Defect and Metal Oxide Control of Schottky Barriers And

DEFECT AND METAL OXIDE CONTROL OF SCHOTTKY BARRIERS AND

CHARGE TRANSPORT AT ZINC OXIDE INTERFACES

DISSERTATION

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

Geoffrey Michael Reimbold Foster

Graduate Program in Physics

The Ohio State University

2018

Dissertation Committee:

Professor Leonard J. Brillson Adviser

Professor Thomas Lemberger

Professor Ilya Gruzberg

Professor Robert Perry

Copyrighted by

Geoffrey Michael Reimbold Foster

2018

ABSTRACT

In recent years ZnO has received renewed interest due to its exciting semiconductor properties and remarkable ability to grow nanostructures. ZnO is a wide band gap semiconductor, allowing many potential future applications including electronic nanoscale devices, biosensors, blue/UV light emitters, and transparent conductors. There are many potential challenges that keep ZnO from reaching a full device potential. The biggest challenge is what the role native point defects play in the fabrication of high quality Ohmic and Schottky contacts. The following work examines the impact of these native point defects in the formation of Schottky barriers and charge transport at ZnO interfaces.

We have used depth-resolved cathodoluminescence spectroscopy and nanoscale surface photovoltage spectroscopy to measure the dependence of native point defect energies and densities on Mg content, band gap, and lattice structure in non-polar, single- phase MgxZn1-xO (0

~52%. This minimum also corresponds to a pronounced change in Schottky barriers reported previously.

DRCLS allows us to probe buried interfaces. Due to a strong Fermi-level mismatch, about 10% of the electrons in a 5-nm-thick highly-Ga-doped ZnO (GZO) layer grown by molecular beam epitaxy at 250 C on an undoped ZnO buffer layer transfer to the ZnO (Debye leakage), causing the measured Hall-effect mobility (H) of the

GZO/ZnO combination to remarkably increase from 34 cm2/V-s, in thick GZO, to 64

2 cm /V-s. From previous characterization of the GZO, it is known that ND = [Ga] = 1.04 x

21 20 -3 10 , and NA = [VZn] = 1.03 x 10 cm , where ND, NA, and [VZn] are the donor,

19 acceptor, and Zn-vacancy concentrations, respectively. In the ZnO, ND = 3.04 x 10 , and

18 -3 2 NA = 8.10 x 10 cm . Assuming the interface is abrupt, theory predicts H = 61 cm /V-s, with no adjustable parameters. The assumption of abruptness in [Ga] and [VZn] profiles is confirmed directly with a differential form of depth-resolved cathodoluminescence spectroscopy coupled with X-ray photoelectron spectroscopy. An anneal in Ar at 500 C for 10 min somewhat broadens the profiles but causes no appreciable degradation in H and other electrical properties.

. Using our ability to probe abrupt buried interfaces, we probed the IrOx/ZnO interface. IrOx and other metal oxides exhibit higher Schottky barriers than their pure metal counterparts, consistent with wider depletion regions and potentially useful for ohmic contacts to p-type semiconductors. DRCLS with I-V and 1/C2-V barrier height and carrier profile measurements showed high zinc vacancy VZn and CuZn defect densities that compensate free carrier densities, increase depletion widths, and form higher effective barriers than Ir/ZnO contacts. Zn-polar versus O-polar ZnO interfaces with IrOx exhibit 40% higher VZn + CuZn interface segregation and lower carrier densities within a iii wider depletion region, accounting for the significantly higher (0.89 vs. 0.67 eV) barrier heights. The depth of VZn density segregation and the Zn-deficient layer thickness measured microscopically both match the depletion width, and applied electric fields comparable to spontaneous polarization fields across similar layers display analogous defect segregation. These results account for the difference in polarity-dependent segregation due to the electric field-driven diffusion of native defects near ZnO interfaces.

By establishing the role that defects play in the effective barrier height we attempted to form Ohmic and Schottky contacts to ZnO nanowires. Ohmic and rectifying metal contacts to semiconductor nanowires are integral to electronic device structures and typically require different metals and process techniques to form. A Pt ion beam alone can form Ohmic, Schottky, or blocking contacts to ZnO nanowires with the same metal on the same wire by controlling native point defects at the intimate metal-semiconductor interface. Spatially- resolved cathodoluminescence spectroscopy both laterally and in depth gauges the nature, density, and spatial distribution of specific native point defects inside the nanowires and at their metal interfaces. Combinations of electron and ion beam deposition, annealing, and outer diameter milling of the same pulsed laser deposited nanowire provide either low contact resistivity (2.6 x 10-3 Ω-cm-2) ohmic, Schottky (Φ >

0.35 eV) or blocking junctions with single Pt deposition, depending on the physical nature and spatial distribution of substitutional Cu on Zn sites, zinc vacancy, and oxygen vacancy defects. These results demonstrate the importance of point defects on electrical

iv properties of metal contacts to ZnO nanowires and present methods to tailor contact electronic properties of nanowires in general.

ACKNOWLEDGMENTS

This work would not be possible if not for the support and mentorship of Dr

Leonard J. Brillson. I would like to thank him for his patience advice and guidance throughout my time as a graduate student.

I would also like to thank, and in no particular order, Thaddeus Asel, Jon Cox,

Brent Noesges, Hantian Gao, and all of the Columbus School for Girls interns who all assisted this work in one way or another. Through our success and failures we have advanced the study of ZnO. Without all of you the quality and speed of this work would have been greatly diminished.

Lastly I would like to thank my family for their continual love and support. My wife Melissa Foster, who put up with the strange hours graduate school sometimes requires you while continually being my biggest cheerleader. My mom, Tracy Reimbold, who in primary school always pushed me to greater academic heights and never let me settle with 'just A's'. And my father, Tom Foster, who constantly encouraged me to take chances and just 'go for it' that it is ok to fail as long as you try.

VITA

2009...... Streetsboro High School

2013...... B.S. Physics, University of Akron

2015...... M.S. Physics, The Ohio State University

2015 to present ...... Graduate Research Associate, Department

of Physics, The Ohio State University

PUBLICATIONS

Foster, Geoffrey M., Hantian Gao, Grace Mackessy, Alana M. Hyland, Martin W. Allen,

Buguo Wang, David C. Look, and Leonard J. Brillson. 2017. “Impact of Defect

Distribution on IrOx/ZnO Interface Doping and Schottky Barriers.” Applied Physics

Letters 111 (10). AIP Publishing: 101604. doi:10.1063/1.4989539.

Foster, G. M., Faber, G., Yao, Y. F., Yang, C. C., Heller, E. R., Look, D. C., & Brillson,

L. J. (2016). Direct measurement of defect and dopant abruptness at high electron mobility ZnO homojunctions. Applied Physics Letters, 109(14). http://doi.org/10.1063/1.4963888.

Dordevic, S V, G M Foster, M S Wolf, N Stojilovic, H Lei, C Petrovic, Z Chen, Z Q Li, and L C Tung. (2016). “Fano Q-Reversal in Topological Insulator Bi2Se3.” Journal of

vii

Physics Condensed Matter 28 (16). https://doi.org/10.1088/0953-8984/28/16/165602.

Foster, G. M., Perkins, J., Myer, M., Mehra, S., Chauveau, J. M., Hierro, a., & Brillson,

L. J. (2015). Native point defect energies, densities, and electrostatic repulsion across

(Mg,Zn)O alloys. Physica Status Solidi (a), 212(7), 1448–1454. http://doi.org/10.1002/pssa.201532285.

Perkins, J., G. M. Foster, M. Myer, S. Mehra, J. M. Chauveau, A. Hierro, A. Redondo-

Cubero, W. Windl, and L. J. Brillson. (2015). “Impact of Mg Content on Native Point

Defects in MgxZn1-xO (0 ≤ X ≤ 0.56).” APL Materials 3 (6):62801. https://doi.org/10.1063/1.4915491.

Dordevic, S. V., G. M. Foster, N. Stojilovic, E. A. Evans, Z. G. Chen, Z. Q. Li, M. V.

Nikolic, Z. Z. Djuric, S. S. Vujatovic, and P. M. Nikolic. (2014). “Magneto-Optical

Effects in 1-xAsx with x=0.01: Comparison with Topo- Logical Insulator Bi1-xSbx with x=0.20.” Physica Status Solidi (B) Basic Research 251 (8):1510–14. https://doi.org/10.1002/pssb.201451091.

FIELDS OF STUDY

Major Field: Physics

viii

TABLE OF CONTENTS

Abstract ...... ii

Acknowledgments...... vi

Vita ...... vii

Publications ...... vii

Table of Contents ...... ix

List of Figures ...... xii

Chapter 1 Introduction ...... 1

1.1 Motivation ...... 1

1.2 Defects in ZnO ...... 2

1.3 ZnO Nanowires ...... 7

Chapter 2 Experimental Methods ...... 10

2.1 Ultra High Vacuum (UHV) ...... 10

2.2 Cathodoluminescence Spectroscopy (CLS) ...... 11

2.3 Atomic Force Microscopy ...... 17

2.4 Kelvin Probe Force Microscopy ...... 19

2.5 Surface Photovoltage Spectroscopy ...... 21

2.6 X-ray Photoemission Spectroscopy ...... 25

2.7 Focused Ion Beam\ Electron Beam Induced Deposition ...... 30 ix

Chapter 3 Ohmic and Schottky Contacts ...... 35

3.1 Introduction ...... 35

3.2 Schottky Theory ...... 36

3.3 Schottky Barrier Measurement Techniques ...... 41

Chapter 4 Native point defect energies, densities, and electrostatic repulsion across

(Mg,Zn)O alloys ...... 44

4.1 Introduction ...... 44

4.2 Experiment ...... 45

4.3 Results ...... 47

4.4 Analysis and Discussion ...... 55

4.5 Conclusions ...... 59

Chapter 5 Direct Measurement of Defect and Dopant Abruptness at High Electron

Mobility ZnO Homojunctions...... 60

5.1 Introduction ...... 60

5.2 Experimental Setup ...... 62

5.3 Results ...... 63

5.4 Conclusions ...... 74

Chapter 6 Impact of Defect Distribution on IrOx/ZnO Interface Doping and Schottky

Barriers ...... 75

6.1 Introduction ...... 75

6.2 Experimental Setup ...... 78

6.3 Results ...... 80

6.4 Conclusions ...... 87

Chapter 7 Defect-Controlled Ohmic, Blocking, and Schottky Contacts to ZnO Micro-

/Nanowires ...... 88

7.1 Introduction ...... 88

7.2 Experimental Setup & Results ...... 89

7.3 Conclusions ...... 101

Chapter 8 Future Work ...... 102

8.1 Conclusions ...... 102

8.2 Future Work ...... 104

References ...... 105

LIST OF FIGURES

Figure 1.1. Above the formation energies for native point defects are shown as a function of Fermi level in ZnO. The valance-band maximum is the zero of the Fermi level.

Changes in slope refer to changes in charge state.7 ...... 4

Figure 1.2. DRCLS defect emission intensities and PAS zinc vacancy densities versus depth. Li-implanted ZnO (a) flash annealed at 1200 °C and (b) furnace annealed at 800

°C for 1-hr. Blue circles are the PAS profile...... 6

Figure 1.3. Energies of defect levels calculated for ZnO...... 7

Figure 1.4. Hexagonal wurzite structure of a single ZnO nanowire on the (0001) plane. .. 8

Figure 2.1. Illustration of the various processes generated by a secondary electron cascade in a solid...... 12

Figure 2.2. Excitation of electron hole pairs by an incident high energy electron beam and the possible routes of recombination...... 14

Figure 2.3. Two-dimensional CASINO simulations for bare ZnO, with the Bohr-Bethe range RB of each probe energy EB noted...... 16

Figure 2.4. Atomic force microscopy (AFM) mode operation for Park System XE-70

Scanning Probe Microscope...... 18

Figure 2.5. Kelvin probe force microscopy (KPFM) mode operation for Park System XE-

70 Scanning Probe Microscope...... 20

Figure 2.6. Experimental setup of SPS measurements. In addition to the standard KPFM setup, an outside source of light is applied to the surface which may excite electrons into

xii a different band changing the potential on the surface. KPFM measures this change in potential and can tell us about surface features and defects...... 22

Figure 2.7. The left shows a photopopulation state of an electron moving into the conduction band from a defect state. The right shows the photo depopulation state as the electron moves from the surface down into the bulk material. KPFM measures the change in the Fermi level in these transitions...... 23

Figure 2.8. The above shows a typical SPS scan with transitions being present at energies

E2 and E1...... 24

28 Figure 2.9. Schematic view of photo ionized electron E1...... 27

Figure 2.10. Electron scattering lengths in a solid. From this curve, electrons only from the first few nanometers of the surface escape elastically.29 ...... 29

Figure 2.11. A schematic diagram showing (a)imaging, (b) milling, and (c) deposition. 31

Figure 2.12. Six-contact structure to a 140 nm diameter ZnO nanowire. Contacts are 300 nm thick Pt patterned by electron beam induced deposition (EBID)...... 33

Figure 2.14. Three-dimensional ion implant distribution for a 30 keV Ga ion beam incident on ZnO...... 34

Figure 2.13. Stopping and Range of Ions in Matter (SRIM) simulations for (left) 5 keV and (right) 30 keV accelerated Ga ions in ZnO...... 34

Figure 3.1. The above shows a simple model for barrier formation between a metal and an n-type semiconductor where ΦM > ΦSC (a) before electrical contact and (b) following alignment of Fermi levels...... 38

xiii

Figure 3.2. Barrier formation between an n-type semiconductor and a metal. The semiconductor has surface state induced band bending (a) before contact. Following contact, an interface dipole occurs from the trapped EF...... 40

Figure 3.3. Forward current density J versus applied voltage (V) for a metal/semiconductor contact. This can be used to determine both barrier height and ideality factor ...... 42

Figure 4.1. Representative DRCLS spectra for each of the five MBE-grown (Mg,Zn)O alloys...... 48

Figure 4.2. Depth profile of I(Defect)/I(NBE) for a-plane MgxZn1-xO films versus EB from 1 to 5 keV and depth of peak e-h pair creation rate UO from the free surface to 72 nm for x = 0, 0.31, 0.44, 0.52, and 0.56. Both segregated and bulk defect densities decrease to minimum values at 44% Mg...... 50

Figure 4.3. MgxZn1-xO normalized intensities of VC and VO – related defects versus Mg alloy content. Error bars for all but x = 0 are within the symbols plotted...... 52

Figure 4.4. MgxZn1-xO band gaps measured versus Mg content x by CL, SPS, and SSPC.

All three techniques show a nearly linear increase up to [Mg] = 52%...... 53

Figure 4.5. SPS spectra of MgxZn1-xO x = 0.31, 0.44, and 0.56 showing onsets of gap state photo-population and depopulation transitions and band gap response. Each panel also shows the corresponding CL data for each alloy composition...... 56

Figure 4.6. Level positions of VC and VO within the MgxZn1-xO (0 < x < 0.56 ) band gap vs. Mg/ (Mg + Zn). The midgap VO and VC levels appear to vary with EV and EC, respectively...... 56

xiv

Figure 4.7. Dependence of MgxZn1-xO lattice parameters a and c versus x for 0 < x <

0.56. Both a and c reach minimum values for x ~ 0.52...... 58

Figure 5.1. Normalized room temperature DDRCL spectra of (a) 25nm and (b) 5 nm

GZO on ZnO. Above bandgap emission in 25nm GZO due to degenerate doping is absent inside ZnO ...... 64

Figure 5.2 Comparison of GZO/ZnO VZn vacancy peak areas versus DRCLS probe depth for GZO thicknesses 0, 5, 25, and 50 nm...... 66

Figure 5.3. Room temperature DRCL spectra of 25 nm GZO on ZnO after 10 min. 500

°C anneal in Ar. Above bandgap emission in GZO decreases in GZO but extends into

ZnO...... 70

Figure 5.4. Integrated NBE peak areas vs. depth of (a) GZO and (b) ZnO before and after annealing. The GZO area profile normalized to the bulk GZO level broadens after anneal.

...... 71

Figure 5.5.( a) XPS survey and Ga spectra of GZO. (b) Measured Ga content versus sputter depth profiling of 25 nm GZO/ZnO pre- vs. post-annealed...... 73

Figure 6.1. (a) DRCL spectra, normalized to the NBE, for 32 nm IrOx on ZnO (0001) at

80 K, showing strong VZn segregation toward the IrOx/ZnO (0001) interface. Note the minimum in the defect emission at beam energy 3.5 keV. (b) EB = 1.0 keV, 80 K DRCL spectrum, normalized to the NBE, of bare Zn (0001) surface. Deconvolutions in both (a) and (b) show both VZn and CuZn defects...... 81

Figure 6.2. Deconvolved VZn (2.01-2.05 eV) and CuZn (2.35 eV) area depth profiles showing higher VZn and CuZn surface segregation toward the Zn- polar (0001) vs. O-polar

(000ī) surface...... 82

Figure 6.3. (a) Barrier height vs ideality factor plots of O- and Zn-polar ZnO/IrOx diodes with slopes and barrier height intercepts shown. (b) Carrier density depth profiles extracted from slopes in 1/C2 –V and relative to their calculated depletion depths W.

Carrier density within the O-polar depletion width increases strongly toward the

IrOx/ZnO interface at shallower depths than for Zn-polar...... 84

Figure 6.4. DRCLS peak intensity ratio of 2.01-2.05 eV VZn and 2.35 eV CuZn peak in

ZnO versus NBE intensity with and without 917 V positive bias applied across a 0.5 mm

ZnO crystal. The electron beam penetrates the 20 nm Pt layer to generate luminescence within the ZnO below. The 18.3 kV/cm bias attracts negatively-charged CuZn and VZn acceptor defects toward the top electrode...... 86

Figure 7.1. (a) Tapered ZnO nanowire with five Pt contact and wire pads. (b) I-V characteristics under in air and in dark between Contacts 1 and 3 versus (c) 3 and 4.

Ohmic behavior between Contacts 1 and 3 versus Schottky behavior between Contacts 3 and 4 shows that Contacts 1, 3, and 4 were ohmic, ohmic, and rectifying, respectively.

Contact 5 was blocking...... 91

Figure 7.2. (a’) 80 K DRCLS deep level defect emissions below ZnO band gap versus EB midway between contacts 3 and 4. (b’) Deconvolved defect integrated areas normalized to NBE area versus excitation depth. (c’) 3.5 keV DRCL spectra normalized to NBE intensity for 500, 700, and 1,000 nm diameter ZnO wires. (d’) I(2.35 eV) CuZn defect

xvi intensities versus nanowire diameter for surface (1.5 keV), sub-surface (3.5 keV), and

“bulk” (5 keV) probe depths showing increasing near-surface segregation with increasing

+ wire diameter. (e’) (a) SEI, (b) HSI, and (c) EB = 5keV CL spectra of 5 keV Ga milled removal of segregated defects in outer annulus versus e-beam annealed defect increase in contiguous regions of a 200 nm ZnO wire on SiO2. HSI color gradient signifies magnitude of integrated 1.5 – 2.75 eV defect areas normalized by NBE integrated area.

Defects increase from milled wire section toward annealed region...... 93

Figure 7.3. The above show the DRCLS spectra and defect profile for the base ZnO with no Ga processing between contacts 1 & 2 and between contacts 2 & 3...... 96

Figure 7.4. DRCL spectra at EB = 5.0 keV immediately adjacent to (a) Contact 3 and (b)

Contact 4 with the corresponding Gaussian deconvolutions. The 2.35 eV CuZn defect decreases significantly at Contact 4. Depth profiles of deconvolved defect areas for (c)

Contact 3 and (d) Contact 4, both showing pronounced minima at ca. 50 nm depth and surface segregation, nearly 4x stronger for Contact 3...... 98

Figure 7.5. Schematic diagrams of band bending at Pt-ZnO nano/microwire contact for

(a) 900 nm, (b) 600 nm, and (c) 400 nm diameter wires linked to the interfaces of their corresponding wires. Lighter shading signifies higher acceptor density and lower electron density with increasing radius. With decreasing diameter, interface acceptor density decreases and contact behavior changes from transport by (a) trap-assisted tunneling to

(b) Schottky rectification, to (c) blocking...... 100

xvii

Chapter 1 Introduction

1.1 Motivation

The study of ZnO has been active for the past 60 years1, and only within recent the last decade has its exceptional semiconductor attributes have emerged. 2,3 Containing a broad direct band gap (3.34 eV) and an exciton binding energy of 60 meV, ZnO is latticed matched to GaN.

ZnO based optoelectronics (blue and ultraviolet) have become a focus of research. The high optical transparency of ZnO, lends itself to be a good choice for transparent thin film transistors

(TFTs). ZnO includes many valuable qualities, including but not limited to, its environmental friendliness, low cost, abundance, and safety to humans. New developments of crystal growth have allowed for better control over the quality and doping of ZnO crystals, potentially leading into additional next generation opto- and microelectronic applications. These include but are not limited to gas sensors, buffer layers for solar cell applications, spintronic devices, and high electron mobility transistors (HEMTs). Nanostructure growth is also possible using ZnO, in the form of many different geometries. Founding a base where new electronic applications may start.4 With advancements in nanostructure growth and processing5, new biomedical sensors which can use ZnO’s piezoelectric ability and high sensitivity to adsorbate charge transfer.6

For ZnO to become a realistic candidate for next generation devices there are still many hurdles that must be overcome. Mainly being the role of native point defects and impurities and how they affect electrical properties in ZnO. The chemistry and physics at the metal/ZnO interface has not been yet adequately established. This understanding in critical for making high quality Ohmic or rectifying (Schottky) contacts.

This work will focus on the role of native point defects and doping in controlling both band gaps and surface states and contact formation. Then later expanded into controlled contact deposition to ZnO nanowires and attempts to control the formation of Schottkey, Ohmic, and blocking contacts to ZnO nanowires.

1.2 Defects in ZnO

To create cutting edge devices, we should comprehend the role of native point defects in

ZnO and their impact in controlling the doping and electrical properties. Similarly, as with other semiconductors, defects can have pronounced effects on the electrical properties, optical properties, and in addition interface reactions and doping. Native point defects are defects that include the constituent atoms Zn and O as it were. In this way, native point defects are just those including Zn as well as O atoms and incorporate opportunities (missing Zn or O atoms), interstitials (Zn or O atoms possessing space between lattice sites), antisites (Zn atoms involving

O lattice sites or the other way around), or their complexes.

Recent research has been made to the formation of native point defects using first- principles calculations using density functional theory (DFT) 7 within the local density approximation (LDA) as well as the LDA+U approach for overcoming the “band-gap problem” with DFT methods.8 Differing DFT calculations can vary a lot using different approximations yielding differing basic heat of formation energies and defect energy-level relative to their band edges. Figure 1.1 shows, as a function of Fermi level (EF), the estimated formation energy position for each of the native point defects in ZnO using the LDA+U formulism. A high formation energy suggests that a defect is unlikely to form whereas a low formation energy

2 suggests a high equilibrium concentration of the defect. Following from this work, a conclusion can be drawn that the most energetically favorable native point defects in ZnO are the oxygen vacancy (VO) and the zinc vacancy (VZn). VZn is an acceptor defect, capable of accepting up to

-2 two electrons (VZn ) while the VO is a donor type defect able to contribute up to two electrons

+2 (VO ). Compensation from native point defects is usually against the main acceptor or donor dopants. Defects that form in p-type materials easier are donors and defects that form in n-type materials act as acceptors. Zn- vs. O-rich growth conditions can affect the formation energies of these point defects. This is shown in Figure 1.1 where, under both O- and Zn-rich conditions, the

EF approaches the valence-band-maximum (VBM) the formation energy of VZn increases.

Conversely, as the VZn formation energy is lowest as EF approaches the conduction-band- minimum (CBM). Under Zn-rich conditions, the VZn is less likely to form given most EF positions. Similarly, under O-rich conditions the VO is less likely to form for all EF except those very close to VBM.

Figure 1.1. Above the formation energies for native point defects are shown as a function of

Fermi level in ZnO. The valance-band maximum is the zero of the Fermi level. Changes in slope refer to changes in charge state.7

One of the main known defect emissions is the "green band" emission in ZnO. This presents itself at around 2.5 eV and is typical a dominant feature. This was originally said to be from Cu impurities,9 with some recent research supporting this classification.10 Other recent work shows a strong correlation between concentrations of VO and this GB emission. These were

4 obtained on various annealing conditions or oxidation studies and have correlated the 2.5 eV

11,12 feature to VO. Different techniques, such as optical absorption, depth resolved cathodoluminescence, photoluminescence (PL), and electron paramagnetic resonance (EPR)

13,14 indicate a strong relationship between the GB feature and VO defects as well . This VO is a mid-gap state at ~2.5 eV above the VBM. It behaves as a deep donor. The issue of whether the

GB emission is caused by VO or some impurity still has not reached a complete consensus in the literature.

There is one more defect emission seen in ZnO. This is a "red" emission in the range of

1.8-2.1 eV. This has been shown to have correlation with isolated and clustered VZn defects in

15 Li-doped ZnO. At the low end of the above energy range, isolated VZn were shown. While between 1.9-2.1 eV clustered VZn were shown. Positron Annihilation Spectroscopy (PAS) and

DRCLS were used to confirm these results. Figure 1.2 shows that the VO and VZn are not related to each other. The PAS profile matches with the VZn profile while the VO (or GB emission) remains independent of the VZn profile. VZn behaves as an acceptor. It has the lowest formation energy of ZnO's native point defects and can accept no more than two electrons.16

Figure 1.2. DRCLS defect emission intensities and PAS zinc vacancy densities versus depth. Li- implanted ZnO (a) flash annealed at 1200 °C and (b) furnace annealed at 800 °C for 1-hr. Blue circles are the PAS profile.

Figure 1.3 shows accepted energy levels within the band gap of most native point defects in ZnO. These Zni act as shallow donors in ZnO. Zni have a large formation energy in n-type

ZnO and can diffuse quickly with a low migration barrier of ~0.57 eV. Zni are unstable at 300K and are easily annealed away at temperatures as low as 170 K.

Figure 1.3. Energies of defect levels calculated for ZnO.

To date there has been no reliable p-type doping of ZnO. The VO native point defect and their complexes all behave as donors. VZn and its complexes behave as acceptors. The impact that these native point defects has on carrier concentrations is recognized. The physical nature of how these defects, acting as donors and acceptors behave in ZnO, has been left unresolved. The challenge remains to correlate features in ZnO emissions to native point defects.

1.3 ZnO Nanowires

ZnO can be grown into various types of ZnO nanostructures such as nanobelts, nanobows, and nanowires, and have all attracted significant attention due to their ease of fabrication, remarkable relative surface area, and low-dimensional nature17,18 . Nanowires of

ZnO can exhibit pinch-off of electrical current with surface charge-sensitive depletion depths that are on the order of the wire radius19,20. Single nanowires have been studied for use in nano- photodetectors, with photoconduction gains as high as 1010 21. The crystal structure of ZnO is wurzite, and the hexagonal crystal facets of a typical ZnO nanowire can be seen clearly via scanning electron microscopy (SEM) in Figure 1.4 below. 7

Figure 1.4. Hexagonal wurzite structure of a single ZnO nanowire on the (0001) plane.

Defects in bulk ZnO have been shown to strongly affect the behavior of metal contacts by modifying interfacial band bending, pinning the Fermi level, and allowing “hopping” transport via trap states through the metal-ZnO Schottky barrier20. By modifying the spatial distribution of defects at the surface and near-surface, the contact behavior is also modified, allowing for a method to control the behavior of a contact without changing the metallization scheme. The outer annulus of the nanowire is known to have a higher relative density of defects 22,23, and can be removed to reveal a more pristine crystalline core for fabrication of contacts with lower interfacial defect density. On the other hand, the surface and near-surface can be degenerately doped by implanting with gallium in order to increase carrier concentration and promote tunneling and hopping transport, thereby reducing the contact resistance 24.

Chapter 2 Experimental Methods

2.1 Ultra High Vacuum (UHV)

To properly study surfaces and interfaces, clean surfaces must be maintained. Through a series of pumps, ultra-high vacuum can be achieved. This is a pressure at or below 10-10 Torr.

Thus, no molecules can absorb onto the surface of our materials in any considerable numbers.

The equipment used in this process includes stainless steel chambers, vacuum pumps to remove contaminants, pressure gauges, quartz view ports, and manipulators to move samples from one section to another. Lastly researchers will want to have analysis tools included in which to take measurements with. The chambers are then joined and sealed together using bolts and copper gaskets. These gaskets will deform to “knife edges” that machined in to a flange. This will minimize gas leakage. Different pumps will operate in isolate portion of a chamber, separated by valves, to act as a stage to achieve UHV conditions. The different pumps used for these purposes include roughing pumps (max 10-3 Torr), turbo molecular pumps (max 10-9 Torr), ion pumps

(max 10-10 Torr), and cryopumps (max 10-10 Torr).

The initial stage one must pump on is done by roughing pumps. This initial chamber evacuation is necessary to reach a pressure where a high vacuum pump could be used to reach

UHV pressures. There are two main variants of roughing pumps known as rotary vane and scroll pumps. Advantages and disadvantages are found in both. Rotary pumps may allow a backflow of lubricating oil into the vaccum chamber, but are considerably cheaper than their scroll pump counter parts. Turbo pumps behave like turbine engines, in that air is pushed though vanes that can rotate between 15-75 kRPM. Cyropumps produce vacuum by condensing molecules and vapor onto a cooled surface. Essentially this is the adsorption of gases onto a condenser coil

10 connected to a source of cryogenic gas. Ion pumps ionize gas within the chambers and uses a strong electric potential, typically 5 kV, which allows the ions to accelerate into and be captured by Ti cathode plates. The ionized gas molecules either bond with the Ti layer and are buried or they knock off additional Ti which can bond with other gas ions.

2.2 Cathodoluminescence Spectroscopy (CLS)

Cathodoluminescence spectroscopy (CLS) involves an electron incident on a semiconductor or insulator that creates a cascade of secondary electrons and ultimately electron- hole pairs that emit a photon when they recombine. This has several advantages over photoluminescence spectroscopy (PL), such as an order of magnitude higher free carrier generation rates and the ability to probe wide bandgap semiconductors and insulators. Although, perhaps the greatest advantage to using an electron beam, is that by controlling the energy of the incident electron beam, the depth that one is probing can be controlled. This allows one to have a spatial resolution on the order of nanometers, allowing for the separate probing of bulk effects, surface effects, etc.

A wide swath of processes occurs when an electron beam irradiates a solid. Figure 2.1 shows these different processes that can occur when an electron beam strikes a solid surface. The irradiation from the electron beam will produce a pear-shaped area, where the following will be generated: (i) Auger electron generation within the first few Å of the free surface, (ii) secondary electrons due to ionization of impacted atoms ranging from the first few nanometers to higher depths depending on the incident electron beam energy, (iii) backscattered electrons at lower depths due to random collisions of electrons that have lost significant kinetic energy, (iv) X-rays

11 characteristic of specific atomic transitions, (v) a continuum of X-ray energies resulting from secondary X-ray excitation, and (vi) optical emission (fluorescent X-rays) due to low energy electrons initially excited by X-rays. The size of this generation volume is determined by several factors. These are the size of the electron beam size, the generation volume size, and the diffusion length of the minority carriers. This is also the determining factor of CLS resolution.

The diffusion lengths of the of the minority carriers are typically very small, e.g., on scal of tens of nanometers or less, for semiconductors, on the nanoscale, making CLS are a great characterization tool.

Figure 2.1. Illustration of the various processes generated by a secondary

electron cascade in a solid.

The cascade of generated electrons makes CLS a powerful technique for characterizing the radiative defect states in semiconductor materials. The large free carrier generation in CLS will create a cascade of electrons that are free to recombine with holes through one or more available channels. Figure 2.2 shows schematically some of the various channels that exist.

These channels are: intra-band transitions due to electrons excited to well above the conduction band edge falling as thermal equilibrium is reached typically resulting in the emission of a phonon, band-to-band transitions, excitonic transitions including free, donor bound, and acceptor bound excitons, transitions involving defect states, and transitions due to radiative de-excitation of centers. It is possible to have some recombination of electron-hole pairs that occur through nonradiative processes, but these are difficult to characterize.25

Figure 2.2. Excitation of electron hole pairs by an incident high energy electron beam and the possible routes of recombination.

Measurements are usually performed with current-voltage product, i.e., power adjusted

for constant power as incident beam voltage varies. When the electron beam energy is changed

the current is also change. With a new accelerating voltage, one can change the depth that the

sample is probed. This depth can be modeled through Monte Carlo simulations. The simulation

is essentially having an electron take a random walk through the material while allowing for

random collisions where the electron gives up energy to create electron hole pairs.26 Simulating

for hundreds of thousands of electrons, the simulation can map the electron cascades to the depth

where they create and excite the electron-hole pairs and the volume which the electrons cascade

to. The results of the Monte Carlo simulations for an infinite slab of ZnO are shown in Figure

2.3. From the Monte Carlo simulation, there are two values of interest: the maximum energy loss

14 per unit depth, Uo, and the Bohr-Bethe range, RB, which is the maximum range over which excitation occurs27.

Figure 2.3. Two-dimensional CASINO simulations for bare ZnO, with the Bohr-Bethe range RB of each probe energy EB noted.

2.3 Atomic Force Microscopy

Atomic force microscopy (AFM) is a useful technique for mapping surfaces on a nanoscale. The concept involves a cantilever with a nanoscale- or atomically-sharp tip

in order to measure along the surface of a material. This will induce an attractive or repulsive force, that is dependent on distance between the sample and that can be used to map surfaces topography. Figure 2.4 shows a schematic of non-contact mode AFM for the Park Systems XE-

70 SPM. Deflection of the cantilever is caused by changes in surface morphology. In essence a laser reflects off a mirror attached to the backside of the cantilever and mirror assembly. The reflected laser light hits a position sensitive detector, which converts that optical information into an electrical signal. A piezo-electric scanner can scan samples in the X-Y-Z direction. The scan proceeds line by line while the PSPD establishes a feedback loop which controls the vertical movement of the scanner as the cantilever is raster scanned across the surface.

Figure 2.4. Atomic force microscopy (AFM) mode operation for Park System XE-70 Scanning

Probe Microscope.

The AFM in use in our lab is has the capability to measure in non-contact mode using a

feedback loop coupled with an independent Z-scanner. At the short distances between sample

and tip there are two main forces, the van der Waals force and static electric repulsive force.

Contact AFM mode operates in the repulsive electric force region while non-contact mode

operates in the van der Waals region. In this van der Waals regime, the forces between the tip

and the sample are quite small, the non-contact mode thus detects changes in either the

vibrational amplitude or the phase of the cantilever. The cantilever oscillates over the surface of

the sample to generate an attractive force so the above changes can be measured. This measures

the surface morphology. More specifically, in non-contact mode, a piezoelectric bimorph is used

to vibrate the cantilever near the cantilever’s intrinsic resonant frequency (f0), usually between

100 kHz and 400 kHz, with an amplitude of a few nanometers. The resonant vibration has a

corresponding spring constant (k0), described by the following equation. 18

As the tip-sample distance decreases, the van der Waals force changes the amplitude and phase of the cantilever’s resonance, resulting in a new effective resonant frequency (feff) and effective spring constant (keff). As the tip-sample distance decreases, keff decreases and feff becomes smaller than f0.

Changes in the amplitude reflect changes in tip-sample distance. By measuring these changes in amplitude at the resonant frequency of the cantilever, the NC-AFM feedback loop compensates for the changes in tip-sample distance by controlling the Z-scanner movement. So, by maintaining constant amplitude and tip-sample distance, the NC-AFM mode can measure surface topography.

2.4 Kelvin Probe Force Microscopy

The AFM also has the capability to perform Kelvin Probe Force Microscopy (KPFM).

This is accomplished by using an enhanced electro static force microscopy (e-EFM) mode. In e-

EFM, an AC bias of frequency w, in addition to an applied DC bias, is applied to the the tip an external lock-in amplifier. The lock-in amplifier will separate the frequency component from the output signal. Figure 2.5 shows a representation of KPFM mode for the Park Systems XE-70

SPM. During KPFM, a feedback loop controls the DC bias by nulling out the ω term in the applied bias. When the tip and the samples reach a distance where they come into electrical contact, a force is generated. With this electrical contact to a semiconductor surface, the Fermi levels align. This shift in the semiconductor vacuum level will produce a force that is proportional to the contact potential difference (CPD).

Figure 2.5. Kelvin probe force microscopy (KPFM) mode operation for Park System XE-70

Scanning Probe Microscope.

A DC bias is then used to zero this force. This bias voltage is a measure of the surface potential. The difference is in the way the signal from the lock-in amplifier is processed. The signal from the lock-in may be expressed by the following equation

where VDC is the DC bias, VS is the surface potential, VAC is the applied AC bias, C is the capacitance of the tip-sample system, and d is the tip sample distance. The ω signal itself can be used to measure the surface potential. The amplitude of the ω signal is zero when VDC = VS or

20 when the DC offset matches the surface potential of the sample. A feedback loop can vary the

DC offset bias such that the output of the lock-in amplifier that measures the ω signal is zero.

This is used to generate and image of the surface potential using the variation of the DC offset bias.

2.5 Surface Photovoltage Spectroscopy

Surface photovoltage spectroscopy (SPS) provides a useful complement to DRCLS since it can determine the nature of the transitions involving states within the band gap. SPS involves measuring the potential difference between a probe tip and the specimen surface. This potential measurement technique is termed Kelvin Probe Force Microscopy (KPFM). When sub-band gap light illuminates a semiconductor, transitions involving gap states cause the band bending and work function to change. Figure 2.6 shows the experimental setup for taking SPS measurements, here using an atomic force microscope. Light from a broad band light source is passed through a monochromator then shone on the surface of our samples under the tip. The light can be controlled by the monochromator which scans the wavelength output. This induces work function changes at the surface of the material.

Figure 2.6. Experimental setup of SPS measurements. In addition to the standard KPFM setup, an outside source of light is applied to the surface which may excite electrons into a different band changing the potential on the surface. KPFM measures this change in potential and can tell us about surface features and defects.

The light applied can excite electrons into different states, as shown in Figure 4 for an n- type semiconductor. As an electron moves from one state to the next, the Fermi level changes and the KPFM measures this change in the Fermi level. In the Figure 2.7, the right image shows a transition from a trap into the conduction band. For n-type band bending, the newly promoted electron in the conduction band feels the band bending electric field that forces it into the bulk of the material, while trapping the hole in an isolated surface reduces the negative charge and the n- type band bending at the surface. The decrease in electron charge at the surface results in less band bending and the Fermi level moving up, i.e., closer to the vacuum level, decreasing the semiconductor surface work function. This results in the work function of the material being reduced, which in our systems results in an increase in the change in the contact potential difference (ΔCPD). As the light applied increases in energy above a certain energy, electrons 22 from the valence band may be promoted into a state with the photon energy as shown in the right side of Figure 2.7. This transition moves electrons into a surface state which leaves a minority carrier hole trapped at the surface that immediately recombines with majority carrier electrons. In turn this increases the density of the electrons at the surface, which increases band bending and increases the work function, moving the Fermi level away from the vacuum level and changing the contact potential different, ΔCPD, between the tip and surface9.

By using a monochromatic external source, one can control the energies of photons incident on the surface, thus allowing an exact energy to be applied to excite an electron into the conduction band from a trap state or from the valence band into the trap state. In Figure 2.8 a

23 typical SPS scan is shown where E2 and E1 show a population state and depopulation state and subsequent change in ΔCPD. The population state and depopulation state are the same as described above. The change in slope at E2 corresponds to the onset of a photodepopulation transition as shown in Fig.2.8, while E1 marks the onset of a photopopulation transition as shown in Fig. 2.8. When the energy is greater than the band gap Eg, electrons will transition into the conduction band from the valence band, producing both free electrons and holes that flatten the

9 band bending and decrease the work function as with transition E2 .

Figure 2.8. The above shows a typical SPS scan with transitions being present at energies E2 and

E1.

Once the energy level of the trap is known, a transient form of SPS enables one to then look at the transient states at those energies, to determine the trap density for this particular defect state. To look at the transient states, light is switched onto the surface of the material for a

period of time until the surface is saturated with electrons. The trap surface density can be estimated at a particular threshold by

(1)

where is the Boltzmann constant, q is the elementary charge, T is the temperature, is the bulk doping density, and ε is the dielectric constant. The 40 is the normalization constant for the

factor of . corresponds to slope change as the light is turned on and corresponds to

the slope just as the light is turned off, is the surface potential CPD without light, is

10 surface potential CPD after being saturated in light, and . Then to find the bulk density from the trap density one must divide the trap density by the surface depletion width

(2)

where is the semiconductor band bending.

2.6 X-ray Photoemission Spectroscopy

X-ray photoemission spectroscopy (XPS) is used to characterize the atomic composition of a material, chemical states of constituent atoms, as well as the surface electronic structure.

XPS is a technique that has a high amount surface sensitive technique. The photoelectric effect is 25 utilized to make XPS possible. As shows in Figure 2.928, a photon of energy hv enters the surface of a material and is absorbed by an electron with binding energy E1. Once the electron is ejected into the vacuum, it will have a kinetic energy equal to one of the core levers or valence band states of the material. The kinetic energy can be determined by the following expression

where EB is the characteristic binding energy of the photoelectron, Evac is the vacuum level of the semiconductor, and EFermi is the Fermi level of the electron spectrometer. The Fermi level is set at the point where the electron emission is zero and is independent of material. Binding energies of emitted is determined from the work function of the material. Thus, the Fermi level is a constant of kinetic energy across photoemission spectra.29 By taking the 80%/20% midpoint of a linear extrapolation of the valance band spectrum of a well-defined metal such as Au, the

Fermi level of the analyzer can be determined experimentally.

28 Figure 2.9. Schematic view of photo ionized electron E1.

After an X-ray enters the surface of a sample, the photoelectron can only travel a short distance before encountering scattering and changing the photoelectrons energy. The electrons that escape at their original energy add to the XPS signal. This short scattering length makes that

XPS very surface sensitive. The average analysis depth is ~50 Å. Figure 2.1030, shows the experimentally scattering lengths of electrons in a solid. This shows that electrons from the first few nanometers can escape elastically. From this curve electrons only from the first few nanometers of the surface escape elastically.

Figure 2.10. Electron scattering lengths in a solid. From this curve, electrons only from the first few nanometers of the surface escape elastically.29

2.7 Focused Ion Beam\ Electron Beam Induced Deposition

To process nanowires for contacts, an FEI Helios Nanolab 600 Focused Ion Beam (FIB) /

Scanning Electron Microscope (SEM) at the NanoSystems Laboratory (NSL) was employed.

This instrument is equipped with electron beam induced deposition (EBID) to pattern Pt metal over a relatively large writing field. This combination allows for Ga ion imaging, milling, and Pt deposition as seen in Figure 2.1131. The prepared ZnO nanowires are loaded into the UHV chamber, and the sample surface is raised to the eucentric height of the instrument. A hollow metal needle is then inserted 50 μm from the sample surface, near the desired nanowire. A Pt- bearing organometallic vapor is injected through the needle, and the electron beam is rastered over the desired area defined by the user in the software suite. deposit the platinum (Pt) metal as contact to nanowires

Figure 2.11. A schematic diagram showing (a)imaging, (b) milling, and (c) deposition.

The organometallic vapor flowing near the sample surface feels the Coulombic attraction of the primary electrons in the beam; the primary electrons also impact the dense vapor and scatter, generating secondary electrons. Like the scission process in electron beam sensitive organic photoresists, the secondary electrons cleave the Pt metal from the organometallic molecule, and the beam carries the Pt with it as it rasters the surface. The impinging Pt metal atom accelerates with energy comparable to the primary electron beam and strongly adheres to the surface. The remaining organic portion of the molecule is quickly pumped away via the system’s turbomolecular pump (TMP), leaving only a pattern of Pt metal on the surface. Contact thickness can be varied down to a lower limit of approximately 10 nm. An example of a nanowire with 150 nm diameter patterned with six Pt contacts of 200 nm width, 300 nm thickness can be seen below in Figure 2.12.

Figure 2.12. Six-contact structure to a 140 nm diameter ZnO nanowire. Contacts are 300 nm thick Pt patterned by electron beam induced deposition (EBID).

The FIB capability of the Helios Nanolab instrument was used to sculpt the outer annulus

of the nanowires using 5 keV Ga ions, as well as implant the near-surface with Ga ions

accelerated at 30 keV. Stopping and Range of Ions in Matter (SRIM) simulations demonstrate

that 5 keV Ga ions have a projected range RP of 4.4 nm, and 30 keV Ga ions have an RP of 14.7

nm, shown below in Figure 2.12. With 5 keV Ga ions, the doping effects are limited to this

narrow region, and the surface is primarily sputtered away with Ga. A three-dimensional

distribution for Ga ions implanted at 30 keV in ZnO can be seen below in Figure 2.13.

Figure 2.13. Stopping and Range of Ions in Matter (SRIM) simulations for (left) 5 keV and

(right) 30 keV accelerated Ga ions in ZnO.

Figure 2.14. Three-dimensional ion implant distribution for a 30 keV Ga ion beam incident on

ZnO.

Chapter 3 Ohmic and Schottky Contacts

3.1 Introduction

The role that metal contacts play in modern contacts play cannot be overstated. Chiefly

Schottky contacts. Many different devices require these contacts including but not limited to transparent thin film transistors, light emitting diodes and lasers, photo detectors, high electron mobility transistors, electronic nanostructures, and spintronic devices. The potential barrier that is formed at some metal semiconductor junctions is exploited to make a rectifying device that is known as a Schottky diode. The advantages this holds towards a p-n junction are a smaller depletion width and a smaller junction voltage. Fabrication is easier with majority carriers controlling the operation of the diode. Schottky diodes can have greater switching speeds than a p-n diode. This is due to Schottky diodes operate on majority carrier operation. These higher switching speeds allows for devices such as RF detectors and power convertors. The other type of contact is an ohmic contact. These ohmic contacts are another metal-on-semiconductor system, but one that is non- rectifying. Ideal ohmic contacts are low resistance. These are crucial for modern electronic devices. The formation of Schottky and ohmic depends on the work functions of both the semiconductor and the metal, the carrier densities of the semiconductor, the composition of surface, and the band gap of the semiconductor. In order to have an understanding of how these electronic properties is dependent on the surfaces and the interfaces of the metal/semiconductors, it is important to under the processes that occur during formation of these contacts.

3.2 Schottky Theory

Metal on semiconductor contacts that exhibit rectifying behavior are Schottky diodes. In the 1960's Schottky barrier studies were conducted to determine the interplay between the Fermi level "pinning" in the band gap and surface states of ZnO and other semiconductors. The model for basic barrier formation between an n-type semiconductor and metals is shown in Figure 3.1.

This illustration shows the case where the work function of the metal is great than the work function of the semiconductor (ΦM > ΦSC). Figure 3.1 (a) shows the energy bands of the semiconductor and metal before contact. Note the different Fermi level difference. After contact, electrons will move into the metal from the semiconductor, thus depleting the surface of electrons. Figure 3.1 (b) shows the Fermi levels of the metal and semiconductor align, after the charge has been transferred. Across this depleted region, on the charge layer of the surface, there is a voltage drop qVB. This is equivalent to the drop in voltage equal to the contact potential difference between the bulk of the semiconductor and the metal. The width of the depletion region layer (W) is given by the following expression:

where εs is the static dielectric constant of the semiconductor, Nq is the bulk concentration of ionized impurities within the surface space charge region, and (V-V0) represents the change in contact potential of the semiconductor surface. The barrier height depicted here is an ideal case which is dependent on the work functions of the metal and semiconductor.

This ideal formulation fails to model the behavior at most metal/semiconductor junctions. The barrier height in actuality is only weakly dependent on the metal work function, that is, the Fermi level position in the semiconductor band stabilizes in a much narrower range of energies than the range of metal work functions.

In 1947, Bardeen32 proposed that localized surface states could accumulate at the metal- semiconductor interface. This accumulation layer would change the contact potential difference between the metal and semiconductor, changing the expected barrier height. In Figure 3.2 (a), acceptor like surface states will have energy band bending upwards. This will move the Fermi level towards the energy level of the surface states (ESS). With a high enough density of surface states, the Fermi level can be "pinned" to the ESS, and cannot be moved due to the metal contact.

This is shown in Figure 3.2 (b). A dipole region is formed at the surface, by a large number of sure face state and not from the layer of surface space charge. This causes charge transfer to the metal. Hence, the barrier height instead may be modeled as:

where Δχ depends on the surface state energy position (ESS) within the semiconductor band gap.

This results in a barrier height that is weakly dependent on the metal work function.33

3.3 Schottky Barrier Measurement Techniques

Current can flow through the interface of a mechanism of a metal-semiconductor interface through one of three methods. These are: 1) thermionic emission, 2) recombination or generation, and 3) field emission (tunneling). For an ideal Schottky diode, thermionic emission over the barrier will be the dominant charge transport mechanism. Drift, diffusion, and emission over the barrier will create a current. This can measure using a current-voltage (I-V) technique.

For thermionic emission the forward current at an applied voltage through a metal/semiconductor interface is given by

where A** is the Richardson’s constant, kB is Boltzmann’s constant, T is temperature and ΦB is the barrier height of the semiconductor, V is the applied voltage and n is known as the “ideality factor”. Forward current density, extrapolated logarithmically to zero applied forward bias, has an intercept at.

Barrier height can thus be extrapolated from a plot of ln J vs. applied voltage (V) as shown in

Figure 3.3. The slope of the ln J vs. V curve also gives the ideality factor (n) of the contact, which is defined by

The “ideality factor” is a measure of how close a diode is to ideal (n=1), and therefore acts as a

measure of diode quality. The factors that determine an ideality factor greater than one are

electron-hole recombination, traps at the interface, and dielectric layers.

Figure 3.3. Forward current density J versus applied voltage (V) for a metal/semiconductor contact. This can be used to determine both barrier height and ideality factor

Measurements techniques, such as capacitance-voltage (C-V) and internal photoemission

(IPE) are also used in the determination of Schottky barrier heights. Those methods, along with

I-V measurement, can each yield a differing measure of the Schottky barrier height. Each

42 technique though, carries its advantages and disadvantages. Some examples are the I-V is sensitive to tunneling and interface recombination and traps, and C-V measurements can be affected by deep level traps. Although C-V is less likely to be affected by tunneling. Ideally, a combination of these techniques should be used to best characterize Schottky barrier heights.

Chapter 4 Native point defect energies, densities, and electrostatic repulsion

across (Mg,Zn)O alloys

4.1 Introduction

MgZnO alloys are emerging as exciting UV optoelectronic materials 34,35,36,37,38 together with ZnO39,40 based on their large exciton binding energies, small lattice mismatch with ZnO substrates, and ability to create heterostructures for quantum well and superlattices by varying

Mg content. Deep level defects in these materials will impact dipole formation, Schottky barriers and heterojunction band offsets,41,42 yet how their physical properties depend on alloy content and lattice structure is only now being investigated. We have used depth-resolved cathodoluminescence spectroscopy (DRCLS) and nanoscale surface photovoltage spectroscopy

(n-SPS) to measure the dependence of native point defect energies and densities on Mg content, band gap, and lattice structure in non-polar MgxZn1-xO (0

Mg content can be used to assess the role of electric fields and strain on the native defect distribution near MgZnO interfaces. In turn, the native point defect type, energies, and densities can be correlated with Schottky barriers measured previously.

Here we present DRCLS spectra corresponding to zinc or magnesium cation (VC) and oxygen vacancies ((VC and VO), respectively, previously identified by positron annihilation spectroscopy (PAS) and electron paramagnetic resonance (EPR), respectively, with significant

44 surface segregation but relatively constant bulk densities. A linear variation of MgxZn1-xO band gap is evident with Mg content up to 52% from both DRCLS and SPS. SPS work function changes at photopopulation and depopulation thresholds provided defect level positions with

43,44 respect to band edges. VO (VC) energy level movements vs. Mg% parallel valence and conduction band edge variations respectively, consistent with their orbital-derived nature. Both

VO and VC defect densities exhibit a pronounced minimum at ~45% Mg corresponding to similar a and c parameter minima at ~52%. Furthermore, reported Schottky barrier heights decrease from ~1.2 to 1 eV above ~45% Mg,45 consistent with stronger Fermi level pinning as defect densities rise, the reduced lattice parameters may inhibit defect formation due to electrostatic repulsion as reflected in DFT calculations that assess the roles of electric fields and strain on the native defect distribution. These results highlight the reduction of native defect densities in

(Mg,Zn)O alloys by Mg incorporation and in particular the systematic effect of lattice structural changes on defect densities across this wide alloy series.

4.2 Experiment

We used a variety of structural and electronic techniques for this study. Five single- phase, one μm-thick MgxZn1-xO films were grown on a-plane sapphire by molecular beam epitaxy (MBE) over a wide (0

(RBS/C) showed that ~93% of all Mg atoms occupied Zn sites homogeneously within the wurtzite lattice structure.49 Alloy compositions were measured directly with RBS to be x= 0,

0.31, 0.44, 0.52, and 0.56. X-ray diffraction (XRD) showed that these crystals were single-phase, high quality wurtzite structure with no cubic inclusions over the entire alloy series studied.16

Atomic Force Microscopy (AFM) maps showed smooth surfaces on a nanometer scale without surface asperities.

DRCLS performed in ultrahigh vacuum (UHV) employed a glancing incidence electron gun and incident electron beam energies EB = 1, 2, 3, 4, and 5 keV. Based on Monte Carlo

50 simulations, these EB corresponded to peak electron-hole (e-h) pair creation depths U0 = 7, 18,

32, 50, and 72 nm, respectively, and Bohr-Bethe maximum range RB = 20, 50, 80, 130, and 180 nm, accordingly. The optical train for DRCLS consists of a CaF2 focusing lens, a sapphire viewport, and an f-number matcher coupled to an Oriel monochromator and CCD detector.

DRCL spectra obtained in this excitation region provide optical emissions due to near band edge

(NBE) and defect level-to-band edge energy level transitions with nanometer depth resolution.51,52,53 Previous DRCLS measurements of ZnO single crystals from various sources demonstrate that such defects exhibit emissions that can vary by orders of magnitude with depth, crystal growth method, supplier, and even batches from the same supplier.54 Such results underscore the need for depth-resolved measurements to obtain representative defect densities of a given crystal.

SPS measurements involved a Park XE-70 AFM operated in Kelvin Probe Force

Microscopy (KPFM) mode. Here the contact potential difference (cpd) between probe tip and sample surface is measured as a function of incident photon energy h from an Oriel monochromator and tungsten light source. Together with DRCLS-measured NBE emission energies, SPS spectra exhibit cpd changes in slope versus h whose sign depends on

46 photoinduced photopopulation or depopulation of states within the band gap with respect to the conduction band EC or valence band EV. Likewise, a pronounced cpd change at h = EG provides a measure of the semiconductor band gap. Furthermore, a transient variation of SPS (t-SPS) provides densities of these deep level defects.55,56 These measurements were compared with deep level optical spectroscopy (DLOS), current-voltage (I-V), Schottky barrier, and steady state photo-capacitance (SSPC) measurements published previously on the same films.57

4.3 Results

4.3.1 DRCLS – Figure 1 presents representative DRCL spectra for each of the five (Mg,Zn)O alloys. The NBE peak in each panel provides an indication of the band gap for each of the alloys.

Defect emissions appear at energies below each of the NBE features. Here second order grating replicas of the NBE emission have been subtracted. Each of the sub-band gap features appears to consist of at least two peaks whose relative intensities change with alloy composition and with excitation depth.

Figure 4.1. Representative DRCLS spectra for each of the five MBE-grown (Mg,Zn)O

alloys.

Previous theoretical studies have determined the most energetically stable native point

58,59,60 defects to be the VO and VZn defects. Complementary PAS and DRCLS studies versus excitation depth have identified optical emissions in the 1.7 – 2.1 eV range with the zinc vacancy

61,62 VZn and VZn clusters, increasing in energy with cluster size. Similarly, a combination of chemical process-dependent and electron paramagnetic resonance (EPR) studies suggest that emission at 2.4-2.5 eV is related to the oxygen vacancy.63 For the ZnO (x = 0) depth-dependent spectra of Fig. 1(a), features at 1.77 and 2.33 eV are evident that support the identification of these defects with VZn and VO, respectively.

For each of the alloys measured in Fig.1, the intensities of the native point defect emissions relative to their respective NBE intensities changes by orders of magnitude with depth of excitation. Figure 2 illustrates plots of these defect ratios I(defect)/I(NBE) for VC and VO as a function of EB and corresponding depth of peak e-h pair creation rate. Figure 2(a) shows a large rise in both I(VZn)/I(NBE) and I(VO)/I(NBE) with decreasing depth. For x = 0, VZn segregation to the surface is much more pronounced than that of VO. For EB > 2 keV, corresponding to a peak e-h excitation depth of ~ 20 nm, these normalized defect intensities appear relatively constant and representative of the bulk films. Segregation of both VC and VO defects is evident in Figs.

2(b) and 2(c) for x = 0.31 and 0.52, respectively. However, the magnitude of both I(VC)/I(NBE) and I(VO)/I(NBE) is strongly reduced with the addition of Mg, both at the surface and in the bulk. Figure 2(b) shows that the surface segregation depth decreases to less than 10 nm with 31%

Mg content. Likewise, the Mg0.52Zn0.48O surface segregation depth in Fig. 2(c) appears to be in the 10-12 nm range. In both Figs. 2(b) and (c), the magnitudes of VC and VO segregation appear to be comparable. Overall, the segregation of VC and VO to the free surface is significant since these films are oriented along a non-polar direction so that piezoelectric fields cannot be contributing to the defect segregation.

Figure 4.2. Depth profile of I(Defect)/I(NBE) for a-plane MgxZn1-xO films versus EB from 1 to 5 keV and depth of peak e-h pair creation rate UO from the free surface to 72 nm for x = 0, 0.31,

0.44, 0.52, and 0.56. Both segregated and bulk defect densities decrease to minimum values at

44% Mg.

Figure 1 also shows that the normalized defect densities in the bulk vary dramatically with Mg content. Figure 3 displays values for EB = 2 keV I(VZn)/I(NBE) and I(VO)/I(NBE) that are characteristic of the bulk films. Here, I(VC)/I(NBE) and I(VO)/I(NBE) exhibit > 100 x decrease and > 30 x decreases, respectively, with the addition of 31% Mg. These normalized densities reach a minimum at [Mg] = 44%, then begin to rise for [Mg] = 52% and 56%. Figure 3 is significant since it shows that the addition of Mg to ZnO produces a strong decrease in mid- gap deep level defects – defects are the dominant recombination centers for free carriers in this 50 oxide system. Figure 1 shows high energy peaks that correspond to NBE edge characteristic of each alloy’s band gap. Peak energies correspond to the highest NBE intensity and increase linearly from x = 0 to 0.52 as shown in Fig. 4.4 (blue line) according to EG(x) = 3.309+1.69*x.

Above x = 0.52, the band gap increases more rapidly, consistent with the much higher MgO end point band gap of 7.8 eV and the crystal strain near the wurtzite-rocksalt crystal crossover.

These band gap variations are consistent with theory64 and experimental reports.65,66,67 Figure 1 also shows that the energies of the native point defects increase with increasing Mg content although much less than the NBE increases.

Figure 4.3. MgxZn1-xO normalized intensities of VC and VO – related defects versus Mg alloy content. Error bars for all but x = 0 are within the symbols plotted.

Figure 4.4. MgxZn1-xO band gaps measured versus Mg content x by CL, SPS, and

SSPC. All three techniques show a nearly linear increase up to [Mg] = 52%.

4.3.2 SPS - In order to determine the absolute energy level positions of defect states in each band gap, we used SPS to measure the threshold energies to photo-populate or depopulate the defect levels within each band gap. For x = 0.31, Fig. 5 shows positive slope changes at 1.85 and 2.5 eV, corresponding to photo-induced depopulations of states 1.85 and 2.5 eV below conduction band EC. Negative slope changes occur at 2.05 and 2.25 eV, corresponding to photo-induced population of states 2.05 and 2.25 eV above the valence band EV. The strong positive slope at

3.60 eV and peak at 4.05 eV are due to band gap transitions and excess free carriers that tend to reduce the band bending. Similar SPS features are evident for x = 0.44 and 0.56 in Fig. 5. Note that the positive and negative slope changes at 1.85 eV and 2.05 eV correspond to depopulation and population, respectively, of the same energy level located ~ 1.85 eV below the conduction band edge and ~ 2.05 eV above the valence band. These two energies add up to nearly the band gap of 4.05 eV determined from the peak NBE intensity with the difference due in part to the energy broadening of the gap state. The CL data shown for each of these alloys illustrates the correspondence between the SPS and CLS transitions and identifies the defect level positions within the band gap responsible for the sub-gap CL emissions.

4.3.3 Technique Comparison – The DRCLS and SPS technique both provide measurements of optical transitions between energy levels. While DRCLS enables depth-dependent measurements, SPS provides energy level positions with respect to the band edges. Both techniques are sensitive to transitions between conduction and valence bands. Previous steady state photocapacitance (SSPC) studies on the same samples also provide measurements of band gap transitions.45 Figure 4.4 illustrates a comparison of band gaps measured by each of these

54 techniques. DRCLS, SPS, and SSPC all exhibit a nearly linear increase of (Mg,Zn)O band gap with Mg content from 0 < x < 0.52) for this wide range of single crystal alloy compositions.

4.4 Analysis and Discussion

4.4.1 Defect Energy Levels – Both Fig.1 DRCLS and Fig. 4.5 SPS results show that the deep level defect energies as well as the band gaps in this alloy series vary with Mg content. Figure

4.6 illustrates the energy levels of the VC and VO-related defects in their band gaps and their variation with Mg content. Here the energy level position of EC is determined to be -4.6 eV relative to the vacuum level EVAC based on the electron affinity of the ZnO (10ī0) surface and EV

68 follows from the measured band gap. The variation of EC and EV with increasing EG follows the well-known 2/3-1/3 rule. The VO-related defect appears to decrease in parallel with the valence band, which is O 2p-derived, while the VC-related defects appear to track with the conduction band, which is derived from Zn 4s states. Thus these native defects seem to reflect their atomic orbital-derived nature

Figure 4.6. Level positions of VC and VO within the MgxZn1-xO (0 < x < 0.56 ) band gap vs.

Mg/ (Mg + Zn). The midgap VO and VC levels appear to vary with EV and EC, respectively.

4.4.2 Defect Density Reduction – Figure 4.3 showed that both VC and VO defect densities as reflected by their normalized DRCLS intensities decrease dramatically with Mg content.

Thermodynamics may play a role in the VC decrease since Mg forms stronger bonds with O than

69 Zn, i.e., the MgO heat of formation -H298(kJ/mole) = 601.6 for MgO versus 350.5 for ZnO.

Thus it is energetically more favourable to form Mg-O rather than Zn-O bond during growth, suggesting that Zn vacancy sites would be filled more readily with increasing Mg content.

The (Mg,Zn)O lattice structure may also impact the defect energetics and equilibrium densities.

Figure 4.7 shows that the measured lattice parameters of these crystals, which vary significantly with Mg content from the a-plane ZnO lattice constants of c = 5.178Å and a = 3.245 Å13 and also reach a minimum in the same alloy range as VC and VO defect densities. Note that the wide alloy range of these samples enables us to identify the correspondence between minima in defect density and unit cell volume.

Figure 4.7. Dependence of MgxZn1-xO lattice parameters a and c versus x for 0 < x < 0.56.

Both a and c reach minimum values for x ~ 0.52.

The 2.5% decrease in unit cell may inhibit lattice vacancy formation due to electrostatic repulsion since, for example, the removal of an oxygen atom between two cation atoms results in more charge on each cation and the absence of screening by the missing anion. The effect of such repulsion is to expand the lattice, an effect reported for complex oxides.70,71 Conversely, a decrease in unit cell volume increases the formation energy for such vacancies, thereby lowering their equilibrium defect densities. Preliminary density-functional theory (DFT) calculations based on free energy changes with lattice dimensions and Mg content are also consistent with the defect variations in Fig. 4.3.72Again, the wide alloy range investigated here enables the identification of the corresponding minima in Fig.4.3 and 4.7.

4.4.3 Schottky Barrier Formation – Previous electronic measurements of Schottky barrier

35 height SB, ideality factor , and series resistance RS for the same (Mg,Zn)O specimens also showed significant changes in the alloy range of VC, VO, and unit cell volume minima. In

Fig.4.3, VC and VO defect densities increase from the x=0.44 minimum to x= 0.56 by 11- and 5- fold, respectively. At x=0.44, Gür et al. measure an Au-(Mg,Zn)O SB that increases to a maximum of 1.26 eV, then decreases to 1.04 and 1.0 eV at x=0.52 and 0.56, respectively.35 The increase in mid-gap defect densities between x=0.44 and 0.56 may account for the pronounced change in SB over this same range. The decrease in SB is consistent with the larger increase in

VC acceptor versus VO donor density, introducing a negative dipole that raises the Fermi level and decreases SB at the junction.

4.5 Conclusions

Combined DRCLS and SPS studies of non-polar, single-phase MgxZn1-xO (0

Chapter 5 Direct Measurement of Defect and Dopant Abruptness at High

Electron Mobility ZnO Homojunctions

5.1 Introduction

Semiconductor heterojunctions have enabled a wide array of microelectronic applications for high speed electronics73,74 as well as their extension to fundamental physical studies.75

Central to the high mobility required for these advances is the ability to spatially separate the impurity dopants that provide free carriers from the channel layer in which these carriers move.

However, the utility of heterojunctions can be limited by lattice mismatch between the constituents that can lead to dislocations, scattering centers, and reduced mobility. Recent work shows that enhanced electron mobility channels are also possible for semiconductor homojunctions, thus bypassing some of the disadvantages of lattice mismatch. Look et al.76 used

Hall-effect measurements to show that layers of ZnO degenerately doped with Ga (GZO) and grown on unintentionally doped ZnO by molecular beam epitaxy (MBE) will produce a significant electron leakage (Debye tail) into the ZnO due to the large Fermi-level EF mismatch between the GZO and the ZnO. Since the ZnO will have a much higher mobility than the GZO the net result is an approximately 2-nm-thick layer of high carrier concentration nZnO and high mobility ZnO. The ZnO Debye layer will have its greatest effect (highest increase in total) on structures with thin GZO layers. For example, a 2-nm Debye layer would not have as much effect on a 300-nm GZO layer as it would on a 5-nm GZO layer, since the former would be completely dominated by the GZO. This is the reason that  increases as dGZO decreases, in contrast to the typical behavior of  in a thin film on a lattice-mismatched substrate in which a poor interfacial region adds low-mobility, instead of high-mobility, electrons4. Our present study 60 includes four samples, with dGZO = 5, 25, 50, and 300 nm. Since a 300-nm-GZO/ZnO structure is little affected by the ~2-nm Debye tail, we can determine donor (ND) and acceptor (NA) concentrations in the GZO by degenerate scattering theory, and it is also known from studies of similar material that ND = [Ga] and NA = [VZn], where VZn is the Zn vacancy. The results are:

21 20 3 [Ga] = 1.04 x 10 and [VZn] = 1.03 x 10 cm . For the thinner GZO layers, 5, 25, and 50 nm in this case, the theoretical analysis requires detailed profiles of n in the GZO and ZnO, using

Poisson theory, and then profiles of , using scattering theory, and finally multilayer Hall-effect theory to add up all the contributions and predict the values of Hall mobility H and sheet concentration ns in the whole GZO/ZnO structure. Although the agreement between experimental and theoretical values of H and ns in the 5, 25, and 50-nm samples was good, as shown in Fig. 1 of Ref. 76, it seemed worthwhile to directly examine the main hypothesis, i.e., the abruptness of the Ga and VZn profiles. To accomplish this task, the VZn profile was determined on a nm scale by a differential form of depth-resolved cathodoluminescence spectroscopy, and the Ga profile by X-ray photoelectron spectroscopy. Then the profiles were modified by an anneal in Ar at 500C for 10 min, and reexamined. For both the as-grown and annealed sample, the profiles were basically abrupt although displaying some interesting non- abrupt features. However, differences in the electrical properties were not large in either case.

This speaks well for practical applications of this technology.

The distribution of zinc vacancies (VZn) is of particular interest since their density is known to be high in degenerately doped GZO, in fact, ~ 1.0 x 1020 cm-3 in these particular samples. This defect becomes increasingly more stable thermodynamically as Fermi levels rise

77 due to the high donor (GaZn) doping. In addition, VZn densities have been observed to increase

78,79 significantly at the free surface of ZnO grown by MBE. VZn and O vacancies (VO) are the two most thermodynamically stable defects in ZnO under high Zn partial pressure and with

Fermi levels high in the band gap.80 Both types of defects exhibit characteristic luminescence

81 82 emissions, VZn at 1.7 to 2 eV increasing with increasing cluster size and VO at 2.43 eV. Both assignments are consistent with chemical reactions observed at metal-ZnO interfaces and their electrically active nature.83

5.2 Experimental Setup

A major challenge in evaluating the spatial distribution of defects near interfaces at these nanoscale thicknesses is determining their depth distribution on a near-nm scale. To address this, we used a differential form of depth-resolved cathodoluminescence spectroscopy (DRCLS) termed DDRCLS to measure these defects in the same samples used in the Look et al. study.

DDRCLS involves normalizing and subtracting out emissions generated by lower beam energies

(EB) from those at higher EB. Previously we showed that this differential technique is capable of resolving electronic features at buried interfaces on a near-nm scale.84 Here we used incident beam energies EB = 0.5 – 5 keV to generate electron-hole (e-h) pairs that recombined to exhibit

DRCL features due to near band edge (NBE) and defect-related electronic transitions. Intrinsic band gap emissions in ZnO appear at 3.27 eV, while NBE conduction-to-valence band transitions extend from ~3 to 3.5 eV in GZO due to Burstein-Moss band filling85, previously

86 reported for thicker GZO layers. VZn emissions appear in both GZO and ZnO at ~1.8 eV.

DDRCLS enabled us to profile the deconvolved VZn density across the GZO/ZnO interface, consistent with Monte Carlo simulations of e-h pair creation and luminescence excitation rate.87

This unique defect depth precision revealed chemical and electronic interfaces that are initially monolayer-abrupt, but with VZn profiles that extended from GZO into the ZnO and increased with GZO thickness. The GZO layers were grown by MBE under growth conditions described previously with widths of 300, 50, 25, and 5 nm on 160 nm ZnO buffer layers on 3 m-thick

GaN on sapphire.76 See Figure 5.1. DRCL spectra were acquired using a glancing incident electron gun in ultrahigh vacuum (UHV). Depth-resolved chemical information was acquired by

X-ray photoemission spectroscopy (XPS) using a glancing incidence 1 keV 0.5 A Ar+ ion gun in a PHI Versaprobe under UHV conditions.

5.3 Results

DRCL spectra exhibit features that provide several measures of interface abruptness including depth profiles of band-edge and native point defect emissions. Figure 1(a) shows

DRCL spectra for GZO thickness of 25 nm with increasing EB from 1.0 to 3.0 keV, corresponding to peak Monte Carlo excitation depths U0 of 7.4 to 34 nm, respectively, and maximum excitation depths RB of 19 to 86 nm, respectively. The second order grating replica is subtracted for clarity. These spectra exhibit a pronounced change in NBE features as excitation extends from the GZO layer into the ZnO. At lower EB and excitation primarily in the GZO,

NBE features are dominated by an emission peak at 3.5 eV with a lower energy shoulder at 2.9 eV. The 3.5 eV peak likely results from excited holes moving up until they reach kFermi, then recombining with free electrons that already fill the conduction band up to kFermi. The 2.9 eV feature may involve conduction band-to- acceptor (e-A) transitions as well as lower band gap

energies due to correlation effects. The intensities of these two features appear to scale together

within the GZO.

1 e- hν 3.27 eV 3.50 eV 1.0kV 3.25 ZnO 0.5 kV ZnO NBE GZO NBE 1.2kV 1.0 kV 1.4kV NBE 1.5 kV 2.92 eV 1.6kV 5 nm GZO/ZnO 2.0 kV 1.8kV 2.5 kV 2.0kV (b) 3.0 kV 0.1 2.2kV 3.5 kV 2.4kV 4.0 kV 2.6kV 4.5 kV 5.0 kV 2.8kV 1.80 eV 3.0kV 1.8 eV

VZn 2.2 eV VZn

0.01 Intensity(A.U.)

25 nm GZO/ZnO

0.001 2.20 eV (a)

2 3 4 5 2 3 4 5 Photon Energy (eV) Photon Energy (eV) Figure 5.1. Normalized room temperature DDRCL spectra of (a) 25nm and (b) 5 nm GZO on ZnO.

Above bandgap emission in 25nm GZO due to degenerate doping is absent inside ZnO

The 3.5 eV peak reaches a maximum at EB =1.2 keV, consistent with an RB = 22 nm.

With increasing EB, these features decrease in intensity, replaced by a dominant peak at 3.27 eV

corresponding to the ZnO buffer layer underneath the GZO. Figure 1(a) also exhibits lower

energy emissions at 1.80 eV attributed to VZn defects and a broad shoulder extending from 2 –

2.5 eV that is due to re-excitation from the underlying GaN substrate.88 Spectral ripples apparent

for energies below 2.5 eV are due to interference fringes within the ZnO layer, confirmed by

64 angle-dependent CL spectra. With increasing EB, the VZn peak intensity decreases rapidly as excitation extends into the ZnO, consistent with the ZnO’s lower doping.

Figure 5.1(b) shows DRCL spectra for 5 nm GZO on ZnO with EB = 0.5 to 5 keV, corresponding to peak Monte Carlo excitation depths U0 = 3 to 77 nm and maximum excitation depths RB = 8 to 185 nm, respectively. Since the 5 nm thickness is less than the diffusion length of minority carrier holes and the valence band of ZnO lies above that of GZO with EF levels aligned, almost all holes recombine inside the ZnO. Since minority carriers determine the spatial location of recombination, the dominant emission features are those of ZnO even at the shallowest (3 – 8 nm) excitation depth. VZn emission for this lowest EB excitation corresponds to excitation within a few nm of the GZO/ZnO interface and is less than that at all depths in Fig.

1(a), decreasing even further with increasing EB and excitation into the ZnO.

Figure 5.2 Comparison of GZO/ZnO VZn vacancy peak areas versus DRCLS probe depth for

GZO thicknesses 0, 5, 25, and 50 nm.

Figure 5.2 illustrates depth profiles of the deconvolved VZn peak areas for the bare ZnO surface and GZO thicknesses of 5, 25, and 50 nm. The deconvolution was done by fitting

Gaussian functions to known transitions in the ZnO CL spectra, then normalizing the areas of the defect transitions to the areas of the NBE transitions. Background is removed during the 66 acquisition process. Hall effect measurements provide a calibration of VZn density in Figure 2.

DRCL spectra within the 300, 50, and 25 nm GZO layers exhibit similar area ratios of A(VZn)/

A(NBE), consistent with the constant volume carrier concentration for all layers.4 The 300 nm sample has a conductance almost entirely determined by the GZO, and with the measured values

20 -3 2 of carrier density n= 8.53 x 10 cm and  = 34.2 cm /Vs, we calculate a donor density ND =

21 -3 20 -3 1.06 x10 cm and acceptor density NA = 1.03 x 10 cm . Since we know that NA = [VZn], then

20 -3 [VZn] = 1.03 x 10 cm within the GZO and more than one order of magnitude lower in the

78 ZnO. As with MBE-grown ZnO reported elsewhere, VZn emission intensities for the ZnO buffer layer without a GZO layer increase near the free surface then decrease by > 4x into the

ZnO bulk. The 5 nm GZO profile shows that [VZn] normalized to NBE emission within a few nm of the GZO/ZnO interface nearly matches that of the bare ZnO buffer well away from the interface. Both the 25 and 50 nm GZO samples exhibit a pronounced VZn increase within the

GZO near the GZO/ZnO interface and comparable densities. The 300 nm GZO exhibits nearly flat [VZn] intensity except within 20 nm of the GZO free surface, where it increases by only 35%.

The 25 nm GZO profile exhibits higher VZn integrated areas than either the 5 nm GZO or the bare ZnO within the first 10-20 nm on the ZnO side of the GZO/ZnO interface. Nevertheless, the sharp VZn drop within the first 10 nm of the 25 nm GZO/ZnO interface is indicative of its abruptness on this nm scale. The 50 nm GZO specimen indicates even higher VZn densities extending into the ZnO and a broadened interface.

The plots in Fig. 5.2, in general, show that VZn defects extend from GZO into the ZnO up to ~ 50 nm, increasing with increasing GZO thickness and overlapping the 2 nm high-mobility

Debye tail extending into the ZnO in which high carrier density transport occurs. The ~ 1019 cm-3

[VZn] density in this region suggested by Fig. 5.2 appears to be somewhat high since we know

19 18 -3 that ND = 3.0 x 10 , and NA = 8.1 x 10 cm (probably also VZn) in the bare ZnO. Adding [VZn]

19 -3 2 =10 cm to NA in this region would reduce the theoretical mobility from 60.7 to 47.7 cm /Vs, with the latter value being well below the experimental mobility of 64.1 cm2/V-s. However, the additional VZn beyond the interface may be clustered, and thus not contributing strongly to scattering. Indeed, VZn clustering is consistent with a peak shift to higher energy, as is known to occur in ZnO81, and also with a broadening of the ~1.8 eV peak, evident in Fig. 1(b) versus 1(a).

Thus, it is likely that the additional VZn observed by CL in the ZnO beyond the interface does not affect the measured mobility significantly.

Hall effect measurements show that carrier densities (and hence VZn densities) are nearly constant over the entire 5 – 300 nm range of GZO film thicknesses. Thus, e.g., since the VZn densities in the 25 and 50 nm GZO films are comparable, any driving force for VZn diffusion at the same temperature (250°C) into the ZnO should be similar, yet VZn densities (including both isolated and clustered) increase measurably for the 50 nm sample. The difference must be related to the growth time, calculated from the growth rate of 3.3 nm per minute. To further study the effects of diffusion, we performed an anneal at 500°C for 10 min. in flowing Ar. Then the VZn profiles were measured by DDRCLS, and the Ga profiles, by XPS.

The DDRCL spectra for the 25 nm GZO/ZnO interface after the anneal are illustrated in

Fig. 5.3. These spectra can be compared with those shown for the same sample at the same energies in Figure 1(a). The above band gap emissions that dominated at low EB [Fig. 1(a)] are no longer present after annealing (Fig. 5.3) and the NBE peak emission energy (Fig. 5.3) now occurs very close to the room temperature value in ZnO [Fig. 1(a)]. VZn defect emissions

68 increase with annealing and are again strongest at EB ~ 1.2 keV. To help understand these optical data, the 300-nm GZO/ZnO sample was subjected to the same anneal and analyzed by Hall effect measurements, since the latter are dominated by GZO, not ZnO, in this thick sample. The results

20 20 -3 2 were: n decreased from 5.91x10 to 4.38x10 cm ; μ increased from 36.0 to 51.8 cm /V-s; ND

20 20 19 19 -3 decreased from 7.59 x10 to 4.86 x10 ; and NA decreased from 8.42 x10 to 2.53 x10 cm .

The decrease in the NBE peak emission energy can be explained by the reduced band filling due to lower n, causing the holes to recombine with electrons at lower values of wave vector k, thus producing lower emission energies. The normalized VZn emission is also reduced, but is a higher fraction of the NBE emission, which is reduced even more. Note that our understanding of the

CL data is enhanced by the Hall effect data.

Figure 5.3. Room temperature DRCL spectra of 25 nm GZO on ZnO after 10 min. 500 °C anneal in Ar. Above bandgap emission in GZO decreases in GZO but extends into ZnO.

Figure 5.4(a) shows the integrated areas of NBE emission for spectra of the 25 nm

GZO/ZnO sample shown in Figures 5.3 and 5.1(a). These areas are obtained from deconvolution of the NBE emissions using the same energy and linewidth parameters. Intensities are normalized to the bulk GZO levels as shown. Comparison of these depth profiles in Figure 5.4(a) shows that the annealed GZO peak area variation across the interface to the ZnO is less abrupt

70 after the anneal. The GZO integrated area decreases by 20 – 30% for depths inside the GZO and increases by similar amounts inside the ZnO; the GZO increase is largest within the first ~10 nm of the ZnO past the interface. Similarly, Figure 5.4(b) shows that the annealed ZnO peak variation is less abrupt after the anneal. The ZnO integrated area decreases by 25-30% inside the

ZnO within 20 nm of the interface. For both GZO and ZnO, the interface width broadened to nearly 30 nm. This broadening is significantly greater than the spatial extent of the high- mobility Debye tail.

Figure 5.4. Integrated NBE peak areas vs. depth of (a) GZO and (b) ZnO before and after annealing. The GZO area profile normalized to the bulk GZO level broadens after anneal.

The broadening of the GZO profile suggests that Ga dopant diffusion occurs at elevated temperatures. Figure 55.5(a) shows a characteristic XPS survey spectrum from the 25 nm

GZO/ZnO sample. Only Zn, O, and Ga are present in this spectrum. The insert shows a Ga

2p3/2/2p1/2 feature at 1118-117 eV binding energy whose intensity corresponds to a nominal 2% surface Ga composition at the free surface of GZO. Depth profiling spectra obtained at a sputter rate of 1 nm per minute provided the variation of [Ga] from the free GZO surface into the ZnO using standardized peak sensitivity factors supplied by the manufacturer. [Ga] peaks were well- resolved at depths extending past the interface more than 3 nm into the ZnO. Before the anneal,

Figure 5.5(b) shows that [Ga] decreases to 0 at the interface (x = 0). The 500 °C, 10 min. anneal increases the [Ga] content within a few nm of the interface. At x = 0, [Ga] increases to 0.457%.

This [Ga] increase indicates that Ga atoms can diffuse out of GZO into the ZnO under these anneal conditions, and combined with the above results it is clear that both native defects and impurity atoms can diffuse at 500 °C. In this case, the electrical properties are not greatly affected, but previous studies have shown that a 600°C anneal strongly reduces the concentration due to Ga atoms being removed from donor states.77

Figure 5.5.( a) XPS survey and Ga spectra of GZO. (b) Measured Ga content versus sputter depth profiling of 25 nm GZO/ZnO pre- vs. post-annealed.

The 0.457% [Ga] measured at the 500°C-annealed GZO/ZnO interface can be used to

gauge the extent of outdiffusion to the interface at the 250°C GZO growth temperature with 1.5

minute elapsed growth time. Based simply on thermally activated diffusion with vacancy-

assisted diffusion and the lowest measured activation energy reported, 89 [Ga] is estimated to be

orders of magnitude lower than [VZn] at the GZO/ZnO interface. Hence, while both Ga and VZn

may contribute to charge scattering at the interface, VZn should have a much higher impact with

annealing.

5.4 Conclusions

In conclusion, these results show that differential DRCLS can resolve near-band-edge

(NBE) and VZn–related optical-emission profiles on a near-nm scale in GZO/ZnO structures grown by MBE at 250 C. Both the NBE and VZn profiles across the GZO/ZnO interface are basically abrupt on a nm scale; however, the abruptness is somewhat diminished with increasing

GZO thickness. Since GZO growth times are proportional to GZO film thickness, the broadened

VZn emission can be explained by a corresponding increase in the diffusion of VZn across the interface. There is also evidence of Ga diffusion, as determined by XPS profiling. A more pronounced broadening of both the VZn and Ga profiles occurs during a 10 min., 500 C anneal in Ar. However, even this high-temperature processing does not strongly degrade the electrical properties, suggesting that the Debye-tail mobility enhancement technology in ZnO is robust.

The ability of differential DRCLS to profile interfaces at a nm scale will be useful for more complex versions of this technology.

Chapter 6 Impact of Defect Distribution on IrOx/ZnO Interface Doping and

Schottky Barriers

6.1 Introduction

ZnO has emerged as a wide band gap semiconductor with a wide range of potential applications in micro- and optoelectronics due to its high free exciton binding energy (60 meV) for low threshold light emission,90 ease of chemical processing,91 availability of large area substrates,91 low cost,90 and biocompatibility.92 For device applications, considerable research has focused on material structures to form the highly rectifying and Ohmic contacts to ZnO needed.93, ,94 Previous work by Allen and Durbin showed that oxidized Ag produced Schottky barrier heights ΦB significantly higher than those obtained with pure metals, and with ideality factors closer to unity.95 Subsequent work demonstrated similar barrier increases for Ir, Pt, and

Pd oxidized metallic contacts.96,97 The intentional oxidation of Ag and other metals during deposition can be interpreted in terms of removal of a commonly observed surface hydroxide OH

98,99 100 layer as well as oxygen vacancies VO at and below the free surface. In the latter case, a remote oxygen plasma (ROP) strongly reduces VO densities at depths up to nearly 50 nm below the surface and avoids oxygen ion damage that could otherwise introduce new defects.98

Different chemical bond terminations on the Zn- versus O-polar surfaces (OH for Zn-polar, H for

- O-polar) as well as the introduction of O2 could also influence band bending and dipole formation.101 This early work also showed an I-V minimum displaced from zero bias for the

97 highest ΦB and lowest ideality factor η diodes, similar to those observed for very reactive

102 metals such as Ta on clean ZnO surfaces where Ta2O5 formation at the interface is evident.

Current- and capacitance-voltage I-V and C-V measurements show that IrOx/ZnO

Schottky barriers are consistently higher for Zn-polar versus O-polar surfaces.98 This is a systematic effect that has also been observed for several other noble metal oxide SCs to ZnO,

98 including PtOx and PdOx. Interestingly, this polarity effect is usually significantly stronger for oxidized SCs compared to their plain metal counterparts. This is because plain metal SCs are more susceptible to Fermi level pinning by interfacial oxygen vacancies, which act to limit the variation of SC barrier height with metal work function and surface polarity.103 The underlying mechanism(s) responsible for the Schottky barrier polarity effect, that has significant implications for the design of ZnO-based unipolar devices such as visible-blind UV Schottky photodiodes and transparent metal semiconductor field effect transistors, is not well understood and in this work we use DRCLS measurements on IrOx/ZnO SCs to propose a defect- segregation model that explains the variation of ZnO SC barrier height with crystallographic polarity. In turn, the extent to which these mechanisms affect ΦB will depend strongly on free carrier density and the resultant width of the surface space charge region. These factors can all be influenced by changes in carrier concentration within the band bending region of the semiconductor induced by a non-uniform distribution of electrically-active defects. Thus oxygen vacancies complexed with H can act as donors104 while zinc vacancies are acceptors that act to

105 compensate the otherwise n-type carriers. Optical signatures for VZn and VO can help identify native defects within the ZnO near its metal interface. From combined DRCLS and positron annihilation spectroscopy (PAS) measurements, luminescence from 1.6 to 2 eV is identified with

106 isolated and clustered VZn, whose energy increases with cluster size. This assignment is consistent with photoluminescence (PL), PAS, and Hall effect studies107 as well as optically

108 detected magnetic resonance (ODMR) studies. Assignment of VO-related luminescence is less well established with emission at 2.43 eV suggested by PL and electron paramagnetic resonance

(EPR) while closely spaced emission at 2.35 eV is associated with Cu impurities on Zn sites.109

Most recent EPR, ODMR, and PL studies indicate that the 2.45 eV emission is indeed associated

110,111 with VO-related defects. Single-crystal ZnO grown by vapor phase transport (VPT) by ZN

Technologies displayed deep levels corresponding to VO-related defects that increased at interfaces with metals that form strong oxygen bonds.100,102 Similarly, evaporated Au and Pd metals on ZnO surfaces displayed higher ΦB for Zn- vs. O-polar surfaces of chemomechanically- polished, low-defect, single crystal surfaces of VPT ZnO.112,113 Corresponding 1/C2-V measurements of carrier concentration versus depth showed lower carrier densities and higher barrier heights for Zn-face vs. O-face ZnO. For those crystals, VO was the dominant defect near the metal-ZnO interface, exhibiting higher densities for the O-face both with and without ROP treatment. Those studies also showed correlations of higher ΦB with lower carrier density within the surface space charge region.

Higher ΦB correlates with lower bulk carrier densities for crystals grown by different methods. Thus hydrothermal (HT) growth has 2 – 3 orders of magnitude lower carrier density than “melt-grown” ZnO due to compensating acceptor Li and Na centers in the hydrothermal case. Consequently, higher resistivity HT ZnO typically exhibits higher Schottky barriers than melt-grown material,114,115 while previous defect studies at metal–VPT ZnO interfaces correlated a reduction in segregated VO with increased ΦB. Here we report VZn as the dominant defect for metal oxides deposited on ZnO and find that VZn segregates more strongly to Zn vs. O-polar surfaces, resulting in lower free carrier densities and higher ΦB. Furthermore, this increase is

77 confined within the ZnO surface space charge region for both Zn- and O-polar interfaces. Indeed,

VZn density exhibits a pronounced decrease at the neutral edge of the space charge region, suggesting that surface space charge fields across the more insulating depletion region drive these defects toward the buried interface. In general, these results suggest that decreased carrier densities can account for the higher effective work function of these n-type metal-oxide contacts.

6.2 Experimental Setup

Arrays of 32 nm thick, 300 μm diameter IrOx Schottky contacts were deposited on Zn- polar and O-polar face samples obtained from opposite sides of the same 0.5 mm thick double- sided polished c-axis ZnO (low-Li Tokyo Denpa Ltd.) single crystal wafer, by the reactive 50 W

RF sputtering of an Ir target (purity 99.95 %) using an O2:Ar plasma (O2:Ar ratio = 10:7 by volume). This conventional hydrothermal ZnO wafer, containing the acceptor Li, was annealed

(post-growth) at 1400 C for 2 hrs, then ground and re-polished to remove the surface segregated impurities, in particular Li, making it significantly less compensated with a carrier density of ~

1017cm-3, as opposed to ~ 1014 cm-3 for unprocessed material. Rutherford backscattering spectroscopy (RBS) determined IrOx stoichiometry to be [Ir]:[O] =1:2.6 rather than 1:2 for

116 IrO2 . A 200 nm thick Ir capping layer web-patterned to permit electron beam excitation and optical access within the diode was deposited by e-beam evaporation with Ti/Au Ohmic half- rectangle contacts to bare ZnO bounding the diodes. See Figure 1 insets. I-V characteristics for these Zn- and O-polar interfaces showed 8.6 and 5.5 orders of magnitude rectification ratios, respectively. The diodes were characterized by current-voltage (IV) and capacitance-voltage

(CV) measurements, at room temperature and in dark conditions, using a HP 4155A parameter

78 analyzer and a Philips PM6304 RCL meter, respectively. The composition of reactively rf sputtered IrOx SCs for different Ar/O2 gas ratios has previously been established using

Rutherford Backscattering Spectrometry (RBS) on films deposited on glassy carbon substrates at the same time as the corresponding SCs 116. These measurements indicate that the IrOx

116 stoichiometry of the SCs in this work was [Ir]:[O] =1:2.6 rather than 1:2 for IrO2 . The composition of reactively rf sputtered IrOx SCs for different Ar:O2 gas ratios under the same sputtering conditions as used here was previously determined by Hyland et al. 116 using

Rutherford Backscattering Spectrometry with 2 MeV He2+ ions at Western Michigan University on IrOx films deposited on glassy carbon substrates at the same time as the corresponding SCs.

Using the reported IrOx composition for an O2:Ar sputtering gas ratio of 7:10 by volume, we infer that the IrOx stoichiometry of the SCs in this work was [Ir]:[O] =1:2.6 rather than 1:2 for

27 IrO2 . The composition of reactively rf sputtered IrOx SCs for different Ar:O2 gas ratios under the same sputtering conditions as used here was previously determined by Hyland et al. 27 using

Rutherford Backscattering Spectrometry with 2 MeV He2+ ions at Western Michigan University on IrOx films deposited on glassy carbon substrates at the same time as the corresponding SCs.

Using the reported IrOx composition for an O2:Ar sputtering gas ratio of 7:10 by volume, we infer that the IrOx stoichiometry of the SCs in this work was [Ir]:[O] =1:2.6 rather than 1:2 for

116 IrO2 . DRCLS measurements involved an incident electron beam to generate a cascade of secondary electrons and ultimately electron-hole (e-h) pairs that recombine, emitting luminescence corresponding to band-to-band and band-to- deep level transitions characteristic of native point defects. For electron beam energy EB increasing from 2.5 to 5 keV, Monte Carlo

117 distributions show that the Bohr-Bethe range of excitation RB increases from 40 to 140 nm.

DRCLS measurements were performed on a JEOL SEM microscope. All EB–dependent DRCL spectra were obtained at 80 K and with constant beam power of 0.41 µW for incident beam energies EB = 1 – 5 keV and proportional currents.

6.3 Results

All EB–dependent DRCL spectra were obtained at 80 K and with constant beam power.

The high stopping power of Ir required EB to be at least 2.5 keV in order to excite spectra in the

ZnO below the IrOx. For the bare ZnO surface adjoining the IrOx metal, EB = 1 keV excitation depths are comparable to EB = 2.5 keV depths through the IrOx

Besides the near band edge (NBE) emission at 3.35 eV, DRCL spectra exhibit sub-band gap features at 2.01 – 2.05 eV and 2.35 eV corresponding to VZn of varying cluster size and CuZn substitutional impurities. Figure 6.1(a) shows a pronounced 2.08 eV defect peak whose intensity decreases with increasing EB and depth of excitation for EB = 2.5, 3.0, and 3.5 eV. Above EB =

3.5 keV, I(2.08 eV) begins to increase. DRCL spectra with higher EB show steady increases of this intensity peak normalized to I(3.35 eV), indicative of a local minimum concentration at intermediate depth. Figure 6.1(b) shows nearly an order of magnitude lower I(2.01 eV)/I(3.35 eV) ratio from the surface of the ZnO close to the IrOx contact. The free surface VZn intensity is low enough so that an overlapping 2.33 eV due to CuZn defects is evident. Figure 1 shows that

IrOx deposition on a Zn-polar ZnO surface strongly increases VZn density near the IrOx interface.

Figure 6.1. (a) DRCL spectra, normalized to the NBE, for 32 nm IrOx on ZnO (0001) at 80 K, showing strong VZn segregation toward the IrOx/ZnO (0001) interface. Note the minimum in the defect emission at beam energy 3.5 keV. (b) EB = 1.0 keV, 80 K DRCL spectrum, normalized to the NBE, of bare Zn (0001) surface. Deconvolutions in both (a) and

(b) show both VZn and CuZn defects.

Figure 6.2. Deconvolved VZn (2.01-2.05 eV) and CuZn (2.35 eV) area depth profiles showing higher VZn and CuZn surface segregation toward the Zn- polar (0001) vs. O-polar (000ī) surface.

Similar defect features are evident for O-polar IrOx/ZnO interfaces although the corresponding VZn peak feature is broadened, extending from 1.95 to 2.15 eV (not shown).

Figure 6.2 shows the normalized A(2.01 eV) /A(3.35 eV) and A(2.35 eV)/A(3.35 eV) area ratios versus maximum excitation depth RB from Monte Carlo simulations for both Zn- and O-polar interfaces. Both VZn (0001) and VZn (000 ī) area ratios increase strongly toward the IrOx/ZnO interface with minima 60-80 nm below. Both VZn and CuZn are higher for ZnO (0001). Together,

VZn + CuZn acceptor areas are nearly 40% larger at the Zn- versus O-polar interface.

O-polar depletion width increases strongly toward the IrOx/ZnO interface at shallower depths than for Zn-polar.

Figure 6.3(a) shows I-V measured barrier height vs ideality factor that extrapolate to 0.67 and 0.89 eV, respectively, for the O- and Zn- polar diodes. Figure 6.3(b) shows carrier densities

2 versus depth obtained from 1/C –V measurements of four IrOx/Zn- and four O-polar ZnO diodes

1/2 where capacitance C = (A/2)[2qεn/(V0 – V)] , n = ND - NA is carrier density, ε is dielectric permittivity, and A is diode area so that

2 2 n = -2/[qεA (d(1/C )/dV)] (1) and corresponding depletion layer widths

1/2 W = [2ε(V0 – V)/ qn] (2)

For the IrOx/O-polar diodes, n increases toward the interface beginning close to W= 43 nm

17 -3 calculated from Eq.2 for 0.67 eV barrier height and n =3.5 x 10 cm . Similarly, IrOx/Zn-polar contacts display constant carrier density to within about 53 nm of the interface, somewhat lower than the 71 nm depletion width calculated for 0.89 eV barrier height and n = 1.5 x 1017 cm-3.

The lower VZn acceptor density and higher O- vs. Zn-polar ZnO carrier densities are consistent with a narrower depletion region. However, the mechanisms underlying this difference in defect density as well as the segregation itself are not immediately evident even though segregation of native point defects are common features of many metal/semiconductor

118 interfaces. For IrOx/ZnO interfaces, two possible mechanisms are: (i) the chemical diffusion and bonding of RF-activated oxygen from the growth process in the outer ZnO layers and (ii) the electric fields associated with ZnO’s spontaneous polarization and associated band bending. RF excited oxygen at the ZnO surface could promote near-surface Zn oxidation, promoting Zn diffusion toward the surface and VZn formation at depths below. However, polarity would not be a dominant factor. The electric field E associated with ZnO’s spontaneous polarization could also account for segregation and the contrast between polarities. Spontaneous polarization Psp in

ZnO produces electric fields along the c-axis that are equal and opposite for Zn- and O-polar

119 2 orientations. Ab-initio calculations yield Psp = - 0.05 C/m , consistent with experiment so that

2 Psp = ε0 (εr - 1) E = - 0.05 C/m (3)

6 and E = P/ ε0 (εr - 1) = 7.92 x 10 V/cm (4)

5 for εr = 8.75 along the c-axis. However, such high fields are reduced by the > 10 V/cm opposing field due to n-type band bending as well as any strain-induced piezoelectric fields.

Figure 6.4. DRCLS peak intensity ratio of 2.01-2.05 eV VZn and 2.35 eV CuZn peak in ZnO versus NBE intensity with and without 917 V positive bias applied across a 0.5 mm ZnO crystal.

The electron beam penetrates the 20 nm Pt layer to generate luminescence within the ZnO below.

The 18.3 kV/cm bias attracts negatively-charged CuZn and VZn acceptor defects toward the top electrode.

To gauge how electric fields can affect defect distributions in ZnO, we obtained DRCL spectra of native defects vs. depth under applied electric fields. Figure 6.4 illustrates the depth distribution of CuZn and VZn acceptors with applied bias. Without bias, DRCLS through a 20 nm

Pt electrode on a 0.5 mm thick ZnO single crystal showed significant increases in CuZn and VZn density from the bulk to the surface, analogous to Fig. 6.2. An applied bias of +917 V increased this segregation significantly, indicating electric field-induced diffusion of CuZn and VZn acceptors by an electric field gradient of 18.3 kV/cm. Similarly, applied bias between diode and

120 ohmic contact in Figure 6.1 produces analogous CuZn lateral segregation. Since both VZn and

CuZn are negatively charged acceptors, strong PSP fields in ZnO are expected to produce more

VZn segregation toward the interface of the Zn- vs. the O-polar junction.

6.4 Conclusions

These results show that DRCLS - measured width of VZn segregation agrees with the

121 previously measured TEM EELS width of depleted Zn at the IrOx/ZnO interface and the depletion widths derived from 1/C2-V Schottky barrier heights and bulk doping densities. The difference in VZn density at the Zn- vs. O-polar ZnO junctions can account for their difference in depletion widths and effective Schottky barrier heights. In turn, spontaneous polarization provides sufficiently high electric fields to drive diffusion of VZn and CuZn defects that accounts for the difference in segregation with ZnO polarity. Such electric field-induced diffusion of electrically-active defects may be significant at other metal-semiconductor interfaces.

Chapter 7 Defect-Controlled Ohmic, Blocking, and Schottky Contacts to ZnO

Micro-/Nanowires

7.1 Introduction

Metal contacts to semiconductor nanowires are integral to a wide range of device applications,122,123,124,125,126,127,128 yet they exhibit electronic properties that depend strongly on their physical size. For metal interfaces to bulk semiconductors, the nature of charge transfer and barrier formation depend sensitively on the chemical, geometrical, and electronic structure at their intimate junction.129,130 For metal interfaces to semiconductor nanowires, the scale and physical geometry of these junctions introduces additional and unique features including three- dimensional depletion regions and space charge limited injection.123,131 The high surface-to- volume nature of these nano- and microwires amplifies these effects of local band bending, the resultant free carrier depletion within wires,132 and their dependence on ambient effects.133,134,135,136 Electrically-active defects at surfaces also can control carrier transport lengthwise inside nanowires by adsorbate-induced charging that alters depletion regions radially.

These defects are present inside as well as at the surfaces of nano- and microwires.137 Their presence has been inferred from electrical138,139,140,141,142 and optical143,144 measurements of devices and their interfaces, but direct localized measurements correlating them with electronic contact behavior have not until now been available. We have now measured these defects inside

ZnO nano-/microwires and at their metal interfaces to show how the physical nature, density, and spatial distribution of these defects dominates the Schottky, ohmic, or blocking behavior of these contacts.

7.2 Experimental Setup & Results

We used depth-resolved cathodoluminescence spectroscopy (DRCLS)145 to measure defect type, density, and spatial distributions, inside the “bulk” of single nano-/microwires and their interfaces with the same Pt metal, whose Schottky and ohmic contacts have been studied

146,147,148,149 previously. Using EB = 0.5 – 5 keV incident beam energies, one can achieve near-nm depth resolution through bare surfaces as well as through thin metal contact layers.150 Monte

Carlo simulations provided profiles of energy-dependent excitation depths in good agreement

151 with experimental measurements. For bare ZnO surfaces and EB = 0.5 - 3 keV, peak electron- hole pair creation rates, ranged from U0 = 1.5 to 22 nm with maximum excitation (Bohr-Bethe) range RB = 18 – 76 nm, respectively. Simulations show that these depths are reduced significantly depending on thickness and atomic density for excitation through metal overlayers.

Nevertheless, excitation through 20-30 nm thick metal overlayers produces luminescence at the semiconductor interface that can pass back through the metal and be collected by an optical train.

DRCLS coupled with the 5-10 nm incident beam spot size of our UHV-scanning electron microscope (SEM) provided near-nm scale spatial resolution in three dimensions.

The ability to measure spatially-localized luminescence inside ZnO nanowires and at their metal interfaces enabled us to address several questions: (i) what is the nature and distribution of native point defects at interfaces of nanostructures; (ii) what effect do these defects have on interface electronic properties; and (iii) can the nature and spatial distribution of these defects be controlled. To address all three issues, we performed DRCLS and current- voltage (I-V) measurements on ZnO nanowires and microwires of varying diameters and with Pt contacts to the same nano/microwire. Figure 7.1 illustrates a ZnO nanowire grown by a carbo-

89 thermal technique152 via evaporation of pressed targets consisting of ZnO and carbon powder in equal parts which were heated to 950 °C under ambient air conditions in an open tube furnace.

ZnO wires of over 1:100 diameter to length ratio were then collected by dragging lint-free cleanroom paper across the sapphire surface and casting them onto 200 nm thick SiO2 on Si substrates. Pt contacts to individual wires were patterned by electron beam induced deposition

(EBID), using the rastered electron beam of an FEI Helios Nanolab 600 Focused Ion Beam (FIB) scanning electron microscope (SEM) to decompose organometallic vapor injected through a hollow needle inserted 50 nm from the wire to pattern Pt pads and contacts to the wire. Figure 1a shows the pattern of Pt wire contacts across a 60 μm long nanowire whose diameter varied linearly from 400 nm to 1 μm. The five Pt contacts were each spaced ~ 10 μm apart with contact area dimensions of 2 μm wide by wire diameter d and contact thickness 300 nm. Each contact was connected to a larger square pad for I-V wire contacts. Each of the contact areas was implanted using the FIB with a 4 μm wide 1016 cm-2 dose of Ga at 30 keV prior to Pt deposition to reduce contact resistance. Stopping and Range of Ions in Matter (SRIM) simulations demonstrate that the 30 keV Ga ions have a projected range Rp of 14.7 nm. Nevertheless, these five contacts exhibited a wide range of contact behavior. Figure 1b exhibits nearly ohmic I-V behavior between Contacts 1 and 3, indicating that both contacts are ohmic in nature. Figure 1c exhibits Schottky behavior with reverse currents decreasing by nearly two orders of magnitude with ±10 V bias. Note the zero bias voltage offset characteristic of a built-in voltage at an insulating interface.153

(a) (b) (c)

Figure 7.1. (a) Tapered ZnO nanowire with five Pt contact and wire pads. (b) I-V characteristics

under in air and in dark between Contacts 1 and 3 versus (c) 3 and 4. Ohmic behavior between

Contacts 1 and 3 versus Schottky behavior between Contacts 3 and 4 shows that Contacts 1, 3,

and 4 were ohmic, ohmic, and rectifying, respectively. Contact 5 was blocking.

Figure 7.2a shows DRCL spectra obtained at 80 K of the bare ZnO nanowire midway

between Contacts 3 and 4. Near band edge (NBE) features at 3.36, 3.312, 3.276, and 3.199 eV

correspond to free exciton, donor-bound exciton, and two phonon replicas, respectively.154 With

deconvolution of the mid-gap structures (see Fig. 7.3), defect features appear at energies below

the NBE peaks corresponding to VZn, VZn clusters (VZn-R), CuZn and VO at 1.8, ~2, 2.35, and 2.5

eV, respectively. Figure 7.2b’ shows that the integrated intensities of these peaks vary with

excitation depth. The 2.34 eV CuZn defect intensity decreases from bulk values to ~ 50 nm, then

increases by > 2x from ~50 nm toward the free surface. The VZn intensity also increases toward

the surface from similar depths, while VZn-R and VO intensities exhibit little or no depth

dependence. CuZn segregation is measured for several different wire diameters. Based on a

155 previous NBE-normalized calibration, near-surface VZn densities are estimated to be ~ 1.2

21 -3 x10 cm . However, calibration of CuZn densities is not yet available

700, and 1,000 nm diameter ZnO wires. (d’) I(2.35 eV) CuZn defect intensities versus nanowire diameter for surface (1.5 keV), sub-surface (3.5 keV), and “bulk” (5 keV) probe depths showing increasing near-surface segregation with increasing wire diameter. (e’) (a) SEI, (b) HSI, and (c)

+ EB = 5keV CL spectra of 5 keV Ga milled removal of segregated defects in outer annulus versus e-beam annealed defect increase in contiguous regions of a 200 nm ZnO wire on SiO2.

HSI color gradient signifies magnitude of integrated 1.5 – 2.75 eV defect areas normalized by

NBE integrated area. Defects increase from milled wire section toward annealed region.

Figure 7.2

The CuZn defect density exhibits not only a dependence on depth but also on nanowire diameter. CL spectra for the bare ZnO surface in Fig. 7.2c’ show that CuZn peak intensity normalized to the NBE intensity, I(2.35 eV)/I(NBE), for EB = 3.5 keV increases with increasing diameter. In Fig. 7.2d’, this ratio also increases linearly for EB = 1.5 and 5 keV. Furthermore, comparison of the EB = 1.5, 3.5, and 5 keV variations with depth reflect the same dip near RB =

50 nm excitation depth.

Because these CuZn and VZn defects segregate toward the surface, it is possible to remove these accumulated defects by removing the outer annulus of the nanowire. Figure 7.2e’(a) shows an SEI of the 700 nm diameter wire reduced to ~400 nm after 5 kV Ga+ milling. The corresponding hyperspectral image of this wire in Fig. 7.2e’(b) shows that the integrated 1.5 –

2.75 eV defect area decreases at the milled portion of the wire. In contrast, electron beam heating increases defect densities in an adjoining length of wire. The EB = 5 keV CL spectra in Fig.

7.2e’(c) confirm this order of magnitude defect decrease due to milling and increase due to local heating. Note that the 5 keV Ga ions have a projected range Rp of only 4.4 nm so that any new defects introduced by the Ar+ milling are minimal. Figure 7.2 illustrates how native point defects in nanowires can vary spatially both radially and lengthwise either as they are grown or as they are locally milled or heated.

Figure 7.3. The above show the DRCLS spectra and defect profile for the base ZnO with no Ga processing between contacts 1 & 2 and between contacts 2 & 3.

Figure 7.4 compares CL spectra of Contacts 3 versus 4 to show how local differences in defects at their interfaces can account for the ohmic versus Schottky I-V characteristics, respectively, in Figure 7.1. Figs. 7.4a and 7.4b show NBE and defect features similar to Fig.

7.2a but with individual defect contributions deconvolved. Instead of A(CuZn)/A(NBE) = 6.6 near the free surface between Contacts 3 and 4, A(CuZn)/A(NBE) is nearly 4x higher at Contact 3 versus Contact 4. These differences can be attributed to the monotonic decrease in bulk and segregated CuZn with decreasing wire diameter. Pronounced minima in both CuZn and VZn depth profiles are evident in Figs. 7.4c and 7.4d, similar to Fig. 7.2b’ between these contacts. Similar minima in CuZn and VZn were recently reported for metallic IrOx contacts to bulk ZnO Zn- and

O-polar crystals, which were interpreted as electric field-driven diffusion of these acceptor defects within the surface space charge region of the contact. 156

Figure 7.4. DRCL spectra at EB = 5.0 keV immediately adjacent to (a) Contact 3 and (b) Contact

4 with the corresponding Gaussian deconvolutions. The 2.35 eV CuZn defect decreases significantly at Contact 4. Depth profiles of deconvolved defect areas for (c) Contact 3 and (d)

Contact 4, both showing pronounced minima at ca. 50 nm depth and surface segregation, nearly

4x stronger for Contact 3.

The densities and spatial distributions of VZn and CuZn defects inside the ZnO nanowire and their monotonic decrease with decreasing wire diameter can explain the striking difference in their measured contact properties. The ohmic I-V characteristics of Contacts 1 and 3 in Fig.

7.1b can be interpreted as trap-assisted hopping transport since the spatial separation of these defects is comparable to their wave function overlap. Assuming a hydrogenic model,157 the wave

½ function extent a for the VZn trap can be expressed as a = ħ / (2m*Et) where the ZnO effective mass m* = 0.30 m0 so that for trap depth Et = 3.365 – 2.54 = 0.83 eV below the conduction band, a = 3.90 Å. For the CuZn trap with Et = 3.365 – 2.35 = 1.015 eV, a = 3.53 Å. Based on the

34 21 previous calibration for VZn, and the normalized VZn amplitude in Fig.2, [VZn] = 1.17 x 10

-3 cm midway between Contacts 3 and 4. Assuming that [CuZn] and [VZn] densities are

21 -3 comparable, then their combined density is Nt = 2.3 x 10 cm and their spatial separation R is

3 obtained from 4π/3·R = (1/Nt) or R = 4.7 Å. Hence, the spatial separation of trap states is comparable to the wave function extent, suggesting significant wave function overlap and trap- assisted hopping. Note that [CuZn] is even higher for Contacts 1 and 3 based on Fig. 7.2d’. The

Schottky barrier I-V characteristic for Contact 4 appears to reflect the nearly 4x (2x) decrease in segregated CuZn (VZn) in Figs. 7.3c and 7.3d, which translates to a 60% increase in spatial separation. Note that residual hopping may contribute to the relatively low Schottky barrier ΦSB

= 0.35 eV compared with literature values158 as well as carbon incorporation in the EBID Pt contacts. Finally, the blocking contact measured for Contact 5 is consistent with the depletion layer width now comparable or greater than the nanowire radius.132 Previous resistivity measurements of similar ZnO nanowires suggest nominally undoped carrier densities n = 1 – 3 x

17 -3 135 10 cm . With compensating VZn and CuZn acceptors reported here, resultant carrier density N

16 -3 should decrease to 10 cm . Together with a V0 = 0.35 eV barrier height and dielectric constant

½ ε = 8.75 mo, the Pt-ZnO depletion width W = [2εV0/qN] = 206 nm so that depletion regions

extending radially from multiple facetted surfaces can almost fully deplete the 400 nm diameter

nanowire. Figure 7.5 illustrates cross sections of the wire at diameters corresponding to Contacts

1, 4, and 5 and their corresponding interface bands and charge transport. Figure 7.5a depicts trap-

assisted transport through the high defect density band bending region. Figure 7.5b shows both

thermionic emission and trap-assisted tunneling, while Figure 7.5c illustrates band bending

regions extending from opposite faces that fully deplete the nanowire interior.

Figure 7.5. Schematic diagrams of band bending at Pt-ZnO nano/microwire contact for (a) 900 nm, (b)

600 nm, and (c) 400 nm diameter wires linked to the interfaces of their corresponding wires. Lighter shading signifies higher acceptor density and lower electron density with increasing radius. With decreasing diameter, interface acceptor density decreases and contact behavior changes from transport by

(a) trap-assisted tunneling to (b) Schottky rectification, to (c) blocking.

100

7.3 Conclusions

In summary, these results illustrate how native point defects in nanowires can vary spatially both radially and lengthwise either as they are grown or as they are locally milled or heated. As previously reported, these defects can be present throughout the nanowires with densities and nature that depend on growth method. As shown here, these spatial variations are different for different defects within the same ZnO nanostructure. The electrical contact properties of a single ZnO nanowire grown by a carbo-thermal method are found to depend on the density and spatial distribution of CuZn and VZn acceptor defects inside nano- and microwires measured by DRCLS, resulting in ohmic, Schottky, or blocking contacts for the same metal on the same ZnO nanowire.

101

Chapter 8 Future Work

8.1 Conclusions

The above work has shown the rather large impact that native point defects have in charge transport and Schottky contact formation. We saw combined DRCLS and SPS studies of non-polar, single-phase MgxZn1-xO (0

Our next results show that differential DRCLS can resolve near-band-edge (NBE) and

VZn–related optical-emission profiles on a near-nm scale in GZO/ZnO structures. Both the NBE and VZn profiles across the GZO/ZnO interface are basically abrupt on a nm scale; however, the abruptness is somewhat diminished with increasing GZO thickness. Since GZO growth times are proportional to GZO film thickness, the broadened VZn emission can be explained by a corresponding increase in the diffusion of VZn across the interface. There is also evidence of Ga diffusion, as determined by XPS profiling. A more pronounced broadening of both the VZn and

Ga profiles occurs during a 10 min., 500 C anneal in Ar. However, even this high-temperature processing does not strongly degrade the electrical properties, suggesting that the Debye-tail mobility enhancement technology in ZnO is robust. The ability of differential DRCLS to profile interfaces at a nm scale will be useful for more complex versions of this technology.

102

For the IrOx/ZnO interface, results show that DRCLS - measured width of VZn segregation agrees with the previously measured TEM EELS width of depleted Zn at the

2 IrOx/ZnO interface and the depletion widths derived from 1/C -V Schottky barrier heights and bulk doping densities. The difference in VZn density at the Zn- vs. O-polar ZnO junctions can account for their difference in depletion widths and effective Schottky barrier heights. In turn, spontaneous polarization provides sufficiently high electric fields to drive diffusion of VZn and

CuZn defects that accounts for the difference in segregation with ZnO polarity. Such electric field-induced diffusion of electrically-active defects may be significant at other metal- semiconductor interfaces.

In ZnO nanowires our results illustrate how native point defects in nanowires can vary spatially both radially and lengthwise either as they are grown or as they are locally milled or heated. These defects can be present throughout the nanowires with densities and nature that depend on growth method. As shown here, these spatial variations are different for different defects within the same ZnO nanostructure. The electrical contact properties of a single ZnO nanowire grown by a carbo-thermal method are found to depend on the density and spatial distribution of CuZn and VZn acceptor defects inside nano- and microwires measured by DRCLS, resulting in Ohmic, Schottky, or blocking contacts for the same metal on the same ZnO nanowire.

All of these results illustrate how DRCLS is a useful technique for characterizing these native point defects and the power of this characterization tool is expanded when combined with other processing techniques. Combined with FIB/EBID and I-V measurements, contacts can be

103 grown and characterized without ever breaking vacuum and exposing the sample to the atmosphere.

8.2 Future Work

The next step in extending this work is to test environmental factors on the electrical properties of the nanowires. To do this the electrical measurements can be conducted in the presence of atmosphere, O2, N2, and high-water vapor. Also, mapping the change in electrical properties as a function of temperature ranging 79 - 300K. Under different ambient conditions, the defects may shift, as field effects change. We can apply a large electric bias and then measure the defects as they shift due to field effects using DRCLS. A further technique is hyper spectral imaging of the nanowires. The benefit of this measurement in a nanowire, over that of a bulk film, is the defects are then only allowed to move in one dimension. Furthermore, a geometric map to be drawn over the wire instead of single point spectra. This allows for a time dependent measurement to be taken, allowing for a time scale to be mapped to the change of defects across the wire. The relationship between defect distribution and applied field must be established before any practical application as a FET can be used.

104

References

1 K. Hummer, Phys. Status Solidi B 56, 249 (1973). 2 D. C. Look, Mater. Sci. Eng.: B 80, 383 (2001) 3 S. J. Pearton, D. P. Norton, L. Ip, Y.W. Heo, and T. Steiner, Prog. Mat. Sci. 50, 293 (2005). 4 Zhong Lin Wang, Materials Today, 26 (2004) 5 Lukas Schmidt-Mend and Judith L. MacManus-Driscoll, Materials Today 10, 40 (2007) 6 Zhong Lin Wang and Jinhui Song, Science 312, 242-245 (2006) 7 Anderson Janotti and C.G. Van de Walle, Phys. Rev. B 76, 165202 (2007) 8 S. Lany and A. Zunger, Phys. Rev. B 78, 235104 (2008) 9 R. Dingle, Phys. Rev. Lett. 23, 579 (1969) 10 Ya. I. Alivov, M.V. Chukichev, and V.A. Nikitenko, Semiconductors 38, 34-38 (2004) 11 K. Vanheusden, W.L. Warran, C.H. Seager, D.R. Tallant, J.A. Voigt, and B.E. Gnade, J. Appl. Phys. 79, 7983 (1996) 12 F.H. Leiter, H.R. Alves, A. Hofstaetter, D.M. Hoffmann, and B.K. Meyer, Phys. Status Solidi B 226, 201 (2003) 13 K. Vanheusden, C.H. Seager, W.L. Warren, D.R. Tallant, and J.A. Voigt, Appl. Phys. Lett. 68, 403 (1996) 14 L.J. Brillson, H.L. Mosbacker, M.J. Hetzer, Y. Strzhemechny, G.H. Jessen, D.C. Look, G. Cantwell, J. Zhang, and J.J. Song, Appl. Phys. Lett. 90, 102116 (2007) 15 Y. Dong, F. Tuomisto, B.G. Svensson, A. Yu. Kuznetsov, and L.J. Brillson, Phys. Rev. B 81, 081201(R) (2010) 16 K.E. Knutsen, A. Galeckas, A. Zubiaga, F. Tuomisto, G.C. Farlow, B.G. Svensson, and A. Yu. Kuznetsov, Phys. Rev. B 86, 121203(R) (2012) 17 Z.L.Wang, Mater. Sci. Eng. R 2009, 64, 33. 18 W.L. Hughes, Z.L.Wang, Appl. Phys.Lett. 2003, 82, 2886. 19 M.S. Amold, Ph. Avouris, Z.W. Pan, Z.L.Wang , J. Phys. Chem. B 2003, 107, 659. 20 L.J. Brillson and Y. Lu, J. Appl. Phys. 2011, 109, 121301.

105

21 B. Bercu, W. Geng, O. Simonetti, S. Kostcheev, C. Sartel, V. Sallet, G. Lérondel, M. Molinari, L. Giraudet, and C. Couteau, Nanotechnology, 2013, 24, 415202. 22 T. Nobi, E. Kaidashev, A. Rahm, M. Lornz, J. Lenzner, M. Grundmann, Nano Lett. 2004, 4, 797. 23 W.T. Ruane, K. Leedy, D.C. Look, G. Farlow, H. von Wenckstern, M. Grundmann, L.J. Brillson, Nanoscale 2016, 8, 7631. 24 J.-H. He, J.-J. Ke, P.-H. Chang, K.-T. Tsai, P.C. Yang, and I.-M. Chan, Nanoscale, 2012, 4, 3399. 25 B.G. Yacobi and D.B. Holt, J. Appl. Phys. 59, R1 (1986) 26 P. Hovington, D. Drouin, and R. Gauvin, Scanning 19, 1 (1997). 27 L.J. Brillson, Surfaces and Interfaces of Electronic Materials, (Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim, (2010)) pp. 289. 28 Y.M. Strzhemechny, H.L. Mosbacker, S.H. Goss, D.C. Look, D.C. Reynolds, C.W. Litton, N.Y. Garces, N.C. Giles, L.E. Halliburton, S. Niki, and L.J. Brillson, J. Elect. Mat. 34, 399-403 (2005) 29 http://www.nt.ntnu.no/users/sraaen/dif4903/elspec.pdf 30 M.P. Seah and W.A. Dench, Surface and Interface Analysis 1, 1-11 (1979) 31 Reyntjens, S, and R Puers. Journal of Micromechanics and Microengineering 11 (4): 287– 300. (2001) 32 Bardeen, J. J. Appl. Phys.,71 717. (1947) 33 L.J. Brillson, Sur. Sci. Rep. 2, 123-326 (1982) 34 U. Ozgür, Y. I. Alivov, L. A. Teke, M. A. Reshchikov, S. Do˘gan, V. Avrutin, S. J. Cho, and H. Morkoc, J. Appl. Phys. 98, 041301 (2005). 35 K. Koike, K. Hama, I. Nakashima, G.-Y. Takada, M. Ozaki, K.-I. Ogata, S. Sasa, M. Inoue, and M. Yano, Jpn. J. Appl. Phys., Part 1 43(10B), L1372 (2004). 36 S. Sasa, T. Maitani, Y. Furuya, T. Amano, K. Koike, M. Yano, and M. Inoue, Phys. Status Solidi A 208, 449 (2011). 37 A. Ohtomo, K. Tamura, M. Kawasaki, T. Makino, Y. Segawa, Z. K. Tang, G. K. L.Wong, Y. Matsumoto, and H. Koinuma, Appl. Phys. Lett. 77, 2204 (2000). 106

38 P. Tao, Q. Feng, J. Jiang, H. Zhao, R. Xu, S. Liu, M. Li, J. Sun, and Z. Song, Chem. Phys. Lett. 522, 92 (2012). 39 D. C. Look, Mater. Sci. Eng.: B 80, 383 (2001). 40 S. J. Pearton, D. P. Norton, L. Ip, Y. W. Heo, and T. Steiner, Prog. Mat. Sci. 50, 293 (2005). 41 L.J. Brillson and Y. Lu, Appl. Phys. Rev. 109, 121301 (2011). 42 Y. Dong, Z.-Q. Fang, D. C. Look, G. Cantwell, J. Zhang, J. J. Song, and L. J. Brillson, Appl. Phys. Lett. 93, 072111 (2008). 43 Z. Zhang, V. Quemener, C.-H. Lin, B. G. Svensson, L. J. Brillson, Appl. Phys. Lett. 103, 072107 (2013). 44 T. A. Merz, D.R. Doutt, T. Bolton, Y. Dong, and L.J. Brillson, Surf. Sci. Lett. 605, L20 (2011). 45 E. Gür, G. Tabares, A. Arehart, J.M. Chauveau, A. Hierro, S.A. Ringel, J. Appl. Phys.112,123709 (2012). 46 J.-M. Chauveau, J. Vives, J. Zuniga-Perez, M. Laugt, M. Teisseire, C. Deparis, C. Morhain, and B. Vinter, Appl. Phys. Lett. 93, 231911 (2008). 47 P. Muret, D. Tainoff, C. Morhain, and J.-M. Chauveau, Appl. Phys. Lett. 101, 122104 (2012). 48 G. Tabares, A. Hierro, B. Vinter, and J.-M. Chauveau, Appl. Phys. Lett. 99, 071108 (2011). 49 A. Redondo-Cubero, A. Hierro, J.-M. Chauveau, K. Lorenz, G. Tabares, N. Franco, E. Alves, and E. Munoz, Cryst. Eng. Comm.14, 1637 (2012). 50 P. Hovington, D. Drouin, and R. Gauvin, Scanning 19, 1 (1997). 51 L. J. Brillson, J. Vac. Sci. Technol. A 6, 1437 (1988). 52 L. J. Brillson, J. Vac. Sci. Technol. B 19, 1762-1768 (2001). 53 L. J. Brillson, J. Phys. D: Appl. Phys. 45, 183001 (2012). 54 L.J. Brillson, H.L. Mosbacker, D. Doutt, M. Kramer, Z.L. Fang, D.C. Look, G. Cantwell, J. Zhang, and J.J. Song, Superlattices and Microstructures 45, 206 (2009). 55 Zhang, V. Quemener, C.-H. Lin, B. G. Svensson, and L. J. Brillson, Appl. Phys. Lett. 103, 072107 (2013). 56 C.-.H. Lin, E. J. Katz, J. Qui, Z. Zhang, U. K. Mishra, L. Cao, and L. J. Brillson, Appl. Phys. Lett. 103, 162106 (2013). 107

57 E. Gür, G. Tabares, A. Arehart, J. M. Chauveau, A. Hierro, and S. A. Ringel, J. Appl. Phys. 112, 123709 (2012). 58 A.F. Kohan, G. Ceder, D. Morgan, and C.G. Van de Walle, Phys. Rev. B 61, 15019 (2000). 59 F.Oba, A. Togo, I. Tanaka, J. Paier, and G. Kresse, Phys. Rev. B 77, 245202 (2008). 60 A. Janotti and C. G. Van de walle, Rep. Prog. Phys. 72, 126501 (2009). 61 Y. Dong, F. Tuomisto, B. G. Svensson, A. Y. Kuznetsov, and L. J. Brillson, Phys. Rev. B 81, 081201(R) (2010). 62 T. M. Børseth, F. Tuomisto, J. S. Christensen, E. V. Monakhov, B. G. Svensson, and A. Y. Kuznetsov, Phys. Rev. B 77,045204 (2008). 63 K. Vanheusden, C. H. Seager, W. L. Warren, D. R. Tallant, and J. A. Voigt, Appl. Phys. Lett. 68, 403 (1996). 64 R. Schmidt-Grund, D. Fritsch, M. Schubert, B. Rheinländer, H. Schmidt, H. Hochmuth, M. Lorenz, C.M. Herzinger, M. Grundmann, AIP Conf. Pro. 772, 201 (2005). 65 M. W. Williams and E. T. Arakawa, J. Appl. Phys. 38, 5272 (1967). 66 A.S. Rao and R.J. Keamey, Phys. Status Solidi (b) 95, 243 (1979). 67 T. Minemoto, T. Negami, S. Nishiwaki, H. Takakura, and Y. Hamakawa, Thin Solid Films 372, 173 (2000). 68 K. Jacobi, G. Zwicker, and A. Gutmann, Surf. Sci. 141, 109 (1984). 69 O. Kubaschewski, C.B. Alcock, and P.J. Spencer, Materials Thermochemistry, 6th ed. (Pergamon, New York, 1993). 70 D.A. Freedman, D. Roundy, and T.A. Arias, Phys. Rev. B 80, 064108 (2009). 71 C.M. Brooks, L. Fitting Kourkoutis, T. Heeg, J. Schubert, D.A. Muller, and D.G. Schlom, Appl. Phys. Lett. 94, 162905 (2009). 72 W.Windl, O. Restrepo, M. Ball, J. Perkins G. M. Foster, M. Myer, S. Mehra, J.M. Chauveau, A. Hierro, A. Redondo-Cubero,and L. J. Brillson, “Modeling-Assisted Measurement of

Chemical Potentials in MgxZn1-xO” (unpublished). 73 M.A. Khan, A. Bhattarai, J. N. Kuznia, and D.T. Olson, “High electron mobility transistor based on a GaN-AlxGa1-xN heterojunction,” Appl. Phys. Lett. 63, 1214 (1993).

108

74 T. Palacios and U.K. Mishra, “AlGaN/GaN High Electron Mobility Transistors,” in Nitride Semiconductor Devices: Principles and Simulation, ed. J. Piprek (Wiley-VCH Verlag GmbH & Co., KGaA, Weinheim, Germany, 2007). 75 D.C. Tsui, H.L. Stormer, and A.C. Gossard, “Two-Dimensional Magnetotransport in the Extreme Quantum Limit,” Phys. Rev. Lett. 48, 1559 (1982). 76 D.C. Look, E.R. Heller, Y.-F. Yao, and C.C. Yang, “Significant mobility enhancement in extremely thin highly doped ZnO films,” Appl. Phys. Lett. 106, 152102 (2015). 77 D.C. Look, K.D. Leedy, L. Vines, B.G. Svensson, A. Zubiaga, F. Tuomisto, D.R. Doutt, and L.J. Brillson, “Self-compensation in semiconductors: the Zn-vacancy in Ga-doped ZnO,” Phys. Rev. B 84, 115202 (2011). 78 J. Perkins, G. M. Foster, M. Myer, S. Mehra, J. M. Chauveau, A. Hierro, A. Redondo-Cubero,

W. Windl and L. J. Brillson, “Impact of Mg content on native point defects in MgxZn1-xO (0≤x≤ 0.56), APL Mat. 3, 062801 (2015). 79 G. M. Foster, J. Perkins, M. Myer, S. Mehra, J. M. Chauveau, A. Hierro, A. Redondo-Cubero, W. Windl and L. J. Brillson,” Native point defect energies, densities, and electrostatic repulsion across (Mg,Zn)O alloys,” Phys. Status Solidi A, 212, 1448-1454 (2015). 80 A.F. Kohan, G. Ceder, D. Morgan, and C.G. Van de Walle, “First-principles study of native point defects in ZnO,” Phys. Rev. B 61, 15019 (2000). 81 Y. Dong, F. Tuomisto, B.G. Svensson, A. Yu. Kuznetsov, and L.J. Brillson, “Vacancy Defect and Defect Cluster Energetics In Ion-implanted ZnO,” Phys. Rev. B, Rapid Communications 81, 081201(R) (2010). 82 K. Vanheusden, C.H. Seager, W.L. Warren, D.R. Tallant, and J. A. Voight, “Correlation between photoluminescence and oxygen vacancies in ZnO phosphors,” Appl. Phys. Lett. 68, 403 (1996). 83 H.L. Mosbacker, C. Zgrabik, A. Swain, S. El Hage, M. Hetzer, D.C.Look, G.Cantwell, J. Zhang, J.J.Song, and L.J. Brillson, “Thermally Driven Defect Formation and Blocking Layers at Metal-ZnO Interfaces,” Appl. Phys. Lett. 91, 072102 (2007). 84 T.J. Asel, H. Gao, T.J. Heinl, D. Adkins, P.M. Woodward, J. Hoffman, A. Bhattacharya, and L. J. Brillson, “Near-Nanoscale-Resolved Energy Band Structure of LaNiO- 109

3/La2/3Sr1/3MnO3/SrTiO3 Heterostructures and their Interfaces,” J. Vac. Sci. Technol. B 33, 04E103, (2015). 85 E. Burstein, “Anomalous Optical Absorption Limit in InSb,” Phys. Rev. 93, 632 (1954). 86 D.R. Doutt, S. Balaz, L. Isabella, D.C. Look, K.D. Leedy and L.J. Brillson, “Role of native point defects and Ga diffusion on electrical properties of degenerate Ga-doped ZnO”, Phys. Status Solidi B 250, 2114 (2013). 87 D. Drouin (2006). CASINO a Powerful Simulation Tool for Cathodoluminescence Applications. Microscopy and Microanalysis, 12 (Suppl. 02), pp 1512-1513. 88 See, for example, S.T. Bradley, L.J. Brillson, J. Hwang, and W.J. Schaff, “Surface Cleaning and Annealing Effects on Ni/AlGaN Interface Atomic Composition and Schottky Barrier Height,” Appl. Phys. Lett. 85, 1368-1370 (2004). 89 T. Nakagawa, I. Sakaguchi, M. Uematsu, Y. Sato, N. Ohashi, H. Haneda, and Y. Ikuhara, "Analysis of Indium Diffusion Profiles Based on the Fermi-Level Effect in Single-Crystal Zinc Oxide" Jpn. J. Appl. Phys. 46, 4099 (2007). 90. Ü. Özgür, Ya.I. Alivov, C. Liu, A. Teke, M.A. Reshchikov, S. Doğan, V. Avrutin, S.- J. Cho,

H. Morkoc, “A comprehensive review of ZnO materials and devices,” J. Appl. Phys. 98, 041301

(2005).

91. E. Ohshima, H. Ogino, I. Niikura, K. Maeda, M. Sato, M. Ito, and F. Fukuda, “Growth of the

2-in-size bulk ZnO single crystals by the hydrothermal method,” J. Cryst. Growth 260, 116

(2004).

92. Z.Li, R. Yang, M. Yu, F. Bai, C. Li, and Z.L. Wang, “Cellular Level Biocompatibility and

Biosafety of ZnO Nanowires,” J. Phys. Chem. C Lett. 112, 21004 (2008).

93. L.J. Brillson and Y. Lu, “ZnO Schottky Barriers and Ohmic Contacts,” Appl. Phys. Rev./ J.

Appl. Phys. 109, 121301 (2011).

110

94 C. Jagadish and S.J. Pearton, “Zinc oxide bulk, thin films, and nanostructures: processing, properties, and applications,” (Elsevier, 2011, Oxford, UK).

95. M.W. Allen, S.M. Durbin, and J.B. Metson, “Silver oxide Schottky contacts on n-type ZnO,”

Appl. Phys. Lett. 91, 053512 (2007).

96. M.W. Allen, M.M. Alkaisi, and S.M. Durbin, “Metal Schottky diodes on Zn-polar and O- polar bulk ZnO,” Appl. Phys. Lett. 89, 103520 (2006).

97. M.W. Allen, R.J. Mendelsberg, R.J. Reeves, and S.M. Durbin, “Oxidized noble metal

Schottky contacts to n-type ZnO,” Appl. Phys. Lett. 94, 103508 (2009).

98. K. Ip, B. P. Gila, A. H. Onstine, E. S. Lambers, Y. W. Heo, K. H. Baik, D.P. Norton, and S. J.

Pearton, “Improved Pt/Au and W/Pt/Au Schottky contacts on n-type ZnO using ozone cleaning,”Appl. Phys. Lett. 84, 5133 (2004).

99. B. J. Coppa, R. F. Davis, and R. J. Nemanich, “Gold Schottky contacts on oxygen plasma- treated, n-type ZnO (000ī),” Appl. Phys. Lett. 82, 400 (2003).

100. H.L. Mosbacker, Y.M. Strzhemechny, B.D. White, P.E. Smith, D.C. Look, D.C. Reynolds,

C.W. Litton, and L.J. Brillson, “Role of Near-Surface States in Ohmic-Schottky Conversion of

Au Contacts to ZnO,” Appl. Phys. Lett. 87, 012102 (2005).

101. M.W. Allen, D.Y. Zemlyanov, G.I. N. Waterhouse, J.B. Metson, T.D. Veal, C.F.

McConville, and S.M. Durbin, “Polarity effects in the x-ray photoemission of ZnO and other wurtzite semiconductors,” Appl. Phys. Lett. 98, 101906 (2011).

102. H.L. Mosbacker, C. Zgrabik, M.J. Hetzer, A. Swain, D.C. Look, G. Cantwell, J. Zhang, J.J.

Song, and L.J. Brillson, “Thermally-driven defect formation and blocking layers at metal-ZnO interfaces,” Appl. Phys. Lett. 91, 072102 (2007). 111

103. Allen and Durbin, “Influence of oxygen vacancies on Schottky contacts to ZnO,” APL 92,

122110 (2008)

104. A. Janotti and C.G. Van de Walle, “Hydrogen multicenter bonds,” Nat. Mat. 6, 44 (2007).

105. D.C. Look, K.D. Leedy, L. Vines, B.G. Svensson, A. Zubiaga, F. Tuomisto, D.R. Doutt, and

L.J. Brillson, “Self-compensation in semiconductors: the Zn-vacancy in Ga-doped ZnO,” Phys.

Rev. B 84, 115202 (2011).

106. Y. Dong, F. Tuomisto, B. G. Svensson, A. Y. Kuznetsov, and L. J. Brillson, “Vacancy defect and defect cluster energetics in ion-implanted ZnO,” Phys. Rev. B 81, 081201 (R) (2010).

107. K.E. Knutsen, A. Galeckas, A. Zubiaga, F. Tuomisto, G.C. Farlow, B.G. Svensson, and A.

Yu. Kuznetsov, “Zinc vacancy and oxygen interstitial in ZnO revealed by sequential annealing and electron irradiation,” Phys. Rev. B 86, 121203(R) (2012).

108. X.J. Wang, L.S. Vlasenko, S.J. Pearton, W.M. Chen, and I.A. Buyanova, “Oxygen and zinc vacancies in as-grown ZnO single crystals,” J. Phys. D: Appl. Phys. 42, 175411 (2009).

109. R. Dingle, “Luminescent transitions associated with divalent copper impurities and the green emission from semiconductor zinc oxide,” Phys. Rev. Lett. 23, 579 (1969).

110. F. Leiter, H. Zhou, F. Henecker, A. Hofstaetter, D.M. Hofmann, and B.K. Meyer, “Magnetic resonance experiments on the green emission in undoped ZnO crystals,” Physica B 308-310, 908

(2001).

111. F. Leiter, H. Alves, D. Pfisterer, N.G. Romanov, D.M. Hofmann, and B.K. Meyer, “Oxygen vacancies in ZnO,” Physica B 340-342, 201 (2003).

112

112. Y. Dong, Z-Q. Fang, D.C. Look, G. Cantwell, J. Zhang, J.J. Song, and L. J. Brillson, “Zn- and O-face polarity effects at ZnO surfaces and metal interfaces,” Appl. Phys. Lett. 93, 072111

(2008).

113. Y. Dong, Z.-Q. Fang, D. C. Look, D. R. Doutt, G. Cantwell, J. Zhang, and J. J. Song and

L. J. Brillson, “Defects at Oxygen Plasma Cleaned ZnO Polar Surfaces,” J. Appl. Phys. 108,

103718 (2010).

114. S. Müller, H. Von Wenckstern, F. Schmidt, D. Splith, R. Heinhold, M. Allen, and M.

Grundmann, “Method of choice for fabrication of high-quality ZnO-based Schottky diodes,” J.

Appl. Phys. 116, 194506 (2014).

115. R. Heinhold, H.S. Kim, F. Schmidt, H. von Wenckstern, M. Grundmann, R.J. Mendelsberg,

R.J. Reeves, S.M. Durbin, and M. W. Allen, “Optical and defect properties of hydrothermal ZnO with low lithium contamination”, Appl. Phys. Lett. 101, 062105 (2015).

116. A.M. Hyland, R.A. Makin, S.M. Durbin, and M.W. Allen. “Giant improvement in the rectifying performance of oxidized Schottky contacts to ZnO,” J. Appl. Phys. 121, 024501

(2017).

117. D. Drouin, A. Réal Couture, D. Joly, X. Tastet, V. Aimez, and R. Gauvin, CASINO V2.42 –

A Fast and Easy-to-use Modeling Tool for Scanning Electron Microscopy and Microanalysis

Users, Scanning 29, 92 (2007).

118. L.J. Brillson, Y. Dong, D. Doutt, D.C. Look, and Z.-Q. Fang, “Massive Point Defect

Redistribution Near Semiconductor Surfaces and Interfaces and Its Impact on Schottky Barrier

Formation,” Physica B 404, 4768 (2009).

113

119. F.Bernadini, V. Fiorentini, and D. Vanderbilt, “Spontaneous polarization and piezoelectric constants of III-V nitrides,” Phys. Rev. B 56, R10024 (1997).

120. H. Gao, T.J. Asel, J.W. Cox, and L.J. Brillson, unpublished.

121. E. Mayes, J Partridge, M. Field, D. McCulloch, S. Durbin, H. Kim, M. Allen, “The interface structure of high performance ZnO Schottky diodes,” Physica B 407 2867 (2012).

122 Schmidt-Mende, L; MacManus-Driscoll, J. L. Materials Today 2007, 10, 40-48.

123 Léonard, F.; Talin A.A. Nat. Nanotechnol., 2011, 6, 773-783.

124 Zekentes, K.; Rogdakis, K. J. Phys. D: Appl. Phys. 2011, 44, 133001.

125 English, C.D.; Shine, G.; Dorgan, V.E.; Saraswat, K.C.; Pop, E. Nano Lett. 2016, 3824-3830.

126 Dasgupta, N.P.; Yang, P. Front. Phys. 2014, 9, 289-302.

127 Agarwal, R. Small 2008, 4, 1872-1893.

128 Sato, K.; Castaldini, A.; Fukata, N.; Cavallini, A. Nano Lett. 2012, 12, 3012-3017.

129 Brillson, L.J. Surf. Sci. Rept. 1982, 2, 123-326.

130 Brillson, L.J. Surfaces and Interfaces of Electronic Materials; Wiley-VCH, 2010.

131 Léonard, F.; Talin, A.A. Phys. Rev. Lett. 2006. 97, 026804.

132 Bercu, B.; Geng, W.; Simonetti, O.; Kostcheev, S.; Sartel, C.; Sallet, V.; Lérondel, G.;

Molinari, M.; Giraudet, L.; Couteau, C. Nanotechnology 2013, 24, 415202.

133 Goldberger, J.; Sirbuly, D.J.; Law, M.; Yang, P. “ZnO Nanowire Transistors,” J. Phys. Chem.

B 2005,109, 9-14. DOI: 10.1021/jp0452599

134 Zimmermann, G.; Lange, M.; Cao, B.; Lorenz, M.; Grundmann, M. “Resistivity control of

ZnO nanowires by Al doping,” Phys. Stat. Sol. 2010, 4, 82-84. DOI 10.1002/pssr.201004015

114

135 Schlenker, E.; Bakin, A.; Weimann, T.; Hinze, P.; Weber, D.H.; Gölzhäuser, A.; Wehmann,

H-H.; Waag, A. Nanotechnol. 2008, 19, 365707. doi:10.1088/0957-4484/19/36/365707

136 Lau, C.S.; Liu, J.; Gao, P.; Zhang, L.; Davidovic, D.; Tummala, R.;Wang, Z.L. “ZnO

Nanobelt/nanowire Schottky Diodes Formed by Dielectrophoresis Alignment across Au

Electrodes,” Nano Lett. 2006, 6, 263-266. DOI: 10.1021/nl052239p

137 Ruane, W.T.; Leedy, K.; Look, D.C.; Farlow, G.; von Wenckstern, H.; Grundmann, M.;

Brillson, L.J. “Defect Segregation and Optical Emission in ZnO Nanowires and Microwires,”

Nanoscale 2016, 8, 7631-7637. DOI: 10.1039/c5nr08248j

138 Hong, W.-K.; Sohn, J.I.; Hwang, D.-K.; Kwon, S-S.; Jo, G.; Song, S.; Kim, S-M.; Ko,, H-J.;

Park,S.-Ju.; Welland, M.E.; Lee, T. “Tunable Electronic Transport Characteristics of Surface-

Architecture-Controlled ZnO Nanowire Field Effect Transistors,” Nano Lett. 2008, 8, 950-956.

139 Li, P.; Liao, Q.; Zhang, Z.; Wang, Z.; Lin, P.; Zhang, X.; Kang, Z.; Huang, Y.; Gu, Y.; Yan,

X.; Zhang, Y. “Investigation on the Mechanism of Nanodamage and Nanofailure for Single ZnO

Nanowires under an Electric Field,” ACS Appl. Mat. Int. 2014, 6, 2344-2349. dx.doi.org/10.1021/am4044079

140 Shao, Z.Z.; Zhang, Z.A.; Wang, Z.F.; Chang, S.L. “Electrical characteristics of Pt-ZnO

Schottky nano-contact,” Sci. China 2010, 53, 64-67. doi: 10.1007/s11433-010-0077-2

141 Weissenberger, D.; Gerthsen, D.; Reiser, A.; Prinz, G.M.; Feneberg, M.; Thonke, K.; Zhou,

H.; Sartor, J.; Fallert, J.; Klingshim, C.; Kalt, H. “Influence of the measurement procedure on the field-effect dependent conductivity of ZnO nanorods,” Appl. Phys. Lett. 2009, 94, 042107. doi:

10.1063/1.3075849

115

142 Lee, T.I.; Choi, W.J.; Kar, J. P.; Kang, Y.H.; Jeon, H.H.; Park, J.H.; Kim, Y.S.; Baik, H.K.;

Myoung, J.M. “Electrical Contact Tunable Direct Printing Route for a ZnO Nanowire Schottky

Diode,” Nano. Lett. 2010, 10, 3517-3523.

143 A. B. Djurišič, A.B.; Leung, Y.H. “Optical Properties of ZnO Nanostructures,” Small 2006, 2,

944-961. DOI: 10.1002/smll.200600134

144 Oh, Y.M.; Lee, K.M.; Park, K.H.; Kim, Y.; Ahn, Y.H.; Park, J.-Y.; Lee, S. “Correlating

Luminescence from Individual ZnO Nanostructures with Electronic Transport Characteristics,”

Nano Lett. 2007, 7, 3681-3685. DOI: 10.1021/nl071959o

145 Brillson, L.J. “Applications of Depth-Resolved Cathodoluminescence Spectroscopy,” J. Phys.

D: Appl. Phys. 2012, 45, 183001-183027. doi:10.1088/0022-3727/45/18/183001

146 Heo, Y.W.; Tien, L.C.; Norton, D.P.; Pearton, S.J.; Kang, B.S.; Ren, F.; J.R. LaRoche, J.R.

“Pt/ZnO nanowire Schottky diodes,” Appl. Phys. Lett. 2004, 85, 3107 (2004). DOI:

10.1063/1.1802372]

147 He, J.-H.; Ke, J.-J.; Chang, P.-H.; Tsai, K.-T.; Yang, P.C.; Chang, I.-M. “Development of

Ohmic nanocontacts via surface modification for nanowire-based electronic and optoelectronic devices: ZnO nanowires as an example,” Nanoscale 2012, 4, 3399. DOI: 10.1039/c2nr30688c

148 Das, S.N.; Kar, J.P.; Myoung, J.-M. “Junction Properties and Applications of ZnO Single

Nanowire Based Schottky Diode,” in Nanowires-Fundamental Research, A. Hashim (Ed.) ISBN:

978-953-307-327-9, InTech.

149 Mao, S.; Shang, T.; Park, B.; Anderson, D.D.; Dillon, S.J. Appl. Phys. Lett. 2014, 104,

153105. doi: 10.1063/1.4871509

116

150 Asel, T.J.; Gao, H.; Heinl, T.J.; Adkins, D.; Woodward, P.M.; Hoffman, J.; Bhattacharya, A.;

Brillson, L.J. “Near-Nanoscale-Resolved Energy Band Structure of LaNiO-

3/La2/3Sr1/3MnO3/SrTiO3 Heterostructures and their Interfaces,” J. Vac. Sci. Technol. 2015, B 33,

04E103. doi:10.1116/1.4922270

151 Drouin, D. “CASINO a Powerful Simulation Tool for Cathodoluminescence Applications,”

Microsc. Microanal. 2006, 12 (Suppl. 02), 1512. DOI: 10.1017/S1431927606069686

152 Dietrich, C.P.; Brandt, M.; Lange, M.; Kupper, J.; Böntgen, T.; von Wenckstern, H.; M.

Grundmann, M. “Defect properties of ZnO and ZnO:P microwires,” J. Appl. Phys. (2011) 109,

013712. doi:10.1063/1.3530610

153 Mosbacker, H.L.; Zgrabik, C.; Hetzer, M.J.; Swain, A.; Look, D.C.; Cantwell, G.; Zhang, J.;

Song, J.J.; Brillson., L.J. “Thermally driven defect formation and blocking layer at metal-ZnO interfaces,” Appl. Phys. Lett. 2007, 91, 072102. DOI: 10.1063/1.2772664

154 Meyer, B.K.; Alves, H.; Hofmann, D.M.; Kriegseis, W.; Forster, D.; Bertram, F.; Christen, J.;

A. Hoffmann, Straβburg, M.; Dworzak, M.; Haboeck, U.; Rodina, A.V. “Bound exciton and donor-acceptor pair recombination in ZnO,” Phys. Stat. Sol. (b) 2004, 241, 231-260. https://doi

10.1002/pssb.200301962

155 Foster, G.M.; Faber, G.; Yao, Y.-F.; Yang, C.C.; Heller, E.R.; Look, D.C.; Brillson, L.J.

“Direct measurement of defect and dopant abruptness at high electron mobility ZnO homojunctions,” Appl. Phys. Lett. 2016, 109,143506. https://doi.org/10.1063/1.4963888

156 Foster, G.M.; Gao, H.; Mackessy, G.; Hyland, H.M.; Allen, M.W.; Wang, B.; Look, D.C.;

Brillson, L.J. “Impact of defect distribution on IrOx/ZnO interface doping and Schotty barriers,”

Appl. Phys. Lett. 2017, 111, 101604. https://doi.org/10.1063/1.4989539 117

157 Look, D.C.; Walters, D.C.; Manasreh, M.O.; Sizelove, J.R.; Stutz, C.E.; Evans, K.R.

“Anomalous Hall-effect results in low-temperature molecular-beam-epitaxial GaAs: Hopping in a dense EL2-like band,” Phys. Rev. B 1990, 42, 3578-3581.

158 L.J. Brillson and Y. Lu, “ZnO Schottky Barriers and Ohmic Contacts,” Appl. Phys. Rev./ J.

Appl. Phys.2011, 109, 121301-. https://doi.org/10.1063/1.3581173

118