growth and characterization of dilute nitride antimonides for long

GROWTH AND CHARACTERIZATION

OF DILUTE NITRIDE ANTIMONIDES

FOR LONG-WAVELENGTH OPTOELECTRONICS

A DISSERTATION

SUBMITTED TO THE DEPARTMENT OF

MATERIALS SCIENCE AND ENGINEERING

AND THE COMMITTEE ON GRADUATE STUDIES

OF STANFORD UNIVERSITY

IN PARTIAL FULFILLMENT OF THE REQUIREMENTS

FOR THE DEGREE OF

DOCTOR OF PHILOSOPHY

Homan Bernard Yuen

March 2006

c© Copyright by Homan Bernard Yuen 2006

All Rights Reserved

ii

iv

Abstract

The incredible explosion of bandwidth capacity in optical fiber networks has been

achieved by a combination of higher speed devices and wavelength division multi-

plexing. The major limitation to access this capacity is limited access by the user

(usually via modem) through local and metro area networks. The demand due to

a surge in internet usage can be met by increasing transmission speed. Although

high speeds are utilized in the fiber backbone, the cost of InP-based lasers is far too

expensive for the home user. The addition of nitrogen to InGaAs, forming GaInNAs,

reduces the bandgap and lattice parameter simultaneously, enabling much lower-cost

optoelectronic devices on GaAs substrates operating at 1.3 and 1.55 µm. However,

high-quality GaInNAs growth by molecular beam epitaxy has many challenges. Anti-

mony was added as a surfactant and an incorporated species during GaInNAs growth,

forming GaInNAsSb, dramatically increasing the material quality, and enabling the

fabrication of high performance 1.3–1.55 µm lasers on GaAs.

GaInNAsSb is a five element (quinary) semiconductor which provides a broad

range of alloy compositions that can be used in the quantum well and confining

barrier regions of quantum well devices. This thesis focuses on finding and under-

standing the optimum materials compositions, growth conditions, and annealing for

low threshold, high performance, long-wavelength lasers on GaAs.

GaInNAsSb quantum wells (QWs) in the laser active regions have three possible

QW barrier materials. GaAs and GaNAs barriers were used for GaInNAs devices,

but GaNAsSb was a new option since antimony was thought to improve all dilute ni-

trides. Further investigation into the growth parameters, the resultant material and

v

optical properties, and the heterojunction band offsets revealed the difficulties of uti-

lizing GaNAsSb for the QW barriers. Additional studies were performed combining

the different barrier materials with GaInNAsSb QWs to determine the advantages

and shortcomings of each option. We determined GaAs barriers to be optimal but

are more difficult to implement. GaNAs barriers, though not ideal, are the best

compromise.

Although antimony was considered a panacea for dilute nitride growth, it did not

always improve material, as with GaNAsSb. Additional investigation to antimony’s

role and proper utilization was performed. GaInNAs, with lower indium content,

has garnered attention for solar cell applications, but high concentrations of defects

make it difficult to implement due to low efficiency. Antimony was added for the first

time to low-indium GaInNAs in hopes of improving the optical quality. Surprisingly,

antimony led to a degradation of optical quality. The different behaviors of antimony

in GaInNAs with high and low indium concentrations were studied and the role of

antimony as a reactive surfactant was confirmed. It was concluded that although

antimony is beneficial in certain situations, minimization of antimony incorporation

in GaInNAsSb is a key parameter in improving optical quality.

In conjunction with these findings as well as several other discoveries, GaInNAsSb

(vertical cavity surface emitting lasers) VCSELs at 1.53 µm were grown. These are

the longest wavelength, monolithic, GaAs-based VCSELs to our knowledge. In ad-

dition, world-record low-threshold and high-power GaInNAsSb edge emitting lasers

operating at 1.55 µm were developed.

Growth of In(N,As,Sb) QWs and quantum dots were also grown by MBE for

the first time to explore the development of new alloys which could be employed in

biosensing applications which require light of wavelengths of 3 µm or longer. The

preliminary findings are presented.

vi

Acknowledgements

First and foremost, I must acknowledge my research advisor, Professor James S. Har-

ris (otherwise known as Coach). His expertise, knowledge, and intuition in semicon-

ductor materials, devices, and molecular beam epitaxy were an invaluable resource

for the work I was able to perform during my graduate career. However, Coach is

more than a fountain of semiconductor wisdom; he is a great mentor on life and a

great friend. It is this well rounded tutelage for which I am most indebted. Occa-

sionally, Coach has given some poor advice: ski and support the Stanford Cardinal.

I also would like to thank the members of my reading committee, Professors

Paul McIntyre and Mark Brongersma, as well as the members of my oral defense

committee, Professors William Nix and Mark Cappelli. Their feedback and sugges-

tions were valuable in further exploring and understanding the many facets of my

research. Professor Nix provided guidance in thin film mechanical processes and

Professor Cappelli, who served as the chair of my orals committee, was helpful in

understanding the nitrogen plasma properties.

I had the honor and privilege of working along side many great coworkers on the

GaInNAsSb project. Mark Wistey is a great source of knowledge on all fronts. He

amazes me in many ways with his wide ranging wisdom and know-how and has been

enormously helpful in lab. Mark’s patience and willingness to help others is a trait

I can only hope to emulate. Seth Bank and I came to Stanford at the same time, so

I had the opportunity to work together with him through all phases of my graduate

career. Seth’s knowledge of semiconductor physics (among many other areas) is

impressive and has been a great resource. In addition, he has been a colleague whom

I have been able to talk about everything and that has really made a big difference

vii

in my research and life. Finally, I would like to thank him for the numerous Chinese

food truck lunches, snowboarding through dense double black tree runs by accident,

and intellectual discussions about chalupas. Hopil Bae is probably one of the hardest

working people I know. He has helped me many times in lab with various tasks, some

of which were not glamorous. Hopil was also the first MBE guru I trained and has

made the mentor look good! I would also like to thank Vincent Gambin and Wonill

Ha for all the patience they had when they were teaching me everything there was

to know about dilute nitrides and molecular beam epitaxy. Vince Lordi was an

exceptional source of information on the theory of dilute nitrides, among many other

areas, and contributed greatly to my general understanding of the physics observed

in my crystal growths. Kerstin Volz, although only at Stanford for a short time, was

very knowledgeable in the behavior of antimony as well as the properties of OMVPE

and MBE dilute nitrides. I would also like to thank Lynford Goddard for his work

on GaInNAsSb laser characterization and to Tim Gugov for his TEM work on the

GaInNAsSb samples. Tomas Sarmiento and Evan Pickett were useful in lab and will

undoubtedly do great work in the future on this project.

There are many other people I would like to thank who were in the Harris Group,

but not on the GaInNAsSb project. Xiaojun Yu was a great source of knowledge

while we were taking the numerous MSE and EE classes our first few years and during

the preparation for the MSE qualifying exam. Seongsin Kim and Fariba Hatami were

helpful coworkers on the In(N,As,Sb) project. Vijit Sabnis, Evan Thrush, Meredith

Lee, Rafael Aldaz, Kai Ma, Qiang Tang, Xian Liu, and Rekha Rajaram were fellow

group mates with whom I had many useful discussions. While she may not be an

expert in III-V epitaxy, Gail Chun-Creech is an expert in making the group run as

efficient and painless as possible for all members in the group. It can be argued that

Gail is more important to the Harris Group than Coach himself! With her assistance

and gracious friendship, I was able to concentrate on my research.

I was also fortunate to have the opportunity to work with many collaborators

outside of Stanford. From Sumitomo Ltd., Akihiro Moto’s perspectives on the mate-

rial growth and devices and financial support were important to much of work found

in this dissertation. Robert Kudrawiec from Wroclaw University in Poland was an

viii

endless source of electronic measurements and results which revealed a great deal

about the dilute nitride materials we grew. I would also like to thank Alan Chin of

Eloret Corp. at NASA Ames for the In(N,As,Sb) PL measurements.

While we no longer see each other as much, my colleagues in my MSE group were

great classmates and friends. They helped me transition from Physics to MSE by

helping me with what I did not know and were also good company. Many thanks to

David Chi and his fantasy football knowledge, Yana Matsushita and her Japanese

candy, Pete Hess and his unending supply of electronic gadgets, Aditi Chandra, Juliet

Risner, Melissa Lai, and Eric Guyer. I cannot say enough about my MSE classmates.

My interests and friends outside of Stanford helped maintain a balance between

research and life. Many of my friends, especially Mark Wong, Cynthia Kao, and

Bryan Tolmachoff, provided support through my graduate career and I am grateful

for their company. I would like to show my appreciation to the California Golden

Bears football and basketball teams for their entertainment and their victories over

the Stanford Cardinal during my time here (something I did not see while at Berke-

ley). I would also like to thank my teammates on the Dragon Warriors dragon

boating club for a great time the last two years. The friendship of people like Karla

Choy, Mike Liu, Gloria Lee, Wendy Lai, Greg Moy, Kristin Sunamoto, Vicki Jew,

Yi-Ling Su, Leslie Loui, Chuck Chen and many others gave me a great sense of

community.

I would now like to thank the people most important in my life. Although I met

my girlfriend, Angie Lin, near the end of my graduate career, her support has been

tremendous during the chaotic times of the oral defense, the writing of this thesis,

and the still-on-going dreaded job search. I am glad she did not laugh at my Mango

Drop drink on our first date. Angie is a beautiful person and I am lucky to be with

her. And finally, I cannot express in words my gratitude towards my parents for

everything they have given me my entire life. From the day I was born to this very

moment I am writing this sentence, they have never been away from my side and

have always shown their unconditional support of my endeavors. Even when I saw

no hope in what I was doing, they were there to hold me up. I can only hope I do

as much good in my life as they have in theirs.

ix

x

Dedication

To Mom and Dad.

xi

xii

Contents

Abstract v

Acknowledgements vii

Dedication xi

1 Introduction 1

1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Semiconductor Lasers . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.2.1 P-i-N semiconductor laser basics . . . . . . . . . . . . . . . . 3

1.2.2 Edge emitting lasers . . . . . . . . . . . . . . . . . . . . . . . 5

1.2.3 Vertical cavity surface emitting lasers . . . . . . . . . . . . . . 6

1.3 Long Wavelength Optoelectronics Materials . . . . . . . . . . . . . . 8

1.3.1 Long wavelength active regions . . . . . . . . . . . . . . . . . 10

1.3.2 Distributed Bragg reflectors . . . . . . . . . . . . . . . . . . . 12

1.3.3 GaInNAs/GaAs . . . . . . . . . . . . . . . . . . . . . . . . . . 14

1.4 Dilute Nitrides . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

1.4.1 Anomalous band gap reduction with nitrogen . . . . . . . . . 15

1.4.2 Advantages of dilute nitrides . . . . . . . . . . . . . . . . . . . 18

1.5 Outline of Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

2 MBE Growth and Characterization of Dilute Nitrides 23

2.1 Molecular Beam Epitaxy . . . . . . . . . . . . . . . . . . . . . . . . . 24

2.1.1 Molecular beam epitaxy system . . . . . . . . . . . . . . . . . 25

xiii

2.1.2 Group-III sources: Al, Ga, In . . . . . . . . . . . . . . . . . . 27

2.1.3 Group-V sources: As, Sb . . . . . . . . . . . . . . . . . . . . . 28

2.1.4 Dopant sources: Be, C, Si . . . . . . . . . . . . . . . . . . . . 31

2.1.5 MBE tools and components . . . . . . . . . . . . . . . . . . . 31

2.1.6 Traditional III-V semiconductor growth . . . . . . . . . . . . . 36

2.2 Growth of Dilute-Nitrides . . . . . . . . . . . . . . . . . . . . . . . . 37

2.2.1 Radio-frequency nitrogen plasma source . . . . . . . . . . . . 38

2.2.2 GaNAs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

2.2.3 GaInNAs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

2.2.4 Surfactant Growth . . . . . . . . . . . . . . . . . . . . . . . . 45

2.2.5 GaInNAsSb . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

2.2.6 GaNAsSb . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

2.3 Characterization Methods . . . . . . . . . . . . . . . . . . . . . . . . 50

2.3.1 Reflection high-energy electron diffraction . . . . . . . . . . . 50

2.3.2 High resolution x-ray diffraction . . . . . . . . . . . . . . . . . 51

2.3.3 Secondary ion mass spectrometry . . . . . . . . . . . . . . . . 55

2.3.4 Photoluminescence . . . . . . . . . . . . . . . . . . . . . . . . 57

2.3.5 Electroreflectance and photoreflectance spectroscopy . . . . . 61

2.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

3 Nitrogen Plasma Pressure Optimization and Characterization 65

3.1 Plasma Physics Basics . . . . . . . . . . . . . . . . . . . . . . . . . . 65

3.2 GaInNAs Quality with Different Gas Flows . . . . . . . . . . . . . . . 66

3.2.1 Structural and compositional analysis . . . . . . . . . . . . . . 66

3.2.2 Photoluminescence measurements . . . . . . . . . . . . . . . . 69

3.3 Effects of Gas Flow Variation on the Nitrogen Plasma . . . . . . . . . 71

3.3.1 Ion count and energy measurements . . . . . . . . . . . . . . . 71

3.3.2 Material quality and plasma properties correllation . . . . . . 75

3.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

4 GaInNAsSb Quantum Well Barrier Investigation 77

4.1 Quantum Well Barrier Choices . . . . . . . . . . . . . . . . . . . . . . 77

xiv

4.2 GaNAsSb Growth Investigation and Characterization . . . . . . . . . 79

4.2.1 Initial growth characterizations . . . . . . . . . . . . . . . . . 79

4.2.2 Arsenic overpressure examination . . . . . . . . . . . . . . . . 85

4.2.3 Growth temperature examination . . . . . . . . . . . . . . . . 87

4.2.4 Antimony reduction for improved luminescence . . . . . . . . 91

4.3 Heterojunction Band Offset Measurements . . . . . . . . . . . . . . . 98

4.4 GaAs Barriers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

4.5 Analysis of Quantum Well Barrier Choices . . . . . . . . . . . . . . . 107

4.6 Quantum Well Barrier Comparisons . . . . . . . . . . . . . . . . . . . 110

4.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

5 Effects and Role of Antimony on GaInNAsSb 115

5.1 Improving GaInNAsSb Luminescence at 1.3 µm . . . . . . . . . . . . 115

5.2 Indium Concentration and Strain Effects on Antimony . . . . . . . . 118

5.2.1 Antimony variation with high indium GaInNAs(Sb) . . . . . . 120

5.2.2 Antimony variation with low indium GaInNAs(Sb) . . . . . . 123

5.2.3 Indium variation with constant antimony flux . . . . . . . . . 126

5.2.4 Antimony as a reactive surfactant . . . . . . . . . . . . . . . . 130

5.2.5 Growth interactions between antimony and indium . . . . . . 131

5.2.6 Minimization of Sb incorporation for improved luminescence . 132

5.3 Annealing Behavior and Lattice Strain . . . . . . . . . . . . . . . . . 133

5.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135

6 Long Wavelength Semiconductor Lasers 137

6.1 Low-Threshold GaInNAsSb QW Edge Emitting Lasers . . . . . . . . 137

6.2 GaInNAsSb Vertical Cavity Surface Emitting Lasers . . . . . . . . . . 142

6.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

7 Additional Applications of Dilute Nitride Alloys 145

7.1 Dilute Nitride for Biosensor Applications . . . . . . . . . . . . . . . . 145

7.1.1 InNSb and GaNSb . . . . . . . . . . . . . . . . . . . . . . . . 148

7.1.2 InNAsSb . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150

xv

7.1.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

7.2 Dilute Nitride for Solar Cell Applications . . . . . . . . . . . . . . . . 155

8 Conclusion and Future Work 161

8.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161

8.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163

Bibliography 176

xvi

List of Tables

3.1 Summary of growth conditions for the samples described in this study.

The gallium and indium growth rates for the three growth rate con-

ditions are listed. The√

’s represent the samples which were grown

with the designated growth rates and nitrogen gas flows. . . . . . . . 67

4.1 XRD and SIMS compositional results of GaNAs and GaNAsSb grown

under the normal 1.3 and 1.55 µm QW growth conditions. . . . . . . 82

4.2 Summary of antimony fluxes utilized, GaNAs(Sb) compositions ob-

tained from SIMS and HRXRD, and strain from HRXRD. . . . . . . 95

4.3 Summary of QW barrier investigation findings. These materials are

attainable in our current MBE system configuration. . . . . . . . . . 113

5.1 A summary of the growth conditions for the samples described in this

study. The intended indium composition and the applied antimony

fluxes are listed. A) Varying antimony under constant “high” indium

flux, B) varying antimony under constant “low” indium flux, C) vary-

ing indium under constant 1.0×10−7 BEP Torr antimony flux. . . . . 120

xvii

xviii

List of Figures

1.1 Example structure design of a ridge waveguide edge emitting laser.

Laser light propagation is in the transverse direction. . . . . . . . . . 5

1.2 Example structure design of a vertical cavity surface emitting laser.

Laser light propagation is in the surface normal direction. . . . . . . . 7

1.3 Maximum transmission distances of a light signal at 850, 1310, and

1550 nm with varying bit rates. . . . . . . . . . . . . . . . . . . . . . 9

1.4 Fiber loss in silica fiber as a function of wavelength. “Wet” refers

to fiber laid in the 1970s and 1980s which contained OH− impurities.

“Dry” refers to newer fiber without this impurity. . . . . . . . . . . . 9

1.5 Material dispersion in silica fiber and chromatic dispersion in band-

width dispersion-shifted fiber. . . . . . . . . . . . . . . . . . . . . . . 10

1.6 Band gap versus lattice constant for a variety of zincblende III-V and

IV semiconductors. Ternary alloys are shown as lines between their

respective binary consituents. . . . . . . . . . . . . . . . . . . . . . . 11

1.7 Band gap versus lattice parameter showing the effects of adding small

amounts of nitrogen to GaAs and InGaAs. . . . . . . . . . . . . . . . 16

1.8 Illustration in k -space of the band anticrossing effects on the nitrogen

level and GaAs conduction band. . . . . . . . . . . . . . . . . . . . . 17

1.9 The effects of different nitrogen concentrations in the dilute regime on

the E+ and E− levels in GaNAs. . . . . . . . . . . . . . . . . . . . . . 17

2.1 A top-view schematic of a Mod Gen II MBE chamber. Important

components are illustrated. . . . . . . . . . . . . . . . . . . . . . . . . 25

xix

2.2 Side-view configuration of the sources found in MBE systems for di-

lute nitride antimonide devices. Blue denotes group-III, green denotes

group-V, and red denotes dopant sources. . . . . . . . . . . . . . . . . 26

2.3 Monomeric antimony fraction as a function of cracker temperature

and antimony sublimator flux. . . . . . . . . . . . . . . . . . . . . . . 30

2.4 Pictures of internal parts of the MBE system. (a) An arsenic coated

source flange with the eight shutters. White shutters are made of PBN.

(b) Side view of a shutter coated with 3 mm of Al due to build-up

over time. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

2.5 Reflectivity of a semi-insulating GaAs wafer at different wavelengths

of light for different substrate thermocouple temperatures. The sharp

transition marks the band gap at that temperature. . . . . . . . . . . 35

2.6 Nitrogen content in a dilute nitride layer showing incorporation even

when the shutter is closed caused by “blow by” when the plasma is

running. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

2.7 Concentration of nitrogen in GaNAs as a function of group-III growth

rate. Plasma conditions are 300 W forward power and 0.5 sccm N2

flow, corresponding to K=0.8. . . . . . . . . . . . . . . . . . . . . . . 43

2.8 PL of GaInNAs(Sb) samples comparing the best 1.3 µm material

grown without antimony and the dramatic improvement in PL at

longer wavelengths by adding antimony. . . . . . . . . . . . . . . . . 49

2.9 Examples of RHEED patterns relating to different surface structures.

(a) Smooth surface, (b) rough surface, and (c) quantum dots. . . . . . 51

2.10 Diagram illustrating the geometry of a symmetric ω/2θ scan of (00l)

planes. ω is the angle between the incident beam and the surface while

2θ is the angle between the diffracted beam and the incident beam.

Q is the diffraction vector. . . . . . . . . . . . . . . . . . . . . . . . . 52

2.11 Diagram illustrating the three axes in the triple-axis configuration.

In a normal ω/2θ scan, the analyzer is not present and the direct

diffracted beam is detected. In triple-axis, a detector in a different

location measures the beam diffracted from the analyzer. . . . . . . . 55

xx

2.12 Diagram illustrating the direction of relaxation for GaNAs when ex-

amining the (224) diffraction peaks. . . . . . . . . . . . . . . . . . . . 56

2.13 Example (224) RSMs of (a) a perfectly coherent 80 A GaInNAsSb

QW on GaAs and (b) a partially relaxed 1 µm GaInNAsSb layer on

GaAs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

2.14 Illustration in momentum space of the basic carrier processes in PL. . 58

2.15 Annealing behavior for a GaInNAsSb QW as a function of anneal

temperature. RTA time was 60 seconds. . . . . . . . . . . . . . . . . 60

3.1 (004) ω/2θ scans of GaInNAs QWs grown at identical growth rates,

but different flow rates. (a) 0.25 sccm, (b) 0.50 sccm, and (c) 0.75

sccm gas flows. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

3.2 Nitrogen incorporation for different gas flow rates for GaInNAs QWs

at the same growth rate. The cracking efficiency is also plotted show-

ing a saturation past 0.50 sccm. . . . . . . . . . . . . . . . . . . . . . 68

3.3 Emission wavelength as measured by PL of different gas flow GaInNAs

QW samples at different annealing temperatures. . . . . . . . . . . . 70

3.4 Peak PL intensity with different anneal temperatures for the different

gas flow samples. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

3.5 Schematic of the Langmuir probe utilized in this study to analyze

plasma properties. The beam flux gauge is rotated towards the ni-

trogen cell and is nominally found in the same position as the wafer

during growth. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

3.6 Langmuir probe measurements of the plasma species exiting the cell

with different gas flows. . . . . . . . . . . . . . . . . . . . . . . . . . . 74

3.7 Maximum ion energies for the ions exiting the plasma cell as a function

of gas flow rate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

4.1 RHEED pictures showing the streaky patterns from (a) GaNAs and

the spotty patterns from (b) GaNAsSb. . . . . . . . . . . . . . . . . . 81

xxi

4.2 (004) ω/2θ HRXRD spectra showing the amount of strain in the sam-

ples. (a) GaN0.029As0.873Sb0.098, (b) GaN0.034As0.867Sb0.099, (c) GaN0.019-

As0.981, and (d) GaN0.027As0.973. (a) and (c) are grown under the 1.3

µm device growth conditions where as (b) and (d) are grown under

1.55 µm device growth conditions. . . . . . . . . . . . . . . . . . . . . 82

4.3 SIMS depth profile of antimony and nitrogen for a GaN0.029As0.873Sb0.098

QW sample. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

4.4 PL results from the GaN0.029As0.873Sb0.098 sample (barrier material

for the 1.3 µm QWs). The blue line shows PL intensity. The red line

shows the peak PL wavelength. . . . . . . . . . . . . . . . . . . . . . 85

4.5 SIMS results from the arsenic overpressure study. . . . . . . . . . . . 87

4.6 PL spectra from the GaN0.029As0.873Sb0.098 sample grown at different

arsenic-to-gallium overpressures. (a) 30×, (b) 25×, and (c) 15×. . . . 88

4.7 (224) reciprocal space map of the GaN0.029As0.873Sb0.098 sample grown

at high temperature (545◦C). No in-plane components from the QW

different from the substrate are seen in the diffraction pattern. . . . . 90

4.8 SIMS results from the growth temperature study. . . . . . . . . . . . 91

4.9 PL spectra from the GaN0.029As0.873Sb0.098 sample grown at different

substrate temperatures. (a) +35◦C (475◦C), (b) +70◦C (510◦C), and

(c) and +105◦C (545◦C). The small peak at 1400 nm is due to water

present in the testing environment. . . . . . . . . . . . . . . . . . . . 92

4.10 (004) ω/2θ HRXRD spectra of the four GaNAs(Sb) layers. (a) GaN0.0063-

As0.9937, (b) GaN0.0071As0.9869Sb0.006, (c) GaN0.008As0.978Sb0.014, and

(d) GaN0.0091As0.9709Sb0.02. The tensile strain decreases with increas-

ing antimony flux. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

4.11 (004) ω/2θ HRXRD spectrum of the GaN0.0063As0.9937 with its corre-

sponding simulated spectrum. . . . . . . . . . . . . . . . . . . . . . . 94

4.12 (224) reciprocal space map of the GaN0.0063As0.9937 sample. . . . . . . 95

4.13 PL spectra of the GaNAs(Sb) samples showing a redshift and increase

in intensity with increasing antimony flux. . . . . . . . . . . . . . . . 97

xxii

4.14 PR spectra obtained from GaN0.02As0.87Sb0.11/GaAs QW samples. (a)

6 nm, (b) 8 nm. Shown are the experimental data, theoretical spectra

fit (in red), and moduli of the PR energy resonances. . . . . . . . . . 100

4.15 Band lineup for the GaN0.02As0.87Sb0.11/GaAs QW samples. The nu-

merical values for the offsets have taken strain into account. . . . . . 101

4.16 The effects of varying antimony concentration on the (a) conduction

band offset ratio Qc and (b) valence and conduction band offsets. . . 101

4.17 Band lineup for Ga0.62In0.38N0.026As0.954Sb0.02/GaAs QW sample. Nu-

merical values have taken strain into account. The energy transitions

for the four confined states are also shown. . . . . . . . . . . . . . . . 102

4.18 Band lineup of the Ga0.61In0.39N0.023As0.957Sb0.02/GaN0.027As0.973/GaAs

stepped QW sample. Numerical values have taken strain into account. 103

4.19 Band offset comparison of GaNAs, GaAs, and GaNAsSb. Both GaNAs

and GaAs are type-I to GaInNAsSb, but GaNAsSb is possibly type-II. 105

4.20 (004) ω/2θ HRXRD of a GaInNAsSb SQW on GaAs with 2.6% lattice

strain. Even without tensile barriers, the material remains structurally

good. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

4.21 SIMS depth profile of a GaInNAsSb/GaAs SQW. The indium profile

defines the QW region. Antimony incorporation is found outside of

the QW. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

4.22 PL intensities of GaInNAsSb SQWs with GaAs, GaNAs, or GaNAsSb

barriers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

4.23 PL intensities of GaInNAsSb/GaAs and GaInNAsSb/GaNAs SQWs

as a function of annealing temperature. All anneals were for 60 s. . . 111

5.1 HRXRD spectra of (004) ω/2θ scans of GaInNAsSb/GaNAs QWs

with different concentrations of nitrogen, indium, and antimony, but

the same 1.3 µm emission wavelength. . . . . . . . . . . . . . . . . . 117

5.2 Annealing behavior on the PL intensity of the GaInNAsSb/GaNAs

QWs. The lower nitrogen concentration sample has much higher in-

tensity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

xxiii

5.3 HRXRD spectra of the (004) GaInNAs(Sb)/GaAs QWs with “high”

indium compositions. . . . . . . . . . . . . . . . . . . . . . . . . . . . 122

5.4 Indium, nitrogen, and antimony compositions as a function of anti-

mony flux. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123

5.5 PL spectra of GaInNAs(Sb) samples under high indium, high strain

conditions with varying antimony flux. . . . . . . . . . . . . . . . . . 124

5.6 HRXRD spectra of the (004) GaInNAs(Sb)/GaAs layers with “low”

indium compositions. . . . . . . . . . . . . . . . . . . . . . . . . . . . 125

5.7 Indium, nitrogen, and antimony compositions as a function of anti-

mony flux utilized during the QW growth in the “low” indium com-

position range. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126

5.8 PL spectra of GaInNAs(Sb) samples under low indium, low strain

conditions with varying antimony flux. . . . . . . . . . . . . . . . . . 127

5.9 HRXRD spectra of the (004) GaInNAs(Sb)/GaAs QWs with varying

indium fluxes under a constant antimony flux. . . . . . . . . . . . . . 128

5.10 Nitrogen and antimony compositions as a function of indium concen-

tration with the antimony flux held constant. . . . . . . . . . . . . . 129

5.11 PL spectra of GaInNAs(Sb) samples with a constant antimony flux

with varying indium concentrations. . . . . . . . . . . . . . . . . . . . 129

5.12 Reduction of the optimal anneal temperature with increasingly strained

GaInNAsSb QWs due to larger (a) antimony or (b) indium concen-

trations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134

6.1 Laser spectrum of a GaInNAsSb/GaNAs/GaAs 1.56 µm EEL at 1.2×threshold. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

6.2 L-I curve and wall plug efficiency for the GaInNAsSb/GaNAs/GaAs

1.56 µm laser under cw conditions. . . . . . . . . . . . . . . . . . . . 140

6.3 Comparison of data from our work and devices found in the literature.

The effect on material and device improvement can be seen in the

reduction of the threshold current density. . . . . . . . . . . . . . . . 141

xxiv

6.4 Laser spectrum of a triple GaInNAsSb/GaNAs QW 1.534 µm VCSEL

at 1.6× threshold. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

6.5 L-I curve for a triple GaInNAsSb/GaNAs QW 1.534 µm VCSEL op-

erating under pulsed conditions. . . . . . . . . . . . . . . . . . . . . . 144

7.1 Band gap versus lattice constant illustrating the region of III-V semi-

conductors which can be used to obtain emission at mid to far-IR

wavelengths. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

7.2 (004) ω/2θ HRXRD spectra of thick InSb and InSb:N layers on GaAs.

There is no difference in diffraction angles. . . . . . . . . . . . . . . . 149

7.3 (004) ω/2θ HRXRD spectra of thin GaSb and GaNSb films on InAs.

While nitrogen was found in GaNSb, the structural quality was quite

poor. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151

7.4 (224) RSM of the GaNSb film on InAs indicating relaxation. . . . . . 152

7.5 (004) ω/2θ HRXRD spectra of 500 A InAsSb and InNAsSb films on

InAs. The nitrogen containing sample appears to be of good structural

quality. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

7.6 Low-temperature PL of an InNAsSb QW on InAs. The QW peak is

located at 4 µm. The dip in luminescence of the QW peak located

between 4.25–4.5 µm is due to CO2 absorption in the ambient. . . . . 154

7.7 Band alignment of strained InAs0.9Sb0.1 on InAs showing a type-II

alignment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154

7.8 The solar spectrum as observed below Earth’s atmosphere (AM 1.5).

(a) The current design of three junction III-V solar cell devices. (b)

A proposed four junction device with an added 1.0 eV junction for

increased efficiency. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156

7.9 (224) RSM of a 2 µm thick GaInNAs layer on GaAs. No relaxation is

observed. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157

7.10 (224) RSMs of relaxed (a) GaInNAs and (b) GaInNAsSb P-i-N struc-

tures on GaAs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158

xxv

7.11 (002) dark field cross-sectional TEM tilted slightly off-axis of the re-

laxed GaInNAsSb sample. A network of misfit dislocations can be

seen on the top and bottom interfaces of the GaInNAsSb layer. . . . 159

xxvi

Chapter 1

Introduction

1.1 Motivation

Growing quantities of information content in combination with an increasingly “wired”

population have led to the tremendous expansion of Internet traffic since the begin-

ning of the millennium. With high quality audio and video streaming becoming

more common and popular, the volume of data transmitted to computers around

the country will lead to a shortage in available bandwidth. Increased data trans-

mission speeds and bandwidth will be required in order to facilitate the burgeoning

demand in future years.

Most of the data traffic in the United States travels through a fiber-optic net-

work. Lasers generate the light signals which pass through the fiber cable, trans-

mitting data as a sequence of photonic packets. Long-haul communications (the

“fiber backbone”), wide area networks (WANs), metro area networks (MANs), and

now even some local area networks (LANs) utilize fiber communications due to its

advantageous speed and bandwidth properties. Unfortunately, the structure of the

fiber-optic network is not optimized for efficient usage. Some LAN and almost all

“last-mile” connections are not fiber enabled. Many end and home users connect to

the Internet using cable modem, DSL, or dial-up services. These technologies are

100-10,000× slower than a fiber connection and create a “bottleneck” which greatly

hinders high-speed data transmission from one computer to another.

1

2 CHAPTER 1. INTRODUCTION

Enabling fiber connections to the home-user would eliminate the fiber bottleneck

and allow for high-speed, high bandwidth communications. The main difficulty with

this realization is the cost required to bring fiber to the home. The cost of current

laser technology is quite high, but is not an impediment in expanding long-haul

communication and WAN capabilities. The sales volume of these lasers is relatively

small and the funding is provided for by large companies. On the other hand, the

high cost does prevent most home users from even considering this technology. While

many grumble about having to wait for a page or a very large file to download

onto their computer, users do not value the “wait” time to be worth the thousands

of dollars it would cost to have a fiber connection. The key to eliminating the

fiber bottleneck is to provide low-cost device technology that is optimized for fiber

communications.

1.2 Semiconductor Lasers

The laser is one of the most remarkable scientific and technological advances of the

20th century. After significant contributions from Albert Einstein, Charles Townes,

Gordon Gould, and several others, Theodore Maiman created the first working laser

using a solid-state flashlamp-pumped synthetic ruby crystal in 1960 at Hughes Re-

search Laboratories. Since then, the field of lasers has diversified extensively, using

many different methods and materials to create lasing action. The semiconductor

laser was first proposed by Basov and Javan and the first laser diode was demon-

strated by Robert Hall at General Electric Laboratories in 1962. This GaAs-based

device emitted light at 850 nm, but required liquid nitrogen cooling and could only

operate under pulsed conditions. The first semiconductor heterojunction laser was

independently developed by Zhores Alferov in the former Soviet Union and Mort

Panish and Izuo Hayashi at Bell Laboratories in 1970, leading to continuous wave,

room temperature operation of the laser diode.

When Maiman created the first laser 45 years ago, no one could have imagined

the wide range of applications or the ubiquity of lasers found in today’s technology.

From grocery checkout stands to communications to nuclear fusion experiments to

1.2. SEMICONDUCTOR LASERS 3

light shows in Las Vegas, the laser has affected almost every facet of life. A laser

differs from other common sources, such as incandescent or fluorescent light bulbs by

emitting light which is coherent and monochromatic. These fundamental properties

of a laser immediately lend it to certain applications (including communications)

that can utilize single wavelengths of light and can be focused down to its diffraction

limit.

1.2.1 P-i-N semiconductor laser basics

The detailed specific operational principles behind different lasers may vary, but

the fundamental theories are the same. A system with a ground state and one or

more excited states exists for electrons to occupy. Typically, the electrons are found

in the lowest energy configuration, or in the ground state. They can be pumped

into higher energy states, but any electrons found in excited states can release their

energy through spontaneous emission and return to the ground state. However, if the

lifetime of the electron in the excited state is long enough such that a critical number

of electrons are found in the higher energy level, a population inversion results.

Throughout the process, electrons in this higher energy level will decay down to the

ground state and emit a photon. With a population inversion present, absorption

is no longer a dominant mechanism and a photon can induce or stimulate another

electron to de-excite back to the ground state with the same coherency. As these

photons pass through the optical gain medium, the process continues, stimulating

additional photons. If the light is allowed to pass through the optical gain medium

repeatedly in a resonant cavity, enough light will be stimulated to create lasing action.

This, in essence, is why a laser stands for light amplification by stimulated emission

of radiation.

Lasers made from semiconductors are utilized for fiber communications because

of their ability to be made in extremely small sizes. All laser devices have three main

components: a pump, an optical gain medium, and a resonant cavity. For semicon-

ductor lasers, the pump creates a population of electrons in the conduction band and

a population of holes (or missing electrons) in the valence band. The generation of


the electrons and holes can be done optically or electrically. In optical excitation,

carriers are created when light of energy larger than the band gap of the semicon-

ductor is absorbed by electrons. The electrons are excited into the conduction band,

leaving a corresponding hole in the valence band. Carriers can also be electrically

injected into a pn junction device with the application of an electrical current and

voltage. Under forward bias, electrons are injected into the gain region from the

n-doped region and holes from the p-doped region.

The optical gain medium of the laser device is the component which creates the

light needed for lasing action. Operational parameters, such as wavelength, are de-

pendent upon the semiconductor material found here, requiring most of the scientific

analysis when initially designing lasers for different applications. Typically, the gain

region is found in the intrinsic region of a P-i-N diode heterostructure. The semicon-

ductor material is of smaller band gap than the adjacent n-type and p-typed doped

layers. One type of heterostructure device consists of a quantum well (QW) struc-

ture. The gain medium is a very thin (≤10 nm) layer of the smaller band gap material

surrounded by the larger band gap material, forming a quantum well in which in-

jected carriers from the doped regions can be locally confined. The recombination

of these confined carriers gives off light (photons) or lattice vibrations (phonons).

Initially, only spontaneously emitted light originates from the active region. With

larger currents, carriers are pumped into higher levels than the band minima and

eventually degeneracy is achieved. A band gap photon cannot be absorbed as it

travels out of the device allowing for stimulated emission.

The last main component of a laser is the resonant cavity. Without this third

portion of a laser, lasing action cannot occur with only the pump and active gain

region. The purpose of the resonant cavity is to provide positive feedback for light

created in the gain medium to stimulate additional emission. A resonant cavity is

made by placing reflective mirrors on both sides of the active gain region along the

path of light propagation. For semiconductor lasers, these mirrors can be the cleaved

facets of wafers (in-plane emission) or repeating stacks of different semiconductor

layers (normal emission). With very high reflectivity mirrors, enough light can be

made to travel back-and-forth through the gain medium. Gain will exceed loss and


Bottom n-type cladding

Top p-type cladding

GaAs n+ substrate

Top metal contact

Bottom metal contact

Active region

Figure 1.1: Example structure design of a ridge waveguide edge emitting laser. Laserlight propagation is in the transverse direction.

lasing occurs. Typically, one of the mirrors will have slightly lower reflectivity to let

the laser light escape.

1.2.2 Edge emitting lasers

A commonly produced class of semiconductor lasers is the edge emitting laser (EEL).

From a design perspective, it is a fairly simple structure to develop and fabricate. One

example of an EEL, a ridge waveguide EEL, is shown in Figure 1.1. It is typically

a P-i-N double heterostructure device utilizing one or more QWs in the intrinsic

region as the active gain medium. Probes are placed onto the top and bottom metal

contacts, providing good electrical connections. Current flows from the p-type top

cladding layers through to the bottom n-type bottom cladding layers and substrate.

Electrons and holes are injected into the QW active region from this applied current.

The resonant cavity is created using the waveguide structure and the cleaved

facets as mirrors. An advantage of the double heterostructure is that the large

bandgap semiconductor has a smaller refractive index compared to small bandgap

active region, which has a larger refractive index. EELs are operated in an in-plane


orientation, so the light created within the active region is index guided along the

direction of propagation. The two cleaved facets at both ends of the waveguide struc-

ture form the mirrors of the Fabry-Perot cavity. With sufficient current densities,

light created in the active region makes many trips in the gain medium, producing

lasing action. Since light travels in the in-plane direction, it passes through several

hundreds of microns of gain region, improving efficiency. Light is also emitted in

the transverse direction, leading to the name of edge emitting laser. Due to the

device structure, the beam profile is an ellipse and can lead to some difficulties when

coupling the device to the fiber.

1.2.3 Vertical cavity surface emitting lasers

Vertical cavity surface emitting lasers (VCSELs) are another class of semiconductor

lasers. As shown in Figure 1.2, the VCSEL structure has several key differences

compared to the EEL structure. Light propagation is along the surface normal

direction, or vertically, rather than the in-plane direction. Correspondingly, the

mirrors which help form the resonant cavity are layers which are grown below and

above the active region containing the gain medium. In a VCSEL, the light travels

through a very small distance through the gain medium in each pass, so the mirrors

must be extremely reflective (≥99.5%) and the active region must have very high

gain. Due to these restrictions, the reflectivity of the semiconductor-to-air interface

is no longer sufficient as it was for EEL design. Even high-reflectivity coatings and

metallic mirrors which can achieve ∼98% reflectivity are inadequate for VCSELs.

Rather than utilizing single interfaces to provide the high reflectivities needed

for VCSELs, repeating layer stacks called distributed Bragg reflectors (DBRs) are

used. Semiconductors and dielectric materials have very low absorption coefficients

for light which have energies below their band gaps. If two of these materials with

different refractive indices are grown on top of each other, light will be reflected at the

interface. The amount of light reflected at this single interface is small. However,

if this coupled layer is grown into a periodic structure consisting of layers with a

thickness of λ/4n, the reflections from these many interfaces will add in phase and


Active Region

Distributed Bragg Reflectors

Metal Contacts

Figure 1.2: Example structure design of a vertical cavity surface emitting laser. Laserlight propagation is in the surface normal direction.

produce a very high reflection coefficient. The number of layers required depends on

the desired reflectivity and refractive index contrast. Other factors such as thermal

and electrical conductivity and epitaxial quality also play an important role. These

repeating periodic structures are the DBRs utilized in VCSEL structures.

VCSELs are often used in fiber communication due to several advantages over

EELs. These lasers have the potential to have much better performance character-

istics and lower production cost. Testing of the devices is also much easier as they

can be tested on the actual wafer without the need for individual packaging as with

EELs, drastically reducing the cost per laser. VCSELs do not suffer from the same

elliptical beam profile due to their structure design and can be fiber coupled more

efficiently. They also have sufficient laser mode spacing for single mode operation

as well as the ability to be grown and operated in array configurations. However,

VCSELs are generally much more difficult to develop and fabricate due to the tighter

tolerances required for efficient operation. The active region must be of very good

quality and have very high gain and the DBR mirrors must be grown to exact spec-

ifications to obtain the reflectivities needed. Additional details will be discussed in

the following sections.

For a more advanced treatment of EELs and VCSELs, please refer to a well-read

work by Coldren and Corzine [1].


1.3 Long Wavelength Optoelectronics Materials

It is very important to consider the wavelength of light used in fiber communication.

The VCSELs most commonly and easily produced today are based on GaAs/AlGaAs

for 850 nm emission and InGaAs/GaAs for 980 nm emission. While these VCSELs

are well commercialized, they are not applicable for high transmission speeds in

fiber. At 850 nm, it can be seen in Figure 1.3 that the transmission distance at high

data transfer speeds suffers due to attenuation and dispersion [2]. This property

of 850 nm light in the fiber limits its usefulness to very low speed transfers (tens

of kilometers below 10 Mb/s) or very short transmission distances (tens of meters

above 1 Gb/s). Longer wavelengths in the range of 1300–1600 nm have much higher

transmission distances due to decreased effects from attenuation and dispersion. At

these wavelengths, even at very high bit rates, the transmission distances are on the

order of a hundred kilometers. This further enhances network speed and efficiency

by reducing the need for additional components such as repeaters or amplifiers.

The reason for the increased transmission distances in this long-wavelength re-

gion, specifically at 1310 and 1550 nm, can be seen in Figures 1.4 and 1.5. At 1550

nm, fiber loss is at a minimum, reducing attenuation of the light signal as it passes

through the fiber. Light can travel long distances before the signal is no longer dis-

cernable from a noise background. There is also a local minimum in loss near 1310

nm. An OH− absorption peak dominates at ∼1380 nm for fiber laid in the 1970s

and 1980s, but is no longer present for newer fiber due to technological advances in

silica purification. 1310 nm light experiences zero material dispersion in standard

single-mode and multimode fiber while 1550 nm light experiences zero chromatic

dispersion in bandwidth dispersion-shifted fiber. This minimizes the degradation of

a small-width pulse as it travels through the fiber. With these advantages, it is clear

VCSELs which operate at 1310 nm and 1550 nm are the desired devices to drive the

push to bring low-cost fiber connections to the home user.

1.3. LONG WAVELENGTH OPTOELECTRONICS MATERIALS 9

0.01

0.1

1

10

100

1000

0.01

0.1

1

10

100

1000

0.1 1 10 100 1000 10000

Attenuation Limited

Dispersion Limited

1 Gigabit

1550 nm

1310 nm

Bit Rate (Mb/s)

Tra

nsm

issio

n D

ista

nce

(km

)850 nm

10 Gigabit

Figure 1.3: Maximum transmission distances of a light signal at 850, 1310, and 1550nm with varying bit rates.

Figure 1.4: Fiber loss in silica fiber as a function of wavelength. “Wet” refers to fiberlaid in the 1970s and 1980s which contained OH− impurities. “Dry” refers to newerfiber without this impurity.


1100 1200 1300 1400 1500 1600-50

-40

-30

-20

-10

0

10

20

301.1 1.05 1 0.95 0.9 0.85 0.8

Dis

pers

ion

(ps/

nm/k

m)

Wavelength (nm)

Chromatic Dispersion

Material Dispersion

Energy (eV)

Figure 1.5: Material dispersion in silica fiber and chromatic dispersion in bandwidthdispersion-shifted fiber.

1.3.1 Long wavelength active regions

To obtain VCSELs at 1.31 and 1.55 µm, the semiconductor must have a band

gap such that radiative carrier recombination results in emission near those two

wavelengths. Shown in Figure 1.6 are several III-V and IV semiconductors which

have band gaps that can emit light at the desired fiber wavelengths: SiGe, InGaAs,

GaAsSb, InAlAs, InAsP, AlGaSb, InAlSb, etc. However, many choices are not fea-

sible. Both silicon and germanium are indirect semiconductors and do not have

efficient radiative recombination processes. High quality growth of the semiconduc-

tor is also an issue to ensure efficient device performance. This eliminates some

mixed group-V alloys (such as phosphide-antimonides) due to miscibility issues. In

addition, the limited availability of substrates dictates the materials which may be

grown coherently. Large differences in lattice constants lead to the introduction of

deleterious mechanical defects. Finally, although not directly related to the active

region itself, the material system chosen for long-wavelength emission must also have

a compatible DBR material system. This issue will be discussed in the next section.


AlAs

AlSb

GaAs

GaSb

InP

GaP

10

2

1

0.5

0.0

0.5

1.0

1.5

2.0

2.5

5.4 5.6 5.8 6.0 6.2 6.4

AlPB

and

Gap

(eV

)

Lattice Constant (Å)

Wav

elen

gth

(µm

)

InSbInAs

Si

Ge

Fiber Wavelengths

Figure 1.6: Band gap versus lattice constant for a variety of zincblende III-V and IVsemiconductors. Ternary alloys are shown as lines between their respective binaryconsituents.

Current long-wavelength technology employs the InGaAsP alloy grown on InP

substrates. InGaAsP has been able to reach 1.55 µm and 1.31 µm, with slightly

more difficulty. While it has had success in EELs, InGaAsP does have disadvan-

tages including cost, performance issues, and VCSEL integration difficulties. The

properties of InGaAsP lasers have strong temperature dependences [3] due to the

heterojunction band alignment to InP. InGaAsP has a relatively small conduction

band offset to InP of ∆Ec=0.4∆Eg [4, 5]. Electrons have low mass and are more

susceptible to escape the confinment of the QW with sufficient thermal energy. As

the temperature increases, as it does during operation, electrons will leak out of the

QW decreasing efficiency and power by reducing the available gain. To compensate

for the decreased gain, additional current is required, further increasing the temper-

ature of the active region. To prevent a thermal run-away process and ensure stable

operation, InGaAsP lasers require external cooling packages. This unfortunately in-

creases the cost and makes monolithic integration with other devices more dfficult.


In addition, as a mixed group-V system, InGaAsP growth is extremely dependent

upon many growth parameters such as growth rate, substrate temperature, and flux

ratios. This high sensitivity decreases yield, further increasing the production cost.

InP substrates cannot be made in large diameters reliably and are expensive com-

pared to GaAs. Thus, the largest problem with widespread distribution of InGaAsP

lasers is cost. These lasers cost several hundreds to thousands of dollars and will

never enter the home-user market.

Several other III-V material systems have been examined for their application in

long-wavelength optoelectronics. InAl(Ga)As on InP has a larger conduction band

offset compared to InGaAsP, but growth is challenging due to miscibility issues,

surface segregation, and atomic ordering [6, 7]. GaAsSb/GaAs continuous wave (cw)

EELs have been demonstrated at 1.3 µm [8], but adding additional antimony to push

to 1.55 µm leads to a type-II band offset. Electron confinement is a difficult or non-

existent in GaAsSb/GaAs systems resulting in poor device performance. InGaAs(Sb)

on GaAs has been shown to reach 1.27 µm emission [9], but pushing to 1.3 µm

requires the addition of more indium and antimony, enlarging the lattice constant

past a critical point at which it can be grown coherently. Alternatively, In(Ga)As

quantum dots (QDs) on InP [10, 11] and GaAs [12, 13] have also been utilized

for 1.3 and 1.55 µm devices, but QD size control and QD wetting layer radiative

recombination degrades their performance. While QD lasers have very low threshold

current densities, their lack of uniformity and low gain make them difficult for use

in VCSELs.

1.3.2 Distributed Bragg reflectors

There are several factors which must be considered when choosing the materials for

the alternating layers in DBR structures. The DBR materials must have a lattice

constant very close to that of the substrate so that several microns can be grown with

very little strain, preventing the formation of structural defects. This generally places

great limitations on the available materials for each substrate. The band gap of the

materials used in the DBR structure must also be larger than the light emitted from


the active region to minimize loss in the laser diode. Finally, the alternating materials

must have sufficient refractive index contrast to minimize the number of mirror pairs

needed. Excessive numbers of mirror pairs creates a very thick DBR structure,

potentially degrading performance due to electrical and thermal conductivity issues.

The InP-based lasers utilizing InGaAsP as the active region have had several

problems with VCSEL integration, including the availability of feasible DBR mate-

rials. Examining Figure 1.6, it can be seen that the range of band gaps for alloys

which can be grown lattice matched on InP and be transparent to light from the

InGaAsP active region is relatively small. A small band gap difference implies a

small refractive index contrast, making it difficult to obtain mirrors with sufficient

reflectivity. For InP based layers, the InGaAsP/InP and InAlGaAs/InAlAs mirror

systems have been most commonly studied and utilized. However, due to a lack of

sufficient refractive index contrast, greater than 50 mirror pairs are required. These

materials have relatively low thermal conductivity and in combination with very

thick DBRs, removing heat from an already temperature sensitive InGaAsP active

region becomes a very high priority requiring external cooling.

To overcome these problems, alternatives have been attempted which do not in-

volve coherent growth of DBR materials: wafer fusion [14], metamorphic growth

[15], and dielectric mirrors [16]. Wafer fusion utilizes GaAs-based DBRs with supe-

rior performance and bonds the DBRs directly to the InGaAsP active region. This

dramatically increases cost due to the processes which must be undertaken and can

potentially hinder device performance due to interface degradation. GaAs-based

DBRs are also used in metamorphic growth on InP layers, but minimization of dis-

location formation from the large lattice mismatch is difficult. In addition, only

the top layer may be produced in this fashion. Dielectric mirrors can be deposited

onto the active regions, but these materials tend of have low thermal and electrical

conductivities, making efficient device operation difficult.

GaAs-based DBR structures have far superior characteristics compared to those

used for InP-based lasers. Fortuitously, AlAs has a lattice constant very close to

GaAs allowing for thick coherent growth of Al(Ga)As layers on GaAs without the

formation of dislocations. In addition, the band gap of AlAs is much larger than


GaAs. The refractive index contrast is sufficient to obtain high reflectivity DBRs

and the Al(Ga)As alloy is transparent to all light with band gaps smaller than GaAs.

Only 20–30 mirror pairs are required to obtain 99.9% reflectivity, many less than the

InP-based DBRs. Since these DBRs are based on binary or ternary alloys rather than

quaternary alloys, thermal conductivity is generally increased. Heat removal from

the active region is improved with both increased thermal conductivity and smaller

thicknesses of the DBR stacks, decreasing effects due to thermal sensitivity of the

active region. AlAs also has a well controlled oxidation reaction forming AlxOy,

enabling current confinement in VCSELs, improving laser threshold currents.

1.3.3 GaInNAs/GaAs

InP-based materials systems are not practical options for low-cost high-performance

VCSELs for fiber communication. Their temperature sensitivity and lack of DBR

alloys lead engineers in search of an alternative material system which can realize

the goal of bringing fiber to the home. Ideally, a GaAs-based system with minimized

temperature sensitivity utilizing the AlAs/GaAs DBR system would be the most ad-

vantageous path towards that goal. As mentioned earlier, GaAsSb is not feasible due

to band alignment issues [17–21] and InGaAs cannot reach the fiber communications

without suffering from lattice relaxation if grown on GaAs. There are no traditional

III-V GaAs-based semiconductors which can be grown coherently and have a small

enough band gap to emit light at 1.3 and 1.55 µm.

In 1992, Weyers et al. discovered that adding small amounts of nitrogen to GaAs

had the strange effect where the emission wavelength of the material redshifted [22].

The reduction of lattice parameter typically increases the band gap, blueshifting the

emission wavelength. In 1996, Kondow et al. discovered adding small amounts of

nitrogen to InGaAs formed GaInNAs, a new material which could be grown coher-

ently on GaAs while emitting light at the desired fiber communication wavelengths

[23]. This discovery enabled the development of GaAs-based long-wavelength opto-

electronics.

1.4. DILUTE NITRIDES 15

1.4 Dilute Nitrides

GaAs-based alloys with small amounts of nitrogen (≤5%), also referred to as dilute

nitrides, have garnered great interest for their non-traditional behavior compared

to other semiconductors. The dramatic reduction of band gap with the addition

of small amounts of nitrogen to GaAs is contrary to the behavior of most other

semiconductors. An increasing lattice parameter typically leads to a smaller band

gap, and vice versa. GaN has a band gap of 3.2 eV and a cubic zincblende lattice

parameter of 4.51 A (GaN is normally found in the wurtzite crystal structure). If

a Vegard’s-like law is assumed for an alloy of GaN and GaAs, one would expect

a larger band gap with increasing nitrogen concentration. Figure 1.7 shows the

relationship of the GaNAs alloy for small concentrations of nitrogen. Although the

entire alloy is not shown, the GaNAs alloy has a very large bowing parameter in

the band gap versus lattice constant relationship due to nitrogen’s unique properties

in the arsenide semiconductor system. Nitrogen’s band gap reducing behavior has

important implications for developing new materials systems and devices for long-

wavelength optoelectronics. By adding specific ratios of nitrogen and indium to

GaAs, forming GaInNAs, the lattice parameter can be kept very close to that of

GaAs while simultaneously reducing the band gap to 1.3 and 1.55 µm (and perhaps

beyond).

1.4.1 Anomalous band gap reduction with nitrogen

The reason behind the large band gap bowing between the binary alloys GaN and

GaAs leading to the anomalous behavior of simultaneous band gap and lattice para-

meter reduction with small amounts of nitrogen is not completely understood. There

have been several theories, models, and experiments which have attempted to explain

and examine nitrogen’s unique behavior in GaAs and other non-nitride semiconduc-

tors. Many of the theories emphasize the large differences between nitrogen and

arsenic, such as size and electronegativity. Of those theories, the band-anticrossing

(BAC) model [24] is the most widely accepted explanation of the anomalous band

gap reduction.


0

0.5

1

1.5

2

2.5

0

0.5

1

1.5

2

2.5

5.4 5.5 5.6 5.7 5.8 5.9 6.0 6.1 6.2

Ba

nd

ga

p (

eV

)

GaNyAs1-yIn1-xGaxAs

InP

GaxIn1-xNyAs1-y

GaAs

AlAs

InAs

Lattice Parameter (Å)

1.3 µm

1.55 µm

Figure 1.7: Band gap versus lattice parameter showing the effects of adding smallamounts of nitrogen to GaAs and InGaAs.

In small quantities, nitrogen in GaAs forms a localized level due to its large

electronegativity. This localized state in real space is located throughout a wide

range in momentum space and is found roughly 200 meV above the bottom of the

GaAs conduction band at k=0, as shown in Figure 1.8. Although the nitrogen level

and GaAs conduction band appear to cross, they do not due to the Pauli Exclusion

Principle. Instead, the two levels repel each other, or anticross, and form two new

hybridized levels termed E+ and E−. The lower E− level effectively becomes the new

conduction band. The band gap is further reduced with the addition of more nitrogen

due to the increased repulsion between the E+ and E− levels, as shown in Figure

1.9. The BAC model primarily focuses on behavior in the conduction band and does

not predict any significant change in the valence band. A lack of significant shift in

the valence band energy levels supports the theory that nitrogen mostly affects the

conduction band.

1.4. DILUTE NITRIDES 17

Figure 1.8: Illustration in k -space of the band anticrossing effects on the nitrogenlevel and GaAs conduction band.

Figure 1.9: The effects of different nitrogen concentrations in the dilute regime onthe E+ and E− levels in GaNAs [25].


1.4.2 Advantages of dilute nitrides

GaInNAs on GaAs has several advantages over InGaAsP/InP technology, making

it a very attractive candidate to enable low-cost fiber connections to the home.

GaInNAs/GaAs has a much larger conduction band offset, ∆Ec=0.7–0.8∆Eg [26],

compared to InGaAsP/InP. In addition, due to the decreased curvature of the new

conduction band in dilute nitrides, the electron effective mass increases from 0.06mo

in GaAs to 0.11–0.12mo [26–29]. With a deeper QW and increased electron mass,

dilute nitrides provide better confinement of the carriers and a better match of the

valence band and conduction band densities of states. These properties lead to

higher To, operating temperature, efficiency, and output power of GaInNAs lasers.

GaInNAs also has better compositional control compared to InGaAsP. InGaAsP

composition and quality is extremely sensitive to growth temperature and exact

As/P flux ratios. As will be discussed in Chapter 2, although GaInNAs is also

a mixed group-V alloy, there is independent control of the nitrogen and arsenic

compositions. This translates to better yield and eases requirements in production

scale-up, reducing production costs. Finally, growth on GaAs leads to its own set

of advantages. Growth on GaAs utilizes cheaper and larger substrates than InP,

employs AlAs/GaAs DBRs, and allows for monolithic integration with other existing

high-speed devices, leading to high-performance low-cost devices.

1.5 Outline of Thesis

This thesis is not meant to be a complete documentation of the research performed

during the duration of the author’s time on the GaInNAs(Sb) project nor is it a

full description of the field of dilute nitride research. Many details will only be

briefly presented, leaving the reader to examine other works for advanced details.

For additional information on different aspects of this research project, the author

highly recommends the doctoral theses of S. G. Spruytte [30], C. W. Coldren [31], V.

F. Gambin [32], W. Ha [33], V. Lordi [34], M. A. Wistey [35], and S. R. Bank [36].

1.5. OUTLINE OF THESIS 19

This chapter, Chapter 1, has presented the motivation behind developing high-

quality dilute nitride materials for long-wavelength semiconductor optoelectronics.

Current technology based on InP has sub-par performance characteristics and is

much too expensive to enable fiber connections to the home. GaAs-based lasers have

several cost and performance advantages over those grown on InP, however there was

a lack of material system which could emit light reliably in the 1.3-1.55 µm range.

The discovery and development of GaInNAs on GaAs has shown great promise in

creating GaAs-based long-wavelength devices for fiber communications.

The next chapter, Chapter 2, will discuss the growth and general characteriza-

tion of the dilute nitride materials. Details of the molecular beam epitaxial (MBE)

growth of this alloy are presented. Although this alloy is promising, there are in-

variably many challenges and issues which must be analyzed and overcome to obtain

high-quality material. One of the unique features of MBE dilute nitride growth

is the usage of a plasma cell to create reactive nitrogen for incorporation. Since

this is a non-traditional III-V semiconductor, many growth parameters required re-

examination. Antimony is added as a surfactant and an incorporated species to

improve dilute nitride growth quality. Post-growth annealing is also necessary to

further improve optical quality. The development and utilization of these dilute

nitride antimonides will be the focus of this thesis. The characterization methods

used to analyze and investigate optical, electrical, and structural properties of dilute

nitrides are presented. Since the ultimate goal is to implement these dilute nitride

materials in lasers, measuring the optical quality is a good indicator of its potential

performance as a laser active region. However, in combination with electrical and

structural characterizations, scientific analysis and feedback is used to improve the

growth techniques.

Chapter 3 discusses aspects of the rf plasma cell optimization. An unfortunate

consequence of plasma generation of reactive nitrogen is the creation of damaging

ion species. Ion energies and counts can be minimized by optimizing the plasma

operating conditions. Adjusting the gas flow into the plasma cell is one method of

altering these conditions and their effects were applied to a series of dilute nitride

growths. In addition, ion counts and energies were measured using a novel technique.


Higher ion counts and energies from non-optimal plasma operating conditions lead

to degraded optically quality material.

Chapter 4 presents a systematic investigation of the materials which surround the

GaInNAsSb QWs, the QW barrier materials. Analysis of these barrier materials is

essential to improving the active region device structure. The structural and optical

properties, heterojunction band offsets, and implementation of the barrier materials

are considered. GaAs, GaNAs, and GaNAsSb materials are examined as potential

barriers for the GaInNAsSb QWs. GaNAsSb growth investigations are presented

since growth of this alloy has not previously received attention. Heterojunction band

offset measurements of these dilute nitride antimonide alloys are made in order to

examine the band lineups of various materials as this is important in device design.

Finally, some practical growth considerations are examined since there are some

limitations in using MBE. GaNAs barriers are preferred for GaInNAsSb laser diodes.

Chapter 5 deals mostly with the GaInNAsSb QW material itself. Antimony us-

age has not been fully studied and more detailed studies on the usage of antimony

are presented. The role and effects of antimony on widely varying compositions

on GaInNAsSb QWs are presented. As expected, growth interactions between an-

timony, indium, and nitrogen are observed with varying compositions. In addition

antimony has drastically different effects on optical quality of GaInNAsSb at low and

high indium concentrations. Antimony improves material quality in highly-strained

materials, consistent with the properties of a reactive surfactant. Minimization of an-

timony incorporation in the presence of an antimony flux is key to the improvement

of GaInNAsSb QWs. Annealing properties are also presented; the results necessitate

a greater understanding of the optimal annealing temperature.

Chapter 6 presents results from EELs and VCSELs utilizing GaInNAsSb QWs

with GaNAs barriers in the active region. Brief performance characteristics of these

lasers will be discussed.

Chapter 7 includes brief descriptions of additional dilute nitride research applied

towards different applications. The In(N,As,Sb) alloy system is analyzed for its po-

tential usage in biosensor applications which require detection of light at 3-15 µm.

Luminescence from InNAsSb QWs is presented for the first time. GaInNAs(Sb)/GaAs

1.5. OUTLINE OF THESIS 21

has also garnered interest for its promise as the 1.0 eV junction in solar cell devices.

Growth details as well as some solar cell device results utilizing GaInNAsSb are

shown.

Chapter 8 concludes this thesis and also presents some directions for future re-

search.


Chapter 2

Molecular Beam Epitaxial

Growth and Characterization of

Dilute Nitrides

Many techniques exist for the growth of semiconductors. In the early history of

compound semiconductor development, liquid phase epitaxy (LPE) was the preferred

method of obtaining high quality GaAs. However, with the advent and development

of the more advanced growth techniques of molecular beam epitaxy (MBE) and

organometallic vapor phase epitaxy (OMVPE), researchers are now fitted with the

technology to fabricate materials more advanced than most could imagine. MBE

tends to be utilized more commonly in the research environment while OMVPE is the

technique used for mass production in industry. Although both MBE and OMVPE

have been employed in the development of dilute nitrides, this thesis focuses on MBE

growth. Thus far, MBE-grown dilute nitrides have shown far superior performance

for 1.55 µm emitting devices. This chapter will present the basics of MBE growth,

the growth of dilute nitrides, and the characterization tools used to analyze theses

materials.

23

24 CHAPTER 2. MBE GROWTH AND CHARACTERIZATION

2.1 Molecular Beam Epitaxy

Alfred Cho developed molecular beam epitaxy (MBE) at Bell Laboratories in the late

1960s and early 1970s as a method to grow high-purity epitaxial GaAs and study

its kinetic growth processes [37, 38]. Since its inception, this versatile technique has

diversified and is now utilized to grow a wide range of materials including magnetic

alloys [39], oxides [40], Si/Ge [41], III-V, and II-VI [42] compound semiconductors.

MBE is preferred by many researchers over other epitaxial growth techniques be-

cause it allows the opportunity to obtain monolayer control of thicknesses, atomi-

cally abrupt interfaces, precise compositions, and, overall, high quality single crystal

thin films. When studying and creating new materials, it is very useful to have such

control over the growth processes.

MBE, also referred to as “mega-buck evaporator,” is at its heart a very sophis-

ticated and expensive evaporation system. In an ultra high vacuum (UHV) envi-

ronment, a flux of atoms or molecules is directed towards a single crystal substrate.

These atomic and molecular species adsorb onto the surface and either bond into

the growing crystal or desorb if an inadequate bonding site is not available. The

fluxes typically originate from very high purity (≥99.9999% pure) solid sources, ei-

ther through evaporation or sublimation. Since the pressure inside the MBE chamber

is ∼10−10 Torr, the mean free path for the atoms and molecules is on the order of

tens of meters. Even during growth when the pressure can reach ∼10−5 Torr, the

mean free path is still on the order of a meter. The distance between source and sub-

strate is 30–50 cm, well below the mean free path even considering growth pressures.

The source fluxes can be referred to as a “beam” since the atoms and molecules do

not encounter any collisions with other species when traveling from the source to

substrate. Growth rates are slow (<1 µm/h) compared to other epitaxial growth

techniques and are accurately maintained by highly sensitive temperature control.

Mechanical shutters block the molecular beam when the source is not needed during

the layer growth. This method is sufficient to obtain atomically abrupt interfaces

since the shuttering time is shorter than the monolayer formation time due to the

slow growth rate.

2.1. MOLECULAR BEAM EPITAXY 25

Effusion cells

Shutters

RHEED gun

Substrate holder & heater

CAR assembly

Cryo shroud

Ion gauge

Figure 2.1: A top-view schematic of a Mod Gen II MBE chamber. Important com-ponents are illustrated.

2.1.1 Molecular beam epitaxy system

Since the commercialization of MBE chambers and equipment, several companies

have created a wide selection of MBE models. The most popular style of MBE

chamber is the model originally created by Varian and currently sold by Veeco: the

Mod Gen II system. A schematic of a Mod Gen II MBE chamber may be seen in

Figure 2.1.

Two coupled Mod Gen II MBE chambers were utilized for dilute nitride anti-

monide research and device growth. One MBE chamber is specifically designed for

dilute nitride growth while the other is used primarily for DBRs and laser cladding

layers. These two chambers are connected by a transfer tube which is also under

UHV. Keeping the transfer tube under UHV prevents contamination of the MBE

chamber itself as well as the wafer when it is transferred from one chamber to an-

other. Wafers enter through a loading chamber which is separated from the transfer

tube by a gate valve and are baked to remove water and other volatile contaminants.

After this initial bake, the wafers are transferred through the transfer tube and are


System5System5©©

AsAs

BeBe SiSi

SbSb

NN

AlAl GaGa

InIn

System4System4©©

AsAs

BeBe SiSi

CC

GaGa

AlAl GaGa

AlAl

Figure 2.2: Side-view configuration of the sources found in MBE systems for dilutenitride antimonide devices. Blue denotes group-III, green denotes group-V, and reddenotes dopant sources.

baked individually in another chamber up to arsenic desorption temperatures to re-

move additional contaminants (such as hydrocarbons) before their introduction into

the MBE chamber. Contaminant minimization is of the utmost importance.

A Mod Gen II MBE chamber can contain up to eight different source materials.

The limited number and configuration of these source ports places restrictions on

the sources which can be placed simultaneously inside a single MBE chamber. As

seen in Figure 2.1, the source flange is perpendicular to the ground. Four sources

are upward facing while four sources are downward facing. Group-III materials

which require evaporation from a liquid melt are constrained to upward facing ports.

Gas sources and sublimated materials can be used in either location. A single MBE

chamber is insufficient to grow both dilute nitride antimonides and the DBRs required

for long-wavelength VCSELs due to the number of upward source ports and total

source ports required. This necessitated the utilization of the dual chamber system

which allowed greater flexibility in growing a wider range of materials. The source

configuration of the two MBE chambers utilized in dilute nitride antimonide device

growth, System4 c© and System5 c©, can be seen in Figure 2.2.


2.1.2 Group-III sources: Al, Ga, In

Aluminum, gallium, and indium are supplied to the substrate through evaporation

from a liquid melt. The three group-III metals have significant vapor pressures at

elevated temperatures past the melting point. To heat and evaporate these mate-

rials, an effusion (or Knudsen) cell is used. Resistive heater coils wrap around a

pyrolytic boron nitride (PBN) crucible to provide the thermal energy necessary to

produce steady beam fluxes. Original crucible and cell designs consisted of a conical

trumpet crucible which held 60 cc of source material (in a true upward configura-

tion). However, there were several disadvantages to this configuration. The conical

shape limited the source capacity, increasing the frequency of system openings due to

source recharging and greatly reducing the up-time of the chamber. Source capacity

becomes an even greater issue when the cell is placed in the shallow upward facing

source ports (for example the indium location in Figure 2.2) due to the tilt angle of

the crucible. Waste is also a problem due to the wide mouthed opening of the trum-

pet crucible as much of the group-III material becomes deposited onto the shutters

or chamber walls. Gallium and indium do not wet PBN and can recondense at the

lip of the crucible, where it is usually cooler than the source melt, forming small

metal beads. These metal beads can then drop back into the hot source melt and

spray droplets of metal onto the substrate, creating oval-defects which damage the

growth surface [43, 44]. Although aluminum does not have this problem since it wets

PBN, it can creep up the walls of the crucible at prolonged elevated temperatures

and can drip out of the crucible and down to the heater coils. This destroys the cell

as the aluminum short-circuits the heater coils and repair is required.

To solve these problems, a dual filament SUMO cell (sold by Veeco) was used for

the group-III sources in both MBE systems. In dual zone heating [45], one heater

coil is placed at the tip of the crucible while another is placed at the base. The

geometry of the PBN crucible is also much different; rather than a conical trumpet,

the crucible is now a straight walled cylinder with a tapered opening. SUMO cell

crucibles can hold up to 400 g (∼400 cc in a true upward configuration) and can also

hold a significant amount in the shallow upward port due to the smaller exit orifice.

The dual filament configuration and smaller tapered exit orifice also enhances flux


stability and reduces oval defect densities. In dual zone heating, the tip is typically

set to a higher temperature (+150–200◦C) than the base. Heat is provided to the

base such that the beam flux exiting the crucible is lower than what is desired without

tip heating. The remaining thermal energy is provided by the tip heater coil and the

intended flux is obtained. The logic behind this arrangement is to provide constant

heat to the bulk of the melt while the tip provides the fast adjustments to the surface

temperature of the melt to account for any transients coming from shutter movement.

The proportional-integral-derivative (PID) controllers are calibrated differently for

each heater coil. A very large integral or “I” value is given to the base so that there

are no rapid adjustments to the bulk temperature while “D” is set to a high value

for the tip allowing it to rapidly adjust to any temperature variance. Since the tip

is quite hot, it prevents recondensation of gallium or indium at the lip and thus

prevents the the problem of oval defects present with the conical trumpet crucibles.

One drawback with effusion cells is the inability to accurately change growth

rates or fluxes during growth. Each change in growth rate requires a recalibration

of the cell since the flux will not always be the same for identical temperatures

due to various reasons. In addition, the stability of the fluxes immediately after a

temperature change is not sufficient for layers which require a high level of uniformity.

Thus compositional grading and layers with multiple compositions is quite difficult

in MBE. Multiple cells with the same source material, but different growth rates

provides some added flexibility, but it must be considered with the limited number

of source ports available. Digital grading is also an option, but puts great stress on

the mechanical shutters. Valved effusion cells have been considered for MBE [46],

however their feasibility remains to be proven.

2.1.3 Group-V sources: As, Sb

Traditional III-V growth (arsenides, phosphides, and antimonides) is generally per-

formed in a group-V rich environment. The group-V species will desorb from the

growth surface unless there is a group-III atom with an available bonding location.

The group-III atoms have unity sticking at normal growth temperatures, so the


group-V is the limiting species in the growth process. An appropriate overpressure

of the group-V flux is required to ensure a group-V atom bonds with a group-III

atom before another group-III atom occupies the location.

During GaAs growth, an overpressure of 15× at 0.5 µm/h corresponds to ∼10−5

Torr beam equivalent pressure (BEP). An arsenic flux is also required for substrate

temperatures ≥350◦C due to arsenic thermal desorption from the surface. With

required beam pressures ≥15× those of typical group-III fluxes, a source with a very

large capacity for a group-V source charge is required. In both MBE systems utilized

in this dissertation, a 500 cc valved arsenic cracker source provided arsenic during

growth.

A large $10,000 cylinder of pure arsenic is placed inside the sublimator zone of

the arsenic cell. As with other group-V sources, arsenic sublimates from the solid

phase and does not need to be melted like the group-III metals to obtain a significant

vapor pressure. A needle valve is placed at the exit of the sublimator for practical

reasons. Since the sublimator zone is quite large, the thermal mass prevents a rapid

change in arsenic fluxes required for most growths. By maintaining the sublimator at

a steady temperature, the arsenic flux is adjusted by controlling the valved opening.

In addition, when an arsenic flux is not required, the valve can be closed, preventing

a waste of arsenic if the sublimator had to be cooled to idle temperatures.

Arsenic sublimes in the form of As4. Growth studies on arsenides have shown

that the usage of As2 rather than As4 leads to better material quality and higher

PL luminescence due to a difference in growth kinetics [47, 48]. As4 requires higher

substrate temperatures as well in order to encourage arsenic incorporation. The As2

molecule is more reactive than As4 and requires a lower substrate temperature. To

obtain As2 from As4, a cracking zone is placed immediately after the valve from

the sublimator. This cracking zone consists of tantalum baffles which are heated to

high temperatures (850–1000◦C) and thermally “crack” the As4 molecule into 2 As2

molecules.

Similar to arsenic, antimony is provided by a 175 cc antimony cracker source

for dilute nitride antimonide growth. However, this source is unvalved and adds

several challenges during growth. To adjust antimony fluxes, the temperature of the


Cracking Zone Temperature (C)

Mo

no

mer

ic S

b F

ract

ion

Figure 2.3: Monomeric antimony fraction as a function of cracker temperature andantimony sublimator flux [49].

sublimator has to be heated and adjusted before growth. Antimony sublimes in the

form of Sb4 and can be thermally cracked to Sb2 or Sb1 depending upon the beam

pressure and cracker temperature as seen in Figure 2.3 [49]. Under typical operating

beam fluxes of 10−7 Torr, the cracker temperature was set to 850◦C, producing mostly

monomeric antimony.

The design and components of an antimony cracker are slightly different than

that of arsenic and phosphorus crackers due to the corrosive nature of antimony

on typical tantalum parts. PBN, which is not corroded by antimony, replaces all

tantalum parts in areas where the temperature is elevated. Unfortunately this leads

to a dramatic increase in price, a major reason why a valved cracker was not available

for the research in this dissertation. Without a valve, it was difficult to adjust the

antimony flux mid-growth for different applications, to be discussed in this thesis.

Antimony waste is a concern since there is no valve to prevent escape from the

sublimator region when not in use, as in the arsenic cracker. In addition, antimony

has been known to severely degrade a UHV chamber’s ability to pump down after

exposure to atmosphere due to an oxide which is neither volatile nor stable at room

temperature. Luckily in both cases, relatively small fluxes (≤10−7 Torr BEP) were


required during dilute nitride antimonide growth since pure antimonides were not

grown. This small flux in the presence of a more dominant arsenic flux, which

covered any antimony which did stick to the chamber walls, prevented source charge

exhaustion and vacuum poisoning. Obtaining a valved cracker is a priority for future

dilute nitride antimonide research.

2.1.4 Dopant sources: Be, C, Si

n-type and p-type material is obtained by providing a flux of dopants during growth.

In MBE, silicon is the typical n-type dopant and beryllium is the p-type dopant.

Since silicon is amphoteric in GaAs, the growth method can affect the site on which

the silicon atom resides. In liquid phase epitaxy (LPE), under certain growth con-

ditions, the silicon atom tends to incorporate into the group-V lattice, making it

an acceptor. However in MBE, growth is performed in group-V rich conditions and

the silicon atom incorporates in the group-III lattice, becoming a donor. A special

dopant source provides silicon and beryllium via sublimation. These small 5 cc effu-

sion cells have silicon or beryllium deposited on the walls of the crucible and provide

the necessary small flux for doping. GaAs can be doped by silicon in a wide range

of concentrations up to 1019 cm3 before crystal quality is degraded. Beryllium can

be located interstitially and has very high diffusivity in GaAs. Due to this difficulty,

at high doping levels (>519 cm3), the doping profile can no longer be abrupt.

Carbon is another p-type dopant used in GaAs growth. It has much lower dif-

fusivity, but is also much harder to incorporate into GaAs. CBr4 sources have been

utilized to enhance incorporation of carbon to high enough doping levels. The CBr4

gas is injected into the chamber and is cracked at the surface of the hot substrate.

The bromine is pumped away, leaving carbon to deposit. Carbon is preferred over

beryllium at high concentrations and is used to dope the laser p-type regions.

2.1.5 MBE tools and components

There are many other tools and components besides the chamber and sources which

make MBE the complex technique we know and love. While the sources provide the


necessary materials to grow a III-V semiconductor, a method is needed to stop the

molecular beam from reaching the substrate when not needed. Mechanical shutters

physically block the line-of-sight between the substrate and the source flux, prevent-

ing any deposition. The atoms or molecules either deposit onto the shutter itself or

bounce off, adsorbing onto the chamber wall.

The shutters consist of a flap of tantalum or PBN ∼0.75 mm thick which is

connected to an appropriately shaped metal rod that extends outside the chamber

allowing for external control. Pictures of shutters can be seen in Figure 2.4. Tantalum

shutters were originally used for all sources, but had several disadvantages. Tantalum

is not transparent to the heat generated from the cell and warms significantly when

the source is hot. This provides some thermal shielding and lowers the current

required to keep the cell at a certain temperature. Upon the opening of the shutter,

the surface is exposed to a colder environment and the surface of the source charge

cools, reducing the flux. PBN is transparent to thermal radiation; the temperature

transients from shuttering operations are reduced, enhancing flux stability. Gallium

and indium “wet” tantalum shutters and tend to stick to them. Eventually, a large

buildup of gallium and indium on the shutter will force it to get stuck against the wall

when closed, causing a failure of the shutter system. With PBN shutters, gallium

and indium generally drip off and do not build up. However, it still remains a minor

problem when arsenic is present. An arsenic coated PBN shutter enhances surface

cohesion of gallium and indium and a buildup occurs. Conversely, a tantalum shutter

is preferred for aluminum since aluminum wets PBN, but not tantalum.

Besides source shutters, another important component of an MBE chamber is

the vacuum pumping system. Maintaining UHV before, during, and after growth is

one dominant factor in the ability to grow high quality materials. UHV technology,

engineering, and maintenance are a large portion of the MBE growers’ responsibil-

ities since it has such an effect on the end product. Even at partial pressures of

10−12–10−10 Torr, contaminants such as oxygen and carbon can significantly degrade

the quality of the thin films. For the MBE systems used in this thesis, a diverse

and specific set of vacuum equipment is employed to maintain an excellent UHV

environment.


(a) Shutters and source flange (b) Shutter coated with 3 mm of Al

Figure 2.4: Pictures of internal parts of the MBE system. (a) An arsenic coatedsource flange with the eight shutters. White shutters are made of PBN. (b) Sideview of a shutter coated with 3 mm of Al due to build-up over time.

All pumps on an MBE system are “dry,” meaning there is no oil present in

any stage or component of the pump. Oil can backstream through the pump and

contaminate the system. A venturi system, utilizing Bernoulli’s Principle, is used to

pump a chamber down from atmosphere to 50–100 Torr. Once those pressures are

reached, sorption pumps cooled to 77 K with liquid nitrogen drop the pressure to

10−5–10−4 Torr. At this point, with a majority of the gas removed from the chamber,

a cryopump can begin to remove the remaining gas. With fins cooled to 77 K and

4 K, this removes all gases effectively except hydrogen and helium. Ion pumps are

also used in conjunction with cryopumps to complement the pumping by removing

hydrogen and helium. With all sources cold, this pumping system can take a system

pressure down to mid 10−10 Torr after a 200◦C chamber bake to remove water and

some oxides. Arguably, the most important pump in the chamber is the cryoshroud.

The cryoshroud, as shown in Figure 2.1, surrounds the substrate holder and is a vessel

inside the MBE chamber which fills with liquid nitrogen. With a large surface area

and cold temperature, it condenses most volatile gases except hydrogen and helium.

In addition, during growth, all source fluxes which do not hit the substrate will

condense on the cryoshroud, preventing them from bouncing around and condensing


elsewhere. The cryoshroud, when the system is idle, can bring the base pressure

down below 10−11 Torr.

Since MBE is performed under an UHV environment, there is a diverse selection

of in situ monitoring tools which are useful in examining the crystal during growth or

ensuring the growth parameters, like temperature or growth rate, are the intended

values. By analyzing the material and recording growth details, a better under-

standing of the processes during growth can be achieved. Three in situ tools were

primarily used for the growth of dilute nitride antimonides: reflection high-energy

electron diffraction (RHEED), band gap thermometry, and optical pyrometry.

RHEED is a powerful tool which can perform a variety of functions during MBE

growth. A more detailed description of RHEED can be found in the characteriza-

tion section of this chapter, but a brief discussion of its uses will be described here.

RHEED can be used to observe the surface reconstruction patterns during growth,

including the commonly recorded [110] and [110] directions. Different surface re-

constructions can alter the growth kinetics and a large field of surface science is

dedicated to that research. In addition, the quality of the RHEED pattern is used to

draw qualitative conclusions about the growth surface. A streaky pattern indicates

a very flat epitaxial surface while a spotty pattern suggests surface roughening or 3D

growth. A very rough or amorphous surface shows up as a ring or hazy luminescence

on the RHEED screen. The RHEED surface reconstructions are also used to mea-

sure (to a relative scale) the temperature of the substrate. Surface oxide desorption

or surface reconstruction changes occur at specific temperatures. For example, the

RHEED pattern changes from hazy to spotty for GaAs when the last amount of sur-

face oxide is desorbed, corresponding to 580◦C. Growth rate calibrations can also be

performed using RHEED. The intensity of the RHEED pattern (when the substrate

is not rotating) is proportional to the smoothness of the surface. Upon layer-by-layer

growth, the intensity is highest when the surface is perfectly flat and lowest when

the layer is 50% complete due to the roughness caused by random atomic steps. This

oscillatory period can then be converted to a growth rate.

Although RHEED can be used to calibrate the substrate temperature with the


Rela

tive R

eflectivity

Wavelength (µm)

Band edge

transitions

Increasing T

in 50C steps

100C

Figure 2.5: Reflectivity of a semi-insulating GaAs wafer at different wavelengths oflight for different substrate thermocouple temperatures. The sharp transition marksthe band gap at that temperature.

real temperature, it does not provide the accuracy required for the small temper-

ature window of optimal dilute nitride antimonide growth. The substrate surface

temperature can be much different than what is expected due to thermal lag and in-

accurate readings from the substrate thermocouple. Band gap thermometry utilizes

the well known Varshni fits, the change in band gap with temperature, for various

semiconductors. One variation of band gap thermometry used in the dilute nitride

MBE system measures the reflectivity of a semi-insulating substrate from a white

light source. As seen in Figure 2.5 [35], a sharp change in reflectivity occurs when

the wavelength (and energy) of light is equal to that of the band gap of the sub-

strate. By measuring the shift in this transition versus substrate temperature, the

corresponding real temperature can be recorded.

However, this technique has its own difficulties. It requires 1–2 minutes to com-

plete a scan and is not sufficient for real-time measurements. In addition, band

gap thermometry does not work with doped substrates due to a smearing of the


band gap transition caused by band gap renormalization and free carrier absorp-

tion. Conventional pyrometry in itself is not a very accurate method of measuring

the temperature and requires a set of temperature references. By utilizing band

gap thermometry with conventional pyrometry, it is possible to measure the surface

temperature accurately in real time for semi-insulating and doped substrates. It has

been observed that the real surface temperature varies greatly from what is read by

the substrate thermocouple due to a variety of reasons including thermal lag, wafer

holder differences, and substrate size. Use of the pyrometer gives the grower the

opportunity to manually and precisely tune the temperature to the desired value.

2.1.6 Traditional III-V semiconductor growth

The growth rate of the III-V semiconductor under normal conditions is dictated by

the group-III growth rate, due to their unity sticking. An overpressure of group-V

flux is maintained during growth since these atoms will not bond unless there is a

group-III atom with available bonds. Before growth, the group-III sources are stabi-

lized at the desired temperatures for a couple hours. Thermal stability is extremely

important since even a 0.1◦C change can lead to a 1% flux difference. This stabiliza-

tion time also eliminates any oscillations which may arise from the PID controller.

The beam fluxes are measured with the ion gauge, as seen in Figure 2.1, pointing

towards the sources obtaining a BEP. Calibrations relating the BEP with growth

rate are usually performed when the sources are refilled.

As discussed earlier in this chapter, before the wafer is put into the growth cham-

ber, it is baked to thermally remove water and hydrocarbons. The wafer is placed on

the substrate holder and heater. The substrate holder can rotate between “loading”

and “growth” positions and can also have continual azimuthal rotation (lending the

name to the entire setup as the “CAR assembly”) for wafer uniformity. With the

arsenic flux incident when the substrate is above 350◦C, the wafer is heated until

oxide desorption at 580◦C is observed using RHEED. It is then heated to 610◦C for

ten minutes to remove any remaining native oxide and then a 0.3 µm GaAs buffer is

grown to smooth the surface and bury remaining surface contaminants or defects.

2.2. GROWTH OF DILUTE-NITRIDES 37

GaAs is optimally grown at or above the oxide desorption temperatures to prevent

any residual oxygen in the chamber from bonding to the surface. Surface, structural,

and optical quality is optimized in this temperature range. Much higher tempera-

tures, while beneficial in removing contaminants, prevent the gallium from having

unity sticking, unpredictably reducing the growth rate, and inducing surface rough-

ening. Colder temperatures also lead to surface roughening as well as the formation

of arsenic antisite defects (arsenic in a group-III site) due to lower arsenic desorption

rates at lower temperatures. Reducing the arsenic flux at lower growth temperatures

(for example 4× overpressure at 440◦C) leads to improved quality. AlAs is grown

at 600–650◦C since aluminum has lower surface mobility than gallium and requires

more thermal energy to ensure a smooth surface. InGaAs is grown 450–550◦C since

indium desorbs from the surface at much lower temperatures. Unity sticking is ob-

tained for temperatures below 480◦C, but this typically leads to poor optical quality

material. Many grow InGaAs of high optical quality at 500–530◦C, but temperature

calibration and stability is very important as the sticking coefficient of indium in this

range is strongly dependent upon the temperature.

2.2 Growth of Dilute-Nitrides

Dilute nitrides or dilute nitride-arsenides are very different III-V semiconductors,

not only for their electronic properties, but also because of the very large miscibility

gap in the alloys, the different crystal structure for the endpoint alloys (zinc-blende

for InGaAs or GaAsSb versus wurtzite for InGaN), and the methods by which they

must be grown. Most ternary semiconductors, such as AlGaAs and InGaAs, have

complete miscibility across the entire alloy range and can be grown by molecular

beam epitaxy using standard Knudsen effusion cells and thermal sublimator and

cracker cells. Nitrogen in its standard state, N2, is an extremely stable molecule with

a dissociation energy of 9.76 eV [49]. Injecting N2 into the MBE chamber would only

lead to a very small amount of interstitially incorporated N2 and a cryopump full

of nitrogen gas. In order to incorporate nitrogen into the lattice, one must create a

more reactive form of nitrogen: atomic nitrogen. The dissociation energies for arsenic


(3.96 eV) and phosphorous (5.03 eV) are small enough to enable thermal cracking.

Unfortunately, nitrogen has a bond much too strong for such standard UHV cracking

methods. Different sources of nitrogen must be considered and many have been used

in the past including ammonia [50], dimethylhydrazine [50, 51], radio frequency (rf)

plasma, electron-cyclotron resonance plasma (ECR) [52], and DC plasma sources

[53, 54]. The rf plasma source is the most popular choice among dilute-nitride

research groups and is also the tool utilized in this thesis. Because the plasma source

is the largest difference between the dilute-nitrides and other III-V alloys, additional

focus must be given to the operation and utilization of the rf plasma source.

2.2.1 Radio-frequency nitrogen plasma source

The most successful method of obtaining reactive atomic nitrogen has been the use of

a radio frequency (rf) plasma source. Dissociated atomic and molecular radicals and

ions are generated in the plasma and escape from a front aperture plate towards the

substrate, depositing in substantial concentrations. Ions generated in the plasma can

be accelerated by the rf fields and cause substantial substrate damage while radical

neutrals should inflict little damage and incorporate on substitutional sites. There

are many different parameters which control the plasma, including forward rf power,

reflected rf power, the nitrogen gas flow rate, the number, size and configuration of

holes in the front aperture. Unfortunately, there are very few analytical techniques

which are available in MBE to measure the plasma properties and determine the ideal

operating conditions directly. It is also not completely known which characteristics

of the plasma are beneficial to high quality crystal growth and which contribute to

detrimental defects. A rf plasma was chosen over other options because of its low

ion count and high atomic dissociation yield [55].

There are several rf plasma cells which are utilized for dilute nitride growth.

We use a highly modified SVT Associates rf plasma cell model 4.5 because of the

circumstances surrounding available equipment during the early stages of GaInNAs

growth. The SVT Associates cell was chosen at the time since it was the only model

which had a removable PBN front aperture plate in which the number, size, and


orientation of holes could be modified to reduce the nitrogen flux as well as control

the deposition behavior. This was particularly important initially since dilute nitride

growth was very new at the time and most nitrogen sources were used for high growth

rates in GaN. Another advantage of the SVT Associates cell was the fully transparent

back viewport which allowed for direct monitoring of the plasma glow.

The operation of the rf plasma was initially optimized to maximize the generation

of atomic nitrogen within the limits of stable plasma operation. The plasma condi-

tions that maximize the amount of atomic nitrogen versus molecular nitrogen can

be roughly determined using the emission spectrum of the plasma. In the spectrum,

peaks originating from molecular and atomic nitrogen are detected. The amount of

atomic nitrogen is proportional to the ratio of the integrated intensities of the atomic

nitrogen peaks to those of the molecular nitrogen peaks. By varying such primary

parameters as the nitrogen gas flow, plasma power, and number and diameter of

holes in the front aperture plate, the optimal conditions can be obtained by maxi-

mizing the ratio of the atomic to nitrogen peaks found in the emission spectrum [56?

]. With the current Y-shaped four-hole aperture, the rf plasma operating under 300

W forward and 7 W reflected power with 0.5 sccm flow resulted in the best ratio of

the atomic to molecular peaks.

This method, however, does not always give the most optimal plasma for crystal

growth of dilute nitrides. Ions and excited neutrals are difficult to distinguish in the

emission spectrum. In addition, the emission spectrum originates from the back of

the plasma cell while the species exiting the cell towards the substrate come from

the front of the plasma cell. More extensive investigations relating plasma conditions

to growth quality are required to fully determine the optimal conditions for dilute

nitride crystal growth. Detailed discussion on the effects of gas flow on dilute nitride

material quality and nitrogen ion generation is presented in Chapter 3.

Plasma stability and operating conditions during low nitrogen flow growth is one

of the largest challenges in growth of dilute nitrides and is probably the greatest cause

of differences not only between wafers grown in the same system, but particularly in

comparisons between different groups who believe they are growing under nominally

identical conditions. Some of this is the result of short-term plasma instabilities in


the source, others due to long-term changes in the source. A second issue is ion

or electron damage to the epitaxial films from the plasma source, which is quite

dependent upon the current operating region of the source. It is also dependent

upon the use of charged particle deflecting plates at the source exit. These charged

deflection plates can steer ions away from the substrate, decrease ion damage, and

lead to an increase in optical quality [57].

The challenge of igniting and maintaining the plasma is a degradation problem

which gets worse with time. This manifests itself by requiring an increase in the

flow rate necessary to ignite the plasma after 1–3 years of operation. Generally, igni-

tion of the plasma becomes more difficult, operational parameters are no longer the

same, material quality decreases, and in worst cases, the plasma would extinguish

intermittently during growth. First, it appears that there is decomposition of the

PBN crucible during operation, creating boron dust as well as plasma etching of the

holes in the front aperture plate. The only solution to this problem is to replace the

crucible when servicing the MBE system and to minimize the duration of plasma

operation. Also, arsenic contamination is a problem in the crucible. During growth

when the plasma cell is off, the cell is not heated and arsenic can condense in or

on the cell. This can either contaminate the inside of the cell which interferes with

the plasma characteristics or coat the outside of the crucible which acts as an elec-

tromagnetic shield and reduces rf coupling into the plasma. A proposed solution to

this problem is the installation of a gate valve to isolate the cell from the rest of

the chamber when nitrogen is not needed. However, this introduces a new set of

difficulties including stability and etching of the gate valve.

Another stability problem relates to the time required for temperature and power

to stabilize in the cell. This is a particular problem when growing laser structures

as there is nitrogen leakage or “blow by” around the shutter. This results in heavy

nitrogen “doping” when the source is operating, but with the shutter closed. We

typically see ∼1019 cm−3 levels of nitrogen in such films as shown in Figure 2.6. This

is a serious problem when growing AlGaAs as even minute quantities of nitrogen dra-

matically increase the trap density and non-radiative recombination rate. However,

removal of the growth of aluminum containing alloys from the dilute nitride MBE


20 40 60 80 100 1201.0x1018

1.0x1019

1.0x1020

1.0x1021

Nitr

ogen

(ato

ms/

cm3 )

Sputter Depth (nm)

IgnitionShutter open

Shutter closed

Figure 2.6: Nitrogen content in a dilute nitride layer showing incorporation evenwhen the shutter is closed caused by “blow by” when the plasma is running.

system has greatly reduced Al/N difficulties. Our current solution to the “blow by”

issue is to run the cell for a period of time before the wafer is loaded and growth

is started. This allows the cell to thermally stabilize before a growth starts. The

source is then extinguished, but with moderate rf forward power on to maintain the

temperature while other non-nitrogen containing layers are grown. The plasma is

re-ignited shortly before nitrogen containing alloys are grown. However, this is not

always a viable solution and the blow by around the shutter is far worse when the

plasma is ignited. These problems can be greatly reduced by using better shutter

designs, placing the source behind a differentially-pumped gate valve, or utilizing an

arsenic cap to protect the surface of the wafer [58].

2.2.2 GaNAs

Incorporation of nitrogen into GaAs is unlike crystal growth of most other III-V

semiconductors. The kinetics of growth are different from that found for arsenides,

phosphides, or antimonides in that the surface is not terminated or stabilized by

one of the aforementioned group-V species. Nitrogen also does not compete in the


same manner as other group-Vs for the group-V lattice site. For example, there is no

simple relation of composition for various arsenic and phosphorus fluxes for a certain

growth rate when growing arsenide-phosphides. The arsenic and phosphorus atoms

compete for the group-V sites in complex ways, which can also be altered by variables

such as growth rate and substrate temperature. The material system requires many

calibration samples beforehand to know the exact concentration obtained during

growth.

During GaNAs growth, nitrogen is independent of the arsenic flux [56? ] and, to

a certain extent, substrate temperature; it is affected only by the group-III growth

rate. To vary the nitrogen composition, the group-III growth rate was adjusted while

maintaining constant plasma parameters to ensure optimal plasma conditions during

deposition. Other groups control nitrogen incorporation by varying flow rate or rf

power. However, varying power or flow rate can greatly modify plasma characteristics

and change material quality; this is discussed in Chapter 3.

Nitrogen concentration follows an inverse dependence to the group-III growth

rate. This inverse linear dependence is valid for concentrations as high as 10%, at

which nitrogen incorporation is difficult to analyze accurately due to compositional

segregation and relaxation [56, 59]. Figure 2.7 plots concentration for a series of

GaNAs samples grown at different growth rates. The data can be fit to the following

equation:

[N%] =K

{GR} (2.1)

where the K is a constant dependent upon plasma operating parameters such as rf

power, gas flow, and front plate aperture design and {GR} is the group-III growth

rate [56? ]. The inverse linear dependence is due to the fact that all incident ni-

trogen (of the correct species) adsorbs onto the GaAs growth front with unity-like

sticking and is then buried by additional gallium and arsenic adatoms. Thus, the

higher the group-III growth rate, the lower the amount of nitrogen found per volume

of material. The growth temperature window for GaNAs was once restricted to low

temperatures (≤475◦C) due to compositional segregation and nitrogen clustering, as

is the case with other dilute nitrides such as GaInNAs. However with improvements


1.5 2.0 2.5 3.0 3.51.2

1.4

1.6

1.8

2.0

2.2

2.4

2.6

0.6 0.5 0.4 0.3

Nitr

ogen

(%)

1/Group III Growth Rate ( m/h)

Group III Growth Rate ( m/h)

Figure 2.7: Concentration of nitrogen in GaNAs as a function of group-III growthrate. Plasma conditions are 300 W forward power and 0.5 sccm N2 flow, correspond-ing to K=0.8.

in growth and plasma operating techniques, GaNAs quality through a larger range

of temperatures is acceptable. With this increased temperature range, it was dis-

covered that nitrogen incorporation is also independent of substrate temperature up

to temperatures close to normal GaAs growth temperatures of 580◦C. At very high

temperatures, compositional segregation remains and the incorporation kinetics are

drastically altered.

2.2.3 GaInNAs

Adding indium to GaNAs introduces several growth related challenges which are not

present in GaNAs. In the presence of indium, the growth window for various parame-

ters decreases significantly, making growth more difficult due to the tight tolerances

required. Unlike GaNAs, the growth temperature window is restricted to a small


range between 430–445◦C. Substrate temperatures colder than 430◦C lead to de-

graded optical quality due to high concentrations of low-temperature growth defects,

such as arsenic antisites and vacancies. Above 445◦C, GaInNAs quality also suffers

with the introduction of surface roughening and clustering of nitrogen and indium

rich areas. The optimal temperature of 440◦C kinetically inhibits compositional seg-

regation [60] and minimizes defects introduced with low-temperature growth. This

growth difficulty is related to the addition of nitrogen into the system as InGaAs

growth does not suffer from these challenges. InGaAs is fully solid soluble for all

compositions.

The degradation in optical quality of GaInNAs is not fully explained with low-

temperature defects. Nitrogen incorporation into (In)GaAs has generally decreased

the electronic and optical quality of the material, leading to the joke in the community

that “there is no material so bad that adding nitrogen cannot make it worse” [35]. A

variety of explanations have been presented describing why dilute nitride materials

suffer from this “nitrogen penalty.” Nitrogen’s large electronegativity difference with

the other atoms in the alloy, its very small atomic size, and the method of nitrogen

incorporation have all been widely suggested as dominant reasons for the degradation

of (In)GaAs material quality with nitrogen. Since both indium and nitrogen decrease

the band gap in GaAs, GaInNAs alloys were generally grown with the maximum

allowable indium concentration, minimizing the amount of nitrogen needed to reach

long wavelength emission. Typical compositions used to obtain 1.3 µm emission

range from 30–33% for indium and 1.5–2.0% for nitrogen.

The maximum indium concentration is primarily dictated by the lattice strain

of GaInNAs on GaAs. For strained epitaxial thin films, there exists a maximum or

critical thickness in which the layer remains perfectly coherent to the substrate. Past

this thickness, the layer will form misfit dislocations to relieve the strain. Dislocations

are defects which lead to non-radiative recombination of carriers, reducing the optical

and electrical quality of the material. Since the thickness of the GaInNAs QWs are

quite thin (6–8 nm), the lattice mismatch to GaAs found in this material can be quite

large (2–3%). Several groups have been able to incorporate up to 35% indium with

1–2% nitrogen. Most of the time, the QW can be grown thicker than theoretically


allowed; due to other factors, such as low growth temperature and fast growth rates,

the kinetics driving relaxation are inhibited. Often times, the limiting factor in

increasing indium concentration in dilute nitrides is compositional segregation of

indium and nitrogen without relaxation [59–61].

2.2.4 Surfactant Growth

Surfactants have played a crucial role in the development of high-quality epitaxial

thin films [62]. The term “surfactant” originated in chemistry and was used to

describe “a substance that lowered the surface or interfacial tension of the medium

in which it is dissolved” [63] It mostly applied to substances which reduced the

surface tension of liquids, such as water. Early in thin film deposition, “surfactant”

was adopted to mean any element which altered the growth mode of the film by

lowering the surface free energy.

The application of surfactants to semiconductor thin film growth was introduced

by Copel et al. He showed that the usage of a single monolayer of arsenic on silicon

could improve the growth of germanium on silicon [64]. The growth of germanium

on silicon was difficult due to the large lattice mismatch, and thus strain, which

existed between the two elements. Growth of germanium on silicon generally begins

in the Stranski-Krastanov (S-K) mode (layer-by-layer) for a few monolayers. How-

ever, the strain energy in the germanium film becomes large enough such that it is

energetically favorable to form 3D islands. Continual growth with the 3D features

leads to highly dislocated and defected material. By adding arsenic, it changed the

thermodynamics and kinetics of the growth by lowering the surface free energy and

restricting the formation of 3D islands. The arsenic also did not incorporate and

continually segregated to the surface of the growth front. Arsenic acted as a surfac-

tant for germanium on silicon growth and thus enhanced the quality and thickness

of the material grown.

One theory in explaining how surfactants work in semiconductor growth was ex-

plained by Massies and Grandjean in 1993 [65] and further described by Tournie et


al. in 1995 [66]. They theorized that surfactants could be separated into two cat-

egories: reactive surfactants and non-reactive surfactants. Non-reactive surfactants

are those used primarily in homoepitaxy in which strain does not play an important

role in affecting the kinetics of growth. These elements surface segregate and do not

react with any of the actual growth species. Their function is simply to enhance the

surface adatom diffusivity. For GaAs growth, it has been found that column-III and

column-IV elements, such as Ga, Sn, and Pb, are generally considered non-reactive

surfactants. Reactive surfactants are used mostly in heteroepitaxy, in which strain

does prevent S-K growth due to lattice mismatches. These elements surface segre-

gate, but also incorporate in dopant to dilute levels in the growing crystal. They

react with the adsorbed species and reduce the surface diffusivity. This is beneficial

in growth of highly-strained material because very high surface diffusivity can lead

to clustering and islanding. Column-V and column-VI elements, such as As, Sb, and

Te, are considered reactive surfactants for GaAs growth.

On an atomistic scale, one can imagine the case for the non-reactive surfactant

at the edge of a step. If an atom adsorbs on top of a layer, it moves around on the

surface until it finds the lowest energy position. When it arrives at the step edge

(which is next to a lower level), it encounters a large energy barrier to drop down

to the lower level because it must break extra bonds to traverse over the edge, down

to the layer. This prevents atoms from growing in an S-K mode and islands form.

However, if a non-reacting atom sits at the step edge due to Van der Waal forces,

it can assist in S-K growth by eliminating the step energy barrier. An adsorbed

atom approaching the step edge can easily move down to the lower level because the

surfactant atom provides extra bonds for the adsorbed atom and thus extra bonds

do not have to be broken. The surfactant atom, since it is non-reactive with the

crystal, simply shifts over one atomic position and lets the adsorbed atom slip into

place. This enhances the surface mobility. For the reactive case, the surfactant

atom actually bonds into the crystal. However, when an adsorbed atom approaches

surfactant atom, it bonds with the surfactant atom and is either held in place or is

buried in the layer beneath it. It is also possible that the surfactant atom, if buried,

can switch places with an adsorbed atom on top of it. Since the surfactant atom is


reactive, it is also possible it does not swap locations and becomes incorporated in

the material. These actions reduce surface mobility.

Antimony has been used fairly widely as a surfactant in many different semicon-

ductor alloys [62, 67–79]. It is one of many elements which have been used, but

primarily in III-V semiconductor growth. Since III-V growth usually involves het-

eroepitaxy, it is desired to have a surfactant which is reactive and does not provide

electrical carrier dopants (such as column-IV elements). Isoelectronic surfactants

(usually column-V) are preferred. Due to electronegativities and bonding potentials,

smaller atoms tend to be overly-reactive and do not provide the surfactant-like effects

which are desired. Antimony and bismuth are the usual candidates for III-V surfac-

tant growth, where antimony is more popular due to economics and the number of

past studies.

2.2.5 GaInNAsSb

Wang et al. in 1999 first proposed the usage of antimony as a surfactant [71] to

overcome the difficulties of obtaining high-quality GaInNAs/GaAs due to the large

miscibility gap [80–82]. Utilizing an effusion cell, a flux of Sb4 was provided during

GaIn(N)As growth. The RHEED improved from a spotty to a streaky pattern with

an antimony flux and the photoluminescence (PL) intensity at 1.2 µm improved by

a factor of five. Antimony incorporation was not reported, leading to the conclusion

that it did act as a surfactant in GaInNAs growth.

Shimizu et al. in 2000 reported 1.6% antimony incorporation in GaInNAs, form-

ing GaInNAsSb [72], when utilizing it as a surfactant. This was surprising since

surfactants do not typically incorporate past dopant concentrations. With anti-

mony, they were able to improve their GaInNAs lasers which emitted at 1.26 µm.

In a later paper by Wang et al. in 2001, they did report 1% antimony concentration

in GaInNAsSb [83], overruling their previous claim that antimony did not incorpo-

rate. They were able to show significant structural improvement via transmission

electron microscopy (TEM). An improvement in PL intensity at 1.5 µm was also

shown, although it was still extremely poor, even with antimony.


More recently, work in this group by Gambin [32] and Ha [33] have shown the

promise that GaInNAsSb has a high quality material which emits light out to 1.6 µm

on GaAs. Figure 2.8 illustrates the effects of adding antimony to GaInNAs grown by

our group. With GaInNAs, we were only able to obtain very high PL intensity for

material out to 1.3 µm. Pushing to longer wavelengths by adding more indium or

nitrogen led to compositional segregation and a significant degradation in material

quality [30, 32, 60]. However, by adding a flux of 1×10−7 Torr BEP monomeric

antimony to the best GaInNAs material at 1.3 µm, we were able to increase the PL

intensity and obtain slightly longer wavelengths. With antimony, more indium and

nitrogen was added without suffering from compositional segregation or relaxation,

even though the strain was much higher than before [32, 84]. With antimony, the

maximum indium concentration which resulted in high-quality coherent material

increased from ∼33% to greater than 42%. PL out to 1.6 µm was reached with

only a 5–10× reduction in intensity at 1.5 µm. GaInNAsSb EELs with the lowest

threshold current densities (2.8 kA/cm2) at the time for 1.49 µm on GaAs were

obtained. After a proof of principle, a great deal of research remained on studying

the effects of antimony and obtaining a greater understanding of how to utilize it

properly.

2.2.6 GaNAsSb

The addition of antimony to GaInNAs led to a dramatic improvement in material

quality. It was also suspected that adding antimony to GaNAs would have the

same effect. However, very little research has been performed on GaNAsSb with the

exception of the work by Harmand et al. at Le Centre National de la Recherche

Scientifique [85, 86]. Examination of GaNAsSb can hopefully provide insights on the

more complex quinary alloy GaInNAsSb.

It was thought there was unity sticking of nitrogen when growing Ga(In)NAs

since the relationship between the inverse of the group-III growth rate and nitrogen

concentration was linear. However, this was found not to be the case when anti-

mony was introduced. The addition of antimony to GaNAs increased the nitrogen


0.75 0.80 0.85 0.90 0.95 1.00

100

101

102

103

1600 1500 1400 1300

More In, Sb, N

GaInNAsSb

GaInNAs

GaInNAs

PL In

tens

ity (a

.u.)

Energy(eV)

Wavelength (nm)

Figure 2.8: PL of GaInNAs(Sb) samples comparing the best 1.3 µm material grownwithout antimony and the dramatic improvement in PL at longer wavelengths byadding antimony.

concentration under identical growth conditions [84, 85]. This was surprising in light

of the theory that nitrogen had unity sticking when growing Ga(In)NAs. Antimony

somehow enhanced the nitrogen incorporation into the material, leading to upwards

of 25–50% increase in composition. The mechanism for the increased nitrogen con-

centration in the material is not known. However, it is thought that the properties

of antimony as a “reactive surfactant” [65, 66] help promote the incorporation of

nitrogen into GaAs. By reducing surface mobility and having strong interactions

with the adsorbing species, it is possible that the antimony prevents any nitrogen

desorption or allows for the incorporation of another nitrogen species such as N∗2 as

opposed to atomic nitrogen. A more detailed discussion and analysis of GaNAsSb

can be found in Chapter 4.


2.3 Characterization Methods

The dilute nitrides pose a number of very interesting and challenging characteri-

zation problems. First, the number of elements in the quaternary (GaInNAs) or

quinary (GaInNAsSb) compounds produces a very wide range of materials com-

positions which can have the same bandgap or lattice constant. Next, the large

differences in atomic mass and radii and miscibility gaps lead to compositional seg-

regation and formation of micro-phases [81, 82]. Finally, there is uncertainty of the

exact mechanisms for the large bandgap changes with nitrogen composition.

A number of characterization techniques commonly used for the dilute nitrides

will be briefly described, starting with reflection high energy electron diffraction

(RHEED) since it is used in situ during MBE growth. After growth, high-resolution

x-ray diffraction (HRXRD) and room-temperature photoluminescence (PL) measure-

ments are made on every sample. More difficult, but insightful, measurements not

performed on every sample, include secondary ion mass spectroscopy (SIMS), elec-

troreflectance (ER), and photoreflectance (PR). These are necessary to understand

both the properties and quality of the material and provide feedback for growth as

well as the fundamental properties that are essential to optimize the design of lasers.

2.3.1 Reflection high-energy electron diffraction

RHEED is an important in-situ technique for MBE growth. Arthur and LePore in

1969 incorporated RHEED with MBE to analyze the surface reconstruction of GaAs

during growth. The uses of RHEED were described earlier during the discussion on

MBE tools, but a brief description of the technique will be presented. RHEED is

highly surface sensitive due to the low penetration depth (a few monolayers) of the

electron beam. In addition, it must be performed in an UHV environment so that

the electrons are not scattered. A beam of electrons roughly 5–10 keV in energy is

incident upon a sample at a grazing angle of 1–3◦. After diffracting off the sample

surface, the electrons strike a phosphor screen and cause a visible fluorescent pattern

corresponding to the structure of the surface.

2.3. CHARACTERIZATION METHODS 51

(a) Streaky (b) Spotty (c) Quantum dots

Figure 2.9: Examples of RHEED patterns relating to different surface structures.(a) Smooth surface, (b) rough surface, and (c) quantum dots.

The RHEED pattern for a planar growth surface is a series of streaks corre-

sponding to the surface reconstruction of the sample. To minimize surface energy,

the atoms reconfigure themselves to reduce excess energy from dangling bonds. If

no surface reconstruction occurs, only primary streaks will be seen. However, if a

repeating surface pattern with a period of two atoms occurs, primary streaks with

less intense secondary streaks will be seen between the primary streaks. GaAs at

580◦C under an arsenic flux of 15× overpressure has a “2×4” surface reconstruction

pattern and can be seen in RHEED as a 2× pattern in the [110] direction and a 4×pattern in the [110] orientation. If the RHEED pattern consists of elongated dots

that do not form streaks, this is an indication of a rough surface with several mono-

layer peak to valley heights. Quantum dots show up as a patterned set of sharp dots.

For a non-coherent surface, RHEED will show rings for polycrystalline material or a

dim haze for amorphous situations (such as native oxides). An example of RHEED

patterns can be seen in Figure 2.9. For an advanced treament of RHEED, please see

References [87] and [88].

2.3.2 High resolution x-ray diffraction

HRXRD is one of the most useful techniques for analyzing epitaxial films. Since the

peaks in an x-ray diffraction pattern are directly related to the atomic distances,

one may obtain information such as strain, film thickness, relaxation, and phase.


2θ ω

Incident beamDiffracted beam

Sample surface

Diffracting planes

Q

Figure 2.10: Diagram illustrating the geometry of a symmetric ω/2θ scan of (00l)planes. ω is the angle between the incident beam and the surface while 2θ is the anglebetween the diffracted beam and the incident beam. Q is the diffraction vector.

Although there can be complications and complexities, HRXRD can also provide

some compositional details rather quickly and nondestructively. One of the most

common scans performed on HRXRD for epitaxially grown semiconductors is the

ω/2θ scan of the (004) planes. A diagram of a symmetric ω/2θ scan can be found

in Figure 2.10. Here, the diffraction vector is perfectly perpendicular to the surface

and thus only measures out-of-plane components.

At certain angles of diffraction depending on the lattice spacing, constructive

interference of x-rays occurs and is described by Bragg’s Law:

nλ = 2dhkl sin θ (2.2)

where n is the order of the diffraction, λ is the wavelength of the incident x-rays, θ

is the scattering angle and dhkl is the spacing between the (hkl) planes. For cubic

materials, dhkl is related to the lattice spacing a by the following relationship:

d2hkl =

a2

h2 + k2 + l2. (2.3)

When measuring bulk films, Bragg’s Law is sufficient if the lattice constant is at

equilibrium. For example, the (004) ω/2θ scan of a fully relaxed InGaAs film can give


compositional information based on the lattice spacing dhkl and film thickness based

on the width of the diffraction peak. For fully coherent strained films, a more complex

analysis is needed. Strained films grown on GaAs, such as GaNAs, are tetragonally

distorted such that the (004) plane spacing is different from its equilibrium value. In

a cubic crystal, the equilibrium unstrained lattice parameter, aeq, can be calculated

from the measured (004) lattice parameter a004 using the relationship

σzz = 0 = 2C12εxx + C11εzz (2.4)

where σzz is the strain in the out-of-plane direction and C11 and C12 are the stiffness

coefficients for the film. εxx, the in-plane strain1 for the film is given by

εxx = εyy =aGaAs − aeq

aeq

(2.5)

and εz, the perpendicular strain for the film is given by

εzz =a004 − aeq

aeq

. (2.6)

where aGaAs is the lattice parameter for GaAs. Using the previous equations together

and solving for aeq, one obtains

aeq =2C12

C11aGaAs + a004

1 + 2C12

C11

. (2.7)

The ratio 2C12

C11is approximately 0.9 for most III-V materials [89]. For GaNAs, using

the unstrained lattice parameter, the nitrogen concentration can be calculated using

Vegard’s Law, which is valid in the dilute regime [81].

One difficulty with using HRXRD to determine compositions is the ambiguity

arising from four or more elements. In a ternary compound, it is simple to analyze

1In thin film mechanics formalism, compressive strain is a negative value while tensile strain isa positive value. Unfortunately, this is differs from the convention used in the optoelectronics andphysics communities by a negative sign. This thesis follows the convention used in the dilute nitridecommunity: Compressive strains are positive values while tensile strains are negative values.


the change in lattice parameter, and thus strain, due to the addition of one element

in the III-V semiconductor. However, once two elements are added, the source of the

strain becomes more complex. For example in GaInNAs, there are an infinite number

of combinations of indium and nitrogen compositions which will give the same lattice

parameter. One can obtain an estimate of the concentrations from past calibrations

or general experience by setting a range in which indium and nitrogen compositions

can be in existence. The quinary GaInNAsSb is even more difficult to obtain a defi-

nite composition from HRXRD. SIMS, in conjunction with HRXRD simulations, is

the best method of obtaining the composition for quaternary or quinary compounds.

In order to obtain an accurate unstrained lattice parameter value from the (004)

ω/2θ scans, it is assumed there is no relaxation. If any relaxation has occurred, this

will shift and possibly broaden the peak leading to incorrect results. GaInNAs(Sb)

QWs are highly compressive (up to +2.6% strain) and have a theoretical critical

thickness of 35–45 A, much thinner than the typical QW width of 75–80 A. Relax-

ation for these materials can be an important issue.

One method to ensure relaxation has not occurred is to perform a reciprocal

space map (RSM). A RSM is a series of ω/2θ scans holding ω constant. The ω

value is increased incrementally and a plot is obtained showing contours of diffracted

intensity in reciprocal space. An additional axis of diffraction is added, making it a

triple-axis configuration. Figure 2.11 shows the differences between a typical ω/2θ

scan and one with the third axis. Without this third axis, the diffraction beam

probe is a line which spans a finite amount in ω. Since the probe is not a point,

it smears signals across in ω, reducing the resolution of the map. With the third

axis, fine diffraction features are able to be seen. This is very useful in examining

any diffraction peaks which may not be obtained from the standard (004) ω/2θ scan

including peaks which are found with in-plane components. RSMs are plotted with

the out-of-plane (00l) direction as the y-axis and the in-plane (hk0) direction as

the x-axis. The most common diffraction direction for RSMs is the (224) set of

planes because of its sensitivities to in-plane strain [90]. If all layers are coherent

on the substrate, the RSM will only show a “line” of diffraction with the same in-

plane value but various out-of-plane values. This is very similar to the pattern one


Sample

MonochromatorAnalyzer

Detectors

1

2

3

Figure 2.11: Diagram illustrating the three axes in the triple-axis configuration.In a normal ω/2θ scan, the analyzer is not present and the direct diffracted beamis detected. In triple-axis, a detector in a different location measures the beamdiffracted from the analyzer.

would observe if the (004) ω/2θ scan were turned on its side and viewed from the

top. However, any relaxation will generate diffraction intensities at different in-plane

values of the substrate. Generally the relaxation is neither complete nor uniform so a

smear of diffraction intensities will develop originating from the unrelaxed diffraction

point. Figure 2.12 illustrates relaxation line if GaNAs were to relax on GaAs. Figure

2.13 also shows an example of a perfectly coherent GaInNAsSb QW and a relaxed

GaInNAsSb thick film on GaAs. Additional information on HRXRD and RSMs can

be found in References [91] and [92].

2.3.3 Secondary ion mass spectrometry

Another widely used and powerful tool used in III-V semiconductor research is SIMS

depth profiling. A focused ion beam sputters atoms from the surface at a controlled

rate such that the user is able to obtain compositional and depth location informa-

tion. The secondary ions generated from sputtering are collected and a compositional

depth profile is generated. When measuring nitrogen concentrations with SIMS, it

does not form a negative atomic ion and thus GaN− or CsN+ ions can be monitored

when sputtering with a cesium ion beam. The GaN− ion is generally produced in

high yields and gives a good minimum detection limit during depth profiling. How-

ever, since this ion is dependent upon the gallium concentration as well, changes in


GaNAs (224)

coherent

GaAs (224)

GaNAs (224)

unstrained

Relaxation

line

(000)

ω/2θ scan

direction

ω scandirection

Q||

Q⊥

Figure 2.12: Diagram illustrating the direction of relaxation for GaNAs when exam-ining the (224) diffraction peaks.

GaAs

GaInNAsSbQW

Q

Q ||

(a) Coherent GaInNAsSb QW

GaAs

GaInNAsSbQ

Q ||

(b) Relaxed GaInNAsSb layer

Figure 2.13: Example (224) RSMs of (a) a perfectly coherent 80 A GaInNAsSb QWon GaAs and (b) a partially relaxed 1 µm GaInNAsSb layer on GaAs.


the amount of gallium will affect the apparent nitrogen signal measured from GaN−.

The CsN+ ion is immune to these effects, but the yield is much lower and thus the

signal is much noisier. This increases the minimum detection limit and reduces the

resolution of the measurement.

SIMS is also greatly affected by matrix effects. The secondary ion yields are

strongly dependent on the electronic properties, such as ionization energy, of the

matrix. This problem can be eliminated by creating several known “standard” sam-

ples to calibrate the signals. In the case of GaInNAs(Sb) material, the nitrogen signal

was calibrated using nuclear reaction analysis Rutherford backscattering (NRA-RBS)

and the antimony signal using particle induced x-ray emission RBS (PIXE-RBS). Ad-

ditional details on SIMS calibration may be found in the doctoral dissertations of S.

G. Spruytte and V. F. Gambin [30, 32].

2.3.4 Photoluminescence

The primary purpose in developing the dilute nitride antimonides is to create an

active region material which has superior lasing characteristics compared to existing

InP-based technology. Measuring and analyzing the optical quality and properties

of these materials is important towards that end. A common and easy, yet powerful,

method of investigating the optical behavior of semiconductors is PL. PL is non-

destructive and requires almost no sample preparation or complex growth structures,

making it one of the most useful techniques when developing a new direct bandgap

material system such as the dilute nitride antimonides.

In general, luminescence describes the emission of radiation from a material when

supplied with energy. There are several types of luminescence which can be used to

study semiconductors depending on the method of carrier generation. Cathodolu-

minescence (CL) utilizes a beam of electrons, electroluminescence (EL) employs an

electric field for carrier injection, and PL relies on the absorption of photons to create

electron-hole pairs. The recombination of these carriers from two different energy

levels, typically the ground states of the conduction and valence bands, results in a

photon of identical energy as the transition.


Conduction

Band

Valence Band

Pump

Emission Defect

Figure 2.14: Illustration in momentum space of the basic carrier processes in PL.

In PL, a pump laser with photon energy larger than the band gap is incident

upon the sample. Carriers are generated via absorption of the light and typically

have energies greater than the ground state in the conduction and valence bands

due to energy conservation. The excited carriers relax non-radiatively to the lowest

available states in the bands. Ideally, the electrons and holes radiatively recombine

and emit light which is collected into a spectrometer and displays the intensity as

a function of wavelength. Figure 2.14 illustrates this process. The carriers can

also non-radiatively recombine through various processses, releasing energy through

phonon creation and causing the sample to heat.

The sample returns to the equilibrium state after a characteristic time τ . τ is

related to the recombination time by

1

τ=

1

τr

+1

τnr

, (2.8)

where τr is the radiative lifetime and τnr is the non-radiative lifetime of the carriers.

Optical processes are most efficient for direct gap semiconductors since the minimum

and maximum of the conduction and valence band, respectively, are located at the

same value in momentum space at the gamma point and photons have almost zero


momentum. For perfect direct gap material, τr is small and the process can be domi-

nated by optical emission. Indirect gap semiconductors such as Si, Ge, and GaP have

extrema at different locations in momentum space, requiring a phonon to facilitate

the transition. This drastically reduces the radiative efficiency and leads to a large

τr, allowing non-radiative processes to occur. What makes PL very useful in com-

paring relative optical qualities with other samples for direct gap semiconductors is

the sensitivity to defects. Dislocations, vacancies, antisites, contaminants, and many

other defects can trap carriers, causing them to recombine non-radiatively. These

processes reduce τnr significantly allowing non-radiative recombination to become

dominant. Although two different samples may appear to be structurally identical,

point defects, which are harder to detect in exact quantities, affect the overall optical

quality of the material. Laser performance is highly sensitive to the optical quality

of the material and almost perfectly tracks PL characteristics.

The main parameters of interest of PL measurements are the peak wavelength,

peak intensity, and linewidth. Wavelength is important since it gives a good indi-

cation of the lasing wavelength when the material is put into the active region of

a laser. Peak intensity and linewidth allow for a relative comparison of material

quality of different samples. The PL intensity is an excellent indicator of optical

quality of the material. As mentioned earlier, lower defect densities lead to higher

PL intensities. As long as the samples being investigated do not have significant dif-

ferences in sample structure, the PL intensities can be compared. The PL linewidth

gives an indication of material quality, including interface quality and alloy disorder.

Occasionally the linewidth can be broadened due to peaks which are near each other.

Performing low-temperature PL can assist in resolving these two peaks.

PL has been integral to the investigation of annealing dilute nitride alloys. The

dilute nitride samples are typically rapid thermal annealed (RTA) for 60 seconds in

a range of temperatures from 700–900◦C. The samples are proximity capped with

a GaAs wafer to prevent arsenic desorption from the surface and are annealed in

a N2 ambient. The annealing behavior of dilute nitrides has a completely unique

behavior [2, 56, 93–99] compared to all similar III-V semiconductor alloys: there is

a dramatic increase (10–50×) in photoluminescence (PL) intensity and a significant


550 700 750 800 850 9000.0

0.2

0.4

0.6

0.8

1.0

1.2

Annealing Temp (C)

PL In

t (a.u.

)

1180

1200

1220

1240

1260

1280

Wav

elen

gth

(nm

)

Figure 2.15: Annealing behavior for a GaInNAsSb QW as a function of annealtemperature. RTA time was 60 seconds.

blueshift (50–150 nm) in wavelength. Figure 2.15 shows an example of the behavior

in PL intensity and wavelength as a function of anneal temperature. The as-grown

luminescence is very poor due to defects related to low-temperature growth, plasma

damage, and possible atomic coordination. All samples must be annealed to obtain

high PL intensity, including the lasers themselves. The PL intensity increases with

hotter anneal temperatures up to the optimal annealing temperature. The degrada-

tion in PL intensity after this point is thought to be attributed to diffusion of defects

from the surface. This optimal anneal temperature, once thought to be a minor

parameter, is actually quite important in the growth of long-wavelength lasers and

will be discussed in Chapter 5.

There have been many studies of this annealing behavior and the initial hypoth-

esis was that the blue shift was due to either indium [100] or nitrogen [2, 74, 93] out

diffusion from the GaInNAs(Sb) QWs. However, as the overall material quality of

GaInNAs(Sb) has improved as a result of better growth techniques, the amount of

outdiffusion has decreased to a negligible value, making the theory outdated. The

annealing behavior appears to be a unique property of alloys with both indium and


nitrogen content. Annealing of InGaAs QWs in GaAs produces absolutely no change

in either PL intensity or wavelength. Annealing GaNAs QWs in GaAs produces a

small (2–4×) increase in PL intensity and little wavelength shift. Neither InGaAs

nor GaNAs exhibit any discernible difference in PL linewidth or lattice constant

from HRXRD. However the situation is entirely different for GaInNAs(Sb) QWs on

GaAs. These QWs show substantial increases in PL intensity and blueshift and no

change in strain from HRXRD. As-grown GaInNAsSb tends to show a higher PL

intensity than as-grown GaInNAs, and subsequently a smaller increase in intensity

after annealing.

It is now believed the dominant mechanism in causing the blueshift of wavelength

after anneal is the change in local atomic arrangement of the indium, gallium, and

nitrogen atoms [95, 96, 101]. Atomic rearrangement has been studied in-depth by

Lordi et al. using XANES and EXAFS [34, 96]. In as-grown material, the atomic

distribution of indium, gallium, and nitrogen is random. In this situation, ∼40% of

nitrogen atoms have one indium nearest neighbor, ∼25% have zero or two nearest

neighbors, and the remaining have three or four nearest neighbors. After annealing,

this distribution changes significantly and a majority of the nitrogen atoms have two

or three indium nearest neighbors. A decrease in chemical energy with increasing

numbers of nitrogen/indium bonds is due to an overall decrease in individual bond

strains. The longer nitrogen/indium bond is less stretched from equilibrium than

the nitrogen/gallium bond. By decreasing the distance between the nitrogen and

indium bonds, the atomic interaction parameter increases, increasing the band gap.

Local atomic rearrangement, however, does not explain the cause of blueshift that

exists in GaNAs(Sb), an indium-free alloy. This is still under investigation.

2.3.5 Electroreflectance and photoreflectance spectroscopy

Photoreflectance (PR) and electroreflectance (ER) spectroscopy are useful tools in

the analysis of semiconductor properties such as the energy band structure, quantum

well depths, and trap levels. In PR, a chopper-modulated laser beam is directed

towards a semiconductor sample which absorbs the light, creates electron-hole pairs,


and causes modulation of the surface electric field, manifested as a measured change

in the reflectivity. A second probe beam, whose energy is scanned, is focused on

the illuminated area to measure the spectral dependence of the change in reflectivity

by using phase-sensitive detection The final data set obtained is a plot of ∆R/R

versus wavelength. The features in the PR spectra are related to the derivatives

of the energy band transitions and thus give valuable information on the interband

transitions of the sample. By using model fitting procedures, one is able to determine

important parameters such as heterojunction band-offsets, band gaps, dopant levels,

and internal electric fields [102, 103]. ER is similar to PR except that rather than

using a laser beam, the electric field in the sample is directly modulated using a

variety of methods of applying voltage to the sample.

Of key importance in dilute nitride antimonide semiconductor laser devices is the

heterojunction band offsets between GaAs, the QW barriers, and the QWs them-

selves. Band offsets are very important in laser device design. One of the primary

advantages of GaInNAs(Sb)/GaAs devices over InGaAsP/InP devices is the higher

To for the GaAs-based lasers due to improved electron confinement in the QW.

There have been several studies on the band offsets of GaNAs and GaInNAs to

GaAs [29, 104–106], however there have not been any experimental measurements of

these values for dilute nitride antimonides. It was unclear what effect adding anti-

mony had on the band offsets of GaNAs and GaInNAs. Traditionally, the addition of

antimony to GaAs mostly affects the valence band by pushing it upwards towards the

conduction band and has a very small effect on the conduction band (also pushing

it upwards). In the dilute nitride antimonides, it was unclear if the antimony would

mostly affect the valence band or if there would be a more complex interaction of

the valence band and conduction band due to effects such as the BAC model.

In heterojunction band offset measurements, the key parameter obtained is Qc,

the conduction band offset. It is defined as

Qc =∆Ec

∆Ec + ∆Ev

. (2.9)

Conversely, Qv is the valence band offset and is simply 1-Qc. These values do not take

2.4. CONCLUSION 63

into account strain, which is generally present in most heterojunctions. Compressive

hydrostatic strain will enlarge the band gap of a semiconductor while tensile strain

will shrink it. Biaxial strain splits the hole band into light and heavy hole bands.

The dilute nitride antimonide samples examined in this article are all under either

compressive or tensile strain and thus the calculations of the actual band offsets must

take strain into account. The numerical values for the band offsets that will be shown

have taken the strain effects into account. For further details on the methodology

and formalisms on calculating band offsets with strain, please see reference [107] and

the references within.

2.4 Conclusion

MBE is a versatile technique, well suited to grow the complex dilute nitride anti-

monide alloy system. Specialized equipment exists for all elemental source materials

to ensure well controlled deposition. There are several important tools which can

be utilized during growth, such as RHEED and pyrometry, for in situ analysis and

further growth refinement.

The major difference in the MBE system for traditional III-V semiconductor

growth and dilute nitride growth is the nitrogen rf plasma cell. There are many

challenges in optimizing the cell for semiconductor growth, but progress has been

made. Growth of dilute nitrides for 1.3 µm emission has proved successful, but

obtaining the 1.55 µm wavelength was difficult. The addition of antimony as a

surfactant has enabled the growth of GaInNAsSb which emits light at high intensity

at 1.55 µm.

As with any new alloy system, a large set of characterization tools are employed to

investigate the properties such that the knowledge can be used to further improve the

material. The primary techniques used in this thesis are RHEED, HRXRD, SIMS,

PL, and PR. The feedback provided by these techniques has enabled the development

of high quality GaInNAsSb lasers at 1.55 µm.


Chapter 3

Nitrogen Plasma Pressure

Optimization and Characterization

3.1 Plasma Physics Basics

A plasma is an ionized gas where at least one electron has been removed from a

significant fraction of the molecules. There are several methods to remove the elec-

trons including the use of an electric field, a magnetic field, electrical discharge, or

resonances at certain electron frequencies. Plasma generation is a repeating process

where a free electron is accelerated and ejects another electron from an atom after a

collision. Eventually the free electron concentration is high enough such that the gas

becomes electrically charged and shields the remaining gas from the applied electric

field. The plasma contains both electrons and ions. In plasmas used for semiconduc-

tor growth, the pressure ranges from 0.1 to 100 mTorr. At these pressures, the mean

free path is short enough that the ions and electrons dissipate much of their energy

via collisions. Lower pressures lead to more efficient generation of ionized species;

however the energies are also higher. The opposite is true for higher pressures.

It is generally believed that the cracking efficiency of a plasma is highest when

the rf skin-depth is less than the radius of the space in which the plasma is contained

[108–110]. However, it is uncertain whether or not the plasma conditions used by

most groups are optimal. Different plasma conditions can lead to varying nitrogen

65

66 CHAPTER 3. NITROGEN PLASMA PRESSURE OPTIMIZATION

fluxes and ion counts. One straightforward method of changing the efficiency of a

plasma is to vary the gas density. By changing the gas flow rate into the plasma cell,

one can change the gas density inside the crucible. The density of gas then affects

the rf skin-depth of the plasma and thus changes the efficiency of the plasma and the

types of species which are generated. Ions are not preferred during plasma growth

due to their damaging nature. They can cause roughening and break or alter bonds

and thus cause defect laden material [108]. Nitrogen ions are also known to etch

GaN during growth [109]. It is suspected that ion damage is one of the dominant

sources of defects in the dilute-nitrides [35, 110].

3.2 GaInNAs Quality with Different Gas Flows

To test the effects of different flow rates on GaInNAs quality, several samples were

grown at a variety of growth rates and gas flows. First, three samples were grown at

typical GaInNAs growth rates with different flow rates to examine feasibility as well

as the incorporation properties of higher (0.75 sccm), normal (0.50 sccm), and lower

(0.25 sccm) gas flows. From those samples, new nitrogen incorporation constants

as described in Equation 2.1 were obtained. With the new calibrations, the growth

rates were adjusted such that the compositions of the slower growth rate/lower flow

rate and faster growth rate/higher flow rate samples had the same composition and

band gap as the normal growth rate/normal flow rate sample. A summary of the

growth conditions for the samples listed above is shown in Table 3.1. The structure

for all samples consisted of a 7 nm GaInNAs quantum well grown on a 300 nm GaAs

buffer and capped by a 50 nm GaAs layer.

3.2.1 Structural and compositional analysis

A series of three samples were grown to first examine the properties and feasibility of

growing GaInNAs materials with different nitrogen gas flow rates. Using the typical

growth rates of 0.34 and 0.16 µm/h for gallium and indium, samples with 0.25,

0.50, and 0.75 sccm were characterized. HRXRD was used to analyze the structural

3.2. GaInNAs QUALITY WITH DIFFERENT GAS FLOWS 67

Gallium (µm/h) Indium (µm/h) 0.25 sccm 0.50 sccm 0.75 sccm

Slow GR 0.21 0.09√

Normal GR 0.34 0.16√ √ √

Fast GR 0.44 0.20√

Table 3.1: Summary of growth conditions for the samples described in this study.The gallium and indium growth rates for the three growth rate conditions are listed.The

√’s represent the samples which were grown with the designated growth rates

and nitrogen gas flows.

quality, strain, and composition of the three samples. Figure 3.1 shows the (004) ω/2θ

scans of the QW samples. All three samples showed distinct Pendellosung fringes

indicating excellent interfaces between the layers, good epitaxial growth, and no

relaxation or compositional segregation. As the gas flow was increased from 0.25 sccm

to 0.75 sccm, the strain of the GaInNAs QW became less compressive, indicating

increased nitrogen incorporation. SIMS confirmed no change in indium concentration

occurred between samples. Since the indium concentration remained constant, it

was possible to simulate the HRXRD scans to obtain the nitrogen concentrations.

Figure 3.2 plots the nitrogen composition corresponding to the different gas flow

rates. Also shown in the figure is a value corresponding to nitrogen composition

divided by the gas flow rate. This value, to first order, gives information on the

nitrogen cracking and incorporation efficiency of the gas flow. The value is higher at

lower flow rates and saturates past 0.50 sccm.

Once the nitrogen concentrations were determined, new relationships similar to

Equation 2.1 were developed for the 0.25 sccm and 0.75 sccm gas flows. In order to

accurately compare the optical qualities of samples with different gas flows, it is im-

portant to obtain samples which all have the same composition and wavelength. The

amount of nitrogen in the sample can drastically affect the quality of the material.

It is well known that there are many nitrogen-related defects in dilute-nitride growth

and any difference in composition will prevent comparison of samples grown at differ-

ent flow rates. The group-III growth rate was adjusted, as shown in Table 3.1, such


31 32 33 3410

100

1k

10k

100k

1M

10M

100M

(c)

(b)

Increasinggas flow

Inte

nsity

(Cou

nts)

/2 (Degrees)

(a)

Figure 3.1: (004) ω/2θ scans of GaInNAs QWs grown at identical growth rates, butdifferent flow rates. (a) 0.25 sccm, (b) 0.50 sccm, and (c) 0.75 sccm gas flows.

0.25 0.50 0.751.0

1.2

1.4

1.6

1.8

2.0

2.2

2.4

Gas flow (sccm)

Nitr

ogen

(%)

3.0

3.5

4.0

4.5

%N

/ sc

cm fl

ow

Figure 3.2: Nitrogen incorporation for different gas flow rates for GaInNAs QWs atthe same growth rate. The cracking efficiency is also plotted showing a saturationpast 0.50 sccm.

3.2. GaInNAs QUALITY WITH DIFFERENT GAS FLOWS 69

that the composition of the samples grown at the other gas flow rates would have

the same composition. SIMS confirmed that all three samples had 30-31% indium

and 1.7% nitrogen.

3.2.2 Photoluminescence measurements

PL measurements were made on the three samples with nominally identical com-

positions but different growth rates and gas flows. Ex situ annealing was also per-

formed to remove various defects in the material and improve the optical quality

[2, 56, 74, 93, 94, 97]. Figure 3.3 shows the blue-shifting of the peak PL wavelength

with annealing. The wavelengths for the three samples are comparable with the 0.25

and 0.50 sccm samples within 10 nm of each other and the 0.75 sccm sample ∼20 nm

away from the other samples for the annealed conditions. The shortened wavelength

of the 0.75 sccm can be attributed to the fact it has 1% less indium than the other

two samples. Since the wavelengths are similar, the PL intensities can be compared

to determine which samples are optically better materials. The degree of blue shift

between samples is notable. For the 0.50 and 0.75 sccm samples, the wavelength

stabilized and no further blue-shift took place. However, for the 0.25 sccm sample,

there was no such stabilization, indicating the existence of defects requiring higher

removal energies.

Figure 3.4 shows the PL intensities with different anneal temperatures for the

three different gas flow samples. The as-grown intensities are shown as the 580◦C data

point (the growth temperature of the GaAs cap). Before anneal, the PL intensity

increases with higher flow rates. For all annealing temperatures, it is seen that as

the nitrogen gas flow is increased into the cell, the GaInNAs luminescence improved.

The luminescence for the 0.50 sccm sample is 3× greater than that of the 0.25

sccm sample. The 0.75 sccm sample is higher than the 0.50 sccm sample, but the

shorter wavelength may contribute to the increased intensity. It is, at minimum, of

slightly higher intensity than the 0.50 sccm sample. The full width at half maximum

(FWHM) of the PL spectra also showed that the 0.25 sccm sample had the widest

peaks of 55-70 meV while the 0.50 and 0.75 sccm samples had very similar FWHMs


550 700 750 800 850

1280

1290

1300

1310

1320

1330

1340

1350

1360

1370

0.97

0.96

0.95

0.94

0.93

0.92

0.91

Wav

elen

gth

(nm)

Annealing Temp (C)

0.25 sccm/slow GR 0.50 sccm/norm GR 0.75 sccm/fast GR

Wav

elen

gth

(eV)

Figure 3.3: Emission wavelength as measured by PL of different gas flow GaInNAsQW samples at different annealing temperatures.

of 35-50 meV. These results agree with the theory that if a plasma is running in a

“sparse” condition where the plasma is starved of gas molecules, higher gas densities

lead to more beneficial species such as atomic N and N∗2 [108–110]. The 0.25 sccm

flow leads to a “sparse” plasma and creates a larger fraction of ions verses active

species. These results can also be compared to the cracking efficiency data shown in

Figure 3.2. Both the intensities and cracking efficiencies of the 0.50 and 0.75 sccm

samples are similar, while the 0.25 sccm sample has different properties. It is possible

that enhanced cracking efficiency also enhances the creation of damaging ions and

highly energetic species. In addition, the greater improvement with anneal of the

higher flow samples suggests the type or amount of damage caused by lower flow

samples cannot be removed as easily as with the higher flow samples. Ion damage is

highly destructive and cannot be removed entirely with thermal annealing.

3.3. EFFECTS OF GAS FLOW VARIATION ON THE NITROGEN PLASMA71

550 700 750 800 8500.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

PL

Int (

a.u.)

Annealing Temp (C)

0.25 sccm 0.50 sccm 0.75 sccm

Figure 3.4: Peak PL intensity with different anneal temperatures for the differentgas flow samples.

3.3 Effects of Gas Flow Variation on the Nitrogen

Plasma

3.3.1 Ion count and energy measurements

Analyzing the properties of the GaInNAs QWs grown at different nitrogen gas flows

and growth rates, but maintaining the same composition, is an effective method of

examining the changes in the plasma and its effect on material quality. However, it

can be argued that increasing or decreasing the growth rates alters the kinetics of

dilute nitride growth. Increasing the growth rate can reduce the amount of conta-

mination per layer, but it also reduces the surface diffusion length. Decreasing the

growth rate can increase contamination, but allow for adatom diffusion to thermody-

namically favored sites. To examine the plasma contribution to the optical properties

of the GaInNAs, ion count measurements were performed. Highly energetic species

and N+, N+2 , or N++

2 ions can damage the surface of the sample and cause a variety of

point defects. Voltage biased deflection plates placed at output end of the rf plasma

cell can be used to eliminate the ions [57], but it would be preferential to reduce the


Figure 3.5: Schematic of the Langmuir probe utilized in this study to analyze plasmaproperties. The beam flux gauge is rotated towards the nitrogen cell and is nominallyfound in the same position as the wafer during growth.

number by first optimizing plasma conditions. Lastly, a voltage bias cannot remove

non-charged energetic species.

The plasma was characterized using a Langmuir probe to measure ion counts

exiting the cell. The Langmuir probe and the setup have been discussed in depth

elsewhere [35, 58], but a brief description will be given and a schematic can be seen in

Figure 3.5. The nude ion gauge filament found behind the substrate heater as shown

in Figure 2.1 can be used to measure ion and electron currents in the line-of-sight

between the cell and substrate. The current measured in the filament can give an

idea of the number of electrons and/or ions impinging upon the substrate surface. A

voltage bias to the filament can be applied to separate ions and electrons and may

also give the energies of the ion species. Standard ammeters and high-voltage power

supplies were used to measure the current and provide filament bias.

3.3. EFFECTS OF GAS FLOW VARIATION ON NITROGEN PLASMA 73

Using the Langmuir probe, a series of ion count measurements were made of the

three different nitrogen gas flows. Figure 3.6 shows the measurements of the filament

current plotted against the bias applied to the filament. Two characteristic values

may be obtained from data shown in the plot: maximum ion energy and relative

ion count. The maximum ion energy may be obtained by analyzing the filament

currents found for positive filament bias voltages. For a small filament voltage bias,

the electric field created by the bias repels ions up to the corresponding ion energy.

If an ion has an energy of 5 eV and the voltage applied to the filament is 10 eV, the

ion will be repelled away from the filament. Once the applied voltage has exceeded

the maximum ion energy, the filament current should theoretically saturate to a

constant value since there is no current caused by an impinging ions. In the data,

the value saturates to a linear value due to the presence of secondary electrons which

contribute to the filament current. The exact source of the secondary electrons is

unknown but can potentially be created with sufficient voltage to pull electrons off

the grid surrounding the filament. Thus, the maxiumum ion energy can be obtained

by extrapolating the point in which the filament current begins to deviate from the

asymptotic value at large currents. Figure 3.7 plots the maximum ion energies with

respect to different flow rates. The maximum ion energy increases as the gas flow

is decreased. Lower flow rates create a condition inside the plasma such that ion

generation is much more efficient than that of higher flow rates. This is consistent

with the general theory that higher gas pressures inside the crucible lead to smaller

rf skin-depths and thus create a more optimal plasma for reactive nitrogen species.

Examination of the filament currents for negative biases indicates the relative

number of ions impinging at the substrate location. The 0.25 sccm flow has more

current than that of the 0.50 and 0.75 sccm flows, which are roughly equal to within

error of the measurement. This indicates the 0.25 sccm flow has more ions impinging

on the surface of the substrate and causes more damage. Higher ion counts are

expected for sparse plasmas since the probability of the ion hitting another ion,

molecule, or electron is lower than that in dense plasmas. The ion can then escape

without interaction to the substrate.


-20 0 20 40 60

-12

-10

-8

-6

-4

-2

0

2

Fila

men

t Cur

rent

(nA

)

Filament Voltage (V)

0.25 sccm 0.50 sccm 0.75 sccm

Figure 3.6: Langmuir probe measurements of the plasma species exiting the cell withdifferent gas flows.

0.25 0.50 0.75

20

25

30

35

40

Ion

Ene

rgy

(eV

)

Flow Rate (sccm)

Figure 3.7: Maximum ion energies for the ions exiting the plasma cell as a functionof gas flow rate.

3.4. CONCLUSION 75

3.3.2 Material quality and plasma properties correllation

The results obtained from the Langmuir probe analysis of the plasma indicate ion

damage is a dominant factor in altering the optical quality of GaInNAs alloys. The

0.25 sccm flow rate created a plasma which created the largest number of ions as

well as the highest energy ions. Although the relative ion counts of the 0.50 and 0.75

sccm flow rates were similar, the maximum ion energies were lower for the 0.75 sccm

flow rate compared to the 0.50 sccm flow rate. These measurements agree with the

PL intensity results; the 0.25 sccm sample had the worst intensity, while the 0.50

and 0.75 sccm samples were substantially improved.

The similarity between the ion counts, PL intensity (taking into account the

slightly shorter wavelength for the 0.75 sccm sample), and similar incorporation

efficiency suggests that the plasma with 0.50 and 0.75 sccm flow rates operates in a

comparable manner. From these results, the gas flow into this rf plasma cell should

be equal to or greater than 0.50 sccm to minimize plasma damage. The necessary

gas flows will vary for each cell design, but it is important to make sure the cell is

not operating in a “sparse” condition as it decreases optical quality due to increased

ion generation.

3.4 Conclusion

Plasma optimization and damage minimization is important in the growth of dilute-

nitride materials. While much focus has gone into non-plasma related growth condi-

tions, the plasma itself is a dominant factor in high-quality MBE GaInNAs growth.

Sparse gas density conditions lead to decreased optical performance. Higher gas flows

reduce the number of damaging ions and maximum ion energies originating from the

plasma.


Chapter 4

GaInNAsSb Quantum Well

Barrier Investigation

4.1 Quantum Well Barrier Choices

Several different materials have been utilized as the layers surrounding the dilute

nitride QWs in long-wavelength optoelectronic devices. The selection of QW barrier

materials is quite numerous as long as it can be grown of very high quality below

and above the QW. However, in MBE, the alloy constituents and compositions of the

QW barrier materials are usually dictated by the sources available in the machine as

well as the QW growth conditions itself. Since most machines have only one source

for each material, changing source beam fluxes mid-growth is impossible without any

sort of valved apparatus. They may not be changed during growth since the sources

must be kept thermally stabilized to ensure a steady flux and are preferentially cali-

brated for the QW composition. This restrictive nature of MBE severely restricts the

available materials selections and compositions and thus forces a closer examination

of how the entire active region is grown.

For GaInNAsSb QWs, there are three easily obtainable alloys which can be uti-

lized as the QW barrier material: GaAs, GaNAs, and GaNAsSb. Initial GaInNAs

devices employed GaAs barriers [111–113]. However, for newer devices, GaNAs has

been the predominant choice as the QW barrier material, especially for emission

77

78 CHAPTER 4. GaInNAsSb QW BARRIER INVESTIGATION

wavelengths longer than 1.4 µm [114–117]. Originally, GaNAs was preferred to GaAs

since it was thought to reduce nitrogen out diffusion from the GaInNAs QW dur-

ing annealing [93]. This is no longer the case with improved dilute nitride growth

quality, but several advantages remain. GaNAs is tensily strained on GaAs and

provides strain compensation for the highly compressive GaInNAs QWs, enabling

MQW active region structures. The GaInNAs QWs are grown past the theoretical

critical thickness, but do not relax in a SQW structure. The addition of a second or

third QW without strain compensation induces the formation of misfit dislocations

to relieve strain energy. GaNAs barriers also lengthen the emission wavelength com-

pared to GaAs barriers due to decreased confinement in the GaInNAs QW, shifting

the energy levels lower. With the addition of antimony, another option developed:

GaNAsSb. It was originally thought that since antimony had improved the material

quality of GaInNAs, it would do the same with GaNAs and lead to better overall

active region quality.

In addition to the previous three materials mentioned, there are other logical

material choices which could be useful for GaInNAsSb QWs. GaAsP QW barriers

have been used in GaInNAs devices grown by OMVPE [118, 119]. GaAsP is tensily

strained on GaAs and thus provides the same strain compensation as GaNAs. How-

ever, its material quality is typically better than GaNAs since there are no nitrogen

related defects in the layer. The increased band gap of GaAsP also provides increased

carrier confinement within the QW structure. GaAsP growth in MBE has not been

observed with GaInNAsSb QWs since it is rare to find a machine with four group-

V sources due to lack of available source ports. AlGaAs has also been utilized for

GaInNAs devices. Although it does not have the tensile strain advantage of GaNAs

and GaAsP, it can be grown of high quality easily and aluminum is found in most

MBE machines. However, there have been several reports of severe material degra-

dation of dilute nitride devices which have aluminum-containing layers adjacent to

dilute nitride layers and/or have both source materials in the same chamber. The

exact cause of the degradation is unknown, but it has been speculated aluminum and

nitrogen may have unexpected bonding properties at the interface or have precursor

interactions which alter the deposition quality of the material. More complicated

4.2. GaNAsSb GROWTH INVESTIGATION AND CHARACTERIZATION 79

structures involving GaInNAs barriers [120] (of different compositions) for GaInNAs

QWs have also been studied, but require additional gallium or indium sources. All

the materials mentioned here have various difficulties, leaving GaAs, GaNAs, and

GaNAsSb as the most applicable options.

The growth properties of GaNAsSb will be discussed in this chapter and the

utilization of this material, in addition to GaNAs and GaAs, as the QW barrier

will be examined. There are several considerations when deciding upon the optimal

QW barrier material. The alloy itself must have good structural and optical quality.

Defects and traps within these layers can lead to carrier leakage away from the QW,

reducing the efficiency of the active region. For MQW devices, it is important to have

strain compensation for the highly compressive GaInNAsSb QWs so that relaxation

is prevented. The band alignment of the QW barrier materials with the QW is

important; a type-I line-up is desired to ensure both electron and hole confinement.

Finally, practical growth issues must be confronted.

4.2 GaNAsSb Growth Investigation and Charac-

terization

4.2.1 Initial growth characterizations

To study the properties of the GaNAsSb material used in barriers of previous 1.3

and 1.5 µm laser devices, the barriers themselves were “converted” into SQW sam-

ples. A series of samples of 20 nm GaNAs(Sb) QWs with GaAs barriers were grown

to examine different growth conditions, such as substrate temperature and arsenic

overpressure. The compositions of GaNAs(Sb) used in this study were chosen to be

similar to those used in 1.3 and 1.55 µm GaInNAsSb devices and are uniquely deter-

mined by the GaInNAsSb growth conditions. The nitrogen content is predetermined

due to its inverse proportional relationship with the group-III flux and the antimony

flux is unchangeable since it is supplied by an unvalved cracker. These combined

conditions do not allow the barrier compositions to be arbitrarily changed.


The GaNAs(Sb) SQWs were all grown at 440◦C (except for the substrate tem-

perature study) and the growth rate was either 0.45 µm/h (to duplicate 1.3 µm QW

barriers) or 0.30 µm/h (to duplicate 1.5 µm QW barriers). The composition of nitro-

gen in GaNAs was determined by HRXRD and nitrogen and antimony in GaNAsSb

by SIMS and HRXRD. As mentioned earlier, the QW thicknesses were all 20 nm

with 50 nm GaAs capping layers. An arsenic-to-gallium overpressure of 20× (except

during the arsenic overpressure study) and an antimony flux of 0.8–1.0×10−7 Torr

beam equivalent pressure (BEP) were supplied during the QW growth.

RHEED is a good method to examine the quality and properties of the growth

surface. During the growth of GaNAs and GaNAsSb samples, RHEED patterns

were recorded to examine any significant differences on the surface in the presence

of antimony. Figures 4.1a and 4.1b show the patterns obtained from the GaNAs and

GaNAsSb samples, respectively. It is seen that the RHEED pattern from GaNAs is

streaky. Although not shown in the figure, the orthogonal [110] directions for GaNAs

grown under the growth conditions mentioned above showed a 2×4 reconstruction.

When antimony was applied to form GaNAsSb, the RHEED pattern changed signifi-

cantly. Instead of a streaky pattern, a spotty pattern emerged suggesting the surface

quality was not as good as GaNAs. This result was somewhat surprising since previ-

ous studies have shown antimony had the opposite effect with InGaAs and GaInNAs

[71, 72, 74]. Again, in examining orthogonal [110] directions, the reconstruction

for GaNAsSb appeared to be 1×4. From the observed RHEED patterns, it can be

concluded the growth surface of GaNAs is smoother than that of GaNAsSb.

Figure 4.2 shows the HRXRD (004) ω/2θ scans from four different GaNAs(Sb)

SQW samples. The nitrogen compositions of the GaNAs samples were determined

by simulation of the HRXRD spectra. For the GaNAsSb samples, simulations were

also performed using values obtained for the antimony and nitrogen compositions

from SIMS to confirm the validity of the SIMS measurements. The results for the

1.3 and 1.55 µm device growth conditions are shown in Table 4.1. As expected in 1.55

µm device growth conditions, there was more nitrogen found in both samples due to

the slower growth rate. For a fixed antimony-to-arsenic flux ratio, variations in the

4.2. GaNAsSb GROWTH INVESTIGATION AND CHARACTERIZATION 81� � � � � �

Figure 4.1: RHEED pictures showing the streaky patterns from (a) GaNAs and thespotty patterns from (b) GaNAsSb.

growth rate did not change the antimony composition. Antimony incorporation ap-

peared to be independent of altered growth conditions, such as the group-III growth

rate and nitrogen incorporation, suggesting the flux ratio of antimony-to-arsenic is

the deciding factor in determining the composition. This is different than what is seen

in GaAsSb growth, where for fixed antimony-to-arsenic flux ratios, different growth

rates lead to different incorporation rates [121]. The increase in nitrogen composi-

tion in GaNAsSb compared to GaNAs in both conditions is consistent with previous

observations [84–86, 122]. The mechanism for the increased sticking of nitrogen in

the material is not known. However, it is thought that the properties of antimony as

a reactive surfactant [65, 66] help promote the incorporation of a different species of

nitrogen into GaAs not normally incorporated. As observed from the HRXRD scans,

the GaNAsSb was either lattice-matched or was slightly compressively strained to

GaAs for both compositions. This property is not advantageous when used with

highly compressively strained GaInNAsSb QWs since the barriers would not provide

any strain compensation in the active regions of MQW devices. However, both com-

positions of GaNAs showed an appreciable amount of tensile strain. The amount

of strain found in GaN0.019As0.981 and GaN0.027As0.973 was –0.38% and –0.55%, re-

spectively. From HRXRD, there does not appear to be a significant improvement or

degradation of material quality upon addition of antimony to GaNAs for either set

of growth conditions.


32.0 32.5 33.0 33.5 34.01

10

100

1k

10k

100k

1M

10M

100M

1G

(d)

(c)

(b)

(a)

Inte

nsity

(a.u.)

(Degrees)

GaAs

Figure 4.2: (004) ω/2θ HRXRD spectra showing the amount of strain in the sam-ples. (a) GaN0.029As0.873Sb0.098, (b) GaN0.034As0.867Sb0.099, (c) GaN0.019As0.981, and(d) GaN0.027As0.973. (a) and (c) are grown under the 1.3 µm device growth conditionswhere as (b) and (d) are grown under 1.55 µm device growth conditions.

Device growth conditions GaNAs GaNAsSb1.3 µm 1.9% N 2.9% N, 9.8% Sb1.55 µm 2.7% N 3.4% N, 9.9% Sb

Table 4.1: XRD and SIMS compositional results of GaNAs and GaNAsSb grownunder the normal 1.3 and 1.55 µm QW growth conditions.


SIMS was performed on the GaNAsSb samples to obtain the compositional values.

Figure 4.3 shows the SIMS depth profile for the GaN0.029As0.873Sb0.098 sample. The

compositions were confirmed with HRXRD data and simulations. An interesting

feature to note in the SIMS depth profile is the top interface of the GaNAsSb layer.

The nitrogen and antimony profiles do not end at the same location within the sample

when shutters for both sources were closed at the same time. This was determined

not to be a measurement or sputtering artifact, as it was repeatable within the same

sample and was seen on all other samples measured with SIMS. Antimony continues

to incorporate∼2 nm beyond the end of nitrogen incorporation. Surfactants typically

surface segregate on the growth front and do not incorporate into the material. As

mentioned in earlier papers relating to GaInNAsSb, antimony appears to act as

a surfactant and group-V component [72, 74, 83, 84]. It is likely that antimony

partially incorporates and partially segregates on the growth front for GaNAsSb as

well. If this is the case, there will be antimony remaining on the growth front after

the shutter is closed to the cell and the residual antimony continues to incorporate or

desorb until the supply is exhausted. This growth artifact could be quite detrimental

to devices since there is a thin layer of GaAsSb which could significantly change the

originally-intended band structure properties due to changes of both composition

and strain.

PL measurements were obtained from the GaN0.029As0.873Sb0.098 sample. No sig-

nal was observed for the GaN0.034As0.867Sb0.099 sample, either as-grown or annealed.

This suggests the material was of very poor optical quality and it was not studied

further. The PL obtained from the as-grown GaN0.029As0.873Sb0.098 sample had a

peak wavelength of 1.316 µm, but was very weak in intensity. This was not sur-

prising since most groups report poor PL intensity for GaInNAs(Sb) samples which

have not been annealed [56, 74, 83]. The sample was annealed at a series of tem-

peratures between 720◦C and 820◦C to study the effect upon the optical quality of

the material. Similar to GaInNAs(Sb), annealing the PL samples led to a dramatic

increase in PL intensity. As seen in Figure 4.4, the PL intensity improved with in-

creasing annealing temperatures until it peaked at 760◦C and decreased beyond this


20 40 60 80 100 1200

1

2

3

4

5

6

7

8

9

10

11

Nitrogen

Antimony

Mole

Frac

tion

(%)

Depth (nm)

Figure 4.3: SIMS depth profile of antimony and nitrogen for a GaN0.029As0.873Sb0.098

QW sample.

point. The PL peak wavelength also blueshifted with increasing anneal tempera-

tures. Compared to the as-grown PL spectrum, the optimal anneal PL signal was

25× higher in intensity and was blueshifted 55 meV, which is similar to the blueshift

found in GaInNAs(Sb). Unlike GaInNAs(Sb) samples, there is no indium in the

samples and thus the blueshifting of the PL wavelength upon annealing cannot be

explained by In/Ga/N rearrangement [95, 96, 101]. Sources of blueshifting likely

include nitrogen outdiffusion, N/As/Sb rearrangement, and nitrogen de-clustering.

SIMS and HRXRD measurements indicated a slight reduction in nitrogen concentra-

tion after annealing for older samples, but newer samples have shown little change in

concentration after anneal. When compared to typical GaInNAs(Sb) PL intensities,

the GaNAsSb intensities are at least 25× lower. The low intensity in comparison

to other nitride-arsenides is likely due to poor optical quality material. One final

point to note is the actual transition energy of the GaN0.029As0.873Sb0.098 sample in

comparison with the QW material it surrounds in devices. If it is assumed that the

PL peak wavelength gives a rough estimate of the bandgap of the material, then it

is seen that the bandgap of the GaN0.029As0.873Sb0.098 is roughly 0.99 eV while the


400 700 750 800 850

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

Anneal Temperature (C)

PL In

tens

ity (a

.u.)

1240

1250

1260

1270

1280

1290

1300

1310

1320

Wav

elen

gth

(nm

)

Figure 4.4: PL results from the GaN0.029As0.873Sb0.098 sample (barrier material forthe 1.3 µm QWs). The blue line shows PL intensity. The red line shows the peakPL wavelength.

QW at 1.3 µm is 0.95 eV. With only ∼40 meV difference in bandgap, there is very

poor confinement of electrons and holes within the QW and it is possible that the

alignment between the GaNAsSb and GaInNAsSb at 1.3 µm is not the desired type-I

alignment. The heterojunction band offsets for GaNAsSb will be discussed later in

this chapter.

4.2.2 Arsenic overpressure examination

In an attempt to study general growth properties of GaNAsSb and potentially im-

prove PL intensity and material quality, a series of samples with varying arsenic-

to-gallium overpressures was grown with the same structure as the previous study.

It is known that in mixed group-V materials, the relative fluxes of each group-V

element play a large role in composition and growth kinetics. In GaInNAs, there is

no significant effect on nitrogen incorporation by different arsenic fluxes due to the

“unity” sticking properties of nitrogen [56]. However in GaNAsSb, it was suspected

that the arsenic and antimony fluxes affected each other since they do not have the


same sticking properties as nitrogen. It is also possible that a variation in antimony

incorporation could affect the nitrogen composition. To test the effects of different

arsenic overpressures on GaNAsSb, the original 20× arsenic-to-gallium flux overpres-

sure was varied between 15×, 25×, and 30× when growing GaN0.029As0.873Sb0.098,

the material used as barriers for 1.3 µm QWs. All other growth conditions were held

constant.

In HRXRD measurements, it was seen that as the arsenic overpressure increased

from 15× to 30×, the strain in the GaNAsSb layer became less compressive. Since

this is a quaternary system, it cannot be determined whether the decrease in com-

pressive strain was due to a reduction in antimony concentration, an increase in

nitrogen concentration, or a combination of both. In all cases, the HRXRD scans

did not show any degradation of material compared to the original 20× sample. To

determine the origin of the strain reduction, SIMS was performed to measure the

composition. Figure 4.5 plots the results obtained from the SIMS analysis. As the

arsenic overpressure is increased from 15× to 30×, the data shows that the anti-

mony concentration drops from 12% to 9% while the nitrogen concentration remains

constant. This explains the decrease in compressive strain with increasing arsenic

flux since a reduction in antimony would decrease the lattice constant of GaNAsSb.

According to the SIMS data, the increase in arsenic flux had a direct effect on the

antimony incorporation rate but had no discernable effect on nitrogen incorporation

(similarly seen in GaNAs). This decrease in antimony incorporation with increas-

ing arsenic flux is seen commonly in GaAsSb growth [121]. In addition, the change

in antimony concentration had no effect on the nitrogen incorporation in agreement

with previously obtained results [84, 122]. The data also show significantly enhanced

nitrogen incorporation in the GaNAsSb. GaNAs grown under the same growth con-

ditions yields 1.8% N, much lower than the observed 2.4–2.9% in GaNAsSb.

PL measurements revealed no significant change in optical material quality be-

tween the different samples. The PL spectra from the varying arsenic overpressure

samples are shown in Figure 4.6 and are from the optimal annealing temperature.

The PL peak wavelengths shifted due to different antimony concentrations. Adjust-

ing the arsenic overpressure from 15× to 30× arsenic overpressure results in a 3%


15 20 25 300

2

4

6

8

10

12

Mol

e Fr

actio

n (%

)

Arsenic/Gallium Overpressure

Nitrogen

Antimony

Figure 4.5: SIMS results from the arsenic overpressure study.

antimony decrease, shifting the peak wavelength from 1.275 to 1.220 µm. The PL

intensities are equal within measurement error and sample repeatability. Changing

the arsenic overpressure does not seem to have any major effect on GaNAsSb except

for the difference in antimony incorporation. Structural and optical quality remains

the same. For GaInNAsSb QWs with GaNAsSb barriers, it would be beneficial to

increase the arsenic overpressure so that the GaNAsSb would have less compressive

strain and have a larger bandgap for increased confinement in the wells.

4.2.3 Growth temperature examination

The substrate temperature during GaNAsSb growth was also varied to examine

the effects on crystal quality and composition. GaInNAs(Sb) was grown at 440◦C

to inhibit compositional segregation and relaxation. One of the driving factors in

segregation in GaInNAs(Sb) is the clustering of indium-rich areas. The GaNAsSb

barriers were also grown at the same temperature to ensure substrate temperature

stability for the QW. However, indium is not present in this material and thus raised

the possibility the material could be grown at a higher temperature. One issue

with dilute nitride growth is the low growth temperature. These low temperatures


1050 1100 1150 1200 1250 1300 1350 1400 14500.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.01.15 1.1 1.05 1 0.95 0.9

(c)(b)

(a)

PL In

tens

ity (a

.u.)

Wavelength (nm)

Energy (eV)

Figure 4.6: PL spectra from the GaN0.029As0.873Sb0.098 sample grown at differentarsenic-to-gallium overpressures. (a) 30×, (b) 25×, and (c) 15×.

introduce defects in GaAs materials, such as arsenic antisites [123, 124] and gallium

vacancies [125, 126]. It is ideal to grow the material as close to 580◦C as possible

to minimize these defects which cause non-radiative recombination [127, 128] and

reduce luminescence. A series of samples with structures and growth conditions

identical those mentioned earlier in this section were grown with varying substrate

temperatures: +35◦C (475◦C), +70◦C (510◦C), and +105◦C (545◦C). Another set of

samples with no antimony (GaNAs) was also grown for comparison.

HRXRD scans of the GaNAs and GaNAsSb showed that temperature did have an

effect on the composition and strain of the material. When the growth temperature

was increased, the strain in the GaNAsSb samples shifted from slightly compressive

to slightly tensile on GaAs. It is suspected that the shift in strain is most likely due

to a reduction in antimony since it is known that antimony tends to desorb more

readily at higher growth temperatures. It is possible that the change in strain is also

due to an increase in the nitrogen incorporation rate, although it is highly unlikely.

To determine whether temperature has an effect on nitrogen incorporation, GaNAs


was grown at the same increased growth temperatures used for GaNAsSb. From

HRXRD scans, it was seen that the nitrogen composition remained the same, except

for the hottest sample in which there was a slight decrease in nitrogen. Nitrogen

incorporation at moderately hotter temperatures is independent of growth tempera-

ture and thus is not the cause for the shift in GaNAsSb strain. The most surprising

result from the HRXRD scans was the fact that the samples did not appear to have

any significant compositional segregation, relaxation, or interface degradation when

grown at higher temperatures. All samples had very well defined QW peaks and

Pendellosung fringes. To confirm that the (004) ω/2θ scans were not missing any

signs of relaxation or compositional segregation, (224) reciprocal space maps (RSMs)

were taken of GaNAs and GaNAsSb. Figure 4.7 shows the RSM of GaNAsSb grown

at 545◦C. The GaNAs RSM, not shown, is very similar in appearance. It is seen in

the figure there are no major diffraction peaks in the in-plane direction away from

the (224) GaAs peak. Even at high temperatures, GaNAsSb can be grown coherently

on GaAs.

SIMS scans were taken of the GaNAsSb samples to measure composition and

depth profiles. As seen in Figure 4.8, there is a large decrease in antimony con-

centration with increasing substrate temperature while the nitrogen concentration

remained roughly constant. The loss of 8% antimony explains the large shift in the

strain observed in HRXRD peaks. Similar to the SIMS data from the arsenic over-

pressure investigation, the nitrogen composition remained constant, even though the

antimony concentration changed.

With this data, one might be encouraged by the fact that at high temperatures,

the material obtained was coherent and had smaller amounts of antimony so that

the bandgap was increased. However, the PL results, displayed in Figure 4.9, were

unexpected. With increasing substrate temperature, the PL spectra blueshifted as

expected due to lower antimony concentration, but also decreased in intensity. There

is also a large shoulder to the PL spectra for all three samples which was not found

in the original substrate temperature sample. This longer wavelength shoulder could

be a result of nitrogen clustering which may not be observed easily in HRXRD. Since

there could be areas of clustering of increased nitrogen concentration within the QW,


Q

Q ||

GaAs

GaNAsSb

Figure 4.7: (224) reciprocal space map of the GaN0.029As0.873Sb0.098 sample grown athigh temperature (545◦C). No in-plane components from the QW different from thesubstrate are seen in the diffraction pattern.


440 460 480 500 520 5400

2

4

6

8

10

12

Mol

e Fr

actio

n (%

)

Substrate Temperature (C)

Nitrogen

Antimony

Figure 4.8: SIMS results from the growth temperature study.

it could lead to areas of luminescence in which the bandgap is smaller, leading to

luminescence of longer wavelength. If these regions occurred only inside the QWs,

they would have no effect on the interfaces and thus would not affect or reduce the

diffraction thickness oscillations. As evidenced by the poor PL results, GaNAsSb

cannot be grown at high temperatures without a large decrease in optical material

quality.

4.2.4 Antimony reduction for improved luminescence

Although antimony improved GaInNAs material quality and surface morphology, it

provided little improvement to GaNAsSb; this may be due to the nature of antimony

incorporation and the effects of adding indium. For the same 1.0×10−7 Torr BEP

flux of antimony, 2% antimony is found in GaInNAsSb while 8–11% is found in

GaNAsSb. The change in antimony concentration is not primarily caused by a change

in growth rates. This was confirmed by the growth of a GaInNAsSb sample and a

GaNAsSb sample, grown with the same total group-III growth rate and antimony

flux, that showed differing antimony concentrations. Therefore, it is indium that

dramatically changes antimony incorporation between GaNAs and GaInNAs. This


1050 1100 1150 1200 1250 1300 1350 1400 14500.0

0.5

1.0

1.5

2.0

2.5

1.15 1.1 1.05 1 0.95 0.9

(c)(b)

(a)

PL In

tens

ity (a

.u.)

Wavelength (nm)

Energy (eV)

Figure 4.9: PL spectra from the GaN0.029As0.873Sb0.098 sample grown at different sub-strate temperatures. (a) +35◦C (475◦C), (b) +70◦C (510◦C), and (c) and +105◦C(545◦C). The small peak at 1400 nm is due to water present in the testing environ-ment.


will be discussed further in Chapter 5. It is possible that an excessive amount of

antimony flux was present during the growth of GaNAs(Sb), negating antimony’s

surfactant properties. To examine this effect, reduced antimony fluxes were utilized

to decrease antimony incorporation and attempt to improve the optical quality. Four

samples were grown under identical growth conditions except the antimony flux. The

first sample was a GaNAs sample with 0.63% nitrogen and no antimony. The next

three samples were grown using the same procedures but included antimony fluxes

of 1.0×10−8, 2.0×10−8, and 2.8×10−8 Torr BEP, respectively.

The HRXRD (004) ω/2θ scans, shown in Figure 4.10, were used to examine

structural quality and to measure the biaxial strain in the four GaNAs(Sb) layers.

The nitrogen composition of the GaNAs sample was determined by simulation of the

HRXRD spectra. The simulated spectrum of the GaNAs sample is shown in Figure

4.11 with the real data for comparison. Simulations were also performed on the

GaNAsSb spectra, but required SIMS for confirmation since quaternary compound

compositions cannot be determined uniquely. All four samples have well-defined

film diffraction peaks as well as distinct Pendellosung fringes. This suggests that all

of the GaNAs(Sb) layers are coherent with the GaAs substrate, have good quality

interfaces, and have no discernable structural defects. Reciprocal space maps also

indicated no relaxation had occurred. The (224) RSM for the GaNAs can be seen in

Figure 4.12. The other three space maps were very similar in appearance. As seen

in Figure 4.12, the strain of the GaNAs layer is tensile in nature as expected. With

increasing antimony fluxes, the tensile strain decreases and approaches a lattice-

matched condition in the GaNAsSb sample with 2.8×10−8 Torr BEP antimony. The

strain of the four samples was calculated from the position of the film diffraction

peak and the results are shown in Table 4.2.

SIMS was performed on the four samples to confirm and correlate the compo-

sitions with those obtained from HRXRD. The results for the antimony containing

samples were used in conjunction with additional HRXRD simulations to refine the

nitrogen values obtained from SIMS. A summary of the results from those measure-

ments is displayed in Table 4.2. Although the same nitrogen flux was used for the

GaNAs sample as for the other three, the amount of nitrogen in the layer increased


32.9 33.0 33.1 33.21k

10k

100k

1M

10M (d) (c) (b) (a)

Cou

nts

(a.u.)

/2 (degrees)

GaAs

Figure 4.10: (004) ω/2θ HRXRD spectra of the four GaNAs(Sb) layers. (a)GaN0.0063As0.9937, (b) GaN0.0071As0.9869Sb0.006, (c) GaN0.008As0.978Sb0.014, and (d)GaN0.0091As0.9709Sb0.02. The tensile strain decreases with increasing antimony flux.

32.9 33.0 33.1 33.2

100

1k

10k

100k

1M

10M

Simulation

Data

Cou

nts

(a.u.)

/2 (degrees)

GaAsSubstrate

GaNAs Film

Figure 4.11: (004) ω/2θ HRXRD spectrum of the GaN0.0063As0.9937 with its corre-sponding simulated spectrum.


Q

Q ||

GaNAs

GaAs

Figure 4.12: (224) reciprocal space map of the GaN0.0063As0.9937 sample.

Antimony Flux BEP (Torr) Nitrogen (%) Antimony (%) Strain0 0.63 0 –0.13%

1.0×10−8 0.71 0.6 –0.10%2.0×10−8 0.80 1.4 –0.03%2.8×10−8 0.91 2.0 –0.005%

Table 4.2: Summary of antimony fluxes utilized, GaNAs(Sb) compositions obtainedfrom SIMS and HRXRD, and strain from HRXRD.


with increasing antimony flux. The enhancement of nitrogen incorporation with an-

timony has been reported previously for larger antimony concentrations [84, 85]. The

effect is different in this compositional regime. The addition of more antimony does

increase the efficiency with which nitrogen incorporates; the increase in nitrogen is

roughly linear with the antimony flux and composition. For the GaNAsSb samples

with larger antimony fluxes, only a constant multiplicative enhancement of nitrogen

was seen with varying antimony fluxes. In the low antimony concentration regime,

the enhanced nitrogen incorporation behavior can be fitted to the following linear

relationship:[N ]

[No]= K[Sb] + 1, (4.1)

where [N] is the nitrogen concentration, [No] is the nitrogen concentration without

antimony present, [Sb] is the antimony concentration in percent, and K is a con-

stant representing the incorporation enhancement. The samples in this work had

a much higher enhancement of K=0.21 compared to the value of K=0.03 reported

by Harmand et al. [86]. In their studies, they utilized Sb2 rather than monomeric

antimony and did not study multiple compositions ≤3%. It is possible that the rise

in nitrogen incorporation efficiency is a rapid phenomenon for small amounts of an-

timony and levels off to a smaller value with antimony concentrations greater than

∼5%. Although it has not been examined, it is suspected that the monolayer cov-

erage saturates above ∼5%, altering the surface kinetics of antimony and nitrogen

incorporation. It is clear that antimony does have a significant effect on nitrogen

incorporation in the 0–2% antimony range.

Figure 4.13 shows PL measurements of the GaNAs(Sb) samples to determine op-

tical quality. The GaN0.0063As0.9937 layer emitted at a peak wavelength of 964 nm.

With the addition of antimony to the GaNAs layer, the peak intensity increased

and the wavelength redshifted with increasing antimony flux. The shift in the peak

wavelength (out to 987 nm in the GaN0.0091As0.9709Sb0.02 sample) was not surprising

as the addition of antimony to GaAs reduces the band gap in the GaAsSb alloy. In

addition, antimony enhanced the incorporation of nitrogen, which further reduced

the band gap. Interestingly, an increase in PL intensity with larger amounts of


900 950 1000 10500

1

2

3

4

5

6

7

1.35 1.3 1.25 1.2

Sb=0 Sb=1.0x10-8

Sb=2.0x10-8

Sb=2.8x10-8

PL In

tens

ity (a

.u.)

Wavelength (nm)

eV

IncreasingSb flux

Figure 4.13: PL spectra of the GaNAs(Sb) samples showing a redshift and increasein intensity with increasing antimony flux.

nitrogen and antimony in the layer was observed, contrary to the behavior of the

GaNAsSb samples with higher antimony fluxes and concentrations. These results

indicate the amount of antimony plays a crucial role in whether or not it improves

the material. The GaN0.0091As0.9709Sb0.02 sample shows 2.3× increase in PL inten-

sity over the GaN0.0063As0.9937 sample as well as a reduction of the FWHM from

56 meV to 44 meV. It was previously believed that the addition of more nitrogen

further degrades the electronic and optical properties of the material. However, it

was observed that the addition of small amounts of antimony negates this “nitrogen

penalty” and material with more nitrogen can be grown with comparable or better

optical quality. With increasing nitrogen content and antimony content, the optical

properties have improved. The improvement of GaNAs with increasing amounts of

nitrogen and antimony shows that the “nitrogen penalty” should be thought of as a

“nitrogen complexity” in which the optimal growth parameters must be rediscovered

for each composition of the alloy.


4.3 Heterojunction Band Offset Measurements

In designing optoelectronic devices, one must give attention to the band line up of

the semiconductor alloys to ensure optimal operation. Spatial carrier confinement,

among other parameters, is important to ensure high efficiency lasers. It is preferable

to have a type-I band line up of the QW with the alloys adjacent to it so that both

electrons and holes are confined. Although type-II “W” QW lasers do exist, they

generally suffer from high threshold currents due to low gain. Since the dilute nitride

antimonides are relatively new, detailed study of their band properties have not yet

been performed. A greater understanding of the heterojunction band offsets and

effective masses would enable improved laser design and a determination of which

QW barrier materials are best suited for GaInNAsSb QWs.

One powerful method of determining band offsets is photoreflectance (PR) spec-

troscopy. By performing theoretical calculations and simulating the energy transi-

tions obtained from PR measurements, a model of the band offsets, energy levels,

and effective masses are obtained. There have been many studies on GaInNAs and

GaNAs band offsets [29, 104–106], however there have been none for GaNAsSb and

GaInNAsSb. Using PR, the band offsets were measured for three different band

line ups: GaNAsSb/GaAs, GaInNAsSb/GaAs, and GaInNAsSb/GaNAs/GaAs. Be-

fore this study, it was unclear what effect the addition of antimony would have on

the GaInNAs and GaNAs band offsets. In the dilute nitrides, it was not known if

antimony would only affect the valence band (as it does in most other III-V semicon-

ductors) or if there would be a more complex interaction of the valence and conduc-

tion bands due to effects such as band anticrossing. The GaNAsSb/GaAs structure

corresponds to the GaNAsSb composition used when applied as GaInNAsSb QW

barriers which operated at 1.3 µm. The GaInNAsSb/GaAs structure corresponds

to material which had an emission wavelength of 1.5 µm. This structure is not

typically used in laser devices but is an important first-step in analyzing the more

complex laser structure. The GaInNAsSb/GaNAs/GaAs stepped QW structure rep-

resents the most technologically relevant sample since it is utilized in developing the

long-wavelength lasers on GaAs.

4.3. HETEROJUNCTION BAND OFFSET MEASUREMENTS 99

Two GaNAsSb/GaAs QW samples were grown with different QW thicknesses

of 6 and 8 nm. Figure 4.14 shows the PR spectra obtained from the two samples.

The shift in energy transitions can be seen due to the different levels of quantum

confinement of the two samples. The simulation of the PR spectra cannot be seen

clearly because it is buried beneath the actual PR signal and indicates an excellent

fit. The moduli of the various PR energy resonances are also plotted to show the

locations of the energy transitions. By analyzing and simulating the spectra and

utilizing Equation 2.9, a Qc, of 0.5 was obtained for GaN0.02As0.87Sb0.11/GaAs. The

numerical values of the offset taking strain into account are shown in Figure 4.15.

The results show that when antimony was added to GaNAs, the valence band was

lifted to higher energies. As in most other III-V semiconductors, antimony pushes

the valence band higher while not affecting the conduction band to any significant

degree [17–21]. In GaNAs, a majority of the band offset is found in the conduction

band (Qc of 0.8–1.0). It is seen that in the GaNAsSb sample, the 11% mole fraction

of antimony lifted the valence band without affecting the conduction band. This

opens up the possibility for more advanced band engineering of the dilute nitride

antimonide alloys for different applications.

To confirm the effect of antimony on the conduction and valence band offsets,

additional GaNAsSb samples were grown with different antimony concentrations

while all other conditions were held constant. Figure 4.16 shows the effects of different

antimony concentrations on Qc as well as the actual energy values after strain is taken

into account. As the antimony concentration increases, Qc decreases since antimony

enlarges the valence band offset. The valence band offset increases from ∼50 to ∼250

meV when up to 11% antimony is added to GaNAs. There is no significant change in

the conduction band with the addition of antimony, similar to the behavior of other

III-V semiconductors.

Before a study of the more complex stepped GaInNAsSb/GaNAs/GaAs QW

structure could be studied, a simpler Ga0.62In0.38N0.026As0.954Sb0.02/GaAs QW sample

was analyzed. A Qc of 0.8 was obtained from analysis of the PR spectra. Figure

4.17 shows the band lineup of the GaInNAsSb/GaAs structure. The value of Qc

for this QW is very similar to that of the antimony-free GaInNAs material. This is


-1.0

-0.5

0.0

0.5

1.0

1.5

0.9 1.0 1.1 1.2 1.3 1.4

-0.5

0.0

0.5

1.0

Exp. data Fit Modulus of individual lines

x0.05

32H

22H11L

11H

GaA

s:N GaA

s

(a)

G

aAs

GaA

s:N

11L

11H 22L

32H 22L

(b)

x0.05

105

R/R

energy (eV)

Figure 4.14: PR spectra obtained from GaN0.02As0.87Sb0.11/GaAs QW samples. (a)6 nm, (b) 8 nm. Shown are the experimental data, theoretical spectra fit (in red),and moduli of the PR energy resonances.


Figure 4.15: Band lineup for the GaN0.02As0.87Sb0.11/GaAs QW samples. The nu-merical values for the offsets have taken strain into account.

2 4 6 8 10 1240

50

60

70

80

90

QC (%

)

Sb Conc. (%)

(a) Conduction band offset ratio

0 2 4 6 8 10 120

50

100

150

200

250

300

350Compressive StrainTensile Strain

Ene

rgy

(meV

)

Sb Conc. (%)

EC*

EVLH

EVHH

(b) Numerical offsets with strain

Figure 4.16: The effects of varying antimony concentration on the (a) conductionband offset ratio Qc and (b) valence and conduction band offsets.


0 10 20 30 40 500.0

0.1

0.2

1.0

1.2

1.4

EV=146 meV

EC=520 meV

1.36

0 eV

1.12

6 eV

0.92

6 eV

0.79

9 eV

Ener

gy (e

V)

Distance z (nm)

GaInNAsSb

Figure 4.17: Band lineup for Ga0.62In0.38N0.026As0.954Sb0.02/GaAs QW sample. Nu-merical values have taken strain into account. The energy transitions for the fourconfined states are also shown.

due to the fact that there is only a small amount (2%) of antimony found within

the QW material. This small amount does not significantly affect the band lineups

and is similar to that of the antimony-free material. There is a very deep electron

confinement in these QWs and would be useful for improved thermal characteristics

due to the reduction of electron leakage during operation.

With knowledge from the previous structure, an analysis of the more complex but

technologically relevant Ga0.61In0.39N0.023As0.957Sb0.02/GaN0.027As0.973/GaAs stepped

QW structure could be undertaken. The band lineup of the GaInNAsSb/GaNAs/GaAs

structure can be seen in Figure 4.18. As mentioned earlier, the majority of the band

offset between GaNAs and GaAs is found in the conduction band. The band lineup

between GaInNAsSb and GaNAs is much different than that of the GaInNAsSb/GaAs

structure. Here, there are only two confined electron states and three hole states.

With only 144 meV in electron confinement, leakage to the QW barrier layers is

not insignificant and is probably one the major contributors to a higher-than-desired

laser threshold current and degraded thermal performance.


0 10 20 30 40 50 60 70

0.00

0.05

0.10

0.15

1.0

1.1

1.2

1.3

1.4

315 meVGaNAs

GaInNAsSb56 meV

127 meV

144 meV

hh1hh2

hh3

e1e2

Ene

rgy

(eV

)

Distance z (nm)

Figure 4.18: Band lineup of the Ga0.61In0.39N0.023As0.957Sb0.02/GaN0.027As0.973/GaAsstepped QW sample. Numerical values have taken strain into account.


A compiled summary of the band offset measurements can be found in Figure

4.19. Both GaAs and GaNAs alloys have the required band alignment to GaInNAsSb

QWs for type-I structures. The band alignment shown for GaNAsSb with respect to

the other alloys is for the composition of GaNAsSb utilized as barrier layers for 1.3

µm emitting GaInNAsSb QWs. Results were not obtained from the GaNAsSb com-

position when it was employed as barriers for 1.5 µm emitting GaInNAsSb QWs due

to very poor optical quality. Growth conditions for the 1.5 µm emitting GaInNAsSb

QWs resulted in GaNAsSb with a higher nitrogen content due to the slower group-III

growth rate. Antimony composition was not changed. This would have lowered the

conduction band by ∼100 meV without any significant change to the valence band

level. Comparing the band alignment of GaNAsSb to GaInNAsSb, the resultant

structure is type-II. Even if a large error of >50 meV is assumed in the valence band

alignment between GaInNAsSb and GaNAsSb to make the structure type-I, hole

confinement would still be an issue as there would be very likely to have even one con-

fined state. This lack of confinement of holes was probably a major contributor to the

very large threshold current densities of previous GaInNAsSb/GaNAsSb/GaAs edge

emitting lasers [129]. From a heterojunction band alignment perspective, GaNAsSb

would not be a good choice of QW barrier material while GaAs and GaNAs would

be acceptable.

4.4 GaAs Barriers

While both GaNAs and GaNAsSb have been utilized as GaInNAsSb QW barrier

materials for laser devices, GaAs has not been employed as a GaInNAsSb QW

barrier material. Early GaInNAs devices used GaAs barriers, but GaNAs barri-

ers became the prevalent choice due to advantages such as strain compensation and

reduced nitrogen outdiffusion from the QWs. However, this nitrogen outdiffusion

from the GaInNAs(Sb) QW has been minimized with improved growth and strain

compensation may not be an issue for GaInNAs(Sb) SQW devices. In addition,

the large band offset would eliminate any carrier leakage which may exist with the

GaInNAsSb/GaNAs structure.

4.4. GaAs BARRIERS 105

GaNAs Barriers GaAs Barriers

GaAs GaAs

GaIn

NA

sS

b Q

W

GaIn

NA

sS

b Q

W

GaNAsSb Barriers

GaAs

GaIn

NA

sS

b Q

W

Figure 4.19: Band offset comparison of GaNAs, GaAs, and GaNAsSb. Both GaNAsand GaAs are type-I to GaInNAsSb, but GaNAsSb is possibly type-II.

To examine any issues with strain or composition, a 1.55 µm emitting, 8 nm

GaInNAsSb/GaAs QW was grown. Figure 4.20 shows an (004) ω/2θ HRXRD scan

of the sample. A well defined QW peak with a very high strain of 2.6% and many

Pendellosung fringes are present, making the presence of any relaxation a low prob-

ability. SIMS was also performed on this sample and the results are shown in Figure

4.21. The feature most apparent in the data is the mismatch of the top interface

of the indium and antimony depth profiles, similar to that found for GaNAsSb in

Figure 4.3. This was not an artifact of the measurement as it was repeated and sim-

ilar results were obtained. For growth of the GaInNAsSb QW, the antimony shutter

was opened ∼1 nm before the indium shutter since antimony is partially a surfactant

which tends to surface segregate. This allowed the bottom interfaces of antimony and

indium incorporation to match up. However, it continued to incorporate in slightly

higher amounts throughout the QW, from 0.7% to 1.2%. The antimony and indium

shutters were closed at the same time, but the antimony continued to incorporate

for ∼2 nm up to 1.5%. This thin GaAsSb layer would be detrimental to the per-

formance of a GaInNAsSb/GaAs device since it would introduce a region adjacent


31 32 33 34101

102

103

104

105

106

107

Cou

nts

(a.u

.)

/2 (Degrees)

GaAs

GaInNAsSb

Figure 4.20: (004) ω/2θ HRXRD of a GaInNAsSb SQW on GaAs with 2.6% latticestrain. Even without tensile barriers, the material remains structurally good.

to the QW with a type-II alignment. The holes could be trapped in this thin layer,

leading to hole leakage from the QW. This issue was resolved by closing the anti-

mony shutter ∼1–2 nm earlier than the indium shutter, leading to an alignment of

the top compositional profiles. However, different antimony fluxes and growth rates

would require additional calibration to determine the shuttering times, making this

a tedious process.

In addition to the difficulty of aligning the antimony incorporation with the QW,

utilizing GaAs barriers also inhibits the ability to grow MQW devices. At 2.6% strain,

attempts to grow a second or third QW were unsuccessful due to the lack of strain

compensation which would have been present with GaNAs barriers. Two or three

QW devices may be possible with lower strain, but this would limit the wavelength to

1.3–1.4 µm devices or necessitate the incorporation of significantly higher amounts

of nitrogen in the QW. Most groups have had difficulties obtaining high quality

material with more than 4–5% nitrogen. The lack of nitrogen in the barriers was also

thought to be an issue with dilute nitride QWs since nitrogen outdiffusion during ex

4.5. ANALYSIS OF QUANTUM WELL BARRIER CHOICES 107

50 60 70 80 90 100 110

0

10

20

30

40 In Sb

Sputter Depth (nm)

In M

ole

Frac

tion

(%)

0.0

0.5

1.0

1.5

Sb M

ole

Frac

tion

(%)

Figure 4.21: SIMS depth profile of a GaInNAsSb/GaAs SQW. The indium profiledefines the QW region. Antimony incorporation is found outside of the QW.

situ annealing was observed, further blueshifting the emission wavelength. However,

improved growth quality has eliminated much of this outdiffusion, although some

may remain. GaAs barriers for GaInNAsSb QWs have their practical difficulties,

but ideally appear to be an option for SQW devices.

4.5 Analysis of Quantum Well Barrier Choices

In choosing a barrier material for a QW, it is desirable to have certain key properties

which will enhance the structure without detrimental effects. The barriers must

have sufficient heterojunction band offset and the correct alignment. If the QW is

strained, it is important to strain compensate it with barriers of the opposite strain.

This will prevent relaxation and the formation of defects by effectively increasing the

critical thickness of the layers. Finally, the material must be of good quality. If it

is not, the defects present will lead to a higher rate of non-radiative recombination,

preventing carriers from recombining radiatively in the QW.


GaNAsSb was originally thought to be an improvement over the previous dom-

inant choice of GaInNAs(Sb) QW barrier of GaNAs since antimony had improved

GaInNAs material quality. However, this was not the case and GaNAsSb at similar

beam fluxes to GaInNAsSb growth suffered from degraded material quality compared

to GaNAs, as determined from RHEED and PL measurements. Lower intensities

generally indicate lower optical quality which is a result of non-radiative recombi-

nation traps. Carrier leakage from the QW to the GaNAsSb barriers becomes a

greater issue compared to GaNAs barriers. GaNAsSb provided no strain compensa-

tion for the highly compressive GaInNAsSb QWs, preventing the growth of MQW

structures necessary for more complex and high power devices. These QWs exceed

the critical thickness and require strain compensation for an active region containing

two or more QWs. GaInNAsSb/GaNAsSb also suffered from a valence band offset

which prevented hole confinement in the QW. By decreasing the beam flux of anti-

mony, the optical quality of GaNAs could be improved with a lower concentration

of antimony. A direct result of a lower antimony concentration is a smaller upward

lifting of the valence band such that the alloy would be type-I to the GaInNAsSb

QW, providing hole confinement. In addition, a step-like QW leads to longer wave-

length, assisting in the push to reach longer wavelengths. However, with smaller

amounts of antimony, the nitrogen enhancement factor is highly sensitive to the an-

timony concentration and any variation in composition would lead to a large change

in nitrogen concentration and thus band gap. Uniformity may become an issue with

a non-stable homogeneous antimony flux or a variation in antimony incorporation

through the layer.

GaNAs barriers have been employed in GaInNAs-based devices as well as the

higher-performance GaInNAsSb laser diodes. Using the same antimony flux as dur-

ing GaInNAsSb QW growth, GaNAs has better material and optical quality than

GaNAsSb. For typical GaInNAsSb QW growth rates, GaNAs has a tensile strain of

–0.5–0.8%, providing sufficient strain compensation for the compressive QWs. This

enables the growth of more complex MQW devices, such as VCSELs, without the dif-

ficulty of exceeding the critical thickness. The GaInNAsSb/GaNAs band alignment

is type-I and there are two and three confined states in the conduction and valence

4.5. ANALYSIS OF QUANTUM WELL BARRIER CHOICES 109

bands, respectively. Although it is a type-I structure, higher temperature operation

may become an issue due to thermal excitation of the carriers since only two states

exist in the conduction band. GaNAs quality is also realtively growth temperature

independent compared to GaNAsSb. This allows an additional parameter to be var-

ied, introducing additional freedom in the active region growth. Though GaNAs

has several advantages, it still does suffer from a non-trivial amount of non-radiative

defects, lowering the optical quality of the material. This again leads to additional

carrier leakage from the QW, reducing the performance of the device.

GaAs barriers were intially used in early GaInNAs QW lasers, but not cur-

rent GaInNAsSb devices. They provide no strain compensation to the compres-

sive GaInNAsSb QWs, limiting the number of QWs to at most two. The antimony

compositional profile is difficult to maintain and can lead to deleterious effects if

a thin GaAsSb layer is adjacent to the GaInNAsSb QW. Although antimony can

incorporate into a GaNAs layer above the QW as it does with a GaAs barrier, the

corresponding thin GaNAsSb layer’s band structure is not as adversely affected.

However, by adjusting the shuttering times with tedious calibrations, this layer can

be eliminated. GaAs barriers also increase the difficulty of pushing out to longer

wavelengths due to the shift in energy levels with increased confinement. However,

the large confinement does allow for four conduction and valence band states, leading

to better thermal stability. In addition, GaAs can be grown to produce much higher

optical and material quality than GaNAs, reducing non-radiative traps and carrier

leakage from the QW.

From the extensive analysis, GaNAsSb grown with the same antimony flux used

for the QW is not a viable option as a barrier material for GaInNAsSb QWs. Re-

duced antimony flux can improve material quality, but uniformity issues negate that

advantage. A valved antimony cracker may help, but incorporation irregularities

may still remain. GaAs barriers are ideal if only one QW needs to be grown. The

antimony compositional profile, while tedious to control, can be optimized. With

our current MBE system restraints, GaNAs barriers are overall the best choice for

GaInNAsSb QWs due to their strain compensation and moderately good material

quality. They are the only option for growth of MQW devices such as VCSELs. The


addition of a phosphorous cell would enable the growth of GaAsP barriers, which

have been proven in low-threshold GaInNAs devices [118, 119] and could be superior

to GaAs or GaNAs. A summary of all these results may be seen in Table 4.3.

4.6 Quantum Well Barrier Comparisons

An ultimate comparison of the three QW barrier materials GaAs, GaNAs, and

GaNAsSb is an examination of GaInNAsSb PL samples with each of the three barrier

materials. 8 nm GaInNAsSb QWs of nominally same composition were grown with

20 nm GaNAs or GaNAsSb barrier layers and 50 nm GaAs cap. The GaAs barrier

sample contained only the 50 nm GaAs cap. Figure 4.22 shows the PL spectra from

the three samples before annealing. The GaInNAsSb/GaNAs sample had the highest

PL intensity indicating the best overall optical quality of the structure. The samples

with GaAs and GaNAsSb barriers were of comparable intensity. The wavelength

shifts are due to the different amounts of confinement found within the GaInNAsSb

QW. Greater QW confinement leads to shorter wavelength emission. Examining the

FWHM, the GaNAsSb barrier sample had the largest value of 57 meV while the

other two samples were smaller at 42–43 meV. PL intensity and linewidth are good

indicators in general of overall material and optical quality. GaNAsSb has inferior

quality as a QW barrier material.

The annealing behavior for the GaAs and GaNAs barrier samples was also ex-

amined and the results shown in Figure 4.23. As grown, the PL intensity of the

GaNAs barrier sample is ∼1.5× higher than the GaAs barrier sample. Upon anneal-

ing, the GaNAs barrier sample improves to much higher intensities than the GaAs

barrier sample, ∼3.5× higher at the optimal annealing point. GaInNAsSb QWs

with GaNAs barriers result in consistently higher PL intensities compared to those

with GaAs barriers. The reduced intensity with GaAs barriers is possibly due to a

difficulty in accurately controlling the antimony concentration profile.

4.6. QUANTUM WELL BARRIER COMPARISONS 111

1400 1500 1600 17000.0

0.2

0.4

0.6

0.8

1.0

0.9 0.85 0.8 0.75

GaNAs GaNAsSb GaAs

PL In

tens

ity (a

.u.)

Wavelength (nm)

Energy (eV)

Figure 4.22: PL intensities of GaInNAsSb SQWs with GaAs, GaNAs, or GaNAsSbbarriers.

550 700 750 8000.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

PL In

tens

ity (a

.u.)

Anneal Temperature (°C)

GaAs Barriers

GaNAs Barriers

Figure 4.23: PL intensities of GaInNAsSb/GaAs and GaInNAsSb/GaNAs SQWs asa function of annealing temperature. All anneals were for 60 s.


4.7 Conclusion

The choice of material for the GaInNAsSb barriers plays a large role in the overall

active region quality. GaAs, GaNAs, and GaNAsSb are the three easily attainable

alloys by MBE for GaInNAsSb QWs. A summary of the findings are listed in Table

4.3. An extensive investigation on the material properties of GaNAsSb was conducted

as it was a relatively unknown material alloy. It was discovered GaNAsSb has poor

material quality under typical growth conditions compared to GaNAs, but can be

improved using smaller amounts of antimony. The heterojunction band offsets were

measured for a variety of dilute nitride antimonide structures. The difficulties of

utilizing GaAs barriers were also studied. For our current MBE system arrangement,

although GaAs barriers may be ideal, GaNAs barriers for GaInNAsSb QWs are the

optimal choice for long-wavelength devices. The addition of a phosphorous cell,

enabling GaAsP growth, could further enhance GaInNAsSb device quality.

4.7. CONCLUSION 113

Pros Cons

GaNAsSb • Small Sb GaNAsSb better

optical quality than GaNAs • Step-like QW leads to

longer wavelength

• Poorer material quality than

GaNAs • No strain compensation

• Band alignment issues

• Small Sb GaNAsSb highly sensitive to Sb%

GaNAs • Better quality than

GaNAsSb

• Strain compensation • OK band alignment

• Step-like QW leads to

longer wavelength

• Growth temperature independent

• Still has non-trivial amount

of non-radiative defects

• Carrier leakage

GaAs • No defects

• Good band alignment

• Hard to control Sb profile

• No strain compensation

• Increased confinement gives shorter wavelength

Table 4.3: Summary of QW barrier investigation findings. These materials are at-tainable in our current MBE system configuration.


Chapter 5

Effects and Role of Antimony on

GaInNAsSb

5.1 Improving GaInNAsSb Luminescence

at 1.3 µm

A majority of the efforts in developing and improving GaInNAsSb QWs have been

focused on obtaining high intensity 1.55 µm luminescence. Once it was shown the

addition of antimony improved GaInNAs at 1.3 µm [74, 130], little was done to

examine the extent to which the luminescence could be increased at this wavelength.

If improved sufficiently, the GaInNAsSb alloy would be the preferred solution for

GaAs-based optoelectronics operating between 1.3 and 1.6 µm.

Adding increasing amounts of nitrogen to the dilute nitrides while maintaining

material quality has been a multifaceted issue. Once thought to be a “nitrogen

penalty,” this “nitrogen complexity” can be overcome with extensive optimization

of the growth parameter space and improvements of the rf plasma operation. How-

ever, the reduction of nitrogen in dilute nitrides, to first-order, can improve material

quality.

Antimony has allowed for higher concentrations of indium in GaInNAs without

suffering from relaxation or compositional segregation [61, 74, 84]. Adding antimony

115

116 CHAPTER 5. EFFECTS AND ROLE OF ANTIMONY ON GaInNAsSb

to optimized GaInNAs at 1.3 µm while holding the growth conditions constant has

led to a dramatic improvement in optical quality. In hopes of further increasing

the PL intensity at the same wavelength, the nitrogen concentration was reduced

significantly while increasing the indium concentration and antimony flux.

A sample with a Ga0.673In0.327N0.016As0.964Sb0.02 SQW (composition measured by

SIMS) and GaNAs barriers had the highest PL intensity for dilute nitrides at 1.3

µm from our group. In a different GaInNAsSb/GaNAs/GaAs sample, the nitrogen

concentration was reduced from 1.6% to 0.8% while the indium was increased to

∼34% and antimony flux from 1.0×10−7 to 1.5×10−7 Torr BEP. The exact concen-

trations for indium and antimony are not known since SIMS was not performed on

these samples. Nitrogen concentrations were extrapolated from the GaNAs barrier

compositions. Although the antimony flux was increased, from previous calibrations,

the concentration is not expected to increase more than a few tenths of a percent.

Figure 5.1 shows the (004) ω/2θ scans of these two samples. Both samples do

not appear to have any structural quality issues. The top spectrum represents the

Ga0.673In0.327N0.016As0.964Sb0.02 QW and has an in-plane strain of 2.1%. By decreasing

the nitrogen concentration and increasing the indium concentration and antimony

flux, the GaInNAsSb QW has significantly increased its in-plane strain to 2.5%.

In Figure 5.2, the effect of reducing the nitrogen content is illustrated by the

dramatic increase in PL intensities for all anneal temperatures. Both QWs emitted

light in the 1.25–1.35 µm range, depending on the amount of blueshift from annealing.

The as-grown luminescence for the reduced nitrogen sample already equaled the

peak luminescence from the previously “best” GaInNAsSb material after annealing.

This was an amazing result. At the optimal annealing temperature for the reduced

nitrogen sample, the PL intensity was found to be 6.5× higher. Even with anneal

temperatures past the optimal anneal, the PL intensity remained much higher than

the GaInNAsSb sample with higher nitrogen content.

A reduction in point defects related to nitrogen incorporation, such as N/N or

N/As split interstitials and interstitial nitrogen, is probably the reason for the dra-

matic increase in optical quality. While this would seem to indicate a breakthrough

5.1. IMPROVING GaInNAsSb LUMINESCENCE AT 1.3 µm 117

30.5 31.0 31.5 32.0 32.5 33.0 33.5 34.0

1

10

100

1k

10k

100k

1M

10M Typical N, In, Sb Decrease N,

increase In, Sb

Cou

nts

(a.u.)

/2 (Degrees)

Increase

GaInNAsSb QW

GaAsSubstrate

GaNAsBarriers

Figure 5.1: HRXRD spectra of (004) ω/2θ scans of GaInNAsSb/GaNAs QWs withdifferent concentrations of nitrogen, indium, and antimony, but the same 1.3 µmemission wavelength.

550 650 700 750 800 8500

1

2

3

4

5

6

7

PL

Inte

nsity

(a.u

.)

Anneal Temp (C)

Decrease N Typical

N, In, Sb

Figure 5.2: Annealing behavior on the PL intensity of the GaInNAsSb/GaNAs QWs.The lower nitrogen concentration sample has much higher intensity.


for GaInNAsSb devices at 1.3 µm, we decided not continue with the process. In addi-

tion to practical difficulties with time and wafer shortages, there was a possibility the

device would not have good performance characteristics. As will be discussed later in

this chapter, low optimal anneal temperatures are not desired for laser active regions

[36]. The optimal anneal temperature for the reduced nitrogen GaInNAsSb QW was

720◦C. GaInNAsSb QWs with higher nitrogen concentrations have optimal anneal

temperatures ranging from 760–800◦C. Lower optimal anneal temperatures appear

to be correlated with GaInNAsSb QWs containing higher lattice strain. However, a

1.3 µm GaInNAsSb device with reduced nitrogen still holds much promise in having

superior performance characteristics.

5.2 Indium Concentration and Strain Effects on

Antimony

GaInNAs(Sb) has been heavily investigated as the material to replace InP-based

lasers operating at wavelengths between 1.3–1.6 µm. GaInNAs has also garnered

interest for high-efficiency solar cell junctions operating at 1.0 eV (∼1.2 µm) [131–

138]. With concentrations of 6–8% indium and 2–3% nitrogen, GaInNAs has a band

gap corresponding to a 1.0 eV band gap with little or no strain when grown on GaAs.

The ability to grow thick coherent layers (≥ 1 µm) is extremely important for high-

quality absorption-based devices, such as solar cells and detectors. Before GaInNAs,

GaAs-based devices requiring thick layers of 1.0 eV band gap material necessitated

techniques such as graded strain relaxation layers or wafer bonding which introduced

performance degrading defects. However, there have been challenges in implementing

GaInNAs for use in solar cells [133, 134]. High acceptor and defect concentrations

reduce carrier mobility and lifetime, decreasing the efficiency of a solar cell device.

These concentrations must be reduced before they become viable 1.0 eV junctions.

Adding antimony has helped improve GaInNAs designed for long-wavelength op-

toelectronics, but it has never been studied with compositions used for solar cell

devices. It is possible that antimony can improve the optical quality and/or defect

5.2. INDIUM CONC. AND STRAIN EFFECTS ON ANTIMONY 119

concentrations in GaInNAs with lower indium concentrations. However, much is still

unknown about its effects and roles with GaInNAs. Adding antimony has made an

already complex alloy even more complicated, forming a five-element quinary sys-

tem. Growth interactions between indium, nitrogen, and antimony are undoubtedly

present. Although antimony improved GaInNAs material quality and surface mor-

phology, it behaved much differently with indium-free GaNAs material, as described

in Chapter 4. This may be due to the nature of antimony incorporation and the

effects of adding indium. It is believed that a reactive surfactant, such as antimony

on GaAs, behaves differently in low-strain versus high-strain materials. The driving

force for roughening is much lower in low mismatch systems than in high mismatch

systems and the presence of the reactive surfactant is not required to prevent rough-

ening. This may lead to a difference in observed properties when utilizing antimony

in the high-strain GaInNAsSb alloys for laser applications as opposed to the low-

strain alloys for 1.0 eV solar cell applications.

To determine the effects and behavior of widely varying compositions on the

structural, optical, and electronic properties of GaInNAsSb, several samples were

grown at a variety of indium and antimony fluxes. A summary of the samples

described below is shown in Table 5.1. The first series consisted of GaInNAs(Sb) QWs

intended to contain 32% indium and 2.0% nitrogen, a typical composition for 1.3 µm

wavelength emission. These samples are considered to be in the “high indium” and

high lattice strain regime. The antimony flux was varied from zero to 1.0×10−7 Torr

BEP while all other growth parameters were held constant. 1.0×10−7 BEP Torr is

the typical flux we have used for all GaInNAsSb laser devices [114, 115, 129, 139, 140].

Next, a series of GaInNAs(Sb) QWs grown with much lower indium, 8%, and 2.0%

nitrogen were analyzed for their properties with varying amounts of antimony. These

“low indium” and low lattice strain samples correspond to the composition typically

used to obtain 1.0 eV band gap for solar cell applications. Finally a set of samples

with constant 1.0×10−7 Torr BEP antimony flux and varying indium concentrations

from the “low” to “high” indium regimes were studied to connect the properties of

the two previous series into a comprehensive understanding on the role of antimony

in GaInNAsSb.


8% In 16% In 24% In 32% In

0 Sb B A2.0×10−8 Torr BEP Sb B A6.0×10−8 Torr BEP Sb B A1.0×10−7 Torr BEP Sb B, C C C A, C

Table 5.1: A summary of the growth conditions for the samples described in thisstudy. The intended indium composition and the applied antimony fluxes are listed.A) Varying antimony under constant “high” indium flux, B) varying antimony underconstant “low” indium flux, C) varying indium under constant 1.0×10−7 BEP Torrantimony flux.

The GaInNAs(Sb) SQWs were all grown at a substrate temperature of 440◦C

measured by pyrometry. An arsenic-to-gallium overpressure of 20× and an anti-

mony flux of 0.2–1.0×10−7 Torr BEP were supplied during the GaInNAsSb QW

growth. The total group-III flux (gallium and indium) was held constant such that

the nitrogen composition nominally remained the same for all samples in this series.

The structure for all samples consists of a 7.5 nm GaInNAs(Sb) QW grown on a 300

nm GaAs buffer capped by a 50 nm GaAs layer. The compositions of the samples

were determined by HRXRD and SIMS. Room-temperature PL measurements were

used to obtain emission wavelength and luminescent intensity.

5.2.1 Antimony variation with high indium GaInNAs(Sb)

The first series of samples studied contained a single GaInNAs(Sb) QW with nom-

inally 32% indium and antimony fluxes which varied between 0 and 1.0×10−7 Torr

BEP. These QWs are typically used for 1.3 µm light emitters and have relatively

high amounts of indium. The strain in the QWs is large and the thickness is past

the critical thickness, but does not relax. The calculated Matthews-Blakeslee crit-

ical thickness is 35–45 Awhile the actual thickness exceeds this value by 30-35 A.

Figure 5.3 shows the HRXRD spectra from these four samples. As will be discussed

later, all of the QWs contained 34% indium rather than 32% indium due to flux


miscalibration. The top spectra in Figure 5.3 represents a GaInNAs QW which is of

poor quality. The QW Pendellosung fringes are severely degraded, peak intensity is

very low, and strain is much lower than expected. This indicates a loss of structural

quality, possibly due to surface roughening, compositional segregation, or a change in

incorporation kinetics. A RSM did not indicate any relaxation had occurred. It has

been a challenge to obtain high quality GaInNAs with indium compositions greater

than 34–35% due to compositional segregation and relaxation and this difficulty is

apparent in this sample. Evidence of compositional segregation may be found in

Reference [74]. The bottom three spectra in Figure 5.3 are from GaInNAsSb QWs

which have recovered their structural quality in the presence of an antimony flux.

The three spectra, with +2.1% strain, are almost identical except for a very small

compressive shift of the QW with increasing antimony flux. The origin of this small

shift requires a detailed compositional analysis in order to accurately determine the

cause. For this compositional range of high indium and high strain, the addition of

antimony greatly improved the structural quality of the GaInNAs. However, from

HRXRD, no significant difference in strain or structural quality of the GaInNAsSb

QWs with different antimony fluxes could be detected.

SIMS was used to determine the composition of the four samples in this series.

Although it is relatively straight forward to obtain depth profiles, obtaining exact

compositional values requires previous calibration due to artifacts, including ma-

trix effects. The SIMS analysis was calibrated using parameters obtained from past

growths analyzed with nuclear reaction analysis Rutherford backscattering (NRA-

RBS) for nitrogen and particle-induced X-ray emission RBS (PIXE-RBS) for anti-

mony [74]. The compositions were consistent with HRXRD data and simulations.

Figure 5.4 shows the indium, nitrogen, and antimony compositions as a function of

the antimony flux. It is unclear if the data from the antimony-free GaInNAs sample

is believable since the QW itself suffered severe material degradation, thus changing

the bonding structure and possibly the incorporation kinetics of the alloy. These

changes may alter SIMS sputtering statistics. The indium concentration is much

lower than anticipated compared to previous growths which were of good quality.

Nitrogen is also higher than expected compared to past growths than the other three


30.5 31.0 31.5 32.0 32.5 33.0 33.5 34.0 34.510

100

1k

10k

100k

1M

10M

100M

1G

10G

100G

0 Sb 2x10-8 Torr Sb 6x10-8 Torr Sb 1x10-7 Torr Sb

Cou

nts

(a.u.)

/2 Degrees

IncreasingAntimony

Figure 5.3: HRXRD spectra of the (004) GaInNAs(Sb)/GaAs QWs with “high”indium compositions.

samples which have antimony since antimony enhances nitrogen incorporation. As

expected, for increasing antimony fluxes, the antimony concentration in the QW rose

from 0.5% to 2.0%. Indium was ∼34% (higher than intended due to flux miscali-

bration) while nitrogen was 2.3–2.4%. There was not much change in the indium or

nitrogen concentrations with varying antimony flux.

Room temperature PL measurements were performed to study the optical prop-

erties of the GaInNAs(Sb) QWs. Figure 5.5 shows the PL intensity of the four QWs

after an ex situ RTA. The relative peak intensities of the samples were all com-

parable before and after RTA. Additional details on the annealing behavior of the

samples may be found later in this chapter. As expected, the GaInNAs QW that

showed poor structural quality in HRXRD produced weak emission centered at 1.32

µm. The addition of a small antimony flux of 2.0×10−8 Torr BEP during the QW

growth dramatically improved the optical quality. Adding 6.0×10−8 Torr BEP fur-

ther increased the PL intensity, but adding the typical 1.0×10−7 Torr BEP used in


0 2x10-8 4x10-8 6x10-8 8x10-8 1x10-730

31

32

33

34

35

36

Indium Nitrogen Antimony

Sb Flux BEP (Torr)

Indi

um (%

)

0.0

0.5

1.0

1.5

2.0

2.5

Nitr

ogen

, Ant

imon

y (%

)

Figure 5.4: Indium, nitrogen, and antimony compositions as a function of antimonyflux.

GaInNAsSb laser growths actually decreased the optical quality. This indicates there

is an optimal antimony flux which produces the highest optical quality GaInNAsSb

QWs. A red shift of the peak wavelength was seen due to the increasing antimony

incorporation.

5.2.2 Antimony variation with low indium GaInNAs(Sb)

The next series of samples contained relatively low amounts of indium (8%) and

antimony fluxes which varied between 0 and 1.0×10−7 Torr BEP. GaInNAs in this

compositional range is used for solar cell junctions which operate at 1.0 eV. Since

thick layers are required for high-efficiency solar cells, the strain must be very small

or non-existent and thus contain a much smaller indium composition for the same

nitrogen concentration. Since the strain in these QW samples is very small, thicker

1000 A samples of identical composition were grown to facilitate strain determination

with HRXRD. The HRXRD spectra of these thicker samples are shown in Figure 5.6.

The HRXRD spectra indicated the samples possessed excellent structural quality.

The oscillations appear to be more strongly damped in samples with high antimony


1.2 1.3 1.4

0.0

0.2

0.4

0.6

0.8

1.0

1.05 1 0.95 0.9

Inte

nsity

(a.u.)

Wavelength ( m)

0 Sb 2x10-8 Sb 6x10-8 Sb 1x10-7 Sb

Energy (eV)

Figure 5.5: PL spectra of GaInNAs(Sb) samples under high indium, high strainconditions with varying antimony flux.

flux, suggesting that the interface quality is not as good as in samples with little or no

antimony flux. In the antimony-free case, the GaInNAs layer had a strain of +0.24%.

The strain increases with larger antimony fluxes, up to +0.54% with 1.0×10−7 Torr

BEP. Even with the slight increase in strain, these values are significantly lower than

the highly-strained (2.0–2.6%) QWs which are used for 1.3 and 1.55 µm wavelength

emission. Additional information from SIMS was required to determine the origin of

the increase in compressive strain with increasing antimony fluxes.

The compositions of the four samples are shown in Figure 5.7. As expected,

an increase in antimony flux used during the QW growth led to significantly higher

antimony incorporation, up to 5.5%. This value is much higher than the high indium

case, where there was only 2.0% incorporation at the highest antimony flux. Nitrogen

content was found to be 2.1% in the case without antimony and 2.8% in the presence

of antimony. This enhancement of nitrogen incorporation was discussed in Chapter 4.

In the antimony-free case, the indium composition was found to be 10.5%. However,

when 2.0×10−8 Torr BEP antimony was applied, the composition dropped to 9.4%,


32.0 32.5 33.0 33.5 34.0

1

100

10k

1M

100M

10G

1T

1x10-7 Torr Sb

6x10-8 Torr Sb

2x10-8 Torr Sb

0 Torr Sb

Cou

nts

(a.u

.)

/2 (Degrees)

GaAs Substrate

GaInNAs(Sb) Film

Figure 5.6: HRXRD spectra of the (004) GaInNAs(Sb)/GaAs layers with “low”indium compositions.

even though the indium flux was held constant. The indium concentration decreased

further to 8.7% at the highest antimony flux. This finding was surprising since it

was unexpected that antimony would affect the incorporation kinetics of indium.

The ultimate goal of this series of samples was to determine whether antimony

would improve the material and optical quality of GaInNAs in the low indium (low

strain) conditions as it did in the high indium (high strain) conditions. Figure

5.8 shows the PL spectra from these samples after ex situ RTA. Adding antimony

degraded the optical quality of the GaInNAs; the higher the antimony flux, the lower

the PL intensity. In addition, the FWHM of the PL peaks also increased with larger

antimony fluxes. Also, the peak wavelength red shifted with increasing antimony

concentration, despite the decrease in indium content. Antimony lowers the band

gap much more rapidly than indium in this low concentration regime. The decrease in

PL intensity is a very surprising result since this is the opposite behavior compared to

samples with high indium, and high strain, in which antimony dramatically improved

the optical quality of GaInNAs. It may be argued that the antimony-containing


0 2x10-8 4x10-8 6x10-8 8x10-8 1x10-7

0123456789

1011

Mol

e Fr

actio

n (%

)

Sb Flux BEP (Torr)

Indium Nitrogen Antimony

Figure 5.7: Indium, nitrogen, and antimony compositions as a function of antimonyflux utilized during the QW growth in the “low” indium composition range.

samples have lower PL intensities due to higher nitrogen content, degradation typical

of the dilute nitrides. However, the three samples with antimony contain almost the

same amount of nitrogen and thus the continuing degradation cannot be explained in

that manner. Finally, it is possible that a smaller antimony flux than 2.0×10−8 Torr

BEP may improve the material quality as it was found with GaNAsSb in Chapter 4.

The behavior is nonetheless very different than the high indium, high strain samples.

5.2.3 Indium variation with constant antimony flux

Finally, in an effort to connect the behaviors observed in the previous two studies, a

series of samples was grown with a constant 1.0×10−7 Torr BEP of antimony while

adjusting the indium concentration. Figure 5.9 shows the HRXRD spectra taken

from the four samples with varying precalibrated indium concentrations. All the

samples have a well defined QW peak and strong Pendellosung fringes, indicating

good structural quality. A large change in strain was also observed with the addition


1.0 1.1 1.2 1.3

0.0

0.2

0.4

0.6

0.8

1.0

1.2 1.15 1.1 1.05 1 0.95

1e-7 Torr Sb6e-8 Torr Sb

2e-8 Torr Sb

0 Torr Sb

Inte

nsity

(a.u.)

Wavelength ( m)

Energy (eV)

Figure 5.8: PL spectra of GaInNAs(Sb) samples under low indium, low strain con-ditions with varying antimony flux.

of more indium as expected. The strain varied from +0.7% at the low indium com-

position to +2.0% in the high indium composition. These values are very similar to

those observed in the previous two studies.

SIMS and HRXRD were employed to determine the compositions of the samples

in this series. A summary of the data is shown in Figure 5.10. A sample containing no

indium (GaNAsSb) grown under nearly identical growth conditions was included in

the plot for additional comparison. The observed indium concentration in the QWs

matched well with the intended compositions, determined from previous calibrations.

The nitrogen composition for all five indium compositions remained constant at

2.5%. The total group-III growth rate (indium and gallium) and antimony flux were

held constant and thus the relation in Equation 2.1 dictates a constant nitrogen

composition as well. Different indium compositions were obtained by changing the

ratio of indium and gallium fluxes. 11% antimony was found in the indium-free

GaNAsSb QW sample. Interestingly, when indium was present at low composition,

the antimony dropped to 7.3% in the GaInNAsSb QW. The antimony concentration


30.5 31.0 31.5 32.0 32.5 33.0 33.5 34.0 34.510

100

1k

10k

100k

1M

10M

100M

1G

10G

100G

32% In

24% In

16% In

8% In

Cou

nts

(a.u

.)

/2 (Degrees)

Figure 5.9: HRXRD spectra of the (004) GaInNAs(Sb)/GaAs QWs with varyingindium fluxes under a constant antimony flux.

continued to decrease with increasing indium fluxes down to 0.8% antimony at high

indium composition. This indicates a very strong interplay of strain, resulting in

local competition between the indium and antimony atoms during growth.

This investigation varying the indium composition under a constant antimony

flux connects the improvement in optical quality with antimony at high indium and

the degradation with antimony at low indium. The PL spectra after ex situ RTA

are shown in Figure 5.11. For the sample with only 8.8% indium, the peak intensity

was weak, similar to that found in the study with low indium. As evident from

the plot, increasing the indium concentration (and also strain) in the GaInNAsSb

QW dramatically improved the optical quality. A red shift in the peak wavelength

can be attributed to the large increase in indium concentration. The shift in peak

wavelength between the 8.8% and 16.7% indium samples was very small. Indium does

not decrease the band gap as rapidly as antimony for those specific compositions.

From this study, low concentrations of antimony improve the material quality when

there is a significant amount of indium and/or strain in the GaInNAsSb QW.


0 8 16 24 320

2

4

6

8

10

12 Nitrogen Antimony

Mol

e Fr

actio

n (%

)

Intended In %

Figure 5.10: Nitrogen and antimony compositions as a function of indium concen-tration with the antimony flux held constant.

1.1 1.2 1.3 1.4

0.0

0.2

0.4

0.6

0.8

1.0

1.1 1.05 1 0.95 0.9

32% In

24% In

16% In

8% In

Inte

nsity

(a.u

.)

Wavelength ( m)

Energy (eV)

Figure 5.11: PL spectra of GaInNAs(Sb) samples with a constant antimony flux withvarying indium concentrations.


5.2.4 Antimony as a reactive surfactant

When antimony was first added to GaInNAs to improve the material quality for

1.3 and 1.55 µm, edge-emitting lasers and VCSELs, little attention was given to the

amount used, in what situations it would be beneficial, and especially how it worked.

It was thought that antimony was a “panacea” to improve the quality of all dilute

nitrides. As shown in previous studies of GaNAsSb presented in Chapter 4 and this

study on GaInNAsSb, it has become apparent that antimony is not a “magical cure”

to improve dilute nitrides and must be utilized correctly for different devices with

different compositions.

From the studies presented in this chapter, there are several factors that con-

tribute to the improvement or degradation of GaInNAs with antimony. The typical

definition of a surfactant is a species that lowers the surface free energy of the growth

front. However, it was later realized that the modification of the epitaxial growth

kinetics was the key effect of surfactants [65]. In two papers by Massies et al. [65]

and Tournie et al. [66], they proposed two classes of surfactants which were utilized

in epitaxial growth: reactive surfactants that decrease the surface diffusion length

(SDL) and non-reactive surfactants that increase the SDL at the growth front. For

homoepitaxy or non-strained heteroepitaxy, a non-reactive surfactant is preferred

since increasing the SDL would improve material quality. However, for strained het-

eroepitaxy in which there is a significant lattice mismatch between the layer and

substrate, minimizing the SDL would be beneficial to reduce the formation of is-

lands and compositional segregation, which are thermodynamically favored. Thus,

reactive surfactants are utilized for strained heteroepitaxial growth.

The issue of non-strained epitaxy versus strained epitaxy is particularly relevant

when comparing GaInNAs used for 1 eV solar cell applications and GaInNAs used in

long-wavelength optoelectronics. For the samples with high indium, and thus high

strain, adding antimony improved the structural and optical quality of GaInNAs.

However, the addition of any amount of antimony to the low indium, low strain

GaInNAs degraded the optical quality. It was also seen that increasing the strain in

the QW by increasing the indium concentration while applying identical antimony

fluxes helped improve the PL intensity of the GaInNAsSb. The reduction of the SDL


in the high indium and high strain case is important in minimizing the formation

of islands and 3D growth due to the high strain and tendency to phase segregate.

Antimony keeps GaInNAs growth 2D and reduces or eliminates compositional seg-

regation [61], improving the optical quality. However, for the thick lattice matched

layers used in GaInNAs solar cells, the strain is minimal or zero and a reactive sur-

factant is undesirable. Suppression of the SDL in this case leads to a high density of

defects because there is no need to prevent strain-induced islanding. TEM would be

useful in providing additional data to this conclusion.

5.2.5 Growth interactions between antimony and indium

Interplay or competition between indium and antimony when incorporating into

GaInNAs is another factor affecting the quality. In the high indium, high strain

case, it was unclear whether the presence of antimony led to a change in indium

concentration since the antimony-free sample was of poor quality. Ignoring the first

sample with poor material quality, there was no observed change in indium concen-

tration with increasing antimony fluxes. It is possible with such small percentages

of antimony incorporation (≤2%), any effect on indium incorporation would not be

detectable. A noticeable change was observed in the low indium, low strain case.

By introducing a small antimony flux (V/III BEP=0.07), the indium concentration

decreased. It continued to decrease with larger antimony fluxes. The ease of an-

timony incorporation with low indium concentrations was evident since there was

5.7% antimony compared to 2.0% antimony with high indium concentrations for the

same maximum flux of antimony and group-III growth rate. For the samples with

constant 1.0×10−7 Torr BEP of antimony and varying indium, the competition be-

tween indium and antimony was apparent. Without indium and at very low strain

conditions, the antimony incorporated at 11% concentration. However, as the indium

and strain increased, the antimony concentration steadily decreased to 0.8%.

Additional investigation is required to determine the exact cause of this compe-

tition, but it is suspected that the local strain associated with the large atomic radii

of antimony and indium in GaAs play a major role. Both atoms induce a large local


strain in the GaAs matrix and incorporating both atoms would not be energetically

favorable due to the large local strain energies they would create. Thus, one species

would preferentially incorporate while the other does not. However, the fact that

indium is a group-III atom and antimony is a group-V atom adds complexity to this

argument as it is not a direct location site competition. In the low indium, low strain

case, there was a larger initial decrease in indium concentration with the introduc-

tion of antimony while there was no such reduction when indium was introduced to

GaNAsSb in the study varying indium. The decrease in antimony from 11% to 0.8%

is most likely caused by a significant increase in indium from 0% to 32.7%.

5.2.6 Minimization of antimony incorporation for improved

luminescence

Minimizing antimony incorporation leads to the best optical quality in GaInNAsSb.

This is the common trend observed in all three studies. It was also observed in

Chapter 4 with GaNAsSb in which dilute antimony incorporation increased optical

quality. Although a surfactant is technically defined as a surface segregated species

that does not incorporate, a reactive surfactant does bond substitutionally [65]. The

surfactant atom then exchanges places with an incoming adatom and continues to

segregate to the surface in that manner. However, it does bond to the matrix and cou-

pled with the fact that dilute nitride alloys are grown at relatively low temperatures,

the antimony atom does not always continue to surface segregate and incorporates.

In the high indium, high strain case, adding antimony up to 1% mole fraction

improved the optical quality while 2% antimony degraded it. For the low indium,

low strain case, the behavior was much different. Adding antimony, even at 1.3%

mole fraction, decreased the PL intensity from the antimony free case. In the in-

vestigation where the indium concentration was varied, although the antimony flux

itself was not varied, the antimony concentration dropped with increasing indium

concentrations. The sample with the lowest antimony concentration of 0.8% had the

highest PL intensity. The amount of antimony flux utilized during the QW growth is

not the key parameter since many factors must be considered including composition

5.3. ANNEALING BEHAVIOR AND LATTICE STRAIN 133

and strain. The PL intensity improved by increasing the indium concentration, forc-

ing the antimony concentration to decrease indirectly. An antimony flux is required

in the high strain case as it is needed to improve the material and optical quality

with antimony’s reactive surfactant qualities in dilute nitrides. However, either no

antimony or a flux much smaller than those studied in this investigation is desired for

the low strain samples since reactive surfactants are not helpful in non-strained het-

eroepitaxy. In the end, it is the amount of antimony incorporated in the GaInNAsSb

which must be given the most attention.

5.3 Annealing Behavior and Lattice Strain

Annealing of dilute nitride material is a procedure performed to obtain improved

luminescence. As described in Chapter 2, the PL intensity increases with hotter an-

neal temperatures until an optimal annealing temperature. Temperatures exceeding

the optimal point lead to lower intensities, but still higher than that for as-grown

material. Depending on conditions such as the layer structure and length of anneal,

the optimal annealing temperature typically ranges from 720–820◦C. This optimal

annealing temperature has been a curious parameter as no one in the dilute nitride

community has been able to make much sense of it. Most groups, including ourselves,

perform an annealing study, locate the optimal anneal temperature, and anneal the

lasers at that temperature so that they may have the best performance characteris-

tics. The optimal anneal temperature was never considered as an important metric

during material characterization.

The GaInNAsSb QWs with varying antimony and indium concentrations were

annealed at a range of temperatures for one minute. Figure 5.12 shows the PL

intensities obtained at each annealing temperature. As expected, a reduction in

defects related to low-temperature and plasma growth occured after annealing and

the intensities of the GaInNAsSb QWs increased dramatically. In Figure 5.12(a), it

is seen that the 0% antimony sample had the highest PL intensity with an optimal

anneal temperature ∼850◦C. (The data point at 840◦C is probably the result of an

error with the PID parameters of the RTA apparatus.) With increasing antimony


550 750 800 850 9000.00

0.05

0.10

0.15

0.20

0.25 0% Sb 1.3% Sb 3.7% Sb 5.4% Sb

PL In

tens

ity (a

.u.)

Anneal Temp (C)

Increasing Sb,

(a) Varying antimony concentration

550 700 750 800 850

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7 8% In 16% In 24% In 32% In

PL In

tens

ity (a

.u.)

Anneal Temp (C)

Increasing In,

(b) Varying indium concentration

Figure 5.12: Reduction of the optimal anneal temperature with increasingly strainedGaInNAsSb QWs due to larger (a) antimony or (b) indium concentrations.

concentrations, the optimal anneal temperature shifts downwards to 820◦C with

5.4% antimony. In the annealing study with varying indium concentrations as seen

in Figure 5.12(b), the optimal anneal temperature of the 8% indium sample could

not be determined due to a lack of wafer material, but appears to be >850◦C. With

increasing indium concentrations, the PL intensity increases and the optimal anneal

temperature shifts downwards to 760◦C with 32% indium.

The results from these GaInNAsSb QWs indicate a clear relationship between

the in-plane strain and the optimal anneal temperature. In both cases, the opti-

mal anneal temperature decreased with increasing lattice strain. The smaller 30◦C

decrease in the varying antimony study compared to the >90◦C decrease in the

varying indium study is probably related to the magnitude of change in composi-

tion (∆Sb=5.4% for antimony and ∆In=24% for indium) and strain (∆ε=0.3% for

antimony and ∆ε=1.3% for indium).

The relation between strain and optimal anneal temperature is also confirmed in

GaInNAsSb EEL growth by Bank et al. at 1.55 µm [36]. In an effort to push from

1.49 µm to 1.55 µm emission, the indium and antimony concentration was increased

while holding nitrogen constant. Nitrogen was not increased to avoid the “nitrogen

5.4. CONCLUSION 135

complexity.” The lasers fabricated utilizing these GaInNAsSb QWs suffered from

relatively high threshold current densities, an indication of high concentrations of

defects. In a second effort, the antimony concentration was held constant while in-

creasing the nitrogen concentration and slightly decreasing the indium concentration.

Lasers using the higher nitrogen and lower indium content QWs had world-record

low threshold current densities for 1.55 µm GaAs-based lasers. For both sets of QWs,

the PL intensities (normally the best indicator of laser performance) at the optimal

anneal temperature were comparable, yet the lasers were drastically different. The

GaInNAsSb QWs with more indium and antimony (and strain) had a much lower op-

timal anneal temperature compared to those with higher nitrogen and lower indium

content and lower strain. The lower optimal anneal temperature indicates the mate-

rial has a lower thermal budget during high-temperature growth of the top cladding

layer. In effect, the QW material becomes over-annealed, leading to degraded device

performance.

The physical cause of the decrease in optimal anneal temperature with strain

is unknown and requires further study. It is suspected that dilute nitride materi-

als which are highly strained have lower activation energies for defect formation or

propagation. Lower temperatures are required to provide sufficient thermal energy

to initiate material degradation. The type of defect is also unclear, although com-

positional segregation or misfit dislocation formation are the main suspects and can

be analyzed using TEM.

5.4 Conclusion

The quinary GaInNAsSb is a very complex alloy. Many different growth parameters

can affect one another, but a systematic study can lead to some important insights. A

reduction in nitrogen content in GaInNAsSb at 1.3 µm lead to amazing PL intensities

compared to previously “optimal” material. In examining varying antimony and

indium concentrations in GaInNAsSb, a competition for incorporation was found,

possibly due to the local lattice strain each atom creates. Antimony was found to be

detrimental to GaInNAs material with low concentrations of indium and low lattice


strain, contrary to the behavior of alloys with high indium and high strain. For

strained materials, it was found minimization of antimony incorporation led to the

highest optical quality material. The optimal anneal temperature, once thought to

be a trivial parameter, has been connected to the strain found in the material and

is an important quantity in GaInNAsSb laser growth.

Chapter 6

Long Wavelength Semiconductor

Lasers

In this chapter, a brief overview of the laser devices utilizing GaInNAsSb QWs with

GaNAs barriers is presented. A dramatic improvement in the EELs and the demon-

stration of VCSELs at 1.55 µm has been due to our greater understanding of the

growth methods and resultant physical properties of dilute nitride antimonides. Most

of the laser device work which is presented was performed by my fellow colleagues on

the dilute nitride project. The world-record low-threshold GaInNAsSb EEL work was

performed primarily by S. R. Bank. Demonstration of the first electrically pumped

1.55 µm VCSELs on GaAs was due to the efforts of M. A. Wistey. Additional details

and discussion may be found in their respective doctoral dissertations [35, 36].

6.1 Low-Threshold GaInNAsSb QW Edge Emit-

ting Lasers

Previous GaInNAsSb EEL devices at 1.50-1.55 µm suffered from high threshold

current densities (≥1 kA/cm2) or were unable to operate under cw conditions [116,

129, 141]. In order to be feasible for telecommunication systems and be advantageous

over InP-based devices, these devices must have high power, low threshold current

137

138 CHAPTER 6. LONG WAVELENGTH SEMICONDUCTOR LASERS

densities, and the ability to operate cw. Through improvements in the active region

by improving the QW and barrier material quality and optimization of various growth

techniques, such as growth temperature and plasma operation, we have developed

lasers which have low threshold current densities, high power, and the ability to

operate in cw conditions at 1.50-1.55 µm.

Separate confinement heterojunction EELs were grown on (100) n-type GaAs

substrates. An example structure may be seen in Figure 1.1. The active region

for these lasers consisted of a single 7.5 nm GaInNAsSb QW surrounded by 22 nm

GaNAs barriers. An AlGaAs/GaAs waveguide surrounds the SQW active region.

The active region was symmetrically embedded in an undoped GaAs layer 460 nm in

thickness. 1.8 µm of Al0.33Ga0.67As formed the bottom n-type cladding. The lower

900 nm was silicon doped at 3×1018 cm−3 and the upper 900 nm at 7×1017 cm−3.

The top p-type cladding was also 1.8 µm of Al0.33Ga0.67As with the lower 900 nm

carbon doped at 7×1017 cm−3 and the upper 900 nm at 3×1018 cm−3. A 50 nm

p-type GaAs contact layer doped at ∼1×1020 cm−3 was grown on top of the laser

structure. Quarter wafers were annealed at 740◦C for one minute in the RTA. Ti-

Pt-Au top contacts were deposited onto the wafer. Ridge waveguides of 5, 10, and

20 µm were etched down into the top GaAs waveguide layer. After wafer thinning,

Au-Ge-Ni-Au bottom contacts were deposited on the thinned backside. Cleaving

was performed to define the Fabry-Perot cavities.

Laser testing was performed at room temperature (controlled by a copper heat

sink), under cw operation, and epi-side up. Bank et al. had initial success in devel-

oping EELs with lasing wavelengths ≥1.50 µm with a laser diode which emitted at

1.50 µm [115]. Changing from GaNAsSb to GaNAs barriers and additional growth

and plasma improvements enabled the development of this set of devices. For a 20

µm × 2150 µm device, the threshold current density was 440 A/cm2 with an effi-

ciency of 51%. This is the lowest threshold current density for GaAs-based lasers

with wavelengths longer than 1.4 µm to our knowledge. A peak output power of 431

mW from both facets was obtained with 16.3% peak wall plug efficiency.

The next EEL success by Bank et al. was the development of a low-threshold

GaInNAsSb laser at 1.55 µm [139]. Development of a laser at this wavelength was

6.1. LOW-THRESHOLD GaInNAsSb QW EDGE EMITTING LASERS 139

1540 1545 1550 1555 1560 1565 15700.0

0.2

0.4

0.6

0.8

1.0

Pow

er (A

.U.)

Wavelength (nm)

Figure 6.1: Laser spectrum of a GaInNAsSb/GaNAs/GaAs 1.56 µm EEL at 1.2×threshold.

more difficult due to a decreasing growth parameter window for optimal quality.

Pushing from 1.50 to 1.55 µm required the addition of more nitrogen to GaInNAsSb,

rather than more indium and antimony as previously thought. In addition, the sensi-

tivity of material quality to the exact growth temperature became more apparent for

this wavelength. Finally, as discussed in Chapter 5, the optimal anneal temperature

of the GaInNAsSb active region became an added concern in the design and growth

of these lasers. Figures 6.1 and 6.2 show the relevant data for the 1.55 µm device.

For a 20 µm x 2400 µm device, the threshold current density was 579 A/cm2 with

an efficiency of 40%. This is the lowest threshold current density for a GaAs-based

laser at 1.55 µm.

Figure 6.3 shows a compilation of various EELs from our own research and from

the literature. Plotting the threshold current density, a primary indicator of device

quality, versus lasing wavelength, several patterns can be observed. Dilute nitride

device development at 1.3 µm has progressed quite well with a continual reduction

in the threshold current density in recent years. However, the initial push to longer

wavelengths by adding more nitrogen and/or indium resulted in a dramatic increase


0 250 500 750 10000

50

100

150

0

2

4

6

CW

Out

put P

ower

(mW

)

Current (mA)

Wal

lplu

g E

ffici

ency

(%)

Figure 6.2: L-I curve and wall plug efficiency for the GaInNAsSb/GaNAs/GaAs 1.56µm laser under cw conditions.

in the threshold current density, indicating poor device quality and performance.

This was dubbed the “nitrogen penalty” due to the perception several years ago

that addition of nitrogen could only make the material quality worse. However, this

“nitrogen penalty” was discovered instead to be a “nitrogen complexity” once the

optimal growth parameters were obtained. Unfortunately, each wavelength requires

a re-examination of such quantities as composition and growth temperature, making

dilute nitride device development more “complex.” As the figure points out though,

with improvements in growth and active region design, the device quality was im-

proved such that we were able to obtain low threshold current densities which were

comparable or even below those of current InP-based lasers. With further improve-

ment and greater understanding in growth of dilute nitrides, it is possible to obtain

better performing lasers.

6.1. LOW-THRESHOLD GaInNAsSb QW EDGE EMITTING LASERS 141

1.1 1.2 1.3 1.4 1.5 1.60.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

Improving MBE Growth

Initially Perceived"Nitrogen Penalty"

Thre

shold

Curr

ent D

ensi

ty (kA

/cm

2)

Lasing Wavelength (µm)

<1.25 µm InGaAs <1.32 µm GaInNAs >1.32 µm GaInNAs(Sb) This Work GaInNAsSb

InP LasersInP Lasers

Figure 6.3: Comparison of data from our work and devices found in the literature.The effect on material and device improvement can be seen in the reduction of thethreshold current density.


6.2 GaInNAsSb Vertical Cavity Surface Emitting

Lasers

Much research has gone into the development of GaAs-based VCSELs operating at

or near 1.3 µm. With the exception of our research group, no one has been able

to produce a GaAs-based electrically pumped VCSEL beyond 1.4 µm due to the

extreme complexities of developing a high quality dilute nitride active region and

associated DBR structures. In 2003, Wistey et al. demonstrated a 1.46 µm VCSEL

utilizing an active region with three 7 nm GaInNAsSb QWs surrounded by 20 nm

GaNAs barriers [114]. Multiple QWs are required to ensure enough gain for lasing

action since light travels perpendicularly to the QW layers. The bottom silicon doped

n-type DBR contained 29 Al0.92Ga0.08As/GaAs mirror pairs for a reflectivity greater

than 99.99%. The top carbon doped p-type DBR contained 24 mirror pairs for a

reflectivity of 99.7%.

At –10◦C operating temperature, the threshold current was 580 mA under pulsed

conditions (0.1% duty cycle) with a peak power of 33 µW. This translated to a current

threshold density of 17 kA/cm2 or 5.7 kA/cm2/QW. Performance characteristics

deviated from expected values due to a misalignment of the cavity resonance with

the gain peak and fabrication errors. The GaInNAsSb QW material emitted light at

1.49 µm at room temperature, but the DBRs were optimized for 1.46 µm emission

due to a miscalibration of source fluxes during DBR growth. To obtain lasing, cooling

the GaInNAsSb/GaNAs active region increases the band gap, thus shortening the

emission wavelength. The emission wavelength can then be made to match the cavity

resonance.

More recently, Wistey et al. has demonstrated a VCSEL which operated at 1.534

µm, the longest wavelength electrically pumped VCSEL on GaAs to our knowl-

edge [140]. These devices also utilized an active region containing three 7.5 nm

GaInNAsSb QWs with 21 nm GaNAs barriers. The bottom silicon doped n-type

DBRs contained 31 AlAs/GaAs and 4 Al0.91Ga0.09As/GaAs mirror pairs for a reflec-

tivity of 99.99%. The carbon doped p-type DBR contained 21 Al0.98Ga0.02As/GaAs

mirror pairs for a reflectivity greater than 99.1%.

6.3. CONCLUSION 143

Figure 6.4: Laser spectrum of a triple GaInNAsSb/GaNAs QW 1.534 µm VCSELat 1.6× threshold.

Figures 6.4 and 6.5 show the output characteristics of these 1.534 µm VCSELs

with a mesa diameter of 28 µm. The devices were operated under pulsed conditions

with 0.1 µs pulse width at a 0.67% duty cycle. The threshold current was 70 mA.

Due to an operational problem with the plasma cell, the plasma properties were

altered during the growth of these VCSELs causing higher amounts of nitrogen to be

incorporated than expected. This caused the GaInNAsSb QW to have an emission

wavelength of 1.585 µm while the cavity resonance was calibrated for 1.54 µm. Due

to this misalignment, the operating temperature of the VCSEL was –48◦C.

6.3 Conclusion

The demonstration of low-threshold GaInNAsSb/GaNAs/GaAs EELs at 1.55 µm

and VCSELs which operate at 1.534 µm illustrates the feasibility of employing the


Figure 6.5: L-I curve for a triple GaInNAsSb/GaNAs QW 1.534 µm VCSEL oper-ating under pulsed conditions.

dilute nitride antimonide material system for long-wavelength optoelectronics. Fur-

ther improvement in material and device quality will continue to reduce the threshold

current density and increase the output power of EELs. Although there were sev-

eral difficulties in producing VCSELs at 1.55 µm, they were not inherent challenges

and can be surely overcome. The development of low-cost GaAs-based VCSELs for

fiber telecommunications utilizing dilute nitride antimonides will continue. When the

market recovers from the recession, commercialization of these devices can occur.

Chapter 7

Additional Applications of Dilute

Nitride Alloys

7.1 Dilute Nitride for Biosensor Applications

Most biological and chemical agents involve bonds between carbon, hydrogen, oxy-

gen, and nitrogen and have strong absorption and resonances in the mid and far

infrared (IR) wavelength range of 3–15 µm. Complex protein molecules including

RNA and DNA are also highly absorbing in the IR. The mid to far IR wavelength

range is important for many biochemical applications.

There has been an increasing need for highly sensitive and multi-functional sen-

sors for biochemical detection, medical applications, and national security against

biological weapons. These multifunctional smart sensors must have several charac-

teristics:

• High sensitivity to small numbers of molecules in a volume of space.

• Multiple wavelength operation designed for chemical and biological agents.

• High selectivity to specific molecular and chemical species.

• Chemical recognition to compounds such as antibodies.

• Full integration with lasers, detectors, optical components, and circuits.

145

146 CHAPTER 7. ADDITIONAL APPLICATIONS OF DILUTE NITRIDES

• Low cost and compact size.

Highly efficient and inexpensive IR sources and detectors are essential to enable

smart sensor systems. Several different types of IR lasers have been developed,

but a cryogenic operating temperature has been a major challenge in widespread

distribution. Recent quantum cascade lasers (QCLs) have been able to operate at

room temperature, but these lasers have difficulties with compact system integration.

Quantum dot lasers have several advantages including:

• Low power consumption due to low threshold current operation.

• Large quantum efficiency due to discrete energy states.

• Array operation and integration.

• Low cost due to simple growth and fabrication techniques.

• Ability to combine materials with large lattice constant mismatches with thick-

ness.

Although visible and near-IR QD lasers have been demonstrated with good de-

vice performance, there have been difficulties in extending their operation beyond

mid-IR wavelengths. Bulk InAs and InSb emit at mid-IR wavelengths, but quantum

confinement from the QD increases the transition energies. As shown in Figure 7.1,

there are no other conventional III-V semiconductors which can emit light of longer

wavelengths, making the realization of a semiconductor-based mid to far-IR detector

difficult. InAsSb with ∼30% antimony has the lowest band gap of conventional semi-

conductors, but it is still too large with quantum confinement for far-IR wavelengths.

Compositional unifromity also make growth difficult [142, 143].

Due to the success of GaAs-based dilute nitrides for long-wavelength optoelec-

tronics, there has been great interest in developing an InAs and/or InSb-based dilute

nitride material system to reduce the band gap so that quantum confinement will not

limit emission wavelengths to the mid-IR. The addition of small amounts of nitrogen

has been shown to reduce the band gap of InAs and InSb, similar to that of GaAs

[144, 145]. Theoretical calculations have also predicted the addition of 2–3% nitrogen

7.1. DILUTE NITRIDE FOR BIOSENSOR APPLICATIONS 147

AlAs

AlSb

GaAs

InSb

GaSb

InP

GaP

InAs10

2

1

0.5

0.0

0.5

1.0

1.5

2.0

2.5

5.4 5.6 5.8 6.0 6.2 6.4

GaNAs

AlP

InAsNSb

Ban

d G

ap (e

V)

Lattice Constant (Å)

Wav

elen

gth

(µm

)

Figure 7.1: Band gap versus lattice constant illustrating the region of III-V semicon-ductors which can be used to obtain emission at mid to far-IR wavelengths.

should result in a zero band gap semiconductor [145]. However, there have not been

many investigations into the growth of dilute nitrides for mid to far-IR applications,

leaving much knowledge to be discovered.

Devices based on In(N,As,Sb) quantum nanostructures have several advantages

including:

• Freedom for band gap engineering to cover the IR spectrum.

• Suppression of Auger recombination in tightly confined structures leading to

higher temperature device operation.

• Fewer constraints exist on the choice of substrate for growth; preferable for

device integration.

There are however several challenges which must be overcome before these devices

become viable options:

• Obtaining high quality dilute nitride antimonide material by MBE.


• Lack of knowledge in the community on material and growth properties.

• Possible different mechanism of nitrogen incorporation in QDs compared to

bulk and quantum well structures.

• Strain distribution changes with nitrogen incorporation.

The following sections describe pioneering work on the development of In(N,As,Sb)

materials grown by MBE.

7.1.1 InNSb and GaNSb

The goal is to develop a material which can emit light at mid to far-IR wavelengths

and be grown as QDs due to the advantages stated earlier in this chapter. The

smallest band-gap binary III-V semiconductor is InSb with a value of 174 meV. By

adding a small amount of nitrogen into InSb, forming InNSb, it has been shown to

reduce the band gap below this value [145].

A systematic investigation into InSb and InSb:N QD growth on InAs and GaAs

was performed. Details will not be presented here, but can be found elsewhere [146].

Exposure of InSb dots to the nitrogen plasma lead to some change in QD density and

morphology, indicating a change in lattice constant and/or surface growth kinetics.

However, low-temperature PL showed no luminescence from the QDs.

The lack of luminescence from the InSb:N QDs necessitated an examination of

the growth technique as well as the quality of the InNSb material itself. Nitrogen

incorporation into non-nitride III-V semiconductors has traditionally been a very

difficult task due to practical and scientific challenges. Nitrogen in its natural state

is non-reactive and requires a method to produce species which will bond with the

semiconductor. In our case, the rf plasma cell has been shown to be a very good

source of nitrogen for GaAs-based dilute nitrides. McConville et al. has been able

to grow InNSb using different methods including ion implantation [147] and ECR

plasma [148]. Nitrogen incorporation is also a difficult process and requires a detailed

growth examination. Ideally, the nitrogen bonds substitutionally to other group-

III atoms. However, it can bond interstitially, as an N-N or N-V split interstitial,


27 28 29 30 31 32 33 34

10

100

1k

10k

100k

1M

10M

Cou

nts

(a.u

.)

/2 (Degrees)

InSb

In"N"Sb

GaAs

Figure 7.2: (004) ω/2θ HRXRD spectra of thick InSb and InSb:N layers on GaAs.There is no difference in diffraction angles.

in clusters, or not at all. Finally obtaining high quality material can be difficult

depending on the growth methods.

Thick relaxed layers of InSb and InSb:N were grown on GaAs to determine the

approximate presence of any nitrogen incorporation. Figure 7.2 shows the (004)

HRXRD spectra from the two samples. If any significant nitrogen concentration

is found in the InSb:N material, the diffraction angle of the bulk material should

be different than that of InSb. However, it is seen that the diffraction angles are

essentially identical, indicating no substitutional nitrogen incorporation.

The method of nitrogen production as well as the large lattice constant and

electronegativity differences between nitrogen and the matrix atoms are possible

causes for the difficulty of incorporating nitrogen. McConville et al. was able to

incorporate up to 2% nitrogen in InNSb [145]. However, greater than 90% of this

nitrogen was found in interstitially. They have utilized ion implantation and ECR

plasmas as a way of incorporating nitrogen into InSb. Both are very high energy

sources which bombard the nitrogen into the material, possibly explaining the very


high interstitial concentrations. The rf plasma is a much lower energy source and

does not implant the nitrogen into the material. In addition, nitrogen is a very small

atom compared to other group-III and group-V atoms. While it is smaller than

gallium and arsenic atoms, it is much smaller compared to indium and antimony

atoms. The size of a nitrogen atom allows it to be placed in the interstitial site of

the InSb crystal, further enhancing the ease of interstitial incorporation. Combined

with the large electronegativity differences, nitrogen incorporation in InSb is quite

difficult.

In an effort to reduce the atomic size and electronegativity differences, indium

was replaced with gallium, forming GaNSb. Figure 7.3 shows the HRXRD spectra

of GaSb and GaNSb grown on InAs. GaSb and InAs are similar in lattice constant

(6.056 A and 6.058 A, respectively) allowing for coherent growth of GaSb on InAs.

It can be see that growth of GaSb on InAs results in nice Pendellosung fringes and a

well defined film peak. However, in the presence of the nitrogen plasma, the fringes

disappear and the peak broadens and shifts towards the substrate peak. Figure 7.4

shows a RSM of the same sample indicating what appears to be a relaxation of the

material. This is surprising since the strain is not very large and the thickness of

the film peak under the critical thickness. SIMS confirmed a concentration of 3±2%

nitrogen. While nitrogen incorporation was accomplished, it appears the growth of

dilute nitrides in pure antimonides is quite difficult. This has also mentioned in the

growth of antimony rich GaInNAsSb alloys [149].

7.1.2 InNAsSb

In light of the difficulties with obtaining nitrogen incorporation and high quality ma-

terial with pure antimonides, arsenic was re-introduced during the growth process.

Incorporation of nitrogen into arsenic-based semiconductors such as GaNAs and In-

NAs has been successfully accomplished. It is possible that the smaller differences in

size and electronegativity between arsenic and nitrogen allow for easier incorporation

compared to antimony and nitrogen. Utilizing the previous growth conditions for

In(N)Sb growth in which nitrogen incorporation was unsuccessful, an arsenic flux


29.0 29.5 30.0 30.5 31.0 31.5

10

100

1k

10k

100k

1M

10M

Cou

nts

(a.u.)

/2 (Degrees)

InAs

GaSb

Ga"N"Sb

Figure 7.3: (004) ω/2θ HRXRD spectra of thin GaSb and GaNSb films on InAs.While nitrogen was found in GaNSb, the structural quality was quite poor.

was introduced of ∼5×10−7 Torr BEP (∼4–5× overpressure) during growth. Figure

7.5 shows the HRXRD spectra of 500 A InAsSb and InNAsSb films on InAs. Both

samples are grown coherently upon InAs with nice Pendellosung fringes. Compared

to the InAsSb sample, the InNAsSb sample has less strain, indicating the possibility

of nitrogen incorporation. As confirmed with SIMS and HRXRD, the InAsSb sample

contained 12.5% antimony while the InNAsSb sample contained 12.5% antimony and

1% nitrogen. The addition of arsenic has enabled nitrogen incorporation.

Low-temperature PL (LT-PL) was performed on a 90 A InNAsSb QW grown on

and capped by InAs to examine the optical properties of this new material. Figure

7.6 shows the spectra obtained from the QW. The InNAsSb QW luminescence can

be seen magnified at 4 µm. This is the first reported luminescence from InNAsSb to

our knowledge. This holds great promise for reaching the mid to far-IR wavelengths

needed for biosensing devices. However, the peak intensity of the InNAsSb QW is

quite low compared to the luminescence of the InAs substrate at 3 µm. In addition,

the PL linewidth is ∼100 meV, quite broad for a LT-PL measurement. The material


InAs

GaNSb

Q

Q ||

Figure 7.4: (224) RSM of the GaNSb film on InAs indicating relaxation.


29.0 29.5 30.0 30.5 31.0

100

1k

10k

100k

1M

10M

100M

1G

Cou

nts

(a.u.)

/2 (Degrees)

InAsSb

InNAsSb

InAs

Figure 7.5: (004) ω/2θ HRXRD spectra of 500 A InAsSb and InNAsSb films onInAs. The nitrogen containing sample appears to be of good structural quality.

quality is certainly a possible cause for such broad linewidth as this was a proof-of-

principle sample and additional work is required to further improve the alloy.

The band lineup of InNAsSb to InAs is also a potential issue. Figure 7.7 diagrams

the band lineups of strained InAs0.9Sb0.1/InAs. It can be seen that this lineup is

slightly type-II. Adding nitrogen should push the conduction band downwards forcing

a type-I alignment, but the value of the electron and hole confinements could be an

issue. With quantum confinement in a QW or QD, the ground state energy could

exceed the band offset values and no carrier confinement would occur, resulting in a

broadened PL peak.

7.1.3 Conclusion

Semiconductor QD nanostructures have several advantages over current technology

for smart sensor technology. However, QDs utilizing conventional III-V semiconduc-

tors have difficulty reaching the mid to far-IR wavelengths needed for these biosensing


0.45 0.4 0.35 0.3

3.0 3.5 4.0 4.5

123456789

1011

Wavelength ( m)

Inte

nsity

(a.u.)

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

Inte

nsity

(a.u.)

Energy (eV)

InAs InNAsSb

CO2 absorption

Figure 7.6: Low-temperature PL of an InNAsSb QW on InAs. The QW peak islocated at 4 µm. The dip in luminescence of the QW peak located between 4.25–4.5µm is due to CO2 absorption in the ambient.

InAsInAs0.9Sb0.1

79 meV

8 meV 283 meV

Figure 7.7: Band alignment of strained InAs0.9Sb0.1 on InAs showing a type-II align-ment.

7.2. DILUTE NITRIDE FOR SOLAR CELL APPLICATIONS 155

devices. It was proposed the addition of nitrogen could be used to further reduce the

band gap of InAs and InSb such that quantum confinement would not increase the

transition energy beyond the desired range. Addition of nitrogen to pure antimonide

materials has proved difficult using an rf plasma cell. Nitrogen incorporation was

made possible by re-introducing arsenic into the material, forming InNAsSb. PL was

obtained for the first time from InNAsSb at 4 µm, giving promise to the development

of devices with mid to far-IR wavelength operation. Future work is needed to further

study and improve the material. To ensure sufficient band alignment, InNAsSb QDs

should be grown on GaAs.

7.2 Dilute Nitride for Solar Cell Applications

As mentioned in Chapter 5, GaInNAs has garnered great interest as material which

can be used in a four junction solar cell [131–138]. Current commercially available

III-V solar cells utilize a triple junction structure consisting of GaxIn1−xP, GaAs,

and Ge layers as shown in Figure 7.8. A 1.0 eV junction placed between the 1.4

eV GaAs and 0.7 eV Ge layers increases device efficiency [150]. GaInNAs with low

concentrations of indium (6-8%) and ∼2% nitrogen can be grown coherently upon

GaAs or Ge while obtaining a 1.0 eV band gap, making it an ideal candidate for the

four junction solar cell. However, GaInNAs material quality is an issue as defects

have limited the minority-carrier diffusion lengths [133, 134], leading to inefficient

performance.

In an effort to improve the material quality and minority carrier diffusion lengths

of GaInNAs, deflection plates and antimony were employed during the dilute ni-

tride growth. Both deflection plates and antimony helped GaInNAs for the long-

wavelength lasers, so it was assumed that it would also help for the solar cell mate-

rial. First, a 2 µm thick GaInNAs P-i-N structure was grown to set a baseline for

performance. At the time (2003), this was the first reported successful growth of a 2

µm thick GaInNAs layer by MBE. Plasma stability was an important improvement

in enabling such a thick growth. Figure 7.9 shows a RSM of this 2 µm thick GaInNAs

layer with 7.7% indium and 1.8% nitrogen. The layer is coherent even though it had


4 5 6 7 8 91

2 3 4

E ner gy (eV )

GaAs

1.4 eV

GaAs

1.4 eV

GaxIn1-xP

1.8 eV

GaxIn1-xP

1.8 eV

Ge

0.7 eV Ge

0.7 eV

New

1.0 eV

(a) (b)

Figure 7.8: The solar spectrum as observed below Earth’s atmosphere (AM 1.5). (a)The current design of three junction III-V solar cell devices. (b) A proposed fourjunction device with an added 1.0 eV junction for increased efficiency.

a compressive strain of 0.19% and a theoretical critical thickness of 58 nm, well be-

low the actual thickness. Absorption and PL measurements indicated a band gap of

1.1–1.2 eV, slightly larger than intended. Optimal anneal temperatures were quite

high (> 820◦C) and agree with the theory presented in Chapter 5 that lower strained

materials have higher optimal anneal temperatures than highly strained alloys. Ini-

tial dark current measurements were quite low compared to previous relaxed InGaAs

detectors and gave very promising results.

The next set of growths consisted of a 1 µm thick GaInNAs and GaInNAsSb

P-i-N structures with a voltage applied to the deflection plates found in front of

the plasma cell to reduce ion damage. By reducing ion damage as well as improving

growth with antimony, it was hoped the material quality would improve and increase

the minority carrier diffusion lengths. Since the original GaInNAs sample did not

reach 1.0 eV band gap, the indium and nitrogen concentrations were increased to

10.5% and 2.3%, respectively. Unfortunately, this increased the compressive strain

found in the thick layers and induced relaxation. The un-relaxed compressive strains


Q

Q ||

GaInNAs

GaAs

Figure 7.9: (224) RSM of a 2 µm thick GaInNAs layer on GaAs. No relaxation isobserved.


GaAs

GaInNAs

(a) 1 µm GaInNAs P-i-N

GaAs

GaInNAsSbQ

Q ||

(b) 1 µm GaInNAsSb P-i-N

Figure 7.10: (224) RSMs of relaxed (a) GaInNAs and (b) GaInNAsSb P-i-N struc-tures on GaAs.

for the GaInNAs layer were 0.38% while the GaInNAsSb layer was 0.50%. The RSMs

of the two samples can be seen in Figure 7.10. The GaInNAs layer showed slight

signs of relaxation while the GaInNAsSb showed increased amounts. Neither sample

was fully relaxed however. Figure 7.11 shows a TEM micrograph of the GaInNAsSb

with a network of misfit dislocations on the top and bottom interfaces of the layer.

Expectations for solar cell performance were low since these two newer samples

had relaxed, introducing significant amounts of non-radiative defects. In addition,

as the results in Chapter 5 demonstrated, antimony with low-strained GaInNAs

degraded the PL intensity. Investigations of the GaInNAsSb structure continued

nonetheless because, as the first reported GaInNAsSb detector, much information

could still be obtained. The band gap of the GaInNAs sample with deflection plates


p-GaAs

GaInNAsSb

n-GaAs

Surf

ace

g00

2

500 nm

Figure 7.11: (002) dark field cross-sectional TEM tilted slightly off-axis of the relaxedGaInNAsSb sample. A network of misfit dislocations can be seen on the top andbottom interfaces of the GaInNAsSb layer.

was ∼1.02 eV while the GaInNAsSb sample had a band gap of ∼0.93 eV, both closer

to the intended 1.0 eV value. In results obtained by A. J. Ptak for D. B. Jackrel

(additional details may be found in his thesis), the baseline GaInNAs device had an

internal quantum efficiency (IQE) of 55% while the GaInNAs with deflection plates

and GaInNAsSb samples had IQEs of 68% and 79%, respectively. The fill factor (FF)

is another metric of solar cell devices and measures the peak power to the product of

the short circuit current and open circuit voltage. Ideal cells have FF of ∼0.9 while

commercial solar cells have a FF of 0.8–0.9. The baseline GaInNAs device had a

FF of 0.51, quite low compared to commercial devices. However, the GaInNAs with

deflection plates and GaInNAsSb devices had FFs of 0.71 and 0.61, respectively.

Surprisingly, both GaInNAs with deflection plates and GaInNAsSb devices showed

superior performance characteristics compared to the baseline GaInNAs sample even

though relaxation occurred in each. In addition, the high IQE and improved FF of

the GaInNAsSb sample compared to the GaInNAs baseline sample appears to con-

tradict the PL results when adding antimony to low strained GaInNAs. There are


a couple theories why the GaInNAsSb sample had improved device performance.

Solar cell researchers do not use PL as a metric of potential device performance and

rely heavily on mobility measurements. In fact, it has been observed with dilute

nitride devices that mobility (and solar cell performance) and PL intensity appear

to be mildly inversely related: the worse the PL intensity, the better the solar cell

performance [151]. The donor/acceptor concentrations were also measured for these

samples. Generally, lower concentrations lead to larger minority carrier diffusion

lengths. The GaInNAsSb sample had much lower donor concentrations compared to

the GaInNAs baseline sample. However, the GaInNAs sample with deflection plates

had higher donor concentrations compared to both [151].

In conclusion, GaInNAsSb has been shown to be a viable option for the 1.0 eV

junction in a four junction solar cell. A great deal of research and discovery remain

as there are several unanswered questions and more optimal samples to be grown.

Chapter 8

Conclusion and Future Work

8.1 Conclusion

Dilute nitride alloys for long-wavelength applications have been grown by MBE.

Adding nitrogen to GaAs allows a simultaneous reduction in band gap and lattice

parameter, contrary to the behavior of traditional III-V semiconductors. This anom-

alous band gap reduction is due to the band anti-crossing of a nitrogen-related energy

level with the GaAs conduction band. These GaAs-based dilute nitrides have sev-

eral advantages over current InP-based technology in the quest to bring low-cost

fiber connections to the home.

Conventional growth techniques are not valid for novel materials like the dilute

nitrides. In order to create reactive nitrogen, a rf plasma cell was required to generate

species which could be incorporated. The properties of the nitrogen plasma greatly

affect the quality of the material. Antimony was also added as a surfactant and in-

corporant to improve dilute nitride quality, enabling the development of GaInNAsSb

materials which were used for 1.55 µm emitting devices.

One particular aspect of plasma cell operation is the gas flow which enters the cell.

This affects the pressure found inside the cell and also has an effect on the quality of

the plasma generated for semiconductor growth. It was found that higher gas flows

(and higher pressures in the cell) led to an improvement in GaInNAs optical quality.

Ion count measurements were also performed to correlate the plasma properties with

161

162 CHAPTER 8. CONCLUSION AND FUTURE WORK

the material improvement. With higher gas flows, ion counts and ion energies were

reduced, limiting the damage done to the surface during growth.

The barrier materials which surround the QWs are of key importance to the

overall active region quality. GaAs, GaNAs, and GaNAsSb are the three materials

which are easily accessible and are used by groups who have the capabilities to grow

GaInNAsSb. It was found GaNAsSb is generally not a good choice as a barrier

material due to poor optical quality, lack of tensile strain compensation for the

highly compressive GaInNAsSb QWs, and potentially type-II band offset alignment.

GaAs is the ideal barrier for a single GaInNAsSb QW since it has much higher

quality than GaNAs. However, it does not provide strain compensation and there

are practical difficulties in controlling the antimony profile. For devices which require

MQW structures, GaNAs is the best choice for GaInNAsSb QWs.

Antimony has been used in the past without much investigation of its role and

effects. Although it has been shown to improve material and optical quality of

GaInNAs alloys for long-wavelength applications, it has never been employed for

compositions used in solar cell devices. Adding antimony to GaInNAs of lower

strain due to lower indium concentrations degraded the optical quality, contrary

to the behavior of high indium, high strain compositions. Antimony is a reactive

surfactant which works well for strained heteroepitaxial growth. This can explain

why antimony was useful for GaInNAs with high indium and high strain, but not

in the low indium and low strain cases. Although antimony is desirable for the

high indium material, it is apparent that minimization of antimony incorporation is

key to improved luminescence. Strain also plays an important role in the optimal

annealing temperature of the GaInNAsSb QWs. Ultimately, material with higher

optimal anneal temperatures and lower lattice strain are preferred for GaInNAsSb

device growth.

GaInNAsSb EELs and VCSELs have been fabricated at the important 1.55 µm

wavelength. The EELs have the lowest current density thresholds for GaAs-based

lasers at this wavelength and are comparable or lower than those of InP-based de-

vices. The VCSELs are the first monolithic and electrically pumped VCSELs at

1.55 µm. These devices have shown the feasibility of utilizing this new material

8.2. FUTURE WORK 163

for low-cost long-wavelength optoelectronics. GaInNAsSb devices at 1.3 µm have

not been fabricated. Typical GaInNAsSb PL intensities are higher than those of

GaInNAs QWs, but the GaInNAsSb alloys with much higher luminescence also con-

tained more strain and much lower optimal annealing temperatures, leading to the

conclusion that they may be overannealed during laser growth.

Dilute nitrides also have several other applications in addition to long-wavelength

optoelectronics. In(N,As,Sb) alloys have been shown to emit light in the mid IR wave-

lengths, useful for biosensing devices. With further study, the maximum wavelength

achievable should increase into the far IR, allowing for a wide range of biological and

chemical detection applications. As mentioned earlier, dilute nitrides have also found

interest for solar cell applications as the 1.0 eV junction. The use of deflection plates

and antimony with GaInNAs have improved the solar cell operational characteristics,

increasing the efficiency of III-V semiconductor solar cells.

8.2 Future Work

The end of this thesis does not signify the end of possible research opportunities

with the dilute nitrides. Much research remains in the discovery, investigation, and

optimization of these materials for its wide variety of applications. It is my hope

that some or all of these questions and points may be answered in the future:

• What species of nitrogen (atomic N, N∗2, etc.) is actually incorporated? Does

the species of nitrogen make a difference in the material quality? How does

one change the plasma properties to obtain the optimal species of nitrogen.

• Further investigations of dilute nitride antimonides with a valved antimony

cracker. Can adjusting the antimony fluxes assist in antimony compositional

profiles?

• What is the exact cause of the nitrogen incorporation enhancement with anti-

mony? Does nitrogen enhance antimony incorporation?

164 CHAPTER 8. CONCLUSION AND FUTURE WORK

• What species of antimony is best for dilute nitride growth? We have used Sb1

and quick investigations with Sb2 and Sb4 have not shown much change, but if

the behavior is similar of that to GaAs quality using As4 and As2, there should

be a difference.

• What other barriers can be used with GaInNAsSb QWs? GaAsP has shown

great promise with GaInNAs devices, but it is difficult to have so many sources

in a single MBE machine. Could GaInNAs strain mediating layers help? Would

having an additional gallium or indium cell help?

• Why exactly does adding antimony to low-strained GaInNAs hurt material

quality? Does the species matter? Antimony’s effects on dilute nitrides are

still mysterious and much work remains.

• Further study and improvement to In(N,As,Sb) materials. It is a very new

material and no one else has grown this material by MBE or MOCVD.

• Why is nitrogen incorporation in pure antimonide materials such a difficult

process? Are nitrogen and antimony just too different from each other?

• Development of high quality GaInNAsSb solar cells.

• Why is PL intensity and solar cell performance inversely related to each other?

GaInNAsSb with low indium had very poor optical quality, yet it had superior

solar cell characteristics compared to GaInNAs.

• Why is our GaInNAs material n-type? Almost every other group in the world

reports their material to be p-type. Is this related to the fact our material is

much higher quality than many others around the world?

Bibliography

[1] L. A. Coldren and S. W. Corzine, Diode Lasers and Photonic Integrated Circuits

(Wiley, New York, NY, 1995).

[2] J. S. Harris, Semicond. Sci. Technol. 17, 880 (2002).

[3] A. F. Phillips, S. J. Sweeney, A. R. Adams, and P. J. A. Thijs, IEEE J. Sel.

Top. Quantum Electron. 5, 301 (1999).

[4] M. Yano, H. Imai, and M. Takusagawa, J. Appl. Phys. 52, 3172 (1981).

[5] N. K. Dutta and R. J. Nelson, Appl. Phys. Lett 38, 407 (1981).

[6] K. Kurihara, M. Takashima, K. Sakata, R. U. M. Takahara, H. Ikeda, H. Na-

mita, T. Nakamura, and K. Shimoyama, J. Cryst. Growth 271, 341 (2004).

[7] A. J. SpringThorpe, M. Extavour, D. Goodchild, E. M. Griswold, G. Smith,

J. K. White, K. Hinzer, R. Glew, R. Williams, and F. Robert, J. Cryst. Growth

51, 760 (2003).

[8] M. Yamada, T. Anan, K. Tokutome, A. Kamei, K. Nishi, and S. Sugou, IEEE

Photon. Technol. Lett. 12, 774 (2000).

[9] H. C. Kuo, Y. H. Chang, H. H. Yao, Y. A. Chang, F. I. Lai, M. Y. Tsai, and

S. C. Wang, IEEE Photon. Technol. Lett. 17, 528 (2005).

[10] C. Paranthoen, C. Platz, G. Moreau, N. Bertru, O. Dehaese, A. L. Corre,

P. Miska, J. Even, H. Folliot, and C. Labbe, J. Cryst. Growth 251, 230 (2003).

165

166 BIBLIOGRAPHY

[11] C. N. Allen, P. J. Poole, P. Barrios, P. Marshall, G. Pakulski, S. Raymond,

and S. Fafard, Physica E 26, 372 (2005).

[12] J. A. Lott, N. N. Ledenstov, V. Ustinov, A. Maleev, A. Zhuov, A. Kovsh,

M. Maximov, B. Volovik, Z. Alferov, and D. Bimberg, Electron. Lett. 36, 1384

(2000).

[13] M. J. DaSilva, A. A. Quivy, S. Martini, T. E. Lamas, E. C. F. DaSilva, and

J. R. Leite, J. Cryst. Growth 251, 181 (2003).

[14] G. Patriarche, F. Jeannes, J. Oudar, and F. Glas, J. Appl. Phys. 82, 4892

(1997).

[15] L. Goldstein, C. Fortin, C. Starck, A. Plais, J. Jacquet, J. Boucart, A. Rocher,

and C. Poussou, Electron. Lett. 34, 268 (1998).

[16] S. Uchiyama and S. Kashiwa, Electron. Lett. 31, 1449 (1995).

[17] R. Teissier, D. Sicault, J. C. Harmand, G. Ungaro, G. LeRoux, and L. Largeau,

Appl. Phys. Lett 89, 5473 (2001).

[18] S. R. Johnson, C. Z. Guo, S. Chaparro, Y. G. Sadofyev, J. Wang, Y. Cao,

N. Samal, C. Navarro, J. Xu, S. Q. Yu, et al., in 2002 International MBE

Conference (San Francisco, CA, 2002).

[19] X. D. Luo, L. F. Bian, Z. Y. Xu, H. L. Luo, Y. Q. Wang, J. N. Wang, and

W. K. Ge, Acta Phys. Sinica 52, 1761 (2003).

[20] S. W. Ryu and P. D. Dapkus, Semicond. Sci. Technol. 19, 1369 (2004).

[21] R. Kudrawiec, G. Sek, K. Ryczko, J. Misiewicz, and J. C. Harmand, Appl.

Phys. Lett 84, 3453 (2004).

[22] M. Weyers, M. Sato, and H. Ando, Jpn. J. Appl. Phys. Part 2 31, L853 (1992).

[23] M. Kondow, K. Uomi, A. Niwa, T. Kitatani, S. Watahiki, and Y. Yazawa, Jpn.

J. Appl. Phys., Part 1 35, 1273 (1996).

BIBLIOGRAPHY 167

[24] W. Shan, W. Walukiewicz, J. W. Ager, E. E. Haller, J. F. Geisz, D. J. Fried-

man, J. M. Olson, and S. R. Kurtz, Phys. Rev. Lett 82, 1221 (1999).

[25] P. J. Klar, H. Gruning, W. Heimbrodt, J. Koch, F. Hohnsdorf, W. Stolz,

P. M. A. Vincente, and J. Camassel, Appl. Phys. Lett 76, 3439 (2000).

[26] J. Misiewicz, R. Kudrawiec, K. Ryczko, G. Sek, A. Forchel, J. C. Harmand,

and M. Hammar, J. Phys. Cond. Mat. 16, 3071 (2004).

[27] C. Skierbiszewski, P. Perlin, P. Wisniewski, W. Knap, T. Suski,

W. Walukiewicz, W. Shan, K. M. Yu, J. W. Ager, E. E. Haller, et al., Appl.

Phys. Lett 76, 2409 (2000).

[28] P. N. Hai, W. M. Chen, I. A. Buyanova, H. P. Xin, and C. W. Tu, Appl. Phys.

Lett 77, 1843 (2000).

[29] M. Hetterich, M. D. Dawson, A. Egorov, D. Bemklau, and H. Riechert, Appl.

Phys. Lett 76, 1030 (2000).

[30] S. G. Spruytte, Ph.D. thesis, Stanford Univ. (2001).

[31] C. W. Coldren, Ph.D. thesis, Stanford Univ. (2005).

[32] V. F. Gambin, Ph.D. thesis, Stanford Univ. (2003).

[33] W. Ha, Ph.D. thesis, Stanford Univ. (2003).

[34] V. Lordi, Ph.D. thesis, Stanford Univ. (2004).

[35] M. A. Wistey, Ph.D. thesis, Stanford Univ. (2004).

[36] S. R. Bank, Ph.D. thesis, Stanford Univ. (2006).

[37] A. Cho, J. Vac. Sci. Technol. 8, S31 (1971).

[38] A. Cho and J. Arthur, Prog. Solid-State Chem. 10, 157 (1975).

[39] J. DeBoeck, W. VanRoy, V. Motsnyi, Z. Liu, K. Dessein, and G. Borghs, Thin

Solid Films 412, 3 (2002).

168 BIBLIOGRAPHY

[40] M. B. Nardelli, F. J. Walker, and R. A. McKee, Phys. Stat. Sol. 241, 2279

(2004).

[41] E. Kasper, H. Kibbel, H. Jorke, H. Brugger, E. Friess, and G. Abstreiter, Phys.

Rev. B 38, 3599 (1988).

[42] L. Kolodziejski, R. Gunshor, N. Otsuka, S. Datta, W. Becker, and A. Nur-

mikko, IEEE J. Quantum Electron. 22, 1666 (1986).

[43] K. Nanbu, J. Saito, T. Ishikawa, K. Kondo, and A. Shibatomi, J. Electrochem.

Soc. 133, 601 (1986).

[44] C. T. Lee and Y. C. Chou, J. Cryst. Growth 91, 169 (1988).

[45] D. G. Schlom, W. S. Lee, T. Ma, and J. S. Harris, J. Vac. Sci. Technol. B 7,

296 (1989).

[46] T. J. Mattord, M. M. Oye, D. Gotthold, C. Hansing, A. L. Holmes, and B. G.

Streetman, J. Vac. Sci. Technol. A 22, 735 (2004).

[47] B. A. Joyce, Rep. Prog. Phys. 48, 1637 (1985).

[48] R. F. C. Farrow, Molecular Beam Epitaxy: Applications to Key Materials

(Noyes Publications, New York, NY, 1995).

[49] P. D. Brewer, D. H. Chow, and R. H. Miles, J. Vac. Sci. Technol. B 14, 2335

(1996).

[50] I. Ecker and S. Menzel, http://www-opto.e-technik.uni-

ulm.de/forschung/jahresbericht/1998/ar98ie.pdf.

[51] Y. Qiu, C. Jin, S. Francoeur, S. A. Nikishin, and H. Temkin, Appl. Phys. Lett

72, 1999 (1998).

[52] D. Korakakis, H. M. Ng, M. Misra, W. Grieshaber, and T. D. Moustakas, MRS

Internet J. Nitride Semicond. Res. 1, 10 (1996).

BIBLIOGRAPHY 169

[53] University of California at Berkeley Microlab Foundry,

http://microlab.berkeley.edu/iml/ichar/Combined.htm.

[54] P. Zhong, L. Lian-He, X. Ying-Qiang, Z. Wei, L. Yao-Wang, Z. Rui-Kang,

Z. Yuan, and R. Xiao-Min, Chinese Phys. Lett. 18, 1249 (2001).

[55] V. Kirchner, H. Heinke, U. Birkle, S. Einfeldt, D. Hommel, H. Selke, and P. L.

Ryder, Phys. Rev. B 58, 15749 (1998).

[56] S. G. Spruytte, M. C. Larson, W. Wampler, C. W. Coldren, H. E. Petersen,

and J. S. Harris, J. Cryst. Growth 227-228, 506 (2001).

[57] M. A. Wistey, S. R. Bank, H. B. Yuen, H. P. Bae, and J. S. Harris, J. Cryst.

Growth 278, 229 (2005).

[58] M. A. Wistey, S. R. Bank, H. B. Yuen, and J. S. Harris, in 2003 North American

MBE Conference (Boulder, CO, 2003).

[59] K. Volz, T. Torunski, and W. Stolz, J. Appl. Phys. 97, 014306 (2005).

[60] X. Kong, A. Trampert, E. Tournie, and K. H. Ploog, Appl. Phys. Lett 87,

171901 (2005).

[61] T. Gugov, V. Gambin, M. A. Wistey, H. B. Yuen, S. R. Bank, and J. S. Harris,

J. Vac. Sci. Technol. B 22, 1588 (2004).

[62] E. Tournie and K. H. Ploog, Thin Solid Films 231, 43 (1993).

[63] L. L. Schram, The Language of Colloid and Interfacial Science (American

Chemical Society, Calgary, Alberta, Canada, 1993).

[64] M. Copel, M. C. Reuter, E. Kaxiras, and R. M. Tromp, Phys. Rev. Lett 63,

632 (1989).

[65] J. Massies and N. Grandjean, Phys. Rev. B 48, 8502 (1993).

[66] E. Tournie, N. Grandjean, A. Trampert, J. Massies, and K. H. Ploog, J. Cryst.

Growth 150, 460 (1995).

170 BIBLIOGRAPHY

[67] D. Rioux and H. Hochst, Phys. Rev. B 46, 6857 (1992).

[68] A. Sakai, T. Tatsumi, and K. Ishida, Phys. Rev. B 47, 6803 (1993).

[69] W. Dondl, G. Lutjering, W. Wegscheider, J. Wilhelm, R. Schorer, and G. Ab-

streiter, J. Cryst. Growth 127, 440 (1993).

[70] H. J. Osten, J. Klatt, G. Lippert, E. Bugiel, and S. Higuchi, J. Cryst. Growth

74, 2507 (1993).

[71] X. Yang, M. J. Jurkovic, J. B. Heroux, and W. I. Wang, Appl. Phys. Lett 75,

178 (1999).

[72] H. Shimizu, K. Kumada, S. Uchiyama, and A. Kasukawa, Electron. Lett. 36,

1379 (2000).

[73] J. K. Shurtleff, S. W. Jun, and G. B. Stringfellow, Appl. Phys. Lett 78, 3038

(2001).

[74] V. Gambin, W. Ha, M. A. Wistey, H. B. Yuen, S. R. Bank, S. M. Kim, and

J. S. Harris, IEEE J. Sel. Top. Quantum Electron. 8, 795 (2002).

[75] L. Zhang, H. F. Tang, J. Schieke, M. Mavrikakis, and T. F. Kuech, J. Appl.

Phys. 92, 2304 (2002).

[76] S. W. Jun, G. B. Stringfellow, J. K. Shurtleff, and R. T. Lee, J. Cryst. Growth

235, 15 (2002).

[77] F. Dimroth, A. Howard, J. K. Shurtleff, and G. B. Stringfellow, J. Appl. Phys.

91, 3687 (2002).

[78] R. R. Wixom, N. A. Modine, and G. B. Stringfellow, Phys. Rev. B 67, 115309

(2003).

[79] T. Kageyama, T. Miyamoto, M. Ohta, T. Matsuura, Y. Matsui, T. Furuhata,

and F. Koyama, J. Appl. Phys. 96, 44 (2004).

[80] G. B. Stringfellow, J. Electrochem. Soc. 119, 1780 (1972).

BIBLIOGRAPHY 171

[81] J. Neugebauer and C. G. VandeWalle, Phys. Rev. B 51, 10568 (1995).

[82] I. Ho and G. B. Stringfellow, J. Cryst. Growth 178, 1 (1997).

[83] X. Yang, J. B. Heroux, L. F. Mei, and W. I. Wang, Appl. Phys. Lett 78, 4068

(2001).

[84] K. Volz, V. Gambin, W. Ha, M. A. Wistey, H. B. Yuen, S. R. Bank, and J. S.

Harris, J. Cryst. Growth 251, 360 (2003).

[85] J. C. Harmand, G. Ungaro, L. Largeau, and G. LeRoux, Appl. Phys. Lett 77,

2482 (2000).

[86] J. C. Harmand, A. Caliman, E. V. K. Rao, L. Largeau, J. Ramos, R. Teissier,

L. Travers, G. Ungaro, B. Theys, and I. F. L. Dias, Semicond. Sci. Technol.

17, 778 (2002).

[87] W. Braun, Applied RHEED: Reflection High-Energy Electron Diffraction Dur-

ing Crystal Growth (Springer-Verlag, New York, NY, 1999).

[88] A. Ichimiya and P. I. Cohen, Reflection High-Energy Electron Diffraction

(Cambridge University Press, New York, NY, 2004).

[89] K. T. Faber and K. J. Malloy, Mechanical Properties of Semiconductors and

Semimetals (Academic Press, San Diego, CA, 1992).

[90] G. Bauer, J. H. Li, and V. Holy, Acta Phys. Pol. A 89, 115 (1996).

[91] B. E. Warren, X-Ray Diffraction (Dover Publications, New York, NY, 1990).

[92] U. Pietsch, V. Holy, and T. Baumbach, High-Resolution X-Ray Scattering

(Springer-Verlag, New York, NY, 2004).

[93] S. G. Spruytte, C. W. Coldren, W. Wampler, P. Krispin, M. C. Larson,

K. Ploog, and J. S. Harris, J. Appl. Phys. 89, 4401 (2001).

[94] A. Pornarico, M. Lomascolo, R. Cingolani, A. Egorov, and H. Riechert, Semi-

cond. Sci. Technol. 17, 145 (2002).

172 BIBLIOGRAPHY

[95] V. Gambin, V. Lordi, W. Ha, M. Wistey, T. Takizawa, K. Uno, S. Friedrich,

and J. Harris, J. Cryst. Growth 251, 408 (2003).

[96] V. Lordi, V. Gambin, S. Friedrich, T. Funk, T. Takizawa, K. Uno, and J. Harris,

Phys. Rev. Lett 90, 145505 (2003).

[97] J. M. Chauveau, A. Trampert, M. A. Pinault, E. Tournie, K. Du, and K. H.

Ploog, J. Cryst. Growth 251, 383 (2003).

[98] E. V. Rao, A. Ougazzaden, Y. L. Bellego, and M. Juhel, Appl. Phys. Lett 72,

1409 (1998).

[99] H. P. Xin, K. L. Kavanagh, M. Kondow, and C. W. Tu, J. Cryst. Growth

201-202, 419 (1999).

[100] W. Li, M. Pessa, T. Ahlgren, and J. Decker, Appl. Phys. Lett 79, 1094 (2001).

[101] P. Klar, H. Gruning, J. Koch, S. Schafer, K. Volz, W. Stolz, W. Heimbrodt,

A. Saadi, A. Lindsay, and E. O’Reilly, Phys. Rev. B 64, 121203 (2001).

[102] M. Cardona, Modulation Spectroscopy (Academic Press, New York, NY, 1969).

[103] O. J. Glembocky and B. V. Shanabrook (Academic Press, 1992), The Spec-

troscopy of Semiconductors.

[104] Z. Pan, L. H. Li, Y. W. Lin, B. Q. Sun, D. S. Jiang, and W. K. Ge, Appl.

Phys. Lett 78, 2217 (2001).

[105] J. Hader, S. W. Koch, J. V. Moloney, and E. P. O’Reilly, Appl. Phys. Lett 76,

3685 (2000).

[106] I. A. Buyanova, G. Pozina, P. N. Hai, W. M. Chen, H. P. Xin, and C. W. Tu,

Phys. Rev. B 63, 033303 (2001).

[107] J. Misiewicz, P. Sitarek, G. Sek, and R. Kudrawiec, Mater. Sci. 21, 263 (2003).

[108] A. J. Ptak, K. S. Ziemer, M. R. Millecchia, C. D. Stinespring, and T. H. Meyers,

MRS Internet J. Nitride Semicond. Res. 4S1, G.3.10 (1999).

BIBLIOGRAPHY 173

[109] A. J. Ptak, M. R. Millecchia, T. H. Meyers, K. S. Ziemer, and C. D. Stinespring,

Appl. Phys. Lett 74, 3836 (1999).

[110] T. Kageyama, T. Miyamoto, S. Makino, F. Koyama, and K. Iga, J. Cryst.

Growth 209, 350 (2000).

[111] M. Kondow, S. Natatsuka, T. Kitatani, Y. Yazawa, and M. Okai, Electron.

Lett. 32, 2244 (1996).

[112] K. Nakahara, K. Kondow, T. Kitatani, Y. Yazawa, and K. Uomi, Electron.

Lett. 32, 1585 (1996).

[113] C. W. Coldren, S. G. Spruytte, and J. S. Harris, J. Vac. Sci. Technol. B 18,

1480 (2000).

[114] M. A. Wistey, S. R. Bank, H. B. Yuen, L. L. Goddard, and J. S. Harris,

Electron. Lett. 39, 1822 (2003).

[115] S. R. Bank, M. A. Wistey, L. L. Goddard, H. B. Yuen, V. Lordi, and J. S.

Harris, IEEE J. Quantum Electron. 40, 656 (2004).

[116] J. A. Gupta, P. J. Barrios, X. Zhang, J. Lapointe, D. Poitras, G. Pakulski,

X. Wu, and A. Delage, Electron. Lett. 41, 1060 (2005).

[117] G. Jaschke, R. Averbeck, L. Geelhaar, and H. Riechert, J. Cryst. Growth 278,

224 (2005).

[118] N. Tansu and L. J. Mawst, IEEE Photon. Technol. Lett. 14, 444 (2002).

[119] N. Tansu, N. J. Kirsch, and L. J. Mawst, Appl. Phys. Lett 81, 2523 (2002).

[120] H. Y. Liu, M. Hopkinson, P. Navaretti, M. Gutierrez, J. S. Ng, and J. P. R.

David, Appl. Phys. Lett 83, 4951 (2003).

[121] X. Sun, S. Wang, J. Hsu, R. Sidhu, X. Zheng, X. Li, J. Campbell, and

A. Holmes, IEEE J. Sel. Top. Quantum Electron. 8, 817 (2002).

174 BIBLIOGRAPHY

[122] L. H. Li, V. Sallet, G. Patriarche, L. Largeau, S. Bouchoule, K. Merghem,

L. Travers, and J. C. Harmand, Electron. Lett. 39, 519 (2003).

[123] M. Kaminska, Z. Liliental-Weber, E. R. Weber, T. George, J. B. Kortright,

F. W. Smith, B. Y. Tsaur, and A. R. Calawa, Appl. Phys. Lett 54, 1881

(1989).

[124] A. Suda and N. Otsuka, Appl. Phys. Lett 73, 1529 (1998).

[125] A. A. Bernussi, C. F. Souza, W. Carvalho, D. I. Lubyshev, J. C. Rossi, and

P. Basmaji, Brazilian J. Phys. 24, 460 (1994).

[126] Z. Liliental-Weber, X. W. Lin, J. Washburn, and W. Schaff, Appl. Phys. Lett

66, 2086 (1995).

[127] B. K. Meyer, D. M. Hofmann, J. R. Niklas, and J. M. Spaeth, Phys. Rev. B

36, 1332 (1987).

[128] J. Dabrowski and M. Scheffler, Phys. Rev. B 40, 10391 (1989).

[129] W. Ha, V. Gambin, M. A. Wistey, S. R. Bank, H. B. Yuen, S. M. Kim, and

J. S. Harris, Electron. Lett. 38, 277 (2002).

[130] S. R. Bank, H. B. Yuen, M. A. Wistey, W. Ha, V. F. Gambin, and J. S. Harris,

in 2003 Electronic Materials Conference (Salt Lake City, UT, 2003).

[131] S. R. Kurtz, A. A. Allerman, E. D. Jones, J. M. Gee, J. J. Banas, and B. E.

Hammons, Appl. Phys. Lett 74, 729 (1999).

[132] K. Volz, J. Koch, B. Kunert, and W. Stolz, J. Cryst. Growth 248, 451 (2003).

[133] A. J. Ptak, S. W. Johnston, S. Kurtz, and D. J. Friedman, J. Cryst. Growth

251, 392 (2003).

[134] A. J. Ptak, D. J. Friedman, S. Kurtz, and R. C. Reedy, J. Appl. Phys. 98,

94501 (2005).

BIBLIOGRAPHY 175

[135] G. Leibiger, C. Krahmer, J. Bauer, H. Herrnberger, and V. Gottschalch, J.

Cryst. Growth 272, 732 (2004).

[136] B. Damilano, J. Barjon, S. W. Wan, J. Y. Duboz, M. Leroux, M. Laugt, and

J. Massies, IEE Proc. Optoelectron 151, 433 (2004).

[137] K. Nishimura, H. S. Lee, H. Suzuki, I. Gono, N. Kojima, Y. Ohshita, and

M. Yamaguchi, in 31st IEEE Photovoltaic Specialists Conference (Lake Buena

Vista, FL, 2005).

[138] A. Ohmae, Y. Shimizu, and Y. Okada, in 3rd World Conference on Photovoltaic

Energy Conversion (Osaka, Japan, 2003).

[139] S. R. Bank, H. B. Yuen, H. P. Bae, M. A. Wistey, L. L. Goddard, and J. S.

Harris, in 2005 North American MBE Conference (Santa Barbara, CA, 2005).

[140] M. A. Wistey (2006), submitted to Electron. Lett.

[141] L. H. Li, V. Sallet, G. Patriarche, L. Largeau, S. Bouchoule, L. Travers, and

J. C. Harmand, Appl. Phys. Lett 83, 1298 (2003).

[142] I. T. Ferguson, A. G. Norman, B. A. Joyce, T. Y. Seong, G. R. Booker, R. H.

Thomas, C. C. Philips, and R. A. Stradling, Appl. Phys. Lett 59, 3324 (1991).

[143] A. G. Norman, T. Y. Seong, I. T. Ferguson, G. R. Booker, and B. A. Joyce,

Semicond. Sci. Technol. 8, S9 (1993).

[144] A. A. ElEmawy, H. J. Cao, E. Zhmayev, J. H. Lee, D. Zubia, and M. Osiski,

Phys. Stat. Sol. 228, 263 (2001).

[145] T. D. Veal, I. Mahboob, , and C. F. McConville, Phys. Rev. Lett 92, 136801

(2004).

[146] H. B. Yuen, F. Hatami, S. Kim, J. S. Harris, A. Chin, M. Ferhat, K. Yoh, and

A. Moto, in 2005 Electronic Materials Conference (Santa Barbara, CA, 2005).

[147] I. Mahboob, T. D. Veal, and C. F. McConville, J. Appl. Phys. 96, 4935 (2004).

176 BIBLIOGRAPHY

[148] I. Mahboob, T. D. Veal, and C. F. McConville, Appl. Phys. Lett 83, 2169

(2003).

[149] W. Li, J. B. Heroux, and W. I. Wang, J. Appl. Phys. 94, 4248 (2003).

[150] S. R. Kurtz, D. Myers, and J. M. Olson, in Projected Performance of Three-and

Four-Junction Devices using GaAs and GaInP (Anaheim, CA, 1997).

[151] A. J. Ptak, private communication, unpublished (2005).

BIBLIOGRAPHY 177

MBE

me

Seth

”We ♥ LATEX!”

”Go Bears!!!”

Documents

growth and characterization of dilute nitride antimonides for long