UNIVERSITY OF CALIFORNIA,

Santa Barbara

Aqueous synthesis of zinc oxide films for GaN optoelectronic devices

A DISSERTATION

submitted in partial satisfaction of the requirements

for the degree of

DOCTOR OF PHILOSOPHY
in Materials

by

Arthur H. Reading

Dissertation Committee:
Professor Steven P. DenBaars, Chair
Professor Nadir Dagli
Professor Shuji Nakamura
Professor James Speck
Mar 2014

 Mar 2014 Arthur H. Reading

The dissertation of Arthur H. Reading

is approved and is acceptable in quality and form for
publication on microfilm and in digital formats:

Committee Chair

ii

University of California, Santa Barbara

Mar 2014

iii

TABLE OF CONTENTS
Page
LIST OF FIGURES

vii

LIST OF TABLES

xii

ACKNOWLEDGMENTS

xiii

CURRICULUM VITAE

xiv

ABSTRACT OF THE DISSERTATION

xvi

Introduction
1.1 GaN/InGaN-based Light Emitting Diodes . . . . . . . . . . . . .
1.1.1 Overview of GaN LED technology . . . . . . . . . . . .
1.1.2 Transparent current spreading layers for GaN LEDs . . . .
1.1.3 Performance metrics for LEDs . . . . . . . . . . . . . . .
1.2 Zinc Oxide . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2.1 Material Properties of ZnO . . . . . . . . . . . . . . . . .
1.2.2 ZnO grown by Metallorganic Chemical Vapor Deposition
(MOCVD) . . . . . . . . . . . . . . . . . . . . . . . . .
1.2.3 ZnO grown by Molecular beam epitaxy (MBE) . . . . . .
1.3 n-ZnO contacts to p-GaN layers . . . . . . . . . . . . . . . . . .

1
. 1
. 1
. 2
. 5
. 10
. 10
. 12
. 14
. 15

Prior Work Using Aqueous Synthesis Method

21
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.2 Hydrothermal ZnO growth chemistry . . . . . . . . . . . . . . . . . 22
2.3 ZnO current spreading layers . . . . . . . . . . . . . . . . . . . . . 23

Undoped ZnO current-spreading layers on c-plane LEDs

27
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2 Experimental methods . . . . . . . . . . . . . . . . . . . . . . . . 29
iv

3.3

3.4

3.5
4

3.2.1 Epitaxial ZnO growth on c-plane GaN-on-sapphire . . . .

3.2.2 Circular Transmission Line Method . . . . . . . . . . . .
3.2.3 LED processing for devices on c-plane sapphire substrates
3.2.4 LED packaging for devices on c-plane sapphire substrates
3.2.5 LED testing . . . . . . . . . . . . . . . . . . . . . . . . .
Characterization of ZnO films . . . . . . . . . . . . . . . . . . .
3.3.1 Thermal Annealing of Undoped Films on c-plane GaN . .
3.3.2 XRD characterization . . . . . . . . . . . . . . . . . . . .
3.3.3 SEM characterization . . . . . . . . . . . . . . . . . . . .
LED characterization . . . . . . . . . . . . . . . . . . . . . . . .
3.4.1 I-V performance . . . . . . . . . . . . . . . . . . . . . .
3.4.2 LED efficiency . . . . . . . . . . . . . . . . . . . . . . .
3.4.3 Current spreading performance . . . . . . . . . . . . . . .
3.4.4 LEDs with etch-roughened ZnO CSLs . . . . . . . . . . .
Summary and Conclusion . . . . . . . . . . . . . . . . . . . . . .

Ga-doped ZnO current-spreading layers on c-plane LEDs

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . .
4.2 Experimental methods . . . . . . . . . . . . . . . . .
4.2.1 ZnO film growth . . . . . . . . . . . . . . . .
4.2.2 Elemental characterization . . . . . . . . . . .
4.2.3 Hall effect measurements . . . . . . . . . . . .
4.2.4 Optical absorption measurements . . . . . . .
4.2.5 LED device fabrication . . . . . . . . . . . . .
4.3 Characterization of Ga-doped ZnO layers . . . . . . .
4.3.1 SIMS elemental Characterization . . . . . . .
4.3.2 Hall Effect Characterization . . . . . . . . . .
4.3.3 Optical Characterization . . . . . . . . . . . .
4.3.4 XRD characterization . . . . . . . . . . . . . .
4.4 c-plane LED results . . . . . . . . . . . . . . . . . . .
4.5 Summary and Conclusion . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

29
32
34
39
43
46
46
50
52
53
53
54
56
57
59

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

64
64
65
65
68
68
71
73
73
73
75
77
80
80
85

Ga-doped ZnO current-spreading layers for semipolar GaN LEDs

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2 Experimental Methods . . . . . . . . . . . . . . . . . . . . . .
5.2.1 Patterning methods . . . . . . . . . . . . . . . . . . . .
5.2.2 Film growth . . . . . . . . . . . . . . . . . . . . . . . .
5.2.3 Device Fabrication . . . . . . . . . . . . . . . . . . . .
5.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3.1 XRD characterization of semipolar ZnO on GaN . . . .

.
.
.
.
.
.
.

.
.
.
.
.
.
.

87
87
88
88
90
95
96
96

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

5.4
6

p-GaN
5.3.2 Contact resistance measurements on 202 1 and 2021
5.3.3 TEM characterization of semipolar ZnO on GaN . . . . .
semipolar GaN LED . . . . . . . . .
5.3.4 LED results on 2021
Summary and Conclusions . . . . . . . . . . . . . . . . . . . . .

and 202 1 GaN lasers with ZnO cladding layers

Semipolar 2021
6.1 Background and Motivation for oxide-clad GaN laser diodes
6.2 GaN lasers with ITO cladding . . . . . . . . . . . . . . . .
substrates . . . . . . . . .
6.3 Optical characterization on 2021
blue laser fabrication . . . . . . . . . . . . . . . . .
6.4 (2021)
blue laser results . . . . . . . . . . . . . . . . . . . .
6.5 (2021)
contact studies . . . . . . . . . . . . . . . . . . . . .
6.6 (202 1)
6.6.1 Experimental method . . . . . . . . . . . . . . . . .
6.6.2 IV characterization . . . . . . . . . . . . . . . . . .
6.7 Summary and Conclusions . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.

98
100
102
103

.
.
.
.
.
.
.
.
.

106
106
108
108
112
114
116
117
118
119

Appendices
122
A
Optical Absorption calculations . . . . . . . . . . . . . . . . . . . 122

vi

LIST OF FIGURES
Page
1.1

1.2

1.3
1.4

3.1

3.2

3.3

Schematics demonstrating potential problems with a current-spreading

layer design. a) High sheet resistance, (b) High absorption, (c) High
contact resistivity, and (d) Optimal current spreading . . . . . . . .
Eye sensitivity function (left axis) and luminous efficacy (right axis).
Retina responsivity reaches a peak value at 555nm for green-yellow
light. Figure taken from Schubert [3]. . . . . . . . . . . . . . . . .
The Wurtzite (B4) crystal structure, which is shared by both GaN
and ZnO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Band diagrams generated using Silense band modeling software. pGaN (left) and n-ZnO junctions with (a) nZnO = 1020 cm-3 pGaN =
5 1018 cm-3 at zero bias,(b) at 1V reverse bias, (c) nZnO = 1020 cm-3
pGaN = 2 1018 cm-3 , (d) at 1V reverse bias, (e) nZnO = 1019 cm-3
pGaN = 2 1018 cm-3 , (f) at 1V reverse bias . . . . . . . . . . . . . .

11
12

17

Diagrams of a TLM structure, with metal in gold, ZnO in cyan,

and p-GaN in red. (a) a linear TLM structure that could be used
to measure metal/ZnO contact resistance and ZnO sheet resistance.
(b) Example plot of resistance vs. contact spacing used to determine
contact and R sh . (c) Diagram of a CTLM structure that could be used
to measure ZnO/p-GaN contact resistance and p-GaN sheet resistance. 33
LED fabrication process. Note that there is always some undercut,
which can be up to approx. 10m after the ZnO wet-etching step.
The metal stacks used were 10/100/10/500nm Ti/Al/Ni/Au for ncontacts, and 10/500nm Ti/Au for the p-pads. . . . . . . . . . . . . 37
White LED package. Due to the poor thermal properties of this
package, it was necessary to do pulsed measurements for measurements at high current-density. However, the encapsulation provides
a substantial boost as compared to an unencapsulated LED chip. . . 42

vii

3.4

CTLM measurements of the ZnO/p-GaN contact resistance showed

that annealing at 500C gave the lowest contact resistance at 7.0
102 cm2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.5 Effect of post-annealing on the sheet-resistance of ZnO films on cplane GaN, measuared using the CTLM method. Error bars show
the std. deviation of the measured samples . . . . . . . . . . . . . . 48
3.6 XRD characterization of a ZnO film grown on a commercial GaN/InGaN
LED wafer on a sapphire substrate. This film had a seed layer
annealed at 500C and a post-deposition anneal at 250C for 15
minutes. (a) 2-scan showing ZnO and GaN (002) reflections. (b)
-scan of ZnO (002) reflection, showing a FWHM of 0.365. (c) reflections, and (d) -rotational scan
rotational scan of ZnO (1011)

of GaN (1011) reflections, confirming heteroepitaxial relationship

between ZnO and GaN layers. . . . . . . . . . . . . . . . . . . . . 49
3.7 SEM micrographs showing formation of roughened c-plane ZnO
surface by etching in 1:100 diluted HCl solution. . . . . . . . . . . 51
3.8 LED IV testing results from devices with ZnO and ITO current
spreading layers. The ITO layer had a thickness of 240nm, and a
sheet resistance of 91/. The ZnO current spreading layers were
2100-2300nm thick and had a sheet resistance of 26/. . . . . . . 53
3.9 External quantum efficiency (EQE) of the 4 types of LEDs. The
dominant wavelength for these devices was at 470nm. Pulsed measurements were made with a 10% duty cycle. . . . . . . . . . . . . 54
3.10 Power efficiency of the 4 types of LEDs. This metric combines
the effects of quantum efficiency with the effects of resistive power
dissipation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.11 Current spreading characteristics demonstrated using a high-aspectratio LED. (a) optical microscope image, with the p-contact on the
bottom. (b) Current spreading performance at currents from 1 to
10mA that have been colorized to show the extent of current spreading. (c) Light-intensity distribution for the ZnO CSL. (d) Lightintensity distribution for the ITO CSL. . . . . . . . . . . . . . . . . 58
3.12 (a) IV curves of c-plane LEDs with ZnO CSLs roughened for different times in dilute HCl. After 40s the films had thinned considerably which resulted in higher-voltage operation. (b) %EQE of
LEDs with roughened ZnO CSLs. . . . . . . . . . . . . . . . . . . 59
4.1

Van der Pauw geometry used to measure carrier concentration &

mobility via the Hall method . . . . . . . . . . . . . . . . . . . . . 70

viii

4.2

Elemental characterization (a) Calibrated SIMS results showing fraction of Ga atoms in doped ZnO films. The left-most side of the
chart represents the surface of the ZnO film, while the right-most
side represents data from near the spinel substrate. (b) % incorporation of Ga in the films with Ga precursor concentration in growth
solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3 Hall effect measurements showing (a) the n-type carrier concentration in the films, (b) the electron mobility in the films, and (c) the
bulk electric conductivity of the films. Annealing was performed
sequentially on the same samples of c-plane ZnO on spinel substrates.
4.4 Optical absorption coefficients calculated from UV-visible transmission spectra. The calculation method by Cisneros [3] was used.
Films thicknesses were measured by cross-sectional SEM to be between 4500 and 5500m. Substrate index was fitted to a Sellmeier
refractive index model. . . . . . . . . . . . . . . . . . . . . . . . .
4.5 Cross-sectional SEM showing (a) voids present in undoped ZnO
film after 250C, and (b) no voids present in the Ga-doped ZnO
film with 2.0mM Ga precursor concentration. SEM figure courtesy
Jacob Richardson. . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.6 TCO Conductivity of absorption figure of merit for Ga-doped
ZnO layers under different annealing and doping conditions. . . . .
4.7 X-ray diffraction spectra of ZnO and ZnO:Ga films grown on spinel
substrates. (a) 2 scan near ZnO (002) peak, (b) 2 scan of an offaxis ZnO (101) peak. (c) 2 scan near ZnO (002) peak after annealing at 250C, and (d) 2 scan of off-axis ZnO (101) peak after
annealing at 250C. . . . . . . . . . . . . . . . . . . . . . . . . . .
4.8 External quantum efficiency of type 3 LEDs, with peak wavelength
at 470nm. Measurements were done on bare chips without silicone
encapsulation to avoid packaging-related inconsistencies. The device area was 1.07mm2 . . . . . . . . . . . . . . . . . . . . . . . .
4.9 Optical microscope images of ITO and 0.5mM Ga current spreading layers. Current spreading was almost completely uniform for
the 0.5mM ZnO:Ga CSLs. . . . . . . . . . . . . . . . . . . . . . .
4.10 Characterization of white LEDs fabricated with ITO and best ZnO:Ga
LED chips. (a) I-V curves showing operating voltages in the packaged white LEDs, and (b) luminous efficacy of white LEDs with
photograph of operating device . . . . . . . . . . . . . . . . . . . .

ix

74

76

78

79
79

81

83

83

84

5.1

5.2

5.3

5.4
5.5

5.6

5.7

5.8

6.1

6.2

SEM micrographs showing a patterned ZnO layer grown using an

GaN substrate.
SiOx hard mask for etch-free patterning on a (202 1)
(a) CTLM pad showing metal, ZnO, and SiOx layers. (b) Some
prismatic ZnO particles did nucleate in solution, which is undesir
able, espectially for long growths, (c) UHR image of 700nm
thick
ZnO layer (SiOx is 200nm thick). (d) top-down image of CTLM
pad showing precise patterning . . . . . . . . . . . . . . . . . . . .
Heating curves for 90mL PTFE vessel with different Variac powers.
Higher power allowance above 75% led to melting of the PTFE
vessel exterior. 90C was reached at 54 min., at which time a 100200nm film is already deposited. . . . . . . . . . . . . . . . . . . .
Diagram of PTFE vessel used for growing ZnO films on small sub11cm free-standing GaN wafers. The variac power was set at 75%
to prevent the PTFE melting. . . . . . . . . . . . . . . . . . . . . .
Test structures allowing measurement of ZnO contact resistance
patterned using SiOx hard mask . . . . . . . . . . . . . . . . . . .
and 202 1 pXRD characterization of ZnO films grown on 2021
GaN templates. (a) 2 scan of ZnO and GaN (201) reflections, and
(b) rocking curve scan showing higher crystal quality, and less
template . . . . . . . . . . .
tilt in the ZnO films grown on the 2021
TEM and CBED characterization of ZnO layers on semipolar p p-GaN, and (b) ZnO on 202 1
GaN templates. (a) ZnO on 2021
p-GaN. TEM courtesy Dr. Feng Wu. . . . . . . . . . . . . . . . . .
bulk GaN subIV performance of green 0.1mm2 LEDs on 2021
strates with different current-spreaing layers (CSLs). ECR-sputtered
ITO layers were deposited by MES AFTY Co. using an ECR
remote plasma sputtering system. E-beam evaporated ITO layers
were deposited on 250C heated substrates. . . . . . . . . . . . . .
GaN substrate.
LED performance of a green LED on bulk 2021
LED packaging used a conventional silicone dome on a silver LED
header. The backside of the GaN substrate was roughened using a
dry etch procedure to enhance light extraction. . . . . . . . . . . . .

91

92

93
96

97

101

103

104

(a) Experimental vs. simulated VASE data showing good fit with
experimental values using a graded-conductivity ellipsometry model.
(b) Complex refractive index data showing dispersion using gradedconductivity model. . . . . . . . . . . . . . . . . . . . . . . . . . . 110
Refractive index values calculated using methods described in Appendix A by transmission spectroscopy. Films were 4.5-6m thick
and annealed at 150C. . . . . . . . . . . . . . . . . . . . . . . . . 111
x

6.3

6.4

6.5

A.6
A.7
A.8

A.9

blue laser pulsed LIV characterization at 1% duty cycle,

(2021)
with (a) 1.6m, (b) 4m, and (c) 8m wide ridges. (d) Device
threshold currents and voltages vs. ridge widths. . . . . . . . . . . .
(a) 1.6m device operating above threshold, and (b) Spectral narrowing for a 1.6m wide laser ridge, for currents below and above
the threshold current, confirming lasing. . . . . . . . . . . . . . . .
GaN laser structures using 50mIV characterization on (202 1)
radius contact pads with backside contacts. Laser structures were
grown by Arash Pourhashemi. . . . . . . . . . . . . . . . . . . . .
(a) Transmission spectrum of bare DSP spinel substrate. (b) Fitted
substrate refractive index function. . . . . . . . . . . . . . . . . . .
(a) Transmission curve with labeled interference extrema. (b) Computed phase function . . . . . . . . . . . . . . . . . . . . . . . . .
Error analysis of absorption coefficient modeled by Cisneros method.
When is changed by 20% from the calculated value there is a
large shift in the modeled data, no longer reproducing experiment at
the interference maxima. The model correctly touches the substrate
spectrum for = 0. . . . . . . . . . . . . . . . . . . . . . . . . . .
Error analysis of film refractive index modeled by Cisneros method.
When nfilm is changed by 20%, the computed transmission value
barely changes at the even-order points. This verifies that the absorption coefficient can be computed at these maxima with little
dependence on the film index. Even when the film index is moved
to less than the substrate index, the model successfully reproduces
the transmission values at the even-order points. . . . . . . . . . . .

xi

115

116

119
126
127

129

130

LIST OF TABLES
Page
3.1

3.2

Electrical properties of deposited current spreading layers calculated from CTLM measurements. Thicknesses were measured using a Dektak profilometer. . . . . . . . . . . . . . . . . . . . . . . 54
Luminous Efficacy comparison between Type 3 LEDs fabricated
with the phosphor cap structure shown in Figure 3.3. Measurements
were done using a 10% duty cycle to avoid irreproducible heating
effects. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

4.1

Electrical properties of ITO, ZnO, and ZnO:Ga current spreading

layers, characterized by CTLM method. . . . . . . . . . . . . . . . 82

5.1

CTLM measurements on ZnO layers grown on 120nm p-GaN/10nm

and 202 1 substrates. . . . . . . . . . . 99
p++ -GaN templates on 2021

xii

ACKNOWLEDGMENTS
I would like to thank my commitee members, Professors DenBaars, Nakamura,
Speck, and Dagli, for their contributions and interest in this work. I would also like
to thank all university personnel who made this work possible. Staff scientists,
technicians, and administrative workers who all contributed greatly to things going
smoothly and helping to get things done. Id like to thank all the great students in
our group at UCSB who provided many great friendships, and many interesting
discussions and ideas.
This work was supported by the Solid-State Lighting & Energy Center at UC
Santa Barbara. This work made use of the MRL Shared Experimental Facilities,
which are supported by the MRSEC program of the NSF under Award No. DMR
1121053; a member of the NSF-funded Materials Research Facilities Network,
and the UCSB nanofabrication facility, part of the NSF-funded NNIN network.

xiii

CURRICULUM VITAE
Arthur H. Reading
EDUCATION
Doctor of Philosophy in Materials
University of California, Santa Barbara

2013
Santa Barbara, California

Bachelor of Science in Physics

Massachusetts Institute of Technology

2009
Cambridge, Massachusetts

Bachelor of Science in Mat. Sci. & Eng.

Massachusetts Institute of Technology

2009
Cambridge, Massachusetts

RESEARCH EXPERIENCE
Graduate Student Researcher
University of California, Santa Barbara

20092013
Santa Barbara, California

Undergraduate Researcher
Massachusetts Institute of Technology

2006-2009
Cambridge, Massachusetts

TEACHING EXPERIENCE
Teaching Assistant
University of California

Fall 2009
Santa Barbara, California

SELECTED HONORS AND AWARDS

Graduate Student Award for Outstanding Research Achievement
Solid State Lighting & Energy Center, UCSB

2011

Graduate Student Award for Outstanding Research Achievement

Solid State Lighting & Energy Center, UCSB

2013

xiv

JOURNAL PUBLICATIONS
High efficiency white LEDs with single-crystal ZnO currentspreading layers deposited by aqueous solution epitaxy
Optics Express

xv

2012

ABSTRACT OF THE DISSERTATION

Aqueous synthesis of zinc oxide films for GaN optoelectronic devices
By
Arthur H. Reading
Doctor of Philosophy in Materials
University of California, Santa Barbara, Mar 2014
Professor Steven P. DenBaars, Chair

GaN-based LEDs have generally made use of ITO transparent contacts as

current-spreading layers for uniform current injection. However, the high raw
material and processing costs of ITO layers have generated interest in potentially
cheaper alternatives. In this work, zinc oxide transparent layers were fabricated by
a low-cost, low-temperature aqueous epitaxial growth method at 90C for use as
transparent contacts to GaN LEDs on c-plane sapphire, and on semipolar bulk
GaN substrates. Low-voltage operation was achieved for c-plane devices, with
voltages below 3.8V for 1mm2 broad-area LEDs at a current density of 30A/cm2 .

Blue-green LEDs on 202 1-plane

GaN also showed low voltage operation below
3.5V at 30A/cm2 . Ohmic contact resistivity of 1.8 102 cm2 was measured for
p-GaN templates. Ga-doped films had electrical conductivities as
films on (2021)
high as 660S/cm after annealing at 300C. Optical characterization revealed
optical absorption coefficients in the 50-200cm-1 range for visibile light, allowing
thick films with sheet resistances below 10/ to be grown while minimizing
absorption of the emitted light. Accurate and reproducible etch-free patterning of
xvi

the ZnO films was achieved using templated growths with SiOx hard masks. A
roughening method is described which was found to increase peak LED
efficiencies by 13% on c-plane patterned sapphire (PSS) substrates. In addition,

ZnO films were successfully employed as laser-cladding layers for blue (2021)
lasers, with a threshold current density of 8.8kA/cm2 .

xvii

Chapter 1
Introduction

1.1
1.1.1

GaN/InGaN-based Light Emitting Diodes

Overview of GaN LED technology

Solid-state lasers and LEDs based on the III:nitride family of semiconductors have
revolutionized the fields of lighting, displays, and data storage. Blu-ray
technology, made possible by low-cost blue GaN/InGaN laser diodes, has enabled
high-definition entertainment products, as well as massive data storage capability
in a compact-disc form factor. LED-based replacements for inefficient
incandescent and halogen lightbulbs are rapidly gaining traction, as cost-saving,
low-maintenance, high color-quality lighting solutions across the globe.
Worldwide concern with carbon-emissions and energy wastefulness has spurred

many incentives for consumers and businesses to use these products, as well as for
researchers and technology businesses to further improve device capability.
GaN-based LEDs and lasers consist of 3 main layers, which are epitaxially formed
on a crystal substrate. These layers are of n-type GaN:Si, p-type GaN:Mg, with a
light-emitting active region in-between. Active regions in these structures are
generally intrinsic GaN with lower-bandgap Inx Ga1-x N quantum wells for
increased carrier localization and recombination efficiency. The prevailing
technique used to grow these structures is metallorganic chemical vapor deposition
(MOCVD), which employs precursor gases which form crystalline layers in the
presence of a heated crystal substrate. Substrates used for MOCVD GaN device
fabrication are sapphire, SiC, bulk GaN wafers, and recently, (111) Si wafers.

1.1.2

Transparent current spreading layers for GaN LEDs

GaN LEDs for most high-power applications are usually broad-area devices which
employ a lateral mesa structure. This design is particularly advantageous if it is
possible to pattern the substrate in order to allow light-extraction from the
backside of the GaN epitaxial layers. In these designs, the n-side contact is formed
by reactive-ion etching the p-GaN and active regions in order to expose areas of
n-type GaN. One important consideration in these types of designs is achieving
uniform current injection across the area of the mesas. This is problematic because
n-GaN is much more electrically conductive than the top-most p-type GaN layers.
Furthermore, it is not possible to grown p-side-down structures due to the tendency

Figure 1.1: Schematics demonstrating potential problems with a current-spreading

layer design. a) High sheet resistance, (b) High absorption, (c) High contact resistivity, and (d) Optimal current spreading
of Mg dopant to continue to be incorporated in subsequent layers after MOCVD
growth of the p-type layers in a device. In the absence of a current-spreading layer,
the current will flow through the active regions directly below the p-side metal
contacts, where much of the light will be blocked (see Figure 1.1).
For maximum device efficiency, it is essential that the current-spreading layer is
highly transparent, and that low-resistance electrical contacts can be formed with
the p-GaN layers. The bulk electrical conductivity of the layers must also be high
such that low sheet-resistances can be achieved at reasonable thicknesses. Chen
et al. [1] have done some theoretical modeling of such structures in order to
determine the optimum tradeoff between the sheet resistances of the n-GaN and
TCO layers, and the contact resistance between the p-GaN and the TCO layer.
One possible figure-of-merit for evaluating such a transparent conducting material

is the absorption of conductivity, which is calculated by

FOMTCO = /
, where is the bulk electrical conductivity, and is the optical absorption
coefficient. The units of this figure are -1 .
Initially, thin metal layers were used as current spreading layers, due to their
low-resistance Ohmic contact with p-GaN, and their high conductivity. In
particular, Ni/Au layers were often used in the earlier generations of GaN LEDs,
due to their ease of deposition by evaporation. However, these layers have rather
high absorption coefficients, often higher than 10000cm-1 , which limits their
thicknesses to a few nanometers to prevent unacceptable amounts of optical
absorption.
ITO is currently the layer of choice for LED current spreading layers, owing to its
high conductivity, and low optical absorption. There is a considerable variation in
the reported properties of ITO films, since its properties vary widely depending on
the deposition method, the deposition surface, annealing conditions, and the
elemental ratios of indium and tin in the films. ITO is an n-type semiconductor
with a wide band-gap of 4.0 eV. However, the large amount of defects in ITO films
lends to high optical absorption in the visible range, which is generally reported to
be in the 650-2000cm-1 range.

1.1.3

Performance metrics for LEDs

LED performance is characterized by a chain of efficiencies, which must all be

high in order for the device to perform efficiently. In other words, a single point of
failure in the device design can cause the entire design to perform inefficiently.
The 3 main efficiencies of interest in an LED are the internal quantum efficiency
(IQE), the external quantum efficiency (EQE), and the light-extraction efficiency
(LEE). For white-emitting LEDs, there is the added issue of the phosphor
wavelength conversion efficiency. However, this is outside the scope of this work,
as it is not affected by the properties of the transparent contacts within the LED
chips.
The internal quantum efficiency is a measure of how efficiently charge carriers are
being converted to light in the active region of the device. Commonly, the IQE is
characterized by the so-called ABC model, which attempts to model the IQE of a
device as a function of carrier concentration in the active layers. The model
compares the non-radiative carrier-recombination rate, An, with the radiative
carrier recombination rate Bn2 , and the Auger recombination rate, Cn3 . The IQE is
given by,

IQE =

Bn2
An + Bn2 + Cn3

(1.1)

As current density increases, there is a tendency for the IQE to reach a peak, after

which efficiency decreases for all increases in the current density. This effect is
known as LED droop. There has been much experimental and theoretical work
exploring the possible causes of droop. Recent work by Iveland et al. [2] has
allowed direct observation of Auger droop, as correlated with loss of device
efficiency. Regardless of the mechanisms associated with droop, it has been
unavoidable in all LED designs, and must be taken into account when designing
devices.
The light-extraction efficiency is the efficiency of an LED chip or package in
allowing the light generated in the active area to escape the chip to the outside
world. The refractive index of GaN is about 2.5 for blue light, which causes much
difficulty associated with light being reflected back into the LED chip, which
causes much of the light to eventually be reabsorbed. High LEEs have been
attainable using surface roughening techniques to increase the amount of light
extracted. However, light extraction still remains a problem in many bulk-GaN
LED designs, due to the tendency of light to multiply reflect within and LED die.
Light-extraction efficiency also includes the effect of absorbing areas in an LED
package, such as metal contacts, heat sinking materials, or wires within the LED
package.
The external quantum efficiency (EQE) of an LED device is defined as the total
number of photons emitted by the device per number of charge carriers allowed to
flow through the device. Thus, it is a measure of how efficiently the device
transforms electric current into useful emitted light. Experimentally, it is
observable by measuring the spectrally-resolved total light output power of the

device. Such a measurement yields a quantity of light emitted vs. the wavelength
of that light. From this we can calculate the total number of photons emitted per
unit time from the device,
Photon emission rate =

Pout
Pout
=
Avg. energy per photon hc/avg.

(1.2)

where h is Plancks constant, c is the speed of light in vacuum, avg. is the average
light wavelength emitted, and Pout is the total power optical power emitted.
Similarly, the total number of charge carriers entering the device are calculated by,

Rate of electron injection =

I
Current in
=
Charge per electron e

(1.3)

The EQE is defined as the ratio of the photon emission rate, and the current
injection rate, which gives,

EQE =

Pout avg. e
hcI

(1.4)

The EQE is an excellent measure of the device performance, since it combines so

many potential sources of efficiency loss. However, it has some drawbacks, which
prevent it from being a comprehensive performance measure. For example, it does
not take into account the resistive losses incurred in device operation. It is also
7

useful to calculate the EQE of a device for which the light-extraction efficiency is
known through experiment and extensive optical modeling. In these cases it is then
possible to calculate the IQE from the observed output power and wavelength.
There are several ways in which the performance of LED devices can be evaluated.
Given the rapid commercialization of LED technology, and the wide array of
applications in which they are used, it is sensible that LEDs that are appropriate
for one type of use will have drawbacks in a different situation. In other words, it
is important to consider the specific application in which an LED will be used
when evaluating its performance or efficiency. For example, street-lighting
applications require low-maintenance, high brightness, high efficiency, but do not
require good color-rendering. In contrast, lighting in art galleries require excellent
color-rendering, but are not sensitive to maintenance costs.
Another measure of the device performance, especially for high-power LEDs, is
known as the power efficiency, or the wall-plug efficiency (WPE). This simple
metric is calculated by,

WPE =

Pout
Optical power out
=
Electrical power in
IV

(1.5)

The WPE is important, because it measures any resistive losses in the LED device.
This is of importance for the analysis of a current spreading layer, since it
measures the amount of light extracted, along with how with how much power is

lost to waste heat. The amount of waste heat generated by an LED is given by,

Pheating = (1 WPE) Pin = IV(1 WPE)

(1.6)

This is an important factor in high-power LED packaging design, since the

operating temperature of a device tends to greatly influence the device lifetime.
Advanced heat sinking designs are necessary the more heat is produced by the
LED, which adds to the cost of the final package.
For evaluating the performance of a high-power white LED used in lighting
applications, the luminuous efficacy is often quoted. A lumen is a measure of
optical power, but weighted by a standardized sensitivity function which measures
how responsive a retina is to light of different wavelengths. The human optical
responsivity is shown in Figure 1.2. The benefit of this metric is that numbers can
be easily compared across lighting technologies, the main ones being conventional
incandescent, halogen and compact fluorescent (CFL) lighting. In order for an
LED to function as a source of white light, it is necessary to couple a higher
energy monochromatic LED emission to a phosphor material, which is excited at a
higher wavelength, but emits light over a range of longer wavelengths. The LED
wavelength and phosphor material(s) can be chosen such that the net light color
appears white.
However, consumer tastes are considerably more complex when it comes to
lighting sources. Pure lighting efficiency is not always the most important factor.
9

Factors like the color temperature, the color-rendering index, or the ability of
a package to be paired with a dimmer switch are often crucial properties which
will determine the success of a commercial LED lighting product. The color
temperature of a white light source is defined as the temperature of a black-body
emitter that would have the same color hue as the white light source. This allows
the categorization of white LEDs as cool white, with color temperatures above
4500K, and warm white, around 3000K, which is considered more visually
appealing by consumers.
Total lumens is another important consideration in LED lighting design. Many
houses have layouts based on using 100W incandescent lightbulbs in each fixture.
A 100W incandescent puts out about 1600 lumens. It has thus far been challenging
for LED producers to make lightbulbs that can output this much light without
generating enough heat to diminish the bulb lifetime. For this reason, the WPE of
the LED chips is of great importance, since it determines how much energy is
being wasted as heat, which is damaging to the device itself.

1.2
1.2.1

Zinc Oxide
Material Properties of ZnO

Zinc oxide is a wide-bandgap semiconductor. Like GaN, it has a wurtzite (B4)

crystal structure, which is shown in Figure 1.3. ZnO has a direct bandgap of 3.3eV
at room temperature [4], which is also close to that of GaN. The a and c lattice
10

Figure 1.2: Eye sensitivity function (left axis) and luminous efficacy (right axis).
Retina responsivity reaches a peak value at 555nm for green-yellow light. Figure
taken from Schubert [3].
parameters of ZnO are 3.25 and 5.2, respectively. There has been much interest
in ZnO as a light-emitting material, due to its high exciton binding energy of
60meV. However, its success in this area has been limited by the difficulty and
irreproducibility of creating p-type ZnO. There has also been much theoretical
work showing that there may be little hope for finding suitable shallow acceptor
impurities that would allow p-type doping [5].
The basal-plane lattice mismatch to GaN is only about 2%, which allows
single-crystal layers of ZnO to form on GaN substrates. Single-crystal layers are
advantageous due to the possibility of high carrier mobilities. ZnO has been
reported to have high transparences, which is important for transparent-conducting
layers on thin-film solar cells, LEDs and laser diodes. Furthermore, it has been
shown that the band-gap of ZnO can be tuned by alloying with Mg or Cd in order

11

Figure 1.3: The Wurtzite (B4) crystal structure, which is shared by both GaN and
ZnO
to widen, or compress, the band-gap, respectively [6].

1.2.2

ZnO grown by Metallorganic Chemical Vapor Deposition

(MOCVD)

MOCVD has been a method of interest in growing ZnO films due to the many
successes in creating device-quality crystalline layers of GaN and other
semiconductors. The Zn precursor generally used is diethyl zinc (DEZn), and
there are a variety of possible oxygen precursors, including O2 and N2 O. One
12

difficulty associated with the MOCVD growth of ZnO films is the propensity of
the precursor gases to pre-react causing the formation of powders in the reactor
chamber, which is highly detrimental to growing smooth crystalline films. For this
reason it is necessary to carefully control gas-flow conditions with the use of a
close-coupled showerhead reactor design, where the precursor gases are prevented
from mixing until they are already close to the heated substrate surface.
Highly-crystalline and conductive ZnO layers were grown by Ive et al. [7] and
Ben-Yaacov et al. [8] on GaN (0001) substrates. These were grown by a two-step
growth using a ZnO buffer layer grown at temperatures below 400C, followed by
a second layer grown at temperatures 750950C. RMS roughness of less than
2.6m were reported, and FWHM of -scans of the (0002) ZnO reflection were
reportedly 0.091, indicating a very high degree of crystallinity. Step-flow growth
was also observed in these reports, indicating well-controlled growth conditions.
An LED was fabricated by Ive et al. [9] using a Ga-doped ZnO transparent contact
grown at 300C with a forward voltage of 4.5V at 20mA, which demonstrates
interesting possibilities of ZnO as a transparent contact. Recently, Horng et al.
[10] have reported Ga-doped ZnO current spreading layers grown by MOCVD
with low contact and bulk resistivities and (8.7 103 cm2 and 3.64 104 cm,
respectively). This was accomplished by annealing the GZO films at temperatures
as high as 800C.
The growth rates reported for MOCVD growth of ZnO are of the order 1.2per
second, meaning it would take about 2.5 hours to grow a film of 1m thickness.
Long growth times add to the cost of the process, due to the amount of precursor

13

chemicals, energy, and time consumed. For this reason it would be of interest to
develop lower-cost methods with potentially faster growth rates.

1.2.3

ZnO grown by Molecular beam epitaxy (MBE)

Heteroepitaxy of ZnO layers on GaN films by MBE has been studied, and is of
interest due to the ability to precisely control growth conditions and to monitor the
film in-situ by RHEED electron diffraction. Ko et al. [11] have shown ZnO films
on GaN/Al2 O3 templates by RF-plasma assisted MBE with (0002)
rocking-curve FWHMs of 0.16in films grown at 510C with a Zn wafer
pre-treatment, and by using a ZnO buffer layer growth at 300C. Growth rates of
0.4-0.75m per hour were reported.
Ga-doping in ZnO films grown by PA-MBE were investigated by Ko et al. [12].
Carrier concentrations as high as 1.13 1020 were achieved, with ZnO (0002)
rocking-curve linewidths in the 0.08-0.25range. Control of the polarity of
wurtzite ZnO crystals grown on (0001) GaN templates was reported by Hong et al.
[13]. It was found that Zn-terminated or O-terminated ZnO films could be grown
heteroepitaxially, with control of crystal polarity achievable by changing wafer
pre-treatment conditions in the MBE chamber. The polarity inversion was made
possible by the formation of an interfacial Ga2 O3 layer between the GaN and ZnO
layers.
Highly conductive Ga-doped ZnO layers were grown by Tamura et al. [14] by
MBE as transparent Ohmic contacts to InGaN/GaN based LEDs. The ZnO
14

contacts were found to increase the efficiency of the LEDs by a factor of 2 times,
as compared to a reference LED with a thin Ni/Au p-electrode. The ZnO:Ga films
in this study had conductivities as high as 5000S/cm, which is comparable to
commercial-grade ITO layers. However these highly-conductive films generally
had poor transparency, as measured by UV-vis spectrophotometryonly 70%
transmittance for 450nm blue light in a 500nm ZnO:Ga film.
It is worth mentioning that there have been reports of p-type ZnO:N grown by
MBE using temperature modulation techniques [15]. There have even been reports
of such material producing LED homojuncitons with blue-violet visible range light
emission [16]. This is a matter of much controversy, as theoretical first-principles
calculations have shown that nitrogen only leads to deep acceptor states in ZnO
with ionization energies as high as 1.3eV, as reported by Lyons et al. [5]. These
theoretical calculations have been confirmed experimentally, using
photoluminescence measurements, by Tarun et al. [17]. It remains to be seen
whether reports of p-type ZnO will lead to reproducible light-emitting devices.

1.3

n-ZnO contacts to p-GaN layers

There have been several reports of fabricating n-ZnO/p-GaN heterojunction LEDs

[18, 19], which can emit violet, or UV-range light under forward bias. However,
this approach has not been found to approach the device efficiencies seen in
n-GaN/p-GaN homojunction LEDs, requiring high operating voltages and low
light emission intensities. In this work, the n-ZnO/p-GaN contact is in reverse
15

bias, and works due to carrier tunneling through a very narrow depletion region. In
an n-ZnO/p-GaN junction under reverse bias, the hole fermi energy of the p-GaN
layer is above the electron fermi energy in the n-ZnO, with only a thin depleted
region in-between, which acts as a quantum-mechanical potential barrier. Under
bias, there is a wavefunction overlap of the carriers with the states on the other side
of this barrier, which are swept away by the applied electric field under reverse
bias. This effect is desirable since it is the main mechanism by which an Ohmic
contact can be formed between the n-ZnO, and the p-GaN.
In Figure 1.4 we see the effect of varying the n-side and p-side doping
concentrations on the effectiveness of the ZnO contact. In general, the smaller the
depletion region between the n-ZnO and p-GaN layers, the lower the resistance of
the Ohmic contact. Notice that under reverse bias (the same bias direction for an
underlying LED at forward bias), the hole Fermi energy in the p-GaN is above the
electron Fermi energy in the n-ZnO, allowing good conduction. In the junctions
with lower doping concentrations, we can see that the depletion region becomes
wider, requiring carriers to tunnel across a greater distance, with lower probability,
resulting in a worse contact behavior.

16

Figure 1.4: Band diagrams generated using Silense band modeling software. pGaN (left) and n-ZnO junctions with (a) nZnO = 1020 cm-3 pGaN = 5 1018 cm-3 at
zero bias,(b) at 1V reverse bias, (c) nZnO = 1020 cm-3 pGaN = 2 1018 cm-3 , (d) at 1V
reverse bias, (e) nZnO = 1019 cm-3 pGaN = 2 1018 cm-3 , (f) at 1V reverse bias

17

References
[1] Chen, J.-C.; Sheu, G.-J.; Hwu, F.-S.; Chen, H.-I.; Sheu, J.-K.; Lee, T.-X.;
Sun, C.-C. Optical Review 2009, 16, 213215.
[2] Iveland, J.; Martinelli, L.; Peretti, J.; Speck, J. S.; Weisbuch, C. Phys. Rev.
Lett. 2013, 110, 177406.
[3] Schubert, E. Light-Emitting Diodes; Cambridge University Press, 2006.
[4] Srikant, V.; Clarke, D. R. Journal of Applied Physics 1998, 83, 54475451.
[5] Lyons, J. L.; Janotti, A.; Van de Walle, C. G. Applied Physics Letters 2009,
95, .
[6] Makino, T.; Segawa, Y.; Kawasaki, M.; Ohtomo, A.; Shiroki, R.; Tamura, K.;
Yasuda, T.; Koinuma, H. Applied Physics Letters 2001, 78, 12371239.
[7] Ive, T.; Ben-Yaacov, T.; de Walle, C. V.; Mishra, U.; DenBaars, S.; Speck, J.
Journal of Crystal Growth 2008, 310, 3407 3412.
[8] Ben-Yaacov, T.; Ive, T.; Van de Walle, C. G.; Mishra, U. K.; Speck, J. S.;
Denbaars, S. P. physica status solidi (c) 2009, 6, 14641467.
18

[9] Ive, T.; Ben-Yaacov, T.; Murai, A.; Asamizu, H.; Van de Walle, C. G.;
Mishra, U.; DenBaars, S. P.; Speck, J. S. physica status solidi (c) 2008, 5,
30913094.
[10] Horng, R.-H.; Shen, K.-C.; Yin, C.-Y.; Huang, C.-Y.; Wuu, D.-S. Opt.
Express 2013, 21, 1445214457.
[11] Ko, H. J.; Chen, Y.; Hong, S. K.; Yao, T. Journal of Crystal Growth 2000,
209, 816 821.
[12] Ko, H.; Chen, Y.; Hong, S.; Wenisch, H.; Yao, T.; Look, D. C. Applied
Physics Letters 2000, 77, 37613763.
[13] Hong, S.-K.; Hanada, T.; Ko, H.-J.; Chen, Y.; Yao, T.; Imai, D.; Araki, K.;
Shinohara, M.; Saitoh, K.; Terauchi, M. Phys. Rev. B 2002, 65, 115331.
[14] Tamura, K.; Nakahara, K.; Sakai, M.; Nakagawa, D.; Ito, N.; Sonobe, M.;
Takasu, H.; Tampo, H.; Fons, P.; Matsubara, K.; Iwata, K.; Yamada, A.;
Niki, S. physica status solidi (a) 2004, 201, 27042707.
[15] Tsukazaki, A.; Ohtomo, A.; Onuma, T.; Ohtani, M.; Makino, T.; Sumiya, M.;
Ohtani, K.; Chichibu, S. F.; Fuke, S.; Segawa, Y. Nature materials 2004, 4,
4246.
[16] Wei, Z.; Lu, Y.; Shen, D.; Zhang, Z.; Yao, B.; Li, B.; Zhang, J.; Zhao, D.;
Fan, X.; Tang, Z. Applied physics letters 2007, 90, 042113042113.
[17] Tarun, M. C.; Iqbal, M. Z.; McCluskey, M. D. AIP Advances 2011, 1, .

19

[18] Alivov, Y. I.; Van Nostrand, J.; Look, D.; Chukichev, M. V.; Ataev, B. M.
Applied Physics Letters 2003, 83, 29432945.
[19] Rogers, D. J.; Hosseini Teherani, F.; Yasan, A.; Minder, K.; Kung, P.;
Razeghi, M. Applied Physics Letters 2006, 88, .

20

Chapter 2
Prior Work Using Aqueous Synthesis
Method

2.1

Introduction

Hydrothermal growth of ZnO has long been established as a reliable method for
growing high-quality bulk single-crystals of ZnO [1, 2]. These methods generally
take place at high temperature and pressure, and produce high-quality crystals with
low carrier concentrations, and low electrical conductivities. The n-type carrier
concentrations of bulk hydrothermal crystals is in the 1014 -1015 cm-3 range [3],
which results in high resistivity. The low carrier concentrations in this type of ZnO
can be attributed to substitutional alkali metal acceptor impurities, which can
compensate free electron carriers [4].

21

2.2

Hydrothermal ZnO growth chemistry

ZnO in the context of GaN optoelectronics generally requires films with high
electrical conductivities. Also, in the case of p-type GaN, it is important that the
underlying substrate is not exposed to any conditions that could compensate or
otherwise damage the sensitive p-type GaN layers. For example, annealing in
hydrogen is a commonly used method to easily increase the carrier concentration
in ZnO, due to its ability to form shallow donor impurities, and its ability to diffuse
throughout a crystal [5, 6]. However, annealing in hydrogen will have severely
detrimental effects on the hole concentration in p-type GaN:Mg, since it can easily
compensate the Mg-acceptor impurities with donor states [7]. Also, techniques
like RF-sputtering of films onto p-GaN surfaces has shown to cause oxygen
plasma damage, which can compensate the holes in p-GaN [8]. For these reasons,
it is important to select a ZnO deposition method which will allow highly
conductive ZnO films to be grown, but mild conditions that will not compensate
holes in p-type GaN.
One method that fulfills these criteria is the low-temperature hydrothermal growth
of ZnO. This type of deposition method has been shown to grow high-quality films
of ZnO at temperatures as low as 90C [9, 10, 11, 12]. The premise behind this
growth method is that allowing films to grow in conditions close to equilibrium
allows the formation of high quality crystals. Also, a large amount of hydrogen is
incorporated in these films as they are grown, resulting in high n-type carrier
concentrations in as-grown films. Importantly, the low temperature conditions of
these growths allow the p-GaN layers to be largely unaffected by the
22

low-temperature growths.
Theoretical modeling work by Demianets et al. [4] and Richardson and Lange [13]
have allowed insight into low-temperature mechanisms which can control the
growth conditions of ZnO films. In particular, it has been made possible to
separately control the propensity of a growth solution to form many crystallites,
through high-nucleation conditions, or to continuously grow ZnO on existing ZnO
surfaces, with very low nucleation of new crystallites. This is an important tool for
optimizing the microstructural characteristics of ZnO films, powders, or
nanostructures. In this work we are primarily interested in the growth of thick
films of wurtzite ZnO on wurtzite GaN layers to act as current-spreading layers,
light-extraction structures, or as laser cladding.

2.3

ZnO current spreading layers

Some previous work by researchers has demonstrated operable LEDs using

hydrothermally grown zinc oxide as current spreading layers. Thompson et al. [14]
showed a large improvement in device efficiency using such ZnO layers, as
compared to a Ni/Au thin-metal contact. At an operating current of 20mA, a 93%
increase in light power was observed, indicating greatly improved
current-spreading layer performance for a blue LED device emitting at 452nm.
The device had a voltage of 3.32V at a current of 20mA, with a light output of
20.7mW, which gave a maximum external quantum efficiency of 37.7%, compared
to 19.5% for the device with a thin metal contact layer. These results were
23

achieved using commercially-acquired LED wafers.

ZnO current spreading layers grown by a similar method were reported on LEDs
grown by MOCVD on 101 1 substrates by Richardson et al. [15]. Here, the peak
EQE of the devices with ZnO current spreading layers was 48%. It was found that
there was an appreciable tilt in the ZnO crystal with respect to the underlying GaN
epilayers. Finally, there have been reports of ZnO patterned using imprint
nanolithography to create patterns of nanorods, which could prove useful in
improving the light-extraction efficiency on a GaN LED structure [16].
For the work in this thesis, the goal is to combine and improve these advances in
order to demonstrate high LED efficiencies utilizing these current spreading
layers, and to fully characterize the relevant material properties of these layers for
use with future LED designs.

24

References
[1] Laudise, R. A.; Ballman, A. A. The Journal of Physical Chemistry 1960, 64,
688691.
[2] Maeda, K.; Sato, M.; Niikura, I.; Fukuda, T. Semiconductor Science and
Technology 2005, 20, S49.
[3] Kassier, G. H.; Hayes, M.; Auret, F. D.; Mamor, M.; Bouziane, K. Journal of
Applied Physics 2007, 102, 0149030149035.
[4] Demianets, L.; Kostomarov, D.; Kuzmina, I.; Pushko, S. Crystallography
Reports 2002, 47, S86S98.
[5] Ip, K.; Overberg, M. E.; Heo, Y. W.; Norton, D. P.; Pearton, S. J.;
Stutz, C. E.; Luo, B.; Ren, F.; Look, D. C.; Zavada, J. M. Applied Physics
Letters 2003, 82, 385387.
[6] Oh, B.-Y.; Jeong, M.-C.; Kim, D.-S.; Lee, W.; Myoung, J.-M. Journal of
Crystal Growth 2005, 281, 475 480.
[7] Nakamura, S.; Iwasa, N.; Senoh, M.; Mukai, T. Japanese Journal of Applied
Physics 1992, 31, 12581266.
25

[8] Cao, X.; Pearton, S.; Zhang, A.; Dang, G. T.; Ren, F.; Shul, R.; Zhang, L.;
Hickman, R.; Van Hove, J. Applied Physics Letters 1999, 75, 25692571.
[9] Kim, J. H.; Kim, E.-M.; Andeen, D.; Thomson, D.; DenBaars, S. P.;
Lange, F. F. Advanced Functional Materials 2007, 17, 463471.
[10] Le, H. Q.; Lim, S. K.; Goh, G. K. L.; Chua, S. J.; Ong, J. Journal of The
Electrochemical Society 2010, 157, H796H800.
[11] Andeen, D.; Loeffler, L.; Padture, N.; Lange, F. Journal of Crystal Growth
2003, 259, 103 109.
[12] Andeen, D.; Kim, J. H.; Lange, F. F.; Goh, G. K. L.; Tripathy, S. Advanced
Functional Materials 2006, 16, 799804.
[13] Richardson, J. J.; Lange, F. F. Crystal Growth & Design 2009, 9, 25702575.
[14] Thompson, D. B.; Richardson, J. J.; DenBaars, S. P.; Lange, F. F. Applied
Physics Express 2009, 2, 042101.
[15] Richardson, J. J.; Koslow, I.; Pan, C.-C.; Zhao, Y.; Ha, J.-S.; DenBaars, S. P.
Applied Physics Express 2011, 4, 126502.
[16] Richardson, J. J.; Estrada, D.; DenBaars, S. P.; Hawker, C. J.; Campos, L. M.
J. Mater. Chem. 2011, 21, 1441714419.

26

Chapter 3
Undoped ZnO current-spreading
layers on c-plane LEDs

This chapter is largely based on a paper by Reading, Richardson, Pan, Nakamura,

and DenBaars [1] that was published in Optics Express.

3.1

Introduction

Prior work on c-plane GaN/InGaN LEDs has made use of both Ni/Au [2, 3] and
ITO current spreading layers [4, 5, 6] . This is due to the high transparency and
conductivity of these layers. These layers are commonly deposited by sputtering
or by electron beam (e-beam) deposition. Prior work has also found that both
Ni/Au, and ITO can form low-resistance Ohmic contacts to p-type GaN, and can

27

be lithographically patterned by wet-chemical etching in dilute HCl solutions. ITO

is the material that is most widely used in commercial devices, especially as
current-spreading layers in broad-area LEDs, as laser cladding layers, and in other
applications.
There have also been efforts to grow ZnO layers on InGaN/GaN LEDs as
transparent current spreading layers[7]. This could provide a low-cost alternative
to ITO if the material properties and deposition methods allow. However, these
attempts have previously suffered from high operating voltage as a result of the
poor electrical contact between the p-type top most layer of the III:nitride epitaxial
layers, and the current spreading layer. One of the aims of this study is to find
deposition and processing conditions under which a ZnO current spreading layer
can form good, Ohmic electrical contact to an underlying p-GaN layer, while
performing well as a current spreading layer.
The annealing and processing conditions of the ZnO layers were optimized first,
followed by the successful fabrication of LED devices. These devices were then
characterized and compared to a reference device with a conventional ITO current
spreading layer.

28

3.2
3.2.1

Experimental methods
Epitaxial ZnO growth on c-plane GaN-on-sapphire

Seed layer growth

For this study, ZnO films were grown using the two-step growth method described
in Kim et al. [8]. This method involves the deposition of a seed-layer onto the
substrate, followed by an epitaxial second-layer growth. This two-step growth is
advantageous because it allows the thin 200nm seed layer to be annealed at higher
temperatures that would cause a thicker film, for example a 1m film to
delaminate from the substrate. The final film (seed layer + 2nd layer) is then
annealed again after the second-layer growth.
Before depositing the seed layer, the wafers are cleaned in acetone, ethanol, and
DI water, using a sonicating bath. For GaN LED wafers, an HCl dip is performed
by fully immersing the wafer in 1:1 HCl:H2 O solution for 10 minutes. All labware
that touches the growth solution is cleaned prior to the preparation of growth
solution using dilute nitric acid (1:4 HNO3 :H2 O). The following solution is then
prepared in sealable PTFE pressure vessels in order to deposit seed layers on each
sample:

24mL DI 18M H2 O (Millipore system)

0.15g Zn(NO3 )2 :6H2 O (Sigma-Aldrich, 99% purity)

29

 0.6g NH4 NO3 (Sigma-Aldrich, 99% purity)

Creating and storing a stock solution is not recommended, as CO2 from the air can
dissolve in the solution and form carbonate precipitates. It is best to prepare fresh
growth solution every time. After the salts are fully dissolved, the solution is
sealed into a PTFE pressure vessel and placed in an oven at 90C for 2-3 hours,
until it reaches the ambient temperature. The sealed vessel is then removed from
the oven. The sample wafer is then placed into the solution, and suspended in such
a way as to not be touching any of the walls of the vessel. This can be
accomplished by using a small section of PTFE tubing that has had small notches
cut into the side to hold a sample. The wafer must be face-down to limit the
amount of powder that falls onto the surface of interest during the seed layer
formation. After the sample has been placed into solution, 0.7mL of 2.9% NH4 OH
is added, using a PTFE syringe, to rapidly raise the pH of the solution, which
immediately lowers the solubility of the Zn2+ complexes and creates a driving
force for the nucleation of ZnO crystallites. It is very important during this stage to
be agitating the growth solution so that the pH change takes place everywhere in
the solution, and nucleates evenly across the surface of the wafer. The sealed
vessel is then put back in the 90C oven for several hours, preferably overnight to
complete the growth. Samples are then removed from the solution, rinsed, dried,
and inspected.
A well-formed seed layer generally will have a distinct iridescent rainbow-like
sheen that is visible to the naked eye. This is due to interference effects in the
200nm crystallites on the surface of the off-white p-GaN. The seed layers were
30

annealed in an AET RTA system at different temperatures, in order to find one

with the lowest contact resistance. This was found to be at 500C, for 15 minutes
in an N2 atmosphere. These annealing conditions were subsequently used for all
seed-layer anneals. The results of this study are shown in Figures 3.4.

Film growth

After the seed-layer has been annealed, a growth solution is prepared, which is
designed to epitaxially grow the ZnO film with no further nucleation of ZnO in the
solution. In this way, the film will gradually get thicker over time, rather than
nucleating everywhere at once, as is the case for a seed-layer growth. The growth
solution is prepared as follows:

24mL DI 18M H2 O
0.05g Na3 C6 H5 O7 (Trisodium citrate, Sigma-Aldrich, 99% purity)
0.1g Zn(NO3 )2 :6H2 O (Sigma-Aldrich, 99% purity)

After the salts have been dissolved, the pH is raised to 11.6 by the addition of
1.25mL of 29% NH4 OH using a PTFE syringe. Initially, some precipitates will be
visible in the solution, but will quickly disappear with some gentle agitation. After
the solution becomes completely clear, the sample wafer is again suspended in the
solution using the PTFE tubing, face down, such that it is not touching the sides of
the PTFE vessel. In this type of growth, the driving force for ZnO crystallization
results from raising the temperature of the solution steadily to 90C. This is
31

accomplished by sealing the PTFE vessel tightly (to prevent evaporation of the
ammonia in the solution), and placing it in the oven overnight, or for about 12
hours.
These growths take a long time, because the PTFE vessels have poor heat transfer
in this case, and heat transfer is only slowly taking place through convective
ambient heating of the vessel exterior. It should also be noted that this method has
relatively poor control over the thickness of the resulting ZnO film, because the
growth is limited by the total available amount of Zn2+ species in the growth
solution, not the total growth time. Another disadvantage of this method, is that
the growths take such a long time, mostly because of simple heat transfer issues.

3.2.2

Circular Transmission Line Method

The transmission-line method, or TLM method is widely used in electronics

engineering to measure the sheet resistance of a layer separately from its contact
resistance to an underlying layer. The procedure works by forming a TLM
structure an isolated bar of material with contact pads at different spacings.
Figure 3.1(a)shows a diagram of such a structure. Current-voltage sweeps are then
measured with different pad spacings, which give linear curves in the case of
Ohmic contact. The slope of each linear current-voltage curve is the total
resistance of 2 contact pads and the underlying transmission line with varying
length. By plotting these total resistances vs. pad spacing, it is possible to
differentiate the contact resistance of the pads from the sheet resistance of the

32

Figure 3.1: Diagrams of a TLM structure, with metal in gold, ZnO in cyan, and
p-GaN in red. (a) a linear TLM structure that could be used to measure metal/ZnO
contact resistance and ZnO sheet resistance. (b) Example plot of resistance vs.
contact spacing used to determine contact and R sh . (c) Diagram of a CTLM structure that could be used to measure ZnO/p-GaN contact resistance and p-GaN sheet
resistance.
underlying transmission line. Figure 3.1(b) shows an example plot of these total
resistances with pad spacing. The sheet resistance, R sh is calculated from the slope
of the line by the equation,
R sh = w

dR
dx

(3.1)

where w is the width of the transmission line, R is the total resistance between
pads, and x is the distance between contacts. When the slope of this line, is
calculated, it can be extrapolated to zero spacing between contact pads, at which
point the value is equal to twice the contact resistance. Thus, the contact resistance
of the pads and the sheet resistance of the transmission line can simultaneously

33

estimated using a linear fit of the form,

R f it = 2Rcontact + R sh x

(3.2)

The specific contact resistivity, with units of cm2 is an intrinsic property of the
interface of two layers, and is independent of the area of the contact. This quantity
is calculated by,
c = Rcontact Ac

(3.3)

, where Ac is the area of the contact pad. In the case of a circular transmission line
structure, shown in Figure 3.1(c), it is necessary to multiply each total resistance
value by a circular correction factor in order to be able to use a linear fit. This
corrects for the geometrical difference between a linear TLM, and a circular TLM
geometry. The correction factor is a function of the outer diameter of the center
contact pad, and the inner diameter of the outer contact pad, R1 and R2 ,
respectively, as shown in Figure 3.1(c). After this correction, the problem can be
treated similarly to the case of a linear TLM.

3.2.3

LED processing for devices on c-plane sapphire

substrates

The device fabrication of LEDs involved 3 basic steps: etching the LED mesas,
patterning the n-contacts, and patterning the p-contacts. Reference LEDs were
34

also manufactured with ITO current spreading layers, which were deposited at
room-temperature in an electron-beam evaporation system at 3.0/s. The ITO
appeared black after deposition at room temperature, and remained so until the
transparency annealing later in the process. The thickness of the ITO layer was
225nm, and was verified by profilometry measurement. A mask was designed with
4 types of LEDs with different lateral contact spacings, in order to study the effect
of current-spreading on the performance of the LEDs. The main features of this
mask design are shown in 3.8. To aid with cleaving later on, it is important to align
the mesas with the crystallographic orientation of the underlying sapphire
substrate. For the mesa etch step, positive photoresist (AZ SPR220-3.0) was spun
onto the wafers for 30s at 3000rpm, to produce a photoresist thickness of
approximately 2.7m. A pre-exposure bake was applied at 115C for 60s on a
hot-plate. The photoresist was exposed in a Karl Suss MJB3-UV400 contact
aligner for 23s, and developed using AZ300MIF photoresist developer for 55s.
These times have been found through experience to give best results in this
process. After exposure, this particular type of photoresist must sit undisturbed for
5 minutes before developing.
After the mesa lithography was finished, the films underwent a 30s O2 plasma
descum step in a Technics PE-IIA O2 plasma cleaning system, with 100W RF
power, at 300mT O2 pressure, to remove any residual photoresist from the surface
of the ZnO layer. This step is essential, because a thin layer of residual photoresist
will confound efforts to remove the ZnO layers from the p-GaN layers via
wet-chemical etching.

35

The next step was to remove the ZnO and ITO from the p-GaN surface away from
the mesa areas by etching in dilute acid. This was one of the more problematic
steps, because of the possibility of etch undercut, where the acid will dissolve ZnO
underneath the photoresist. This leads to a smaller effective device area, and, if
gone too far, can cause all the ZnO to be completely removed. The ITO was etched
using the same solution with no issues. The etch solution used was a mixture of
HCl:H2 O2 :H2 O in a ratio of 1:1:20. The samples were immersed in the solution
for 10 seconds at a time, and were checked either visually, or in an optical
microscope, to see whether the ZnO had been fully removed. During the etching,
the ZnO film becomes very rough, causing it to appear very dark in an optical
microscope, or hazy-white to the naked eye. When the smooth p-GaN is visible, it
is very obvious, as the p-GaN has mirror-like smoothness which appears bright
white in the microscope. If the ZnO is not fully etched away, the device processing
is highly likely to fail due to rough n-contacts. The etch-rate in this solution for
undoped, non-annealed ZnO is approximately 50-100nm/s. It should also be
mentioned that the thicker the ZnO film is, the more difficult this step is likely to
be, especially if the ZnO layer has been post-annealed, which reduces the etch rate.
The mesas were formed using a Panasonic E640 Inductively-Coupled Plasma
(ICP) etching system. The etch conditions had previously been developed by
UCSB researchers, and were comparable to those in Vartuli et al. [9], with an etch
chemistry of 80%Ar and 20%Cl2 . The sapphire LED wafers were mounted on a
4-inch Si carrier wafer using small droplets of vacuum oil to allow good
conduction. The etch rate for GaN was estimated to be 400600nm per minute.
The wafers in this study required a 1.5 minute etch to allow contact to the highly
36

Figure 3.2: LED fabrication process. Note that there is always some undercut,
which can be up to approx. 10m after the ZnO wet-etching step. The metal stacks
used were 10/100/10/500nm Ti/Al/Ni/Au for n-contacts, and 10/500nm Ti/Au for
the p-pads.

37

conductive n-type GaN layer. The samples were then unmounted and cleaned in
acetone, isopropanol, and deionized water.
After mesa etching, the current spreading layers were annealed in a Rapid Thermal
Annealing (RTA) system. The ITO layers were annealed according to Vampola
et al. [6], for 10min in N2 /O2 ambient followed by 3min in N2 ambient. After this
anneal the ITO films appeared clear. The ZnO films were annealed at 225C in N2
ambient, as this was found to lower their sheet resistances, as shown in Figure 3.5.
The n-metal electrodes were patterned using negative AZ nLof-2020 photoresist.
The resist was spun onto the wafers at 3000rpm for 30s, to give a photoresist
thickness of approx. 1.8m. For the nLof-2020 lithography, an exposure time of 6s
was used. A pre-exposure bake at 110C on a hot-plate was used, followed by a
60s post-exposure bake at 110C. The resist was developed for 50s in AZ300MIF
developer. Prior to metal deposition, the wafers were exposed to an O2 plasma
descum, followed by a 30s HCl dip in 1:1 H2 O:HCl solution in order to remove
any residual oxide on the n-GaN surface. Ti/Al/Ni/Au contact were chosen
because they were previously found to have very low-resistance Ohmic contact to
n-type GaN [10]. The metal thicknesses were 10/100/10/500nm, respectively, and
were deposited at a pressure below 3.0 106 Torr in an electron-beam evaporation
system. The metal liftoff was done in heated AZ 1165 photoresist stripper solution
and left overnight on a hot-plate at 80C.
The p-metal electrodes were deposited in an identical way to the n-metal
electrodes, with the exception that there was no HCl dip, since the concentrated
HCl would immediately etch away the ZnO contacts. It was found that a simple O2
38

plasma descum prior to metal deposition allowed for excellent contact to both ZnO
and ITO current spreading layers using a 10/500nm Ti/Au metallization. This had
also been shown previously for ITO layers by Margalith et al. [4].
Optionally, the ZnO layers could then be roughened by covering the entire device
in photoresist, except for the ZnO areas, and immersing the solution in very dilute
HCl. The dilution used was 1:100 HCl:H2 O. It was found that the surface of
c-plane ZnO was fully roughened after 30 seconds of etching. However, on the
LEDs used in this study, the roughening did not have an appreciable effect on
light-extraction because of the underlying patterned sapphire substrate, which
already allowed for good light-extraction from the wafer.

3.2.4

LED packaging for devices on c-plane sapphire substrates

After the LED wafers have been processed, it is necessary to connect the device
to an outside circuit using a package structure. The main steps involved for this
type of mesa-LED are thinning the substrate, polishing the backside, cleaving the
die apart using a diamond wafer scribe, bonding the die onto a reflective header
via gold wire, and finally, encapsulating the LED chip in a silicone dome to
improve the light extraction from the chip.
Sapphire substrates require several hours of grinding in order to thin the roughly
400m thick substrate down to a thickness of approximately 180m, which is thin
enough to be cleaved with high reliability. First, the wafer is mounted using wax,
face-down, on a ceramic plate, for use with a Tecdia abrasion system. The sample
39

must be mounted face-down using mounting wax. This is done by heating the
ceramic plate on a hot-plate and melting a small amount of wax onto the top
surface. To prevent scratching, a piece of weighing paper is put in the melted wax,
and more wax is melted onto the weighing paper. The sample wafer is then placed
face-down on the weighing paper. A razor blade can then be used to remove all
wax and weighing paper from everywhere around the sample, as well as removing
all wax from the backside of the sample. During this step it is essential that the
wafer surface is continuously lubricated, and that a slow feed rate is selected. If
the grinding tool is thinning the wafer too slowly, it must be re-surfaced using a
whetstone block. During this step, the thickness of the wafer can be monitored
using a wafer thickness gauge.
After thinning the wafer to 160-180m, it is necessary to polish the wafer so that
clean scribes can be made across the back surface using the wafer scriber tool.
This was done on a polishing wheel at 24rpm using a 1m Hyprez diamond slurry,
for 10-15 minutes. The wafer is still thinned appreciably during this step, so it is
important to monitor the initial thickness during grinding. The wafers are then
unmounted from the ceramic polishing plate by heating on a hotplate at 100C
until the mounting wax completely liquefies. At this point the wafers are fragile
and must be handled with care. The samples should then be soaked in acetone to
dissolve all residual wax, followed by a sequential solvent cleaning using acetone,
isopropanol, and DI water, where the sample is fully immersed during each step. If
the wax is not completely removed, the wire-bonding step is unlikely to be
successful, and can damage the wire-bonding tool due to soot accumulation.

40

After polishing the backside of the wafers, they are prepared for scribing. The
method used here is to cut a square hole into a sheet of weighing paper, placing a
small sheet of adhesive plastic facing inwards through the hole. The protective
clear plastic from the adhesive sheet can be used to cover the wafer surface for
protecting the samples during storage. This should leave a window in the center
with an adhesive surface. The wafer is placed face-down on this surface leaving
the backside exposed for scribing. Wafer scribing was done on a Tecdia grinding
tool using a diamond-studded tip. The pressure used for a sapphire wafer was
approx. 80-120 units on the tool pressure gauge, and usually about 3-4 scribes
would cause the wafer to cleave reliably. If the wafer did not get sufficiently
thinned during polishing, it is possible to eventually get the wafers to cleave using
higher pressures and repeated scribing, but this will quickly wear out the
diamond-studded scribing tips, in the case of sapphire substrates.
The next step in the packaging is to bond the individual LED chips onto a silver
header. A header is a piece of reflective metal with 2 isolated leads that can be
easily wire-bonded to the LED chip. For this step, it is necessary to have a small
amount of GE silicone already prepared. A small droplet of the silicone is placed
onto the center of the header using a toothpick. Then, under a microscope, an
individual LED chip can be picked off the adhesive tape and gently floated onto
the surface of the droplet, using surface tension to prevent from being fully
immersed. It is vital that the LED chip does not become fully immersed in the
silicone, because this will make wire-bonding to the device impossible. After the
LED chips have been placed on their headers, the header-plus-chip assemblies are
placed into an oven at 120C to cure, for 30 minutes.
41

Figure 3.3: White LED package. Due to the poor thermal properties of this package, it was necessary to do pulsed measurements for measurements at high currentdensity. However, the encapsulation provides a substantial boost as compared to an
unencapsulated LED chip.
After the silicone has cured, the samples must be wire-bonded to the header leads.
This step is often highly troublesome, and it is common for metal pads to become
delaminated, or for the wire-bonds to become disconnected. Best results were
obtained using the lightest-contact setting, and moving the bonding tip extremely
slowly and gently at all times. Samples were all bonded using gold wire. At this
point, the LEDs can be tested, but are highly fragile, and must be handled very
carefully, but it is possible to test them at this point.
Encapsulation of the devices is very important for obtaining good light-extraction
and for being able to compare LEDs directly to each other. An LED sample left
unencapsulated is also likely to become damaged and inoperable over time. Some
of the research at UCSB employed a so-called transparent stand package, which
involved a strip of ZnO single-crystal substrate wafer, made by dicing a ZnO
42

substrate, and then mounting the LED chip onto the side of the substrate, so that
the chip is emitting light from further away from the header [11]. This was found
to increase light-extraction. However, this procedure was not used for this work, as
it was not found to be reproducible, and did not give isotropic emission in all
directions. For this study, all LEDs were encapsulated in simple silicone domes,
made by immersing the LED header head-first into a brass mold filled with liquid
silicone, and left to cure for 3 hours in an oven at 120C.
For making white LEDs, it is necessary to combine blue-wavelength emission
from the LED chip with broad yellow and red emission from various phosphor
materials. The most commonly used phosphor was YaG: Ce, which emits yellow
light, along with some proprietary red phosphor materials from Mitsubishi
Chemical Co. The white-LED package consisted of a regular silicone dome with a
phosphor-impregnated silicone cap surrounding the dome, as shown in Figure 3.3.
These are the same type of structures that were modeled by Lalau Keraly,
Kuritzky, Cochet, and Weisbuch [12] using LightTools ray-tracing software to
calculate their light-extraction efficiencies (LEE).

3.2.5

LED testing

LEDs were usually tested on-wafer before packaging using a probe station. The
main information that could be gathered before packaging were the I-V
performanace, and to test whether there was any leakage in the devices. Leakage
can be caused by many different things, for example growth defects, bad metal

43

contact alignment, or a faulty lithography step. However, on-wafer testing is

usually more of an error-check for the LED process, since a real LED device must
be isolated and wire-bonded to a package. It was generally found that a device that
was wire-bonded displayed lower resistance than would be indicated by an
on-wafer test, most likely due to the lack of probe resistance.
The optical power output of the LEDs was measured in an integrating sphere
setup. An integrating sphere is simply a large, spherical cavity with a
non-absorbing, white BaSO4 coating on all walls. In the center of the cavity there
is a stand with electrical connections so that an LED can be operated inside the
sphere while no external light is able to enter the cavity. An optical fiber with a
monochromator and photodetector allow light to be collected from the sphere, in
such a way that the output signal is directly related to the output power of the
device. This allows the signal to be calibrated to calculate the output power of the
device. The output signal can also be analyzed based its spectral characteristics.
The SpecWin light software package also allows calculation of white-LED
metrics, such as total lumens, luminous efficacy, color-rendering index, color
temperature, and the CIE chromaticity coordinates.
DC measurements and the I-V curves were obtained using a Keithly 2420 Source
meter controlled by a Labview computer program. These were generally done
prior to complete encapsulation to mitigate heating effects during continuous
operation without a heat sink. Pulsed measurements are very important because
the packaging scheme described above does not allow for good heat transfer out of
the LED chip, and the chip is completely surrounded by insulating silicone. The

44

only heat transfer possible is through the gold wire leads themselves, which leads
to heating of the LED chip during operation. By pulsing the input signal as a
square wave with only a 5-10% duty cycle, it is possible to measure the light
output without heating the chip, and without having to construct elaborate
heat-sinking mechanisms. Pulsed measurements are made using a pulse generator
and an oscilloscope to monitor the input current. The input current was monitored
by measuring the voltage across a 50 resistor that was connected in series with
the LED testing stage.
In order to measure the lateral current spreading ability of the ZnO and ITO layers,
images were acquired using a confocal microscope, with an attached CCD camera.
Since the light from the LED chips was so bright, it was necessary to use both a
neutral-density filter, and a polarizing filter, to prevent the CCD camera from being
saturated. Even with this optical filtering, it was necessary to operate only at low
currents in order to measure the current-spreading performance. Current spreading
performance was measured on a special high-aspect-ratio LED which exacerbate
the current-spreading problem, and allowed estimation of the characteristic current
spreading lengths.

45

Figure 3.4: CTLM measurements of the ZnO/p-GaN contact resistance showed that
annealing at 500C gave the lowest contact resistance at 7.0 102 cm2

3.3
3.3.1

Characterization of ZnO films

Thermal Annealing of Undoped Films on c-plane GaN

Seed layer annealing

It was also noted that undoped ZnO films with as-grown seed-layers would often
fully delaminate from the p-GaN surface during growth, indicating poor surface
adhesion. This made the seed-layer annealing a necessary step in forming
high-performance films. ZnO contacts to p-GaN are asymmetrical tunnel
junctions, which depend strongly on there being a high concentration of carriers in
each material, as well as having an abrupt interface between the p and n sides of
the junction. For this reason, it is theorized that the carrier concentration on either
side of the junction must be high in order to form a good contact. The CTLM
46

measurements shown in Figure 3.4 showed 2 main trendsa decrease in contact

resistance after annealing at temperatures up to 500C, but a subsequent increase
after annealing at temperatures greater than 500C.
Shi et al. [13] reported that ZnO films that were unintentionally doped with
hydrogen showed an activation of hydrogen into a shallow-donor state after
annealing at temperatures greater than 400C, which could be an explanation for
the improved contact behavior, due to the sensitivity of a tunnel junction to carrier
concentration. Zhang and Brodie [14] showed that annealing undoped ZnO films
above 500C generally caused an increase in resistivity due to the material
transitioning from a degenerately-doped n-type semiconductor, to a thermally
activated wide-bandgap semiconductor after annealing in air at temperatures
upwards of 500C. It is suspected that when annealing undoped layers higher than
500C, there is a decrease in the carrier concentration, which would explain the
increased contact resistance under these annealing conditions.

Annealing fully-coalesced films

CTLM measurements on the fully-coalesced ZnO films showed a drastic decrease

in sheet resistance after annealing above 200C. It has been shown that annealing
of ZnO generally leads to an increase in the crystal quality, and therefore the
carrier mobility of the films. Song et al. [15] showed that the FWHM of the crystal
orientation is already decrease by annealing at temperatures as low as 200C,
which could be a cause of the increased sheet resistance. Laurent et al. [16]
conducted a study of the effect of annealing on undoped ZnO films grown by
47

Figure 3.5: Effect of post-annealing on the sheet-resistance of ZnO films on c-plane

GaN, measuared using the CTLM method. Error bars show the std. deviation of the
measured samples
electrodeposition, and observed that a photoluminescence peak at 350nm, which is
associated with Zn(OH)x completely disappears after annealing in Ar at 200C. At
the same time, they noticed an increase in the ZnO PL peak at 368nm, which
suggests an increase in the crystal quality. This could be due to residual Zn(OH)x
in solution-deposited ZnO films becoming fully dehydrated to ZnO after annealing
at 200C. Annealing the films at higher temperatures led to cracking and haziness
of the ZnO films. The annealing of the seed layers did not appear to have any
systematic effect on the sheet resistance of the final film.

48

Figure 3.6: XRD characterization of a ZnO film grown on a commercial

GaN/InGaN LED wafer on a sapphire substrate. This film had a seed layer annealed at 500C and a post-deposition anneal at 250C for 15 minutes. (a) 2-scan
showing ZnO and GaN (002) reflections. (b) -scan of ZnO (002) reflection, show reflections, and (d)
ing a FWHM of 0.365. (c) -rotational scan of ZnO (1011)

-rotational scan of GaN (1011) reflections, confirming heteroepitaxial relationship

between ZnO and GaN layers.

49

3.3.2

XRD characterization

Figure 3.6 shows X-ray diffraction results from ZnO films grown on MOCVD
GaN layers on c-plane sapphire substrates. X-ray data was gathered on a Philips
XPert PRO X-ray diffractometer in triple-axis mode, with Cu-K radiation. The
2 scan shown in Figure 3.6(a) shows the wurtzite GaN and ZnO (0002)
reflections, indicating they are both aligned in the same direction. No additional
phases were observed. However this is not sufficient to confirm a heteroepitaxial
relationship between the two films. Many reports have shown polycrystalline ZnO
films with preferred (0002) orientation, in films deposited by sputtering [17],
aqueous solution growth [18], sol-gel methods [19]. This is simply due to the
lower surface energy of the close-packed Zn-terminated basal (0001) plane of the
wurtzite crystal structure. For our purposes it is important to confirm that the films
are entirely single-crystal, with the grains all sharing the rotational () crystal
alignment to the underlying GaN layers. In Figures 3.6(c) and (d), we see that this
is the case, as the 6-fold symmetry of the (0002) surface is identical in both the
GaN and ZnO layers. This is in agreement with previous results by Kim et al. [20].
Finally, it was found that there was a considerable amount of mosaicicity in the
ZnO films, with a FWHM in an scan being 0.365. This is due to the defective
nature of the thick heteroepitaxial films.

50

Figure 3.7: SEM micrographs showing formation of roughened c-plane ZnO surface by etching in 1:100 diluted HCl solution.

51

3.3.3

SEM characterization

Roughening of semiconductor surfaces is of much interest for increasing the

light-extraction capability of an LED chip. Simple, highly-symmetrical geometries
allow for the possibility of multiple internal reflections, which causes much of the
light to be reabsorbed by the LED quantum wells, or by other absorption
mechanisms in the LED die. One advantage of having thick current spreading
layers is that they can be roughened without any appreciable decrease in their
family of planes exhibits
current-spreading ability. In wurtzite ZnO, the {1011}
stability during wet, and dry-etch processes. This suggests it is possible to apply
mild surface treatments to a ZnO surface to cause roughening.
Figure 3.7 shows scanning electron microscopy (SEM) images of roughened of
c-plane ZnO layers grown on GaN LEDs. The surfaces were roughened by
dipping the sample in a highly-diluted solution of HCl (1:100) concentration, for
times up to 1 minute. It was found that the surface already showed some change
after 5 seconds of etching, with small etch pits beginning to form. As the etching
time increased, the film became increasingly rough, and eventually did not appear
to change with roughening time after about 35s.

52

Figure 3.8: LED IV testing results from devices with ZnO and ITO current spreading layers. The ITO layer had a thickness of 240nm, and a sheet resistance of
91/. The ZnO current spreading layers were 2100-2300nm thick and had a
sheet resistance of 26/.

3.4
3.4.1

LED characterization
I-V performance

The bulk resistivity of the ZnO and ITO films was calculated from the sheet
resistance and the thicknesses, and are shown in Table 3.1. It should be mentioned
at this point that there have been reports of ITO with substantially higher
conductivities, by depositing the films at elevated temperatures. However, for our
purposes these layers should suffice as an initial comparison for the ZnO layers.
LEDs with different contact spacings were tested at currents up to 350mA, as is
shown in Figure 3.8. It was found that the devices with ZnO current spreading
layers had lower operating voltages than ITO current spreading layers. This held
true for devices with all contact geometries that were tested. Also, the devices with
larger contact spacings showed a larger discrepancy between the ITO and ZnO
operating voltages, implying the spreading resistance was considerably higher for
the ITO layers. The average current densities can be calculated from the current

53

Table 3.1: Electrical properties of deposited current spreading layers calculated

from CTLM measurements. Thicknesses were measured using a Dektak profilometer.
Material
ZnO
ITO
Thickness (nm)
2200
240
Sheet Resistance (/)
26
91
3
2.28 103
Bulk Resistivity (cm) 5.72 10
Conductivity (S/cm)
174.8
439.5

Figure 3.9: External quantum efficiency (EQE) of the 4 types of LEDs. The dominant wavelength for these devices was at 470nm. Pulsed measurements were made
with a 10% duty cycle.
and the mesa areas, which were 0.97, 1.02, 1.07, and 1.09mm2 for Type 1, 2, 3,
and 4 devices, respectively. A drive current of 350mA corresponded to current
densities of 36.1, 34.3, 32.7, and 32.1 A/cm2 on the 4 types of devices tested. The
difference in areas was due to the increased wafer area consumed by the n-contacts
in the devices with lower contact spacing.

3.4.2

LED efficiency

In Figure 3.9 we show how the current spreading material and the device geometry
changes the peak efficiency and droop behavior of the devices. All devices showed
substantial decrease in efficiency at higher currents, a well-known phenomenon,
54

Figure 3.10: Power efficiency of the 4 types of LEDs. This metric combines the
effects of quantum efficiency with the effects of resistive power dissipation.
particularly in devices grown on c-plane GaN. However, when there is a large
amount of efficiency droop at high current densities, it is important to have good
current spreading. If the contacts are spaced too far apart for the current to be
spread across the mesas, there is generally current-crowding, where most of the
current is injected underneath or near the p-side top metal contacts. Since most of
the current is being injected within a small region of the device, the effective
current density can be much larger, which will exacerbate the internal quantum
efficiency (IQE) droop. This effect is most pronounced in the Type 4 devices, with
the largest contact spacing.
The power efficiency measurements shown in Figure 3.10 show how the effects of
higher voltage, and lower current-spreading caused the Type 4 ITO devices to have
drastically decreased efficiency, compared to the devices with ZnO current
spreading layers. The performance was much more similar in the Type 1 devices
with closely spaced contacts, again confirming the role of current-spreading in the
performance of the devices. The Type 1 devices also had very similar I-V
behavior, as was shown in Figure 3.8.

55

Table 3.2: Luminous Efficacy comparison between Type 3 LEDs fabricated with
the phosphor cap structure shown in Figure 3.3. Measurements were done using a
10% duty cycle to avoid irreproducible heating effects.
Current Density (A/cm2 ) 0.5
2.0
10
35
Luminous Efficacy: ZnO (lm/W) 157 142
124 84.8
Luminous Efficacy: ITO (lm/W) 126 115 93.0 56.2
Color Temp.: ZnO (K) 4496 4479 4466 4442
Color Temp.: ITO (K) 4458 4466 4443 4437
CRI: ZnO 65.4 65.2 65.4 65.2
CRI: ITO 67.1 66.0 64.9 65.2
The luminous efficacy of the films is shown in Table 3.2. Luminous efficacy
depends on many factors, as it is easy to achieve high-efficacy devices with poor
color-rendering or a high color temperature. The CRI of these devices is quite low,
as the phosphor cap was calibrated for use with a 450nm blue emitting LED,
instead of the 470nm dominant wavelength that was observed for these devices.
However, since the same exact phosphor cap was used for all devices, it is possible
to directly compare the performance of the resulting white LEDs. The peak
efficacy of the devices with ZnO current spreading layers was 24% higher than that
of the device with an ITO current-spreading layer. This peak performance was at a
fairly low current density (0.5A/cm2 ). At the highest current density tested,
35A/cm2 , the ZnO current spreading layer outperformed the ITO layer by 50%.

3.4.3

Current spreading performance

The degree of current spreading in the devices was directly observed by operating
high-aspect-ratio LEDs under an optical microscope. By operating the CCD in

56

grayscale mode, it was possible to obtain a signal that was proportional to the
power emitted by the chip. Images of the chip were colored by their intensity to
show the degree of current spreading, shown in Figure 3.11(b). The light-intensity
distribution across the wafer is plotted in Figure 3.11(c)&(d). From this figure it
becomes clear that the enhanced current spreading of the ZnO layer allows much
more light to be emitted away from the top-metal contacts, which leads to more
light escaping the chip. In addition, the average current density is lower, due to the
current spreading more evenly across the chip. This also explains why the
outperformance of the ZnO samples was greater in devices with greater contact
spacing.

3.4.4

LEDs with etch-roughened ZnO CSLs

LEDs with roughened ZnO current-spreading layers showed higher output powers,
with films roughened in 1:100 HCl for 10s showing a 13% enhancement as
compared to LEDs with unroughened ZnO CSLs. After roughening for 40s there
was a large increase in the operating voltage due to thinning of the ZnO layers,
which would increase their sheet resistances. The film morphologies are shown in
Figure 3.7. The I-V and EQE performance of LEDs with roughened ZnO CSLs is
shown in Figure 3.12. Unfortunately there is no ITO comparison available in this
case, due to sample loss during sapphire wafer thinning step.

57

Figure 3.11: Current spreading characteristics demonstrated using a high-aspectratio LED. (a) optical microscope image, with the p-contact on the bottom. (b)
Current spreading performance at currents from 1 to 10mA that have been colorized
to show the extent of current spreading. (c) Light-intensity distribution for the ZnO
CSL. (d) Light-intensity distribution for the ITO CSL.

58

Figure 3.12: (a) IV curves of c-plane LEDs with ZnO CSLs roughened for different
times in dilute HCl. After 40s the films had thinned considerably which resulted in
higher-voltage operation. (b) %EQE of LEDs with roughened ZnO CSLs.

3.5

Summary and Conclusion

In summary, we have demonstrated that ZnO films deposited by aqueous solution

deposition could be a viable competitor to conventional ITO for use as
current-spreading layers, particularly in the case of GaN/InGaN LEDs, where
single-crystal ZnO films can be used, as was confirmed by X-ray diffraction.
Annealing conditions were optimized for the ZnO nucleation layers, as well as for
the final films. LEDs fabricated from a single epiwafer with different contact
geometries confirmed that devices with ZnO current-spreading layers had lower
operating voltages, as well as higher light-output, which, combined, caused the
devices with ZnO current spreading layers to have considerably higher luminous
efficacies24% greater at peak efficiency, and 50% greater at higher drive currents.
The bulk conductivity of the ZnO layers were found to be 174.8 S/cm, compared
59

to 439.5 S/cm for the ITO layers. However, the ZnO films could be much thicker,
due to their high transparency. This enabled the ZnO layers to have lower sheet
resistance26/ for a 2200nm ZnO layer vs. 91/ for a 240nm ITO layer.
Additionally, it was found that the ZnO films could be roughened using a simple
wet-etching procedure. However, due to the already-high light-extraction
efficiency from patterned sapphire chips, this effect did not lead to increased
performance in a systematic way. It is expected that roughening methods would
play an important role in other types of LED chips, for example in the case of
GaN-on-Si or on bulk GaN substrates.

60

References
[1] Reading, A. H.; Richardson, J. J.; Pan, C.-C.; Nakamura, S.; DenBaars, S. P.
Opt. Express 2012, 20, A13A19.
[2] Ho, J.-K.; Jong, C.-S.; Chiu, C. C.; Huang, C.-N.; Chen, C.-Y.; Shih, K.-K.
Applied Physics Letters 1999, 74, 12751277.
[3] Sheu, J. K.; Su, Y.; Chi, G.; Koh, P. L.; Jou, M.; Chang, C. M.; Liu, C. C.;
Hung, W. C. Applied Physics Letters 1999, 74, 23402342.
[4] Margalith, T.; Buchinsky, O.; Cohen, D. A.; Abare, A. C.; Hansen, M.;
DenBaars, S. P.; Coldren, L. A. Applied Physics Letters 1999, 74,
39303932.
[5] Kim, D.; Sung, Y.; Park, J.; Yeom, G. Thin Solid Films 2001, 398, 8792.
[6] Vampola, K. J.; Fellows, N. N.; Masui, H.; Brinkley, S. E.; Furukawa, M.;
Chung, R. B.; Sato, H.; Sonoda, J.; Hirasawa, H.; Iza, M.; DenBaars, S. P.;
Nakamura, S. Phys. Stat. Solidi (a) 2009, 206, 200202.
[7] Thompson, D. B.; Richardson, J. J.; DenBaars, S. P.; Lange, F. F. Applied
Physics Express 2009, 2, 042101.
61

[8] Kim, J.-H.; Kim, E.-M.; Andeen, D.; Thomson, D.; DenBaars, S. P.;
Lange, F. F. Adv. Funct. Mater. 2007, 17, 463471.
[9] Vartuli, C. B.; MacKenzie, J. D.; Lee, J.; Abernathy, C.; Pearton, S.; Shul, R.
Journal of Applied Physics 1996, 80, 37053709.
[10] Fan, Z.; Mohammad, S. N.; Kim, W.; Aktas, O.; Botchkarev, A. E.;
Morkoc, H. Applied Physics Letters 1996, 68, 16721674.
[11] Pan, C.-C.; Koslow, I.; Sonoda, J.; Ohta, H.; Ha, J.-S.; Nakamura, S.;
DenBaars, S. P. Japanese Journal of Applied Physics 2010, 49, 080210.
[12] Lalau Keraly, C.; Kuritzky, L.; Cochet, M.; Weisbuch, C. In III-Nitride Based
Light Emitting Diodes and Applications; Seong, T.-Y., Han, J., Amano, H.,
Morkoc, H., Eds.; Topics in Applied Physics; Springer Netherlands, 2013;
Vol. 126; pp 231269.
[13] Shi, G. A.; Saboktakin, M.; Stavola, M.; Pearton, S. Applied Physics Letters
2004, 85, 56015603.
[14] Zhang, D.; Brodie, D. Thin Solid Films 1994, 238, 95 100.
[15] Song, H.; Kim, J.-H.; Kim, E. K. Journal of the Korean Physical Society
2008, 53, 258261.
[16] Laurent, K.; Yu, D. P.; Tusseau-Nenez, S.; Leprince-Wang, Y. Journal of
Physics D: Applied Physics 2008, 41, 195410.
[17] Fujimura, N.; Nishihara, T.; Goto, S.; Xu, J.; Ito, T. Journal of Crystal
Growth 1993, 130, 269279.
62

[18] Yamabi, S.; Imai, H. J. Mater. Chem. 2002, 12, 37733778.

[19] Ohyama, M.; Kouzuka, H.; Yoko, T. Thin Solid Films 1997, 306, 78 85.
[20] Kim, J. H.; Kim, E.-M.; Andeen, D.; Thomson, D.; DenBaars, S. P.;
Lange, F. F. Advanced Functional Materials 2007, 17, 463471.

63

Chapter 4
Ga-doped ZnO current-spreading
layers on c-plane LEDs

The work in this chapter is based on a paper submitted in November 2013 to

Crystal Growth & Design by Reading, Richardson, Estrada, and DenBaars [1].

4.1

Introduction

The results from undoped ZnO films detailed in the prior chapter indicated that
there was much promise using ZnO for LED current spreading layers. However,
there are many ways in which the properties of these films can be improved by
optimizing the synthesis methods. One such way is by introducing dopant atoms
which could act as donors, and increase the n-type carrier concentration of the

64

films. n-type doping of ZnO has been widely studied. Group III elements (Al, Ga,
and In) have been shown to allow AZO, GZO, IZO, and combinations thereof to
reach greatly improved bulk conductivities that were found above for undoped
films. Al, Ga, and In have all been confirmed to act as shallow donors in ZnO with
low ionization energies.
Gallium-doping is especially interesting for single-crystal films, because Ga3+ ions
have similar ionic radius to Zn2+ , which means they can be incorporated into a
ZnO lattice with only small lattice distortions, which presumably would mean a
small impact on the carrier mobility of these types of films. In this chapter, we will
show that this is indeed possible, and led to greatly improved conductivities in the
ZnO films while maintaining excellent transparency. Le et al. [2] have shown that
highly Ga-doped films can be grown hydrothermally using a very similar method
as was shown in the undoped layers of the last chapter, and were able to achieve
conductivities greater than 700S/cm using 2% molar fractions of Ga.

4.2
4.2.1

Experimental methods
ZnO film growth

For this study, ZnO films were grown in much the same way as described in
Chapter 3, using a nitrate salt solution chemistry. The same batch-growth method
was used, in which growths were left to continue overnight for both the seed
layers, and the second layers. However, this time, a gallium nitrate salt was added
65

to the growth solutions in order to allow Ga atoms to be incorporated into the ZnO
films during growth. Also, to aid in the materials characterization, films were
grown on insulating, highly-transparent MgAl2 O4 substrates. Films were also
deposited on c-plane LED wafers supplied by Walsin Lihwa Co., for a second
comparison study with ITO current spreading layers.
The seed layer growths for this study were identical to those in Chapter 3, and
were all undoped ZnO seed layer growth solutions. For growths on the double-side
polished (DSP) spinel substrates, the substrates were annealed in air at 600C for
several hours after cleaning in a heated acid piranha (H2 SO4 + H2 O2 ). This was in
order to completely clean the surfaces of any organic material, and to reconstruct
the spinel (111) surfaces to their equilibrium state.
The second-layer growth solutions contained small amounts of Ga(NO3 )3 :xH2 O.
The formula mass of the Ga-nitrate salt was calculated to be 330g/mol. 0-2
millimolar (mM) concentrations of the Ga-nitrate were dissolved into the growth
solutions, with the goal of maintaining an unperturbed ZnO layer growth, but in
the presence of available Ga-precursor species. The 2nd-layer growth solutions
used in this study were prepared as follows:

24mL 18MDI H2 O
0.1g Zn(NO3 )2 :6H2 O
0.05g Na-citrate
varying amounts of Ga(NO3 )3 :xH2 O
66

The Ga-nitrate concentrations were selected such that the Ga-ion concentrations in
solution were 0, 0.1, 0.5, 1.0, and 2.0mM, compared to the Zn precursor
concentration of 14mM. This gave Ga:Zn precursor concentration ratios of .007,
.035, .071, and .142, respectively. The wafers were placed into the solutions and
heated in a 90C oven, as they were in Chapter 3. There was no noticeable change
in growth rate, although using the batch method, there was already a fair amount
of variation in the final film thickness, which generally decreased with larger-area
wafers. This suggested that there was significant depletion of precursors in the
solution during the long growths. A typical 5m ZnO film on an approx. 2x2cm
sapphire wafer contains 1.4 104 mol of Zn, which equates to 42% of the
3.3 104 mol of Zn in the growth solution. This means that the final thickness in
these growths was most likely being limited by the total amount of available
precursor, which explains why smaller substrates tended to have thicker films.
ZnO films grown on GaN-on-sapphire wafers generally only grew on the GaN
wafer surface, and did not tend to grow on the backside of the sapphiremost
likely because of the substantially larger lattice mismatch between ZnO and
sapphire. However, on the spinel substrates, there was no such restriction, and the
ZnO seed layers would nucleate on both sides of the wafer. However, since powder
particles from the seed layer solution would fall on the top (back) surface, this film
would be highly-defective and rough. For this reason, all films on spinel substrates
were fully covered in thick photoresist, and dipped in HCl in order to completely
clean all ZnO growth from the backside of the wafers. This resulted in smooth
films on smooth substrates, which was necessary for the optical characterization.

67

4.2.2

Elemental characterization

Doping the ZnO films using Ga-nitrate has been shown to work by several
previous researchers, but it was important to exactly determine the amount of Ga
dopant being incorporated into the films by a direct method, since the exact growth
conditions could vary. This was accomplished using Secondary-Ion Mass
Spectrometry (SIMS), and was calibrated by fully dissolving a large quantity of
films in HCl and performing solution inductively-coupled plasma atomic emission
(ICP-AE) spectroscopy. The ICP spectroscopy signal was in turn calibrated to
standard solutions made using high-purity oxide powders such that the exact
concentrations of Zn and Ga were known. Films grown for these calibration
purposes were grown under exactly the same conditions as the Ga-doped films
with 2.0mM of Ga precursor in the growth solution. SIMS measurements were
done using a Physical Electronics 6550 Dynamic SIMS system using an oxygen
ion beam, in order to detect the Zn and Ga cations. The solution spectroscopy was
done on a Thermo iCAP 6300 spectroscopy system, which is capable of
determining elemental concentrations in solution down to 10 parts per billion.

4.2.3

Hall effect measurements

Hall effect measurements are a very common method for characterizing the
properties of semiconductor materials. The electrical conductivity of a

68

semiconductor is given by,

= e n + h p

(4.1)

where h and e are the hole and electron mobilities in Vs/cm2 , and p and n are the
hole and electron concentrations per cm3 . This equation becomes simpler when
the semiconductor is n-type, or p-type, and current is overwhelmingly conducted
by one type of majority carrier. In the case of our ZnO films, this is n-type
conductivity, with the expression being simplified to,
ZnO = e n

(4.2)

Hall measurements seek to decouple the mobility and carrier concentration

contributions to the bulk conductivity of the films. This is accomplished by
applying a large magnetic field perpendicularly to the direction of current flow, and
out of the plane of the film. Simultaneously, current is flown through the sample.
The strong magnetic field caused the current carriers to be deflected, which creates
a measurable voltage across the film, in the transverse direction to current flow.
The van der Pauw method is a common technique by which this Hall voltage can
easily be calculated by calculating the resistances between 4 contacts arranged in a
cloverleaf geometry, as shown in Fig. 4.1. Measurements are made by flowing a
constant current I between two pads diagonally, and measuring the voltage across
the other two pads. For example, while applying a positive magnetic field, current
could flow between pads 1 & 3, while measuring the voltage V24,P . This is repeated

69

Figure 4.1: Van der Pauw geometry used to measure carrier concentration & mobility via the Hall method
for the other diagonal direction to obtain V13,P . Then the magnetic field is reversed
and the process repeated. Finally, the Hall voltage can be calculated by,

VH = (

V24 + V13 + V31 + V42

V24 + V13 + V31 + V42
)P (
)N
8
8

(4.3)

After the Hall voltage has been obtained, the sheet concentration of carriers, n sh
can be calculated by,
n sh =

IBz
e|VH |

(4.4)

The sheet resistance, R sh , is calculated by a series of measurements across the pads.

70

Finally, the carrier concentration, and mobility of the films can be calculated,
e =

1
n sh R sh e

(4.5)

and
n=

n sh
t

(4.6)

where t is the thickness of the layer.

The Hall samples were patterned using photolithography and subsequent wet
etching of the ZnO layers in dilute HCl. The annealing of the Hall samples at
temperatures up to 300C were done sequentially in between the Hall
measurements. The thicknesses of the films were measured by Dektak
profilometry.

4.2.4

Optical absorption measurements

The optical power absorption coefficient, , measured in cm-1 , is one of the most
important material properties that will determine the effectiveness of a transparent
conducting film. The loss in any bulk media, for a wave propagating in the
x-direction, is described by the Beer-Lambert law,
I = I0 ex

(4.7)

71

where I0 is the initial power intensity of light, and I is the power intensity at a
point x. There two main methods to measure this quantityUV-visible absorption
spectroscopy, and spectroscopic ellipsometry. Spectroscopic ellipsometry works
by measuring the distortion of a circularly polarized beam hitting the sample
surface at different angles. An optical model is constructed, and the unknown
complex refractive index (n + ik) of the film can be calculated, where n is the
refractive index, and k is the so-called extinction coefficient. The relationship
between extinction and absorption coefficients is,
=

4k

(4.8)

However, one of the drawbacks of ellipsometry measurements is that the measured

reflective signal is quite small, which makes it difficult to measure the optical
absorption of films below about 1000cm-1 . It is also necessary to rely on optical
models, in which the absorption and reflection of the substrate are known.
UV-visible transmission spectrometry is somewhat simpler experimentally. A film
is deposited on a double-side polished transparent substrate, and light is shone
both through a sample, and through a lossless cavity. The ratio of intensity
between the sample beam, and the lossless cavity is then measured. This value is
known as T . It is also possible to modify this setup to measure reflection, but this
depends on a high-quality lossless mirror, which presents certain difficulties. Also,
distortion of the beam can complicate these reflection measurements significantly.
Cisneros [3] presented a paper describing a method whereby the absorption
coefficient can be calculated from the transmission spectra alone, without relying
72

on any reflection data. The caveat is that the thickness of the films must be known,
but this was simply determined by cross-sectional SEM. The details of the
calculations made can be found in Appendix A.

4.2.5

LED device fabrication

LED fabrication for evaluating the performance of the ZnO:Ga films as current
spreading layers proceeded according to Section 3.2.3 for the LEDs with undoped
ZnO current spreading layers. However, this time a 300nm ITO layer was
deposited as a comparison, instead of 240nm. The LED fabrication and packaging
was otherwise exactly the same as in the study of the undoped ZnO current
spreading layers. ZnO seed layers were annealed at 500C, with a 225C
following the final film deposition. ITO layers were annealed at 600C in an N2 /O2
ambient for 10 min, followed by a 3 minute anneal at 600C in pure N2 .

4.3
4.3.1

Characterization of Ga-doped ZnO layers

SIMS elemental Characterization

SIMS depth profiles were measured throughout the thickness of the films before
and after annealing at 250C, although there was no appreciable change in
elemental distribution with annealing. The ratio of the 69Zn to the 72Ga SIMS
peaks were monitored. Meanwhile, the 2.0mM films were dissolved in HCl and
73

Figure 4.2: Elemental characterization (a) Calibrated SIMS results showing fraction
of Ga atoms in doped ZnO films. The left-most side of the chart represents the surface of the ZnO film, while the right-most side represents data from near the spinel
substrate. (b) % incorporation of Ga in the films with Ga precursor concentration
in growth solution
characterized using ICP-AE solution spectroscopy. It was found that the average
ratio of Ga to Zn in these films was 0.0132. This fact was used to calibrate the
average value of the peak ratio in the SIMS data for the 2.0mM films. This
calibration was used to rescale all the SIMS data to an actual elemental ratio
between Ga and Zn atoms.
The composition profiles in Figure 4.2 show that the amount of Ga in the films
increases with the amount of Ga precursor in the growth solutions. It is also
evident that the Ga concentration tends to increase closer to the surface of the
films. This indicates that the incorporation rate tended to increase during the film
growth. One possible reason for this could be that the incorporation rate is
temperature dependent. The growth solutions for these films were heated in a 90C
oven, which might mean that the temperature was gradually approaching 90C
74

throughout the growth, which would cause the Ga to be incorporated faster

towards the end of the growth. In Figure 4.2(b), it is shown that the fraction of Ga
incorporated increases with the concentration in solution, but that this effect is
sublinear and tends to roll off at higher concentrations of Ga. Presumably this
quantity would approach a limit as the solid solubility fraction of Ga in ZnO is
approached.

4.3.2

Hall Effect Characterization

Hall effect measurements performed by Dr. J. Richardson on the ZnO films grown
on insulating spinel substrates are shown in Figure 4.3, and show a number of
interesting trends. Generally, the carrier concentration increased with thermal
annealing at higher temperatures, with the largest increases taking place between
anneals at 200C and 250C. This is in accordance with the results from Chapter 3,
where the sheet resistance was found to be greatly decreased after annealing at
225C. Also, it appears that the undoped films are the most conductive in their
as-grown state, but are surpassed by all films containing Ga after
higher-temperature annealing. This is due to the considerably higher carrier
concentrations in the doped films after annealing. However, the undoped films did
generally have higher electron mobilities, which is expected due to lattice
distortions that are caused by introducing impurity atoms.
Annealing at higher temperatures was not possible, as it caused failure of the metal
contacts needed to perform the Hall measurements. The highest conductivity value

75

Figure 4.3: Hall effect measurements showing (a) the n-type carrier concentration in
the films, (b) the electron mobility in the films, and (c) the bulk electric conductivity
of the films. Annealing was performed sequentially on the same samples of c-plane
ZnO on spinel substrates.

76

obtained was 660S/cm, for the 2.0mM Ga concentration film, which corresponds
to a bulk resistivity of 1.5 103 cm.

4.3.3

Optical Characterization

The visible-range optical absorption coefficients for the films on transparent spinel
substrates are shown in Figure 4.4. The as-grown films generally exhibited
excellent transparency, with the pure ZnO film being the most highly transparent.
The as-grown, undoped film had an absorption coefficient of around 200cm-1 in the
violet, which dropped to below 100cm-1 in the green wavelengths. The absorption
was only around 70cm-1 at wavelengths of 650nm and higher. The as-grown films
generally became less transparent with increasing dopant concentration. The
2.0mM Ga film exhibited absorption in the range 130300cm-1 across the
wavelength range, which was considerably higher than the undoped film.
After annealing at 150C the films generally became slightly more transparent, but
this effect was quite subtle. After annealing at 250C, we saw a decoupling in the
trend among the films. The undoped, and less heavily doped films exhibited some
type of material failure at these temperatures, and became noticeably hazy. This
resulted in the films scattering much more light. The undoped film had absorption
coefficients in the range 300cm-1 900cm-1 after this anneal. However, this was not
the case for the more heavily doped films. The films with 0.5 and 1.0mM Ga only
showed a small increase in absorption, while the film with 2.0mM Ga actually
showed a decrease in absorption after annealing at this temperature. The SEM

77

Figure 4.4: Optical absorption coefficients calculated from UV-visible transmission

spectra. The calculation method by Cisneros [3] was used. Films thicknesses were
measured by cross-sectional SEM to be between 4500 and 5500m. Substrate index
was fitted to a Sellmeier refractive index model.
images of voids in the annealed undoped films are shown in Figure 4.5.
In order to measure the absolute performance of a TCO layer, it is important to
compare both the intrinsic transparency, and the electrical conductivity. One figure
of merit for these qualities is known as the conductivity of absorption, which is
simply the electrical conductivity divided by the optical absorption coefficient.
The units for this quanitity are inverse Ohms, -1 . Figure 4.6 shows these
calculated values for the films, based on the absorption measurements shown in
Figure 4.4 and the conductivity measurements shown in 4.3(c).
The highest figures of merit were obtained for the most heavily doped films after
annealing at 250C. It is suspected that the properties could improve further with
higher temperature annealing, provided that the samples do not become hazy
during the annealing step. The haziness was attributed to the formation of voids
within the undoped, and lightly doped films. This effect has been studied
previously, and has to do with the dehydration of the films during annealing. Gas
78

Figure 4.5: Cross-sectional SEM showing (a) voids present in undoped ZnO film
after 250C, and (b) no voids present in the Ga-doped ZnO film with 2.0mM Ga
precursor concentration. SEM figure courtesy Jacob Richardson.

Figure 4.6: TCO Conductivity of absorption figure of merit for Ga-doped ZnO
layers under different annealing and doping conditions.

79

that cannot escape will collect in the films, forming voids. This was previously
found by Richardson et al. [4]. However, it appears that this effect is substantially
suppressed by the addition of Ga dopant.

4.3.4

XRD characterization

X-ray diffraction characterization of the films revealed some interesting behavior

in the doped films. For the XRD study, only the undoped, 0.5mM Ga, and 2.0mM
Ga samples were investigated, before and after annealing at 250C. The results are
shown in Figure 4.7.
Figures 4.7(a) and (c) show the behavior of the c lattice parameter with addition of
Ga dopant, which appears to decrease upon addition of the dopant. Figures 4.7(b)
and (d) show the (101) off-axis peaks after annealing, where the lattice again has
been contracted by the addition of Ga impurity atoms. The peaks indicate % and
% contractions for the 0.5mM Ga and 2.0mM Ga samples, respectively.

4.4

c-plane LED results

The LEDs that were fabricated were packaged and tested using the same setup as
in Chapter 3. An ITO comparison LED was fabricated, but this time with a thicker
layer, which was 300nm. The sheet resistances and conductivities of the current
spreading layers are summarized in Table 4.1. The conductivities measured were
in line with Hall measurements described in Section 4.2.3 for devices annealed at
80

Figure 4.7: X-ray diffraction spectra of ZnO and ZnO:Ga films grown on spinel
substrates. (a) 2 scan near ZnO (002) peak, (b) 2 scan of an off-axis ZnO (101)
peak. (c) 2 scan near ZnO (002) peak after annealing at 250C, and (d) 2 scan of
off-axis ZnO (101) peak after annealing at 250C.

81

Table 4.1: Electrical properties of ITO, ZnO, and ZnO:Ga current spreading layers,
characterized by CTLM method.
CSL material
ITO
ZnO
ZnO:Ga
ZnO:Ga
Ga precursor conc. (mM)
n/a
0
0.5
2.0
R sh (/)
55
18
11
9.1
Thickness (nm)
300
2500
2400
2500
Resistivity (cm)
1.53 103 4.50 103 2.64 103 2.28 103
Conductivity (S/cm)
653
222
279
438
225C.
The external quantum efficiency of the devices, shown in Figure 4.8 showed the
sample with intermediate Ga doping having the highest peak efficiency of all the
devices tested. This demonstrates the there is a tradeoff between conductivity and
optical absorption when choosing a doping concentration for the films. The 0.5mM
Ga and 2.0mM Ga ZnO samples had sheet resistances of 11 and 9.1, respectively,
which are likely to be comparable to that of the n-GaN layer below the active
layers of the device. If this is the case, there is no longer any benefit to lower the
sheet resistance of the current spreading layer any further, and the increased optical
absorption of a more highly doped film will tend to decrease the total amount of
light able to escape the devices. The current spreading performance of the ITO and
0.5mM Ga CSLs is shown in Figure 4.9. It appeared that the emission from the
0.5mM Ga CSL sample was highly uniform, with barely any current-crowding
effects visible near the p-side metal contact. The ITO in this study performed
significantly better than the thinner ITO layers used in Chaper 3, having a lower
sheet resistance of 55/, which was still higher than the thicker ZnO layers.
Finally, white LEDs were fabricated using the same phosphor cap method shown
82

Figure 4.8: External quantum efficiency of type 3 LEDs, with peak wavelength at
470nm. Measurements were done on bare chips without silicone encapsulation to
avoid packaging-related inconsistencies. The device area was 1.07mm2

Figure 4.9: Optical microscope images of ITO and 0.5mM Ga current spreading
layers. Current spreading was almost completely uniform for the 0.5mM ZnO:Ga
CSLs.

83

Figure 4.10: Characterization of white LEDs fabricated with ITO and best ZnO:Ga
LED chips. (a) I-V curves showing operating voltages in the packaged white LEDs,
and (b) luminous efficacy of white LEDs with photograph of operating device
in Figure 3.3. However, there was an important change to the phosphor material
used. The white LED results previously achieved used only YAG:Ce phosphor,
which produces white light in combination with 450nm blue light from the LED
chip. The LED wafers used in these studies, however, had a peak emission
wavelength at 470m. This time, a combination of the YAG:Ce with a proprietary
red-emitting phosphor from Mitsubishi Chemical Co. allowed much higher quality
white light to be achieved, with a color-rendering index (CRI) score of 81 for a
cool-white color temperature of 4500K. The results from these LEDs are shown in
Figure 4.10

84

4.5

Summary and Conclusion

In summary, we have demonstrated greatly improved characteristics in epitaxial

ZnO layers for use as current-spreading layers on GaN/InGaN LEDs through
n-type doping using Ga. This was achieved through a simple modification of the
low-temperature aqueous growth solution: adding gallium nitrate salt in addition
to the zinc nitrate salt. The films exhibited greatly increased conductivity, more
than 2 higher, than the highest conductivities observed in the undoped ZnO
layers. The films were shown by X-ray diffraction methods to be single-crystalline
heteroepitaxial layers on the GaN epiwafers, and a distortion in the crystal lattice
was observed as a result of the impurity doping. Hall results showed a lower
electron mobility in the doped layers, but much higher carrier concentrations after
annealing at 200C, which continued to increase after annealing at 300C. The
optical absorption coefficients were measured by transmission spectroscopy to be
below 300cm-1 for the most highly doped films, and below 120cm-1 for the
undoped films, which compares to 800-1200cm-1 for ITO films.

85

References
[1] Reading, A. H.; Richardson, J. J.; Estrada, D. M.; DenBaars, S. P. Cryst.
Growth Des. Submitted Nov. 2013,
[2] Le, H. Q.; Lim, S. K.; Goh, G. K. L.; Chua, S. J.; Ong, J. Journal of The
Electrochemical Society 2010, 157, H796H800.
[3] Cisneros, J. I. Appl. Opt. 1998, 37, 52625270.
[4] Richardson, J. J.; Goh, G. K. L.; Le, H. Q.; Liew, L.-L.; Lange, F. F.;
DenBaars, S. P. Crystal Growth & Design 2011, 11, 35583563.

86

Chapter 5
Ga-doped ZnO current-spreading
layers for semipolar GaN LEDs

5.1

Introduction

The majority of commercially available GaN LEDs and lasers are based on GaN
grown on mismatched substrates, such as sapphire, SiC, or even (111) silicon
wafers. This has a number of downsides such as having a large amount of
detrimental threading dislocations, time-consuming buffer-layer growths, and a
mismatched thermal expansion coefficient between GaN and the substrate
material. Also, for c-plane devices there is the issue of the quantum-confined stark
effect causing electric fields within the devices quantum wells, which separate the
electron and hole wavefunctions, causing severe limitations on the quantum well

87

thickness. This, in turn, limits the allowable thickness of the quantum wells. This
makes it necessary to employ many quantum wells which leads to time-consuming
MOCVD growths.
Recent work has demonstrated great potential in using free-standing bulk GaN
substrates to greatly increase the material quality of GaN epilayers, and to allow
and 202 1
devices grown on different planes of the wurtzite GaN crystal. The 2021
planes are of particular interest for high-performance blue and green LEDs and
laser, and has allowed devices with low efficiency droop at current densities up to
200A/cm2 by Zhao et al. [1]. These types of devices were shown by Feezell et al.
[2] to have improved performance based on the polarization fields within the
device quantum wells, and very low wavelength shift at high current densities.
Previous work has shown that single-crystal ZnO could potentially be used as a

current-spreading layer on 101 1-plane

InGaN/GaN LEDs by Richardson et al. [3],
obtaining EQEs as high as 48%, but high-voltage operation was observed on these
devices, which lowered their total power efficiency.

5.2
5.2.1

Experimental Methods
Patterning methods

For large-area devices, where the patterning does not need to be very accurate, it is
possible to use dilute acid to wet-etch the ZnO films. However, for bulk GaN
wafers, LED chips must be much smaller due to the small substrate size, (0.1mm2
88

vs. 1mm2 ), which means the undercut problem makes device fabrication
impossible. For a 0.1mm2 device, whose lateral dimensions are 350m285m, a
5m undercut represents a 6.2% reduction in the device area. A 15m undercut
results in an 18% reduction in device area, which becomes problematic quite
quickly, especially when trying to etch thick ZnO layers. For this reason, a new
processing method was established in which the ZnO is deposited using an SiO2 or
photoresist mask.
The bulk GaN epiwafers with a top-most surface of p-GaN were first cleaned in
boiling aqua regia to remove all indium metal from the surface, which is necessary
for post-MOCVD characterization. Then, 200nm SiOx are deposited using a
Plasma-Therm 790 plasma-enhanced chemical vapor deposition system at 250C
using SiH4 and N2 O gases. This SiOx layer is then removed in certain areas using
photolithography followed by a 30s O2 plasma descum, and then a 3.5 minute wet
etch in 1:1 H2 O:buffered HF solution. This exposed areas of undamaged p-GaN
on which ZnO can be grown. After this step, the photoresist is completely
removed using heated AZ 1165 photoresist stripper.
The wafers must then be completely cleaned of organic material, since even a tiny
amount of photoresist scum can prevent ZnO films from ever nucleating on the
p-GaN surface at low growth temperatures. This is because the growth conditions
used must favor very low nucleation densities in order to prevent non-epitaxial
ZnO to begin to form on the surface. Non-epitaxial powders will quickly start to
form needles that interfere with the goal of obtaining a smooth, coalesced film.
The cleaning step used is known as the RCA SC 1, for Standard Clean 1. This

89

step involves a 10 minute immersion, with agitation, in a 1:1:5 by volume solution

of 29% NH4 OH:20% H2 O2 :H2 O. This step should oxidize all carbon-containing
compounds remaining from the photoresist-stripper. After this organic clean, it is
necessary to do an oxide removal in 1:1 HCl:H2 O solution, for about 5 minutes.
The wafers are then ready for the ZnO film growth. Note that these cleaning steps
are only necessary when growing directly on a p-GaN surface, not when growing a
second layer of ZnO.
This procedure should allow ZnO to grow only in the windows where p-GaN is
exposed, with little or no ZnO material nucleating and growing on the SiOx layer.
When growing thick films > 2.0m, it is important to realize that there will be
some amount of lateral overgrowth, particularly in the c+ projection of the wafer.
GaN
An image of a successfully patterned film grown by this method on a 2021
wafer is shown in Figure 5.1.

5.2.2

Film growth

Growing ZnO films on bulk GaN semipolar substrates required a number of

significant changes to the growth conditions that were used for growth on c-plane
GaN on sapphire, as was the case in Chapters 3 and 4. The small size of bulk GaN
substrates also greatly increases the problems associated with holding the sample
suspended in solution without causing any restriction of flow across the wafer
surface. A smaller wafer size also means a faster growth rate, which means less
control of the final film thickness. Better control of the heating rate was necessary

90

Figure 5.1: SEM micrographs showing a patterned ZnO layer grown using an SiOx
GaN substrate. (a) CTLM pad showhard mask for etch-free patterning on a (202 1)
ing metal, ZnO, and SiOx layers. (b) Some prismatic ZnO particles did nucleate
in solution, which is undesirable, espectially for long growths, (c) UHR image of

700nm
thick ZnO layer (SiOx is 200nm thick). (d) top-down image of CTLM pad
showing precise patterning

91

Figure 5.2: Heating curves for 90mL PTFE vessel with different Variac powers.
Higher power allowance above 75% led to melting of the PTFE vessel exterior.
90C was reached at 54 min., at which time a 100-200nm film is already deposited.
to enable growth rate calibration and film thickness targeting. The temperature
profile at different variac settings is shown in Figure 5.2.
All growths on semipolar bulk GaN substrates were done in a 90mL PTFE vessel
with an external heating coil wrapped around the outside. The temperature of this
heating unit was limited with a power Variac to prevent the PTFE melting, which
takes place at temperatures higher than 200-250C. A thermocouple coated with
PTFE was inserted into a side-port of the vessel and connected to the heater power
supply to enable good temperature control of the solution. It was found that the
vessel could reliably reach the growth temperature at 90C after 54 minutes. The
nominal temperature ramp time was 30 minutes to ramp the controller from room
92

Figure 5.3: Diagram of PTFE vessel used for growing ZnO films on small sub11cm free-standing GaN wafers. The variac power was set at 75% to prevent the
PTFE melting.
temperature to 90C. The film thickness at 54 minutes, immediately after the
thermocouple reached 90C was measured by Dektak to be about 100-200nm.
During growths, the PTFE vessel was constantly stirred using a small magnetic stir
bar on a magnetic stirring plate.
In order to allow uniform film growths, it is necessary that the samples face down
and are positioned horizontally in the growth vessel. This is important, because if
the sample is positioned at an angle, vertically or face-up, there could be turbulent
flow of growth solution across the sample surface as a result of the stirring. The
samples are suspended in the PTFE vessel using a PTFE ring that has had notches
cut into it. The ring then hangs from a section of PTFE tubing which pressed
against the interior walls of the vessel. A diagram of the growth setup is shown in
Figure 5.3.
Two-step growths were still necessary to ensure good adhesion to the p-GaN

93

surface. The first step involved growing a thin layer by stopping the growth
immediately upon reaching 90C. The film was then annealed in an RTA system at
500C, which was the same conditions used to achieve good electrical contact and
adhesion for seed layers on c-plane GaN, as described in Chapters 3 and 4. If a
film is thicker than about 500nm, it can delaminate during annealing at 500C, so
it is important to do this step while the film is still quite thin. Then the samples
were reinserted into the PTFE growth vessel and allowed to reach the film growth
temperature at 90C. Growth times were calculated as,
T = 54 + T growth

(5.1)

since the growth rate only became constant after the solution has reached growth
temperature, i.e. there is a gradual turn-on effect as the growth solution
temperature reaches 90C.
The growth solution used for both first, and second layers on semipolar bulk GaN
samples was,

90mL 18M DI H2 O
0.1875g Na citrate (tribasic), Sigma-Aldrich, 99.9% purity
0.75g Zn(NO3 )2 :6H2 O, Sigma-Aldrich, 99.9% purity
0.039g Ga(NO3 )3 :xH2 O, Sigma-Aldrich, 99.9% purity
4.67mL NH4 OH 29% concentration, Sigma-Aldrich, reagent-grade

94

The citrate is still necessary to slow down the growth in the c+ crystal direction,
which can lead to rougher films, and formation of needles nucleating from powder
particles. The Ga-concentration used corresponds to the 1.0mM samples from
Chapter 4, which were found to have high TCO figures-of-merit on c-plane GaN
devices. Also, at these concentrations, there is no appreciable change in the
solution concentrations when growing on small bulk GaN wafers when there is
90mL of growth solution. All films were annealed at 225C to increase their
conductivity as was described in Chapter 4.

5.2.3

Device Fabrication

For this study, ZnO layers were deposited on p-GaN test structures, and on GaN
LEDs, for use as current-spreading layers. The p-GaN test structures were 150m
and 2021
substrates.
of p-type GaN:Mg deposited on insulating bulk GaN 2021
These test structures were both co-loaded in the MOCVD reactor, such that the
growth temperatures, and gas flows were the same for both samples. Therefore we
can assume that the properties of the p-GaN layers are the same, except for their
crystal polarity.
TLM structures were made using the test structure samples by patterning the ZnO
according to Figure 3.1(c) and by using the patterning method described in Section
5.2.1. The test structures used are pictured in Figure 5.4. LEDs were fabricated in
the same way as in Chapters 3 and 4, with a few modifications. Due to the use of
the SiOx hard mask, it is necessary to do an additional SiOx dry etching procedure

95

Figure 5.4: Test structures allowing measurement of ZnO contact resistance patterned using SiOx hard mask
in the Panasonic ICP etch system prior to the Ar/Cl2 mesa etch.
The LED packaging for bulk GaN wafers required some additional steps to
enhance light extraction. This step involved thinning the wafer to a thickness of
approx. 200m, patterning a regular pattern of photoresist dots on the back side,
and then conducting another ICP etch with isotropic etching in the ICP etch
system. This was found by Zhao et al. [4] to create a periodic pattern of conical
roughening on the backside, which greatly aids light extraction for bulk-GaN
wafers, providing a 180% increase in light extraction from the chips. During the
backside roughening procedure, the thinned wafer is mounted face-down on a
sapphire carrier wafer.

5.3
5.3.1

Results
XRD characterization of semipolar ZnO on GaN

and 202 1 test structures in

X-ray diffraction measurements were done on the 2021
order to confirm that the ZnO layers were growing heteroepitaxially on the GaN
96

 and 202 1 p-GaN

Figure 5.5: XRD characterization of ZnO films grown on 2021
templates. (a) 2 scan of ZnO and GaN (201) reflections, and (b) rocking curve

scan showing higher crystal quality, and less tilt in the ZnO films grown on the 2021
template

97

layers, and to measure the crystallinity of the layers. The results of XRD 2 and w
scans are shown in Figure 5.5. The (201) reflections in the 2 scan showed that the
ZnO (201) and GaN (201) planes were aligned, indicating heteroepitaxial layers.
However, it is important to note that the polarity of the (201) films is impossible to
and a 202 1 have exactly the same lattice
detect by X-ray diffraction, since a 2021
planes and spacings, which are what determines the Bragg condition required for
an XRD peak.
template
The rocking curves in Figure 5.5(b) show the ZnO film on the 2021
with a FWHM of 0.133, while the ZnO film on the 202 1 had a considerably more
defective crystal, with a FWHM of 0.304. This large difference in crystal
mosaicicity and quality indicates that there could be a fundamental difference in
the film growths on these planes since all other aspects of the film growth were
identicalthe GaN layers were grown simultaneously with the substrates
co-loaded in the MOCVD reactor, and the ZnO layers were similarly co-loaded in
the PTFE growth vessel.

5.3.2

Contact resistance measurements on 202 1 and 2021

p-GaN

CTLM measurements were made on both samples to determine the specific

contact resistivity of the GaN/ZnO interface, and to measure the conductivity of
sample,
the films. Table 5.1 shows the results of TLM measurements on the 2021
showing Ohmic contact with a specific contact resistivity of 0.018cm2 , which

98

Table 5.1: CTLM measurements on ZnO layers grown on 120nm p-GaN/10nm

and 202 1 substrates.
p++ -GaN templates on 2021
substrate 202 1 substrate
ZnO Property 2021
Appearance
Smooth
Smooth
Sheet Resistance (/)
28
35
Specific Contact Res. (/cm2 )
0.018
n/a
Thickness (nm)
900
700
Resistivity (cm)
2.52 103
2.45 103
was lower than was seen on commercial c-plane p-GaN, as was shown in Chapter
3 in Figure 3.4. The properties of the TLM measurements are shown in Table 5.1.
The sheet resistances of the films were different, but this was explained by the film
being slightly thicker than the film grown on the 202 1 p-GaN. They
on 2021
corresponded quite well to each film having the same bulk conductivity, which
would make sense given that the films underwent the same growth and annealing
conditions.
The biggest difference between the electrical properties of the films was that the
film grown on 202 1 displayed highly rectifying behavior when measured in the
TLM structure, indicating that a tunnel junction did not form. In fact, the effective
barrier was very high, requiring over 5-6V in order to allow any measurable
amount of current to flow. This indicates that instead of forming a tunnel junction
with Ohmic characteristics, the junctions are acting simply as reverse-biased p/n
heterostructures. As such, it would be impossible to form any kind of efficient

device using these types of films on 202 1-plane

GaN devices, given the high
voltages required to drive current.

99

5.3.3

TEM characterization of semipolar ZnO on GaN

In order to understand the crystallographic nature of the interface, it was necessary

to perform cross-sectional transmission electron microscopy (TEM) on the films
grown on the p-GaN templates. By using convergent-beam electron diffraction
(CBED), it is possible to simultaneously measure the crystal plane spacings and
the polarity of the crystal, which allows determination of the specific orientation in
a polar material. The TEM images and diffraction patterns are shown in Figure 5.6.
TEM samples were cut from the same CTLM structure described in Section 5.3.2,
crystal direction, which allows the crystal
and were cut perpendicular to the [1120]

polarity to observed most easily. The CBED patterns were simulated for both 2021
film matched
and 202 1 planes. It was found that the CBED patterns from the 2021
ZnO film on 2021
p-GaN, which was
the simulations most closely for a 2021
expected, since the ZnO had good crystallinity, and good electrical contact.
However, it was found that for the ZnO film grown on 202 1 p-GaN, the ZnO film
ZnO film on
still had an opposing crystal polarity, i.e. the ZnO film was a 2021
202 1 p-GaN.

The polarity inversion that was observed by CBED on 202 1-plane

GaN indicates
that it could be quite difficult to obtain low-resistance contacts to these types of
GaN devices using solution-grown ZnO. ZnO crystals tend to grow most quickly
in the c+ crystallographic direction, which is opposite to the direction it would

need to grow to form a 202 1 layer. This is why the ZnO film will grow as 2021
or a 202 1 GaN layer. At the same time, the
whether it is grown on a 2021

100

Figure 5.6: TEM and CBED characterization of ZnO layers on semipolar p-GaN
p-GaN, and (b) ZnO on 202 1 p-GaN. TEM courtesy
templates. (a) ZnO on 2021
Dr. Feng Wu.
polarization fields in the ZnO and GaN layers likely need to be aligned, or at least
in the same direction, in order to form a good electrical contact. When there is a
polarity inversion, with a sharp interface, it could lead to a widening of the
depletion region or an increase in the polarization barrier between the n-type ZnO
and the p-type GaN layers, which leads to an exponential decrease in the carrier
tunneling probability, disallowing the formation of an Ohmic tunnel junction.

101

5.3.4

semipolar GaN LED

LED results on 2021

substrates by
Green emitting GaN/InGaN LEDs were fabricated on bulk 2021
MOCVD, and with ZnO:Ga current spreading layers. The mask used had a
0.1mm2 device area, which measured 350 285m. The LED structures used to
fabricate these devices are the same to those used in Yamamoto et al. [5]. The I-V
characteristics of these devices are shown in Figure 5.7. The contact resistance for
the ZnO:Ga CSL was measured previously to be 0.018cm2 , which translates to
the addition of 1.8V to the operating voltage at 100A/cm2 . This value can then be
used to estimate the contact resistances of the 2 types of ITO current spreading
layers, since we can claim the underlying LED has an operating voltage of
4.3V 1.8V = 2.5V at 100A/cm2 . By this method, the ECR-sputtered ITO film
has a contact resistance of 0.013cm2 , while the 250C E-beam ITO has a contact
resistance of about 0.025cm2 . However, it is likely that all contact resistances
could be substantially lowered using a p-InGaN layer at the topmost surface of the
p-GaN, which enhances tunnel junction formation by increasing the hole
concentration near the top-contact interface, as has been established in the
literature [6, 7].
The LED performance was characterized using an integrating sphere setup. The
EQE performance of these packaged devices was substantially lower than other
reports of semipolar green LEDs, but this was attributable to the high variability
introduced by the backside roughening procedure, making direct comparison
between this particular LED and previous results on similar devices with ITO
current-spreading layers problematic. However, direct observation of the current
102

 bulk GaN substrates

Figure 5.7: IV performance of green 0.1mm2 LEDs on 2021
with different current-spreaing layers (CSLs). ECR-sputtered ITO layers were deposited by MES AFTY Co. using an ECR remote plasma sputtering system. Ebeam evaporated ITO layers were deposited on 250C heated substrates.
spreading in the devices indicated that the ZnO layers performed adequately, with
uniform emission visible from the top-side of the devices.

5.4

Summary and Conclusions

In summary, the performance of ZnO layers as a potential alternative for ITO was
examined in the context of high-power, high-efficiency LED devices grown on the
and 202 1-plane

free-standing bulk GaN 2021

substrates. The conductivity of the

ZnO layers was found to be 400S/cm. The contact resistivity on 2021-plane

p-GaN templates was found by the CTLM method to be 0.018cm2 , which is in
the same range as for ITO layers on semipolar planes. This would indicate that
devices.
ZnO is a viable material for use as a transparent contact on 2021
103

 GaN substrate. LED

Figure 5.8: LED performance of a green LED on bulk 2021
packaging used a conventional silicone dome on a silver LED header. The backside
of the GaN substrate was roughened using a dry etch procedure to enhance light
extraction.
It was found that ZnO grown by aqueous hydrothermal method tends to form
layers on 202 1 p-GaN templates, which prevented the formation of
inverted 2021
an Ohmic tunnel junction contact. This led to high voltages required to drive
current through the interface, resembling a reverse-biased p/n diode rather than an
Ohmic contact. For this reason, it seems that it is not possible to use single-crystal
ZnO layers deposited hydrothermally as transparent contacts on 202 1 devices, as
grown by the methods in this chapter.

104

References
[1] Zhao, Y.; Tanaka, S.; Pan, C.-C.; Fujito, K.; Feezell, D.; Speck, J. S.;
DenBaars, S. P.; Nakamura, S. Applied Physics Express 2011, 4, 082104.
[2] Feezell, D. F.; Speck, J. S.; DenBaars, S. P.; Nakamura, S. J. Display Technol.
2013, 9, 190198.
[3] Richardson, J. J.; Koslow, I.; Pan, C.-C.; Zhao, Y.; Ha, J.-S.; DenBaars, S. P.
Applied Physics Express 2011, 4, 126502.
[4] Zhao, Y.; Sonoda, J.; Pan, C.-C.; Brinkley, S.; Koslow, I.; Fujito, K.; Ohta, H.;
DenBaars, S. P.; Nakamura, S. Applied Physics Express 2010, 3, 102101.
[5] Yamamoto, S.; Zhao, Y.; Pan, C.-C.; Chung, R. B.; Fujito, K.; Sonoda, J.;
DenBaars, S. P.; Nakamura, S. Applied Physics Express 2010, 3, 122102.
[6] Kumakura, K.; Makimoto, T.; Kobayashi, N. Japanese Journal of Applied
Physics 2003, 42, 22542256.
[7] Chang, S.; Chen, C.; Chang, P. C.; Su, Y. K.; Chen, P. C.; Jhou, Y. D.;
Hung, H.; Wang, S. M.; Luang, B. Electron Devices, IEEE Transactions on
2003, 50, 25672570.
105

Chapter 6
and 202 1 GaN lasers
Semipolar 2021
with ZnO cladding layers

The work in this chapter is based on recent unpublished work done with
researchers Anisa Myzaferi and Arash Pourhashemi for MOCVD growth and laser
processing.

6.1

Background and Motivation for oxide-clad GaN

laser diodes

Conventional GaN laser structures use Alx Ga1-x N cladding layers in order to
confine the optical modes such that their maxima are centered in the quantum
wells of the laser structure. However, the index contrast in such structures is quite
106

low, resulting in the optical modes extending some distance outside the active
region and cladding layers. To prevent the laser mode from overlapping with the
highly absorbing metal contacts of the device, it is usually necessary to have a
thick p-GaN layer in order to provide sufficient distance from the quantum well
regions to the metal contacts. This presents several growth-related issues. For one,
the p-GaN layers must be grown at fairly high temperatures in order to be of good
material quality. This extended period at high temperature can lead to degradation
of the Inx Ga1-x N quantum wells, especially when targeting longer wavelength
devices.
Recently, there has been progress in using oxide cladding layers with only thin
p-GaN layers. ITO and ZnO have refractive indices of 1.9 and 2.1, at 450nm,
respectively, allowing much higher index contrast than would be achievable using
Alx Ga1 xN cladding with thick p-GaN spacers. Even a thin layer of ITO or ZnO
(50-200nm) allows the optical mode to be fully attenuated by the time it reaches
the absorbing metal contacts to the device. This could in theory allow less
absorption in the laser structure, resulting in lower threshold currents and higher
device efficiencies. ITO and ZnO have the advantage of deposition at temperatures
250and below, which would not cause any damage to the device quantum wells,
and do not involve time-consuming MOCVD growth of thick p-GaN layers.

107

6.2

GaN lasers with ITO cladding

The first university demonstration of ITO-clad blue lasers by Hardy et al. [1] has
spurred interest in replacing the ITO layer with cheaper, or less lossy alternative
materials. The ITO-clad structures had a lasing wavelength of 471nm,
The deposition method used in [1] used a commercial ECR-sputtering system by
MES AFTY Co. The elaborate nature of this method, and great expense of the
required equipment, as well as the difficulty in patterning the films after deposition
makes ZnO a highly attractive alternative due to its higher transparency, low-cost
processing methods, and its ability to be patterned without etching via
SiOx -templated growths on GaN. For use as TCO layers, ITO is able to compete
with ZnO due to its high electrical conductivity. However, in the case of a laser
cladding layer, this bulk conductivity is much less important since current passes
through the layer vertically, over a short distance of 50-200nm, and does not need
to be spread over a large area.

6.3

substrates
Optical characterization on 2021

For designing laser diode structures, it is important to have reliable values of the
refractive index of all materials involved. This is necessary in order to calculate
the confinement factor of the laser modes, which is a crucial parameter in the
design of an efficient, high-power laser. Spectroscopic ellipsometry was used in
this study to determine values of the refractive index for Ga-doped ZnO films
108

 GaN substrates. Precisely measuring the absorption coefficient was

grown on 2021
difficult by this method, because of the difficulty in decoupling absorption in the
bulk GaN substrate and in the ZnO films. The absorption coefficients measured by
ellipsometry were consistently measured to be greater than 1000cm-1 , which did
not agree with values for ZnO mentioned in the literature. However, it is assumed
that the refractive indices of these films were accurate. The refractive index of a
material can vary depending on defect concentrations and the growth method,
which means it may be unwise to assume that ZnO films grown by a hydrothermal
method at 90C would have the same refractive indices as for literature values
from films grown by high-temperature techniques such as PLD, MBE, or even
high-pressure hydrothermal crystal growth. For this reason, ellipsometry
measurements were performed using a Woolam M2000DI Spectroscopic
Ellipsometer.
Modeling done by N. Hong at Woolam Co. showed the ZnO:Ga refractive index to
be at 1.95 for 450nm wavelengths, decreasing to 1.80 at 1000nm. This is in
agreement with literature data for solution-deposited ZnO:Ga films. The model
used involved a higher conductivity of the film closer to the film surface. The
SIMS data showing Ga-incorporation in thick ZnO films shown in Figure 4.2(a)
shows higher incorporation of Ga towards the top surface of the films which may
give rise to heterogeneous conductivity throughout the film. However, the
extinction coefficients were found to be much higher than would make any sense
based on transmission spectroscopy measurements shown in Figure 4.4 and
Appendix A. The index values as calculated from transmission spectroscopy are
shown in Figure 6.2.
109

Figure 6.1: (a) Experimental vs. simulated VASE data showing good fit with experimental values using a graded-conductivity ellipsometry model. (b) Complex
refractive index data showing dispersion using graded-conductivity model.

110

Figure 6.2: Refractive index values calculated using methods described in Appendix A by transmission spectroscopy. Films were 4.5-6m thick and annealed
at 150C.

111

Based on these data, it is expected with a good degree of agreement among

measurement methods that the refractive index of these ZnO:Ga films is between
1.95-2.00 for 450nm blue light. Incindentally, this is very close to values reported
for ITO films at these wavelengths. The reported loss in the ITO layers, assuming
an absorption coefficient of 2000cm-1 was calculated to be 3.5cm-1 , owing to the
0.17% modal overlap in the ITO layer. If the ZnO film had the same refractive
index, but an absorption coefficient of only 200cm-1 , this would correspond to a
modal loss in the ZnO film of only 0.35cm-1 .

6.4

blue laser fabrication

(2021)

Laser structures were grown by MOCVD on semipolar bulk GaN substrates

supplied by Mitsubishi Chemical Co. The laser structure used in this study is the
same as was reported by Hardy et al. [1], and used a In0.06 Ga0.94 N waveguiding
layer on either side of the quantum wells. The structure used a series of 3 quantum
wells asymmetrically positioned within the waveguide region to maximize their
interaction with the confined mode. 1200m long laser ridges were formed using a
Cl2 plasma etching process. Self-aligned SiOx was used to isolate the ridge
sidewalls using a conformal coating deposited by IBD, followed by liftoff of the
ridge photoresist.
Samples were then cleaned in an RCA SC1 (1:1:5 NH4 OH:H2 O2 :H2 O) solution
at 80C to remove all organic residue, followed by a dip in 1:1 HCl solution for 10
minutes to remove all oxide residue. Undoped ZnO films were grown on the laser
112

 substrates in Chapter 5. ZnO

ridges by the methods described for growing on 2021
patterning was easily achieved due to the propensity of ZnO to grow only on the
exposed p-GaN of the laser ridges, eliminating the need for post-deposition
etching procedures. The structures were annealed for improved conductivity in an
RTA system at 225C in N2 ambient. The ZnO thickness was measured by Dektak
profilometry to be over 400nm, which was substantially thicker than necessary to
ensure the lasing mode is fully attenuated before reaching the metal contacts.
Metal contacts were deposited using e-beam evaporation, using 10/1000nm Ti/Au
for the top-metal contacts to the ZnO cladding layer, and on the back-side of the
GaN substrate using 10/1000nm of Al/Au.
The substrates were then cut using a dicing saw for a target device length of
1200m to form bars of aligned laser ridges. These were mounted onto aluminum
chucks using mounting wax with sapphire shims to protect the device during
polishing. The laser facets were polished using diamond lapping films with 6, 3, 1,
0.3, 0.1, 0.05, and 0.02 grit sizes. The 0.05m and 0.02m lapping films consisted
of Al2 O3 and SiO2 , respectively. This polishing procedure was found previously
by Hsu et al. [2] to give well-polished, vertical facets with high reflectivities. The
devices were tested after polishing without any facet coatings, although it is
expected that facet coatings would substantially reduce the threshold currents of
the devices. SEM micrographs of the laser facets can be seen in Figure ??.

113

6.5

blue laser results

(2021)

Lasers with different ridge widths were tested. It was found that devices with
1.6m, 4m, and 8m wide ridges lased at threshold currents of 8.95, 8.79, and
10.4kA/cm2 , respectively, with threshold voltages of 10.8, 11.2, and 12.3V.
Generally, the threshold currents and voltages appeared to increase with increasing
ridge width. The lasers showed coherent emission as verified by optical
interference patterns that were observable on a white card held behind the laser
under operation above threshold, as shown in Figure 6.4(a).
The high threshold currents were most likely the result of poor confinement,
possibly due to MOCVD growth rates having drifted since calibration, resulting in
thinner than expected p-GaN. If the p-GaN is too thin, the optical mode reaches a
maximum lower than the quantum well positions, which would lead to higher
threshold current densities. The emission spectra above threshold for a 1.6m
wide ridge revealed the peak wavelength at 452nm, shown in Figure 6.4(b).
It is expected with further optimization of the ZnO cladding layers, potentially by
using Ga-doping to increase the conductivity, or Mg-doping to lower the refractive
index, it should be possible to increase the confinement in these structures, and to
obtain lower threshold currents.

114

 blue laser pulsed LIV characterization at 1% duty cycle, with

Figure 6.3: (2021)
(a) 1.6m, (b) 4m, and (c) 8m wide ridges. (d) Device threshold currents and
voltages vs. ridge widths.

115

Figure 6.4: (a) 1.6m device operating above threshold, and (b) Spectral narrowing
for a 1.6m wide laser ridge, for currents below and above the threshold current,
confirming lasing.

6.6

contact studies
(202 1)

One of the most promising planes for next-generation high-power GaN devices is
the 202 1 plane. Pourhashemi et al. [3] showed devices of this type with threshold
currents as low as 5.3kA/cm2 , and peak output powers of 2.15W at 28.1kA/cm2 for
8 900m ridges. One of the contributing factors to these high efficiencies was
the optimization of the p-GaN doping which lowered the absorption near the
quantum wells. It is expected that these structures could be further optimized
through the use of oxide cladding layers. Unfortunately, the results of Chapter 5

indicated that growing epitaxial ZnO films on 202 1 p-GaN led to an inverted 2021
crystal layer, which had very high voltages, behaving more like a reverse-biased
p/n junction than an Ohmic contact. However, it is possible that by using the same

116

two-step growths that were used in Chapters 3 and 4, a polycrystalline c-textured

film could be fabricated, which would have the same projected crystal polarization
as the underlying 202 1 p-GaN layers.

6.6.1

Experimental method

ZnO contacts were deposited on laser diodes grown by MOCVD on 202 1 bulk
GaN substrates supplied by Mitsubishi Chemical Co. The same patterning method
was used as in Chapter 5 to form well-defined circular 50m radius pads, using a
CTLM mask. The TLM method could not be used, because it only applies to
Ohmic contacts with constant specific contact resistivities. 3 samples were
prepared. One with Pd/Au metal contacts, one with a ZnO contact deposited by the
methods in Chapter 5 to form single-crystalline layers, and one with a two-step
growth method using a seed layer. The samples were cleaned in aqua regia, RCA
SC1 solution, and HCl prior to ZnO or Pd/Au deposition.
The seed layer growth proceeded in much the same way as was described in
Chapter 3, except the process was greatly accelerated by growing the films in a
glass beaker heated on a hotplate, instead of making use of a 90C oven, which
took hours to reach a growth-ready temperature. It was found that a hotplate set to
280C will cause a 50mL flat-bottomed beaker with 40mL growth solution to
reach a growth temperature of 90C, with the temperature being stable between
85C and 90C for long enough times to complete a seed layer growth. Seed layer
growths were found to reach completion due to depletion of Zn precursor through

117

powder formation after only a few minutes upon raising of the solution pH by
ammonia addition. Seed layers were annealed at 500in N2 , followed by a
second-layer growth of ZnO for a growth time of 1 hour. 10nm/500nm of Ti/Au
metal were then deposited by standard e-beam deposition on the top surface of the
ZnO pads, followed by 30nm/500nm of Al/Au metal on the backside of the
substrates.

6.6.2

IV characterization

It was found that the ZnO films grown with seed layers had drastically decreased
voltages as compared to the ZnO films grown epitaxially. IV curves of the Pd/Au,
seed-layer, and epitaxial ZnO films are shown in Figure 6.5. At a current density
of 500A/cm2 , the epitaxial ZnO films showed a voltage of 8.8V, compared to only
5.0V for the polycrystalline film, and 4.4V for the Pd/Au contact. At a current
density of 1500A/cm2 this difference was even larger10.4V, vs 6.0 and 5.0 for the
polycrystalline and Pd/Au layers, respectively. This represented a drastic
improvement in these types of contacts, and should allow ZnO to be used as a
cladding layer for these types of devices. While the voltages were not as low as
ITO layers deposited by e-beam evaporation at 250C, it would still be possible to
produce a functioning device using this type of layer. Furthermore, it could well be
possible to lower this voltage further using a Ga-doped seed layer, or a thin layer
of ZnO deposited by another deposition method entirely, such as MBE, or ALD.
In the case of a laser cladding layer, the current needs only to travel vertically

118

 GaN laser structures using 50m-radius

Figure 6.5: IV characterization on (202 1)
contact pads with backside contacts. Laser structures were grown by Arash
Pourhashemi.
through the film, meaning that the higher resistivity of a polycrystalline layer
would not lead to a significantly decreased performance vs. a single-crystal
cladding layer, provided that the optical absorption characteristics are similar.

6.7

Summary and Conclusions

In summary, ZnO, deposited by aqueous solution epitaxy, was successfully

demonstrated as a cladding layer for blue GaN-based laser diodes on (2021)

substrates. The ZnO was epitaxially grown on laser ridges using selective-area
119

growth, which simplified processing as compared to ITO layers which are

deposited by line-of-sight methods. 1200 1.6m blue laser ridges exhibited a
threshold current density of 8.9kA/cm2 and a threshold voltage of 10.8V. ZnO
contact layers were found to have greatly improved operating voltages on (202 1
laser structures when deposited using a two-step growth, using a high-nucleation
seed layer step, indicating promise that they could eventually be used as cladding
layers on this plane.

120

References
[1] Hardy, M. T.; Holder, C. O.; Feezell, D. F.; Nakamura, S.; Speck, J. S.;
Cohen, D. A.; DenBaars, S. P. Applied Physics Letters 2013, 103, .
[2] Hsu, P. S.; Farrell, R. M.; Weaver, J. J.; Fujito, K.; DenBaars, S. P.;
Speck, J. S.; Nakamura, S. Photonics Technology Letters 2013, 25,
21052107.
[3] Pourhashemi, A.; Farrell, R.; Hardy, M.; Hsu, P.; Kelchner, K.; Speck, J.;
DenBaars, S.; Nakamura, S. Applied Physics Letters 2013, 103,
151112151112.

121

Appendices

Optical Absorption calculations

The method used to calculate optical absorption coefficients is described by

Cisneros (1998). Many reports of TCO transparency only report the total
transmittance of a film of some thickness, which is wholly inadequate for use in
device modeling and simulation. This is because the transmittance, defined as the
ratio of light power transmitted through a structure to the amount incident on it,
does not have a simple relationship to any of the intrinsic material properties of the
structure. For example, a 100nm layer might exhibit 95% or higher transmittance,
but only 70% or less at a thickness of 1m. Furthermore, in films with thickness
on the order of the transmitted light, the transmission is greatly influenced by the
specific wavelength of transmitted light, due to interference effects within the
structure. In performing optical simulations of a laser or LED, it is of utmost
importance to have a reliable value available for the intrinsic optical absorption
coefficient in a film. This allows modeling of films with any thickness, and
therefore with any sheet resistance value. In theory, by using the bulk absorption
122

coefficient, the refractive index, the bulk electrical conductivity, and the contact
resistivity to the underlying device layers, it should be possible to develop accurate
models for LED transparent contact layers of any thickness, which is necessary for
the design of optimal devices.
A common method to approximate the absorption coefficient of films is to measure
the amount of light transmitted and reflected, and then to calculate the amount of
light absorbed by,

A=1T R

(A.1)

The amount of light absorbed is then equated to the amount of light that would be
absorbed in a diffuse, non-reflective medium, according to the Beer-Lambert Law:

A = 1 et

(A.2)

where t is the thickness of the film of interest. The absorption coefficient is

calculated by,

ln(T + R)
ln(1 A)
=
t
t

(A.3)

123

The difficulty with this method is in the inaccuracy of the reflection measurement.
To obtain precise values for R, it is necessary to calibrate the spectrophotometry
measurement to the reflection by a perfect mirror, and to assume that the
reflected beam shape will be identical to that of the reference beam. In general, the
accuracy of this measurement is within a few percent. In the example of a TCO
film with T + R = 95%, this A value is only 5%, where a 1% error amounts to a
20% error in the estimated A value. Suppose now that this 95% transmitting film
has 200nm thickness. The range of possible absorption coefficients, with the 20%
error in the A value, is about 2000cm-1 3100cm-1 . Now consider a different film
where the measured T + R = 98%. The 1% error now represents a 50% error. A
200nm film with T + R = 98% would have an absorption coefficient in the range
500cm-1 1500cm-1 . Evidently the 1% (or more) error added by the reflection
measurement creates an unacceptable range for the possible absorption
coefficients, as a simulated 500cm-1 film will diverge widely in performance as a
TCO layer from one with 1500cm-1 absorption. For this reason it is necessary to
only include the measured T values in the calculation of the absorption
coefficients, as this value is generally much more accurate.
Since there is no closed-form relationship between the transmittance of a film and
the optical absorption coefficient of a film, it is necessary to perform a
reverse-search by simulating the transmission of a film with varying absorption
coefficients, and fitting this simulated data to experimental transmission spectra.
The problem with this method is that the refractive index of both the film and the
transparent substrate will affect the amount of light reflected, and therefore not
transmitted by the structure. However, if there are interference fringes visible in
124

the spectrum, it is possible to make use of the fact that the film index value is most
sensitive to the transmission interference minima, while the absorption coefficient
is most sensitive to the transmission interference maxima. Crucially, the
transmission interference maxima are only weakly affected by the film index,
which allows the film index and absorption coefficients to be separately fitted to
the transmission minima and maxima, respectively.
The first step in the fitting procedure is to determine the refractive index of the
substrate. For this study, highly transparent MgAl2 O4 substrates were used, for
which we can assume no absorption. The index of the substrate in the absence of
absorption, and assuming incoherent reflections in the thick substrate, can be
calculated form the transmission spectrum of the bare substrate, T s , by,
1
1
+ 2 1
ns =
Ts
Ts

!1/2
(A.4)

where n s is the substrate index. The substrate index was then fitted to a Sellmeier
refractive index function to obtain a continuous fitted refractive index value across
the wavelength range of interest. The substrate transmission and refractive index
fitting are shown in Figure A.6(a) and (b).
After the substrate index function is determined, it will be used for all subsequent
modeling, assuming all substrates are identical. The next step in the process is to
determine an approximate value for the film index. The value only needs to be
approximate, because the sensitivity of the computed transmission function is only
weakly sensitive to the film index at the maxima of the film+substrate transmission

125

Figure A.6: (a) Transmission spectrum of bare DSP spinel substrate. (b) Fitted
substrate refractive index function.
spectrum. Since only an approximate index is needed, we can calculate the index
from the film transmission spectrum assuming no absorption in the film, which
again gives a closed-form expression, using the minima values of the transmission
spectrum, T min



1/2 1/2
nfilm = + 2 n2s

(A.5)

2n s 1 + n2s

T min
2

(A.6)

We now have approximate values for nfilm computed at the minima of the

126

Figure A.7: (a) Transmission curve with labeled interference extrema. (b) Computed phase function
interference fringes in the transmission spectrum. These values are used to, again,
fit a Sellmeier refractive index dispersion model, giving a continuous function for
the film index. This procedure is done independently for each film, since the
refractive index could be different between films. At this stage, it is also helpful to
index all the minima and maxima of the transmission spectrum with sequential
odd and even indices, respectively. By interpolating between these integer values,
it is possible to generate a phase function, which allows the generation of
continuous simulated transmission spectra based on the index and absorption
values calculated at the minima and maxima, respectively. Figure A.7(a) and (b)
show a labeled transmission curve, and a computed phase function used to
generate continuous simulated transmission spectra.
Finally, the transmission model is rearranged to isolate the film absorption
coefficient into a single term with exponential dependence. The transmission

127

through the substrate+film structure is given by,

T=

Aed
Be2d + Ced + D

(A.7)

A, B, C, and D in this case are values computed from the film and substrate
refractive indices, as well as the film absorption coefficient. Details concerning
how to compute these constants are in the Cisneros paper, but they are readily
computable based on the information we already have, with the exception of the
film absorption coefficient. Rearranging Equation A.7 we obtain,

1/2
!
!2
A

1 A

C +
C 4BD
=
2B T max
T max

(A.8)

When computed in this form, the results iterate rapidly. For the first iteration, the
A,B,C and D coefficients are calculated based on the film index, substrate index,
and a film absorption coefficient of zero. This gives a value for ed . By taking the
log and dividing by d, the film thickness, an initial estimate for the absorption
coefficient is obtained. This first estimate is then plugged back into the A, B, C,
and D coefficients to re-calculate the absorption coefficients, until the absorption
coefficient no longer changes upon further iteration. It was found that these values
converged very quickly, with only 2 or 3 iterations being necessary. The
absorption coefficient plots calculated in Figure 4.4 can then be plotted.
In order to understand the sensitivity of the calculated absorption coefficients to
errors in the experimental transmission values, and the estimated index values, the

128

Figure A.8: Error analysis of absorption coefficient modeled by Cisneros method.

When is changed by 20% from the calculated value there is a large shift in the
modeled data, no longer reproducing experiment at the interference maxima. The
model correctly touches the substrate spectrum for = 0.
transmission spectra were replicated with 20% errors in the film index and
absorption coefficient. These analyses are shown in Figures A.9 and A.8.

129

Figure A.9: Error analysis of film refractive index modeled by Cisneros method.
When nfilm is changed by 20%, the computed transmission value barely changes at
the even-order points. This verifies that the absorption coefficient can be computed
at these maxima with little dependence on the film index. Even when the film index
is moved to less than the substrate index, the model successfully reproduces the
transmission values at the even-order points.

130