind. The synthesis

of ZnO nanowires discussed in this section was conducted in a three-zone furnace held at a

pressure of 210 Pa. The source precursors were ZnO and C (graphite) powders at a mass ratio of

1:1. 1 g of the precursor powder was used in a round crucible in the central zone of the furnace

which was held at 950°C during growth. Substrates were placed downstream in the furnace end-

zone, face up on a plate within an inner quartz-tube and held at 930°C. 73.5 sccm of Ar was

flown as the furnace was heating up, during growth and during cool-down. O2 was only flown

during growth at 1.2 sccm, unless otherwise noted. The growth conditions were held for 20 min,

unless otherwise noted.

4.3.1. Nanowire morphology shifts with seed-particle size

The first step in controlling the morphology of nanowires is to demonstrate control over the

nanowire diameter, which in VLS growth can be tuned via the size of the seed-particle. In our

case, Au seed-particles were prepared by first depositing Au films onto the substrates via

70

electron-beam evaporation. The thickness of the film is estimated by the quartz crystal monitor

in the evaporation chamber. For this study, 0.5 nm, 1 nm, 5 nm and 10 nm films were deposited

on a-plane sapphire substrates after substrates were cleaved and cleaned via sonication in

acetone, methanol and DI water. Growth was then carried out via VTC under the typical

conditions described above. Upon annealing to the high temperatures and introducing the vapor

precursors, the Au film becomes liquid and de-wets, forming droplets on the substrate surface.

These droplets act as the seed-particle for VLS growth. The distribution of diameters observed

shifted towards higher diameters with thicker Au films, as demonstrated by Figure 4-7. The

distribution of nanowire diameters also widened as the Au-film thickness increased, indicating a

wider range of particle sizes upon the Au-film de-wetting. While a narrower size distribution

may be possible using colloidal Au particles, these results demonstrate the ability to tune

nanowire diameter with the size of the seed-particle, consistent with other reports[100].

In addition to the diameter modulations, a density decrease with increasing Au-film thickness

was also observed from these growths. Figure 4-8 shows that the nanowire density produced

from 10 nm thick Au films is only half of the areal density produced from 0.5 nm thick Au films.

As thicker films are heated and the Au de-wets, the density of Au droplets that form is expected

to decrease[178]. To assess the approximate number of seed-particles which result in nanowire

growth, we assume that the seed-particles form spherically at a diameter-distribution equal to

nanowire diameter-distribution. Only about 6% of the Au droplets result in nanowire growth for

the 10 nm Au film based on this assumption and assuming the volume deposited to be equivalent

to a uniform 10 nm film. This assumption provides an estimate of the proportion of Au droplets

that result in nanowire growth, as shown against the right vertical axis in Figure 4-8. The

proportion of nanowires formed from the expected number of Au droplets decreases more

dramatically for the thinner than the thicker films, unlike the linear decrease observed in

nanowire densities.

71

Figure 4-7. ZnO Nanowire diameters as a function Au film thickness on a-sapphire.

Histograms of the diameters observed for VTC-grown ZnO nanowires on a-plane sapphire with

(a) 0.5 nm, (b) 1 nm, (c) 5 nm and (d) 10 nm Au-films. As the Au-film thickness increases, so

does the distribution of nanowire diameters. SEM images at 30°-tilt are shown in the inset.

The large drop in nanowires per expected Au droplets likely coincides with the transition of

the Au film forming individual droplets upon heating to forming larger merged particles. This

effect is known to occur at Au thicknesses larger than 1 nm for Au films on Si[178], which

corresponds well with the observation of a steep decrease in the proportion of nanowires. In a

diffusion-limited growth regime, the larger seed-particles formed from merged Au-seeds would

grow at a much slower rate and would effectively not result in nanowire formation. Thus, the

decrease in density for thicker Au-films can be explained by more of the Au forming large

particles that do not result in significant nanowire growth.

72

Figure 4-8. Nanowire areal density decrease with increasing Au-film thickness. The areal

density ( - along the left axis) of ZnO nanowires grown on a-plane sapphire with Au is shown

to decrease as a function of Au-film thickness deposited. The number of nanowires per expected

Au droplets ( - along the right axis) decreases much more rapidly than the density.

To observe the effect of larger Au particles at the substrate, a shorter nanowire growth of

only 10 min was conducted using 1 nm Au film, so that nanowires were still short enough to

image the underlying substrate. The SEM image (Figure 4-9) shows many large Au particles at

the nanowire base, as expected in the above analysis. The average nanowire diameter for this

shorter growth was about 56 nm, while the Au particles observed at the substrate were much

larger, averaging about 83 nm in diameter. Thus, our results indicate that the observed decrease

in nanowire areal density with increasing Au-film thickness arises due to the thicker Au-films

forming merged Au droplets upon de-wetting which are too large for nanowire growth relative to

the smaller unmerged Au droplets. It should be noted that the density could be more directly

controlled by using colloidal Au particles, however this approach requires functionalization of

the Au colloid particles[179].

73

Figure 4-9. Large Au particles at substrate in ZnO nanowire growth. The large Au particles

at the base relative to the smaller diameter nanowires indicates that merged Au droplets at the

substrate do not result in significant nanowire growth, which explains the decreasing density

trend observed with increasing Au-film thickness.

4.3.2. Discussion of using Au for synthesis

Initial growth attempts of ZnO nanowires on Si (111) substrates utilized Si substrates which

were cleaved and cleaned prior to Au deposition. Immediately prior to the 1 nm Au deposition,

the Si was stripped of any native oxide in hydrofluoric acid, (HF). One of such substrates, shown

in Figure 4-10(a) was compared with a substrate that was again stripped of any potential oxide

formation during the Au deposition by an additional HF treatment immediately before nanowire

growth to strip any oxide over-layer[162] which may have formed during the Au deposition,

shown in Figure 4-10(b). A cleaved and cleaned Si substrate with the native oxide was treated

with poly-L-lysine and 100 nm Au colloids immediately prior to growth, shown in Figure

4-10(c). None of the substrates yielded nanowire growth, though the beginning stages of

nucleation were evident in all cases. A similar growth mode in which the Au particles act as a

collection site but not as a seed-particle for growth has been previously reported, leaving globs of

Au-Zn alloy at the base of each nanowire[180]. Since Zn adatoms have a limited mobility on Si

substrates, the nanwires can only form within the Au particle and result in poor vertical

alignment relative to the growth substrate, which is not ideal for the fabrication nanowire array-

based LEDs.

74

Figure 4-10. Growth attempts of ZnO nanowires on Si with Au seeds. SEM images at 45°-tilt

of a typical ZnO nanowire growth performed on (a) bare Si(111) with 1 nm Au film, (b) bare

Si(111) with 1 nm Au film, HF treated immediately prior to growth and (c) native Si with 100

nm Au colloids.

In summary, ZnO nanowires using Au seed-particles have been synthesized on a-plane

sapphire with control over both the nanowire diameter and areal density, but leaving large Au

particles at the substrate. On Si, initial results indicated that a large amount of Au would be left

at the nanowire base under the given growth conditions. We observed growth of nanowires via

the VLS mechanism; however similar growths of ZnO nanowires via VTC on Au-coated a-plane

sapphire are known to also result in VS growth. In fact, previous work from our group by

Brewster et. al. showed that the oxygen partial pressure controls the role of the Au-particle,

which can transition from enacting VLS growth to becoming a passive Au particle on the

nanowire sidewall as the oxygen partial pressure drops during the growth[54]. Furthermore, this

work observed in cathodoluminescence that Au particles quench luminescence at nanowire tips

and nanowire sidewalls via the transfer of excited electrons over the Schottky barrier formed at

the Au/ZnO interface, thus the particle-assisted growth method is likely deleterious for the

optical properties of the ZnO nanowire arrays. Au is known to form a deep-level trap in both n-

type and p-type Si[181], which can be detrimental to electrical properties. Thus, while Au seed-

particles can allow for a great deal of control of the nanowire morphology[179], there are many

aspects of using Au to grow ZnO on Si that could negatively affect the final device performance.

To avoid the negative effects of using Au, we next consider the growth of ZnO nanowires on

ZnO seed-layer film via VTC.

75

4.4. VTC ZnO nanowire synthesis on silicon with continuous ZnO seed-layers

In this section, we discuss the use of a ZnO seed-layer that allows for the growth of ZnO

nanowires on Si via VTC, without the need for Au, which has the disadvantages previously

discussed. The seed-layer employed in this work is similar to the seed-layers typically used in

HTS (discussed in Section 4.2. Hydrothermal synthesis of ZnO nanowires), however VTC allows

a greater degree of control over the nanowire morphology, and the high temperature growth

provides nanowires with superior optical properties (see Chapter 5: ZnO Nanowire Optical

Properties). We demonstrate that by tuning the seed-layer film thickness and the oxygen flow,

the nanowire areal density can be tuned. There are two distinct regimes of nanowire growth,

depending on the growth parameters, in which the nanowire length is either directly or inversely

proportional to the nanowire diameter. Taking advantage of these two growth regimes, a two-

stage growth is developed to produce nanowires of uniform length.

Prior to discussing the effect of the different growth parameters on ZnO nanowire

morphology, it is important to discuss reproducibility in the system. Figure 4-11 demonstrates

previous growths using Au seed-particles on a-plane sapphire, where the growth represented by

Figure 4-11(a) is synthesized under the same nominal conditions (pressure, flow rate and

temperature) as the growths represented in Figure 4-11(b) and (c), however the latter growths

occurred in succession (i.e. – after running that same growth multiple times), whereas the growth

in Figure 4-11(a) was the first growth conducted after the system had last been used for lower

temperature processes. In the initial growth after the low temperature processes most nanowires

are growing inclined from the substrate, representing a different epitaxial relationship with the

substrate or a different crystallographic growth direction than the vertical nanowires seen in

Figure 4-11(c). As the growth is repeated, the results shift towards the expected vertical

nanowire growth. This trend is likely to a decreased effective O2 flow in the first few growths

relative to the later growths, due to effects from a coating on the quartz tube. After a growth, as

the quartz tube cools to room temperature, the precursor sources adsorb to the inside of the

quartz tube, and in various zones of the tube a coating is developed (likely ZnOx). To achieve

reproducibility, the coating should be in the same state (composition and thickness) prior to

every growth. To achieve this, the coating within tube is oxidized prior to growth.

76

Figure 4-11. Demonstration of shifting reactor conditions. SEM images (at a 45° tilt) of the

same growth after the system had been used for lower temperature processes: (a) initial growth,

(b) after three repetitions and (c) after four repetitions. The initial growth contains very few

nanowires growing vertically relative to the substrate. As the number of growth runs increases,

the growth approaches consistently vertical nanowires.

Figure 4-12. The role of oxidation in reproducibility. SEM images of repetitions of the same

growth: (a) initially after tube oxidation (45°-tilt), (b) the third growth after the initial oxidation

(top-down) and (c) immediately after another tube oxidation (45°-tilt), demonstrating the ability

of the tube oxidation to mitigate the effects of shifting reactor conditions.

To confirm the role of the oxidation, the same growth (5 sccm O2 flowed for growth on 5

coatings of the seed-layer on Si, as described below) was performed sequentially after the

oxidation of the furnace tube was performed with 55 sccm of O2 flowing at 950°C for one hour.

The first growth immediately after oxidation resulted in dense nanowire growth (Figure 4-12(a)),

but by the third growth after the oxidation, the growth yielded no nanowires (Figure 4-12(b)).

77

Confirming the role of the oxidation, similar results to the first growth were achieved by

performing another tube oxidation before growth (Figure 4-12(c)). Thus, to achieve reproducible

results and mitigate the effect of shifting reactor conditions due to the buildup of film deposition

within the tube, an oxidation was performed prior to each growth reported in this section.

ZnO nanowires discussed in this section were synthesized first by preparing a ZnO seed-

layer on Si (111) substrates. ZnO seed-layers were prepared by spin-coating a 0.3 M zinc acetate

solution, prepared by dissolving zinc acetate dihydrate in 2-methoxyethanol and adding 18 µL of

ethanolamine per mL of solution to encourage dissolution of the zinc acetate. The solution was

mixed in a glove box with a nitrogen environment and heated on a hotplate at 100°C for 30 min,

followed by sonication for 30 min. Finally the solution was filtered through a 0.1 µm PTFE filter

to remove any additional undissolved zinc acetate. The solution was spun on to die-sawed ½ inch

by ½ inch Si (111) pieces that had been cleaned via sonication in acetone, methanol and DI

water, followed by a chemical clean in piranha solution composed of H2SO4:H2O2 (2:1) for at

least 10 min. Immediately after these cleaning steps, an additional rinse in DI water and drying

via blown compressed air, the substrates were used to spin-on the zinc acetate solution at 3500

rpm for 60 s followed by an anneal on a hotplate in dry air at 175°C-250°C. The spinning and

anneal steps were repeated from 2-6 times to create a thicker seed-layer film. Growth was

typically conducted at 950°C (unless otherwise noted) at a pressure of 350 Pa in the central zone

of the furnace, with a boat-shaped crucible used to contain the 1 g of ZnO:c (1:1) powder and

substrates mounted seed-layer side down directly on the crucible. 73.5 sccm Ar was flowed

during heating, growth and cooling stages, while a varied amount of O2 was flowed only during

the growth stage, which was 20 min unless otherwise noted.

4.4.1. Density Control via O2 Flow/Seed Layer Thickness

The seed-layer film thickness (as controlled by the number of coatings during substrate

preparation) and the oxygen flow during growth are found to have a large role in determining the

nanowire areal density. Increasing the oxygen flow increases the rate of reduction of the ZnO

powder. It is known that the carbothermal reduction of ZnO powder is most effective via CO, as

opposed to just solid graphite[182, 183]. Thus the reaction takes the route:

78

ZnO() + CO() → Zn() + CO () Equation 4-1

The products of the reduction then create more CO to further feed the reduction:

CO () + C() → 2CO() Equation 4-2

Figure 4-13. Ellingham diagram of relevant reactions in ZnO reduction. The Gibb’s free

energy of reaction is plotted against temperature for the relevant reactions to ZnO reduction,

demonstrating that ZnO reduces via CO, which initially need O2 to form from C powder.

Compiled from [184, 185].

It is informative to consult the Ellingham diagram of the relevant oxides (Figure 4-13) when

considering the reaction. CO has a lower Gibb’s free energy than CO2 at temperatures used for

growth, while the Gibb’s free energy of ZnO is even lower, thus the reaction likely proceeds via

CO formed with O2 flow:

79

2C() + O () → 2CO() Equation 4-3

Since CO is not available to reduce the ZnO powder until oxygen is introduced, the growth does

not effectively start until oxygen flow is initiated. Thus, increasing the oxygen flow, should

increase the CO available for the reduction of ZnO, and thus increase the effective Zn source

flux to the growth substrate. The overall equation for the formation of Zn vapor can thus be

written as:

ZnO() + 3C() + O () → Zn() + 3CO() Equation 4-4

Figure 4-14. Nanowire areal density dependence on seed-layer thickness and O2 flow. SEM

images (at 45° tilt) of ZnO nanowires grown on ZnO seed-layers of varying thickness with

varying O2 flow, demonstrating that increasing either results in increasing nanowire areal

density.

80

Increased seed-layer thickness and oxygen flow during growth increase the resulting

nanowire areal density, as summarized in Figure 4-14. It has been shown in sputtered ZnO seed-

layers on Si used in VTC growth, that increasing the seed-layer thickness increases the

continuity of the seed-layer film, providing more area to seed nanowire growth; however past the

point of a continuous film (~70 nm) the density is no longer increased with thicker seed-

layers[186]. The seed-layer produced with 6 coatings has been measured to have an approximate

thickness of 90 nm, indicating that it is likely a continuous film, which explains why there is not

a large difference in the areal density between 5 coatings and 6 coatings. Since 6 coatings

guarantees a continuous film, from this point on all of the substrates used are coated 6 times.

Figure 4-15. Fine-tuning nanowire areal density with O2 flow. The nanowire areal density is

plotted as a function of the O2 flowed during growth, demonstrating a control over the nanowire

density from sparse growth at 0.8 sccm (~0.4 /µm2) to dense growth at 1.1 sccm (~8.2 /µm2).

The effect of increasing the oxygen flow is more dramatic and also allows for more control

than the seed-layer thickness. The nanowire areal density can be fine-tuned by changing the O2

flow for growths between 0.8 and 1.1 sccm. SEM images of the increasing areal density with O2

flow in this range are shown in Figure 4-15. Low areal densities of 0.4 nanowires/µm2 are

achieved at 0.8 sccm O2, intermediate areal densities of 7 nanowires/µm2 are achieved at 0.9

81

scmm O2 and higher densities (and nanowire growth rates) at which nanowires begin to become

entangled occur at 1.1 sccm O2. The next section discusses the growth kinetics and explains the

effect of O2 flow on the rate of nanowire nucleation and growth.

4.4.2. Continuous nucleation growth regime at low nanowire areal density

The first step to understanding the role of O2 flow in tuning the nanowire areal density is to

understand how the nanowire morphology evolves throughout the growth. Figure 4-16(a)

demonstrates the linear relationship between the nanowire length and diameter, as measured

from SEM images, indicating a fixed aspect ratio. Similar trends are observed in the

hydrothermal growth of ZnO nanowires on ZnO seed-layers[95], in which case the top c-plane

facet and sidewall m-plane facets grow with a fixed rate, where the c-plane facet has a higher

growth rate due the high surface energy from both dangling bonds and the inherent polarity in

the crystallographic c-direction[23, 102, 187]. At these growth conditions (0.9-1.4 sccm O2 flow)

the aspect ratio is found to be 35, i.e. – the c-facet grows about 35 times faster than the m-facets.

As the oxygen flow increases for growths performed for 20 min, the nanowires tend to have both

larger diameters and lengths. As previously discussed the oxygen flow increases the Zn source

flux and thus the super-saturation of Zn in the vapor phase, increasing the driving force for

forming ZnO on the substrate (i.e. – increasing the nanowire growth rate of both facets).

The growth evolution over time was investigated by performing a growth series with a

varying growth window time (8, 12 and 20 min), as shown in Figure 4-16(b). These results show

the same linear relationship as those in Figure 4-16(a), indicating that the aspect ratio is still 35.

However, the shorter growths result in relatively uniform nanowires such that only short and thin

nanowires are observed whereas a broader range of diameters and lengths are observed as the

growth time is increased. For example, after a typical 20 min growth, nanowires in the entire 750

nm – 7 µm length range are present. This result indicates that nanowires nucleate throughout the

growth and the shorter, thinner nanowires are formed near the end of the 20 min growth window,

while the longer, thicker nanowires were formed during the very beginning of the growth

window and the nanowire sidewall and top facets grow with a fixed relative rate. Such a growth

mode necessitates that nanowires can continuously nucleate. To understand the continuous

nucleation regime, it is helpful to observe how the nanowires nucleate and grow from the seed-

layer.

82

Figure 4-16. Linear dependence of nanowire length on nanowire diameter. Nanowire length

as a function of diameter for varying (a) O2 flow and (b) growth time indicate a roughly linear

dependence. In this growth regime the top and sidewall facets of the nanowire grow at fixed

relative rates, with the top-facet growing 35 times faster than the sidewalls. Increased O2 flow

increases these growth rates and the growth time series indicates a continuous nucleation regime.

Figure 4-17. Evolution of nanowire and seed-layer morphology with growth time. SEM

images of ZnO nanowires grown with 1.4 sccm O2 at typical growth conditions for (a) 8 min, (b)

12 min and (c) 20 min. Samples grown for more time contain longer and thicker nanowires and

the seed-layer film evolves from many rough grains to a coalesced film.

The series of SEM images in Figure 4-17 display how the nanowire and seed-layer

morphology evolve with growth time. For the 8 min growth in Figure 4-17(a), the seed-layer has

evolved into a rough film of grains with varying shapes. Between each nanowire and the

substrate there are hexagonal pyramidal bases. Many of the other grains in the seed-layer film

83

have similar morphologies to these pyramidal bases, indicating they may be potential sites for

nanowires which have not yet nucleated. As the growth time is increased the seed-layer film

eventually coalesces (see Figure 4-17(c)), indicating that these grains are also growing as the

nanowire growth progresses. The formation of the pyramidal bases can provide insight to the

growth mechanism of the nanowires. Such pyramidal bases have been observed in the literature,

but usually at different conditions. For example, much larger mounded pyramids at the nanowire

bases have been attributed to the formation of liquid Zn particles that oxidize throughout the

growth at 500°C[188]. The nanowires in this work are grown at much higher temperatures,

nominally above the boiling point of Zn (907°C), thus the oxidation of liquid Zn particles is not

likely the mechanism for pyramidal base formation here. From top-down SEM images of vertical

nanowires, it is observed that the pyramidal facets are inclined from the c-axis, but aligned along

the sidewall m-plane facet, indicating that they are higher order planes of the type110, as

shown in Figure 4-18. Such higher order facets have also been observed to grow at the tip of

ZnO nanowires during cool-down while sources are still flowing[102].

Figure 4-18. Pyramidal faceting observed at ZnO nanowire bases. The 101 type faceting

observed at the base of ZnO nanowires shown in (a) SEM and (b) in the schematic, where red

shaded 110 facets intersect the 1100 m-plane facets, shaded blue.

To understand the evolution of the seed-layer film and how it exposes the c-plane facets on

which nanowires nucleate, we must consider the relative growth rates of different facets. In the

growth of crystals, facets with the highest surface energies and thus the highest growth rates,

tend to grow themselves out of existence. In general, from theoretical work[187] and observed

84

ZnO structures[103], it is believed that the c-facet is the fastest growing low-order plane;

however higher order planes of the form110 as observed at the nanowire base may grow

faster than the c-facet, due to the high density of dangling bonds and the semi-polar nature of the

facet. It should be noted that these 110 facets are not observed at our nanowire tips, indicating

that under the growth conditions used in this work, the observed 110 facets grow faster than

the c-plane facet. Thus, these exposed 110 facets on the seed-layer film grow faster than the

c-facet to the point at which the top c-facet is exposed, allowing for the nucleation of nanowires

on the top c-facet of these pyramids.

Figure 4-19. Diameter of pyramidal c-facet and corresponding nanowire length. The

diameter of the pyramidal c-facet vs. the corresponding nanowire length for O2 flows of (a) 0.3

sccm and (b) 1.4 sccm showing the critical size of the c-facet necessary for nanowire growth.

The critical size decreases from ~60 nm to ~40 nm with the increased oxygen flow, due to the

increase driving force for island nucleation.

We observe that there is a critical size for the top-facet above which nanowires form (Figure

4-19), consistent with island nucleation on these c-plane facets. While such observations are not

widely reported for ZnO nanowire growth, similar island nucleation has been seen in the case of

plasma-assisted nanowire growth[189]. As oxygen flow is increased during the growth, the

critical size for the nucleation of nanowires decreases from ~60 nm at 0.3 sccm O2, as shown in

Figure 4-19(a), to only 40 nm for 1.4 sccm O2, as shown in Figure 4-19(b). This evidence

supports the theory of island nucleation, since the increase in O2 flow will increase the driving

85

force for island nucleation (by increasing the supersaturation of Zn vapor above the growth

substrate), which in turn decreases the critical nucleation size. Once an island nucleates on the

top c-plane facet of the pyramid, only low-order c-plane and m-plane facets of the island are

exposed, which then grow at their fixed growth rate in the reaction-limited continuous nucleation

regime. Since the nanowire density is low enough such that nanowires do not compete for source

adatoms, the pyramids can continue to grow as well as the nanowires throughout the growth. As

the higher-order facets of the pyramids continue to grow to expose c-plane facets, more

nanowires can nucleate, explaining the linear dependence of nanowire length on the nanowire

diameter: nanowires that nucleate earlier in the growth are longer and have larger diameters than

nanowires that nucleate later in the growth, though they maintain the same aspect ratios. This

growth regime is demonstrated schematically in Figure 4-20. However, as the density of

nanowires increases such that the nanowires are now competing for the flux of source adatoms,

the growth regime shifts from continuous nucleation in the reaction-limited regime to a diffusion

limited regime, discussed in the next section.

Figure 4-20. Continuous nucleation growth regime. Schematic of how nanowire growth

proceeds on continuous seed-layers via continuous nucleation: (a) the continuous seed-layer with

through-thickness grains of ZnO, (b) which facet and coarsen as the system is heated. When O2

flow begins, the faceted grains can grow as in (c) with 101 facets growing faster than c-plane

facets, allowing for nucleation to occur atop the pyramidal grain. Finally, (d) nanowire growth at

a fixed aspect ratio can occur due to the c- and m-facets growing at fixed rates.

86

4.4.3. Surface-diffusion limited growth regime at higher nanowire areal density

At higher effective oxygen flows, nanowire nucleation on the seed-layer films proceeds in

the same way as discussed above, but due to higher source flux and thus effectively lower

barriers to nanowire nucleation, more nanowires grow per a given area. At the growth conditions

used here (350 Pa) above 20 nanowires/µm2 the threshold areal density is reached, at which point

the growth transitions to a diffusion limited regime. Since more nanowires are growing per unit

area, a higher source flux would be necessary to maintain reaction limited growth. Instead,

nanowires are competing for source and the characteristic inverse correlation between growth

rate and diameter for the diffusion limited growth of nanowires[173, 190] is observed. To

characterize the growth regime more completely, a temperature series of growths was performed

at 930°C, 910°C, 890°C and 880°C, as shown in the SEM images in Figure 4-22(a)-(d),

respectively. In general it is observed that thinner nanowires grow more quickly, as fewer source

adatoms are necessary to grow each monolayer than for thicker nanowires. The relationship

between the nanowire lengths and diameters was analyzed, as shown in Figure 4-22(e) according

to a growth model in which the growth rate, ! ", is dependent on the arrival of source adatoms,

which depends on the collection area:

#$#% = Ω '"()*+,- + Ω './*+,- Equation 4-5

The arrival rate to the top facet Rtop and sidewall facets Rsw are characterized by the product of

the ZnO molecular volume, Ω, and the flux of source adatoms, J (in adatoms m-2 s-1), assuming

the same flux to the nanowire top and sidewall facets. Since we are concerned with the nanowire

length, Arxn (the reaction area) is the top nanowire facet, while Atop and Asw are the collection

areas on the top c-facet and the sidewall m-facets, respectively. Thus, the first term of Equation

4-5 represents the growth rate as contributed by direct flux onto the top nanowire facet, and the

second terms represents the growth rate of adatoms that diffuse along the nanowire side-wall

facets to the nanowire top-facet to react and growth the length of the nanowire. To simplify the

problem the nanowires are approximated as cylindrical, with diameter, d. The different physical

parameters of the model are schematically shown in Figure 4-21 and the top-facet collection and

reaction areas are equivalent:

87

"() = *+, = 0 '#2-

and the sidewall collection area depends on the surface-diffusion length of source adatoms, LSD:

./ = 0# ∙ $23 Equation 4-7

Integration with respect to time yields the form of the model fit to the data, where t is the growth

time (in s):

$ = Ω% '1 + 4$23# - Equation 4-8

Figure 4-21. Surface-diffusion limited growth model schematic. In the surface-diffusion

limited growth regime, adatoms which contribute to growth are collected on the top-facet

(shaded red) and within one diffusion length (LSD) of the nanowire tip on the nanowire sidewalls

(shaded blue). The gray area in the schematic does not collect adatoms which contribute to

nanowire growth.

88

Figure 4-22. Surface-diffusion limited growth of ZnO nanowires. Cross-sectional SEM

images of the growths used to analyze the surface-diffusion limited growth regime are shown for

(a) 930°C, (b) 910°C, (c) 890°C and (d) 880°C. The experimental results are displayed as

markers in (e), while the fit to the surface-diffusion limited growth model is represented by the

solid lines.

89

By fitting Equation 4-8 to the data presented in Figure 4-22(e) the surface-diffusion length

and effective source flux to the substrate at the various temperatures were calculated. Before and

after the nanowire growth, the source-powder was weighed to determine how much of it was

consumed during growth. The overall amount of Zn vapor flux (in adatoms m-2 s-1) from the

crucible, Jcrucible, was estimated by considering the oxygen induced formation of Zn vapor

(Equation 4-4), the crucible area and the growth time. To estimate the actual flux arriving at the

substrate a fitting parameter σ was used to represent the fraction of Jcrucible that actually reached

the substrate, such that the arrival flux of source adatoms from Equation 4-8 is defined as

= 5 ∙ 6*7689:;. Thus the only fitting parameters for the model were the surface-diffusion length

at each temperature and σ, assuming σ is temperature independent. It was found that σ was

approximately 0.0001, indicating only 0.01% of the source powder evaporated reached the

substrate.

It was found that both the diffusion length and effective source flux increased with increasing

temperature (Figure 4-23), from 169 nm at 880°C to 482 nm at 930°C, consistent with an

increase in adatom mobility at higher temperatures. These surface diffusion lengths are similar to

those reported by others, ranging from 100 to 180 nm specifically for ZnO nanowires[173, 191].

In other systems diffusion lengths as large as 10 µm have been reported for nanowire

sidewalls[192]. The increase of LSD with temperature is indicative that the surface diffusion is in

the clustering-limited regime. The maximum growth rates in this work incorporate about 1000

ZnO molecules/s. Here we will focus on Zn adatoms, as opposed to oxygen, because the

Knudsen equation based on the oxygen partial pressure, PO2, predicts much higher oxygen flux:

< = =< >2?@9A Equation 4-9

The sticking coefficient is a measurement of the percentage of arriving adatoms of a given

species which actually adsorb to the substrate, and can limit the delivery of precursors, especially

for diatomic gases. Considering the sticking coefficient for O2), as found empirically by [193], in

combination with the Knudsen equation (Equation 4-9) under the growth conditions the oxygen

flux is predicted to be in the range of 1016 – 1017 atoms m-2 s-1. With effective fluxes estimated

from the model to be in the 1015 atoms m-2 s-1 range, the growth is likely limited by Zn adatoms.

Desorption rates for m-plane facets at these temperatures are as high as 1014 m-2 s-1 for oxygen

90

and 5×1013 m-2 s-1 for Zn[194], which is less than 1% of the high Zn source fluxes observed in

this work, equating to a build-up of ~100 Zn adatoms/s on a typical nanowire’s sidewall facets

which are not incorporated into the growth of the nanowire via top-facet. This allows for a fast

buildup of adatoms on the nanowire side-walls, further confirming the expected adatom

clustering limited regime. In this regime the diffusion length depends on the surface diffusion

activation energy, ∆GSD, and the flux, J:

$23 = >BCDCEFGHIJ KL> Equation 4-10

Here N0 is the number of possible lattice sites per area, D0 is the base diffusion coefficient, k is

the Boltzmann constant and T is the temperature (in K). Using this equation for clustering-

limited diffusion and fitting to the diffusion lengths and fluxes from the model, a rough

estimation of the surface diffusion activation energy is 3.2 eV, which is relevant for predicting

surface diffusion at other temperatures.

Figure 4-23. Temperature dependence of surface diffusion length and source flux. The

surface diffusion length, LSD, and the effective source flux as derived from the surface-diffusion

limited growth model for ZnO nanowires grown at high areal densities. The surface diffusion

activation energy, ∆GSD, is approximately 3.2 eV based on these results.

91

Besides the change in surface-diffusion length with increasing temperature, there are also

shifts in the nanowire morphology. In general, at higher temperatures the nanowires are both

longer and thicker for a given growth time. As expected, the growth rate increases with

increasing temperature. The nanowire diameters also increase at higher temperatures due to a

higher necessary critical radius for island nucleation with increasing temperature, as has been

seen in the heteroepitaxy of Ge islands on Si and of homoepitaxy of Ag islands on Ag films[195,

196].

It has been observed that ZnO nanowire growth on continuous ZnO seed-layers in VTC can

occur in two regimes: either in the continuous nucleation regime or the surface-diffusion limited

regime. In general, the growth begins with through-thickness grains of ZnO seed-layer films on

Si. As the furnace is heated these grains facet and the 101 facets grow out to expose larger

areas of c-plane facets. A critical diameter exists, dependent on oxygen flow, for the nucleation

of islands on these c-plane facets. At higher oxygen flows, the driving force for nucleation (the

supersaturation of vapor precursor) is higher, and thus the critical size for stable island nuclei is

smaller allowing for a higher nucleation rate. At high enough oxygen flows the nanowire density

will be in a regime in which nanowires are competing for source precursors very soon after the

nucleation. At lower oxygen flows the continuous nucleation occurs as previously discussed,

with new nanowires nucleating as the growth continues. The two growth regimes are

summarized in the schematic in Figure 4-24.

92

Figure 4-24. Growth regime summary for VTC ZnO nanowires on continuous seed-layers.

The growth of ZnO nanowires in VTC on continuous seed-layers can proceed via continuous

nucleation with a linear dependence of nanowire length on diameter (left side) or surface-

diffusion limited with an inverse dependence of nanowire length on diameter (right side).

93

For further control over the nanowire morphology, the pressure is another parameter that can

be utilized. Figure 4-25(a) and (b) compares the SEM images of ZnO nanowires grown at 350 Pa

and 101,325 Pa (i.e. 1 atm), respectively. For atmospheric growths 283 sccm of Ar (the

maximum flow) was flowed during both heat up and cool-down to flush the system. Otherwise

all growth parameters besides the pressure were the same for the comparison in Figure 4-25. The

nanowire length to diameter relationship is compared in Figure 4-25(c). By increasing the growth

pressure to 1 atm the growth transitions from the continuous nucleation regime to a regime with

a large spread of nanowire diameters and lengths, likely in between the continuous nucleation

and surface-diffusion limited regimes. This shift also includes a large shift in nanowire diameter.

In general, atmospheric growths should result an increased partial pressure of O2, since there is

no vacuum pump to aid in flushing the system, thus increasing the overall source flux. An

increased source flux should result in a higher driving force for the nucleation of ZnO nanowires,

and thus should decrease the critical nanowire formation diameter; however, since the furnace is

only designed to be sealed by the vacuum formed when pumping, it is likely that small amount

of O2 is available as the system is ramping up to the growth temperature and only Ar is flowing.

This was confirmed by analyzing the seed-layer after only the heat-up and cool-down phases of

the growth, with no oxygen flowed (results not shown here). The small amount of O2 available

during the heat-up phase of the growth allows the seed-layer to undergo enough higher-order

facet growth for pyramidal structures to form and expose c-plane facets. Thus, for atmospheric

growths, nanowires begin to grow prior to the intentionally flowed O2. Once the intentional O2

flow is started, there are already many islands nucleated and thus the growth during the actual

growth window is likely to transition very quickly to the surface-diffusion limited growth.

During the ramp-up, however, since only whatever O2 leaks into the system is being provided,

the critical diameter for nanowire formation should be much larger, providing a pathway to

achieving growth of thicker nanowires.

94

Figure 4-25. Comparing low pressure and atmospheric growth. SEM images of nanowires

grown at (a) 350 Pa and (b) 101,325 Pa (equal to 1 atm) and (c) the nanowire length as a function

of diameter for the two growths indicate that at atmosphere the nanowire growth transitions from

continuous nucleation toward the surface-diffusion limited growth regime and larger diameters.

4.4.4. Two-Step Growth Technique

The morphology necessary for the fabrication of devices, especially light-emitting devices,

requires an array of vertical nanowires of uniform length with control over the nanowire areal

density. The densities desired with nanowire spacing near the emission wavelength (~8

nanowires/µm2) are observed in the continuous nucleation regime of nanowire growth, where the

nanowire length is linearly dependent on the nanowire diameter. In this growth regime it is not

95

possible to obtain uniform nanowire lengths without strictly controlling the nanowire diameter.

In the surface-diffusion limited regime at high enough nanowire diameters, the inverse diameter

dependence of length approaches an asymptote at large enough diameters, but the densities are

much higher. To achieve lower nanowire areal densities with uniform lengths it is possible to

begin growth in the continuous nucleation regime for short growth time, allowing the formation

of nanowires near the critical size at the desired densities, and then switching the growth

parameters to push the growth to the surface-diffusion limited regime. If the critical diameter for

nanowire formation is large enough, then most of the nanowires would be in the range of

diameters where their growth rates are not substantially different.

Figure 4-26. Growth parameters used for the two stage growth. The Ar (dashed line) and O2

flow rates (solid line) as a function of growth time for the two stage growth to control nanowire

length uniformity. During heating, cooling and transition stages, only Ar was flowed, while

during the growth stages, first at 885°C and then at 935°C, both O2 and Ar were flowed.

Thus, for the two-step growth technique, we chose to use atmospheric growths at low

temperatures to form the nanowire nucleation sites at larger diameters, and then increase the

temperature for a faster growth rate. Though lower temperatures likely lead to smaller critical

nanowire formation diameters, the effect from growing at atmospheric pressure is enough to

96

form thicker nanowire “stubs”, which are very uniform in height due to the low growth rate at

lower temperatures. These stubs then act to template the growth for larger diameter nanowires at

higher temperatures, where the growth rate dependence on diameter is minimized. The two stage

growth was carried out at atmospheric pressure, first at 885°C for 25 min, then at 935°C for 20

min. During the growth phase 2.5 sccm and 73.5 sccm O2 and Ar were flowed, respectively,

while during the heating and cooling phases only 283 sccm of Ar was flowed, as summarized in

Figure 4-26. Figure 4-27(a) and (b) are the SEM images of the nanowire stubs that form in stage

one of the two-step growth, while Figure 4-27(c) and (d) are the SEM images of the full two-

stage growth. The plot of nanowire length as a function of diameter in Figure 4-27(e)

demonstrates the uniformity of the nanowire length, only slightly dependent on the nanowire

diameter, which is likely the effect of the initial stage of the growth where the length depends

linearly on the nanowire diameter. This approach of the two-step growth is a novel way to grow

arrays of ZnO nanowires with uniform nanowire lengths, arrived at by a thorough understanding

of the underlying growth mechanisms; however, control over the nanowire areal density in this

growth method is lacking. In the next section, we discuss a growth method which combines the

concept of the two-step growth with the addition of control over the nanowire areal density.

97

Figure 4-27. Two-step growth process for nanowire arrays with uniform lengths. Top-down

(a) and (c) cross-sectional (b) and (d) SEM images of the first stage and the full two-step growth

and (e) the mostly independent nanowire length as a function of nanowire diameter,

demonstrating the effectiveness of the two-step growth process for uniform nanowire lengths.

98

4.5. VTC ZnO nanowire synthesis on silicon with sparse ZnO seed-layers

To further control the overall nanowire morphology, specifically the nanowire areal density

while maintaining the uniform nanowire lengths achieved with the two-step growth, sparse ZnO

seed-layers were used. Sparse-seed layers were prepared such that the ZnO layer does not form a

continuous film, but rather small particles of ZnO. These particles then act to seed the growth in

the same way that the nanowire “stubs” in the two-step growth did, by defining the nanowire

diameter in the heating stage while the available oxygen is very low. In the case of the sparse

seed-layers no pyramidal bases were observed, indicating that the nanowires nucleate directly as

islands on the silicon substrate. While the growth of the pyramidal base to expose c-plane facets

does not occur, there is still a critical size for the nucleation of ZnO islands that is larger at lower

oxygen flows (lower supersaturation). Thus, the sparse seed-layers can be used in place of the

first step in the two-stage growth if the diameters of the nucleated nanowires are large enough

such that the growth rate is independent of nanowire diameter, and the preparation conditions of

the seed-layer can be used to directly control the nanowire areal density. It should be noted that

this method relies on having areal nanowire densities that are high enough so the growth is in the

surface-diffusion limited regime.

Sparse ZnO seed-layers were prepared on Si (111) substrates in the same way as dense seed-

layers were processed, except the zinc-acetate solution was further diluted to a concentration of

only 0.01 M and the spin-speeds were varied from 1,000 to 10,000 rpm. To achieve

reproducibility the solution was dispensed dynamically, after the spin-coater had fully ramped to

the appropriate speed and only one layer was deposited and annealed. The growths were

performed at atmospheric pressure with the effluent bubbled through a water trap to avoid any

back-flow from the ambient. Similar to the other atmospheric growths, 283 sccm of Ar was

flowed during heating and cooling, with 73.5 sccm of Ar supplied during the growth window.

The oxygen was only flowed during the growth window (at either 2.5 or 5 sccm), which lasted

25 min at either 885°C or 900°C. No oxidation was performed before these growths; instead, a

new quartz outer tube was used, which had no coating built up to shift growth conditions.

99

4.5.1. Controlling nanowire areal density with sparse seed-layers

The nanowire areal density control was accomplished by varying the spin-speed at which the

zinc-acetate solution was dispersed during seed-layer preparation. In general, at higher spin-

speeds fewer zinc acetate particles adsorbed onto the substrate surface as the solution was spun

off of the substrate more quickly. This simple mechanism allowed for control over the nanowire

areal density over three orders of magnitude: from 12 nanowires/µm2 (at 1000 rpm) to 0.05

nanowires/µm2 (at 10,000 rpm). Figure 4-28(a) demonstrates the densities achieved at the

various spin-speeds. This method cannot be extended without further diluting the zinc-acetate

seed-layer solution as the areal densities achieved approach a plateau both at lower and higher

spin-speeds.

Figure 4-28(b)-(e) shows cross-sectional SEM images of selected growths on sparse seed-

layers prepared at spin-speeds of 1000, 2000, 3000 and 10000 rpm, respectively. Nanowire

arrays of uniform length are achieved in most cases except at lower nanowire areal densities. In

these cases there are typical nanowires with uniform lengths with long narrow tips which are

likely due to an excess amount of source available for growth, even during the cooling stage. In

these cases it is possible that the bases of the nanowires grow during the heating stage due to

unintentional leakage at atmosphere when very little O2 is available. Once the oxygen is

intentionally flowed during the growth window, the supersaturation is much higher such that the

critical island size for nucleation is very small, allowing the nucleation of new nanowires on the

top facet of nanowire which have already grown. Such formations are commensurate with a high

Ehrilich-Schwoebel barrier for diffusion off of the top-facets of islands, which have been

observed to be significant in ZnO at temperatures as high as 550°C[29].

100

Figure 4-28. Fine-tuning the nanowire areal density with sparse seed-layers. (a) Nanowire

areal density as a function of the spin-speed used to process sparse ZnO seed-layers,

demonstrating control over the nanowire areal density over three orders of magnitude. The

nanowire morphology changes as a function of areal density as seen in the cross-sectional SEM

images of growths with sparse seed-layers prepared at (b) 1000, (c) 2000, (d) 3000 and (e) 10000

rpm.

101

4.5.2. Controlling nanowire morphology on sparse seed-layers

The demonstrated tunability of the areal density provides further control over the nanowire

morphology as summarized in Figure 4-29. In general, both the average nanowire diameter and

length decrease with increasing nanowire areal density, as controlled by the processing

conditions of the sparse ZnO seed-layers. The decrease in nanowire diameter and length as areal

density increases is due to competition for the source precursors. The standard deviation of

nanowire diameters is smaller at higher areal densities (as represented by the error bars in Figure

4-29), because only at low areal densities is there enough source available for substantial side-

facet growth, which is also why the diameters are much larger at lower areal densities. The

deviation of nanowire length is larger at lower nanowire areal densities, likely due to the larger

spread of nanowire diameters and possibly to fluctuations in local areal density. Overall, this

demonstrates the use of sparse seed-layers to control nanowire morphology, where arrays of

uniform nanowire lengths are achieved above areal densities of 10 nanowires/µm2.

Figure 4-29. Average diameter and length of nanowires on sparse seed-layers. The average

(a) nanowire diameter () and (b) length (), both decrease as a function of the nanowire areal

density. The top axis estimates the average inter-nanowire spacing. Error-bars are one standard

deviation.

The transition from longer nanowires with a larger standard deviation of length to nanowire

arrays with uniform lengths occurs just above 10 nanowires/µm2, or at an inter-nanowire spacing

of about 300 nm. This indicates that uniform nanowire lengths are prevalent when all of the

nanowires are competing with others for source adatoms, indicating that the source adatoms have

102

a diffusion length of approximately 300 nm on the substrate surface at these growth conditions.

As more and more nanowires are competing for source in a given area, the role of direct

impingement of source adatoms onto the top-facet of the nanowire is more dominant, allowing

all of the nanowires, regardless of their diameter, to grow with approximately the same growth

rate. Thus, control over the production of nanowire arrays with uniform lengths and tunable areal

densities has been achieved using sparse ZnO seed-layers. These nanowire arrays are grown

directly on Si, and are ideal for the application of nanowire-based light emitting diodes.

4.6. Summary

In this chapter we demonstrated the growth of ZnO nanowires (1) via low-temperature

hydrothermal synthesis on silicon substrates and flexible substrates and (2) via high-temperature

vapor-transport and condensation on sapphire substrates with Au seed-particles and directly on

silicon substrates using continuous and sparse ZnO seed-layers. Hydrothermal synthesis provides

a low-temperature method for the synthesis of ZnO nanowires and allows for growth on flexible

substrates, but is limited by control over nanowire morphology, specifically the nanowire areal

density. Using high-temperature vapor-transport and condensation with Au-seed particles on

sapphire substrates, control over nanowire diameters and areal density can be achieved; however

the Au-seed-particles remain not only at the nanowire tips, but also along the substrate, which is

not suitable for device fabrication. To overcome the issues associated with foreign seed-particles,

high-temperature growth of ZnO nanowires on continuous seed-layers was investigated. It was

found that growth occurred in two regimes: the continuous nucleation regime at low areal

densities, where the nanowire length is linearly dependent on the nanowire diameter and the

surface-diffusion limited regime at high areal densities, where the nanowire length is inversely

proportional to the nanowire diameter. Control over the areal density within the continuous

nucleation regime was achieved by varying the O2 flow in the system. It was observed that a

critical diameter for nanowire formation decreased with increasing O2 flow, due to the specifics

of the seed-layer growth and the necessary nucleation of nanowires on c-plane facets. In the

surface diffusion limited regime, growth at different temperatures and modeling the kinetics

allowed for the determination of surface-diffusion lengths for the nanowire side-walls, which

was shown to increase with increasing temperature. Knowledge of the nanowire length

dependence on diameter in these regimes was used to develop a novel two-step growth for

103

nanowire arrays of uniform length. To further tune the nanowire areal density, sparse ZnO seed-

layers were developed to mimic the concept of the two-step growth, with explicit control over

the nanowire areal density. Nanowire areal density control was achieved over three orders of

magnitude and nanowire arrays of uniform lengths, ideal for light-emitting devices, were grown

at the higher areal densities. These investigations into the growth of ZnO nanowires allowed for

the development of novel techniques to reliably grow controlled, uniform arrays of nanowires on

Si for the use in nanowire-based devices.

The work in this chapter on controlling the VTC-growth of ZnO nanowires on Si with ZnO

seed-layers is being prepared for publication as a journal article.

104

Chapter 5: ZnO Nanowire Optical Properties

In this chapter we compare the optical properties of ZnO nanowires synthesized using

various methods described in Chapter 4. The synthesis method of ZnO nanowires and additional

processing parameters can have a significant effect on the optical and electrical properties of

ZnO nanowires. We find that high-temperature grown ZnO nanowires have negligible defect

luminescence and significantly higher luminescent efficiency relative to the GaN nanowires

grown in this work as described in Appendix A. Low temperature hydrothermally synthesized

ZnO nanowires have larger defect luminescence and overall reduced luminescence as compared

to those grown at high temperatures via vapor transport and condensation. We also observe that

increased oxygen flow during ZnO nanowire growth results in a slight red-shift of the near band-

edge luminescence. We find that annealing ZnO nanowires in nitrogen results in broad green

defect luminescence and cathodolouminescence (CL) in scanning transmission electron

microscopy (STEM) reveals that it is a bulk emission. The potential sources of this emission and

its potential usefulness in the fabrication of nanowire-based LEDs are discussed. Finally, the

photoluminescence of ZnO nanowires grown at different areal densities are compared and

discussed.

5.1. Optical properties of ZnO nanowires

ZnO is an excellent material for applications that require ultraviolet emission with high

luminescence efficiency due to the high exciton binding energy (~60 meV)[197], such that

excitonic emission is important even at room temperature[198-201]. Near-band edge

luminescence (attributed to free-excitons or electron-hole recombination, dependent on the

excitation[201]) is generally near the band-gap (3.37 eV) in bulk single-crystal ZnO[61, 202].

The near-band luminescence of ZnO nanostructures grown via high temperature methods, on the

other hand, tends to be observed near 3.27 eV [60, 62, 198, 203, 204]. This red-shift in

nanowires as compared to the bulk is likely due to local heating from the excitation source and

the lack of pathways for the diffusion of thermal energy, as proposed by Kong et. al.[205]. In

general, we observe a red-shift of the near-band edge luminescence with increasing excitation,

consistent with this proposition. Defect luminescence is also common in ZnO nanostructures,

both in the green and yellow regimes of the spectrum. Green-luminescence (GL) is commonly

105

observed around 2.4-2.5 eV and is generally attributed to recombination at surface states[198,

203, 204] and has recently been specifically designated to the formation of Zn-vacancies at the

non-polar nanowire sidewall facets, for ZnO nanowires grown in oxygen-rich conditions[60].

Yellow-luminescence (YL) is more commonly observed in hydrothermally synthesized ZnO

nanostructures around 2.0-2.1 eV and has been attributed to adsorbed OH groups at nanowire

sidewall facets or oxygen-interstitials [70, 71, 198].

The transition from GaN nanowires to ZnO nanowires for light-emitting applications in this

work is discussed in Appendix A, but here we emphasize the difference in optical properties. We

note here that the GaN nanowires grown in this work are anomalous in their poor optical quality,

as GaN is generally an excellent material; however we attribute this to a sub-optimal reactor for

GaN growth in which we have been unable to eliminate impurities. Figure 5-1(a) shows the

photoluminescence (PL) spectrum of VTC-grown ZnO nanowires on Si with an integration time

of only 0.15 s, demonstrating a luminescence about three orders of magnitude higher than for

GaN nanowires grown on Si with and without the use of a Zr oxygen-getter, as described in

Section A.4. Optical properties of GaN nanowires, with an integration time of 120 s (Figure

5-1(b)). The nanowires compared are of similar size, areal density and alignment, indicating that

most of the difference in luminescence should come from luminescence efficiency, indicating

that the ZnO nanowires have a luminescence efficiency approximately six orders of magnitude

higher than the GaN nanowires grown in the presence of oxygen. Of these two options, the ZnO

nanowires in this work are more viable materials for the fabrication of light-emitting devices.

5.2. Photoluminescence of ZnO nanowires

The method of synthesis plays a significant role in the optical properties of ZnO nanowires.

Figure 5-2 compares the PL spectra of hydrothermally synthesized ZnO nanowires on Si and

PEN. The hydrothermally synthesized nanowires generally exhibit near band-edge luminescence

at 3.3 eV, higher than the near-band edge luminescence seen from VTC-grown nanowires (3.21

eV). This is again likely due to a thermal effect from the excitation, where the nanowires grown

on Si are relatively short as compared to the VTC-grown wires discussed in reference to Figure

5-2, thus in the nanowires the heat generated from thermalization of the excited carriers can only

dissipate through the nanowires, while in the shorter HTS nanowires, the heat can dissipate

through the Si substrate. Similarly, the band-edge-luminescence from nanowires grown on PEN

106

also undergoes a significant red-shift as compared to nanowires grown on Si. Because nanowires

grown on Si have a higher areal density than nanowires grown on PEN and Si has a much higher

thermal conductivity than PEN – allowing for more heat dissipation and distribution across

nanowires –thermally induced red-shifts are more pronounced in ZnO nanowires on PEN.

Figure 5-1. Comparison of luminescence efficiency of GaN and ZnO nanowires.

Photoluminescence spectra of (a) VTC-grown ZnO nanowires and (b) GaN nanowires grown in

a sub-optimal reactor in the presence of impurities demonstrate the improved luminescence

efficiency of these ZnO nanowires due to a much more optimal growth reactor for the materials

system. The spectrum in (a) was collected over 0.15 s, while the spectrum in (b) was collected

over 120 s and is magnified 1,000 times relative to (a).

107

The broad yellow luminescence from hydrothermally synthesized ZnO nanowires on both Si

and PEN peak is centered at 2.17 eV, and is notably stronger from nanowires grown on PEN

than nanowires grown on Si. Yellow luminescence has been attributed to OH groups adsorbed to

the nanowire surface from the hydrothermal synthesis, and has been shown to be mitigated with

annealing or surfactant treatments of the nanowires[198]. Alternatively, it has been proposed that

the energy level of yellow luminescence corresponds to oxygen interstitials, which are mid-gap

levels in ZnO[206]. For nanowires hydrothermally synthesized on PEN, there is also a peak in

the blue region of the spectrum at ~2.92 eV, which is fluorescence from the PEN substrate as

confirmed by PL of bare PEN. While these nanowires emit over the visible spectrum, it would be

ideal to control the luminescence within a narrow range, as seen with the VTC-grown nanowires.

Figure 5-2. PL spectra of hydrothermally synthesized ZnO on Si and PEN. The PL spectra

of HTS ZnO nanowires on Si (dashed line) and PEN (solid line) demonstrate yellow

luminescence in both cases, and nanowires on PEN have an additional luminescent peak around

2.95 eV, due to the background fluorescence from the PEN substrate.

108

ZnO nanowires grown at high temperatures via the VTC method generally have excellent

optical qualities, with relatively narrow near-band-edge luminescence and no defect

luminescence in the green or yellow. However, the green luminescence can be seen in some

cases with low nanowire areal density, indicating the green-luminescence from our nanowire

arrays may only originate from the seed-layer. Figure 5-3 demonstrates how the PL spectra of

ZnO nanowires grown on continuous seed-layers evolve for samples grown at different oxygen

flows. The nanowire areal density and nanowire growth rate increases with increasing oxygen

flow, as previously discussed, in Chapter 4: ZnO Nanowire Synthesis on Si. For comparison the

nanowire morphologies of these samples is summarized in Table 5-1. The red-shift of the near-

band-edge luminescence can be attributed to the morphological changes of the nanowires and the

associated thermal effects. In the sample grown at 0.8 sccm O2 there are fewer nanowires per

area and they are much shorter, thus heat from the excitation can be conducted through the ZnO

seed-layer film and the Si substrate. In the more dense samples with longer nanowires, however,

the heat will be mostly generated at the nanowire tips, where most of the absorption of photons

will occur, causing more heating as nanowires get longer. Furthermore, power-dependent

photoluminescence of these samples exhibit a red-shift with increasing excitation power, which

is consistent with local heating. A temperature increase-induced red-shift is known for band-to-

band transitions from the valence to conduction band, or for the free-exciton emission[198]. To

understand the necessary temperature shifts that would cause such a change in the band-gap, we

can use the Varshni equation, shown as Equation 5-1, where Eg(T) is the temperature-dependent

band gap, Eg0 is the band-gap at absolute zero, α and β are material properties (in eV/K and K,

respectively) and T is the temperature in kelvins.

MN(A) = MNC + OA A + P Equation 5-1

The parameters of this equation have been previously studied[207] and for ZnO Eg0 is 3.375 eV,

α = 9.5 × 10−4 eV/K and β =912 K. The increase in temperature necessary to cause the 22 meV

redshift (between emission from nanowires grown at 0.8 sccm O2 and 1.3 sccm O2) is 43 K. The

laser power used in the PL setup has an average power density of 10.95 kW/cm2 (with a pulse

power density of 500,000 kW/cm2). The thermal conductivity of ZnO (κZnO) has been reported as

1.35 W/cm·K[208]. At this wavelength, each photon contributes a thermal energy of about 1.48

109

eV, which means 31% of the laser power is converted to heat. In similar ZnO nanostructures to

the dense forest-like growth of the nanowires grown at 1.3 sccm, laser-induced heating has been

shown to cause a red-shift to emission energies as low as 3.200 eV[209] at a laser power density

of 5 kW/cm2.

Table 5-1. Summary of nanowire morphologies compared in Figure 5-3.

Oxygen Flow Nanowire Areal Density Typical Nanowire Length

0.8 sccm 0.4 /µm2 ~2 µm

1.1 sccm 8.2 /µm2 ~6 µm

1.3 sccm 13 /µm2 ~10 µm

Figure 5-3. PL spectra of ZnO nanowires grown with increasing oxygen flow. The

normalized PL spectra of ZnO nanowires grown via VTC on Si exhibit a decrease in green

luminescence and a red-shift of the near-band-edge luminescence with increasing O2 flow.

110

The high density of the samples grown at 1.1 and 1.3 sccm O2 also means that very few of

the excitation photons reach the substrate. The green-luminescence is only observed in the

sample grown at 0.8 sccm O2, where the seed-layer emission substantially contributes to the

spectra. PL of seed-layer films which have undergone the growth process, but at oxygen flows

too low for nanowire growth have similar broad green emission, indicating that the green

luminescence observed is from the seed-layer. Since the seed-layer film after the growth is

mostly highly faceted mound-like structures composed of non-polar and semi-polar facets, the

green luminescence is likely coming from these surfaces[60]. For samples that exhibit green

luminescence, the ratio of the integrated intensities of the near band-edge to green luminescence

is often cited as a measure of optical quality. For the sample in this work that exhibit green

luminescence, this ratio is 6.35, which is higher than most reports in the literature, where the

green luminescence is often even greater than the near-band-edge luminescence[60, 70, 71, 198].

For further insight into control over the green-luminescence, ZnO nanowires were annealed

in an inert nitrogen atmosphere at 800°C for 30 min and then at 1000°C for 60 min, with the

resulting PL spectra shown in Figure 5-4. The near-band-edge luminescence undergoes a small

red-shift upon annealing, but does not shift further at higher temperature annealing. Similar

trends have been observed [198] and are proposed to be from developing strain at the nanowire

interface with the substrate[93]. Increasing the annealing temperature induces more green-

luminescence. While green luminescence has been frequently associated with the formation of

Zn vacancies on the non-polar nanowire sidewalls, cathodoluminscence (CL) in scanning

transmission electron microscopy (STEM) reveals that the emission is not strictly from the

surface (Figure 5-5). The green luminescence is very bright in CL-STEM at ~100K and follows a

line profile that indicates bulk emission, brighter in the center of the nanowire than at the edges

as shown in . We note that there is in fact a surface region that emits at the near band-edge (365

nm), but not for the green luminescence (500 nm). This bulk defect emission could be from

alternative defects which emit in the green, though it could still be due to the formation of Zn

vacancies throughout the nanowire. It is unlikely that it is due to oxygen vacancies, however, as

the peak would be expected to be at 2.25 eV rather than at 2.35 eV, which matches with

recombination from a shallow donor to the Zn vacancy acceptor level[200]. The extent of the

anneal (time and temperature) also increases the relative intensity of the green luminescence,

which could mean that the Zn vacancies form at the nanowire surface, but the elevated

111

temperatures for extended periods allows the vacancies to diffuse into the bulk of the nanowire.

By controlling the anneal time and temperature it should be possible to tune the emission ratio of

the near band-edge and green luminescence to effectively emit white light. Since the nanowires

remain bright and such nanowires have the potential to be used directly for white nanowire based

LEDs in the form of a heterojunction with substrate.

Figure 5-4. Annealing-induced green-luminescence in ZnO nanowires. PL spectra of ZnO

nanowires before and after annealing at 800°C and 1000°C in nitrogen demonstrates induced

green-luminescence with higher temperature annealing as well as a red-shift of near-band edge

emission upon annealing.

112

Figure 5-5. CL of an annealed ZnO nanowire demonstrating bulk defect luminescence. CL

from ZnO nanowires annealed in N2 indicate a bulk like emission as demonstrated in (a) the

composite CL map overlaid on the STEM image of the nanowire mapped where green

luminescence at 365 nm is green and near band-edge luminescence is purple and the red arrow

indicates the linescan demonstrated in (b) of normalized CL intensity for both peaks mapped

against the position along the linescan. The green luminescence is more centered than the UV

luminescence demonstrating that the green emission is from the bulk of the nanowire.

The ZnO nanowires grown using sparse ZnO seed-layers are the most promising nanowire

arrays for use in fabricating devices due to their excellent length uniformity and control over

areal density, thus it is important that they maintain the excellent optical properties observed for

nanowires grown on continuous seed-layers. Figure 5-6 demonstrates that while the intensity of

luminescence from the nanowire arrays decreases with decreasing density, as expected, no defect

luminescence is observed. Furthermore, the peak of the near band edge luminescence does not

shift with areal density since the nanowires are vertically aligned and thus allow similar amounts

of heat dissipation independent of areal density. The approximate PL quantum yield of these

nanowires is 15%, indicating that these nanowires grown on sparse seed-layers are the best-

suited for devices since they retain high luminescence efficiency while the morphology of

vertical nanowires with controllable areal density allows for enhanced directed emission, as

discussed in Section 2.3.1.

113

Figure 5-6. PL spectra of ZnO nanowires on sparse seed-layers. The PL spectra of nanowires

grown via sparse seed-layer to control the nanowire areal density (at densities of 12, 10 and 4

nanowires/µm2) exhibit excellent optical properties with high brightness from the UV near-band-

edge emission and no defect luminescence. The inset demonstrates the increase in integrated

intensity with increasing nanowire areal density.

5.3. Summary

In this chapter, we demonstrated the excellent optical properties of VTC-grown ZnO

nanowires, especially relative to GaN grown in the presence of oxygen. HTS-grown nanowires

demonstrated broad yellow defect luminescence. In comparison, VTC-grown ZnO nanowires on

both sparse and continuous seed-layers exhibited bright near band-edge luminescence with no

defect luminescence. A redshift in the near-band edge emission due to thermal heating from the

laser excitation used for PL measurements was observed for nanowires grown in a very dense

forest-like growth regime. Finally, bulk green luminescence was induced in ZnO nanowires by

annealing in nitrogen at 1000°C. The bulk-nature of the luminescence was confirmed in CL-

STEM and the high brightness of these samples indicates that they are good candidates for

simple single-heterojunction white nanowire-based LEDs.

114

Chapter 6: Towards Scalable Nanowire-Based LEDs on Si

In this chapter we discuss the outlook for scalable nanowire-based LEDs on Si. We address

the potential challenges and review approaches to doping ZnO nanowires, and present a platform

developed to characterize the electrical properties of nanowires. Electron beam lithography and a

nanowire alignment process are used to demonstrate contact to individual nanowires. We discuss

the methods used to extract the electrical properties from individual nanowire devices, including

the conductivity, carrier mobility and carrier concentration. The role of surface states is

discussed as well as approaches to overcoming the strong effects of the surface depletion region.

We show demonstrative examples of individual nanowire measurements of as-grown and H2-

annealed ZnO nanowires, which demonstrate p-type and n-type behavior, respectively. Finally,

different approaches to device fabrication are assessed relative to the nanowires grown in this

work.

6.1. Individual nanowire electrical characterization

Electrical characterization of nanowires is critical to understanding dopant incorporation in

nanowires. To achieve accurate measurements of carrier concentration it is necessary to make

contact to individual nanowires, allowing for normalization by volume to estimate carrier

concentrations and resistivity. The process flow for contacting individual nanowires is outlined

in Figure 6-1. After nanowire growth, the growth substrate is submerged in ethanol and sonicated

briefly to remove the nanowires from the substrate and suspend them in solution. The nanowire

solution is then cast onto pre-patterned gold grids on oxidized silicon. After the ethanol dries,

poly-methyl methacrylate (PMMA), 4% by weight in anisole, is spun on as an electron-beam

resist. The locations of nanowires are mapped with dark-field optical microscopy relative to the

grid to create a mask file used in the scanning electron beam lithography (SEBL). Contacts are

deposited via electron beam evaporation and treated with a rapid-thermal anneal to fabricate

individual nanowire devices. In this work Ti/Au contacts were used, as they are the most

commonly accepted contact metallization for as-grown n-type ZnO[81, 210, 211], while Ni is

common for p-type ZnO[10]. The individual nanowire devices can be used to thoroughly

characterize the nanowire electrical properties.

115

Figure 6-1. Method for individual nanowire electrical characterization. This schematic steps

through the processes of creating a grid to map nanowires for SEBL (step 1), dispersing and

mapping the nanowire locations (steps 2-5) and making contact to the nanowires (steps 6-7) to

form individual nanowire devices to be characterized (step 8).

Individual nanowire devices can then be used to ascertain electrical characteristics by being

probed in various setups. For longer nanowires that allow the deposition of four contacts, four-

point measurements can be used to remove the effects of contact resistance. In four-point

measurements current is passed between the outermost contacts and the voltage is measured

across the two inner-most contacts to avoid measuring the voltage drop associated with contact

resistance. This measurement is the most accurate way to assess the actual electrical properties of

a nanowire. Without four-point measurements the contact resistance would be indiscernible from

the nanowire resistance, and thus the resistivity from two-point measurements is an over-

116

estimate of the nanowire resistivity. In principle, contact resistance can be measured using the

transmission line method, which involves measuring the resistivity through multiple two point

measurements at varying separations and plotting the resistance as a function of separation. The

slope of resistance plotted against contact separation is proportional to the resistivity, while the

intercept is the sum of the contact resistances between the two contacts. The nanowires

characterized in this work used four-point measurements when possible, but otherwise two-point

contact was made and the contact resistance was neglected. Obtaining high yields of correctly-

oriented long nanowires for four-point measurements can be challenging due to breakage of the

nanowires during sonication.

Mobility and doping-type can also be measured by back-contacting the silicon and using the

oxide to create effective field effect transistors. These transfer characteristics measure the change

in source-drain current ISD (at a constant source-drain bias VSD) as a function of the applied gate

voltage, VG. The gate bias attracts carriers to the interface with the dielectric, creating a

conductive channel that becomes more conductive as the bias is increased. For example, if a

positive bias is applied and the nanowire is n-type, the free-electrons will be attracted to the

interface and create a conductive n-type channel. As the gate bias is further increased more

carriers occupy the channel and the conductivity increases. However, if the opposite gate bias is

applied to the same nanowire, the conductivity does not change significantly, since there are not

enough free holes to act as a conductive channel. Thus, the sign of the transconductance (gm), the

change of the measured ISD with applied VG, is indicative of the carrier-type. If the

transconductance is positive (i.e. – ISD increases with +VG) then the material is n-type as in the

example discussed above. Alternatively, a negative transconductance indicates that the material

is p-type. The transconductance depends on the carrier mobility µ, the capacitance of the oxide

Cox, the applied source-drain bias VSD and the length of the channel between the two contacts, L,

according to the field effect transistor charge control model[212, 213] as shown Equation 6-1

and is used to estimate carrier mobility from the transfer characteristics.

QR = S23(+T$ Equation 6-1

If the carrier mobility is known, the carrier concentration can be estimated from the conductivity

according to Equation 6-2, where σ is conductivity, ρ is resistivity, R is the measured resistance,

117

L is the contact separation, A is the nanowire cross-sectional area, q is the electron charge, ND

and NA are the donor and acceptor concentrations, respectively, and µe and µh are the electron and

hole mobilities, respectively. Usually one carrier concentration is dominant, in which case the

approximation that the opposite carrier concentration is negligible can be made (i.e. – in n-type

materials, ND >> NA).

5 = 1U = $V = (B3T; + BWTX) Equation 6-2

These methods of electrical characterization result in useful estimates of the electrical

properties; however there are many effects unique to nanostructures, typically due to the large

surface-to-volume ratio, that can affect these measurements. Contact resistance is usually

assumed to be the same at each contact made to a material. However, nanowire electrical

properties can vary along the growth axis. Furthermore, the actual contact area under different

pads often varies in nanowire contacts. Thus, it may be more useful to define a specific contact

resistance which scales with the actual interface between the metallic contact and the

semiconductor nanowire[214]. It is also important to correctly estimate the effective oxide

capacitance, Cox. Since the interface between the nanowire affects the coupling between the

nanowire and the oxide in the case of nanowires lying flat on SiO2 an effective dielectric

constant can be used as a correction factor for different nanowire cross-sections[215].

The surface of the nanowire plays a critical role in the electronic properties due to the large

surface-to-volume ratio in nanowires. Surface states such as adsorbed OH- groups (that act as

acceptors on ZnO surfaces[211]) can deplete a shell near the nanowire surface of its free-

electrons. Fermi-level pinning at the surfaces results in band-bending from the surface inward

into the normally n-type nanowire such that the surface depletion region depends on the surface

barrier potential φs at the surface and the doping level in the nanowires. The surface depletion

width, W, the depth from the surface that is depleted due to Fermi-level pinning is dependent on

the surface barrier potential φs (which is a property of the type of surface states and their

density), the doping level ND, the dielectric constant of ZnO εZnO and the electron charge q as

expressed in Equation 6-3[216].

118

Y =Z2[\,<].B3 Equation 6-3

If this surface depletion region is large enough relative to the nanowire thickness, then it is

possible for the entire nanowire to be depleted due to this effect. This surface depletion region

can also cause an effective Schottky contact, since contact must be made through the nanowires

sidewalls and thus through the surface depletion region[211].

The ambient pressure has been shown to affect the extent of the surface depletion width, as it

would be expected because at higher pressures there are more species that can adsorb to the

nanowire surface. Therefore, at low pressures the surface depletion width is smaller and

nanowires tend to be more conductive[211, 217, 218]. There are two general approaches to

mitigate the extent of the surface depletion region in nanowires: heavy doping or passivation.

The heavy doping strategy considers Equation 6-3, which demonstrates that the surface depletion

region will decrease with the donor concentration, ND. Alternatively, the density of surface states

is reduced via passivation, which reduces the surface barrier potential φs and thus the surface

depletion width. Annealing in forming gas (5% H2 in N2) is a common approach to increase the

conductivity of ZnO nanowires. The effect of ambient species and pressure seems to be mostly

eliminated via both H2 anneals and H2 plasma treatment[211, 217, 219], though it is unclear

whether hydrogen is passivating surface states or acting as a shallow donor. Other passivation

approaches include coating the nanowires with a higher band-gap material, such as Al2O3 or

PMMA. Al2O3 coating does not seem to be very effective, likely due to a poor interface with the

ZnO nanowires[219]. PMMA on the other hand works quite well, having a similar effect to

annealing in H2[220].

6.2. Electrical properties of ZnO nanowires

From ZnO nanowires grown via sparse-seed layers we interestingly observe p-type as-grown

behavior. All of the nanowire devices that resulted in ohmic contact (only about 20% of as-

grown ZnO nanowire devices) resulted in transfer characteristics with a negative

transconductance, indicating p-type behavior. A representative source-drain measurement from

these nanowires with the device SEM image inset and representative transfer characteristics are

shown in Figure 6-2(a) and (b), respectively. Extracting the mobility using Equation 6-1 for the

119

linear part of the transconductance (generally V < 0), the ohmic ZnO nanowire devices

demonstrated an average mobility of 225 cm2/V·s with a standard deviation of 5 cm2/V·s, similar

to reports of high mobility p-type ZnO achieved in thin-films[85]. It should be noted that the

transconductance becomes much less negative at small positive gate biases. This is an indication

of ambipolar behavior, indicating that there are still donors active in the material[221].

The conductivity of the nanowires and thus the predicted carrier concentration (Equation 6-2)

varied with nanowire diameter. In general, the p-type carrier concentration NA decreased with

increasing nanowire diameter from an estimated 2×1014 /cm3 at a nanowire diameter of 158 nm

to 1×1019 /cm3 at a nanowire diameter of 158 nm, suggesting that the p-type doping may be

related to the surface. It has been suggested that Zn-vacancies act as acceptors, but their

formation energies are high in p-type ZnO[78] and they generally cause green-luminescence at

the nanowire sidewall facets[60], which is not observed in PL of these samples. It is unlikely that

any common p-type dopant could have been introduced to the growth system; however carbon is

used in the carbothermal reduction of ZnO, and has been shown to act as a shallow acceptor if it

can form a substitution on an oxygen site, CO, for which oxygen vacancies are necessary[222].

Furthermore, it has been shown that oxygen vacancies have a very low formation energy on non-

polar nanowire sidewall facets under oxygen poor conditions[60], which is consistent with the

oxygen-deficient conditions for initial stages of nanowire growth discussed in Section 4.5 - VTC

ZnO nanowire synthesis on silicon with sparse ZnO seed-layers. It is thus possible that CO

acceptors form preferentially at the nanowire surface, in which case the p-type conductivity

would increase with nanowire diameter. It is also possible that during the Ar plasma clean prior

to contact deposition, that air was not adequately flushed from the gas-lines, resulting in some

nitrogen plasma, which has been previously shown to induce p-type doping[223] and would be

consistent with the dependence on nanowire diameter.

120

Figure 6-2. Electrical properties of individual ZnO nanowires. Representative current-

voltage curves for a ZnO nanowire grown via sparse seed-layers: (a) the source-drain

characteristics indicate roughly ohmic behavior with the SEM image of the nanowire device

inset and (b) the transfer characteristics indicate p-type behavior.

Upon annealing as-grown ZnO nanowires via an RTA in 5% H2 (in N2) at 500°C for 3 min

the general nanowire electrical characteristics shifted. The percentage of nanowire devices

displaying ohmic behavior increased from only about 20% to 50%, likely because the contact

metallization was optimized for n-type nanowires and the H2 anneal produced nanowires with n-

type doping as deduced from the transfer characteristics. Though n-type nanowires accounted for

about 80% of the ohmic devices produced with the H2 anneal, p-type nanowire devices similar to

the one shown in Figure 6-2 and more ambipolar behavior was also observed. The n-type

nanowire devices demonstrated an average mobility of 249 cm2/V·s, in the range of reported

electron mobilities in nanowires[224], with a standard deviation of 67 cm2/V·s, and the opposite

trend of conductivity with nanowire diameter as compared to the p-type nanowire devices. For

the n-type nanowire devices the estimated carrier concentrations increased with increasing

nanowire diameters from an estimated 9×1013 /cm3 at a nanowire diameter of 143 nm to 3×1017

/cm3 at a nanowire diameter of 309 nm, indicative of a surface depletion region. A representative

source-drain measurement from these n-type nanowires with the device SEM image inset and

representative transfer characteristics are shown in Figure 6-3(a) and (b), respectively.

121

Figure 6-3. Electrical properties of ZnO nanowires after H2 anneal. Representative current-

voltage curves for a ZnO nanowire grown via sparse seed-layers after H2 RTA: (a) the source-

drain characteristics indicate roughly ohmic behavior with the SEM image of the nanowire

device inset and (b) the transfer characteristics indicate n-type behavior.

These results demonstrate both n-type and p-type behavior in ZnO nanowires, indicating a

promising pathway towards ZnO homojunction LEDs; however it should be noted that due to the

low yield of ohmic devices, the number of nanowires characterized in each case is relatively

small (less than 10 devices). Thus, while further work is necessary to demonstrate control over

both dopant types, these preliminary results indicate that it should be possible. It would likely be

advantageous to approach the doping with extrinsic defects, such as group III elements for n-type

dopants and group V elements for p-type dopants so that the concentrations can be controlled

unlike the as-grown defects observed here.

6.3. Fabrication methods for nanowire array devices

In any implementation of nanowire devices on a production scale, it will be necessary to use

scalable techniques developed for typical semiconductor processing. Thus, as previously

discussed in Chapter 1, for nanowire arrays the vertical alignment of the nanowires with respect

to the substrate must be excellent. To make electrical contact to the nanowires on a large scale,

the nanowires must be able to be contacted through the substrate and at their tip. Using Si

substrates to grow ZnO nanowires provides a substrate with tunable electrical properties. To

contact the nanowire tips for high device yield requires not only a high degree of vertical

alignment, but also nanowire arrays with uniform lengths. Furthermore, to prevent an electrical

122

short when depositing the top contact, it is necessary to fill the gaps between nanowires with a

dielectric that exposes only the nanowire tip. Depending on the dielectric filler material and the

deposition process, it can be useful to have control over the nanowire areal density to ensure that

the dielectric penetrates the nanowire gaps and results in a uniform morphology at the top of the

film near the nanowire tips. This control over density is especially necessary when using a spin-

on glass or polymer-based filler.

Different approaches and materials used as the dielectric filler have different advantages and

disadvantages. The potential material deposition process can be in one of two categories: blanket

coatings or spin-on materials. Blanket coatings conform to the morphology of the nanowire

array, equally coating the substrate, the nanowire sidewall facets and the nanowire top facets.

This coating is usually amorphous. One key advantage to this method is that control over the

nanowire areal density is not strictly necessary. Since the coating is conformal, the deposition

must be long enough to deposit enough material to fully cover the nanowires and result in a top

surface. Devices fabrication requires the exposure of nanowire tips, so the coating must be

etched back to expose the nanowires. An example of such a coating is plasma-enhanced

chemical vapor deposited (PECVD) SiOx. SEM images of the process are shown in Figure 6-4.

Filling in all of the space between the nanowires can be time consuming, and due to the

conformal coating, requires a mechanical polish to planarize the surface prior to reactive ion

etching (RIE) to expose the nanowire tips. This RIE can be harsh and may damage the

nanowires, causing poor contact. The nanowires are shown along the steps in this process in

Figure 6-4 (a)-(c). Alternatively, less SiOx can be deposited and etched, as shown in Figure

6-4(d)-(f), leaving nanowires with a conformal shell and exposed tips, but also leaving empty

space between the nanowires which would result a non-continuous contact film if a transparent

conductor were sputtered and/or contact metallization was deposited. These SiOx-coated

nanowires may be useful for passivating surface states or creating gate-all-around transistors,

however these approaches were not explored here, since we were focused on the fabrication of

large-scale nanowire-based LEDs.

123

Figure 6-4. PECVD SiOx as a dielectric filler material. SEM images at 45° tilt of the coated

nanowires at different steps throughout the process: (a) as-grown ZnO nanowires (on a

continuous seed-layer), (b) after PECVD of 1.5µm of SiOx, (c) after PECVD of another 1.5 µm,

mechanical polishing and RIE of 750 nm of SiOx to expose nanowire tips, (d) PECVD of only

350 nm SiOx on as-grown nanowires similar to those shown (a), (e) 175 nm RIE after step shown

in (d), and (f) the top-down view of (e).

The alternative to using blanket coatings to contact materials is the use of a spin-on

dielectric. Spin-on glass dielectrics (usually silica-based) are materials that are viscous at room

temperature, but when annealed form a solid dielectric with excellent interfacial properties for

passivation and planarization. Unlike PECVD of dielectric materials, the spin-on dielectrics can

simply be applied at the proper spin-speed during spin-coating to create a uniform film which

fills the gaps between the nanowires but leaves the tips exposed for contact. Similarly, polymers

that can be cured, such as PMMA, have been used for this purpose, but do not allow for thermal

processing due to the breakdown of polymer functionality at their glass-transition temperatures.

Though spin-on dielectrics are typically expensive relative to polymers and have a short shelf-

124

life (usually 6 months), they provide a more robust approach to fabricating scalable nanowire-

based LEDs than either PECVD SiO2 or polymer-based dielectrics.

After filling the gaps in nanowire arrays, a transparent conductor can be sputtered to provide

electrical contact to the nanowire tips and contact metallization can be deposited to make an

ideal contact to the transparent conductor. This design allows for a bias to be applied through the

axial direction of the nanowire, suitable for axial heterostructures. As seen in Chapter 3, axial

heterostructures can compete with radial heterostructures in terms of efficiency. This thesis has

laid the groundwork for the fabrication of nanowire array-based LEDs and with the achievement

of controlled doping, single heterostructures of n-ZnO nanowires on p-Si can be readily

fabricated. To enhance the efficiency of such devices a nanowire homojunction could also be

formed either by changing dopant during growth or by using spin-on dielectrics to protect the

bottom part of nanowires while the tops of the nanowires are exposed to ex-situ doping (such as

vapor-phase doping or surface-monolayer doping). Finally, as QW materials, such as CdZnO, are

further developed, it may be possible to grow such structures for axial heterostructure-based

devices. Alternatively, to achieve simple white LEDs, the induced green-luminescence discussed

in Chapter 5 can be tuned such that the green and blue/UV combination results in white light.

125

Chapter 7: Conclusions

This thesis focused on the development and fabrication of scalable wide band-gap nanowire-

based LEDs. Nanowire-based LED designs were compared and their tradeoffs were discussed

relative to inherent properties of the design and the material properties, advancing the

understanding of how the materials properties and device design intersect and affect the overall

LED efficiency. A path towards the synthesis of nanowire-array based LEDs was outlined and

the critical control over key characteristics of the ZnO nanowire arrays was demonstrated:

tunable nanowire areal density, a high degree of vertical alignment relative to the substrate and

uniform nanowire lengths by using a two-step growth on continuous seed-layers and by using

sparse seed-layers. The optical properties of ZnO nanowire arrays were shown to be excellent

with no defect luminescence and PL quantum yields of ~15%. We demonstrate a way to control

the defect luminescence which may be a possible route for tuning the emission properties of ZnO

nanowires. Finally, we outlined the approach for the synthesis of nanowire-based LEDs from

these ZnO nanowire arrays.

7.1. Summary and implications of this work

Solid state lighting has tremendous potential in terms of energy and costs saving as compared

to traditional lighting technologies. Wide band-gap materials such as GaN and ZnO are

necessary to achieve white emission required for solid state lighting. These materials are

challenging to grow in a bulk form and epitaxial films can have high densities of extended

defects such as threading dislocations. Nanowires of these wide-band gap materials offer some

unique advantages toward solid state lighting. Due to the large surface-to-volume ratio nanowire

growth can accommodate high lattice mismatch between the substrate and nanowire material via

facile strain relaxation, resulting in dislocation-free single crystalline nanowires on a variety of

substrates. Silicon substrates are an ideal candidate for scaling nanowire growth toward a

scalable production-level whereas control over the electrical properties in Si allows for unique

nanowire-based LED designs.

In chapter 2 we discussed the interplay of materials properties specific to wide band-gap

nanowires, focusing on the case of GaN nanowires with InGaN QWs. The general trade-offs

between axial heterostructures for m-directional and c-directional nanowires as well as radial

126

heterostructures for c-directional nanowires were the focus of the comparison. The active area of

the device can be dramatically increased in radial heterostructures for high-brightness

applications; however, we focus mainly on comparing LED efficiency, which is composed of

two factors: light extraction efficiency and IQE. We emphasize the importance of the nanowire

wave-guiding properties as a function of the nanowire cross-section. Furthermore, the coupling

to the wave-guided modes considering the role of polarization anisotropy from QW emission and

the resulting extraction efficiency was also investigated. Finally, we discussed the role of the

QCSE in the IQE of different designs for nanowire-based LEDs.

In chapter 3 we sought to answer the outstanding questions brought up in Chapter 2. Namely,

we compared the modal properties of polar c-directional and non-polar m-directional GaN

nanowires. To understand the extraction from the different device designs, we used FDTD to

calculate the directed light extraction efficiency, measuring the emission from the guided modes.

Focusing on single-mode nanowire waveguides, we observed that the primary factors affecting

light extraction are the orientation of the QW relative to the nanowire wave-guiding axis and the

QW position. The directed extraction is optimized at QW positions which maximize coupling to

the fundamental mode of the nanowire waveguide. Fully simulating light extraction from these

designs showed that m-directional axial heterostructures had the highest potential for directed

efficiency. Furthermore, to understand the role of surface states from different nanowire facets

and QW orientations inherent in the different designs, the A-B-C model was used to estimate the

IQE based on known material properties. While surface-states are higher for m-directional axial

heterostructures, the QCSE plays a larger role in diminishing the IQE for c-directional axial

heterostructures, thus the highest overall IQEs are also seen for m-directional axial

heterostructures. By combining the calculated extraction efficiency and IQE, it was determined

that the m-directional axial heterostructure has the highest overall efficiency. Finally, we also

looked at the possibility of using diameter modulation to create a distributed Bragg reflector at

the base of the nanowire, with FDTD predicting almost 90% reflectivity with only 10 diameter

modulations (between large and small segments) after optimizing the parameters. This work was

the first to take an approach to directly compare different nanowire device designs and materials

properties, highlight the route for the synthesis of such heterostructures.

The work in chapter 3 will guide the development of nanowire-based LEDs designed as axial

heterostructures, specifically m-directional axial heterostructures, which have yet to be fabricated

127

in the field. A direct experimental comparison between m-directional and c-directional axial and

radial heterostructures would be the next step towards the advancement of choosing the proper

design for highest efficiencies. With control over the InGaN composition in the QW it may be

possible to design MQW nanowire-based LED which emit as RGB emitters, bypassing the need

for phosphor materials (and thus energy lost to down-conversion). Such designs may even have

higher efficiency than the current state-of-the-art GaN thin-film LEDs.

At the end of Chapter 3 we present a device outlook for GaN nanowire-based LEDs. In

Appendix A we discuss the challenges of synthesizing the designs laid out in Chapter 3 with the

available tools. The main issues in the synthesis of GaN nanowires was the lack of alignment

with the Si substrate, regardless of the many growth parameters explored, and the low

luminescence efficiency of the resulting nanowires, likely due to the presence of oxygen in the

growth system, which could not be eliminated without a substantial system upgrade. Thus,

alternatives to GaN nanowires were next considered. Here we note that the growth of high

quality GaN nanowires on Si should be possible and has been demonstrated with excellent

optical properties without seed-particle mediated growth, using PAMBE[132], selective area

epitaxy in MOCVD[225] and via an anisotropic growth method in MOCVD[226]. Furthermore,

p-type GaN nanowires grown on n-type Si have even been used as a solar cell[98]. These

successful implementations of GaN nanowires grown on Si are indicative of the great promise

for GaN nanowire-based LEDs on Si, if they can be grown under the proper conditions.

To further explore nanowire-based LEDs, we next turned to the synthesis of ZnO nanowires

in Chapter 4. To demonstrate the wide range of possibilities for ZnO nanowires we demonstrated

the low-temperature HTS growth of ZnO nanowires on both Si and the flexible plastic, PEN. The

HTS method allows for synthesis with good vertical alignment on an array of substrates, but the

length and nanowire areal density are not easily tuned. Thus, high-temperature VTC growth was

chosen as the synthesis method for this work. First, Au seed-particles were used for synthesis,

demonstrating a VLS growth mechanism. The VLS mechanism with ZnO growth did not provide

the usual control for other materials systems due to the anisotropic growth of ZnO and the

agglomeration of Au particles in the early stages of growth. For these reasons, and to avoid

potential deleterious effects of Au on ZnO nanowire optical properties, the synthesis method was

refined to growth on ZnO seed-layers. The growth mechanism of ZnO nanowires on continuous

ZnO seed-layers solution processed on Si substrates were investigated in depth. We

128

demonstrated control over the nanowire areal density via the O2 flow due to shifting nucleation

conditions on the seed-layer film. At low nanowire areal densities, the ZnO nanowire growth was

shown to be in a continuous nucleation regime in which nanowires continuously nucleate

throughout the growth and have fixed axial and radial growth rates. At higher nanowire areal

densities the growth was observed to be surface-diffusion limited such that the nanowire growth

rate depends inversely on the nanowire diameter. By developing this understanding we were able

to improve the length uniformity of ZnO nanowire arrays by employing a two-step growth. The

first step of the two-step growth, grown in the continuous nucleation growth regime at low

growth rates, defines large nanowire diameters. In the second stage of growth, grown in the

surface-diffusion limited regime, the nanowire growth rates are only slightly dependent on the

diameters, since the diameters set in the first-step are large enough to be relatively independent

of changes in diameter. This approach was extended to also control the nanowire areal density

via a sparse seed-layer, in which individual ZnO particles define the nanowire diameter. This

work developed the synthesis of scalable ZnO nanowire arrays with excellent vertical alignment

relative to the Si substrate, excellent length uniformity and control over the nanowire areal

density via an in-depth understanding of the growth mechanism. These nanowire arrays meet the

necessary requirements for nanowire array-based LEDs on Si, which has not been previously

reported with such control.

In Chapter 5, we investigated the optical properties of the ZnO nanowires synthesized in this

work. In general, though the HTS ZnO nanowires demonstrate broad defect luminescence, the

ZnO nanowires synthesized via VTC have excellent optical properties with bright near band-

edge luminescence and no defect luminescence. We observe some instances in which very dense

nanowire growth results in a 22 meV redshift due to a thermally induced shift in the band-edge

due to heating from the laser. For nanowire arrays grown on sparse ZnO seed-layers, the PL

spectra are independent of the nanowire areal density, though denser arrays are generally

brighter. Furthermore, typical ZnO nanowires grown with sparse seed-layers demonstrated a PL

quantum yield of ~15%. To induce green defect luminescence as a potential way to tune optical

emission from ZnO nanowire for white light, the ZnO nanowire arrays were annealed in N2. The

annealing temperature and time could be used to control the relative intensity of the green

luminescence, which is attributed to Zn vacancies. CL-STEM confirmed that the green

luminescence was a bulk property of the nanowire and was not just emitting from the surface,

129

indicating that this method could be a robust way to control the optical properties, allowing for

the wave-guiding of the emission to enhance extraction. This induced luminescence provides a

tunable alternative approach to QWs, which are not well-established in the ZnO materials.

Finally, in Chapter 6 we laid out the plan for device fabrication with the achievement of

controllable growth of ZnO nanowire arrays with excellent, tunable optical properties as a

critical starting point. We discussed the electrical properties in ZnO nanowires and developed an

approach to measuring these electrical properties. In this work both n-type and p-type doping

was observed due to processing and intrinsic defects. Though controllable extrinsic doping has

not yet been achieved in this work, the demonstration of both dopant types exemplifies the

promise of ZnO nanowires for LEDs. Approaches to filling the nanowire gaps with a dielectric

material to allow for top-contact deposition were also explored and discussed. The conclusion of

this exploration and discussion is that the best approach for overall fabrication purposes is the

use of a spin-on dielectric, which is more scalable than blanket deposition methods and allows

for further thermal processing, such as contact anneals, unlike spin-on polymers. With this work,

a clear path toward the scalable fabrication of ZnO nanowire array-based LEDs is laid out and

the groundwork for such fabrication is complete with the synthesis of ZnO nanowire arrays with

excellent length uniformity, vertical alignment and control over the nanowire areal density as

well as a general approach to understanding the efficiency and trade-offs for different wide band-

gap nanowire-based LED designs.

7.1.1. The ideal GaN nanowire-based LED

This work has focused on and discussed many of the practical advancements necessary to

achieve nanowire-based LEDs, however it has also explored the important roles of the design of

nanowire-based LEDs. Here, we present the ideal nanowire-based LED without considering the

technical challenges in the actual production of such a device. The ideal design of GaN

nanowire-based LEDs on Si begins with the vertical growth of m-directional GaN nanowires on

Si with a controlled areal density of ~15 nanowires/µm2 (with an approximate inter-nanowire

spacing of 260 nm). The nanowires in this design are grown as axial heterostructures, as shown

in Figure 7-1, with high carrier concentrations (n and p) to reduce resistive and heating losses,

likely on the order of 1018 cm-3. At higher carrier concentrations free-carrier absorption can

become significant[227]. Furthermore, the quantum-well should be placed not only to maximize

130

extraction efficiency, but also to limit free-carrier absorption, which is stronger from holes[227],

thus the p-type portion of the nanowire should be the top-most portion of the nanowire, which

should be much shorter than the n-type portion. Since the n-type portion is the bottom, the

devices should be grown on a highly doped n-type Si substrate. From the simulation work, to

maximize extraction efficiency for the case of an air superstrate, the quantum-well should be

placed 700 nm from the substrate. To maintain the high index of refraction contrast, the top-

contact should be deposited with an air-bridge[228]. The top-contact should also be a p-type

transparent conducting oxide, such as NiO, which has been used a hole-transport layer in organic

photovoltaics and can even be solution processed[229].

To tune the emission for solid-state lighting, a yellow phosphor is necessary to down-convert

the blue emission to yellow. This dichromatic approach to solid-state lighting depends on the

relative intensity of emission at each wavelength. For emission near 405 nm, as expected from

the InGaN quantum-wells modelled in this work, the optimal phosphor for producing high CRI

white-light would emit over a broad range in the green and yellow portion of the visible

spectrum. β-Sr2SiO4:Eu2+ has been demonstrated to fit the ideal emission spectrum to pair with

emission near 405 nm[230]. The thickness and coverage of the phosphor layer should be tuned to

maximize the CRI. The schematic for such a design is summarized in Figure 7-1.

Figure 7-1. Schematic of the ideal nanowire-based LED on Si. GaN-based nanowires with

the InGaN QW situated at 700 nm from the base at an inter-nanowire spacing of 260 nm.

131

7.2. Future Work

7.2.1. Controllable electrical properties via extrinsic dopants

The development of any nanowires-based LED design besides the single heterostructure (via

the nanowire/substrate interface) requires control over both n-type and p-type doping in ZnO

nanowires. While this work demonstrated examples of intrinsic defects and defect introduced by

processing causing both n-type and p-type dopants, for the synthesis of nanowire

heterostructures it is necessary to have a level of control not achievable with intrinsic defects,

which can relate to surface states and other nanowire properties as opposed to the synthesis and

processing itself. To achieve in-situ doping during the nanowire growth, the furnace used for

VTC synthesis would need to be upgraded to add the capability of introducing a dopant source.

The best approach here may be for the development of a solid source receptacle which can be

pushed into the active region of the furnace via a magnetic connection through the tube.

Alternatively, combinatorial doping could be used. In such an approach the base nanowire would

be grown n-type, using either the addition of Al or Ga2O3 powder to the precursor powder

mixture. The group III source has been shown to make group V elements such as N or P more

soluble in the ZnO system[231], thus using a capping approach to protect the nanowire base, the

tops of the nanowires could be doped with N, for example, by gas-phase doping using NH3 in a

separate furnace[85] (such as the MOCVD furnace used for GaN growth in this work).

Furthermore, the role of the hydrogen anneal will still be important, as it has been shown that

complexes formed with hydrogen, such as the N-H complex, is the actual acceptor in p-type

ZnO[232]. While controllable doping has been previously accomplished in ZnO nanowires, and

even some devices of nanowire homojunctions, the devices were limited in the application of

advanced designs and scalability due to the poor vertical alignment and areal density

control[233]. We believe combining our approach to nanowire synthesis with controllable

doping is the best path towards nanowire-based LEDs that have the potential for production

level.

7.2.2. Controlled defect luminescence for SH white nanowire-based LEDs

Another route to nanowire-based LEDs only requires developing control over one type of

dopant. By using the N2 anneal to control the ratio of the green to the near-UV luminescence of

the near band-edge, the CRI of emission from the nanowires can be tuned to a certain extent.

132

These devices could work with a heterojunction of n-type nanowires on a p-type Si substrate.

The key challenge in these devices is the electrical generation of carriers at the

substrate/nanowire interface. In this case it would be desirable to have a very heavily-doped Si

substrate, to encourage more of the depletion region of the heterojunction to be within the ZnO

nanowire, where the emission is desired.

7.2.3. Development of QW materials and nanowire-based LED design

Though ZnO is an earth-abundant, robust material system with high luminescent efficiency,

the development of ZnO for optoelectronics is limited by the lack of QW material with the

ability to tune emission via the composition of the material. The most viable material for QW

synthesis seems to be ZnCdO and the emission from these materials has reached as low as ~2.1

eV[234, 235]. With the further development of these materials, including the study of potential

polarization anisotropy and the extent of the QCSE across c-directional QWs a similar approach

to that taken in chapter 2 can be used to approach comparing the efficiency of feasible ZnO

nanowire-based LED designs. This design comparison could then guide the synthesis of large-

scale ZnO nanowire-based LEDs.

133

References

[1] Conti J J, Holtberg P D, Beamon J A, Schaal A M, Sweetnam G E and Kydes A S 2010 Annual Energy Outlook. (Washington, D.C.: Energy Information Administration)

[2] Schubert E F and Kim J K 2008 Transcending the replacement paradigm of solid-state lighting Optics Express 16 21835-42

[3] Humphreys C J 2008 Solid-State Lighting MRS Bulletin 33 459-70 [4] Schubert E F 2006 Light-Emitting Diodes (Cambridge, UK: Cambridge University Press) [5] Teke A and Morkoç H 2007 Group III-Nitrides vol 753-804: Springer US) [6] Look D C and Claflin B 2004 P-type doping and devices based on ZnO physica status

solidi (b) 241 624-30 [7] Ryu Y R, Lee T S, Leem J H and White H W 2003 Fabrication of homostructural ZnO p–

n junctions and ohmic contacts to arsenic-doped p-type ZnO Applied Physics Letters 83 4032

[8] Huang J, Chu S, Kong J, Zhang L, Schwarz C M, Wang G, Chernyak L, Chen Z and Liu J 2013 ZnO p-n Homojunction Random Laser Diode Based on Nitrogen-Doped p-type Nanowires Advanced Optical Materials 1 179-85

[9] Sang N X, Beng T C, Jie T, Fitzgerald E A and Jin C S 2013 Fabrication of p-type ZnO nanorods/n-GaN film heterojunction ultraviolet light-emitting diodes by aqueous solution method physica status solidi (a) 210 1618-23

[10] Yuan G D, Zhang W J, Jie J S, Fan X, Zapien J A, Leung Y H, Luo L B, Wang P F, Lee C S and Lee S T 2008 p-Type ZnO Nanowire Arrays NanoLetters 8 2591-7

[11] Lu J G, Fujita S, Kawaharamura T, Nishinaka H, Kamada Y, Ohshima T, Ye Z Z, Zeng Y J, Zhang Y Z, Zhu L P, He H P and Zhao B H 2007 Carrier concentration dependence of band gap shift in n-type ZnO:Al films Journal of Applied Physics 101 083705

[12] Tanaka H, Fujita S and Fujita S 2005 Fabrication of wide-band-gap MgxZn1−xO quasi-ternary alloys by molecular-beam epitaxy Applied Physics Letters 86 192911

[13] Lupan O, Pauporté T, Le Bahers T, Viana B and Ciofini I 2011 Wavelength-Emission Tuning of ZnO Nanowire-Based Light-Emitting Diodes by Cu Doping: Experimental and Computational Insights Advanced Functional Materials 21 3564-72

[14] Seo K, Lim T, Kim S, Park H-L and Ju S 2010 Tunable-white-light-emitting nanowire sources Nanotechnology 21 255201

[15] Echresh A, Chey C O, Shoushtari M Z, Nur O and Willander M 2014 Tuning the emission of ZnO nanorods based light emitting diodes using Ag doping Journal of

Applied Physics 116 193104 [16] Li J, Wei S-H, Li S-S and Xia J-B 2006 Design of shallow acceptors in ZnO: First-

principles band-structure calculations Physical Review B 74

[17] Jewell S and Kimball S M 2014 Mineral commodity summaries 2014. (Washington, D.C.: U.S. Geological Survey) p 196

[18] Lu I L, Wu Y-R and Singh J 2010 A study of the role of dislocation density, indium composition on the radiative efficiency in InGaN/GaN polar and nonpolar light-emitting diodes using drift-diffusion coupled with a Monte Carlo method Journal of Applied

Physics 108 124508 [19] Feltin E, Beaumont B, Vennéguès P, Riemann T b, Christen J, Faurie J P and Gibart P

2001 Epitaxial Lateral Overgrowth of GaN on Si (111) Physica status solidi (a) 188 733-7

134

[20] Wuu D S, Wang W K, Wen K S, Huang S C, Lin S H, Huang S Y, Lin C F and Horng R H 2006 Defect reduction and efficiency improvement of near-ultraviolet emitters via laterally overgrown GaN on a GaN/patterned sapphire template Applied Physics Letters 89 161105

[21] Liang C H, Chen L C, Hwang J S, Chen K H, Hung Y T and Chen Y F 2002 Selective-area growth of indium nitride nanowires on gold-patterned Si(100) substrates Applied

Physics Letters 81 22 [22] Mårtensson T, Svensson C P T, Wacaser B A, Larsson M W, Seifert W, Deppert K,

Gustafsson A, Wallenberg L R and Samuelson L 2004 Epitaxial III-V Nanowires on Silicon NanoLetters 4 19817-901

[23] Laudise R A and Ballman A A 1960 Hydrothermal Synthesis of Zinc Oxide and Zinc Sulfide Journal of Physical Chemistry 64 688-91

[24] Kozuka Y, Tsukazaki A and Kawasaki M 2014 Challenges and opportunities of ZnO-related single crystalline heterostructures Applied Physics Reviews 1 011303

[25] Lim S K, Crawford S and Gradečak S 2010 Growth mechanism of GaN nanowires: preferred nucleation site and effect of hydrogen Nanotechnology 21 345604

[26] Zhou X, Chesin J, Crawford S and Gradečak S 2012 Using seed particle composition to control structural and optical properties of GaN nanowires Nanotechnology 23 285603

[27] Guo M, Diao P and Cai S 2005 Hydrothermal growth of well-aligned ZnO nanorod arrays: Dependence of morphology and alignment ordering upon preparing conditions Journal of Solid State Chemistry 178 1864-73

[28] Park H, Chang S, Jean J, Cheng J J, Araujo P T, Wang M, Bawendi M G, Dresselhaus M S, Bulović V, Kong J and Gradečak S 2013 Graphene Cathode-Based ZnO Nanowire Hybrid Solar Cells Nano Letters 13 233-9

[29] Tang H, Chang J C, Shan Y, Ma D D D, Lui T-Y, Zapien J A, Lee C-S and Lee S-T 2008 Growth mechanism of ZnO nanowires via direct Zn evaporation Journal of Materials

Science 44 563-71 [30] Conley J F, Stecker L and Ono Y 2005 Directed assembly of ZnO nanowires on a Si

substrate without a metal catalyst using a patterned ZnO seed layer Nanotechnology 16 292-6

[31] Henneghien A-L, Gayral B, Désières Y and Gérard J-M 2009 Simulation of waveguiding and emitting properties of semiconductor nanowires with hexagonal or circular cross-sections Journal of the Optical Society of America B 26 2396-403

[32] Lee B G, Chen X, Biberman A, Liu X, Hsieh I-W, Chou C-Y, Dadap J I, Xia F, Green W M J, Sekaric L, Vlasov Y A, Osgood J, Richard M. and Bergman K 2008 Ultrahigh-Bandwidth Silicon Photonic Nanowire Waveguides for On-Chip Networks IEEE

Photonics Technology Letters 20 398-400 [33] Chesin J and Gradecak S 2014 Comparing directed efficiency of III-nitride nanowire

light-emitting diodes Journal of Nanophotonics 8 083095 [34] Diedenhofen S L, Janssen O T A, Hocevar M, Pierret A, Bakkers E P A M, Urbach H P

and Rivas J G 2011 Controlling the directional emission of light by periodic arrays of heterostructured semiconductor nanowires ACS Nano 5 5830-7

[35] Chesin J, Zhou X and Gradečak S 2014 Wide Band Gap Semiconductor Nanowires -

Heterostructures and Optoelectronic Devices, ed V Consonni and G Feuillet (London, UK: Wiley-IST) pp 161-78

135

[36] Akasaki I and Amano H 1992 Perspective of the UV/blue light emitting devices based on GaN and related compounds Optoelectronic Devices and Technology 7 49-56

[37] Cheng G, Kolmakov A, Zhang Y, Moskovits M, Munden R, Reed M A, Wang G, Moses D and Zhang J 2003 Current rectification in a single GaN nanowire with a well-defined p–n junction Applied Physics Letters 83 1578

[38] Umebu I 1993 The effects of dopant concentration on the properties of single-heterostructure AlGaAs LEDs with diffused and grown junctions Semiconductor Science

and Technology 8 283-9 [39] Huang Y, Duan X and Lieber C M 2004 Nanowires for Integrated Multicolor

Nanophotonics Small 1 142-7 [40] Speck J S and Chichibu S F 2009 Nonpolar and Semipolar Group III Nitride-Based

Materials MRS Bulletin 34 304-14 [41] Qian F, Gradečak S, Li Y, Wen C-Y and Lieber C M 2005 Core/Multishell Nanowire

Heterostructures as Multicolor High-Efficiency Light-Emitting Diodes NanoLetters 5 2287-91

[42] Brantley W A, Lorimor O G, Dapkus P D, Haszko S E and Saul R H 1975 Effect of dislocations on green electroluminescence efficiency in GaP grown by liquid phase epitaxy Journal of Applied Physics 46 2629

[43] Herzog A H 1972 High-Efficiency Zn-Diffused GaAs Electroluminescent Diodes Journal of Applied Physics 43 600

[44] Gibart P 2004 Metal organic vapour phase epitaxy of GaN and lateral overgrowth Reports on Progress in Physics 67 667-715

[45] Bennett S E 2010 Dislocations and their reduction in GaN Materials Science and

Technology 26 1017-28 [46] Lester S D, Ponce F A, Craford M G and Steigerwald D A 1995 High dislocation

densities in high efficiency GaN-based light-emitting diodes Applied Physics Letters 66 1249

[47] Look D C 2005 Electrical and optical properties of p-type ZnO Semiconductor Science

and Technology 20 S55-S61 [48] Sugahara T, Sato H, Hao M, Naoi Y, Kurai S, Tottori S, Yamashita K, Nishino K,

Romano L T and Sakai S 1998 Direct Evidence that Dislocations are Non-Radiative Recombination Centers in GaN Japanes Journal of Applied Physics 37 398-400

[49] Cherns D, Jiao C G, Mokhtari H, Cai J and Ponce F A 2002 Electron Holography Studies of the Charge on Dislocation in GaN physica status solidi (b) 234 924-30

[50] Cherns D and Jiao C 2001 Electron Holography Studies of the Charge on Dislocations in GaN Physical Review Letters 87

[51] McCluskey M D and Jokela S J 2009 Defects in ZnO Journal of Applied Physics 106 071101

[52] Schubert M F, Chhajed S, Kim J K, Schubert E F, Koleske D D, Crawford M H, Lee S R, Fischer A J, Thaler G and Banas M A 2007 Effect of dislocation density on efficiency droop in GaInN⁄GaN light-emitting diodes Applied Physics Letters 91 231114

[53] Look D C 2001 Recent advances in ZnO materials and devices Materials Science and

Engineering B 80 383-7 [54] Brewster M M, Zhou X, Lim S K and Gradečak S 2011 Role of Au in the Growth and

Nanoscale Optical Properties of ZnO Nanowires The Journal of Physical Chemistry

Letters 2 586-91

136

[55] Hierro A, Hansen M, Boeckl J J, Zhao L, Speck J S, Mishra U K, DenBaars S P and Ringel S A 2001 Carrier Trapping and Recombination at Point Defects and Dislocations in MOCVD n-GaN Physica status solidi (b) 228 937-46

[56] Reshchikov M A and Morkoç H 2005 Luminescence properties of defects in GaN Journal of Applied Physics 97 061301

[57] Lim J H, Kang C K, Kim K K, Park I K, Hwang D K and Park S J 2006 UV Electroluminescence Emission from ZnO Light-Emitting Diodes Grown by High-Temperature Radiofrequency Sputtering Advanced Materials 18 2720-4

[58] Chattopadhyay S, Ganguly A, Chen K-H and Chen L-C 2009 One-Dimensional Group III-Nitrides: Growth, Properties, and Applications in Nanosensing and Nano-Optoelectronics Critical Reviews in Solid State and Materials Sciences 34 224-79

[59] Stern E, Cheng G, Cimpoiasu E, Klie R, Guthrie S, Klemic J, Kretzschmar I, Steinlauf E, Turner-Evans D, Broomfield E, Hyland J, Koudelka R, Boone T, Young M, Sanders A, Munden R, Lee T, Routenberg D and Reed M A 2005 Electrical characterization of single GaN nanowires Nanotechnology 16 2941-53

[60] Fabbri F, Villani M, Catellani A, Calzolari A, Cicero G, Calestani D, Calestani G, Zappettini A, Dierre B, Sekiguchi T and Salviati G 2014 Zn vacancy induced green luminescence on non-polar surfaces in ZnO nanostructures Scientific Reports 4

[61] Zeuner A, Alves H, Hofmann D M, Meyer B K, Heuken M, Bläsing J and Krost A 2002 Structural and optical properties of epitaxial and bulk ZnO Applied Physics Letters 80 2078

[62] Shalish I, Temkin H and Narayanamurti V 2004 Size-dependent surface luminescence in ZnO nanowires Physical Review B 69

[63] Kappers M J, Moram M A, Sridhara Rao D V, McAleese C and Humphreys C J 2010 Low dislocation density GaN growth on high-temperature AlN buffer layers on (0001) sapphire Journal of Crystal Growth 312 363-7

[64] Arslan E, Duygulu Ö, Kaya A A, Teke A, Özçelik S and Ozbay E 2009 The electrical, optical, and structural properties of GaN epitaxial layers grown on Si(111) substrate with interlayers Superlattices and Microstructures 46 846-57

[65] Yang W, McPherson S A, Mao Z, McKernan S and Carter C B 1999 Single-crystal GaN pyramids grown on (111) Si substrates by selective lateral overgrowth Journal of Crystal

Growth 204 270-4 [66] Andeen D, Kim J H, Lange F F, Goh G K L and Tripathy S 2006 Lateral Epitaxial

Overgrowth of ZnO in Water at 90 °C Advanced Functional Materials 16 799-804 [67] Setiawan A, Vashaei Z, Cho M W, Yao T, Kato H, Sano M, Miyamoto K, Yonenaga I

and Ko H J 2004 Characteristics of dislocations in ZnO layers grown by plasma-assisted molecular beam epitaxy under different Zn⁄O flux ratios Journal of Applied Physics 96 3763

[68] Zhang K, Hao M R, Guo W, Heeg T, Schlom D G, Shen W Z and Pan X Q 2012 Low temperature electron transport in phosphorus-doped ZnO films grown on Si substrates Physica B: Condensed Matter 407 2825-8

[69] Pearton S J, Zolper J C, Shul R J and Ren F 1999 GaN: Processing, defects, and devices Journal of Applied Physics 86 1

[70] Djurišić A B, Leung Y H, Tam K H, Ding L, Ge W K, Chen H Y and Gwo S 2006 Green, yellow, and orange defect emission from ZnO nanostructures: Influence of excitation wavelength Applied Physics Letters 88 103107

137

[71] Li D, Leung Y H, Djurišić A B, Liu Z T, Xie M H, Shi S L, Xu S J and Chan W K 2004 Different origins of visible luminescence in ZnO nanostructures fabricated by the chemical and evaporation methods Applied Physics Letters 85 1601

[72] Nakamura S 1996 III-V nitride-based light-emitting diodes Diamond and Related

Materials 5 1996 [73] Chung B C and Gershenzon M 1992 The influence of oxygen on the electrical and optical

properties of GaN crystals grown by metalorganic vapor phase epitaxy Journal of

Applied Physics 72 651 [74] Matilla T and Nieminen R M 1996 Ab initio study of oxygen point defects in GaAs, GaN

and AlN Physical Review B 54 16 676-82 [75] Nakamura S, Iwasa N, Senoh M and Mukai T 1992 Hole Compensation Mechanism of p-

type GaN Films Japanes Journal of Applied Physics 31 1258-66 [76] Lyons J L, Janotti A and Van de Walle C G 2010 Carbon impurities and the yellow

luminescence in GaN Applied Physics Letters 97 152108 [77] Neugebauer J and Van de Walle C G 1994 Atomic Geometry and Electronic Structure of

Native Defects in GaN Physical Review B 50 8067-70 [78] Janotti A and Van de Walle C G 2009 Fundamentals of zinc oxide as a semiconductor

Reports on Progress in Physics 72 126501 [79] Kushwaha A and Aslam M 2013 Hydrogen-incorporated ZnO nanowire films: stable and

high electrical conductivity Journal of Physics D: Applied Physics 46 485104 [80] Mohanta A, Simmons J G, Everitt H O, Shen G, Margaret Kim S and Kung P 2014 Effect

of pressure and Al doping on structural and optical properties of ZnO nanowires synthesized by chemical vapor deposition Journal of Luminescence 146 470-4

[81] Yuan G D, Zhang W J, Jie J S, Fan X, Tang J X, Shafiq I, Ye Z Z, Lee C S and Lee S T 2008 Tunable n-Type Conductivity and Transport Properties of Ga-doped ZnO Nanowire Arrays Advanced Materials 20 168-73

[82] Tay C B, Tang J, Nguyen X S, Huang X H, Chai J W, Venkatesan V T and Chua S J 2012 Low Temperature Aqueous Solution Route to Reliable p-Type Doping in ZnO with K: Growth Chemistry, Doping Mechanism, and Thermal Stability The Journal of

Physical Chemistry C 116 24239-47 [83] Zeng Y J, Ye Z Z, Xu W Z, Li D Y, Lu J G, Zhu L P and Zhao B H 2006 Dopant source

choice for formation of p-type ZnO: Li acceptor Applied Physics Letters 88 062107 [84] Kao T-H, Chen J-Y, Chiu C-H, Huang C-W and Wu W-W 2014 Opto-electrical

properties of Sb-doped p-type ZnO nanowires Applied Physics Letters 104 111909 1-5 [85] Kalyanaraman S, Thangavel R and Vettumperumal R 2013 High mobility formation of p-

type Al doped ZnO:N films annealed under NH3 ambient Journal of Physics and

Chemistry of Solids 74 504-8 [86] Gao J, Zhang X, Sun Y, Zhao Q and Yu D 2010 Compensation mechanism in N-doped

ZnO nanowires Nanotechnology 21 245703 [87] Kang H S, Ahn B D, Kim J H, Kim G H, Lim S H, Chang H W and Lee S Y 2006

Structural, electrical, and optical properties of p-type ZnO thin films with Ag dopant Applied Physics Letters 88 202108

[88] Sanford N A, Blanchard P T, Bertness K A, Mansfield L, Schlager J B, Sanders A W, Roshko A, Burton B B and George S M 2010 Steady-state and transient photoconductivity in c-axis GaN nanowires grown by nitrogen-plasma-assisted molecular beam epitaxy Journal of Applied Physics 107 034318

138

[89] Simpkins B S, Mastro M A, Eddy C R and Pehrsson P E 2008 Surface depletion effects in semiconducting nanowires Journal of Applied Physics 103 104313

[90] Yoo J, Chon B, Tang W, Joo T, Dang L S and Yi G-C 2012 Excitonic origin of enhanced luminescence quantum efficiency in MgZnO/ZnO coaxial nanowire heterostructures Applied Physics Letters 100 223103

[91] Azize M and Palacios T 2011 Top-down fabrication of AlGaN/GaN nanoribbons Applied

Physics Letters 98 042103 [92] Chang S-W, Chuang V P, Boles S T, Ross C A and Thompson C V 2009 Densely Packed

Arrays of Ultra-High-Aspect-Ratio Silicon Nanowires Fabricated using Block-Copolymer Lithography and Metal-Assisted Etching Advanced Functional Materials 19 2495-500

[93] Fortuna S A and Li X 2010 Metal-catalyzed semiconductor nanowires: a review on the control of growth directions Semiconductor Science and Technology 25 024005

[94] Debnath R K, Meijers R, Richter T, Stoica T, Calarco R and Lüth H 2007 Mechanism of molecular beam epitaxy growth of GaN nanowires on Si(111) Applied Physics Letters 90 123117

[95] Greene L E, Yuhas B D, Law M, Zitoun D and Yang P 2006 Solution-Grown Zinc Oxide Nanowires Inorganic Chemistry 45 7535-43

[96] Fang F, Zhao D X, Zhang J Y, Shen D Z, Lu Y M, Fan X W, Li B H and Wang X H 2008 The influence of growth temperature on ZnO nanowires Materials Letters 62 1092-5

[97] Kang S M, Kang B K and Yoon D H 2011 Growth and properties of vertically well-aligned GaN nanowires by thermal chemical vapor deposition process Materials Letters 65 763-5

[98] Tang Y B, Chen Z H, Song H S, Lee C S, Cong H T, Cheng H M, Zhang W J, Bello I and Lee S T 2008 Vertically Aligned p-Type Single-Crystalline GaN Nanorod Arrays on n-Type Si for Heterojunction Photovoltaic Cells Nano Letters 8 4191-5

[99] Largeau L, Dheeraj D L, Tchernycheva M, Cirlin G E and Harmand J C 2008 Facet and in-plane crystallographic orientations of GaN nanowires grown on Si(111) Nanotechnology 19 155704

[100] Yang P, Yan H, Mao S, Russo R, Johnson J, Saykally R, Morris N, Pham J, He R and Choi H-J 2002 Controlled Growth of ZnO Nanowires and their Optical Properties Advanced Functional Materials 12 323-31

[101] Wu J-J and Liu S-C 2002 Catalyst-Free Growth and Characterization of ZnO Nanorods Journal of Physical Chemistry B 106 9546-51

[102] Liu D F, Xiang Y J, Zhang Z X, Wang J X, Gao Y, Song L, Liu L F, Dou X Y, Zhao X W, Luo S D, Wang C Y, Zhou W Y, Wang G and Xie S S 2005 Growth of ZnO hexagonal nanoprisms Nanotechnology 16 2665-9

[103] Greene L E, Law M, Tan D H, Montano M, Goldberger J, Somorjai G and Yang P 2005 General Route to Vertical ZnO Nanowire Arrays Using Textured ZnO Seeds NanoLetters 5 1231-6

[104] Guillemin S, Rapenne L, Roussel H, Sarigiannidou E, Brémond G and Consonni V 2013 Formation Mechanisms of ZnO Nanowires: The Crucial Role of Crystal Orientation and Polarity The Journal of Physical Chemistry C 117 20738-45

[105] Kim H-J, Sung K, An K-S, Lee Y K, Kim C G, Lee Y-H and Kim Y 2004 ZnO nanowhiskers on ZnO nanoparticle-deposited Si(111) by MOCVD Journal of Materials

Chemistry 14 3396

139

[106] Baxter J B and Aydil E S 2005 Epitaxial growth of ZnO nanowires on a- and c-plane sapphire Journal of Crystal Growth 274 407-11

[107] Lupan O, Guérin V M, Tiginyanu I M, Ursaki V V, Chow L, Heinrich H and Pauporté T 2010 Well-aligned arrays of vertically oriented ZnO nanowires electrodeposited on ITO-coated glass and their integration in dye sensitized solar cells Journal of Photochemistry

and Photobiology A: Chemistry 211 65-73 [108] Gardner N F, Kim J C, Wierer J J, Shen Y C and Krames M R 2005 Polarization

anisotropy in the electroluminescence of m-plane InGaN–GaN multiple-quantum-well light-emitting diodes Applied Physics Letters 86 111101

[109] Kolbe T, Knauer A, Chua C, Yang Z, Einfeldt S, Vogt P, Johnson N M, Weyers M and Kneissl M 2010 Optical polarization characteristics of ultraviolet (In)(Al)GaN multiple quantum well light emitting diodes Applied Physics Letters 97 171105

[110] Jarašinas K, ajev P, Nargelas S, Aleksiejnas R, Leach J, Paskova T, Okur S, Ozgür U and Morkoc H 2012 Recombination and diffusion processes in polar and nonpolar bulk GaN investigated by time-resolved photoluminescence and nonlinear optical techniques(San

Francisco, CA, (vol 8262) ed J-I Chyi, et al.: SPIE) p 82620G-1-10 [111] Mochida N, Honda T, Shirasawa T, Inoue A, Sakaguchi T, Koyama F and Iga K 1998

Crystal orientation dependence of p-type contact resistance of GaN Journal of Crystal

Growth 189–190 716-9 [112] Park S-H 2002 Crystal orientation effects on electronic properties of wurtzite

InGaN/GaN quantum wells Journal of Applied Physics 91 9904 [113] Yeh T-W, Lin Y-T, Stewart L S, Dapkus P D, Sarkissian R, O’Brien J D, Ahn B and Nutt

S R 2012 InGaN/GaN multiple quantum wells grown on nonpolar facets of vertical GaN nanorod arrays Nano Letters 12 3257-62

[114] Guo W, Zhang M, Banerjee A and Bhattacharya P 2010 Catalyst-free InGaN/GaN nanowire light emitting diodes grown on (001) silicon by molecular beam epitaxy Nano

Letters 10 3355-9 [115] Nguyen H P T, Cui K, Zhang S, Fathololoumi S and Mi Z 2011 Full-color InGaN/GaN

dot-in-a-wire light emitting diodes on silicon Nanotechnology 22 445202 [116] Qian F, Gradečak S, Li Y, Wen C-Y and Lieber C M 2005 Core/multishell nanowire

heterostructures as multicolor, high-efficiency light-emitting diodes Nano Letters 5 2287-91

[117] Qian F, Li Y, Gradečak S, Wang D, Barrelet C J and Liebe C M 2004 Gallium nitride-based nanowire radial heterostructures for nanophotonics Nano Letters 4 1975-9

[118] Henneghien A-L, Gayral B, Désières Y and Gérard J-M 2009 Simulation of waveguiding and emitting properties of semiconductor nanowires with hexagonal or circular sections J. Opt. Soc. Am. B 26 2396-403

[119] Chang S J, Chang C S, Su Y K, Lee C T, Chen W S, Shen C F, Hsu Y P, Shei S C and Lo H M 2005 Nitride-Based Flip-Chip ITO LEDs IEEE Transactions on Advanced

Packaging 28 273-7 [120] Fujii T, Gao Y, Sharma R, Hu E L, DenBaars S P and Nakamura S 2004 Increase in the

extraction efficiency of GaN-based light-emitting diodes via surface roughening Applied

Physics Letters 84 855 [121] McGroddy K, David A, Matioli E, Iza M, Nakamura S, DenBaars S, Speck J S,

Weisbuch C and Hu E L 2008 Directional emission control and increased light extraction in GaN photonic crystal light emitting diodes Applied Physics Letters 93 103502

140

[122] Gardner N F, Kim J C, Wierer J J, Shen Y C and Krames M R 2005 Polarization anisotropy in the electroluminescence of m-plane InGaN–GaN multiple-quantum-well light-emitting diodes Appl. Phys. Lett. 86 111101

[123] Kolbe T, Knauer A, Chua C, Yang Z, Einfeldt S, Vogt P, Johnson N M, Weyers M and Kneissl M 2010 Optical polarization characteristics of ultraviolet (In)(Al)GaN multiple quantum well light emitting diodes Appl. Phys. Lett. 97 171105

[124] Ghosh S, Waltereit P, Brandt O, Grahn H T and Ploog K H 2002 Polarization-dependent spectroscopic study of M-plane GaN on γ-LiAlO2 Appl. Phys. Lett. 80 413

[125] Park S-H 2002 Crystal orientation effects on electronic properties of wurtzite InGaN/GaN quantum wells J. Appl. Phys. 91 9904

[126] Connors B J 2011 Simulation of Current Crowding Mitigation in GaN Core-Shell Nanowire LED Designs. In: Electrical and Computer Engineering, (Atlanta: Georgia Institute of Technology) p 67

[127] Heckel R W 1997 ASM Handbook Volume 20: Materials Selection and Desing, ed G E Dieter: ASM International)

[128] Dolmanan S B, Teo S L, Lin V K, Hui H K, Dadgar A, Krost A and Tripathy S 2011 Thin-film InGaN⁄GaN Vertical Light Emitting Diodes Using GaN on Silicon-On-Insulator Substrates Electrochemical and Solid-State Letters 14 H460

[129] Meneghini M, Tazzoli A, Mura G, Meneghesso G and Zanoni E 2010 A Review on the Physical Mechanisms that Limit the Reliability of GaN-Based LEDs IEEE TRansactions

on Electron Devices 57 108-19 [130] Wybourne M N 1998 Properties of Crystalline Silicon, ed M R Brozel (London: The

Institution of Electrical Engineers) [131] Yeh T-W, Lin Y-T, Stewart L S, Dapkus P D, Sarkissian R, O’Brien J D, Ahn B and Nutt

S R 2012 InGaN/GaN multiple quantum wells grown on nonpolar facets of vertical GaN nanorod arrays Nano Lett. 12 3257-62

[132] Guo W, Zhang M, Banerjee A and Bhattacharya P 2010 Catalyst-free InGaN/GaN nanowire light emitting diodes grown on (001) silicon by molecular beam epitaxy Nano

Lett. 10 3355-9 [133] Nguyen H P T, Cui K, Zhang S, Fathololoumi S and Mi Z 2011 Full-color InGaN/GaN

dot-in-a-wire light emitting diodes on silicon Nanotech. 22 445202 [134] Qian F, Gradečak S, Li Y, Wen C-Y and Lieber C M 2005 Core/multishell nanowire

heterostructures as multicolor, high-efficiency light-emitting diodes Nano Lett. 5 2287-91 [135] Qian F, Li Y, Gradečak S, Wang D, Barrelet C J and Liebe C M 2004 Gallium nitride-

based nanowire radial heterostructures for nanophotonics Nano Lett. 4 1975-9 [136] Maslov A V, Bakunov M I and Ning C Z 2006 Distribution of optical emission between

guided modes and free space in a semiconductor nanowire J. Appl. Phys. 99 024314 [137] Seo M-K, Yang J-K, Jeong K-Y, Park H-G, Qian F, Ee H-S, No Y-S and Lee Y-H 2008

Modal Characteristics in Single NW Cavity with Triangular Cross-Section Nano Lett. 8 4534-8

[138] Witzigmann B, Kölper C, Sabathil M, Witzigmann B, Römer F, Bergbauer W, Strassburg M, Henneberger F, Arakawa Y and Freundlich A 2011 Optical properties of individual GaN nanorods for light emitting diodes: influence of geometry, materials, and facets Proc. SPIE 7933 793314

141

[139] Kölper C, Sabathil M, Römer F, Mandl M, Strassburg M and Witzigmann B 2012 Core-shell InGaN nanorod light emitting diodes: Electronic and optical device properties Phys.

Stat. Sol. A 209 2304-12 [140] Chutinan A, Ishihara K, Asano T, Fujita M and Noda S 2005 Theoretical analysis on

light-extraction efficiency of organic light-emitting diodes using FDTD and mode-expansion methods Org. Elec. 6 3-9

[141] Chesin J, Zhou X and Gradečak S 2012 Light extraction in individual GaN nanowires on Si for LEDs Proc. SPIE 8464 846703

[142] Chesin J, Zhou X and Gradečak S 2012 Light extraction in individual GaN nanowires on Si for LEDs Proc. of SPIE 8467 846703 1-10

[143] Lim S K, Crawford S, Haberfehlner G and Gradečak S 2013 Controlled Modulation of Diameter and Composition along Individual III–V Nitride Nanowires Nano Letters 13 331-6

[144] Krames M R, Shchekin O B, Mueller-Mach R, Mueller G O, Zhou L, Harbers G and Craford M G 2007 Status and future of high-power light-emitting diodes for solid-state lighting J. Disp. Tech. 3 160-75

[145] Bulashevich K A, Khokhlev O V, Evstratov I Y and Karpov S Y 2012 Simulation of light-emitting diodes for new physics understanding and device design Proc. SPIE 8278 827819

[146] Nguyen H P T, Cui K, Zhang S, Djavid M, Korinek A, Botton G A and Mi Z 2012 Controlling electron overflow in phosphor-free InGaN/GaN nanowire white light-emitting diodes Nano Lett. 12 1317-23

[147] Nguyen H P T, Djavid M, Cui K and Mi Z 2012 Temperature-dependent nonradiative recombination processes in GaN-based nanowire white-light-emitting diodes on silicon Nanotech. 23 194012

[148] Shen Y C, Mueller G O, Watanabe S, Gardner N F, Munkholm A and Krames M R 2007 Auger recombination in InGaN measured by photoluminescence Appl. Phys. Lett. 91 141101

[149] Jarašinas K, Ščajev P, Nargelas S, Aleksiejūnas R, Leach J, Paskova T, Okur S, Özgür Ü and Morkoç H 2012 Recombination and diffusion processes in polar and nonpolar bulk GaN investigated by time-resolved photoluminescence and nonlinear optical techniques Proc. SPIE 8262 82620G-1-10

[150] Chen H-Y, Chen R-S, Rajan N, Chang F-C, Chen L-C, Chen K-H, Yang Y-J and Reed M 2011 Size-dependent persistent photocurrent and surface band bending in m-axial GaN nanowires Phys. Rev. B 84

[151] Gorgis A, Flissikowski T, Brandt O, Chèze C, Geelhaar L, Riechert H and Grahn H 2012 Time-resolved photoluminescence spectroscopy of individual GaN nanowires Phys. Rev.

B 86

[152] Simpkins B S, Mastro M A, Eddy C R and Pehrsson P E 2008 Surface depletion effects in semiconducting nanowires J. Appl. Phys. 103 104313

[153] Ni X, Lee J, Wu M, Li X, Shimada R, Özgür U, Baski A A, Morkoç H, Paskova T, Mulholland G and Evans K R 2009 Internal quantum efficiency of c-plane InGaN and m-plane InGaN on Si and GaN Appl. Phys. Lett. 95 101106

[154] Chyi J-I, Park S-H, Ahn D, Nanishi Y, Morkoç H, Piprek J, Yoon E and Fujioka H 2013 Nonpolar and semipolar GaN, optical gain and efficiency Proc. SPIE 8625 862511

142

[155] Lee M, Cheng J-H, Lee Y C, Seghete D, George S M, Schlager J B, Bertness K and Sanford N A 2009 Packaging and interconnect technologies for the development of GaN

nanowire-based light emitting diodes vol 978 ed A Shapiro and A Rosiewicz: IEEE) p 843-8

[156] Su Y-K, Chen K C, Lin C-L and Hsu H-C 2011 Enhancement in Light Extraction of GaN-Based Light-Emitting Diodes With High Reflectivity Electrodes IEEE Phot. Tech.

Lett. 23 1793-5 [157] Tambe M J, Ren S and Gradečak S 2010 Effects of gold diffusion on n-type doping of

GaAs nanowires Nano Lett. 10 4584-9 [158] Henneghien A-L, Tourbot G, Daudin B, Lartigue O, Désières Y and Gérard J-M 2011

Optical anisotropy and light extraction efficiency of MBE grown GaN nanowires epilayers Opt. Exp. 19 527-39

[159] Crawford S, Lim S K and Gradečak S 2013 Fundamental Insights into Nanowire Diameter Modulation and the Liquid/Solid Interface Nano Letters 13 226-32

[160] Jeon H C, Park S H, Lee S J, Kang T W and Ihm G 2013 Effects of applied electric fields on the optical properties of CdZnO quantum well structures EPL (Europhysics Letters) 104 67003

[161] Hebal H and Abid H 2014 Crystal orientation effects on electronic and optical properties of wurtzite ZnO/CdZnO quantum well lasers Superlattices and Microstructures 75 866-78

[162] Jagannathan H, Nishi Y, Reuter M, Copel M, Tutuc E, Guha S and Pezzi R P 2006 Effect of oxide overlayer formation on the growth of gold catalyzed epitaxial silicon nanowires Applied Physics Letters 88 103113

[163] Schmid H, Björk M T, Knoch J, Riel H, Riess W, Rice P and Topuria T 2008 Patterned epitaxial vapor-liquid-solid growth of silicon nanowires on Si(111) using silane Journal

of Applied Physics 103 024304 [164] Lim S K, Tambe M J, Brewster M M and Gradečak S 2008 Controlled Growth of

Ternary Alloy Nanowires Using Metalorganic Chemical Vapor Deposition NanoLetters 8 1386-92

[165] Tambe M J, Lim S K, Smith M J, Allard L F and Gradečak S 2008 Realization of defect-free epitaxial core-shell GaAs/AlGaAs nanowire heterostructures Applied Physics Letters 93 151917

[166] Wacaser B A, Dick K A, Johansson J, Borgström M T, Deppert K and Samuelson L 2009 Preferential Interface Nucleation: An Expansion of the VLS Growth Mechanism for Nanowires Advanced Materials 21 153-65

[167] Dick K A 2008 A review of nanowire growth promoted by alloys and non-alloying elements with emphasis on Au-assisted III–V nanowires Progress in Crystal Growth and

Characterization of Materials 54 138-73 [168] Zhao D, Andreazza C, Andreazza P, Ma J, Liu Y and Shen D 2004 Temperature-

dependent growth mode and photoluminescence properties of ZnO nanostructures Chemical Physics Letters 399 522-6

[169] Wang G Z, Ma N G, Deng C J, Yu P, To C Y, Hung N C, Aravind M and Ng D H L 2004 Large-scale synthesis of aligned hexagonal ZnO nanorods using chemical vapor deposition Materials Letters 58 2195-8

[170] Song J and Lim S 2007 Effect of Seed Layer on the Growth of ZnO Nanorods Journal of

Physical Chemistry C 111 596-600

143

[171] Rajabi M, Dariani R S and Iraji zad A 2013 Effect of growth conditions on zinc oxide nanowire array synthesized on Si (100) without catalyst Materials Science in

Semiconductor Processing 16 171-8 [172] Dayeh S A and Picraux S T 2010 Direct Observation of Nanoscale Size Effects in Ge

Semiconductor Nanowire Growth Nano Letters 10 4032-9 [173] Saunders R B, Garry S, Byrne D, Henry M O and McGlynn E 2012 Length versus Radius

Relationship for ZnO Nanowires Grown via Vapor Phase Transport Crystal Growth &

Design 12 5972-9 [174] Guo Z, Andreazza-Vignolle C, Andreazza P, Sauvage T, Zhao D X, Liu Y C, Yao B,

Shen D Z and Fan X W 2011 Tuning the growth of ZnO nanowires Physica B:

Condensed Matter 406 2200-5 [175] Tian J-H, Hu J, Li S-S, Zhang F, Liu J, Shi J, Li X, Tian Z-Q and Chen Y 2011 Improved

seedless hydrothermal synthesis of dense and ultralong ZnO nanowires Nanotechnology 22 245601

[176] Xu S, Lao C, Weintraub B and Wang Z L 2008 Density-controlled growth of aligned ZnO nanowire arrays by seedless chemical approach on smooth surfaces Journal of

Materials Research 23 2072-7 [177] Liu J, She J, Deng S, Chen J and Xu N 2008 Ultrathin Seed-Layer for Tuning Density of

ZnO Nanowire Arrays and Their Field Emission Characteristics Journal of Physical

Chemistry C 112 11685-90 [178] Li M-Y, Sui M, Kim E-S and Lee J 2014 Droplets to Merged Nanostructures: Evolution

of Gold Nanostructures by the Variation of Deposition Amount on Si(111) Crystal

Growth & Design 14 1128-34 [179] Guo D L, Tan L H, Wei Z P, Chen H and Wu T 2013 Density-Controlled Synthesis of

Uniform ZnO Nanowires: Wide-Range Tunability and Growth Regime Transition Small 9 2069-75

[180] Kar A, Low K-B, Oye M, Stroscio M A, Dutta M, Nicholls A and Meyyappan M 2011 Investigation of Nucleation Mechanism and Tapering Observed in ZnO Nanowire Growth by Carbothermal Reduction Technique Nano Express 6 9

[181] Tavendale A J and J P S 1983 Deep level, quenched-in defects in silicon doped with gold, silver, iron, copper or nickel Journal of Physical Chemistry C 16 1665-73

[182] Bafghi M S, Karimi M and Adeli M 2013 A Kinetic Study on the Carbothermic Reduction of Zinc Oxide from Electric Arc Furnace Dust Iranian Journal of Materials

Science and Engineering 10 18-30 [183] Kim B-S, Yoo J-M, Park J-T and Lee J-C 2006 A Kinetic Study of the Carbothermic

Reduction of Zinc Oxide with Various Additives Materials Transactions 47 2421-6 [184] Kubaschewski O and Evans E L 1951 Metallurgical Thermochemistry (New York:

Academic Press) [185] Ragone D V 1995 Thermodynamics of Materials vol 1: Wiley) [186] Fang F, Zhao D X, Zhang J Y, Shen D Z, Lu Y M, Fan X W, Li B H and Wang X H 2007

Growth of well-aligned ZnO nanowire arrays on Si substrate Nanotechnology 18 235604 [187] Meyer B and Marx D 2003 Density-functional study of the structure and stability of ZnO

surfaces Physical Review B 67 035403 [188] Fan H J, Bertram F, Dadgar A, Christen J, Krost A and Zacharias M 2004 Self-assembly

of ZnO nanowires and the spatial resolved characterization of their luminescence Nanotechnology 15 1401-4

144

[189] Baxter J B, Wu F and Aydil E S 2003 Growth mechanism and characterization of zinc oxide hexagonal columns Applied Physics Letters 83 3797

[190] Dubrovskii V, Cirlin G, Soshnikov I, Tonkikh A, Sibirev N, Samsonenko Y and Ustinov V 2005 Diffusion-induced growth of GaAs nanowhiskers during molecular beam epitaxy: Theory and experiment Physical Review B 71

[191] Oh S, Jung M, Koo J, Cho Y, Choi S, Yi S, Kil G and Chang J 2010 The mechanism of ZnO nanorod growth by vapor phase transportation Physica E: Low-dimensional Systems

and Nanostructures 42 2285-8 [192] Jensen L E, Bjork M T, Jeppesen S, Persson A I, Ohlsson B J and Samuelson L 2004

Role of Surface Diffusion in Chemical Beam Epitaxy of InAs Nanowires NanoLetters 4 1961-4

[193] Hosseini Teherani F, Rojo J C, Liang S, Chen H, Dudley M and Litton C W 2006 Physical vapor transport crystal growth of ZnO Proc. of SPIE 6122 61220Q-Q-8

[194] Göpel W 1978 Reactions of oxygen with ZnO–1010-surfaces Journal of Vacuum Science

and Technology 15 1298 [195] Chaparro S A, Zhang Y, Drucker J, Chandrasekhar D and Smith D J 2000 Evolution of

Ge/Si(100) islands: Island size and temperature dependence Journal of Applied Physics 87 2245

[196] Tersoff J, Denier van der Gon A and Tromp R 1994 Critical island size for layer-by-layer growth Physical Review Letters 72 266-9

[197] Jagadish C and Pearton S J eds 2006 Zinc Oxide Bulk, Thin Films and Nanostructures -

Processing, Properties and Applications (Oxford, UK: Elsevier) [198] Djurišić A B, Leung Y H, Tam K H, Hsu Y F, Ding L, Ge W K, Zhong Y C, Wong K S,

Chan W K, Tam H L, Cheah K W, Kwok W M and Phillips D L 2007 Defect emissions in ZnO nanostructures Nanotechnology 18 095702

[199] Pauzauskie P J and Yang P 2006 Nanowire photonics Materials Today 9 36-45 [200] Rodnyi P A and Khodyuk I V 2011 Optical and Luminescence Properties of Zinc Oxide

Optics and Spectroscopy 111 776-85 [201] Vanmaekelbergh D and van Vugt L K 2011 ZnO nanowire lasers Nanoscale 3 2783 [202] Maeda K, Sato M, Niikura I and Fukuda T 2005 Growth of 2 inch ZnO bulk single

crystal by the hydrothermal method Semiconductor Science and Technology 20 S49-S54 [203] Wang Z G, Zu X T, Zhu S and Wang L M 2006 Green luminescence originates from

surface defects in ZnO nanoparticles Physica E: Low-dimensional Systems and

Nanostructures 35 199-202 [204] Zhou X, Kuang Q, Jiang Z-Y, Xie Z-X, Xu T, Huang R-B and Zheng L-S 2007 The

Origin of Green Emission of ZnO Microcrystallites: Surface-Dependent Light Emission Studied by Cathodoluminescence Journal of Physical Chemistry C 111 12091-3

[205] Kong Y C, Yu D P, Zhang B, Fang W and Feng S Q 2001 Ultraviolet-emitting ZnO nanowires synthesized by a physical vapor deposition approach Applied Physics Letters 78 407

[206] Li L, Qiu X and Li G 2005 Correlation between size-induced lattice variations and yellow emission shift in ZnO nanostructures Applied Physics Letters 87 124101

[207] Liao Z-M, Zhang H-Z, Zhou Y-B, Xu J, Zhang J-M and Yu D-P 2008 Surface effects on photoluminescence of single ZnO nanowires Physics Letters A 372 4505-9

145

[208] Ozgur U, Gu X, Chevtchenko S, Spradlin J, Cho S-J, Morkoç H, Pollak F H, Everitt H O, Nemeth B and Nause J E 2006 Thermal Conductivity of Bulk ZnO after Different Thermal Treatments Journal of Electronic Materials 35 550-5

[209] Yang Y, Yan H, Fu Z, Yang B, Xia L, Xu Y, Zuo J and Li F 2006 Photoluminescence Investigation Based on Laser Heating Effect in ZnO-Ordered Nanostructures Journal of

Physical Chemistry B 110 846-52 [210] Chang P-C, Fan Z, Chien C-J, Stichtenoth D, Ronning C and Lu J G 2006 High-

performance ZnO nanowire field effect transistors Applied Physics Letters 89 133113 [211] Liao Z-M, Liu K-J, Zhang J-M, Xu J and Yu D-P 2007 Effect of surface states on

electron transport in individual ZnO nanowires Physics Letters A 367 207-10 [212] Parkinson P, Joyce H J, Gao Q, Tan H H, Zhang X, Zou J, Jagadish C, Herz L M and

Johnston M B 2009 Carrier Lifetime and Mobility Enhancement in Nearly Defect-Free Core-Shell Nanowires Measured Using Time-Resolved Terahertz Spectroscopy NanoLetters 9 3349-53

[213] Jiang X, Xiong Q, Nam S, Qian F, Li Y and Lieber C M 2007 InAs/InP Radial Nanowire Heterostructures as High Electron Mobility Devices NanoLetters 7 3214-8

[214] Mohney S E, Wang Y, Cabassi M A, Lew K K, Dey S, Redwing J M and Mayer T S 2005 Measuring the specific contact resistance of contacts to semiconductor nanowires Solid-State Electronics 49 227-32

[215] Wunnicke O 2006 Gate capacitance of back-gated nanowire field-effect transistors Applied Physics Letters 89 083102 1-3

[216] Hong W-K, Sohn J I, Hwang J S, Kwon S-S, Jo G, Song S, Seong-Min K, Ko H J, Park S J, Welland M E and Lee T 2008 Tunable Electronic Transport Characteristics of Surface-Architecture-Controlled ZnO Nanowire Field Effect Transistors NanoLetters 8 950-6

[217] Song S, Hong W-K, Kwon S-S and Lee T 2008 Passivation effects on ZnO nanowire field effect transistors under oxygen, ambient, and vacuum environments Applied Physics

Letters 92 263109 [218] Heo Y W, Tien L C, Norton D P, Kang B S, Ren F, Gila B P and Pearton S J 2004

Electrical transport properties of single ZnO nanorods Applied Physics Letters 85 2002 [219] Chen C, He H, Lu Y, Wu K and Ye Z 2013 Surface Passivation Effect on the

Photoluminescence of ZnO Nanorods ACS Applied Materials & Interfaces 5 6354-9 [220] Hong W-K, Jo G, Kwon S-S, Song S and Lee T 2008 Electrical Properties of Surface-

Tailored ZnO Nanowire Field-Effect Transistors IEEE Transactions on Electron Devices 55 3020-30

[221] Prasad A, Pandey A, Kunapuli V K, Bergstrom P L and Yap Y K 2012 Ambipolar Behaviors of Hydrogen-Incorporated ZnO Nanowires The Journal of Physical Chemistry

C 116 8210-5 [222] Jiang M, Xue X D, Chen Z Q, Liu Y D, Liang H W, Zhang H J and Kawasuso A 2013

Defects and acceptor centers in ZnO introduced by C+-implantation Journal of Materials

Science 49 1994-9 [223] Nagar S and Chakrabarti S 2014 Realization of reliable p-type ZnO thin films by nitrogen

implantation using plasma immersion ion implantation Superlattices and Microstructures 75 9-16

[224] Soci C, Zhang A, Xiang B, Dayeh S A, Aplin D P R, Park J, Bao X Y, Lo Y H and Wang D 2007 ZnO Nanowire UV Photodetectors with High Internal Gain NanoLetters 7 1003-9

146

[225] Park J-H, Navamathavan R and Lee C-R 2012 Size effects of nano-pattern in Si(111) substrate on the selective growth behavior of GaN nanowires by MOCVD Materials

Research Bulletin 47 836-42 [226] Ra Y-H, Navamathavan R, Cha J-H, Song K-Y, Lim H-C, Park J-H, Kim D-W and Lee

C-R 2010 Effect of H2Carrier Gas on the Growth of GaN Nanowires on Si(111) Substrates by Metalorganic Chemical Vapor Deposition Japanese Journal of Applied

Physics 49 045004 [227] Kioupakis E, Rinke P, Schleife A, Bechstedt F and Van de Walle C G 2010 Free-carrier

absorption in nitrides from first principles Physical Review B 81

[228] Bryllert T, Wernersson L-E, Löwgren T and Samuelson L 2006 Vertical wrap-gated nanowire transistors Nanotechnology 17 S227-S30

[229] Steirer K X, Chesin J P, Widjonarko N E, Berry J J, Miedaner A, Ginley D S and Olson D C 2010 Solution deposited NiO thin-films as hole transport layers in organic photovoltaics Organic Electronics 11 1414-8

[230] Sun X, Zhang J, Zhang X, Luo Y and Wang X 2008 A green-yellow emitting β-Sr2SiO4:Eu2+ phosphor for near ultraviolet chip white-light-emitting diode Journal of

Rare Earths 26 421-4 [231] Singh A V, Mehra R M, Wakahara A and Yoshida A 2003 p-type conduction in codoped

ZnO thin films Journal of Applied Physics 93 396 [232] Reynolds J G, Reynolds C L, Mohanta A, Muth J F, Rowe J E, Everitt H O and Aspnes D

E 2013 Shallow acceptor complexes in p-type ZnO Applied Physics Letters 102 152114 [233] Chen M-T, Lu M-P, Wu Y-J, Song J, Lee C-Y, Lu M-Y, Chang Y-C, Chou L-J, Wang Z

L and Chen L-J 2010 Near UV LEDs Made with in Situ Doped p-n Homojunction ZnO Nanowire Arrays Nano Letters 10 4387-93

[234] Cheng C L, Bo, Sie E J, Zhou W, Zhang J, Gong H, Huan C H A, Sum T C, Sun H and Fan H J 2010 ZnCdO/ZnO Coaxial Multiple Quantum Well Nanowire Heterostructures and Optical Properties Journal of Physical Chemistry C 114 3863-6868

[235] Sadofev S, Blumstengel S, Cui J, Puls J, Rogaschewski S, Schäfer P and Henneberger F 2006 Visible band-gap ZnCdO heterostructures grown by molecular beam epitaxy Applied Physics Letters 89 201907

[236] Lee S-K, Choi H-J, Pauzauskie P, Yang P, Cho N-K, Park H-D, Suh E-K, Lim K-Y and Lee H-J 2004 Gallium nitride nanowires with a metal initiated metal-organic chemical vapor deposition (MOCVD) approach physica status solidi (b) 241 2775-8

[237] Li Q, Lin Y, Creighton J R, Figiel J J and Wang G T 2009 Nanowire-Templated Lateral Epitaxial Growth of Low-Dislocation Density Nonpolara-Plane GaN onr-Plane Sapphire Advanced Materials 21 2416-20

[238] Wang G T, Talin A A, Werder D J, Creighton J R, Lai E, Anderson R J and Arslan I 2006 Highly aligned, template-free growth and characterization of vertical GaN nanowires on sapphire by metal–organic chemical vapour deposition Nanotechnology 17 5773-80

[239] Huang C-T, Song J, Lee W-F, Ding Y, Gao Z, Hao Y, Chen L-J and Wang Z L 2009 GaN Nanowire Arrays for High-Output Nanogenerators Journal of the American

Chemical Society 132 4766-71 [240] Ahl J P, Behmenburg H, Giesen C, Regolin I, Prost W, Tegude F J, Radnoczi G Z, Pécz

B, Kalisch H, Jansen R H and Heuken M 2011 Gold catalyst initiated growth of GaN nanowires by MOCVD physica status solidi (c) 8 2315-7

147

[241] Gottschalch V, Wagner G, Bauer J, Paetzelt H and Shirnow M 2008 VLS growth of GaN nanowires on various substrates Journal of Crystal Growth 310 5123-8

[242] Calarco R, Meijers R, Debnath R K, Stoica T, Sutter E and Luth H 2007 Nucleation and Growth of GaN Nanowires on Si(111) Performed by Molecular Beam Epitaxy NanoLetters 7 2248-51

[243] Jang E-S, Ra Y-H, Lee Y-M, Yun S-H, Kim D-W, Navamathavan R, Kim J-S, Lee I-H and Lee C-R 2009 GaN Nanowires with Au+Ga Solid Solution Grown on an Si(111) Substrate by Metalorganic Chemical Vapor Deposition Japanese Journal of Applied

Physics 48 091001

148

Appendix A: GaN Nanowires - Synthesis and Properties

This chapter discusses the development of GaN nanowires grown via MOCVD and CVD for

light emitting applications. As a starting baseline we consider the growth of GaN nanowires on r-

plane sapphire, demonstrating m-directional nanowire growth with fixed orientations relative to

the substrate and vapor-liquid-solid growth of the nanowires using Au seed-particles. To extend

the growth of nanowires on silicon substrates, we consider the role of the native oxide and the

interaction of the Au seed-particles and the oxidation of the Si substrate surface. The effect of

growth parameters, such as the V/III ratio and the growth temperature, could be used to tune

density, but not nanowire alignment on Si substrates. We further employ strategies to control the

alignment of nanowires relative to the substrate surface, including the use of differently sized Au

seed-particles and the use of ZnO buffer-layer films, finding that the strategy of controlling the

growth direction with the size of the Au seed-particle is effective on sapphire substrates, but not

on Si. Finally, we summarize our findings relative to the fabrication of nanowire-based light

emitting diodes.

The growth of GaN nanowires on silicon substrates was be achieved using MOCVD and

particle-mediated growth. MOCVD growth of GaN nanowires was be conducted at temperatures

between 800°C and 1000°C in a FirstNano EasyTube 3000 CVD reactor (see Figure A-1). Tri-

methyl gallium and ammonia (NH3) were used as the III and V source, respectively. H2 carrier

gas is used, primarily to promote delivery of the metal-organic source to the substrate by forming

GaH, however H2 also inhibits growth[25], thus it is diluted with N2 flow. The system is also

equipped with tri-methyl indium (TMI), which can be used for InxGa1-xN growth, as well as with

SiH4 for silicon doping. Solid sources, such as metallic Ga, were also be used with the solid

source heater and injector. The system baseline pressure was typically on the order of 150

mTorr, however every time the system vented directly to the room ambient, likely allowing for

the introduction of impurities which caused the poor optical qualities and high electrical

conductivity of our GaN nanowires.

149

Figure A-1. (MO)CVD system setup used for the growth of GaN nanowires. Schematic

demonstrating the capabilities of the reactor used to grow GaN nanowires via both CVD and

MOCVD. Figure adapted from S.K. Lim.

A.1. Growth on sapphire

The growth of GaN nanowires on sapphire substrates is relatively well-understood, thus we

use it here as a baseline for comparison to growths on Si substrates. The growth conditions

within the reactor, the size and composition of the seed-particle used and the substrate and its

orientation all play a role in determining the growth direction of nanowires via seed-particle

mediated growth, such as the vapor-liquid-solid mechanism[93]. GaN nanowires grown on c-

plane sapphire substrates have been observed to grow in the a-direction [236-238], the m-

direction [25, 26, 239] and c-direction [97, 240, 241], likely due to different growth conditions

(V/III ratio, temperature, environmental composition, etc.) within the reactor. Based on the

results from Section 3.3. External quantum efficiencies, considering the design of nanowire-

based light emitting diodes, it would likely be advantageous to grow nanowires along the non-

polar m-direction to avoid the quantum-confined Stark effect and promote light extraction due to

the polarization anisotropy observed in InGaN quantum wells. Growth of GaN nanowires on r-

plane sapphire results in well-aligned m-directional nanowires, growing at an angle of ~56.8°

from the plane of the substrate (~33.2° from the substrate normal), as previously seen by work in

our group[26].

150

Figure A-2 shows SEM images of the CVD grown GaN nanowires on r-plane sapphire,

grown at typical growth conditions used in this work. The nanowires demonstrate the two-fold

symmetry in the top-down view (Figure A-1(b)) and alignment angle relative to the substrate

indicative of m-directional nanowire growth (Figure A-1(a)). The nanowires grow via a seed-

particle mediated mechanism using 1 nm thick Au films and the Au seed-particle is observed at

the nanowire tips. Despite using a Au thin film as opposed to colloidal seed-particles the

nanowire diameter distribution is narrow and the nanowires grow straight with no kinks or

substantial tapering. These m-directional nanowires grown on r-plane sapphire serve as a control

for the growth of GaN nanowires directly on Si substrates.

Figure A-2. m-directional GaN nanowires on r-plane sapphire. SEM images of CVD-grown

GaN nanowires (a) at 45°-tilt and (b) top-down indicate the preferred m-directional growth under

typical growth conditions.

GaN nanowire growth on c-plane sapphire typically results in vertical c-directional

nanowires [97, 240]. However, we observe six-fold symmetry from the top-down view of GaN

nanowires grown with 1 nm Au thin films on c-plane sapphire via MOCVD, which suggests that

the nanowires grow in the m-direction and that the c-direction of the nanowires is not normal to

the c-plane sapphire, as previously seen [239]. An SEM image of this six-fold symmetry of GaN

nanowires grown on c-plane sapphire is shown in Figure A-3. The persistence of m-directional

GaN nanowire growth on both c-plane and r-plane sapphire in both CVD and MOCVD suggests

that the growth conditions in the reactor strongly favor m-directional growth.

151

Figure A-3. Six-fold symmetry of GaN nanowires grown on c-plane sapphire. SEM image of

GaN nanowires grown via MOCVD on c-plane sapphire demonstrates six-fold symmetry

indicative of m-directional nanowire growth.

A.2. Growth on silicon

We prepared Si (111) substrates by first stripping any native oxide in 48% HF and ensuring a

hydrophobic surface indicative of bare-Si prior to depositing the 1 nm Au film used for the seed-

particles via electron beam evaporation. As a control, 1 nm Au was also deposited onto native Si

with no HF pre-treatment. The bare Si and native Si samples with Au were then used to attempt

to grow GaN nanowires via MOCVD with similar parameters used for the growth on sapphire

and the resulting nanowires are demonstrated in the SEM images of Figure A-4(a) and (c),

respectively. It was observed that no substantial growth was achieved on the bare nanowire

substrate, likely due to the formation of an oxide over-layer during the Au evaporation, where a

SiOx layer grows enough to engulf the Au particles[162] and prevent them from absorbing the

Ga source. To eliminate the effect of the oxide over-layer an addition HF etching step was

performed immediately prior to growth with no rinse to ensure a hydrogen terminated Si surface

to prevent re-oxidation. This resulted in dense nanowire growth, similar to growth on the control,

shown in the SEM image in Figure A-4(b). We note these nanowires are m-directional (see the

diffraction from transmission electron microscopy, TEM, in Figure A-5) and show no relative

alignment to the (111) plane Si. For device fabrication, it is necessary to have nanowire

alignment for high yield of the top contact, thus we next varied growth parameters to attempt to

achieve nanowire alignment.

152

Figure A-4. Growth of GaN on Si substrates with different pre-growth treatment. SEM

images of GaN nanowires grown on Si (111) with various substrate pre-processing: (a) Si

substrate stripped of native oxide prior to Au seed-particle film deposition, demonstrating the

effect of the oxide overlayer which forms during or after this deposition, (b) Si substrate

processed as in (a), but also HF strip of the oxide overlayer immediately prior to nanowire

growth, and (c) control sample with Au seed-particle deposited on Si with native oxide.

Figure A-5. TEM of m-directional GaN nanowire grown on bare Si. TEM of GaN nanowires

grown on Si (111): (a) Au nanoparticle at nanowire tips indicating that growth proceeded by the

particle-mediated mechanism, (b) nanowire used for HRTEM shown in (c), with the diffraction

pattern inset indicating growth in the m-direction, and (c) HRTEM of GaN nanowire, indicating

single crystalline nature of the nanowires.

In an attempt to gain control over the nanowire alignment relative to the Si substrate, a

variety of growth parameters were explored. Alignment of GaN relative to Si substrates has only

153

been achieved for c-directional nanowire growth, which grow similar to VTC growth of ZnO

nanowires, without a seed-particle[242, 243]. The V/III ratio of the source precursor is a

common parameter which can control the morphology of nanowires. The effect of the V/III ratio

on nanowire alignment is shown to be minimal, as over a wide range of V/III ratios from about

58 – 610, the nanowire alignment for GaN nanowires grown on Si was still absent as shown in

Figure A-6. Changes in the growth temperature up to 950°C and growth at higher pressure also

had no significant impact on the alignment of nanowires relative to the substrate. Though other

methods of alignment could have been attempted, such as seed-particle size and composition,

improving the optical properties of the GaN nanowires became the priority, due to their low

luminescence efficiency (see Section A.4. Optical properties of GaN nanowires).

Figure A-6. GaN nanowires grown on bare Si at different V/III ratios. SEM images show no

change in nanowire alignment relative to the substrate with changing V/III ratio from (a) 58, (b)

116 and up to (c) 610.

A.3. Electrical characterization of GaN nanowires

Using the individual nanowire characterization discussed in Section 6.1. Individual nanowire

electrical characterization, GaN nanowires grown on native Si via CVD and via MOCVD on r-

sapphire and a-GaN were compared. In general the as-grown nanowires were conductive with an

average carrier concentration on the order of 1018 cm-3 for all three samples. Transfer

characteristics indicated that the as-grown GaN is n-type with mobility ranging from 47 cm2/V·s

to 170 cm2/V·s. Examples of these measurements are shown in Figure A-7. The high level of

carriers for as-grown nanowires indicates these nanowires can be used for heterojunctions with a

p-type substrate, however, it will also make p-type doping difficult to achieve.

154

Figure A-7. Individual nanowire electrical characterization of GaN nanowires. Low contact

resistances are demonstrated by the comparison of the 2-point and 4-point measurements in (a)

and (b) the transfer characteristics indicate n-type behavior.

A.4. Optical properties of GaN nanowires

The optical properties of the GaN nanowires grown in this work were poor relative to other

materials, such as ZnO nanowires grown in this work, as evidenced by the 120 s collection time

necessary to collect a PL spectrum. Section 5.1. Optical properties of ZnO nanowires

demonstrates the approximately six order of magnitude difference between the luminescence

efficiency of ZnO and GaN nanowires grown on Si. We believe the poor optical quality of the

GaN nanowires arises from the presence of oxygen in the growth system. The source of oxygen

contamination was likely due to adsorbed water vapor in the quartz-ware, which is difficult to

eliminate since the system vents to the room ambient. It has been observed that oxygen can

greatly increase the conductivity of as-grown GaN[73] and thus may be responsible for the high

levels of conductivity observed in the individual nanowire electrical characterization.

Furthermore, the defect luminescence observed in Figure 5-1(b) may be indicative of intrinsic

defects which are more easily formed in the presence of oxygen. This theory is supported by the

use of metal Zr foils introduced into the reactor, meant to act as an oxygen-getter. The Zr

oxygen-getter should act to reduce the oxygen present in the vicinity of the substrates during

GaN nanowire growth. The use of the Zr oxygen-getters did results in an increase in

luminescence and decrease in the relative defect luminescence; however, the increase was only

about two-fold, indicating that more than an oxygen getter is necessary to improve the optical

155

properties of these GaN nanowires. Since an upgrade to the system to eliminate sources of

oxygen impurities was impractical in the time-frame of this work, the work was refocused on the

development of ZnO nanowires towards nanowire-based LEDs.