Document [original]

Metalorganic Chemical Vapor

Deposition of

High-Performance GaAs-Based

Quantum-Dot Lasers

vorgelegt von

Diplom-Physiker

Roman Sellin

aus Ulm a. d. Donau

von der Fakultät II

– Mathematik und Naturwissenschaften –

der Technischen Universität Berlin

zur Erlangung des akademischen Grades

Doktor der Naturwissenschaften

- Dr. rer. nat. -

genehmigte Dissertation

Arbeit eingereicht am 6. Mai 2003

Tag der wissenschaftlichen Aussprache: 2. Juli 2003

Vorsitzender der Prüfungskommission: Prof. Dr. P. Zimmermann

Berichter: Prof. Dr. D. Bimberg und Prof. Dr. A. Krost

Berlin 2003

D83

iii

Abstract

In this work, Metalorganic Chemical Vapor Deposition (MOCVD) of

novel GaAs-based semiconductor laser structures with self-organized In-

GaAs/GaAs Stranski-Krastanow quantum dots (QDs) as active medium was

advanced with regard to the laser characteristics. The three-dimensional

morphology of self-organized QDs leads to a significant roughening of thin

cap layers on top of QD sheets. Smoother QD cap layers are required, how-

ever, to reduce the distance between stacked QD layers and thus to increase

the QD volume density for larger modal gain of QD lasers. Hence, the

growth of QD lasers was complemented by an in-situ annealing step flatten-

ing such corrugated surfaces. The strain of lattice-mismatched self-

organized QDs and the untypically low QD deposition temperatures around

500°C lead to dislocations and point defects in QD heterostructures. The

density of such defects was strongly reduced by in-situ annealing. Lasers

with in-situ annealed QDs exhibit room-temperature transparency current

densities around 6 A/cm2 per QD sheet at emission wavelengths between

1.14 and 1.16 µm. The internal quantum efficiency was increased to beyond

90 %. Lasers based on 6-fold stacks of such in-situ annealed QDs show

room-temperature peak output powers of 11.7 W in quasi-continuous-wave

mode and 4.7 W under continuous-wave operation. This was the first dem-

onstration of optical output powers of QD lasers beyond 10 W. The charac-

teristics of such QD lasers did not exhibit significant changes during life-

time measurements of more than 3000 h at 50°C and output powers of 1.0 -

1.5 W.

Arsine, widely used as arsenic precursor in MOCVD, is strongly toxic

and was therefore replaced in the course of this work by the alternative pre-

cursor tertiarybutylarsine (TBAs). The growth of QDs had to be recali-

brated as the physical and chemical properties of TBAs differ from those of

arsine. The worldwide first QD laser grown using alternative-precursor

MOCVD could be demonstrated. Different techniques to grow QDs emit-

ting at the commercially important data communication wavelength of

1.3 µm were developed and evaluated. Such QD structures were investi-

gated using photoluminescence spectroscopy and transmission electron mi-

croscopy. Using InGaAs QDs overgrown with gallium-rich InGaAs quan-

tum films, the room-temperature lasing wavelength could be extended to

1.24 µm. The growth of laser structures for the fabrication of QD-based sur-

face emitting lasers (VCSELs) with Al(Ga)Ox/GaAs oxide mirrors and

nine-fold stacks of InGaAs/GaAs QDs as active region was implemented. A

VCSEL with a 3.5 µm aperture and four top-DBR pairs exhibited a maxi-

mum output power of 0.68 mW at 1.1 µm, a threshold current of 280 µA,

and a differential efficiency of 43 %.

Zusammenfassung

Im Rahmen dieser Arbeit wurde eine Verbesserung der metallorgani-

schen Gasphasenepitaxie (MOCVD) neuartiger GaAs-basierter Halbleiter-

Laserstrukturen mit selbstorganisierten InGaAs/GaAs Stranski-Krastanow-

Quantenpunkten (QPen) als aktivem Medium hinsichtlich der Lasereigen-

schaften erzielt. Die dreidimensionale QP-Morphologie führt zu einer Auf-

rauhung dünner Bedeckungsschichten auf den Quantenpunkten. Glatte Be-

deckungsschichten sind jedoch für kleine Abstände zwischen gestapelten

QP-Schichten erforderlich, um die Volumendichte an QPen und damit den

modalen Gewinn von QP-Lasern zu erhöhen. Daher wurde ein in-situ Tem-

perverfahren zur Glättung solch rauer Oberflächen eingeführt. Die Ver-

spannung von QP-Strukturen und die niedrigen QP-Wachstumstempe-

raturen um 500°C führen zu Versetzungen und Punktdefekten. Durch den

Temperschritt wurde die Dichte solcher Defekte drastisch reduziert. Laser

mit getemperten QPen zeigen bei Raumtemperatur Transparenzstromdich-

ten um 6 A/cm2 pro QP-Schicht bei Emissionswellenlängen zwischen 1.14

und 1.16 µm. Die interne Quanteneffizienz solcher Laser liegt über 90 %.

Laserdioden mit Sechsfachstapeln getemperter QPe zeigen Ausgangsleis-

tungen von 11.7 W im Quasi-Dauerstrichbetrieb und 4.7 W im Dauerstrich-

betrieb. Damit konnten für QP-Laser erstmalig Ausgangsleistungen von ü-

ber 10 W erreicht werden. Die Charakteristik solcher Laser blieb während

einer über 3000-stündigen Lebensdauermessung bei 50°C und 1.0-1.5 W

optischer Ausgangsleistung unverändert.

Das als Arsen-Vorläufer weit verbreitete Arsin ist hochtoxisch und wur-

de daher im Laufe dieser Arbeit durch den alternativen Vorläufer Tertiärbu-

tylarsin (TBAs) ersetzt. Das Wachstum von QPen musste daraufhin neu ka-

libriert werden, da TBAs andere physikalische und chemische Eigenschaf-

ten als Arsin besitzt. Weltweit wurden erstmals QP-Laser mit dem alterna-

tiven Precursor TBAs hergestellt. Verschiedene Verfahren zum Wachstum

von QPen wurden entwickelt, die bei der kommerziell wichtigen Telekom-

munikations-Wellenlänge von 1.3 µm emittieren. Solche QP-Strukturen

wurden mit Photolumineszenzspektroskopie und Transmissions-

elektronenmikroskopie untersucht. Mit InGaAs/GaAs QPen, die mit Galli-

um-reichen InGaAs Quantenfilmen überwachsen wurden, konnte die

Raumtemperatur-Lasingwellenlänge zu 1.24 µm ausgedehnt werden. Des

weiteren wurden Verfahren zum Wachstum von QP-Strukturen für die Her-

stellung von oberflächenemittierenden Lasern (VCSEL) mit

Al(Ga)Ox/GaAs-Oxidspiegeln entwickelt. Ein VCSEL mit einer 3.5 µm

breiten Apertur und vier oberen Spiegelpaaren zeigte eine maximale Aus-

gangsleistung von 0.68 mW bei 1.1 µm und eine differenzielle Effizienz

von 43 %. Der Schwellstrom betrug 280 µA.

vii

Parts of this work have been published in:

R.L. Sellin, I. Kaiander, D. Ouyang, T. Kettler, and D. Bimberg, Alterna-

tive-precursor metalorganic chemical vapor deposition and laser applica-

tion of self-organized InGaAs/GaAs quantum dots, Appl. Phys. Lett. 82,

841 (2003).

A. Lenz, R. Timm, H. Eisele, C. Hennig, S. K. Becker, R.L. Sellin, U.W.

Pohl, D. Bimberg, and M. Dähne, Reversed truncated cone composition dis-

tribution of In0.8Ga0.2As quantum dots overgrown by an In0.1Ga0.9As layer in

a GaAs matrix, Appl. Phys. Lett. 81, 5150 (2002).

R.L. Sellin, C. Ribbat, D. Bimberg, F. Rinner, H. Konstanzer, M.T. Kele-

men, and M. Mikulla, High-reliability MOCVD-grown quantum dot laser,

El. Lett. 38, 883 (2002).

C. Ribbat and R. Sellin, High Power Quantum Dot Lasers in Nano-

Optoelectronics, ed. by M. Grundmann (Springer, Berlin, Heidelberg,

2002), p. 353.

R. Sellin, N.N. Ledentsov, D. Bimberg, V.M. Ustinov, and Z.I. Alferov,

Growth and optical characterization of long-wavelength quantum dots for

low-threshold current lasers, Proc. of the 6th Int. Symp. Adv. Phys. Fields,

Tsukuba, Japan, March 6-9 (2001), p. 49.

R.L. Sellin, C. Ribbat, M. Grundmann, N.N. Ledentsov, and D. Bimberg,

Close-to-ideal device characteristics of high-power InGaAs/GaAs quantum

dot lasers, Appl. Phys. Lett. 78, 1207 (2001).

D. Bimberg, M. Grundmann, N.N. Ledentsov, M.H. Mao, C. Ribbat,

R. Sellin, V.M. Ustinov, A.E. Zhukov, Z.I. Alferov, and J.A. Lott, Novel

Infrared Quantum Dot Lasers: Theory and Reality, phys. stat. sol. (b) 224,

787 (2001).

C. Ribbat, R. Sellin, M. Grundmann, and D. Bimberg, High Power Quan-

tum Dot Lasers at 1160 nm, phys. stat. sol. (b) 224, 819 (2001).

R. Sellin, N.N. Ledentsov, and D. Bimberg, Verfahren zur Verbesserung

der Effizienz von epitaktisch hergestellten Quantenpunkt-Halbleiterbauele-

menten mit einer oder mehreren Quantenpunktschichten, patent pending,

ref. no. 10044040 (DE) (2000).

R. Sellin, F. Heinrichsdorff, C. Ribbat, M. Grundmann, U.W. Pohl, and D.

Bimberg, Surface flattening during MOCVD of thin GaAs layers covering

InGaAs quantum dots, J. Crystal Growth 221, 581 (2000).

viii

Further publications (only refereed journals):

C. Ribbat, R. Sellin, I. Kaiander, F. Hopfer, N.N. Ledentsov, D. Bimberg,

A.R. Kovsh, V.M. Ustinov, A.E. Zhukov, and M.V. Maximov, Complete

suppression of filamentation and superior beam quality in quantum-dot la-

sers, Appl. Phys. Lett. 82, 952 (2003).

F. Guffarth, R. Heitz, M. Geller, C. Kapteyn, H. Born, R. Sellin, A. Hoff-

mann, and D. Bimberg, Radiation hardness of InGaAs/GaAs quantum dots,

Appl. Phys. Lett. 82, 1941 (2003).

S. Rodt, R. Heitz, A. Schliwa, R. L. Sellin, F. Guffarth, and D. Bimberg,

Repulsive exciton-exciton interaction in quantum dots, Phys. Rev. B 68,

035331 (2003).

F. Guffarth, R. Heitz, A. Schliwa, O. Stier, M. Geller, C.M.A. Kapteyn, R.

Sellin, and D. Bimberg, Few-particle interactions in charged InGaAs/GaAs

quantum dots, Phys. Rev. B 67, 235304 (2003).

L. Müller-Kirsch, N.N. Ledentsov, R. Sellin, U.W. Pohl, D. Bimberg, I.

Häusler, H. Kirmse, and W. Neumann, GaSb quantum dot growth using

InAs quantum dot stressors, J. Crystal Growth 248, 333 (2003).

P. Borri, W. Langbein, S. Schneider, U. Woggon, R.L. Sellin, D. Ouyang,

and D. Bimberg, Coherent Light-Matter Interaction in InGaAs Quantum

Dots: Dephasing Time and Optical Rabi Oscillations, phys. stat. sol (b)

233, 391 (2002).

P. Borri, W. Langbein, S. Schneider, U. Woggon, R.L. Sellin, D. Ouyang,

and D. Bimberg, Temperature-Dependent Time-Resolved Four-Wave Mix-

ing in InGaAs Quantum Dots, phys. stat. sol. (a) 190, 517 (2002).

P. Borri, W. Langbein, S. Schneider, U. Woggon, R. Sellin, D. Ouyang, and

D. Bimberg, Relaxation and dephasing of multiexcitons in semiconductor

quantum dots, Phys. Rev. Lett. 89, 187401 (2002).

P. Borri, W. Langbein, S. Schneider, U. Woggon, R.L. Sellin, D. Ouyang,

and D. Bimberg, Rabi oscillations in the excitonic ground-state transition

of InGaAs quantum dots, Phys. Rev. B 66, 081306R (2002).

P. Borri, W. Langbein, S. Schneider, U. Woggon, R.L. Sellin, D. Ouyang,

and D. Bimberg, Coherent Light-Matter Interaction in InGaAs Quantum

Dots: Dephasing Time and Optical Rabi Oscillations, phys. stat. sol. (b)

233, 391 (2002).

V.A. Haisler, F. Hopfer, R.L. Sellin, A. Lochmann, K. Fleischer, N. Esser,

W. Richter, N.N. Ledentsov, D. Bimberg, C. Möller, and N. Grote, Micro-

Raman studies of vertical-cavity surface-emitting lasers with AlxOy/GaAs

distributed Bragg reflectors, Appl. Phys. Lett. 81, 2544 (2002).

D. Ouyang, R. Heitz, N.N. Ledentsov, S. Bognár, R.L. Sellin, C. Ribbat,

and D. Bimberg, Lateral-cavity spectral hole burning in quantum dot la-

sers, Appl. Phys. Lett. 81, 1546 (2002).

C. Ribbat, S. Bognár, R.L. Sellin, and D. Bimberg, Spectral mode dynamics

of short cavity quantum-dot lasers, Appl. Phys. Lett. 81, 147 (2002).

D.S. Sizov, M.V. Maksimov, A.F. Tsatsul’nikov, N.A. Cherkashin, N.V.

Kryzhanovskaya, A.B. Zhukov, N.A. Maleev, S.S. Mikhrin, A.P. Vasil’ev,

R. Sellin, V.M. Ustinov, N.N. Ledentsov, D. Bimberg, and Z.I. Alferov,

The Influence of Heat Treatment Conditions on the Evaporation of Defect

Regions in Structures with InGaAs Quantum Dots in the GaAs Matrix,

Semiconductors 36, 1020 (2002).

S. Rodt, A. Schliwa, V. Türck, R. Heitz, O. Stier, R.L. Sellin, M. Strass-

burg, U.W. Pohl, and D. Bimberg, Few-particle effects in self-organized

quantum dots, phys. stat. sol. (b) 234, 354 (2002).

P. Borri, W. Langbein, S. Schneider, U. Woggon, R.L. Sellin, D. Ouyang,

and D. Bimberg, Exciton relaxation and dephasing in quantum-dot amplifi-

ers from room to cryogenic temperature, J. Sel. Top. Quantum El. 8, 984

(2002).

S. Bognár, M. Grundmann, O. Stier, D. Ouyang, C. Ribbat, R. Heitz,

R. Sellin, and D. Bimberg, Large Modal Gain of InAs/GaAs Quantum Dot

Lasers, phys. stat. sol. (b) 224, 823 (2001).

P. Borri, W. Langbein, S. Schneider, U. Woggon, R.L. Sellin, D. Ouyang,

and D. Bimberg, Ultralong Dephasing Time in InGaAs Quantum Dots,

Phys. Rev. Lett. 87, 157401 (2001).

H.Y. Ryu, Y.H. Lee, R.L. Sellin, and D. Bimberg, Over 30-fold enhance-

ment of light extraction from free-standing photonic crystal slabs with In-

GaAs quantum dots at low temperature, Appl. Phys. Lett. 79, 3573 (2001).

C. Ribbat, R. Sellin, M. Grundmann, D. Bimberg, N.A. Sobolev, and M.C.

Carmo, Enhanced radiation hardness of quantum dot lasers to high energy

proton irradiation, El. Lett. 37, 174 (2001).

Acknowledgements

I would like to thank Prof. Dr. Dieter Bimberg who has offered me to

work in his group on a very fascinating and challenging topic with increas-

ing importance for applied physics and device research, and for his vivid

interest in the progress of my work. Indefatigable fundraising has consid-

erably facilitated the costly business of MOCVD growth. He has given me

numerous opportunities to present my work at international conferences.

During my work in the management and administration of several third-

party-funded projects, I was given the chance to cooperate with very skillful

and interesting researchers from all over the world.

Prof. Dr. Marius Grundmann accompanied my work at the beginning,

and I would like to thank him for many interesting discussions. Prof. Dr.

Nikolai N. Ledentsov provided assistance to my work with many concrete

suggestions and an ampleness of ideas. I benefited from his large experi-

ence with MBE-grown QD lasers. Dr. Frank Heinrichsdorff was my prede-

cessor as an MOCVD quantum-dot device grower, and also my tutor during

the first months of my work. During our joint period in the MOCVD labo-

ratory, he conveyed a precious deal of his practical experience to me. I

would like to thank Dr. Udo W. Pohl for his commitment in the configura-

tion of the new Aixtron 200/4 MOCVD machine as well as in the estab-

lishment of the new MOCVD laboratory. His didactical skills have been a

great help in writing manuscripts and in preparing conference abstracts and

talks. I want to thank Dr. Robert Heitz for many interesting discussions re-

garding quantum-dot physics. His personal views and his open critical notes

have given me valuable impulses.

Dipl.-Krist. Kathrin Schatke has technically assisted the MOCVD ex-

periments, and I would like to thank her for her unselfish commitment. I

would also like to thank Ilona Gründler for her assistance. I have further-

more benefited a lot from the large MOCVD experience of Dr. Armin

Dadgar who supported me with numerous tips and tricks. I would like to

thank Dr. André Strittmatter for his help on atomic force microscopy and x-

ray diffractometry. I am also grateful to Dr. Lutz Müller-Kirsch for many

interesting and helpful discussions.

I am particularly indebted to Dipl.-Phys. Ilia Kaiander for the intensive

and fruitful cooperation in the field of epitaxy during the last year of my

work. I enjoyed the cooperation with Dipl.-Phys. Thorsten Kettler on In-

GaAsN quantum dots and with Dipl.-Phys. Konstantin Pötschke on InAsSb

quantum-dot structures. I want to thank Dipl.-Phys. Florian Guffarth and

Dr. Robert Heitz for help and discussions regarding photoluminescence

spectroscopy. I am very grateful to Dr. Nikolai Zakharov and Dr. Peter

Werner from the Max-Planck Institute of Microstructure Physics, Halle,

xii

Germany, for the steady and speedy supply of numerous transmission elec-

tron micrographs. I would like to thank Dr. Christian Ribbat, M.Sc.

Dongxun Ouyang, Dipl.-Phys. Oliver Schulz and Dipl.-Phys. Thorsten Ket-

tler for the fabrication and characterization of quantum-dot edge-emitting

lasers, and Dipl.-Phys. Friedhelm Hopfer and Dipl.-Phys. Anatol Lochmann

for processing and characterizing surface-emitting lasers.

I would like to thank Dipl.-Phys. Sven Rodt for careful proofreading of

the manuscript and Prof. Dr. Alois Krost for preparing the second expert

opinion on my thesis. I would like to express my gratitude to Dipl.-Phys.

Sven Rodt, Dr. Volker Türck and Dr. Armin Dadgar for administering and

maintaining a very reliable computer network. I enjoyed sharing my office

with Andrei Schliwa, Florian Guffarth and Ilia Kaiander. Parts of this work

were funded by the EU project DOTCOM, the German Federal Ministry of

Education and Research (bmb+f), and the collaborative research center

Sfb 296 of the German Research Foundation (DFG). Last not least I am in-

debted to Agilent Technologies for a grant within the frame of an external

research programme.

xiii

Table of contents

1. Introduction....................................................................................................... 1

2. Lasers and quantum dots ................................................................................. 3

2.1. Semiconductor lasers ................................................................................... 3

2.2. Quantum dots ............................................................................................... 6

2.3. Properties of quantum-dot lasers................................................................. 7

2.4. MOCVD for quantum-dot lasers................................................................ 10

2.5. Objectives and methods of this work.......................................................... 11

3. Metalorganic chemical vapor deposition ...................................................... 15

3.1. Principle..................................................................................................... 15

3.2. Conventional and alternative precursors................................................... 16

3.2.1 Conventional-precursor MOCVD........................................................ 17

3.2.2 Alternative-precursor MOCVD ........................................................... 18

3.2.3 Tertiarybutylarsine ............................................................................... 20

3.2.3.1 Surface stabilization...................................................................... 22

3.2.3.2 V/III ratios for GaAs and AlGaAs growth.................................... 23

3.3. Modular optimization of laser structures................................................... 24

3.3.1 Waveguides and distributed Bragg reflectors ...................................... 25

3.3.1.1 Calibration of growth rate ............................................................. 25

3.3.1.2 Lateral homogeneity of layer thicknesses..................................... 27

3.3.1.3 Calibration of oxidation rate for Al(Ga)Ox/GaAs DBRs .............. 30

3.3.2 Optimization of quantum-dot active regions........................................ 31

4. Self-organization of quantum dots ................................................................ 35

4.1. The equilibrium crystal shape.................................................................... 35

4.2. Strained heteroepitaxy of thin films ........................................................... 36

4.3. Thermodynamic models of 3D island arrays ............................................. 37

4.4. Kinetic description of island formation in one dimension ......................... 39

4.5. Island size and density ............................................................................... 40

4.5.1 Role of temperature.............................................................................. 41

4.5.2 Impact of deposition amount................................................................ 43

4.5.3 Influence of growth interruption .......................................................... 43

4.5.4 Importance of growth rate.................................................................... 43

5. MOCVD of quantum-dot structures for laser diodes.................................. 45

5.1. In-situ annealing of QD structures ............................................................ 46

5.1.1 Flattening of the growth front .............................................................. 47

5.1.2 Optical properties of annealed QD structures ...................................... 52

5.2. Alternative-precursor MOCVD of InGaAs QDs........................................ 54

Table of contents

5.3. Redshift of the quantum-dot emission wavelength..................................... 56

5.3.1 Overgrowth of InGaAs QDs by InGaAs QWs using arsine ................ 57

5.3.1.1 Spectroscopic characterization...................................................... 59

5.3.1.2 Structural characterization ............................................................ 61

5.3.2 Advantage of TBAs for redshifting the QD emission wavelength ..... 67

5.3.3 Wavelength shifting using nitrogen ..................................................... 68

5.3.3.1 Simultaneous deposition of As, Ga, In and N............................... 72

5.3.3.2 Nitridation ..................................................................................... 74

5.3.4 Wavelength shifting using antimony ................................................... 76

5.3.4.1 Deposition amount ........................................................................ 76

5.3.4.2 TESb and TBAs during the growth interruption........................... 78

6. Quantum-dot lasers......................................................................................... 85

6.1. Edge emitters.............................................................................................. 85

6.1.1 Lasers with in-situ annealed quantum-dots.......................................... 85

6.1.1.1 Threshold reduction and increase of efficiency ............................ 86

6.1.1.2 High-Power Operation .................................................................. 89

6.1.1.3 Lifetimes ....................................................................................... 92

6.1.1.4 Characteristic temperature ............................................................ 93

6.1.2 Edge emitters grown with alternative precursors................................. 96

6.1.3 Long-wavelength (>1.24 µm) QD lasers ............................................. 97

6.2. Quantum-dot vertical-cavity surface emitters............................................ 99

6.2.1 Quantum-well VCSEL....................................................................... 101

6.2.2 Quantum-dot VCSEL......................................................................... 105

7. Summary and outlook................................................................................... 111

List of acronyms ................................................................................................ 115

List of figures..................................................................................................... 117

Bibliography ...................................................................................................... 125

1. Introduction

The importance of semiconductor lasers has increased dramatically dur-

ing the last decades. In compact disc and DVD players, they have become

inherent parts of everyday life. As compared to light emitted by other light

sources, laser light is highly monochromatic and by orders of magnitude

more intensive. Since laser light can more easily be focused, for example,

lasers are particularly suited as signal sources for optical fiber communica-

tion. Today, the main application of semiconductor lasers is optical data-

and telecommunication1. Their small size enables, for example, the fabrica-

tion of dense laser arrays for wavelength division multiplexing, or space di-

vision multiplexing for optical interconnects. For semiconductor lasers with

quantum dots2 (QDs) as active medium, Arakawa and Sakaki predicted al-

ready in 1982 reduced threshold current densities that are far less dependent

on operation temperature than conventional quantum well (QW) lasers3.

Larger differential gain of QD lasers was foreseen a few years later by

Asada et al.4, making them attractive as uncooled direct modulators with

larger cutoff frequencies.

QDs used as active medium for laser diodes are small coherent three-

dimensional semiconductor insertions with a fundamental band gap lower

than that of the matrix. They confine trapped carriers in all three spatial di-

rections on a length scale smaller than the exciton Bohr radius and are

therefore electronically zero-dimensional. The density of states in QDs is δ-

function-like, resulting in atom-like electronic properties. QDs thus com-

bine the advantages of semiconductors (electrical and thermal conductivity,

energy gaps in the right range etc.) with those of atoms (spectral purity,

symmetric gain spectrum).

An important advantage of QDs is the possibility to extend the emission

wavelength on gallium arsenide (GaAs) substrates up to 1.3 µm, a standard

telecommunication wavelength. Lasing emission at 1.3 µm is particularly

interesting for broadband metropolitan area fiber communication due to a

dispersion minimum of conventional glass fibers at 1.3 µm. With conven-

tional QWs, the room-temperature emission wavelength on GaAs substrates

is limited to about 1.2 µm5, 6. The demand of 1.3 µm telecom emitters has

so far been satisfied using QW lasers grown on expensive indium

phosphide (InP) substrates. Lower fabrication costs of optoelectronic de-

vices grown on GaAs substrates make the replacement of InP-based QW

lasers by GaAs-based QD lasers attractive. In addition, GaAs technology

opens the possibility to grow highly effective AlAs/GaAs distributed Bragg

reflectors (DBRs). This enables the monolithic fabrication of vertical-cavity

surface-emitting lasers (VCSELs)7. AlAs/GaAs heterostructures can selec-

tively be oxidized to obtain AlOx/GaAs DBRs. Oxide DBRs exhibit higher

reflectivities so that less AlOx/GaAs pairs are needed for a VCSEL mirror.

1. Introduction

Layer structures for the fabrication of oxide-DBR VCSELs are thinner and

require less epitaxial effort. The combination of telecom wavelengths,

GaAs technology and surface-emitting geometry in a monolithic design

makes oxide-DBR GaAs-based 1.3 µm QD VCSELs one of the most inter-

esting optoelectronic devices.

QDs for the application in optoelectronic devices are nowadays fabri-

cated using a self-organization concept2, 8: A thin semiconductor film is epi-

taxially deposited on a substrate having a different lattice constant. Driven

by the strain arising from the lattice mismatch, three-dimensional QDs are

formed by material redistribution and agglomeration at spontaneously de-

fined nucleation sites on the surface. The discovery of this so-called Stran-

ski-Krastanow (SK) growth mode was a major breakthrough in the devel-

opment of QD lasers since only self-organized QDs have the crystalline

quality required for the realization of the theoretical advantages of QD la-

sers9. Modern technologies of semiconductor crystal growth like Metalor-

ganic Chemical Vapor Deposition (MOCVD) or Molecular Beam Epitaxy

(MBE) enable the controlled, reproducible and homogeneous deposition of

ultrathin strained semiconductor films required for the fabrication of self-

organized QDs. MOCVD is the most important fabrication standard for op-

toelectronic semiconductor device epitaxy in Europe. Whereas MOCVD-

grown QD lasers have been demonstrated in the past2, 10-13, their perform-

ance was not yet close to the theoretically predicted limits. The advance-

ments of the MOCVD growth in this work have brought QD lasers further

to their theoretical limits: Threshold current densities could be reduced fur-

ther, internal quantum efficiencies of close to the ideal limit of 100 % have

been achieved, output powers of more than 10 W were demonstrated, and

lifetime tests of such lasers have proven their high reliability. Different ap-

proaches to achieve QD luminescence at 1.3 µm are introduced and dis-

cussed, including an extensive characterization of the crystalline and optical

properties of the corresponding structures. Whereas MBE-grown GaAs-

based QD lasers emitting at 1.3 µm have already been demonstrated (cf.

Ref. 14 and references therein), MOCVD-grown QD lasers with emission

wavelengths beyond 1.2 µm are presented in this work for the first time15.

MOCVD growth of QD laser structures for vertical light emission was de-

veloped within this work and has led to the demonstration of the first

MOCVD-grown oxide-DBR QD VCSEL worldwide.

The advancements of MOCVD growth of QD lasers included the substi-

tution of the highly toxic As precursor arsine and the explosive Si precursor

silane by alternative organic precursors that are significantly less hazardous

and have vapor pressures far below the atmospheric pressure. The success-

ful application of alternative-precursor MOCVD is an important step to-

wards the establishment of a safe and environment-friendly fabrication

technology of novel optoelectronic devices.

2. Lasers and quantum dots

The working principle of a laser is light amplification by stimulated

emission of radiation (laser). In a laser, the photons are not emitted inde-

pendently: a photon that is spontaneously generated in the active region of a

laser stimulates other photons to be emitted with the same phase and wave

vector. In semiconductor lasers, photons are generated by the recombination

of electrons in the conduction band with holes in the valence band. Lasing

can only set in if sufficient electrons (holes) are located in the conduction

(valence) band. This is usually achieved by electrical current injection. In a

semiconductor laser, the light-emitting active medium is centered in a reso-

nator waveguide along which the light amplification takes place. Mirrors at

both ends of the cavity provide optical feedback and lead to the formation

of a stationary light wave. Lasing can only occur if the number of photons

generated per length of the resonator is larger than the number of photons

lost by absorption and scattering, that is, if the gain overcomes the losses.

Most of today’s semiconductor lasers are based on thin low-band-gap

semiconductor insertions in the waveguide, such as nanometer-thin QWs in

which the carrier recombination takes place. A novel approach is the use of

nanometer-sized three-dimensional QDs as light-emitting centers. QD la-

sers were predicted already in the nineteen-eighties to have superior proper-

ties as compared to QW lasers3, 4. A considerable number of these advan-

tages have been demonstrated during the last years. It is the aim of this

chapter to give an introduction to the working principle of semiconductor

lasers, to the benefits of QDs as light-emitting medium, and to the ad-

vancements of the QD laser growth technology achieved during this work.

2.1. Semiconductor lasers

The idea of a current-injection semiconductor laser goes back to a publi-

cation by John von Neumann in 195316. The first injection lasers were dem-

onstrated in 1962 simultaneously by different groups17-19 using GaAs p-n-

junctions as active regions. A further breakthrough in optoelectronics was

achieved when the heterostructure laser was developed20, 21. Heterostructure

lasers differ from the earlier heterojunction lasers by vertical light-wave

confinement between two cladding layers that have a refractive index dif-

ferent from the optical-confinement layer. Yet another step towards an im-

provement of device performance was the separation of charge carrier con-

finement from optical confinement. In such separate confinement het-

erostructure (SCH) lasers, carriers are confined within a few nanometer

thick low-band-gap semiconductor QW, placed in the optical confinement

layer. Quantum-dot lasers are SCH lasers in which the carriers are confined

in nanometer-sized, low-band-gap QDs.

2. Lasers and quantum dots

Most important for good laser characteristics are large optical gain and

low optical losses. The modal gain gmod of a laser is defined as the negative

light absorption coefficient of the resonator waveguide. In the lasing mode,

gmod is positive and must compensate optical losses. In current-injection la-

sers, the lasing threshold is reached if the current passes a certain threshold

value. At the lasing threshold, the modal gain gmod provided by the pumped

active region equals the optical losses. This is the case if

totimmod

ααα

=+=g (2.1)

The total losses

tot consist of the internal losses

i originating from scatter-

ing and free-carrier absorption in the waveguide, and the mirror losses

which depends on the mirror reflectivities R1 and R2, and the cavity length













RRL

(2.2)

For as-cleaved facets, R1 and R2 have the same value. However, R1 and R2

can independently be adjusted by facet coatings.

Edge emitters

Edge emitters or Fabry-Perot lasers are obtained if the light wave is

guided parallel to the substrate and two cleaved facets act as mirrors. A

schematic overview of an edge-emitting laser is given in Fig. 1. The vertical

optical confinement is provided by cladding layers that have different opti-

cal indices than the waveguide material. Lateral light-wave confinement is

Fig. 1: Schematic diagram of the cross-sectional view on the laser facet of a typi-

cal, fully processed edge-emitting laser diode. The different layers are: 1. Ti/Pt/Au

top contact. 2. SiNx insulating layer. 3. p++ GaAs contact layer. 4. p+ AlGaAs top

cladding. 5. Undoped optical-confinement GaAs layer with the active region in

the center. 6. n+ AlGaAs bottom cladding. 7. Substrate. 8. Ni/AuGe bottom con-

tact.

2.1. Semiconductor lasers

achieved, for example, by lithographic definition and etching of ridges into

the as-grown laser structure. The SiNx insulation layer as depicted in Fig. 1

restricts current injection to the ridge area.

VCSELs

In VCSELs, the light wave is guided perpendicularly to the substrate.

DBRs are used as mirrors. VCSELs enable easy coupling to optical fibers

due to their circular beam profile, leading to good mode matching with both

multimode and single-mode fibers. VCSELs are considered as the most im-

portant devices for optical interconnects, enabling ultra-parallel information

transmission in computer systems7, 22.

The working principle of VCSELs is analog to that of edge emitters.

However, the emission perpendicular to the growth plane and the ultra short

cavity length as compared to edge emitters require a completely different

layer design. As can easily be deduced from Eq. 1.2, short cavities require

ultra-high mirror reflectivities. In fact, mirror reflectivities of more than

99 % are required for VCSELs. Such reflectivities can only be provided by

DBRs. Fig. 2 shows a schematic diagram of a pillar VCSEL. The center of

the device is magnified and shows the active layer consisting of QD sheets.

Lateral gain guiding is achieved by a top oxide aperture. The oxide of the

aperture mainly acts as electrical insulator, allowing pumping of the active

medium only below the unoxidized center of the aperture. The diameter of

the aperture has a decisive impact on threshold current and lateral mode

profile. The active layers are centered in an antinode of the stationary wave

so that a maximum overlap of optical wave and active region is achieved.

Fig. 2: Layer and structure design of a GaAs-based full-oxide-DBR VCSEL with

intracavity p- (top) and n-contacts (bottom). The active region as shown schemati-

cally in the magnification can alternatively consist of a multi QD layer stack or

QWs. After Ref. 14.

2. Lasers and quantum dots

The n and p intracavity contact layers are placed in nodes of the stationary

light wave to minimize free-carrier absorption.

If all-semiconductor AlAs/GaAs DBRs are used, the number of periods

for each DBR must be in the order of 30. Using oxide-mirror technology, 5

to 7 Al(Ga)Ox/GaAs DBR pairs yield sufficient reflectivity, depending on

the mirror quality and also on the gain of the active medium. Oxide-DBR

VCSELs can thus be realized with significantly less epitaxial effort.

2.2. Quantum dots

The use of QDs as active material of semiconductor lasers can lead to

superior device characteristics as compared to standard quantum-well based

devices. The basic properties of QDs and their impact on laser characteris-

tics are briefly outlined in this section.

Nanometer-sized In(Ga)As QDs in a GaAs matrix represent localization

centers for both electrons and holes since the band gap of In(Ga)As is lower

than that of the surrounding GaAs matrix and InGaAs/GaAs interfaces are

type-I heterojunctions. Self-organized InGaAs/GaAs QDs are typically 2 to

6 nm high and 10-25 nm wide. Carriers trapped by QDs are confined in all

three spatial directions and can occupy only discrete energy states. Since

the lateral extension of self-organized QDs are comparable to the exciton

Bohr radius, the electronic levels of both electron and hole states are sub-

stantially determined by the size, shape and chemical composition of the

QDs23. Therefore, QDs have more in common with the electronic structure

of atoms than of solids, which have quasi-continuous energy levels.

The original prediction of temperature-independent threshold currents of

QD lasers was made for QDs with a single electron (hole) level and infinite

potential barriers. Since QDs have typically more than one electron (hole)

level, and since the potential barriers are not infinite in real systems, QDs

hold the predictions only if the spacing between energy levels is larger than

the thermal energy kBT. Otherwise, a significant fraction of carriers are

found in higher energy states, or are thermally excited to the GaAs matrix

where recombination does not take place at the target wavelength.

The electronic density of states in zero-dimensional objects like QDs is

δ-function-like (for T = 0):

()

∑

−= n

LLL

zyx

)(

δρ

(2.3)

However, whereas the density of states is extremely large for E = Ei, the

number of confined electronic states present in a QD sheet scales with the

2.3. Properties of quantum-dot lasers

QD area density. This value is decisive for the performance of QD lasers.

Both threshold current density and modal gain of QD laser diodes increase

with the number of QDs.

Due to the δ-function-like density of states, ultra-narrow cathodolumi-

nescence lines of 0.15 meV could be observed for single InAs/GaAs QDs at

low temperatures24; the value of 0.15 meV was limited by the spectral reso-

lution of the setup. A typical ensemble of self-organized QDs, however, ex-

hibits a finite size distribution, leading to an inhomogeneous broadening of

the transition line, the full width at half maximum (FWHM) of which typi-

cally ranges from 25 meV to 80 meV.

Electron and hole levels in QDs as well as transition probabilities be-

tween these levels can be calculated within an 8-band k·p framework25. A

decisive difference between QDs and atoms is the lower spatial symmetry

of QDs, which are not spherically symmetric. Moreover, the strain in self-

organized QDs induces a piezoelectric field, owing to the strain-related

relative displacements of the anion and cation sublattices. This further re-

duces the QD symmetry. None of the transition selection rules valid for at-

oms exists for QDs, so that transitions from any electron state to any hole

state are essentially possible25. The transition matrix elements are in good

approximation proportional to the spatial overlap integral of the respective

electron and hole wavefunctions.

Whereas the QD electron ground state is two-fold degenerate, the first

excited electron state can be occupied by four electrons. Therefore, the mo-

dal gain of a QD laser due to exciton recombination from the first excited

QD electron state is twice as large. If the ground-state gain of a QD laser

cannot overcome losses since, for example, the QD density is not sufficient

or a large fraction of the QDs is dislocated, excited-state lasing sets in.

2.3. Properties of quantum-dot lasers

The first QD-like laser was demonstrated in 1982 by applying a strong

magnetic field to a QW laser perpendicular to the substrate3. An increase of

the characteristic temperature T0 from 144 K to 313 K was measuredi. Since

then, carrier localization was achieved by the fabrication of freestanding

QDs via lithographical patterning of QWs. QDs have also been fabricated

by selective intermixing using ion implantation or laser annealing. Lateral

charge carrier confinement could also be achieved by strain gradients in

QWs and also via growth on pre-patterned substrates (cf. Ref. 2 for a re-

view, and references therein). First QD laser operation was achieved with

QDs realized by lithographic patterning of QWs26. However, lasers based

i For a definition of T0 cf. Eq. 6.3.

2. Lasers and quantum dots

on such lithographic QDs were still troubled by extremely high threshold

current densities (7.6 kA/cm2 at 77 K), most likely due to process-induced

damages that cause high losses. A new era started in 1994 with the demon-

stration of the first semiconductor laser based on self-organized QDs27.

These devices exhibited significantly lower threshold currents and large T0

values. No lithographic steps are needed for the fabrication of such QDs so

that the danger of process-induced damage drops out. Fig. 3 shows the de-

velopment of the lowest threshold current densities of different semicon-

ductor laser types with the years. The lasers based on self-organized QDs as

active region show the lowest threshold current densities. The reduction of

threshold current densities for each of the respective laser types is due to the

improvement of growth and process technology.

To date, mostly InP-based optoelectronic components are available in the

wavelength regime of 1.3-1.55 µm as required for datacom applications.

This lasing wavelength regime can likewise be reached with InGaAsN QWs

on GaAs substrates. Threshold current densities of such dilute-nitride lasers

are rather high, however. It has been shown that low-threshold lasing in the

1.3-1.55 µm lasing wavelength regime can be reached on GaAs using self-

organized QDs8, 44. The wavelength of InGaAs QWs on GaAs, for example,

is limited to about 1.2 µm due to the onset of misfit dislocations5, 6. Since

1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010

101

102

103

104

105

Room Temperature

GaAs p-n

DHS

Threshold current censity (A/cm

)

Year

Fig. 3: Lowest threshold current densities reported for double-heterostructure

(DHS) lasers, QW lasers, and QD lasers, respectively, versus publication year. Af-

ter Ref. 28.

Miller et al.29

Alferov et al.30 Dupuis et al.31 Ledentsov et al.32

Kirstaedter et al.27

Alferov et al.33

Hayashi et al.34 Tsang35

Tsang36

Ledentsov et al.37

Alferov et al.38, 39 Chand Liu et al.40

et al.41 Eliseev

et al.42

Grundmann (theory)43

2.3. Properties of quantum-dot lasers

the height of self-organized InGaAs QDs can be larger than the thickness of

InGaAs QWs, QDs can exhibit larger emission wavelengths. QD emission

wavelengths of more than 1.4 µm at room temperature have been re-

ported45. Lateral electronic coupling in dense QD agglomerates can increase

the emission wavelength to beyond 1.75 µm46.

The overgrowth of self-organized InAs QDs with a thin Ga-rich InGaAs

layer has been demonstrated as an effective approach in MBE to tune the

QD ground-state transition up to 1.3 µm47-49. The overgrowth of InAs QDs

with such QWs results in a larger amount of indium located in or around the

QDs and to a strain reduction within the overgrown QDs (cf. sect. 5.3.1).

As active medium of laser diodes, such QD structures have led to promising

characteristics of both edge emitters8, 50-52 and VCSELs53 at 1.3 µm. Alter-

natively to overgrowing QDs with Ga-rich InGaAs films, InAs QDs have

been inserted into such QWs to reach room-temperature luminescence and

lasing at 1.3 µm. This technique is referred to as the DWELL (Dots in a

WELL) concept40, 42, 48, 54-58. QWs of about 10 nm thickness and a typical in-

dium fraction of 15 % to 20 % are chosen as matrix for the QDs. Liu et al.40

have found that the DWELL concept leads to significantly higher QD den-

sities as compared to direct deposition on GaAs. The lowest transparency

current of 13 A/cm2 ever achieved for a semiconductor laser diode at room

temperature was obtained using a single-DWELL as active medium42, 58.

For 1.3 µm DWELL lasers, lasing activity at operation temperatures beyond

150°C has been be demonstrated59.

QD lasers emitting at shorter wavelengths are also in the focus of scien-

tific interest. MOCVD growth of InGaAs/GaAs QD lasers showing ground-

state lasing with close to 4 W output power at 1100 nm has been demon-

strated12. Such devices can be used as efficient pump sources for Thulium-

doped up-conversion fiber lasers emitting in the blue spectral range60, 61.

MBE growth of QD lasers at 980 nm62 (pump sources for erbium-doped fi-

ber amplifiers) and 940 nm63 (pump sources for ytterbium-doped yttrium

aluminium garnet solid-state lasers and neodymium/erbium-doped fiber la-

sers) have also been reported. Red-light-emitting InP/InGaP QD lasers have

been grown on GaAs substrates using MOCVD64, 65.

Higher differential gain of QD lasers, as foreseen already in 1986 by

Asada et al.4, implies ultra short gain recovery times66, potentially enabling

higher cut-off frequencies for directly modulated QD lasers and higher

modulation frequencies of QD semiconductor optical amplifiers67, 68. Lower

-factors of QD lasers as compared to QW lasers and lower carrier diffu-

sion in QD sheets lead to suppression of beam filamentation of narrow-

stripe lasers69. The enhanced beam quality enables to reduce coupling losses

to optical fibers.

2. Lasers and quantum dots

Whereas the original prediction for QD lasers was a temperature insensi-

tivity of the threshold current at any temperature, the thresholds of real QD

lasers are still temperature-dependent. This is mainly attributed to thermal

excitation of charge carriers from the electron and hole ground states to

higher energy states. Ideally, the energy spacing between the discrete QD

states is larger than kBT, preventing higher energy states to be significantly

populated. This would lead to a complete temperature insensitivity of the

threshold current. However, the energy spacing particularly between hole

states can be smaller than kBT. QD lasers emitting between 1.1 and 1.2 µm

still exhibit temperature sensitivities comparable to those of InP-based QW

lasers. This issue is addressed in more detail in sect. 6.1.1.4.

A drawback of QD lasers is the relatively low modal gain gmod per QD

sheet. Although QDs exhibit a material gain that can exceed the material

gain of corresponding QW devices by far70, the modal gain of a single QD

sheet is much lower due to the small optical confinement. gmod is related to

the material gain gmat of the light-emitting medium according to

gmod =

·gmat (2.4)

is the optical confinement factor, representing the overlap of the optical

wave with the active region. The optical confinement

is best expressed as

xy·

z where

z is the vertical confinement and

xy the area coverage

with active material. Whereas

xy = 1 for QWs, it has typical values of only

0.01-0.03 for QD sheets. For a typical QD laser,

is in the order of 10-4.

This is about ten times lower than a typical optical confinement in QW la-

sers with comparable waveguide design. To grow QD lasers with high mo-

dal gain, both large QD sheet densities and close multiple stacking of QD

layers with thin spacer layers are decisive.

2.4. MOCVD for quantum-dot lasers

MOCVD is the most important commercial large-scale fabrication tech-

nology for electronic and optoelectronic semiconductor devices in Europe.

It is used by leading optoelectronic-device manufacturers like Agilent Tech-

nologies, Bookham, Osram Optosemiconductors und JDS Uniphase. Using

MOCVD, very high deposition rates can be achieved, and since no UHV is

needed like in MBE where extensive baking-out of substrates previous to

growth is required, a larger throughput is yielded.

In spite of these advantages, MBE has long been favored among re-

searchers for the development of QD growth technologies as well as for the

investigation of QD growth physics. The lasers of which record threshold

current densities are shown in Fig. 3 are all grown using MBE. In MBE,

growth conditions can very accurately be defined, controlled, and moni-

2.5. Objectives and methods of this work

tored. The reactants can be supplied in elemental form by molecular beams

that are directed perpendicularly onto a heated substrate placed in a UHV

chamber. The growth process is controlled by substrate temperature and

molecular flows only. Reflection high-energy electron diffraction (RHEED)

is one of the most important in-situ characterization tools and allows to

monitor surface reconstructions, surface smoothness on an atomic scale,

surface diffusion lengths and deposition rates. Although QD self-

organization has been investigated as long using MOCVD, results on

MOCVD-grown QD lasers have hitherto been published only by few re-

search groups8. For a couple of reasons, MOCVD growth is more difficult

to control. For example, the reactants are supplied as organic or hydride

precursors that are thermally cracked by the heated substrate susceptor into

growth-reactive species, making the fraction of reactants in the gas phase a

function of temperature. Aside from the choice of precursor molecules, total

pressure and total gas flow are additional parameters that have to be con-

trolled in contrast to MBE. Since MOCVD is not a UHV technique, in-situ

monitoring like RHEED is not available. Only recently, reflectance and re-

flectance-anisotropy spectrometers were launched as in-situ monitoring

tools by Laytec GmbH, Berlin, Germany71. The promising potential of these

techniques for manufacturing QD-based optoelectronic devices is currently

being explored.

Another reason for the rare use of MOCVD for the growth of QD lasers

might be the observation that MOCVD-grown QDs exhibit an enhanced

tendency to form plastically relaxed defect clusters72, most likely due to the

complex nature of MOCVD precursor molecules, leading to an enhanced

mobility of surface adatoms. Parasitic non-radiative recombination occur-

ring in defective QDs drastically reduces optical gain in QD lasers.

Whereas enhanced adatom mobilities are beneficial in MOCVD for the fast

growth of thick layers and the realization of smooth interfaces, which is

particularly important for the realization of low-loss waveguides for edge

emitters and high-reflectivity DBRs for VCSELs, MOCVD of QD lasers is

still a challenge.

2.5. Objectives and methods of this work

The device characteristics of real QD lasers have not yet been close to

the ideal limits outlined by the early predictions. In addition, there have

been arrears in the development of MOCVD QD growth due to the larger

number of parameters to be controlled with respect to MBE and due to an

enhanced tendency of defect formation. This work contributes to advance

MOCVD of novel GaAs-based lasers based on self-organized QDs as active

medium.

2. Lasers and quantum dots

• A technique of high-temperature in-situ annealing was developed to flat-

ten rough growth fronts. This enables a significant reduction of the

spacer thickness between stacked QD layers without forfeit of the crys-

talline quality of the QDs (cf. sect. 5.1.1). The number of QD sheets in a

laser waveguide was limited for a given QD sheet density and a given

waveguide thickness by the minimum thickness of the spacing layers

between the QD sheets. Thinner GaAs layers on top of QDs typically

exhibit corrugated surfaces73, owing to the underlying 3D morphology

of the QDs, and to the low growth temperatures around 500°C at which

SK QDs are deposited. Such corrugated surfaces are inappropriate for

the deposition of subsequent QD sheets.

• Due to the enhanced tendency of defect formation in MOCVD, thresh-

old current densities of MOCVD QD lasers have long been larger than

those of MBE-grown lasers. In-situ annealing, initially developed for

surface flattening, is shown to improve the crystalline quality of QD

structures by reducing the density of non-radiative recombination cen-

ters in the matrix (cf. sect. 5.1.2). The enhancement of the crystal quality

leads furthermore to an increase of internal quantum efficiencies of QD

laser diodes (cf sect. 6.1.1).

• The origin of the finite temperature stability of QD lasers was investi-

gated (cf. sect. 6.1.1.4). A theoretical approach74 assigns the finite tem-

perature stability of threshold currents of state-of-the-art QD lasers to

thermal excitation of charge carriers to GaAs matrix states and to subse-

quent non-radiative recombination, owing to low matrix crystal quali-

ties. Radiative recombination efficiencies in the GaAs matrix could sig-

nificantly be improved in this work by in-situ annealing73. However, the

temperature behavior of the threshold currents could not be increased

this way. QD lasers were grown in this work with thin AlGaAs diffusion

barriers around the QD active region to prevent charge carriers from

thermal excitation to the matrix. The results obtained with such struc-

tures were compared to quantum-well lasers with and without carrier

diffusion barriers. The temperature stability of quantum-well lasers

could be improved. However, diffusion barriers have only a minor in-

fluence on the temperature stability of quantum-dot lasers. These find-

ings suggest that non-radiative recombination in the matrix is not the

primary reason for the finite temperature stability of QD lasers.

• Strongly toxic gaseous hydride precursors such as arsine or explosive

hydrides like silane widely used in commercial MOCVD constitute a

considerable threat to man and environment. The suitability of the alter-

native organic precursor tertiarybutylarsine (TBAs) for the growth of

QD laser structures was therefore explored (sects. 5.2 and 5.3.2). The

changeover to alternative-precursor MOCVD required the adjustment of

2.5. Objectives and methods of this work

MOCVD growth parameters according to the specific precursor proper-

ties (sects. 3.2.2 and 3.2.3). The demonstration of QD lasers grown us-

ing alternative-precursor MOCVD (sect. 6.1.2) is an important step to-

wards the establishment of a safe and environment-friendly growth

technology of QD-based optoelectronic devices.

• In contrast to MBE where 1.3 µm lasers have already been demon-

strated, lasing wavelengths of MOCVD-grown QD lasers have not

reached 1.3 µm. QDs emitting at this wavelength are rather large and

thus contain high strain energies. The danger of defect formation in such

strained layers is particularly high in MOCVD. A number of alternative

techniques to redshift the QD emission wavelength have been investi-

gated with regard to their suitability for MOCVD-grown long-

wavelength QD lasers (sect. 5.3). These techniques comprise the over-

growth of InGaAs QDs with Ga-rich InGaAs QWs (sects. 5.3.1 and

5.3.2). Strain redistribution and increase of the effective QD size lead to

a redshift of the QD emission wavelength. This technique is currently

used in MBE to grow 1.3 µm QD lasers. In addition, experiments to red-

shift the emission wavelength by nitrogen (sect. 5.3.3) and antimony in-

sertions (sect. 5.3.4) are carried out and discussed. Nitrogen and anti-

mony were supplied during deposition of InGaAs QDs and/or during the

subsequent growth interruption (GRI) that is typically performed after

deposition of the QD material in order to allow QD formation and QD

size evolution. Long-wavelength QD structures were investigated using

photoluminescence (PL) spectroscopy and transmission electron mi-

croscopy images (TEM). QD lasers with emission wavelengths beyond

1.2 µm are presented, based on InGaAs QDs overgrown with Ga-rich

InGaAs QWs (sect. 6.1.3).

• GaAs-based QD VCSELs at 1.3 µm are particularly attractive opto-

electronic devices for data- and telecommunication. An MOCVD

growth process for the fabrication of all-oxide DBR QD VCSELs was

therefore developed within this work (sect. 6.2). To achieve vertical

lasing, large modal gain of the active region and high-reflectivity DBR

mirrors are needed. The comparably low modal gain of QD sheets as

compared to QW sheets requires dense vertical stacking of QD layers

located in the antinodes of the optical wave in the VCSEL cavity. The

gallium content of Al-rich AlGaAs layers for the fabrication of

AlGaOx/GaAs DBRs has a strong influence on the oxidation dynamics.

The Ga fraction of such AlGaAs layers was calibrated, and parameters

for a reproducible growth of such AlGaAs layers were found

(sect.3.3.1). The design of thin Ga-rich AlGaAs buffer layers inserted

between GaAs and Al-rich AlGaAs layers is decisive for the oxide qual-

ity75, 76. The insertion of such buffer layers is related with a roughening

of the DBR hetero-interfaces if TBAs is used. The development of

2. Lasers and quantum dots

VCSEL growth and the development of the fabrication process took

place simultaneously. On the way to the demonstration of a QD VCSEL,

a test VCSEL based on a single InGaAs QW as active medium was

grown and processed (sect. 6.2.1). Valuable feedback to the ongoing de-

velopment of both growth and process technology for the QD VCSEL

were obtained from this experiment.

3. Metalorganic chemical vapor deposition

3.1. Principle

The growth experiments were initially performed on an Aixtron Aix200

machine, equipped with conventional hydride precursors. Within the course

of this work, a new Aix200/4 machine was put into operation. From then

on, growth was only carried out on this machine, using weakly toxic alter-

native precursors.

Both MOCVD machines have horizontal quartz-glass reactors with gra-

phite susceptors and gas-foil rotation of the substrate plate. The graphite

susceptors are heated using an radio-frequency (RF) heater (Aix200) and an

infrared heater (Aix200/4), respectively. The reactor of the Aix200 enables

the growth on 2” wafers. The Aix200/4 has a larger reactor and permits to

grow on single 4”, 3” and 2” wafers, or on three 2” wafers simultaneously if

a satellite-rotation susceptor is used. The structures fabricated within this

work were exclusively grown on single 2” wafer.

Fig. 4 shows a simplified schematic of the MOCVD setup, valid for both

machines. Highly purified carrier gases (H2 or N2) are conducted to the

bubblers containing the organic-compound precursors. The bubblers are

connected to the machine by 4-way valves. Using such valves, unused bub-

blers can be bypassed. The bubblers are kept in thermostats filled with a

mixture of water and ethylene glycol, and are stabilized at temperatures

ranging from –10°C to 20°C. The temperature stabilization guarantees con-

stant equilibrium vapor pressures of the precursors inside the bubblers. In

addition, the total bubbler pressures (metalorganic + carrier gas) are ad-

justed by pressure controllers at predefined values larger than the precursor

vapor pressure, typically between 200 and 1800 mbar. The carrier gas is

conducted through a dip tube to the bottom of the bubbler and ascends

through the precursor liquid or granulate. While the carrier gas is conducted

through the bubbler, it is saturated with precursor molecules.

The 5/2-way “vent/run” valves conduct the mixture of precursor and car-

rier gas alternatively to the reactor or to the vent line that bypasses the reac-

tor. This way, stabilized and thus well-defined precursor flows are perma-

nently available, enabling short and reproducible growth sequences. The

group-V and group-III precursors are conducted separately to the reactor in

order to avoid precursor prereactions. In the Aix200, dopants are preferably

conducted to the reactor through the group-V line. The Aix200/4 has a

separate dopant line that finally meets the reactor few centimeters in front

of the reactor gas inlet.

3. Metalorganic chemical vapor deposition

The Aix200 was operated at 20 mbar with a total flow (carrier gas + me-

talorganic precursors + hydrides) of 5.56 standard liters per minuteii (slm).

These parameters have been found to yield optimum lateral layer-thickness

homogeneity. Moreover, the flow velocity achieved with these parameters

allows ultra short reproducible growth sequences with valve opening times

of 0.1 s. The situation is different for the larger Aix200/4 machine. The re-

actor of the Aix200/4 is twice as wide. To obtain sufficient lateral homoge-

neity for the growth of VCSEL structures, a total gas flow of up to 15.0 slm

has to be adjusted. The large gas volumes in reactor and supply pipes re-

quire a larger total pressure in order not to overstrain the process pump. A

total pressure of 100 mbar is hence used with this machine. The dependence

of lateral layer thickness homogeneity on parameters like total pressure, to-

tal gas flow as well as the group-V to group-III inlet flow ratio is discussed

in detail in section 3.3.1.2.

3.2. Conventional and alternative precursors

In conventional MOCVD of III-V compound semiconductors, only the

group-III precursors are supplied in the form of liquid or granular organic

ii Standard conditions are defined as T = 0°C, p = 1.013 bar.

TBAs TESb DMHy

TMGa TMIn TMAl

process pump

reactor

mass flow controller

5/2 way valve

4 way valve

Fig. 4: Simplified schematic of the MOCVD setup of both the Aix200 and

Aix200/4 machine. The two kits basically differ by the size of the quartz-glass re-

actor. The Aix200 disposes of a hydride lines for arsine instead of the TBAs-

bubbler line. Dopant lines and pressure controllers are not shown.

3.2. Conventional and alternative precursors

compound precursors. The group-V elements are usually supplied as gase-

ous hydrides. The danger connected with the toxicity and the high pressure

of hydrides like arsine and phosphine can be overcome if organic group-V

compounds such as TBAs and tertiarybutylphosphine are used. The major

benefit of alternative precursors for the epitaxy of QD-based optoelectronic

devices is their large decomposition efficiency77 at low growth temperatures

around 500°C required for the formation of SK QDs2. On the other hand,

the different chemical properties of alternative group-V precursors lead to

different layer properties and require other growth parameters, as discussed

in the following sections.

3.2.1 Conventional-precursor MOCVD

For conventional-precursor MOCVD on the Aix200 machine, arsine

(AsH3) was used as arsenic precursor. Since arsine hardly decomposes at

typical QD growth temperatures of 480-520°C, large V/III ratios must be

used. Trimethylgallium (TMGa, (CH3)3Ga), triethylgallium (TEGa,

(C2H5)3Ga), trimethylaluminium (TMAl, (CH3)3Al) and trimethylindium

(TMIn, (CH3)3In) were used as group-III precursors.

Whereas TMGa was used for thick GaAs and AlGaAs layers and also for

InGaAs QD sheets and QWs, TEGa was used for the AlGaAs layers of

VCSEL structures for the fabrication of oxide DBRs. Reproducible ultra

low Ga-fractions of 1 to 4 % are needed in such AlGaAs layers to precisely

achieve target oxidation rates (cf. sect. 3.3.1.3). This requires extremely low

Ga molar flows between 4×10-7 and 1.6×10-6 mol/min. Such low TMGa

flows could be achieved with a double dilution configuration. However, no

such configuration was available for Ga on the Aix200 machine. The prob-

lem was solved using TEGa instead of TMGa. TEGa has a much lower va-

por pressure than TMGa (pTMGa(20°C) = 243 mbar, pTEGa(20°C) = 6.7 mbar)

so that larger source flows can be used for the same molar Ga flow, and lar-

ger source flows can be controlled more precisely.

For the deposition of ternary InGaAs QDs around 500°C, TMGa was

used. It has been reported that the use of TMGa at such growth tempera-

tures leads to strong carbon incorporation and thus to significant intrinsic p-

doping78 which would strongly decrease the radiative efficiency of InGaAs

QD layers. Intrinsic p-doping is reported not to occur if TEGa is used in-

stead78. Therefore, TEGa is sometimes preferred for MOCVD growth at

low deposition temperatures, for example for high-quality InGaAsN QWs79.

A comparative study of InGaAs QDs grown with TMGa and TEGa was un-

dertaken within this work and has shown that the radiative efficiency of the

layers grown with TMGa is not inferior to that of TEGa-grown structures.

3. Metalorganic chemical vapor deposition

A mixture of SiH4 (2 %) and H2 (98 %) was used as n-dopant, conducted

through a double dilution configuration. Dimethylzinc (DMZn, (CH3)2Zn)

was used as p-dopant for the top AlGaAs cladding layers of pin-diode struc-

tures for edge-emitting lasers. Zn shows strong diffusion in GaAs78, 80, how-

ever, so that for the growth of VCSELs with GaAs intracavity contacts,

carbon tetrabromide (CBr4) was used. With this precursor, p-doping levels

ranging from 1016 to 1019 cm-3 can be achieved81, 82.

3.2.2 Alternative-precursor MOCVD

For alternative-precursor MOCVD on the Aix200/4 machine, only low-

pressure organic-compound precursors were used. Tertiarybutylarsine

(TBAs, (C4H9)AsH2) was used as substitute for AsH3. The most important

properties of TBAs as well as the readjustment of growth parameters con-

nected with the use of TBAs are described in more detail in the next sec-

tion. For the growth of InGaAsN QDs, unsymmetrical dimethylhydrazine77

(UDMHy or DMHy, (H2N-N(CH3)2) was used as nitrogen precursor.

DMHy is completely decomposed at 420°C (Ref. 78, p. 258). The decom-

position of the standard N precursor ammonia (NH3), commonly used for

the growth of GaN, is only 15 % at 950°C (Ref. 78) and is negligible

around 500°C. It has long been difficult to manufacture high-purity UD-

MHy due to the large reactivity of N, leading to bonds with undesired

chemical elements. Today, however, the production of UDMHy with high

purity is possible77. UDMHy is likewise usable for the growth of GaN83.

For antimony insertions, triethylantimony (TESb) was used.

The same group-III precursors were used on the Aix200/4 as for the

Aix200. Instead of using TEGa for the Al-rich AlGaAs layers of VCSEL-

DBR structures, however, TMGa was used. A double dilution line for

TMGa was available on the new machine, allowing highly reproducible ul-

tra-low Ga molar flows.

N-doping of AlxGa1-xAs is a general problem since donors like Si or Te

form DX centers84 in AlGaAs. Depending on the Al fraction x, the donor

activation energy is up to 95 meV for Te85 and up to 270 meV for Si86. Fig.

5 shows the electron concentration in AlxGa1-xAs:Te and AlxGa1-xAs:Si lay-

ers as a function of Al fraction x. The data were taken from two different

publications. As one can see for x = 0 (pure GaAs) where the donor activa-

tion is close to 100 % for both Si and Te, the doping level of the

AlxGa1-xAs:Te layer is twice as high as the Si doping level of the

AlxGa1-xAs:Si sample. Between x = 0.6 and x = 0.8, however, the fraction

of ionized Te donors is more than one order of magnitude larger than the

fraction of ionized Si donors. At x = 0.6, the room-temperature activation of

Si donors is roughly 10-3, i.e. a doping level of 1020 cm-3 would be required

3.2. Conventional and alternative precursors

to achieve an electron concentration of 1017 cm-3. Such high doping levels

can lead to low crystal qualities and to rough surfaces. Tellurium donors

build shallower DX centers in AlGaAs so that their activation is higher and

target electron concentrations can be achieved with lower doping levels.

Tellurium was therefore preferred to silicon. Diethyltelluride (DETe) was

used as tellurium precursor.

Tellurium is reported to be a critical n-dopant due to memory effects:

Tellurium, adsorbed by the reactor walls during intentional Te doping, can

desorb during the growth of subsequent, nominally undoped layers and lead

to unintentional Te doping. Memory effects are, however, not necessarily

observed87. It was found in this work that memory effects are only signifi-

cant if, for example, GaAs:Te layers with large electron concentrations of

more than 1019 cm-3 are grown. For Te doping levels leading to electron

concentrations in GaAs of 1018 cm-3 and less, memory effects were not

measured. In addition, it could be shown that memory effects after the

growth of highly doped layers can be suppressed if GRIs during 5-15 min at

650-700°C are introduced immediately after deposition of the Te-doped

layers.

Silicon has no memory effects and should actually be preferred to Te as

n-dopant for GaAs. Within the doping experiments carried out in this work

using alternative-precursor MOCVD, GaAs could successfully be doped

with silicon using the alternative silicon precursor ditertiarybutylsilane88.

0.00.20.40.60.81.0

1E15

1E16

1E17

1E18

1E19

AlxGa1-xAs:Te

AlxGa1-xAs:Si

Electron concentration (cm

-3

)

Aluminium fraction x

Fig. 5: Electron concentration of AlxGa1-xAs:Te and AlxGa1-xAs:Si as functions of

the aluminium fraction x, determined by Hall measurements at room temperature.

The data for AlxGa1-xAs:Te were taken from Ref. 87, the data for AlxGa1-xAs:Si

are from Ref. 86.

3. Metalorganic chemical vapor deposition

Since the Aix200/4 has only one n-dopant line, however, Te was used as n-

dopant for both AlGaAs and GaAs.

3.2.3 Tertiarybutylarsine

The large hazardous potential of the highly toxic arsine was strongly re-

duced on the Aix200/4 machine by using the much less toxic organic-

compound precursor tertiarybutylarsine (TBAs)89. The toxicity of arsenic -

containing molecules is essentially due to the number of As-H functions.

By replacement of AsH3 by TBAs, the number of As-H functions per mole-

cule is reduced from 3 to 2 (cf. Fig. 6), so that the toxicity of TBAs is not

much lower than that of arsine. A significant reduction of the toxicity is

only observed for arsine-trialkyle compounds like trimethylarsine (TMAs)

or triethylarsine (TEAs)89. These molecules do not contain any As-H bonds

and would be the optimum solution with regard to toxicity. However, the

use of TEAs or TMAs leads to a high carbon incorporation even in GaAs78,

90-92, caused by unsaturated methyl radicals of the group-III sources so that

these precursors are applicable only as acceptor dopant sources93.

The decisive advantage of TBAs with regard to safety is its low vapor

pressure of 148 mbar at 20°C. The low vapor pressure drastically reduces

the risk of spreading in the case of an accidental leak in the MOCVD setup.

Since vapor pressure likewise adds to the dangerousness of a precursor,

Fig. 6: Schematic diagram of the principal cracking mechanisms of arsine and ter-

tiarybutylarsine (TBAs) during pyrolysis in MOCVD. Homolytic fission and β-

elimination are competing decomposition mechanisms of TBAs. Homolytic fis-

sion produces a reactive AsH2 radical. β-elimination generates an arsine molecule

and an inert isobutene molecule.

3.2. Conventional and alternative precursors

Stolz and Whitaker defined a safety figure of merit which is given by the

LC50

iii value divided by the vapor pressure of the chemical at 20°C (Ref.

89). On the scale of this figure of merit, TBAs is by two to three orders of

magnitude less dangerous than AsH3.

The decomposition of TBAs molecules takes place at high temperatures

via two main cracking mechanisms, as shown in the schematic diagram of

Fig. 6. In this diagram, the cracking of TBAs is compared to the decompo-

sition of arsine. The homolytic fission of TBAs directly leads to a reactive

AsH2 radical. An isobutene molecule (C4H8) and an arsenic-trihydride

molecule are obtained after β-elimination. The decomposition of arsine

simply occurs by cleavage of one of the hydrogen atoms. Since in the case

of β-elimination, a conventional arsine molecule is obtained and since iso-

butene is a rather inert molecule which is unlikely to participate in later sur-

face processes, the differences between conventional- and alternative-

precursor MOCVD is mainly attributed to the homolytic decomposition

process. Using quadrupole mass spectroscopy, the β-elimination efficiency

of TBAs was estimated to be 55 %. However, Raman measurements of the

decomposition products have identified the homolytic fission to be the main

decomposition mechanism90.

Growth studies of arsenide bulk layers using TBAs showed that crystal

qualities equivalent to those of AsH3-grown layers is obtained at much

lower V/III ratios, owing to the high cracking efficiency of TBAs at low

temperatures77, 94, 95. Fig. 7 shows the decomposition efficiencies of arsine

and TBAs as a function of reactor temperature. Above 450°C, the decom-

position efficiency of TBAs is close to 100 %, whereas the decomposition

efficiency of arsine is close to 100 % only beyond 650°C. Particularly at

QD deposition temperatures around 500°C, the arsine decomposition effi-

ciency is as low as 10 %.

In contrast to bulk-layer epitaxy, QDs grown in the SK mode are formed

by lateral material transport from a strained 2D wetting layer (WL) to co-

herent 3D islands. Sufficient adatom mobility is hence essential for material

migration. Using TBAs, a multitude of bulky organic molecules are ad-

sorbed at the crystal surface during the growth process90 which could a pri-

ori hinder the QD formation or affect the crystal quality by carbon incorpo-

ration. It will be shown later, however, that QDs and QD lasers of high

quality can be grown using TBAs.

iii LC50 is the concentration of a chemical that is lethal for 50% of a representative

animal population after an exposure of typically 4 hours.

3. Metalorganic chemical vapor deposition

3.2.3.1 Surface stabilization

TBAs ranks among the most expensive consumables in MOCVD

growth. First experiments with the new Aix200/4 machine therefore aimed

at the minimization of TBAs consumption wherever this is possible, i.e.

during heat-up at the beginning of an epitaxy run, cool-down at the end of a

run, and during GRIs when the surface is stabilized against As desorption.

In order not to work in the As desorption regime, experiments were con-

ducted with sample pieces that were heated up to 700°C and stabilized un-

der different TBAs flows. The epilayers of these samples are simply

300 nm thick homoepitaxial buffer layers, deposited on semi-insulating (SI)

GaAs(001). The heatup procedure is typical at the beginning of every

growth run and aims at the removal of the oxide layer and other contami-

nants from the substrate surface. The substrate pieces were immediately

cooled down after the bake-out and measured ex-situ by atomic force mi-

croscopy (AFM). The AFM images are shown in Fig. 8. The surface stabi-

lized with the largest TBAs partial pressure of pTBAs = 2×10-2 mbar (A)

shows flat monolayer (ML) terraces. The sample stabilized with a lower

TBAs partial pressure of pTBAs = 2×10-3 mbar (B) shows ML terraces with

small ML-deep holes, presumably owing to congruent desorption of GaAs

from these sites. The surface without As stabilization (C) also shows ter-

races. Here, the terrace planes show small 3D objects, most likely gallium

droplets.

For heatup, GRIs and cooldown of further growth experiments, the

TBAs partial pressure of Fig. 8 (A) was chosen. This TBAs partial pressure

Fig. 7: Comparison of thermal decomposition of TBAs and arsine in an atmos-

pheric pressure reactor. After Stringfellow et al.96.

3.2. Conventional and alternative precursors

is small as compared to the TBAs partial pressure (pTBAs = 0.1-0.3 mbar)

typically used for the growth of high-quality epilayers.

3.2.3.2 V/III ratios for GaAs and AlGaAs growth

If TBAs is used instead of AsH3, the V/III ratio during growth of GaAs

can be much lower to achieve comparable layer qualities77, 95. This is due to

the more efficient pyrolysis of TBAs as compared to AsH3 (cf. Fig. 7). For

GaAs, a V/III ratio of 15 was chosen for the growth of PL test structures.

For the growth of devices, it was raised to 20, however. GaAs layers grown

with V/III=15 are n-type with an electron concentration of only 1014 to 1015

cm-3.

Whereas it is possible to grow

quasi-SI GaAs with a small V/III

ratio, it is shown in sect. 3.3.1.3

that Al-containing layers are

highly p-doped due to incorpora-

tion of carbon from the precursor

organyls. Fig. 9 shows the depend-

ence of the hole concentration of

1 µm thick Al0.4Ga0.6As layers as a

function of V/III ratio. It is in-

versely proportional to the V/III

ratio77. Even for a V/III ratio of 45,

the hole concentration is as high as

2 × 1017 cm-3. Since carbon atoms

in AlGaAs can be passivated by

hydrogen atoms97, 98, likely stem-

Fig. 8: AFM images (3 × 3 µm2) of GaAs substrate pieces heated up to 700°C and

cooled down under different TBAs partial pressures (A: 2 × 10-2 mbar, B:

2 × 10-3 mbar, C: no stabilization).

10 20 30 40

0.0

0.5

1.0

1.5

2.0

Al0.4Ga0.6As

Hole conc. (10

-3

)

TBAs/III ratio

Fig. 9: Hole concentration of 1 µm thick,

nominally undoped Al0.4Ga0.6As layers

as a function of V/III ratio. The hole

concentration was measured by the van-

der-Pauw method.

3. Metalorganic chemical vapor deposition

ming from the carrier gas, the C doping level may be even higher than the

hole concentration.

The decrease of carbon incorporation is particularly important for the

growth of n-doped AlGaAs layers. Here, the unintentional carbon p-doping

must in any case be overcompensated by n-doping. It was found out in this

work that values of n = 5 × 1017 cm-3 needed for bottom AlGaAs cladding

layers of edge-emitting lasers can only be achieved with TBAs/III ratios of

at least 60.

3.3. Modular optimization of laser structures

Modern semiconductor lasers such as SCH lasers with multi-QD/QW-

sheet active regions have rather complex layer structures, as depicted sche-

matically in Fig. 1 and Fig. 2 (sect. 2.1). To work on the performance of

such lasers, it is useful to subdivide the laser structure into different mod-

ules and to separately optimize these modules using appropriate test struc-

tures. The active region, the optical confinement layer, the cladding layers

(edge emitters), and DBRs (VCSELs) are the most important modules.

Important for all modules is the lateral homogeneity of the layer thick-

nesses. Devices processed from different regions of a wafer can only have

the same characteristics if their epilayers have the same thicknesses. Such

homogeneity is particularly important for the growth of VCSELs where a

detuning of the cavity resonance wavelength from the gain maximum of the

active region by more than 1 % can make a VCSEL structure unusable.

Calibration of growth rates and the adjustment of layer thickness homoge-

neity are described in the following section.

The QD active region is the most delicate part of a QD laser: the emis-

sion wavelength must be calibrated, the crystalline quality has to be opti-

mized to obtain maximum radiative efficiency, and growth surface corruga-

tions related with the presence of three-dimensional QDs must be controlled

in order to achieve close vertical QD-layer stacking and also low-loss opti-

cal waveguides. A suitable test structure for the optimization of QDs using

PL spectroscopy is described in sect. 3.3.2. For the optimization of QD lay-

ers, good layer thickness homogeneity is prerequisite. Otherwise, QDs from

different parts of the wafer differ in emission wavelength, area density and

crystalline quality.

The doping profile within the cladding layers, buffer and contact layers

need to be optimized with regard to minimum optical waveguide loss and

minimum series resistance. High doping levels mean low series resistance

but give rise to large optical losses via free-carrier absorption.

3.3. Modular optimization of laser structures

3.3.1 Waveguides and distributed Bragg reflectors

3.3.1.1 Calibration of growth rate

The pyrolysis of MOCVD precursors is a function of temperature. Pre-

cursors must be stable at room temperature, whereas at growth temperature,

the decomposition should be as efficient as possible. Since the decomposi-

tion efficiency varies within a precursor-specific temperature range, growth

rates generally depend on temperature. Such dependences are relevant for

the growth of lasers since the reactor temperature is varied during the

growth run. The choice of growth temperature for a particular layer depends

on its vertical position in the laser structure: Layers below QDs can actually

be grown at any appropriate temperature. Growth temperatures for layers

above the QDs are limited, however, since intermixing of In from the QDs

with Ga of the surrounding matrix sets in above 600°C and alters the struc-

tural and electronic properties of the QDs99-101. This leads essentially lead-

ing to a blueshift of the QD ground-state emission wavelength100.

Underneath the QD layers, all GaAs and Al(Ga)As layers can be depos-

ited at optimum growth temperatures. These are 600-650°C for GaAs and

700-750°C for Al(Ga)As layers for the growth with AsH3. If the deposition

temperature for Al(Ga)As layers is too low, carbon and oxygen incorpora-

tion increase due to a decrease of the effective V/III ratio102, related to the

strong temperature dependence of the arsine pyrolysis78. Large impurity

concentrations can effect poor crystal quality. The efficient low-temperature

cracking of TBAs enables to grow both GaAs and AlGaAs layers at 625°C

with reasonable quality103. At this temperature, a good compromise is found

between layer quality and TBAs consumption.

Mandatory for the growth rate calibration is a sufficient reproducibility

of the layer thicknesses. The run-to-run reproducibility of both MOCVD

machines used is better than 1 %. However, growth rate jumps of up to 4 %

are observed after regeneration of liner tube, graphite susceptor and sub-

strate plate. The usefulness of a VCSEL structure of which the layer thick-

nesses deviate from the nominal values by more than 1 % is highly endan-

gered. For a coarse determination of growth rates and their temperature de-

pendence, AlGaAs/GaAs heterostructures were grown in a single sample at

different temperatures. A coarse calibration of the thicknesses of GaAs and

AlGaAs layers can be performed using cross-sectional AFM, as described

in the following subsection. The fine-tuning of the calibration rate is done

with optical techniques as described subsequently.

3. Metalorganic chemical vapor deposition

Cross-section Atomic Force Microscopy

Cross-section AFM is very useful to determine growth rates for layers of

which the thickness is not so critical for the device characteristics of the la-

ser structure. This applies, for example, to cladding layers and the optical

confinement layer of edge emitters or to the thin AlGaAs buffer layers that

are inserted between the GaAs and AlGaAs λ/4 layers of oxide DBRs in

VCSEL structures, or to the oxide-aperture layers of VCSELs. If the nomi-

nal thicknesses of oxide-aperture layers are not perfectly met, the thickness

of undoped GaAs layers during cavity-thickness calibration can be varied.

Cross-sectional Atomic-Force Microscopy allows to get an overview of

the temperature dependence of GaAs and Al(Ga)As growth rates for a large

number of different temperatures, growth rates and V/III ratios from one

sample. To this end, arbitrary sequences of alternating GaAs and Al(Ga)As

layers are deposited under different growth conditions. Due to the spatial

protrusion of Al-containing layers, owing to the oxidation of Al and swell-

ing of the AlOx in ambient air, the Al(Ga)As/GaAs interfaces can be identi-

fied using cross-sectional AFM on cleaved surfaces. Depending on the Al

fraction, the thickness of the natural oxide ranges from few MLs to tenths

of nanometers. This permits not only to determine the growth rates of the

deposited layers but also their Al fractions104, 105. For the determination of

Al fractions, X-ray diffraction spectra of thick AlGaAs reference samples

were evaluated in this work. Using this method, Al fractions with errors far

below 1 % can be measured106. GaAs is also subject to oxidation in ambient

air. The protrusion due to oxidation is negligible, however. Fig. 10 shows

the temperature dependence of GaAs and Al0.65Ga0.35As growth rates,

evaluated from cross-sectional AFM measurements.

The accuracy of the AFM method is estimated to be within an error of 5-

10 % since the boundaries between different layers are frayed in AFM im-

ages. In addition, the geometry of the AFM tip might also falsify the results.

Reflection spectra of AlGaAs-GaAs DBRs

For the growth of Al(Ga)As/GaAs structures for the wet-thermal fabrica-

tion of oxide DBRs for VCSELs, the evaluation of optical reflection spectra

allows the calibration of layer thicknesses with a precision of better than

0.2 %. Optimum layer thicknesses are important to have the stop band of

the oxide DBRs at the gain maximum. For this purpose, the reflection spec-

trum of such Al(Ga)As/GaAs structures is recorded using a white light

source. A transfer matrix method107 is used to simulate the GaAs and

Al(Ga)As layer thicknesses, respectively. The simulations are subsequently

fitted to the data.

3.3. Modular optimization of laser structures

As mentioned earlier, the optical thickness of the cavity is the most deli-

cate issue of a VCSEL structure. For the calibration of the cavity thickness,

test structures are grown containing two bottom-DBR pairs, the entire cav-

ity with QDs, contact layers and apertures, and two top-DBR pairs. Oxida-

tion of the DBR layers is not required for the calibration of the cavity thick-

ness. Cavity thickness calibrations are performed previous to each VCSEL

growth run. The optical method is also used as quality test of as-grown

VCSEL structures. Fig. 11 shows a measured and simulated reflection spec-

trum of a complete VCSEL structure.

3.3.1.2 Lateral homogeneity of layer thicknesses

The lateral homogeneity of layer-thicknesses of samples grown with the

Aix200 machine, achieved with the standard growth parameters used from

the very beginning of the experiments, has turned out to be sufficient for the

growth of VCSEL structures. The standard parameters are a total pressure

of 20 mbar and a total gas flow of 5.56 l/min, equally split over the group-

III and group-V gas inlet. The growth experiments on the Aix200 were ini-

Fig. 10: GaAs and AlGaAs growth rates as determined from cross-sectional AFM

images, plotted as a function of growth temperature. The sample consisted of al-

ternating AlxGa1-xAs/GaAs layers grown at different temperatures, using TMGa,

TMAl and AsH3 as precursors. The TMGa and TMAl flows were 7.4 and

5.8 µmol/min, respectively; the V/III ratio was 180 for all layers. Error bars origi-

nate from statistical treatment of 10 thickness measurements of each layer; dashed

lines are guides to the eye. An Al fraction of x = 65.1 % was determined from an

x-ray diffraction spectra of a 1 µm thick AlGaAs epilayer grown at 720°C with

the given precursor flows. The Al fraction may slightly change with decreasing

growth temperature.

3. Metalorganic chemical vapor deposition

tially performed using an infrared lamp heater. In the course of the experi-

mental work, the heater was replaced by an RF heater. This improved the

layer-thickness homogeneity further, most likely due to an enhanced homo-

geneity of the lateral temperature profile.

Since the alternative-precursor-MOCVD growth runs described in this

work were the first experiment to be carried out on the new Aix200/4 ma-

chine, parameters such as total pressure and total gas flow were to be ad-

justed first in order to get sufficient layer thickness homogeneity. The main

difference of the Aix200/4 with regard to the Aix200 is the width of the re-

actor. Whereas the 4” reactor of the Aix200/4 is not significantly longer

than the 2” reactor of the Aix200, it is twice as wide. This means that the

precursors must homogeneously spread to over twice the width on their way

from the gas-mixing head to the susceptor. To adjust the layer thickness

homogeneity, total pressure, total flow as well as the group-III-line / group-

V-line inlet flow ratio (FIII/FV) were varied. A schematic diagram of the po-

sition of group-III and group-V inlet in the mixing head is given in Fig. 12.

For the determination of layer thickness homogeneities, similar struc-

tures as for the optical measurement of the growth rates (cf. 3.3.1.1) were

used. To simplify matters, however, superlattices consisting of only binary

AlAs and GaAs layers were grown. Since the area of the circular light spot

used for the reflection measurements was less than 1 mm2, layer thicknesses

could be resolved spatially as a function of lateral position on the wafer.

iv By courtesy of Friedhelm Hopfer.

800 1000 1200 1400 1600 1800

0.0

0.2

0.4

0.6

0.8

1.0 experiment

simulation

Reflectivity

Wavelength (nm)

Fig. 11: Reflection spectrum of an unoxidized, as-grown VCSEL structure. The

DBRs consist of six periods of GaAs / Al0.98Ga0.02Asiv.

3.3. Modular optimization of laser structures

Layer thicknesses of the AlAs layers as a function of distance from the wa-

fer center are shown in Fig. 13 for different total pressures, total gas flows

and FIII/FV ratios. The values for the GaAs layers were omitted as they be-

have analogous to those of AlAs. The AlAs thicknesses of the samples were

normalized to their respective average value. A 1 %-error window in which

the thickness values should be contained is also depicted.

The filled circles and triangles in Fig. 13 show the normalized thick-

nesses of the AlAs layers grown at total pressures of 100 mbar and 50 mbar,

respectively, with a total gas flow of 10 slm, equally split over the group-III

and group-V inlet. It is obvious that the homogeneity of the sample grown

at 50 mbar is worst. This is contradictory to the general belief that the

growth rate homogeneity can be improved by lowering the total pressure.

For all other growth runs, the total pressure was set to 100 mbar, which is

the total pressure recommended by the manufacturer Aixtron. Raising the

total pressure to larger values like 200 mbar should in principle increase the

homogeneity further. However, the gas velocity in the system is then half if

all other parameters are kept constant. This is disadvantageous for the im-

plementation of short growth sequences. Low gas velocities also prevent

from quick parameter ramping during the growth of one layer.

For the sample represented by filled squares in Fig. 13, an asymmetric

gas inlet was used, keeping a total flow rate of 10 slm. FIII was reduced to

3 slm, FV was increased to 7 slm. The asymmetric gas inlet slightly im-

proves the homogeneity. To maintain a reasonable functioning of the ma-

chine, however, FIII should not be reduced further.

The best result was achieved with FIII = 3 slm and FV = 12 slm. As

shown by the empty squares in Fig. 13, the homogeneity is better than 1 %.

These parameters were used for all following structures. It must be noted

that the growth rate for FIII = 3 slm and FV = 5 slm is about 30 % lower than

for FIII = FV = 5 slm. Using the 3 × 2” satellite configuration, special liner

FIII

= group III precursors + H

= group V precursors + H

gas

inlet

Exhaust

Fig. 12: Strongly simplified gas inlet schematic of the Aix200/4 reactor with sus-

ceptor and substrate. Group-V precursor gases are conducted separately from

group-III precursors. In order to avoid prereactions, they are not mixed before

they reach the entrance of the reactor. Dopants are added to the group-III line.

3. Metalorganic chemical vapor deposition

tubes with additional gas deflection plates close to the gas inlet can be used

to achieve good lateral thickness homogeneities108. No experiments with the

satellite rotation system were carried out in this work.

3.3.1.3 Calibration of oxidation rate for Al(Ga)Ox/GaAs DBRs

In order to fabricate Al(Ga)Ox/GaAs DBRs for VCSELs, the

Al1-xGaxAs/GaAs superlattices above and below the microcavity undergo a

wet-thermal oxidation process. Fig. 14 shows the oxidation rate at 420°C of

a 100 nm thick Al1-xGaxAs layer, cladded by 100 nm thick GaAs layers, as a

function of the AlAs mole fraction. The oxidation rates of binary AlAs and

Al0.96Ga0.4As differ by one order of magnitude, so the oxidation rates in the

Al1-xGaxAs layers critically depend on the Ga content x. The effect is used

to separately adjust the oxidation speeds for VCSEL-DBR mesas of differ-

ent diameters. This is particularly important for the one-stage oxidation

process for the fabrication of QD VCSELs as described in sect. 6.2.2. Here,

the oxidation of small top DBRs must be completed within the same time as

0 5 10 15 20 25

0.96

0.98

1.00

1.02

Norm. layer thickness

Radial distance (mm)

Fig. 13: Thickness of the AlAs layers of AlAs/GaAs superlattices, normalized to

1.0 for a radial distance of 0.5 inch = 12.7 mm. For the sake of clearness, the val-

ues for GaAs were omitted. The behavior of GaAs layer thicknesses is analogous

to those of AlAs. The AlAs layers were grown with a V/III of 30 at a growth rate

of 2 µm/h.

() ptot = 100 mbar, FIII = FV = 5.0 slm (cf. Fig. 12).

(▲) ptot = 50 mbar, FIII = FV = 5.0 slm.

( ■ ) ptot = 100 mbar, FIII = 3.0 slm, FV = 7.0 slm.

( □ ) ptot = 100 mbar, FIII = 3.0 slm, FV = 12.0 slm.

3.3. Modular optimization of laser structures

the oxidation of the large bottom DBRs. Moreover, the quality of fast-

oxidizing binary AlAs layers is very poor: Mesas containing thick binary

AlAs layers delaminate along the oxide/semiconductor interfaces109. The

admixture of few percents of Ga considerably enhances the mechanical sta-

bility110. This improvement is attributed to the removal of volatile oxidation

products such as As and As2O3 from the oxidizing layers111. The removal of

such products is not efficient enough if oxidation rates are as high as for bi-

nary AlAs.

Once a parameter window for the oxidation process is determined using

test structures with different values of x, the Ga content must be kept as

precisely as possible for all subsequent growth runs. This is a challenge to

the epitaxy since Ga contents in this range are difficult to control. The prob-

lem was solved using TEGa as gallium precursor for growth experiments

using the Aix200 machine and TMGa in combination with a double dilution

line using the Aix200/4 machine (cf sects. 3.2.1 and 3.2.2).

3.3.2 Optimization of quantum-dot active regions

The most important characterization tool for QD structures is room-

temperature PL. PL spectra reveal QD ground- and excited-state emission

as well as WL transition wavelengths. PL also enables to assess and com-

pare radiative efficiencies of QD layers. The influence of growth-parameter

changes on the properties of the QD layers can fast and comfortably be

monitored. Transmission electron microscopy images can reveal QD shapes

Fig. 14: Al1-xGaxAs oxidation rate at 420°C versus composition for 100-nm-thick

layers cladded by 100 nm thick GaAs layers. From Ref. 109.

3. Metalorganic chemical vapor deposition

and QD area densities and provide an important feedback during the devel-

opment of long-wavelength QDs. In contrast to PL, however, sample prepa-

ration and processing of TEM images are very time-consuming.

All samples for PL characterization were grown on undoped substrates.

The layer schematic of a typical quantum-dot test structure is shown in

Fig. 15. After a bake-out of the substrate at 700-750°C during several min-

utes, a GaAs buffer layer is deposited. The comparably large buffer thick-

ness of 300 nm is chosen to reduce the concentration of contaminants that

are present on the substrate surface and show diffusion into GaAs epilayers.

A thick AlGaAs layer of typically 60-70 % Al content is subsequently de-

posited. This layer acts on one side as a final diffusion barrier for contami-

nants from the substrate. On the other side, it suppresses the diffusion of

photo-excited carriers from the vicinity of the QDs to the substrate where

the lower crystal quality leads to non-radiative recombination. The QDs are

deposited after an intermediate GaAs layer grown at 600°C. After deposi-

tion of the QDs, a GRI is applied for the nucleation and the development of

the QDs12, 37. The first 2-3 nm of the GaAs cap are deposited at the QD

growth temperature. The reactor temperature is then ramped to 600°C. In

multilayer QD structures, QD and cap growth are repeated as indicated in

the schematic. The last QD cap layer is followed by a thin AlGaAs layer of

typically 30-40 % Al content. The low Al content allows growing all layers

above the QDs also at the QD deposition temperature. The top AlGaAs

layer is finally capped with a thin GaAs film to protect it from oxidation in

ambient air.

50 nm

20 nm

2-3 nm

100 nm

GaAs

In(Ga)As

AlGaAs

700°C

470-500°C

20-30 nm

10 nm

600°C

loop

growth temperature

300 nm

SI-GaAs (100) substrate

Fig. 15: Typical structure of an undoped sample containing one or more QD lay-

ers for PL measurements. AlGaAs diffusion barriers below and above the QDs

suppress carrier diffusion to the substrate and to the surface, respectively.

3.3. Modular optimization of laser structures

The samples are excited with the 514.5 nm line of an Ar+ ion laser,

driven at 500 mW output power. The excitation density can be varied with

neutral filters from typically 5 W/cm2 to 5000 W/cm2. A LN2-cooled Ge de-

tector is used for detection. Low temperature measurements are performed

using a closed-cycle cryostat (T = 8 - 300K).

Two room temperature PL spectra are usually taken directly after the

sample growth. From the comparison of low (Pexc = 5 W/cm2) and high

(Pexc = 5000 W/cm2) excitation density PL spectra with spectra of other

samples, the sample quality can be assessed. At low excitation densities

around 5 W/cm2, only few electron-hole pairs are generated per unit time,

the QDs are only partly filled if the QD density is not too low. For QD area

densities nQD > 1010 cm-2, only QD ground-state luminescence is visible in

the setup used. The FWHM of the PL peak is then a probe for the size dis-

tribution of the QD ensemble. The integrated QD ground state lumines-

cence is a measure of the crystalline sample quality: if QDs or the surround-

ing matrix have defects, carriers are trapped faster by these defects and re-

combine via non-radiative channels.

At high excitation-power density (Pexc = 5000 W/cm2), practically all

QDs are filled. Saturation of the QD ground state is inferred from the ap-

pearance of excited-state transitions in the spectra. Using appropriate refer-

ence samples, the peak intensities of the QD ground state luminescence can

be used to compare QD area densities. Fig. 22 shows typical examples of

low and high excitation PL spectra.

4. Self-organization of quantum dots

Self-organization of QDs can be observed if a thin QW is brought onto a

substrate with a different lattice constant. The strain energy of the epilayer

related to the lattice mismatch can be reduced if three-dimensional islands

are formed. QD formation is only observed if the energy gain connected

with the formation of such 3D islands is larger than the energy cost neces-

sary to create island facets and island edges. Self-organized In(Ga)As/GaAs

QDs as used in the present work have typical heights of 3-6 nm and lateral

dimensions of 15-25 nm. The area density of the QDs ranges between 1010

and 1011 cm-2.

For a fixed chemical composition, the emission wavelength of a QD en-

semble depends on the QD sizes and shapes. Whereas the emission wave-

length of QWs can easily be tuned by the QW thickness, QD size and shape

are complex functions of growth parameters such as reactor temperature,

deposition rate, and V/III ratio. The QD sheet density, likewise depending

on the growth conditions, is an additional parameter of QD layers with re-

spect to a QW. Since the gain of QD-based laser diodes directly scales with

the number of QDs in the cavity, the sheet density of QDs must pass a cer-

tain threshold value to enable laser operation.

The precise shape of SK QDs is still a debated issue. Most likely, the QD

shape depends on growth parameters and thus varies. Different shapes of

self-organized QDs such as pyramids, truncated pyramids and lens shapes

have been reported. A strictly pyramidal shape of InAs/GaAs QDs in both

MBE112 and MOCVD113, a multi-facetted dome structure for MBE-grown

InAs islands114 as well as a lens shape for MBE-grown InGaAs QDs115, 116

have been reported.

In the following, the basic mechanisms of quantum-dot self-organization

with respect to size, shape, and sheet density are described in order to sup-

port the understanding of the growth experiments conducted within this

work.

4.1. The equilibrium crystal shape

Whereas the energy balance of liquid surfaces is basically determined by

the surface tension, a multitude of contributions to the energy balance of

crystal surfaces exists. First, crystals are anisotropic. This implies that the

energy required to create a crystal surface depends on the surface orienta-

tion. Surfaces of different indices can have strongly differing energies. Sec-

ond, crystal surfaces exhibit intrinsic surface stress even in the absence of

lattice mismatch. This is due to the particular coordination of surface atoms

in contrast to bulk atoms. Surface stress at edges is analogous to the

4. Self-organization of quantum dots

Laplace pressure of curved liquids117. For these two reasons, the equilib-

rium crystal surface is not necessarily flat. This is impressively illustrated

by the theorem of Herring (1951)118: “If a given macroscopic surface of a

crystal does not coincide in orientation with some portion of the boundary

of the equilibrium shape, there will always exist a hill-and-valley structure

which has a lower free energy than a flat surface, while if the given surface

does occur in the equilibrium shape, no hill-and-valley structure can be

more stable.” This is in clear contrast to the case of liquids where the sur-

face tension always acts towards the establishment of a flat liquid surface.

Intrinsic surface strain originates from the fact that the equilibrium lattice

constant of surface atoms is generally different from that of the bulk since

surface atoms have a smaller number of direct neighbors. Being adjusted to

the bulk lattice constant, the atoms of the surface layer experience an in-

plane surface stress. If the equilibrium surface lattice constant is smaller

than in the bulk, the surface exhibits tensile strain, otherwise it is compres-

sively strained. The intrinsic surface stress cannot relax on a flat surface

since the surface stress tensor has non-zero elements only in the surface

plane. Another situation is given at crystal edges, however, where the sur-

face stress tensor has discontinuities. Here the intrinsic surface strain can

partially relax, either by compression or expansion of the near-edge bulk

crystal along the direction of the resulting surface stress. Intrinsic-strain re-

laxation gives a long-range energy contribution

∆

Eelastic to the overall en-

ergy.

An illustrative example of the relaxation of intrinsic surface stress is the

formation of nanofacets. The minimization of the surface energy leads to an

equilibrium period which is much larger than the lattice constant119.

4.2. Strained heteroepitaxy of thin films

If epilayers are regarded that are lattice-mismatched to the substrate, two

new aspects come into play. Firstly, the interface energy between the two

materials enters the energy balance. The values of the surface energy of

substrate (

S) and epilayer material (

E) as well as the interface energy (

determine if the epilayer wets the surface. If surface wetting takes place and

the lattice-mismatch strain is sufficient, QDs can form to reduce the total

energy of the system.

The energy gain per unit area upon deposition of the epilayer is given by

∆σ

E +

I -

On a lattice-matched substrate, uniform two-dimensional growth takes

place if

∆σ

< 0. This is called the Frank-van-der-Merve120 growth mode. If

4.3. Thermodynamic models of 3D island arrays

∆σ

> 0, three-dimensional coherent crystal islands are formed. This is

analogous to the formation of water droplets on oilcloth. This mode is

called the Volmer-Weber121 growth.

In the Stranski-Krastanow growth mode122 as used for the fabrication of

QDs in this work,

∆σ

is always negative, i.e. the lattice-mismatched epi-

layer wets the surface. Only above a critical layer thickness, a 2D-3D tran-

sition takes place. Coherent islands are formed that grow in size at the ex-

pense of the WL thickness. Whereas the WL is pseudomorphically strained

by the substrate, the lattice at the top of the QDs is rather relaxed. In the SK

mode, the WL can be thinned down to 1 ML by material transport to the 3D

islands but usually does not vanish123. An overview of the different growth

modes and their dependences on the sign of

∆σ

and the lattice mismatch is

given in Tab. 1 where the three growth modes are briefly compared with

liquids on solid surfaces.

Energy

gain

Liquid Crystal

∆σ

> 0 droplet

formation

Volmer - Weber growth

Substrate

lattice

match

Frank-

v.d.Merve

growth

Substrate

∆σ

< 0 wetting of

substrate

surface

lattice

mismatch

Stranski-

Krastanow

growth

SubstrateSubstrate

Tab. 1: Different growth modes of crystal epilayers on lattice-matched and lat-

tice-mismatched substrates compared to the deposition of liquids on solid sur-

faces. For the Stranski-Krastanow growth, surface wetting and a large lattice

mismatch are necessary.

4.3. Thermodynamic models of 3D island arrays

The energy of a strained QW can be reduced by the formation of 3D is-

lands. Following a theoretical model proposed by Shchukin et al.119, the

free energy gained by the formation of islands is made up of the elastic

strain relaxation, the surface energy related to the augmentation of the over-

all surface due to the creation of island facets, and the energy required to

create island edges. The energy gained by the formation of islands is ex-

pressed as

∆

Eisl =

∆

Efacets + Eedges −

∆

Eelastic.

4. Self-organization of quantum dots

The different energy terms on the right are explained as follows.

∆

Efacets describes the change of the surface energy due to the creation of

island facets; the total surface of a QD ensemble is larger than

that of a plane surface. Moreover, the facet planes have different

crystal orientations than the substrate. The surface energy of the

facets might differ from that of the substrate due to the general

anisotropy of the crystal surface energy.

Eedges is the short-range energy that equals the energy cost necessary to

create the islands edges.

∆

Eelastic is the energy related to elastic strain relaxation of the islands.

This term includes the relaxation of both lattice-mismatch strain

and intrinsic surface strain, edges

elastic

E∆, which relaxes at the island

edges.

The energy per atom of the strained layer gained by the 2D-to-3D transi-

tion is plotted in Fig. 16 as a function of the island base length L for a di-

lutev array of islands, and for different values of the parameter

is the

ratio of the change of surface energy,

∆

Esurf, to the absolute value of the en-

ergy contribution of intrinsic-strain relaxation at the island edges, edges

elastic

E∆.

If α ≤ 1, an optimum island size L0 exists, corresponding to a global mini-

v In a dilute island array, the island area density is low enough that island-island

interaction via long-range elastic energies, mediated by wetting layer and sub-

strate, can be disregarded.

Fig. 16: Free energy gained by the 2D to 3D transition of a pseudomorphically

strained WL as a function of island base length. Curves for different values of the

paramter α (cf. text) are plotted. From Ref. 119.

4.4. Kinetic description of island formation in one dimension

mum of the free energy. In case α > 1, an equilibrium island size does not

exist. The reduction of the free energy corresponds to a continuous increase

of the island size. Mass conservation stipulates that the island density de-

creases. Larger islands grow at the expense of smaller islands. This is re-

ferred to as Ostwald ripening124.

Whereas the theory by Shchukin is based on continuum elasticity only,

calculations by Wang and coworkers125 additionally regard atomistic as-

pects using density-functional theory. They found that the energy contribu-

tion of edges is too small to be decisive for the existence of an equilibrium

island size. In their opinion, the equilibrium island size is rather the result of

an energetic balance that governs the material transport between the WL

and the islands: They found that the surface energy of the WL,

WL, is a

function of the film thickness:

WL decreases significantly between 0 and

2 ML and is constant only above 2 ML. The non-zero gradient of

WL as a

function of layer thickness constitutes a force counteracting the thinning of

the WL during the 2D-3D transition.

4.4. Kinetic description of island formation in one dimen-

sion

The thermodynamic models outlined in the previous section describe the

morphology of a thin lattice-mismatched epilayer with the lowest surface

energy. The process of the 2D-3D transition and the subsequent size in-

crease of the QDs towards an eventual equilibrium size cannot be accounted

for within a thermodynamic model. Barabási et al. have introduced an at-

omistic model in one dimension126, describing the process of QD self-

organization. Although this model is subject to controversially debated

simplifications, such as the restriction to one dimension, a number of basic

properties of SK QDs are described which can be observed experimentally.

In the model of Barabási, atoms are successively deposited with a con-

stant deposition rate on a substrate with a smaller lattice constant than the

epilayer, thus accounting for compressive strain. The case of a larger lattice

constant is not discussed within this model. The hopping probability of an

atom to neighboring lattice sites is proportional to exp[µ(x)/kBT] with

≅)x(

))(n( s0n xEEE −+− being the chemical potential. n is the number of

neighboring atoms, En is the energy of the bonds with nearest atoms, E0 is

the diffusion barrier seen by an isolated atom on a stress-free substrate sur-

face, and Es(x) is the strain energy of the surface at the position of the con-

sidered adatom. The equilibrium position of each lattice atom is iteratively

recalculated after each hopping event. Numerical simulations show that for

a lattice mismatch above 5 %, an island size distribution with a narrow peak

evolves. In Fig. 17, the strain energy of a surface atom as a function of posi-

4. Self-organization of quantum dots

tion x is plotted. The strain energy has a maximum at the island edges, the

smallest strain energy value is found at the island top.

Since µ(x) only depends on the strain energy Es, a net surface current

(x)

−∇=jis generated, pointing away from the island edge. Since the peak

strain energy at the island edges grows with the island size, j increases

likewise. The increasing net surface current j suggests a self-limitation of

the island size so that a seemingly stable array of equisized islands exists.

The model suggests that the size increase of the QDs slows down with their

size. An equilibrium size might therefore be assumed asymptotically in

time. Whereas QDs form very quickly after the critical layer thickness is

reached, they might need a rather long time to achieve equilibrium.

4.5. Island size and density

The right value of the emission wavelength and large modal gain are the

most important properties of a QD laser. These properties are directly re-

lated to the QD size and QD sheet density, respectively. The models de-

scribed above do not show how these quantities depend on the growth con-

ditions. The impacts of QD growth temperature as well as of deposition

amount, deposition rate and GRI after deposition of the QD material on size

evolution and equilibrium sizes are outlined in this section.

Fig. 17: Strain energy around a spontaneously formed island in a one-dimensional

atomistic model by Barabási126. Es is the strain energy of an atom placed on the

top of the substrate or on the island. Es is largest at the island edge. From

Ref. 126.

4.5. Island size and density

4.5.1 Role of temperature

A two-dimensional Monte-Carlo simulation of compressively strained

submonolayer islands introduced by Meixner et al.127 reveals the main fea-

tures of the temperature dependence of island size and density. The results

of this simulation are in good agreement with experimental findings for 3D

SK QD systems so that the portability of the main aspects of the 2D sub-

monolayer island system to the 3D island system is likely128.

Fig. 18 shows the calculated average island diameter as a function of

time for deposition temperatures of T = 675K, 700K and 725K. The QD

material was deposited within 0.04 s. The average island diameter is calcu-

lated from island-size distribution histograms taken at different times during

the GRI following the deposition of the island material. It is evident from

Fig. 18 that during the first 5 seconds of the GRI, the size distribution is ki-

netically controlled: Larger islands of a lower density are preferred at

higher temperatures since surface diffusion lengths are larger: Adatoms

rather attach to already existing islands than form new nuclei. In the kineti-

cally controlled regime where the system is not given enough time to

equilibrate, an increase of the temperature thus leads to an increase of the

island size, and, due to mass conservation, to a lower island density. For

long GRIs where thermodynamics prevails, smaller island sizes and larger

island densities are preferred at higher temperatures. At low temperatures,

Fig. 18: Monte-Carlo simulations of the temporal evolution of 2D submonolayer

islands. A coverage of 4 % was deposited randomly on the surface at a flux of

1 ML/s. Every 0.01 s, a histogram of the island size distribution is recorded. The

average island diameter 〈N〉 is plotted as a function of GRI duration for differ-

ent temperatures. The hopping probability p of an adatom is proportional to

exp[-(E0 + nEn - Es) / kBT] where E0 is the adatom diffusion barrier, En the chemi-

cal bond energy to a neighbor atom, n the number of chemically bound neighbors,

and Es the strain energy at the position of the considered adatom. From Ref. 127.

4. Self-organization of quantum dots

islands are larger and hence their sheet density is lower. The dependence of

island size and island density on temperature in the kinetically controlled

regime and in the thermodynamically controlled regime is exactly opposite.

During the GRI where the QD system moves towards equilibrium, a cross-

over of kinetics and thermodynamics is observed.

The decrease of the island size with temperature was experimentally

verified for InAs submonolayer islands (30 % coverage) buried in a GaAs

matrix129. PL as well as high-resolution TEM images evaluated according to

the DALI method130 were used as probes. Oshinowo et al.113 observed an

increase of the diameter of MOCVD-grown InGaAs/GaAs QDs from 15 to

20 nm if the QD deposition temperature is raised from 500 to 550°C. The

QD sheet density concurrently decreases from 1011 cm-2 to 1010 cm-2. The

larger QD density at 500°C can be explained with reduced surface kinetics

as compared to Tgr = 550°C. The apparent inconsistency between kinetic

theories and thermodynamic predictions occasionally found in literature

might simply be attributed to the fact that only time is short for the system

under consideration to assume the equilibrium shape, i.e. the crossover re-

gion might not have been reached in those cases.

In MOCVD, temperature is not a suitable parameter to tune the QD den-

sity since the crystal quality is very sensitive to temperature changes at

typical In(Ga)As/GaAs QD deposition temperatures around 500°C. Growth

temperatures lower than 500°C would lead to a drastic decrease of the crys-

tal quality. This effect is particularly pronounced in MOCVD where the or-

ganic parts of the metalorganic precursors can lead to significant carbon

doping. The QD deposition temperature of about 500°C is already very low

as compared to temperatures of more than 600°C used for GaAs layers to

obtain optimum crystalline quality. Low crystal quality leads to a reduced

radiative efficiency of the QDs. At significantly higher temperatures than

500°C, self-organization of QDs in the SK growth mode is not possible due

to thermodynamic arguments. In addition, even a slight increase of the tem-

perature may lead to a drastic enhancement of the probability of cluster

formation. The temperature window for the fabrication of high-quality QDs

is therefore rather small. Particularly in MOCVD of QDs for optoelectronic

applications, a compromise between growth temperature, QD crystalline

quality and QD size must be found. GRIs are often chosen during which the

system might not have enough time to reach the thermodynamically con-

trolled regime.

Within the present work, a QD equilibrium size could not be found in

any of the described experiments. A steady size increase of uncapped QDs

with time was always observed. For kinetic reasons as described in sect.

4.4, the size increase of the QDs slows down with time and would make the

assumption of an equilibrium size very time-consuming. Moreover, the time

4.5. Island size and density

the system can be given to assume an equilibrium surface morphology is

limited by the onset of dislocation-cluster formation setting on few minutes

after the deposition of the QD material. The formation of dislocation clus-

ters can be verified using TEM or AFM images.

4.5.2 Impact of deposition amount

It has been shown that the amount of deposited material has a strong in-

fluence on the QD density131-134. For MOCVD of InGaAs/GaAs using ar-

sine, Leon and coworkers131 have observed that after the critical layer

thickness is reached, the QD density quickly increases and then reaches a

saturation density after deposition of about 1.1 times the critical thickness.

The QD density does not increase significantly if QD material deposition is

continued. The saturation density was found to be larger for smaller V/III

ratios. The rapid increase of the area density of MOCVD-grown InGaAs

QDs with the amount of deposited material was confirmed within this work:

AFM of an In0.8Ga0.2As QD sheet capped with a 2 nm thick GaAs layer pre-

vious to cooling shows a QD density of (2.2 ± 0.1) × 1010 cm-2. The thick-

ness of the In0.8Ga0.2As deposited was roughly the critical layer thickness.

Increasing the deposition amount by only 6 % leads to a QD density of

(3.3 ± 0.3) × 1010 cm-2, i.e. to a density increase of about 50 %.

4.5.3 Influence of growth interruption

The wavelength of the QD ground-state transition increases with the GRI

after the deposition of the QD material. During the GRI, the QDs grow in

size at the expense of the WL thickness. This behavior is analogous to the

size increase of two-dimensional submonolayer islands as described in sect.

4.5.1, and corresponds also to the one-dimensional model introduced in

sect. 4.4.

It is noteworthy that the duration of the GRI after deposition of the QD

material was found not to influence the QD density for MOCVD-grown

In0.8Ga0.2As QDs using arsine. This finding suggests that kinetic effects

dominate the adjustment of the island density during – or shortly after – the

deposition of the QD material.

4.5.4 Importance of growth rate

The QD density was furthermore found to depend on the growth rate.

This is valid for both the InAs/GaAs133, 135 and the InP/GaAs136 materials

system. A rate equation approach for InP island on GaAs substrates was re-

cently reported137 in which both kinetic and thermodynamic properties are

4. Self-organization of quantum dots

considered. It was found that the nucleus density is higher for larger growth

rates.

Experiments were carried out in this work to tune the QD density and

also the size dispersion by variation of the growth rate. Unfortunately, the

PL efficiency of the QD layers is very sensitive to a variation of this pa-

rameter. The QD layers used for the growth of laser diodes were generally

optimized to have a maximum radiative efficiency. The influence of the

deposition rate on the laser-diode characteristics were not systematically

investigated during the optimization procedures. The inhomogeneous lumi-

nescence broadening of the QD ensembles used for the growth of the laser

diodes is between 60 and 70 meV at room temperature. The QD sheets used

in the active region of the laser diode structure with the lowest transparency

current density have a FWHM of around 48 meV.

Scanning tunneling microscopy (STM) studies of uncapped MBE-grown

InAs/GaAs dots show that QD density and also the QD size dispersion de-

crease significantly when the growth rate is reduced138. PL studies per-

formed on MBE-grown buried InAs/GaAs QD samples show that the emis-

sion wavelength increases with decreasing growth rate, reaching a maxi-

mum around 1.3 µm with the inhomogeneous broadening decreasing from

44 to 27 meV139. For the growth of 1.3 µm QDs, an extremely low growth

rate of 0.003 ML/s was used. The low growth rate enables material redistri-

bution towards a very narrow size dispersion. The QD density of these

1.3 µm QDs, however, is clearly below 1 × 1010 cm-2. A very small value of

the inhomogeneous line broadening of 26 meV for MBE-grown InAs QDs

emitting at 1.25 µm has also been achieved at higher growth rates but with

punctuated deposition of the QD material140.

It can be concluded from these findings that the QD density is highest for

large deposition rates. To achieve large QDs with small size dispersions by

material agglomeration from the WL and by material redistribution between

the islands, the system must be given a certain time. This can be achieved

with long-enough GRIs.

5. MOCVD of quantum-dot structures for laser di-

odes

Although quantum dot material gain is known to exceed quantum well

material gain by far70, the low area coverage of QDs leads to lower modal

gain per QD sheet. For large modal gain of QD lasers, high QD densities

are required12. This makes close stacking of QD layers necessary, thus in-

creasing the QD volume fill factor. Due to the three-dimensional morphol-

ogy of the QDs and the low temperatures at which QDs and subsequent

GaAs capping layers on top of QDs must be grown, GaAs deposited on QD

sheets exhibits significant surface corrugations. It is not possible to deposit

high-quality QDs on such rough surfaces. Therefore, an in-situ annealing

step was developed within this work during which such growth fronts are

flattened (cf. sect. 5.1.1).

The threshold currents and quantum efficiencies of MOCVD-grown la-

sers based on self-organized QDs were not yet close to the ideal limits72.

This is primarily related to the general difficulty to grow strained het-

erostructures at low temperatures with high crystalline quality. The strain

arising from the lattice mismatch implies the danger of plastic strain relaxa-

tion by defect formation. In MOCVD, defect formation is additionally fa-

vored by enhanced surface kinetics141 as compared, for example, to MBE.

In MBE, the As background pressure for the growth of defect-free InAs

QDs at 480°C can be detuned37 by ± 50 % from the standard value of

0102 −

×=

p Torr = 2.7 × 10-6 mbar, regardless of the deposition rate. Only

for extraordinarily arsenic-rich growth conditions of AsAs pp 0

3×= , forma-

tion of large defect clusters at the expense of density and size reduction of

the remaining QDs is reported.

During MOCVD of QDs, the V/III ratio and also the deposition rate have

to be adjusted carefully37. Detuning the V/III ratio by 50 % from the value

for optimum luminescence efficiency usually leads to a decrease of the lu-

minescence efficiency by up to an order of magnitude. A possible explana-

tion is based on the fact that MOCVD growth of QDs takes place under

comparably arsenic-rich conditions: The AsH3 partial pressure of typically

1-3 × 10-1 mbar is by five orders of magnitude higher than the typical As

pressure in MBE. Although an extremely small cracking efficiency of AsH3

around 10 % can be assumed at 500°C, the partial pressure of reactive As

species is probably still much higher in MOCVD than in MBE. This argu-

ment likewise applies to TBAs. The cracking efficiency of TBAs at 500°C

is 100 %, i.e. up to 10 times larger than that of AsH3 (cf. Fig. 7). However,

the partial pressure of TBAs supplied is typically 10 times lower than that

of AsH3

77. The AsH3 / TBAs partial pressure cannot simply be reduced in

MOCVD since high AsH3 partial pressures are nevertheless mandatory for

5. MOCVD of quantum-dot structures for laser diodes

the formation of QDs with a reasonable sheet density and sufficient optical

quality142. The enhanced tendency of MOCVD-grown QDs to form defects

might also be associated with the reactants used. Particularly, atomic hy-

drogen stemming from the decomposition of AsH3 is expected to affect ki-

netics of QD formation. Hydrogen radicals are known to interact with the

surface by breaking bonds due to their high reactivity and thus affect the

surface adatom concentration141. Moreover, hydrogen atoms are not the

only radicals observed during MOCVD; alkyl radicals formed during pyro-

lysis of the metalorganic precursor molecules may also play an important

role90. The in-situ annealing step described in sect. 5.1.1 that has initially

been introduced to flatten corrugated growth fronts has a beneficial impact

on the crystal quality of QD structures as described in sect. 5.1.2. The

threshold current densities of QD lasers could significantly be lowered by

in-situ annealing as described later in chapter 6.

Sect. 5.2 reports the fabrication of QD sheets with the alternative arsenic

precursor tertiarybutylarsine (TBAs) and works out differences with regard

to the growth with arsine.

Lasing emission at 1.3 µm from MOCVD-grown QD lasers has not been

reported yet. In sect. 5.3, experiments carried out to increase the QD emis-

sion wavelength towards 1.3 µm are presented. Sect. 5.3 comprises a brief

review of alternative wavelength shifting techniques. These experiments

were carried out using alternative-precursor MOCVD.

5.1. In-situ annealing of QD structures

The growth of GaAs-based SK QDs must take place in a finite tempera-

ture window, roughly between 430 and 530°C. At lower temperatures, sur-

face kinetics is insufficient to achieve the 2D-3D transition. At higher tem-

peratures, entropy effects tend to flatten the three-dimensional morphology.

Typical growth temperatures of the GaAs material systems, however, are

600°C and more. At QD growth temperatures, surface corrugations are

caused by the presence of the underlying three-dimensional QDs. Such cor-

rugations are only slowly planated during the growth of the GaAs cap on

top of the QDs. If a sample should contain more than one QD layer, rees-

tablishing a flat growth front of the GaAs matrix before depositing a subse-

quent QD layer is strongly required. QDs deposited on a corrugated surface

show very broad low-intensity PL spectra, indicating a large QD size distri-

bution and the presence of defects. Such QDs are not suited for the deploy-

ment in active regions of lasers.

5.1. In-situ annealing of QD structures

5.1.1 Flattening of the growth front

Several methods were evaluated to flatten the growth front of GaAs cap

layers on top of self-organized QDs. For cap thicknesses of more than

50 nm, ML-flat surfaces can be reestablished even if the cap is grown at

around 500°C. To increase the QD volume density by close vertical stack-

ing, however, flat surfaces should be achieved for much thinner cap layers.

Increasing the growth temperature during growth of the cap significantly

enhances the surface flatness. The best results were achieved, however,

with GRIs at high temperatures73, 143.

Below the QD sheets, the sample structure used for the flattening ex-

periments is the same as for the PL structures as shown in Fig. 15 (sect.

3.3.2). The In0.8Ga0.2As QDs were deposited at 490-495°C on 100 nm GaAs

which was grown at 600°C. Subsequently, a GRI of 60 s without arsine was

applied for the nucleation and the development of the QDs12, 37. GaAs caps

of 7 nm and 30 nm thickness were deposited on the QDs under different

growth conditions. Growth was stopped hereafter and the samples were

cooled under As stabilized conditions. The surface morphology was subse-

quently investigated by contact-mode atomic force microscopy (AFM). In

order to avoid degradation of the QDs, the growth temperature was not in-

creased before 3 nm GaAs were deposited on the QDs at the QD growth

temperature. A reference sample (A) was grown in which the QD layer was

omitted. In this reference sample, 30 nm GaAs were directly deposited at

490°C on 100 nm GaAs grown at 600°C.

Fig. 19 (a) shows a surface AFM image of sample A. The image shows

monolayer terraces and a few monolayer-high islands that are elongated in

the <110> direction. Fig. 19 (b) shows a surface AFM image of a QD sam-

ple (sample B). In this sample, 30 nm GaAs were deposited at 490°C on a

single sheet of QDs that were also grown at 490°C. The surface of sample

B exhibits a morphology which differs significantly from that of sample A.

Up to 10 nm high and elongated hillocks are visible with mean lateral ex-

tensions of 80 nm × 280 nm, all elongated in the <110> direction.

Fig. 19 (c) shows a plan-view TEM image of a sample with a single sheet

of QDs buried with GaAs (sample C). The QD density of this sample is

nQD = 3.4 × 1010 cm-2. The QDs of sample C were deposited under the same

growth parameters as those of sample B. Thus the QD density of sample B

is also nQD = 3.4 × 1010 cm-2.

The hillock density at the surface of sample B amounts to 2.5 × 109 cm-2

which is by a factor of 12.8 lower than the QD density. One hillock thus

buries roughly 13 QDs. The hillock height and the QD height (5-10 nm (cf.

Ref. 144)) are very similar whereas the lateral extension of the hillocks is

about 5 – 10 times larger than that of the QDs. It can be concluded from the

TU4472

5. MOCVD of quantum-dot structures for laser diodes

difference of QD density and hillock density that the hillock morphology is

not a continuation of the QD morphology. It is likely that the cap surface

morphology is due to the spontaneous formation of nano-facets. Such nano-

faceting has been observed by Kasu et al. during MOCVD of GaAs on vici-

nal GaAs(001) surfaces145. Although exactly oriented GaAs(001) substrates

were used in the present experiments, the QDs lead to local variations of the

surface orientation so that nano-faceting can locally take place. At growth

temperatures of 600°C and more, nano-faceting in this material system is

thermodynamically unfavorable. Facets that have formed at low tempera-

tures blur above 600°C (cf. Ref. 146).

The ML-high islands on top of sample A as well as the large hillocks of

sample B are significantly elongated in the <110> direction. This is likely

due to the influence of surface kinetics. Two effects can account for this

phenomenon: An anisotropy of the surface diffusion coefficient DS (for ex-

ample DS><110 > DS>< 101 ) or an anisotropy of the sticking probability

of migrating adatoms to >< 101 -oriented steps (A-steps) and ><110 -

oriented steps (B-steps). Kasu and Kobayashi147 attributed island elonga-

tions similar to those of Fig. 19 (a) and (b) to an anisotropy of the sticking

probability of migrating adatoms. They determined the sticking probability

to be larger at A-steps than at B-steps. The difference is considered to be

vi By courtesy of N.D. Zakharov and P. Werner, Max-Planck Institute of Micro-

structure Physics, Halle, Germany.

Fig. 19: (a) Reference sample without QDs. A 30 nm thick GaAs layer was di-

rectly deposited at 490°C on 100 nm GaAs grown at 600°C. (b) AFM image of a

30 nm thick QD-burying GaAs cap deposited at 490°C on QDs that were also

grown at 490°C. The hillock density amounts to 2.5 × 109 cm-2. The picture has

been published in Ref. 73 (c) Plan-view TEM imagevi of a single-sheet QD sam-

ple. The QD density is about 3.4 × 1010 cm-2. The underlying QDs were deposited

under the same parameters as those of sample (b).

5.1. In-situ annealing of QD structures

due to the anisotropy of the arsenic-rich c(4×4) surface reconstruction

which is the preferred reconstruction on exactly oriented GaAs(001) sur-

faces grown under As-rich conditions such as in MOCVD. Calculations by

Ito et al.148 confirm the anisotropy of the sticking probability. Heller and

Lagally149 have investigated MBE-grown GaAs(001) surfaces. In contrast

to the case of MOCVD-grown GaAs(001) surfaces, they determined a lar-

ger sticking probability to B-steps which they attribute to the anisotropy of

the (2×4) reconstruction they determined using STM. Employing first-

principles total-energy calculations, Kley et al.150 have reported a lower

adatom diffusion barrier in the >< 101 direction than in the ><110 -

direction for the MBE-typical β2(2×4) surface reconstruction. According to

their calculations, the adatom diffusion is faster parallel to the As dimers of

the β2(2×4) unit cell, favoring nucleus elongation in the >< 101 direction.

The references cited in this paragraph refer to ML-high nuclei as one can

see in Fig. 19 (a). However, the above considerations are likely to be valid

also for the case of elongated hillocks such as in Fig. 19 (b).

A first step towards reestablishing a flat growth front was to enhance the

surface mobility of adatoms of the cap surface by increasing the growth

temperature. However, before increasing the temperature for the growth of

the cap layer, the QDs were covered by 2-3 nm GaAs at the QD growth

temperature. An earlier rise of the temperature results in a decreased PL

signal from the sample, presumably due to defect formation in the QD

layer. The growth temperature was kept below 600°C to avoid intermixing

effects of In from the QDs with Ga of the matrix. This would result in an

undesired blueshift100 of the QD emission wavelength.

The surface morphology of a sample for which the growth temperature

of the cap was ramped to 590°C is shown in Fig. 20 (a). It is completely dif-

ferent from that of sample B. It exhibits a ML step landscape showing ter-

races and ML islands. As compared to sample A (Fig. 19 (a)), the terraces

are significantly frayed, however, and the ML island density of

2.8 × 108 cm-2 is about five times larger. A sample grown at the intermedi-

ate temperature of 535°C exhibits a surface similar to sample B but with

half the hillock height.

It was also tried to improve the result shown in Fig. 20 (a) by insertion of

a short-period GaAs/AlGaAs superlattice. Similar AlGaAs/GaAs superlat-

tices are typically used in MBE of GaAs-based edge emitting lasers with

AlGaAs cladding layers of high aluminium content to reduce surface corru-

gations related with the thick AlGaAs layers. In such layer structures, short-

period superlattices are grown directly below and above each cladding

layer. A QD sample was grown within this work which differs from that

shown in Fig. 20 (a) by the insertion of 3 periods of a GaAs (2.7 nm) / Al-

GaAs (1.8 nm) superlattice, keeping an overall cap thickness of 30 nm. The

5. MOCVD of quantum-dot structures for laser diodes

temperature was ramped to 590°C before the growth of the superlattice.

AFM of the surface was performed revealing an insignificant reduction of

the ML island density by 14 % to 2.4 × 108 cm-2, and an increase of the av-

erage ML island area by 40 % to 3.3 × 104 nm2. So the insertion of an

Al0.65Ga0.35As/GaAs superlattice does not lead to an improvement of the

surface flatness which is considered relevant for the subsequent deposition

of QDs.

The most efficient flattening is achieved if the growth is stopped after the

deposition of the cap and if GRIs of several minutes are applied subse-

quently at 600°C. This method is very efficient even for only 7 nm thick

caps. In the experiment, the first 2-3 nm of the GaAs caps were grown at

the QD deposition temperature. The growth temperature was then ramped

to 600°C during the last 4-5 nm. Three samples were grown with GRIs of 0,

10, and 20 min. The samples were weakly arsenic-stabilized during the

GRIs with an AsH3 partial pressure of 5.4 × 10-2 mbar. For the as-grown

case (GRI = 0), the surface shows hillocks with an average height of 4 nm

and a hillock density of 8.2 × 108 cm-2, similar to sample B. However, these

hillocks have an atomically flat top of which the average area amounts to

2.0 × 104 nm2. Fig. 20 (a) shows an AFM image of the surface annealed

during 10 min. The surface exhibits ML steps again, the ML island density

is as low as 1.5 × 108 cm-2 and the average island area amounts to

8.5 × 104 nm2. These values are not significantly improved for 20-30 min

annealing.

3 nm

0 nm

3 nm

0 nm

(a) (b)

Fig. 20: (a) AFM surface image of a sample where the growth temperature of the

30 nm thick GaAs cap was ramped in-situ to 590°C. The picture has been pub-

lished in Ref. 73. (b) AFM surface image of a sample where the QDs were capped

with 7 nm GaAs. Subsequently, the surface was annealed during 10 min at 600°C

under As stabilization.

5.1. In-situ annealing of QD structures

The surface of the sample with a 7 nm thick cap annealed during 10 min

(Fig. 20 (b)) is comparable to the surface of the 30 nm thick cap shown in

Fig. 20 (a). The decisive advantage of the annealing step as compared to

simple temperature ramps or the integration of a superlattice is the ex-

tremely small thickness of the cap for which a ML-step surface profile can

be reestablished. This has a strong impact on the layer qualities in dense

QD stacks.

Fig. 21 shows a cross-sectional dark-field TEM image of a five-fold

In0.8Ga0.2As QD stack grown with 10 min in-situ annealing at 600°C after

each QD layer was capped with 7 nm GaAs. The vertical sheet-to-sheet dis-

tance is about 18-20 nm. For such thin GaAs spacers, the strain fields gen-

erated by the lattice-mismatched QDs lead to vertical QD coupling. The

vertically correlated QDs marked with dashed white lines in Fig. 21

broaden from layer to layer. The vertical correlation is due to a minimum of

the elastic-energy density on the spacing-layer surface exactly above buried

QDs151. This minimum exhibits a fine structure of four shallow minima ar-

ranged in cloverleaf geometry above the strained QDs as shown in Fig. 1 of

Ref. 151. This fine structure can possibly account for the increase of the QD

widths from sheet to sheet.

PL spectra from such vertically correlated stacked QD sheets show very

broad transition lines, most likely due to the size increase of the QDs from

sheet to sheet. This is disadvantageous for laser diodes where gain is to be

maximized within a narrow spectral region. For the application as active

medium in lasers, the spacer of such QD layers is increased until correlation

is completely lost. Then the QDs of all layers have similar sizes and shapes

and the inhomogeneous broadening decreases. The occurrence of vertical

QD correlation which is a function of the QD size, the QD strain and the

sheet-to-sheet distance limits the number of QD sheets that can be used in

an optical confinement layer of a given thickness.

vii by courtesy of N.D. Zakharov and P. Werner, Max-Planck Institute of Micro-

structure Physics, Halle, Germany.

50 nm

correlation

TU5338

Fig. 21: Cross-sectional dark-field TEM imagevii of a five-fold In0.8Ga0.2As QD

stack grown with in-situ annealing after each QD layer was capped by 7 nm

GaAs. The vertical sheet-to-sheet distance is 18-20 nm.

5. MOCVD of quantum-dot structures for laser diodes

To conclude, a strong reduction of growth front corrugations of the GaAs

cap layer on top of InGaAs QDs and the reestablishment of a smooth sur-

face were demonstrated. An elevated overgrowth temperature which can be

adjusted after the QDs are covered by 2-3 nm of GaAs is essential. The best

results were obtained for a cap layer thickness of only 7 nm with subse-

quent annealing during 10 min at 600°C.

5.1.2 Optical properties of annealed QD structures

The influence of the annealing step on the optical properties of the QDs

was also investigated. For this purpose, samples were grown as described in

the previous section. The samples were additionally capped with 20 nm of

Al0.33Ga0.67As, however, acting as a charge-carrier diffusion barrier to sup-

press non-radiative surface recombination. The measurements were carried

out at room temperature. Low excitation densities were used to probe the

emission wavelength and the optical quality of the QDs. High excitation

densities where both QD and WL states are saturated were used to investi-

gate the radiative recombination in the GaAs matrix.

Fig. 22 shows PL spectra of undoped test structures for four different an-

nealing times (tA = 0, 10, 20, 30 min) taken at low (a) and high (b) excita-

tion density. The low-excitation spectra show a blueshift of the QD emis-

sion wavelength and a slight reduction of the peak intensity for annealing

during 10 min. The blueshift is attributed to some In-Ga intermixing100. The

decrease of the peak intensity, however, cannot be explained using this ar-

gument. When the annealing time is extended to 30 min, a monotonic in-

crease of the peak intensity of the QD luminescence is observed which is

attributed to a reduction of defect-related non-radiative recombination

channels152. The high-excitation spectra of Fig. 22 (b) show that the GaAs

matrix intensity increases by a factor of about two upon annealing for 10

and 20 min. A GRI of 30 min, however, leads to a reduction of the PL-

intensity of the GaAs matrix. In the high-excitation spectra, the saturated

PL intensity from the QDs remains quasi unchanged for all GRI durations.

As an explanation for the increase of the GaAs luminescence, a reduction

of two different kinds of defects is suggested that are both related to the

growth of the QD layers. First, misfit dislocations might be formed in a

very small fraction of the QDs, most likely in very large ones37. These dis-

locations proceed into the GaAs on top of them and strongly reduce the

GaAs luminescence, acting as non-radiative recombination centers. In con-

trast to Ref. 6 where the dislocation density is enhanced upon post-growth

annealing, an in-situ annealing step with the surface being very close to the

spatial origin of the dislocations enables strong reduction of the number of

dislocations, as demonstrated in Ref. 152 by deep level transient spectros-

5.1. In-situ annealing of QD structures

copy. However, as the number of dislocation-containing or -inducing QDs

is small compared to the number of QDs in which recombination occurs ra-

diatively, no significant improvement of the QD luminescence is observed

upon annealing. It is therefore assumed that the QDs are of high crystalline

quality even without annealing. Secondly, point defects in the first 2-3 nm

of the low-temperature GaAs grown at 490°C immediately on top of the

QDs may act as non-radiative recombination centers. Typical MOCVD

growth temperatures for low-defect-density bulk GaAs are in the range of

600 to 650°C (Ref. 153). GaAs grown at lower temperatures tends to build

non-equilibrium point defects.

The annealing step described here is also useful to eliminate large dislo-

cated clusters from MOCVD-grown QDs emitting at 1300 nm154. In this

case an improvement of the QD PL intensity by a factor of two was found.

The selective deposition of AlAs on defect-free regions of cluster-

Fig. 22: Room-temperature PL spectra of annealed QD samples at low excitation

density (a) (5 W/cm2) and at high excitation density (b) (5 kW/cm2). In (a), re-

combination occurs essentially on the QD ground state. Luminescence from the

WL and the matrix is not visible. In (b), radiative recombination mainly takes

place in the GaAs matrix. The stars denote transitions from the excited QD states.

This figure has been published in Ref. 73.

5. MOCVD of quantum-dot structures for laser diodes

containing InGaAs surfaces enables the selective evaporation of such de-

fect-containing sites if annealing is applied subsequently155.

5.2. Alternative-precursor MOCVD of InGaAs QDs

It will be shown here that QDs grown using TBAs in a suitable growth

regime also have excellent optical properties, very similar to QDs grown

using AsH3. Conclusive evidence of the high quality of self-organized QDs

grown with alternative-precursor MOCVD is supplied by the realization of

QD lasers as reported in sect. 6.1.2, proving the general suitability of this

QD growth process also for other optoelectronic devices.

For PL studies, undoped structures containing only a single

In0.67Ga0.33As/GaAs QD sheet were grown under various growth conditions,

applying a GRI of 60 s after deposition of the QD material. The In fraction

of x = 0.67 was determined using x-ray diffraction of a ten-fold InGaAs

QW stack. The QWs of this stack were deposited under the same growth

parameters as the QDs but the QW thickness was kept below the critical

value for the formation of SK QDs. Room-temperature PL spectra of an

InGaAs QD test structure grown at 485°C using a V/III ratio of 1.5 are

shown in Fig. 23. For high excitation power, emission from the QD ground

state and first excited state as well as transitions from the WL and the ma-

trix is clearly visible. At low excitation power, PL mainly originates from

the QD ground state, the inhomogeneous broadening being 70 nm

(65 meV). A very large QD density of 9.6 × 1010 cm-2 is determined from

the plan-view transmission electron micrograph shown in the inset of

Fig. 23. Both PL and transmission electron micrograph show a unimodal

coherent QD growth without defect clusters. These results are comparable

to data of QDs grown using AsH3 and used as active medium for high-

power lasers156, 157.

Using AsH3, V/III ratio and deposition temperature are crucial growth

parameters for good optical and structural properties of InGaAs QDs142.

Therefore, the peak wavelength of the ground-state transition of

In0.67Ga0.33As QDs grown with TBAs as a function of these parameters is

studied using low-excitation-density PL (Fig. 24). An increase of the V/III

ratio leads to a pronounced redshift of the peak wavelength, indicating an

increased size of the QDs. Similar observations were reported for MOCVD

of InGaAs QDs using AsH3

158 and for MBE of InAs QDs37, indicating a

general behavior which is independent of growth technology and chemistry

of the As precursor. The dependence can be explained by considering the

surface energy: Calculations show that an increase of the arsenic pressure

leads to a decrease of the surface energies of both GaAs(001)159 and

InAs(001)160. It is natural to assume that a strained InGaAs WL on

5.2. Alternative-precursor MOCVD of InGaAs QDs

GaAs(001) shows a similar tendency. Calculations of the thermodynamics

of QD formation indicate that the QD equilibrium size rises if the surface

energy of the WL is reduced119.

In contrast to the dependence on the V/III ratio, the variation of tempera-

ture has different impacts on the emission wavelength of QDs grown using

either AsH3 or TBAs. The PL of TBAs-grown QDs shows a slight blueshift

with increasing temperature above 485°C (Fig. 24). For AsH3-grown In-

GaAs QDs a strong redshift of the QD ground-state transition is always ob-

served142 if the temperature is increased. This is most likely related to a in-

crease of the QD size. The equilibrium size of SK QDs has been predicted

to decrease with temperature, using thermodynamic equilibrium argu-

ments119. Actual growth processes, however, usually imply too short time

intervals for formation and overgrowth of such QDs for reaching an equilib-

rium size. This applies particularly for lower growth temperatures at which

the equilibrium size is large and the adatom mobility is low, as it is the case

for the sample in Fig. 24 grown at 470°C. The emission wavelength of this

sample is expected to be further in the red as compared to the other sam-

ples. Thus, decrease and increase of QD size with temperature may occur,

depending on the balance between thermodynamics and kinetics127.

Using AsH3, however, only a redshift of the emission peak with increas-

ing temperature was observed. This particularity is likely related to the

0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5

PL intensity (arb. u.)

Wavelength (µm)

T = 19 °C

λλ

λexc = 514.5 nm

5 kW/cm2

5 W/cm2

QD*

GaAs nQD= 9.6x1010cm-2

200 nm

Fig. 23: PL spectra of In0.67Ga0.33As QDs, deposited at 485°C. “QD”, “QD*”, and

“WL” denote transitions from the QD ground state, the first excited QD state, and

the WL, respectively. Inset: Plan-view dark-field transmission electron micro-

graph of the same QD layer. This figure has been published in Ref. 161.

5. MOCVD of quantum-dot structures for laser diodes

strong temperature dependence of the AsH3 pyrolysis between 450°C and

650°C (cf. Fig. 7 on p. 22). Increasing the temperature leads to a larger frac-

tion of reactive As species. The V/III ratio hence rises and causes an in-

crease of the QD size via lowering the surface energy, as described above.

For the QDs grown using TBAs, this effect is not observed since the crack-

ing efficiency of TBAs is nearly 100 % at 500°C (cf. Fig. 7 on p. 22). The

use of TBAs thus allows a nearly independent tuning of the growth parame-

ters temperature and V/III ratio. As compared to AsH3, the growth of QDs

using TBAs is less dependent on the chemical properties of the arsenic pre-

cursor.

5.3. Redshift of the quantum-dot emission wavelength

As shown in the previous section, the emission wavelength of TBAs-

grown InGaAs QDs can easily be tuned to above 1200 nm just by increas-

ing the V/III ratio during deposition of the QD material (cf. Fig. 24). In-

GaAs QDs grown using conventional AsH3 at high V/III ratios and elevated

temperatures emitting even beyond 1.3 µm have been demonstratedviii.

Wavelengths close to 1.3 µm can also be achieved using MBE of large InAs

QDs49. Unfortunately, all these long-wavelength QDs have low sheet densi-

ties and are therefore unsuited for the application in laser diodes. Achieving

viii Frank Heinrichsdorff, unpublished.

0 5 10 15

470 480 490 500 510 520

1.15

1.20

1.25

V/III ratio

T = 300 K

Pexc = 5 W / cm2

λλ

λexc = 514.5 nm

Peak wavelength (µm)

QD growth temperature (°C)

Fig. 24: PL peak emission wavelengths of the ground-state transitions of

In0.67Ga0.33As QDs as a function of growth temperature and V/III ratio. QDs of the

V/III series were deposited at 485°C, the V/III ratio of the temperature series was

kept at 1.5. The reproducibility of the peak positions is within 10 nm. This figure

has been published in Ref. 161.

5.3. Redshift of the quantum-dot emission wavelength

sufficient QD sheet densities at 1.3 µm poses a general problem. This is in

agreement with the general observation that QD sheet densities decrease

with increasing QD size. To obtain QD ground-state lasing, however, large

QD densities are required. Different techniques were explored within this

work to redshift the emission wavelength of high-density QDs without for-

feit of their number.

Approaches to obtain QD luminescence at 1.3 µm and beyond generally

make use of high-density short-wavelength (1.1 – 1.2 µm) In(Ga)As

stressor QDs. The density of these stressors is typically in the upper

1010 cm-2 regime. To achieve 1.3 µm luminescence, such stressors are sub-

sequently capped with thin films of a material having a lower band gap than

GaAs. These materials can be lattice-matched to GaAs like InxGa1-xAsyN1-y

with corresponding fractions x and y, or exhibit smaller (GaAsxN1-x) or lar-

ger lattice constants (InxGa1-xAs) than GaAs. QD emission at 1.3 µm has

been demonstrated, for example, by overgrowing InAs QDs with 40 nm

In0.03Ga0.97As0.99N0.01 using MBE162. Due to the large band gap bowing of

dilute nitrides163, the band gap of the In0.03Ga0.97As0.99N0.01 is about 1.2 eV

(1033 nm) at room temperature. Therefore, the confinement of the electron

wavefunction is lower for the quaternary barrier than for GaAs, leading to

lower exciton recombination energies. Overgrowth of InGaAs QDs with

GaAsN is expected to have a similar effect. In addition, the tensile strain of

GaAsN164, 165 in a GaAs matrix would lead to a reduction of the overall

strain of the QD structure. An overview of the different techniques is given

in the flow chart of Fig. 25. The admixture of Sb could possibly further in-

crease the emission wavelength. The DWELL concept described in sect. 2.3

has turned out to lead to a comparably large decrease of the luminescence

efficiency in MOCVD and was therefore not pursued within this work.

Three approaches to shift the QD emission to longer wavelengths will be

discussed in this section. InGaAs QDs overgrown with Ga-rich QWs are

described in sect. 5.3.1. This technique has turned out to be very successful

for the fabrication of 1.3 µm QD lasers in MBE. The admixture of nitrogen

(sect. 5.3.3) and antimony (sect. 5.3.4) to InGaAs QDs has been explored as

other means to redshift the emission wavelength.

5.3.1 Overgrowth of InGaAs QDs by InGaAs QWs using arsine

In this section, the overgrowth of short-wavelength InGaAs stressors

with Ga-rich InGaAs layers using conventional precursors is discussed as

an approach to redshift the QD emission wavelength to 1.3 µm. This tech-

nique has previously been explored in MBE and is referred to as Activated

Alloy Phase Separation (AAPS)49. The increase of the strain energy related

to this technique enhances the danger of plastic relaxation by formation of

5. MOCVD of quantum-dot structures for laser diodes

dislocations and precipitates so that a careful choice of the stressor QD size

and the indium fraction of the overgrown QW is required49, 166. The advan-

tage of TBAs with regard to AsH3 for the deposition of such compound

structures is outlined in sect. 5.3.2.

The In0.8Ga0.2As stressor QDs used here were grown with conventional-

precursor MOCVD. They have a lateral density of about 5 × 1010 cm-2 and

have proven their crystalline quality as active medium of laser diodes156.

These devices exhibit differential efficiencies of almost 100 %, and the

threshold currents are exceptionally low (more details are given in sect.

6.1.1). The InGaAs stressor dots used for the overgrowth experiments were

deposited under the same growth conditions as the QDs of the laser diodes.

They were eventually overgrown with gallium-rich InGaAs QWs of varying

thickness and composition and finally capped with GaAs.

Depending on thickness and composition of the QW, the overgrowth

leads to a distinct redshift of the QD emission wavelength. This is ascribed

to changes of the electronic properties of the InGaAs stressor QDs. Two

main effects come into consideration to explain these changes:

1. The overgrown InxGa1-xAs cap layer causes a strain redistribution

that leads to a partial strain relaxation of the stressor dots167, 168. The

band gap of the InGaAs stressor dots decreases. This is discussed in

sect. 5.3.1.1.

Fig. 25: Flow chart of different approaches to achieve lasing emission at 1.3 µm

and beyond by overgrowing small In(Ga)As QDs. Structures in bold scripture

were grown and characterized within this work.

5.3. Redshift of the quantum-dot emission wavelength

2. Strain-driven surface kinetics leads to the decomposition of the

InxGa1-xAs QW during growth55, 169, 170 due to preferential agglom-

eration of indium at already existing stressors, thus increasing the ef-

fective QD size. The structural changes of the QDs upon overgrowth

are described in sect. 5.3.1.2.

The overgrown QD structures were investigated using PL, TEM, and cross-

section scanning tunneling microscopy (XSTM).

5.3.1.1 Spectroscopic characterization

Fig. 26 shows PL spectra of overgrown In0.8Ga0.2As QDs, recorded at

9 K. The stressors were overgrown with 5 nm thick InxGa1-xAs QWs with

indium fractions x ranging from 0 to 25 % (the In fraction was calibrated

using x-ray diffraction of ten-fold InGaAs QWs). The spectra were normal-

ized to equal peak intensities in order to facilitate the comparison of the re-

spective line shapes.

The emission wavelength of the QD structures shifts monotonously into

the red with increasing xIn. QDs consisting of lattice-mismatched InGaAs

are considerably strained. The hydrostatic strain of the QDs strongly raises

the QD-material band gap as compared to unstrained InGaAs. If an addi-

tional InGaAs QW is inserted between QDs and GaAs matrix on top of the

800 900 1000 1100 1200 1300

PL intensity (arb. u.)

Wavelength (nm)

010 15 20 25

xIn(%) =

T = 9 K

Pexc = 5 W / cm2

Fig. 26: Low-temperature PL spectra of single In0.8Ga0.2As QD layers overgrown

with 5 nm InxGa1-xAs QWs of varying indium fraction x. The spectra were nor-

malized to equal peak intensities.

5. MOCVD of quantum-dot structures for laser diodes

dots, the strain of QDs decreases in growth direction. The strain redistribu-

tion of a similar QD structure was calculated in the framework of contin-

uum mechanics, using a finite-differences method167. It is shown in Ref.

167 that the hydrostatic strain of the QW increases with increasing indium

concentration, whereas the strain of the QDs decreases. This leads to a low-

ering of the confining potential, following the strain dependence of the In-

GaAs band gap, and leads to a reduction of the exciton recombination en-

ergy.

Up to x = 15 %, the PL lines of Fig. 26 exhibit rather symmetric shapes

as expected for mono-dispersed ensembles of self-organized QDs at low

temperatures171. The sample with x = 25 % shows a shoulder on the long-

wavelength side, suggesting that a number of larger QDs have formed that

might lead to a bimodal size distribution if x is raised further.

The peak wavelengths of the spectra in Fig. 26 are plotted in Fig. 27 as a

function of the indium fraction x of the overgrown InxGa1-xAs QW. Calcu-

lated values167 of transition wavelengths previously obtained for compara-

ble structures are likewise plotted. In these calculations, truncated InAs

pyramids capped with a 4 nm thick InxGa1-xAs layer of varying composition

were used. These QDs were assumed to be on top of a 0.36 nm thick

InAs WL and to have {101}-type side facets and a base length / height of

0102030

1000

1050

1100

1150

1200

1250

PL peak wavelength at T=9K, Pexc = 5 W/cm2

8 band k·p calculation

Peak wavelength (nm)

Indium fraction (%)

Fig. 27: Full squares: PL peak wavelength of In0.8Ga0.2As QDs overgrown with

5 nm InxGa1-xAs of denoted indium fraction x, recorded at 9 K. From TEM im-

ages, the QDs are estimated to be 3-5 nm high and 20-25 nm wide. Hollow

squares: Ground-state transition wavelengths of a 2.6 nm high truncated-pyramid-

shaped InAs QD overgrown with an InxGa1-xAs QW of denoted indium fraction x,

calculated within an eight-band k·p framework.

5.3. Redshift of the quantum-dot emission wavelength

11.3 nm / 2.6 nm. The localized electron and hole states in these QDs were

then calculated within an eight-band k·p framework. Exciton formation is

treated self-consistently in the Hartree approximation. Calculations for

truncated-pyramid QDs are described in Ref. 167, calculations for pyrami-

dal InAs/GaAs QDs can be found in Ref. 172.

Fig. 27 indicates that the measured emission wavelengths are larger than

the calculated ones. A deviation of the experimental data from the calcu-

lated values is not surprising since the respective QD structures are slightly

different. Whereas ternary In0.8Ga0.2As QDs were overgrown with 5 nm

InxGa1-xAs, the calculations are based on binary InAs QDs capped with only

4 nm InxGa1-xAs. Binary InAs is nevertheless a good approximation for

QDs formed upon deposition of In0.8Ga0.2As since it has been shown that

the In fraction of SK QDs formed upon deposition of InxGa1-xAs QDs is

usually much larger than the nominal fraction x of the 2D layer173. This ef-

fect is attributed to lateral decomposition of the ternary InGaAs driven by

strain effects. From TEM images not shown here, the shape of the

In0.8Ga0.2As stressor QDs cannot be determined in detail: A possible lens

shape and a truncated-pyramid shape can hardly be distinguished. However,

lens-shaped QDs and truncated-pyramid QDs of comparable spatial dimen-

sions have similar electron and hole levels.

In addition to the vertical offset of the experimental data from the theo-

retical values in Fig. 27, a superproportional increase of the emission wave-

length with increasing indium fraction is observed. This observation gives

reason to assume that an additional phase separation of the InGaAs QW has

taken place, additionally pushing the emission wavelength of the QD struc-

ture further into the red. A structural investigation of such AAPS structures

is given in the next section.

5.3.1.2 Structural characterization

Structural investigations of AAPS structures were carried out using high-

resolution cross-section TEM and XSTM. TEM images were processed

from a series of samples that contain single In0.8Ga0.2As stressor sheets

overgrown with d = 0, 1, 2 and 3 nm of In0.25Ga0.75As. The samples were

grown using conventional precursors. Fig. 28 shows low-excitation room-

temperature PL spectra of these QD structures. The QD emission wave-

length shifts from below 1200 nm (d = 0 nm) to beyond 1300 nm (d =

3 nm). This qualitatively corresponds to the overgrowth experiments using

InGaAs QWs of constant thickness but varying composition. The FWHM

of the QD ground-state transition line increases from 60 meV to more than

80 meV, indicating a broadening of the QD size distribution. The integrated

PL intensity decreases by a factor of two. This is evidence of an increase of

5. MOCVD of quantum-dot structures for laser diodes

the defect density upon overgrowth. Although the target wavelength has

been reached, defects must still be reduced in these structures to obtain QD

structures with sufficient quality for the application in devices.

Fig. 29 shows Fourier-filtered high-resolution TEM images of the struc-

tures in cross-sectional geometry. QDs with a representative size were cho-

sen for the TEM analysis. Plan-view TEM images of these structures do not

show any plastically relaxed clusters. The dotted white lines depict the

boundary between InGaAs and GaAs. These boundaries were determined

according to the reversal of the In-As and Ga-As image contrast, occurring

at an indium fraction of about 15 %174, 175. The TEM images indicate that

the QD height increases with the thickness of the overgrown InGaAs QW.

The plateaus of the truncated pyramids become wider. The layer on both

sides of the QDs, now consisting of the original WL and the overgrown In-

GaAs QW, also gains in thickness. Interestingly, the width of the QDs

hardly varies.

The different states of overgrowth are depicted schematically in Fig. 30.

An as-grown, freestanding InGaAs QD is depicted in (a). During over-

growth (b), In atoms are assumed to attach preferably at the InGaAs islands,

whereas Ga atoms prefer the pseudomorphically strained WL regions be-

tween the QDs176. A similar observation has been made in MBE during

overgrowth of InAs QDs by binary GaAs: gallium atoms were shown to

migrate away from the QDs towards the pseudomorphically strained WL

regions between the QDs that have an in-plane local lattice constant equal

0.80.91.01.11.21.31.41.51.6

Wavelength (µm)

PL intensity (arb. u.)

Pexc= 5 W / cm2

λλ

λexc = 514.5 nm

d = 0 nm

d = 1 nm

d = 2 nm

d = 3 nm

Fig. 28: Room-temperature PL spectra of In0.8Ga0.2As QDs overgrown with

In0.25Ga0.75As QWs of varying thickness d.

5.3. Redshift of the quantum-dot emission wavelength

to that of unstrained GaAs177. Fig. 30 (c) shows the complete GaAs-capped

structure.

As suggested by the TEM images of Fig. 29, the In-enriched material

grows both on top and on the facets of the QDs. It is likely that the In-Ga

phase separation along the QD facets is stronger at the upper part of the

QDs, close to the plateau. It is known from strain calculations178 that the

strain of uncapped pyramidal InAs QDs is minimum at the pyramid tip.

This means that at the QD bases, the local lattice constant is equal to that of

GaAs and increases towards the apex of the island, approaching the lattice

constant of unstrained InAs. The strain situation of truncated-pyramid is-

lands is qualitatively the same; compositionally homogeneous truncated

InGaAs pyramids should have the largest lattice constant on their plateaus.

Therefore, the upper part of such an island is most attractive for In attach-

ment. The increase of the alloy separation efficiency is reflected by the cu-

neiform shape of the dark-grey shaded region at the QD facets in Fig. 30 (b)

123

Height (nm)

InGaAs QW thickness (nm)

(d)

Fig. 29: (a)-(c) Fourier-filtered cross-sectional high-resolution TEM images of

In0.8Ga0.2As QDs covered with (a) 1 nm, (b) 2 nm, and (c) 3 nm In0.25Ga0.75As.

The dotted white lines depict the boundary between InGaAs and GaAs. The

boundaries were determined after the reversal of the In-As and Ga-As image con-

trast, occurring at an indium fraction of about 15 %. (d) Heights of the QDs and

the QW near the QD bases. A generous error of 1 nm is assumed for each value,

taking uncertainties of the determination of the InGaAs/GaAs boundary into ac-

count.

5. MOCVD of quantum-dot structures for laser diodes

and (c), depicting In-enriched InGaAs stemming from the overgrown QW.

The slowly increasing base widths of the overgrown QD structures (cf.

Fig. 29) can be explained by a reduced phase-separation efficiency at the

QD bases.

In addition to the stressor-induced separation of the InGaAs QW into In-

rich and In-depleted regions, a net material transport from the QW to the

QDs can be observed, so that for all thicknesses of the overgrown QW, the

QD plateaus significantly loom over the QW/WL compound. The strength

of net material transport is likely to depend on the In fraction of the over-

grown QW. For an indium fraction of only 15 %, the effect has been shown

to be negligible: Fig. 31 shows a cross-sectional dark-field TEM image of

InAs QDs formed upon MBE of 2 ML InAs at 485°C. These QDs were

subsequently overgrown with 5 nm In0.15Ga0.85As and eventually capped

with GaAs. The three dark regions in the QW are strain contrasts due to the

QDs. The image shows a significant increase of the effective QD height

from usually 4-5 nm for GaAs-capped dots to about 10 nm. However, the

QDs do not loom over the compound of WL and overgrown InGaAs QW.

The room-temperature emission wavelength of this QD structure is 1.3 µm.

The QD increase of the effective height of the MBE-grown InAs QDs

capped with In0.15Ga0.85As is analogous to the MOCVD-grown sample

where the indium fraction of the QW was 25 %. However, the interface be-

tween QD structure and subsequent GaAs cap is not as smooth as for the

MBE QDs (Fig. 31). A net material transport in the MBE-grown sample

Fig. 30: Schematic diagrams illustrating the separation of a ternary InGaAs alloy

into indium-rich and gallium-rich phases, activated by the strain-relaxed surface

of an underlying In(Ga)As stressor QD. (a) Initial In(Ga)As stressor QD, assumed

to have the shape of a truncated pyramid. (b) Partial decomposition of the InGaAs

alloy into indium-rich and gallium-rich regions. (c) Final structure, buried by

GaAs.

5.3. Redshift of the quantum-dot emission wavelength

can therefore be excluded. In the limit of binary InAs used for overgrowth

of InAs QDs, the QW material entirely attaches at the stressors, as has been

demonstrated in MBE140.

Quantitative statements about the indium distribution inside QDs and

WL cannot be made from the TEM images of Fig. 29. The enhanced image

contrast inside the QD only suggests that the indium fraction is larger in the

center of the QD than in the QW region, as one expects due to the large dif-

ference between the nominal In fraction of 80 % for the QDs and 25 % for

the InGaAs QW. This is also consistent with Fig. 31.

Cross-sectional scanning tunneling microscopy (XSTM) was used to ob-

tain quantitative information on the In distribution of In0.8Ga0.2As QDs

overgrown with 3 nm In0.1Ga0.9As179. XSTM is a method to obtain direct

structural and chemical information of buried QDs180-183. Fig. 32 shows an

empty-state XSTM image of a QD representative for the dots of the ensem-

ble, taken at a bias voltage of VS = +2.1 V. For zinc-blende type III-V semi-

conductors, it has been shown that at positive sample bias voltages VS, the

empty states of the group-III atoms are imaged184, 185. Fig. 32 shows such an

XSTM image. The height of the QD amounts to 5 nm and the width to

about 16 nm. The image contrast is partly an In-Ga selective chemical con-

trast. Strain relaxation out of the cleavage plane, however, also contributes

to the image contrast. In order to separate the chemical and topographic

contrast and to determine the local In concentration quantitatively, the lat-

eral distances between neighboring zigzag chains is evaluated186. A maxi-

mum In concentration of 60 % inside the QD is concluded. The data from

strained InGaAs QWs were used to gauge the In concentration. Since QDs

are stronger strained than QWs, however, the maximum In concentration is

assumed to be higher. This is consistent with the nominal In concentration

of 80 %.

Evaluation of the atom chains in the center of the dot (Fig. 32 (a)) shows

that the center of weight of the In distribution is located at the upper part of

the QD, close to the truncated-pyramid plateau. According to the schematic

of Fig. 30 (c), however, a maximum of the In distribution is rather expected

Fig. 31: Cross-sectional dark-field TEM image of QDs formed upon MBE of

2 ML InAs at 485°C, overgrown with 5 nm In0.15Ga0.85As. From Ref. 169.

5. MOCVD of quantum-dot structures for laser diodes

to be located in the lower region of the QD. It is conceivable that during

overgrowth of the In0.8Ga0.2As QDs, kinetic effects lead to a redistribution

of In and to an In segregation to the upper part of the QDs173, 187. An analy-

sis179 of the indium distribution of the 4 nm thick InGaAs compound layer

(Fig. 32 (b)) reveals that the weight of the In distribution is located at the

bottom of the structure, close to the lower GaAs/InGaAs interface. The In

fraction starts at 40 % and decreases linearly to about 15 %. From In segre-

gation effects as observed for InGaAs QWs, the opposite behavior is ex-

pected188, 189. This inverse behavior is ascribed to the decomposition of the

ternary InGaAs alloy of the overgrown QW where In atoms are attracted by

the QD stressors, leaving a Ga-rich phase.

The reverse truncated cone shape of the In-rich QD center is the most

remarkable feature of Fig. 32. XSTM studies of InAs QDs and InGaAs QDs

capped with GaAs have also revealed inhomogeneous In composition pro-

files186, 190-192. An inverse-pyramid shaped indium distribution inside MBE-

grown In0.5Ga0.5As QDs is reported in Ref. 193. It is surprising that a simi-

lar result is obtained for QDs overgrown with a rather thick InGaAs QW of

only 10 % indium fraction. It is conceivable that an initially truncated-

pyramid-shaped InGaAs QD was transformed to the inverted truncated-

pyramid shape as in Fig. 32 by indium depletion near the baselines of the

initial dot. In that case, the depleted region would correspond to the triangu-

lar areas (c) and (d) in Fig. 32. Refs. 194 and 195 report models of the elec-

tronic structure of such inverted pyramid QDs. An inverted electron-hole

alignment is found with respect to a pyramidal QD.

Fig. 32: Empty-state XSTM image taken at VS = +2.1 V. The contours of the dot

are indicated by dotted lines, and those of its In-rich zone by dashed lines. The In

distribution was evaluated along the intersection lines (a) and (b) by evaluation of

atom-chain distances (cf. text). After Lenz et al.179.

5.3. Redshift of the quantum-dot emission wavelength

To conclude, overgrowing In(Ga)As QDs with Ga-rich InGaAs QWs en-

ables to redshift the QD emission wavelength to more than 1.3 µm at room

temperature. Calculations show that renormalization of the band gap inside

the stressor QDs by strain redistribution via the overgrown InGaAs can ac-

count for this redshift. Structural analyses using cross-sectional TEM and

STM images reveal that alloy phase separation of the overgrown QW as

well as segregation of the indium within the QDs takes place, leading to an

inverse truncated pyramid structure of the indium-rich QD core. From com-

parison with other growth experiments, the strength of this material redis-

tribution effects is assumed to increase with the indium composition of the

overgrown QW. The decrease of the luminescence efficiency upon over-

growth with InGaAs requires defect reduction (cf. sect. 5.1) to obtain de-

vice-quality 1.3 µm QD structures. Lasing from InGaAs QDs overgrown

with a Ga-rich InGaAs QW at 1.24 µm is reported in sect. 6.1.3.

5.3.2 Advantage of TBAs for redshifting the QD emission

wavelength

It was shown in sect. 5.2 that due to the different decomposition mecha-

nisms of TBAs with respect to AsH3, the QD formation using TBAs as ar-

senic precursor leads to different dependences on the growth parameters.

Optical in-situ investigations on the formation of InAs QDs have shown141

that the danger of dislocation formation is enhanced for AsH3, probably due

to a larger number of hydrogen radicals.

The properties of QD ensembles as a function of the V/III ratio were in-

vestigated using PL. Fig. 33 shows room-temperature PL measurement for

two sample series. The samples in (a) were grown using AsH3

142; the sam-

ples in (b) were deposited using TBAs as arsenic precursor. From the de-

pendence of the PL spectra on increasing V/III ratio, it can be concluded

that increasing AsH3 partial pressure leads to a bimodal size distribution.

Whereas the crystal quality for V/III = 40 is worst due to a lack of As spe-

cies on the surface, and probably no QDs were formed, the spectrum for

V/III = 80 is dominated by small QDs emitting around 1200 nm. With fur-

ther increased V/III ratio, QDs with a larger size are created, so-called high

QDs142, emitting around 1300 nm. The formation of such QDs is accompa-

nied by large dislocated clusters as one can see on plan-view TEM images

not shown here. The QDs grown using TBAs show a mono-dispersed size

distribution for all V/III ratios. One is tempted to argue that all TBAs/III ra-

tios are lower than the AsH3/III ratios and that, if the same V/III ratios were

used in both cases, the same results would be obtained. At typical QD depo-

sition temperatures of 500°C, however, the cracking efficiencies of the two

precursors strongly (cf. Fig. 7 and Ref. 77). Moreover, the PL excitation

densities are different. Whereas a rather high excitation density was used to

5. MOCVD of quantum-dot structures for laser diodes

excite the AsH3-grown samples, the TBAs-grown QD structures were ex-

cited at a much lower level. Spectra obtained from the TBAs-grown QDs

with higher excitation density, however, do not show the formation of a

second QD size.

Using TBAs, the V/III ratio can thus be tuned within one order of magni-

tude without changing the unimodal character of the size distribution. The

formation of a second size mode using AsH3 has also been observed when

the temperature is raised142. A temperature increase apparently has the same

effect as the increase of the AsH3/III ratio. Using TBAs, the QD size in-

crease is attributed to a decrease of the surface energy. This results in an in-

crease of the QD size due to thermodynamic arguments119. No dislocated

clusters were observed on TEM images of QD sheets grown with

TBAs/III = 16.5. These findings are promising for low defect densities of

1.3 µm QD structures grown using alternative-precursor MOCVD.

5.3.3 Wavelength shifting using nitrogen

Dilute nitrides such as Ga(In)AsN attract growing interest since they ex-

hibit much smaller band gaps than N-free Ga(In)As alloys, even for N frac-

tions of only a few percents. Self-organized GaInAsN QDs could therefore

have longer emission wavelengths than InGaAs QDs. Ga(In)AsN/GaAs

QWs emitting around 1.3 µm are currently attractive for telecom lasers

since the incorporation of N strongly reduces the lattice constant with re-

spect to N-free Ga(In)As164, 165 so that low-band-gap InGaAsN layers lat-

tice-matched to GaAs can be grown196. High-quality InGaAsN/GaAs QWs

are used as active layers in low-threshold 1.3 µm lasers in both edge-

emitting197, 198 and vertically emitting geometry199. Fig. 34 shows a flow

chart of different approaches to achieve QD emission at 1.3 µm and beyond

Fig. 33: Room-temperature PL spectra of InGaAs QDs grown with (a) AsH3

(from Ref. 142) and (b) TBAs, deposited around 500°C. The respective V/III ra-

tios are marked in the viewgraphs.

5.3. Redshift of the quantum-dot emission wavelength

using nitrogen (and also antimony) insertions in In(Ga)As QDs that initially

emit around 1.1 µm. Structures as shown in Fig. 34 were grown and charac-

terized; the experiments carried out using antimony are described in sect.

5.3.4.

If two binary III-V materials AB and AC are mixed, the band gap of the

alloy AB1-xCx follows the formula200, 201

)x1(xbx)x1()x( AC

gap

gapgap −⋅⋅−⋅+⋅−= EEE (5.1)

where b is the so-called bowing parameterix. In case b is small, the compo-

sition dependence of the band gap can be approximated by linear interpola-

tion between AB

gap

E and AC

gap

E. For larger bowing parameters, the composition

dependence of the band gap follows a parabola. The band gaps of

GaAs1-yNy and In0.47Ga0.53As1-yNy as functions of the lattice parameter are

depicted in Fig. 35. A recent publication reports a bowing parameter of

40 eV for GaAsxN1-x. The bowing parameters of nitrogen-free III-V com-

pound semiconductors is much smaller (InxGa1-xAs202: bΓ = 0.32-0.6 eV;

AlxGa1-xAs202: bΓ = 0.14 to 0.66 eV).

The large bowing parameter of dilute nitrides is ascribed to microscopic

interactions of the nitrogen atoms with the matrix. It has been shown that

doping of conventional group III–V semiconductors with low, impurity-like

concentrations of nitrogen introduces highly localized acceptor levels203.

Although a reduction of the band gap by more than 100 meV per atomic

percentage was observed in InGaNAs alloys, the electronic band structure

of the host crystal is actually not affected by such low nitrogen concentra-

tions, owing to the highly localized nature of the nitrogen perturbations.

ix The formula actually applies only to ternary alloys. It is applicable to InGaAsN

if A = InyGa1-y, B = As, and C = N are chosen.

Fig. 34: Flow chart of different approaches to achieve QD emission at 1.3 µm by

addition of antimony and nitrogen to In(Ga)As QDs or InGaAs QWs.

5. MOCVD of quantum-dot structures for laser diodes

The band gap decrease is rather attributed to a strong interaction between

the conduction band and a narrow resonant band formed by the nitrogen

states. This results in a splitting of the conduction band and thus to an effec-

tive reduction of the fundamental band gap204.

The principal difficulty about growing N-containing alloys is that the

small covalent radius of N implies extremely low solubilities, leading to

large mixing enthalpies

∆

H. The free energy F of an alloy with composition

x and temperature T is given by

F(x,T) =

∆

H(x)-T

⋅

S(x) (5.2)

where S is the configuration entropy. F(x) is plotted for a hypothetical

semiconductor alloy at fixed temperature in Fig. 36 (a). It exhibits a compo-

sition range with a large upward bowing. In this region, the second deriva-

tive of F is below zero. If a hypothetical alloy of composition xa is assumed,

the system can reduce its free energy by

∆

1 if this alloy decomposes into

two phases b and c. xb and xc are interdependent since due to mass conser-

vation, the average composition must be constant. The phase separation will

continue until it stops to be energetically favorable. Points C and D are the

deflection points of F ( 0

22 =∂∂ xF ) and are called spinodal points. Alloys

with x between the spinodal points are unstable; this composition range is

called miscibility gap. Outside the miscibility gap, 22 xF ∂∂ is positive. A

spontaneous alloy phase separation will not occur outside the miscibility

gap since F would then increase: If alloy d decomposes into phases e and f,

for example, F would increase by

∆

2. Alloys outside the miscibility gap are

Fig. 35: Band gap energy versus lattice parameter of GaAsN, InGaAs, and

In0.53Ga0.47As1-yNy. From Ref. 205.

5.3. Redshift of the quantum-dot emission wavelength

stable even if they do not coincide with the binodal points A and B, which

are the two local minima of F. The growth of quaternary InGaAsN can be

understood as a combined alloying of In and Ga on the group-III sublattice,

and of As and N on the group-V sublattice, as depicted schematically in

Fig. 36 (b).

With increasing temperature, the miscibility gap shrinks due to the en-

tropy contribution to the free energy (Eq. 5.2), and vanishes above a critical

temperature Tc. All III-V compound semiconductors have positive mixing

enthalpies, so any III-V semiconductor is thermodynamically unstable at

T = 0. If

∆

H(x) is sufficiently small for a given pair of semiconductor mate-

rials to be mixed, however, the critical temperature Tc is below the growth

temperature range, and any composition x is stable. Miscibility can – in

GaAs InAs

GaN InN

Ga In

1-x x

Ga In

1-x x 1-y y

As N InAs N

1-y y

GaAs N

1-y y

Ga In

1-x x

(0,0) (1,0)

(0,1) (1,1)

(x,y)

∆

composition x

ener

00.5 1

(a)

(b)

Fig. 36: (a) Free energy F versus solid composition for a hypothetical semicon-

ductor alloy with a large positive mixing enthalpy H. The points labelled A and B

are the binodal points, the inflection points C and D are the spinodal points. The

miscibility gap is the composition range between xC and xD. (b) The fabrication of

InxGa1-xAsyN1-y can be achieved by separately mixing Ga and In on the group-III

sublattice with parameter x, and As and N on the group-V sublattice with parame-

ter y.

5. MOCVD of quantum-dot structures for laser diodes

principle - be overcome by raising the growth temperature to beyond Tc.

However, this would require extremely high growth temperatures. Instead,

particularly low growth temperatures are commonly chosen to grow dilute

nitrides, allowing to kinetically freeze alloys with thermodynamically un-

stable compositions.

The high indium fraction of the In0.8Ga0.2As QDs makes the nitrogen in-

corporation very difficult. For ternary GaAsN, a maximum N fraction of

6.7 % was reached206. The maximum N fractions reported for InGaAsN

QWs are much lower: InGaAsN QW emitting at 1.3 µm usually have in-

dium contents around 40 % and N fractions of typically 1 %142. However,

the minimum indium fraction needed for the growth of InGaAs QDs in the

SK mode is 40 %142. Upon N incorporation, this fraction must even be

higher since N incorporation leads to a decrease of the lattice constant,

counteracting the driving force for SK growth of QDs. Due to the high in-

dium fraction, arsenic is more easily incorporated than N. This means that

the N/As ratio must be very high in the gas phase. To avoid As vacancies,

however, the partial pressure of reactive As species must not fall below a

certain value. To achieve large N/As ratios at the same time, huge flows of

DMHy have to be provided. To reduce the effect of the miscibility gap, low

temperatures and high growth rates are used for the growth experiments as

described in the following. Low temperatures reduce the adatom mobility

and large growth rates shorten the available time for decomposition and dif-

fusion processes.

5.3.3.1 Simultaneous deposition of As, Ga, In and N

The first experiments to fabricate InGaAsN QDs aimed at the simultane-

ous deposition of the four constituting elements In, Ga, As, and N. For these

experiments, an InGaAs QD sample was used as reference and starting

point. In order to obtain InGaAsN QDs, the same growth parameters as for

the reference sample were used. All samples of this series were grown at

500°C with a nominal growth rate of 1 µm/h. The TBAs/III ratio was con-

stantly 1.5. The reference sample exhibits a QD density of 2.2 × 1010 cm-2

and a QD ground-state PL at 1208 nm at room temperature. The

TMIn/TMGa ratio was 2.5, corresponding to an In fraction of 70-80 % in

the InGaAs reference sample. During the deposition of the InGaAs QD ma-

terial, however, different flows of DMHy were added to the gas phase. The

pDMHy/pTBAs ratio was varied from 0 to 360, larger DMHy flows were not

supported by the DMHy bubbler for technical reasons. After the deposition

of InGaAsN, a GRI of 60 s was performed during which all precursors were

switched off.

5.3. Redshift of the quantum-dot emission wavelength

Fig. 37 (a) shows low-excitation PL spectra of the sample series, re-

corded at room temperature. Fig. 37 (b) shows the peak intensities of the

QD ground-state transitions. PL spectra obtained under a very high excita-

tion power density of 5 kW/cm2 (not shown here) were used for this evalua-

tion. The high excitation power leads to a complete saturation of the QD

ground-state PL. The peak intensities then scale with the QD density and

can thus be used for relative estimation of the latter.

The topmost curve shows the PL of the reference sample. The position of

the peak is at 1208 nm, the FWHM is 63 meV. If DMHy is now switched

on and successively increased during the deposition of the InGaAs, a blue-

shift of the PL peak by 86 nm can be observed up to a pDMHy/pTBAs ratio of

50. This blueshift is accompanied by a significant increase of the low-

excitation PL (a), and also by a strong increase of the QD density, possibly

by a factor of two, as suggested by Fig. 37 (b). The presence of DMHy in

this range apparently leads to the formation of a much larger density of QDs

that are smaller than those of the reference sample. With further increasing

pDMHy/pTBAs, a redshift of the PL peak sets in, the PL intensity is reduced

and the QD density drops. A plan-view TEM image of the sample grown

with DMHy/TBAs=360 shows a QD density of 4.9 × 109 cm-2 which is by a

factor of 4.5 lower than that of the reference sample.

The following explanations can be given for these effects:

• As described in sect. 4.5, the QD density is strongly determined by

kinetic effects. It is conceivable that the presence of DMHy reduces

Fig. 37: (a) Room-temperature low-excitation spectra of undoped samples with

single InGaAsN QD sheets. (b) PL peak intensities of the QD ground-state (GS)

PL, plotted as a function of pDMHy/pTBAs. A power density of 5 kW/cm2 was used

for excitation. The inset shows the spectra belonging to the first data point at

pDMHy/pTBAs = 0.

5. MOCVD of quantum-dot structures for laser diodes

the surface mobility so that a larger number of nuclei are formed.

The smaller QD size can then be explained with mass conservation.

• Using thermodynamic arguments, the reduced QD size could also be

explained by an increase of the surface energy119, leading to a

smaller equilibrium size. In this case, N would act as anti-surfactant.

A reduction of the QD size and an increase of the QD density upon N

supply has also been observed during chemical-beam epitaxy of InGaAsN

QDs 207, 208: The QDs had a nominal N content of 1 %, roughly estimated

using the N2 flow rate and the growth rate, and exhibited a height / base

length of 4 (30) nm as compared to a height (base) length of 6 (45) nm of

the reference sample. An increase of the QD density by a factor of 3 was

measured. Upon a further increase of the N2 flow, the QD size increased

again and the QD density dropped.

The only publication of InGaAsN QDs grown using MBE reports a

lower QD density if nitrogen is switched on. The QDs are reported to keep

their height if N is supplied209. The emission wavelength, however, in-

creased immediately and could be extended to 1.52 µm. The PL intensity

decreased strongly during the redshift. The In fraction of the QDs was

70 %, and an N fraction of 4 % is reported. This high value was not meas-

ured, but estimated from InGaAsN QWs deposited under similar growth

conditions.

Alternatively to the use of InGaAs QDs as starting point to grow In-

GaAsN as described in this section, compressively strained long-

wavelength emitting InGaAsN QWs could be grown and used for QD self-

organization. Upon deposition of a supercritical thickness of this strained

material, self-assembly of SK dots may set in. As mentioned earlier, the ap-

proach is problematic with regard to the strain situation of InGaAsN: In-

GaAsN QWs emitting at 1.3 µm usually have a maximum indium fraction

of around 40 %. For larger In fractions, the N incorporation efficiency is

most likely not large enough to achieve this wavelength. It is known from

the growth of SK QDs, however, that at least 40 % of indium are required

to obtain SK QDs (cf. p. 63 of Ref. 142).

5.3.3.2 Nitridation

As an alternative to the simultaneous deposition of all four elements, ter-

nary InGaAs QDs were grown with a subsequently performed N/As ex-

change reaction. Ternary InGaAs QDs are fabricated by deposition of a su-

percritically thick InGaAs QW and a subsequent GRI of 1 min without

TBAs stabilization. Different DMHy flows were now applied during the

GRI. Fig. 38 shows PL spectra of such InGaAsN-QD test structures for

5.3. Redshift of the quantum-dot emission wavelength

which different DMHy partial pressures were used. Fig. 38 (a) shows that a

partial pressure of 8 Pa DMHy during the GRI leads to a wavelength shift

from 1150 nm (InGaAs QDs) to 1226 nm (InGaAsN QDs). This shift is lar-

ger than the maximum redshift achieved by simultaneous deposition of qua-

ternary InGaAsN (cf. sect. 5.3.3.1).

The most noticeable feature of the PL spectra of Fig. 38 (a) is the tre-

mendous loss of PL intensity associated with the supply of DMHy. This

phenomenon is known for both GaAsN210 and InGaAsN211, 212 QWs. It is

attributed to the deterioration of the crystal quality due to the small size and

the high reactivity of atomic nitrogen. Fig. 38 (b) shows a strong decrease

of the QD ground-state PL peak intensity, indicating also a decrease of the

QD density. The PL intensity can be improved by post-growth annealing

treatments that partially eliminate the N-induced crystal defects in both In-

GaAsN213, 214 and GaAsN210 layers. Annealing was not applied to the pre-

sent samples in order to maintain the comparability to ternary InGaAs QDs.

In view of the large intensity loss observed for the InGaAsN QDs obtained

via As/N exchange reaction, it is remarkable that the PL intensity of simul-

taneously deposited QDs as described in sect. 5.3.3.1 hardly decreases. Al-

though the method of simultaneous deposition is not as effective with re-

spect to the extension of the emission wavelength, the crystalline quality is

almost maintained.

Fig. 38: PL spectra of InGaAsN QD samples. The InGaAs QD samples were ni-

trided by adjusting the denoted DMHy partial pressures in the reactor chamber

during a GRI of 1 min after deposition of the InGaAs QD material. (a) Low-

excitation room-temperature PL spectra. The spectra of the nitrided samples are

smoothed. (b) High-excitation low-temperature spectra of the same samples, loga-

rithmically plotted.

5. MOCVD of quantum-dot structures for laser diodes

5.3.4 Wavelength shifting using antimony

Experiments were conducted to grow long-wavelength QDs by reducing

the band gap of InAs QDs via Sb insertions. The growth of InAsSb QDs via

Sb supply during the deposition of InAs QDs was explored. The investiga-

tions were carried out similarly to the experiments on N incorporation into

InGaAs QDs, as described in the previous section. Undoped test structures

containing single InAs(Sb) QD sheets were grown and investigated using

PL, PL-excitation spectroscopy (PLE) and high-resolution TEM.

The impact of Sb supply during deposition and/or GRI on the structural

and optical properties of the QDs was investigated on the basis of two sam-

ple series. For the samples of the first series (sect. 5.3.4.1), In, As and Sb

were supplied simultaneously. Then a GRI after deposition of the QD mate-

rial was applied to allow QD formation. These samples are compared to an

analogous series of InAs QD reference samples. In the second sample series

(sect. 5.3.4.2), pure InAs QDs were deposited. Different TESb flows were

supplied during a GRI after deposition of the InAs QDs.

5.3.4.1 Deposition amount

A series of samples with InAsSb layers with thicknesses between 1.6 and

2.0 ML was grown at 485°C. The TBAs/III ratio was kept at 1.5, the TESb /

(TBAs+TESb) ratio was 0.3. The QD material was deposited at a growth

rate of 0.4 ML/s. The TESb flow of 4.4 µmol/min supplied during QD

deposition was also kept during the GRI of 5 s. PL spectra of these samples

are compared to a reference series of InAs QDs in Fig. 39.

The deposition amount of the InAs QD samples was varied between 1.5

and 2.1 ML. In the case of binary InAs, the critical layer thickness is be-

tween 1.5 and 1.7 ML as one can see in Fig. 39 (a). This agrees well with

published values 132, 215. The wavelength right above the critical layer thick-

ness is 1.12 µm and shifts to almost 1.2 µm with increasing deposition

amount. For the sample with 2.1 ML InAs, a low energy shoulder emerges

at about 1.27 µm. This shoulder is ascribed to large coherent QDs occurring

in a low density, suggesting the formation of a bimodal size distribution.

The maximum PL signal comes from the sample with 1.8 ML, most likely

due to a high density of coherent InAs QDs and a low density of defects.

For larger deposition amounts, the signal decreases, most likely due to an

increase of the defect density. Disregarding the low-energy shoulder of the

2.1 ML sample, the emission wavelength of InAs could not be extended to

beyond 1.2 µm at room temperature. At 485°C, the maximum emission

wavelength of InAs QDs is widely unaffected by a variation of the V/III ra-

tio.

5.3. Redshift of the quantum-dot emission wavelength

The QD sample series with TESb supply during deposition and GRI

shows a 2D-3D transition between 1.6 and 1.7 ML, i.e. in the same regime

as for the InAs QDs. The behavior of the PL emission lines with increasing

deposition amount is completely different as compared to the binary InAs

QDs, as one can see in Fig. 39 (b). After the onset of the 2D-3D transition,

the emission wavelength immediately shifts to values around 1.3 µm and

beyond. The PL signal does not decrease as in the case of the InAs samples.

However, the average peak intensity of the InAsSb samples is by a factor of

4 lower than for the InAs QDs. The QD density in the sample with 1.7 ML

is probably very low, leading to a saturation of the QD ground-state transi-

tion and to a contribution of the first excited state, leading to the shoulder

on the high-energy side. The peak emission wavelength oscillates around

1.3 µm. This behavior suggests a minor reproducibility within the series.

It can be concluded from the comparison of Fig. 39 (a) and (b) that the

critical layer thicknesses for both InAs and InAsSb layers are comparable.

This means that the strain in both materials is similar, so that the fraction of

incorporated Sb is likely to be very low. For this reason, it is all the more

remarkable that the emission wavelength is strongly redshifted by about

160 meV with regard to the binary InAs QDs. In cross-sectional high-

Fig. 39: Room-temperature PL spectra of (a) InAs QDs and (b) InAsSb QD

sheets. In both cases, the deposition amount was varied over the same range. Both

InAs and InAsSb QDs were grown at Tgr = 485°C. A GRI of 5 s was introduced

after deposition of the QD material to allow QD formation. The InAsSb QDs were

deposited with an additional TESb flow of 4.4 µmol/min during deposition and

the subsequent GRI. The excitation power density was 5 W/cm2.

5. MOCVD of quantum-dot structures for laser diodes

resolution TEM images not shown here, diffraction contrast from Sb atoms

could not be detected. These findings indicate that Sb incorporation cannot

be the principal reason for the strong redshift. However, Sb incorporation

into InAs in the range of few percents has been reported elsewhere78. As

described below, the presence of Sb mainly accelerates the size increase of

the InAs QDs during the GRI, accounting for the longer emission wave-

lengths of the QDs grown with Sb supply.

5.3.4.2 TESb and TBAs during the growth interruption

Experiments to grow InAsSb QDs with TESb supplied only during the

GRI after deposition of the QDs surprisingly revealed that a similarly strong

redshift as described for the simultaneous deposition of In, As, and Sb can

be obtained. A sample series of InAs QDs with constant deposition amount

was grown. The InAs was again deposited at 485°C at a growth rate of

0.4 ML/s, using a TBAs/III ratio of 1.5. Various flows of TESb ranging

from 2.2 to 8.8 µmol/min were supplied during the subsequent GRI of 5 s.

Fig. 40 shows a PLE contour plot of such a QD layer, grown with a

TESb flux of 8.8 µmol/min. The bold curve is a PL spectrum of the same

sample (cf. also Fig. 43). Both PL and PLE spectra were recorded at 7K.

The PL spectrum shows a large number of peaks. Since the QD density is in

the 1010 cm-2 range as deduced from TEM images, and the excitation power

is very low, the high-energy peaks are unlikely due to excited-state transi-

tions. The PLE contour plot helps to identify the origin of this multiplet. In

a PLE contour plot, both excitation and detection energies are varied. For

every detection energy value, the excitation energy is varied from the detec-

tion energy to beyond the GaAs band gap energy. Vertical cuts show QD-

size selective PLE spectra. Signal from excited QD states is maximum at a

given detection energy if the excitation energy is resonant with the respec-

tive QD transition energy. In this case, the signal additionally scales with

the number of QDs of which the ground-state transition matches the detec-

tion energy.

One can see on the PLE contour plot that the PLE signal originating from

the first excited QD states oscillates with the detection energy. It can thus

be concluded that the size distribution of QDs is multimodal. From the ge-

ometry of the experimental setup, it can be ruled out that these peaks are

Fabry-Perot oscillations. The PL peak multiplet can neither be ascribed to

multiphonon relaxation of localized carriers since the energetic distances

between the peaks do not coherently match the GaAs LO-phonon energy of

36 meV (the energetic distances between the respective peaks as marked

with numbers 1 to 6 in Fig. 40 are

∆

E12= 34 meV,

∆

E23= 42 meV,

∆

E34=

44 meV,

∆

E45= 44 meV, and

∆

E56= 34 meV).

5.3. Redshift of the quantum-dot emission wavelength

Fig. 41 shows a cross-sectional high-resolution TEM image of the QD

structure. Flat, truncated-pyramid islands with sharp interfaces between

GaAs and WL and also between island top and matrix can be identified.

The white lines were drawn along the regions of group-V- / group-III-atom

contrast reversal. The contrast reversal of InxGa1-xAs occurs at around x =

0.15 as mentioned earlier. The TEM image suggests that the islands have

well-defined heights of integer numbers of atomic MLs. It is likely that the

discrete nature of island heights is the reason for the multiplet visible in the

PL and PLE spectra of Fig. 40. The single lines of the multiplet are rather

broad since the transition energies of the islands are not solely determined

by the height of the QDs. The lateral carrier confinement, being a function

of the QD width, likewise contributes to the transition energy. Since the lat-

eral extension of the islands is large, the contribution of lateral confinement

is rather weak, however.

x PL and PLE measurements by courtesy of Florian Guffarth and Robert Heitz.

1.20 1.15 1.10 1.05 1.00 0.95 0.90

1.05 1.10 1.15 1.20 1.25 1.30 1.35

1.1

1.2

1.3

1.4

1.5

1.6

Np295

T=7K

Detection Energy (eV)

Excitation Energy (eV)

PL intensity (arb. u.)

Wavelength (nm)

Fig. 40: Low-temperature PLE contour plot of an InAs sample with a TESb flow

of 8.8 µmol/min during the GRI of 5 s, subsequent to the InAs deposition (com-

pare to Fig. 43). PL was excited using a white-light source and a double mono-

chromator. The excitation density is below 5 mW/cm2. The intensity is plotted on

a logarithmic scale, the thin lines are equi-intensity lines. The bold black curve

shows a PL spectrum, excited at 514.5 nm and an excitation power density of

5 W/cm2. The logarithmic scale of the PL intensity is given on the right. The

peaks of the PL spectrum are partially numberedx.

5. MOCVD of quantum-dot structures for laser diodes

An 8-band k·p calculation qualitatively illustrates the relation between

QD height and transition energy. Fig. 42 shows QD ground-state transition

energies for truncated pyramids as a function of their respective height. A

constant QD base width of 13.6 nm and a facet angle of 45° were assumed

for all simulated islands. It is shown that the ground-state transition energy

monotonously decreases with increasing island height. The calculations fur-

ther show that the energetic distance between neighbored ground-state tran-

sition energies decreases with increasing aspect ratio. A similar observation

is made for the present samples, indicating that the aspect ratios of islands

with lower emission energies are higher.

In the following, the dependence of the QD-layer morphology on the

amount of TESb supply during the GRI is studied. In addition, the influence

of TBAs during the GRI is compared to that of TESb. Fig. 43 shows low-

temperature low-excitation PL spectra of InAs QD samples with varying

TBAs (a) and TESb supply (b). The right curve of Fig. 43 (a) shows the PL

spectra of an InAs QD sample without any precursor flow during the GRI.

Supply of the same TBAs flow during the GRI as used for the deposition of

InAs (QTBAs = 20.7 µmol/min) leads to a redshift of all transition energies.

The multiplet structure disappears, indicating an increase of the average as-

pect ratio: The islands gain in height and the size distribution becomes nar-

rower. Fig. 43 (b) shows PL spectra of InAs QD samples for which differ-

ent TESb flows are supplied during the GRI. With increasing TESb flow,

xi TEM image by courtesy of N.D. Zakharov and P. Werner, Max-Planck Institute

of Microstructure Physics, Halle, Germany.

8ML 7 ML 5 ML

Fig. 41: High-resolution cross-section transmission micrograph of a single InAs

QDs sheet grown at 500°C. A GRI of 5 s under TESb flux of 8.8 µmol/min was

performed after the deposition of the InAs. The lower InAs/GaAs interface as well

as the top interfaces of the truncated-pyramidal InAs islands are depicted with thin

white lines. The boundaries are determined according to the reversal of the group-

III/group-V-atom contrast, occurring at an indium fraction x of about 15 % for

InxGa1-xAs.xi

5.3. Redshift of the quantum-dot emission wavelength

the PL peak of the largest QDs strongly shifts to longer wavelengths. In

contrast to the supply of TBAs, however, the size distribution does not be-

come narrower. On the contrary, the ML splitting of the ground-state transi-

tions becomes far more pronounced.

Under the influence of both TBAs and TESb, the QDs exhibit an accel-

erated ripening behavior, suggesting that surface diffusion of In atoms is

enhanced. In addition, the size of the largest QDs of an ensemble is in-

creased. The increase of the average QD size upon supply of group-V pre-

cursors during the GRI can also be taken as a sign of a larger equilibrium

size of the QD ensemble. Thermodynamical calculations show that the

equilibrium size of a QD ensemble is a function of the surface energy119. If

the surface energy is lowered, the equilibrium size is increased. The reason

for an enhanced QD size after TESb / TBAs supply can be an enhanced

equilibrium QD size, suggesting that both TBAs and TESb act as surfac-

tants and lower the surface energy. The redshift during TBAs supply is not

as strong as in the TESb case but the ensemble width decreases. TESb is

more effective with regard to the QD size increase of the largest QDs. The

size distribution broadens, however, and the PL intensity decreases

strongly, most likely due to the formation of defects.

xii Calculations by Oliver Stier and Andrei Schliwa.

0.00 0.05 0.10 0.15 0.20 0.25

024681012

1.08

1.10

1.12

1.14

1.16

Transition energy (eV)

Island height (ML)

T = 0K

Aspect ratio

Fig. 42: Transition energies of truncated-pyramidal InAs islands as a function of

island height, calculated using 8-band k·p-theory. The InAs islands have constant

base length of 13.6 nm and 45° facets. With increasing island height, the shape

changes from a truncated pyramid towards a complete pyramid. The aspect ratio

of a complete pyramid with 45° facet angle is 0.5xii.

5. MOCVD of quantum-dot structures for laser diodes

From the above findings, it can be concluded that the strong redshift of

InAs QDs upon supply of Sb in the GRI is rather due to a size increase than

due to a change of the band gap of the QD material, related to Sb insertions.

Supply of TESb during the GRI causes a pronounced redshift of the QD PL,

but the size distribution of the islands and thus the spectral width of the

ground-state transition PL strongly increases. An acceleration of the QD

size increase during the GRI can be inferred from the fact that the duration

of the GRI applied after QD deposition was the same for all QD samples of

this series. It is conceivable that Sb acts as surfactant216, lowering the en-

ergy of the InAs surface. A lower surface energy leads to larger equilibrium

islands size in the SK growth mode, following thermodynamic arguments

by Shchukin et al.119.

Fig. 43: Low-temperature low-excitation PL spectra of undoped test structures,

each containing a single QD sheet. 1.7 ML InAs were deposited at 485°C with a

V/III ratio of 1.5. During the subsequent GRI of 5s, different precursors (TBAs

and TESb) were switched on. The impact of TBAs supply is shown in (a), the in-

fluence of TESb during the GRI is shown in (b). The rightmost curves in (a) and

(b) are reference InAs samples, grown without any precursor flow during the GRI.

They are slightly different in (a) and (b) since they stem from different growth se-

ries. The nominal deposition parameters are the same, however.

5.3. Redshift of the quantum-dot emission wavelength

The 1.3 µm QDs of which PL is shown in Fig. 43 (b) are not suitable for

the application as active medium in 1.3 µm lasers. A narrower QD size dis-

tribution is prerequisite for large modal gain, low threshold current densities

and high efficiencies. Other approaches such as the growth of an alternating

layer sequence of InAs and GaSb QWs are conceivable to obtain high-

density defect-free InGaAsSb QDs emitting at 1.3 µm with a small size dis-

persion. Although the thickness of each InAs and GaSb is below the critical

value for QD formation, the overall strain of the superlattice can exceed the

critical value and SK QD formation sets in. A similar observation has been

made for an alternating sequence of In0.5Ga0.5As and GaAs layers in

MOCVD217, leading to large self-organized InGaAs QDs.

6. Quantum-dot lasers

In this chapter, the growth of edge-emitting and vertically emitting injec-

tion laser diodes with stacks of QDs as active medium is described. The

layer structures of the edge-emitting diodes follow the state-of-the-art de-

sign of separate-confinement double heterostructures with multilayer active

regions. In sect. 6.1.1, the improvement of laser characteristics of edge

emitters upon in-situ annealing of QDs is described. These devices were

grown with conventional-precursor MOCVD. The growth of QD devices

with alternative precursors is described in sect. 6.1.2. Room-temperature

lasing at wavelengths beyond 1.24 µm from QD lasers grown using alterna-

tive-precursor MOCVD is reported in sect. 6.1.3. Sect. 6.2 reports growth

and characterization of QD VCSEL structures. In sect. 6.2.1 a preliminary

study of an InGaAs QW VCSEL is reported. The first MOCVD-grown QD

VCSEL world-wide has been demonstrated and is described in sect. 6.2.2.

Standard techniques were applied to process broad-area ridge-waveguide

lasers. Although processing of the laser structures is briefly described and

device characteristics are presented and discussed, processing and charac-

terization of the lasers were not part of this work. Processing of the edge

emitters is described in detail in Ref. 218. The lasers described in sects.

6.1.1.2 and 6.1.1.3 were processed and characterized at the Fraunhofer In-

stitute of Applied Solid State Physics in Freiburg, Germany. All other edge

emitters were processed at the Technical University of Berlin. Design,

modeling and processing of the oxide-DBR VCSELs as described in sect.

6.2 were carried out at the Technical University of Berlin and are the sub-

ject of a forthcoming Ph.D. thesis by F. Hopfer76.

6.1. Edge emitters

6.1.1 Lasers with in-situ annealed quantum-dots

The characteristics of previous devices grown using MOCVD were not

yet close to the ideal limits of low threshold and high quantum efficiency as

expected theoretically2-4, 11 and reported for a few exceptional MBE-grown

devices42, 58. Drastic improvements of the QD laser characteristics are re-

ported here, based on in-situ annealing of QD layers. The impact of anneal-

ing on surface morphology and QD PL is described in sect. 5.1. The influ-

ence of this technological step on the performance of laser diodes is dis-

cussed in the following sections. In-situ annealing of QD structures is con-

sidered to have general importance for the optimization of QD lasers143.

6. Quantum-dot lasers

6.1.1.1 Threshold reduction and increase of efficiency

The impact of in-situ annealing of QD lasers on threshold current and

quantum efficiency was investigated using a series of annealed laser struc-

tures grown using AsH3. The results are compared to QD lasers grown

without the annealing step.

Laser structures based on triple stacks of In0.8Ga0.2As QDs were grown

on n-GaAs(001) substrates using conventional-precursor MOCVD. The

GaAs below the QDs was grown at 600°C. Before deposition of the InGaAs

QDs, a sufficiently long GRI of 90 s was performed to reach thermal equi-

librium and a very smooth GaAs surface at the QD growth temperature of

490°C. After deposition of the QD material, a GRI of 60 s was performed in

order to achieve the desired emission wavelength12. From TEM images of

comparable QDs, an area density of nQD ≈ 1011 cm-2 is estimated. Subse-

quent to the QD formation, the QDs were covered by 7 nm of GaAs. The

first 2-3 nm of the GaAs cap were grown at the QD deposition temperature

of 490°C. Then the temperature was ramped to 600°C. This value was

reached after about 60 s, just when the growth of the 7 nm GaAs cap was

completed. Ramping the temperature to 600°C before starting the growth of

the GaAs cap results in structural degradation of the QDs. After deposition

of the cap, GRIs of varying duration (tA = 0, 10, 20, 30 min) were per-

formed under As-stabilized conditions (pAsH3 = 7.2 × 10-2 Torr = 9.6 ×

10-2 mbar). Then the GaAs cap growth was resumed. The active medium of

the present lasers eventually consists of three layers of In0.8Ga0.2As QDs,

separated by 35 nm GaAs, each grown by the described procedure. 1.1 µm

thick Al0.66Ga0.34As layers were used as cladding layers.

Ridge-waveguide lasers as depicted schematically in Fig. 1 were proc-

essed as follows: Broad area mesas are defined by optical lithography, the

300 nm thick contact layer and the 1200 nm thick top cladding are etched

down on both sides of the ridge to about 200 nm above the interface be-

tween optical confinement layer and top cladding. 100-150 nm dielectric

medium is subsequently sputtered on the surface except on top of the ridge.

Afterwards, the Ti/Pt/Au p-contact is brought onto the structure by vacuum

deposition. The wafer is thinned from originally 450-500 µm to 100-

150 µm. Finally, the Ni/AuGe n-contact is deposited on the back side of the

structure. For characterization of the laser structures, ridge-waveguides of

75, 100 and 200 µm width were processed in the present case. For widths

larger than 50 µm, current spreading hardly alters the threshold / transpar-

ency currents. Below this value, the threshold currents become dependent

on the etch depth and the quality of the interface between top cladding and

dielectric medium. Facets were fabricated by cleaving and were not coated.

6.1. Edge emitters

From the slope of the light-current (L-I) curve Popt(I) above threshold,

the differential quantum efficiencies

d (also called external or slope effi-

ciencies

ext) were calculated for different cavity lengths.

d represents the

change of the number of photons emitted on both facets upon a variation of

the number of injected electrons in the lasing regime:

()

opt

dII

−

} (6.1)

The differential quantum efficiency is affected by optical losses and thus

changes with cavity length and facet coating. The internal quantum effi-

ciency

i is independent from optical losses. It is as related to

d via













(6.2)

If the inverse differential quantum efficiency 1/

d is plotted as a function of

the device length, extrapolation to L = 0 yields the inverse internal quantum

efficiency 1/

i. The internal loss

i can be determined from the slope of

d as a function of L. Transparency current density, internal optical losses

and internal quantum efficiency are the most important parameters giving

access the quality of a laser structure.

Fig. 44 shows the L-I-characteristics of a 200 µm wide and 1.3 mm long

device. A maximum output power of 3.7 W is achieved in pulsed mode

(1 kHz, duty cycle 1:1250). Threshold current density and slope efficiency

d are 110 A/cm2 and 85 %, respectively. In undoped test structures, single

sheets of QDs deposited under the same conditions as for the present lasers

showed a PL maximum around 1190 nm. During the growth of the upper

cladding layer of the laser structures, which is typically performed at

620°C, the emission wavelength of such QDs is blue-shifted to 1150 nm

(cf. inset of Fig. 44, EL-spectrum). The inset of Fig. 44 shows that the

lasing modes emerge on the high-energy flank of the QD ground-state

emission. This wavelength shift is attributed to gain contributions from first

excited QD states219. For longer cavity lengths fabricated from the same wa-

fer, lasing occurs at 1160 nm220. Fig. 45 depicts the threshold current den-

sity versus the total optical loss (cf. Eq. 1.1).

Transparency currents of 17 A/cm2 and 20 A/cm2 for the devices with

tA = 10 min were extrapolated. The stripe widths are 75 µm and 100 µm,

respectively; the samples are from different growth runs. These values cor-

respond to a transparency current density of about 6 A/cm2 per QD sheet.

This is the smallest value ever reported for a semiconductor laser emitting

around 1100 nm. Annealing during 30 min increases the transparency cur-

6. Quantum-dot lasers

rent significantly. This is consistent with the decrease of the PL intensity of

the GaAs matrix (Fig. 22 (b)). The inset of Fig. 45 shows the L-I curves of

lasers grown with tA = 0, 10, and 30 min. The annealing step leads to a re-

duction of the threshold current by a factor of about 2.6 for tA = 10 min with

regard to tA = 0 min, indicating an increase of the quantum efficiency of

spontaneous recombination below threshold.

Fig. 46 depicts 1/

d as a function of the cavity length for tA = 0, 10, and

30 min. Internal quantum efficiencies of

i = 47 % (0 min),

i = 92 %

(10 min), and

i = 62 % (30 min) are obtained in correlation with the GaAs

matrix PL signal in Fig. 22 (b).

i is strongly enhanced up to 92 % for

tA = 10 min. The internal losses

i derived from the slope of the fits are

2.1 ± 0.5 cm-1 (0 min), 1.5 ± 0.3 cm-1 (10 min), and 2.1 ± 0.4 cm-1 (30 min).

To conclude, the introduction of in-situ surface annealing has shown to

drastically reduce the transparency current densities of QD lasers and to in-

crease the internal quantum efficiency to nearly 100 %. Device structures

with the lowest transparency currents ever reported for MOCVD-grown QD

lasers were demonstrated. As compared to QD lasers grown using conven-

tional-precursor MOCVD without QD annealing12, the threshold currents

were reduced by more than a factor of two. These improvements in the de-

vice characteristics are attributed to a strong reduction of non-radiative re-

combination centers in the GaAs matrix near the QDs.

012345678

1100 1150 1200

T=10°C

Output power (W)

Drive current (A)

(2)

(1)

EL (arb. u.)

Wavelength (nm)

Fig. 44: Output power (two facets) of a QD laser (10 min annealing step) versus

injection current (pulsed). Inset: electroluminescence of a laser-structure at

0.9 × jth (1) and 1.01 × jth (2). The stripe width of the device is 200 µm, the cavity

length is 1.3 mm. The threshold current density and slope efficiency are

110 A/cm2 and 85 %, respectively.

6.1. Edge emitters

6.1.1.2 High-Power Operation

Concern has been raised about the potential of QD lasers for high-power

emission because of their small volume filling factor, leading to lower mo-

dal gain than for conventional QW devices. In addition, refill times were

conjectured to be high due to slow intra-dot carrier relaxation, owing to the

phonon bottleneck effect221, 222. However, the first QD laser with an output

power beyond 10 W157 and very promising lifetime data recorded under

high-power operation at elevated temperatures is demonstrated and de-

scribed below.

The n-i-p type laser structure was grown on n-GaAs (001) substrate us-

ing conventional precursors. The active region inside the 0.4 µm thick SI

GaAs waveguide consists of six layers of self-organized In0.7Ga0.3As QDs,

grown at 485°C and separated by 30 nm GaAs. A more detailed description

of the QD growth is found in Ref. 156. Bottom and top cladding layers con-

sist of n- and p-doped Al0.3Ga0.7As, respectively. All layers were grown at

600°C, except for the QDs and the bottom cladding that was deposited at

720°C. Broad area devices with 150 µm wide ridges were processed using

0 2 4 6 8 10 12 14 16

100

1000

0 200 400 600

100

10 min, jtr=17 ± 4 A/cm2

10 min, jtr=20 ± 1 A/cm2

30 min, jtr=48 ± 3 A/cm2

thr

(A/cm

)

Optical loss (cm-1)

0 min

30 min

10 min

Output (mW)

Current density ( A/cm2)

Fig. 45: Threshold current density versus optical loss for devices with 10 min and

30 min annealing and stripe-width of 75 µm. Triangles show data for a laser with

QDs annealed during 10 min. Filled and empty triangles are data from the same

laser structure but from different growth runs. The stripe width is 100 µm for

these lasers. Straight lines are exponential fits. The stated errors are from the nu-

meric regression fit only, a general error of 10 % for the measurements can addi-

tionally be assumed, however. Inset: L-I-characteristics of devices grown with

annealing durations of tA = 0 min (jth = 340 A/cm2), 10 min (jth = 130 A/cm2), and

30 min (jth = 180 A/cm2). Stripe widths are 75 µm, cavity lengths are 1.4 mm (tA =

0 min) and 1.3 mm (tA = 10, 30 min). This figure has been published in Ref. 156.

6. Quantum-dot lasers

standard etching and metallization techniques. 2 mm long devices were

cleaved from the laser bars. The facets of the devices were high-reflection

(HR) and anti-reflection (AR) coated, leading to reflectivities of 95 % and

1 %, respectively. The laser diodes were finally mounted p-side down on

copper heat sinks using indium solder.

Output power versus injection current is depicted in Fig. 47 for two rep-

resentative broad area lasers emitting at 1135 nm for a stripe geometry of

2 mm × 150 µm. Device A shows an output power of 4.7 W in continuous-

wave (CW) operation at a heat sink temperature of 20°C. The lasing thresh-

old of this device is 650 mA (216 A/cm2), the maximum differential effi-

ciency equals 57 %, and the maximum optical power density at the front

facet is 7.5 MW/cm2. Device B was driven in quasi-CW (50 µs pulses with

50 Hz repetition frequency) at a heat sink temperature of 20°C. A maxi-

mum output power of 11.7 W was observed for the same stripe geometry.

Device B passes the lasing threshold at 478 mA (159 A/cm2) and exhibits a

maximum differential efficiency of 62 %. The optical power density of the

front facet reaches 19.5 MW/cm2. These are the highest CW output powers

of quantum-dot lasers reported to date. Both lasers fail due to catastrophic

optical mirror damage (COMD), attested by optical interference contrast

microscopy. Device A reaches the COMD level earlier since it is driven in

CW mode and thus heats stronger. Due to the finite temperature sensitivity

of the devices, a larger fraction of the drive current is converted to heat as

compared to quasi-CW operation. Values of the characteristic temperature

0.00.51.01.52.0

1.0

1.5

2.0

2.5

30 min

0 min

10 min

1 / η

Cavity length (mm)

Fig. 46: Inverse differential quantum efficiency as a function of cavity length for

different annealing durations. Dashed lines are linear fits. Stripe widths are

75 µm. An estimated error of about 10 % can be assumed. This figure has been

published in Ref. 156.

6.1. Edge emitters

for both devices are T0 = (73 ± 2) K between 20 and 50°C and T0 =

(53 ± 1) K between 50 and 80°C.

The COMD level of about 19.5 MW/cm2 ranks among the highest values

reported for QD semiconductor laser diodes and is comparable to the best

values reported for conventional QW laser diodes grown on GaAs sub-

strates. For Al-free lasers with InGaP claddings, 20 MW/cm2 have been re-

ported for as-cleaved facets and QDs as gain medium (Theatsink = 20°C)12.

18.5 MW/cm2 have been observed for a QW as active layer and passivated

facets (Theatsink = 10°C)223. 19 MW/cm2 were achieved with an InGaAs QW

as gain medium, AlGaAs claddings and AR/HR coatings (Theatsink =

10°C)224. 20 MW/cm2 is probably close to an upper limit for COMD of

conventionally mounted laser diodes with stabilized heat sink temperature.

The COMD level of an InGaAs/AlGaAs QW laser could be extended to be-

yond 30 MW/cm2 using a special mounting technique combined with tem-

perature stabilization of the laser chip225.

In conclusion, a maximum output power of 4.7 W was obtained in CW

mode, 11.7 W were achieved in quasi-CW operation. The output power of

these lasers is larger than the output power of the lasers described in sect.

6.1.1.1. This is attributed to the active region of the present lasers, consist-

ing of a six-fold QD stack. The devices presented in sect. 6.1.1.1 have only

three QD layers as active medium. The output power of the present lasers

0 5 10 15 20

12 T = 20 °C

(A) CW mode

(B) 50 µs, 50 Hz

Output power (W)

Drive current (A)

Wall-plug efficiency (%)

Fig. 47: Front facet output power of two different devices, driven in CW (A) and

quasi-CW mode (B), respectively. The stripe geometry is 2 mm × 150 µm. The

heat sink temperature was stabilized at 20°C. Device (A) shows CW operation up

to 4.7 W. Device (B) was driven up to 11.7 W in quasi-CW mode with 50 µs

pulses and 50 Hz repetition frequency. This figure has been published in Ref. 157.

6. Quantum-dot lasers

was limited by catastrophic optical mirror damage, occurring at a power

density of about 19.5 MW/cm2 on the front facet.

6.1.1.3 Lifetimes

Lifetime measurements of lasers with six-fold QD stacks in the active

region were performed. Growth and characteristics of these devices are de-

scribed in the previous section. The laser facets were passivated with AR

and HR coatings. Lifetime measurements during 3040 h at 50°C are shown

in Fig. 48 and give evidence of the high reliability of InGaAs / GaAs QD

lasers.

The measurements were started at 1.0 W output power. After 910 h, the

lasers were examined for changes of the device characteristics. No signifi-

cant changes were found. The measurements were resumed at 1.5 W output

power. No significant changes of the conversion efficiency are observable

during the second measurement period, either. A maximum output power of

9 nW/QD in quasi-CW mode was calculated. This corresponds to an upper

limit of 19.5 ps for the refill time. These values are comparable to numbers

previously published for an MOCVD-grown laser diode based on a three-

0 500 1000 1500 2000 2500 3000 3500

0.0

0.5

1.0

1.5

2.0

I = 2.5 A

I = 3.5 A

Output power (W)

Time (h)

T = 50°C

1.10 1.15 1.20

-80

-70

-60

-50

-40

I = 0.6 A

Intensity (dBm)

λλ

λ (µm)

I = 2.5 A

Fig. 48: Lifetime measurements of six 2 mm long and 150 µm wide laser stripes,

driven in CW mode at a heat sink temperature of 50°C at two different drive cur-

rents. The measurement was interrupted after 910 hours. The depiction of the

temporal offset between both measurements is arbitrary. The inset shows lasing

spectra near threshold at 0.6 A, and at 2.5 A. This figure has been published in

Ref. 157.

6.1. Edge emitters

fold stack of comparable InGaAs QDs12, indicating that carrier capture is

not influenced when the number of QD layers is doubled.

6.1.1.4 Characteristic temperature

Temperature insensitivity of the threshold currents of ideal QD lasers has

been predicted3 assuming a zero-dimensional density of states in the QDs

(Eq. 2.3), QD confinement energies larger than kBT at room temperature

and single confined electron and hole levels. This is in contrast to threshold

currents of QW lasers that increase exponentially with temperature. For

temperatures up to 150 K, the prediction of such a temperature insensitivity

holds indeed. Even a decrease of the threshold current with temperature is

observed for QD lasers up to typically 150 K. This is due to the transition

from a non-equilibrium to an equilibrium carrier distribution function43

within the QD ensemble.

In the case of more than one electron and hole level per QD and small

energetic distances between these levels in real QDs, the threshold current

becomes temperature-sensitive above 150 K. The energy separation be-

tween the electron ground state and the first excited state in QDs is typically

between 50101 and 100 meV226, 227, depending on size, shape and chemical

composition of the QDs. The separation between the discrete hole levels,

however, is estimated to be in the order or 10 meV only228 so that at room

temperature, thermal spreading of the holes over closely lying hole states

within the QDs can dominate the gain characteristics at room tempera-

ture229. In the temperature-sensitive regime, the threshold current of a QD

laser therefore increases with temperature and can be approximated by

)/exp()( 0th TTTj ∝ (6.3)

where T0 is the so-called characteristic temperature and is used as a general

measure of the temperature stability of the threshold currents of laser di-

odes.

It is conceivable that charge carriers are thermally excited to higher-lying

electronic levels from which non-radiative recombination takes place. Non-

radiative recombination might particularly occur via matrix states. The car-

rier density in the matrix increases with temperature, and if a radiative effi-

ciency of, e.g., only 5 % is assumed in the matrix, a T0 value of 54 K is ob-

tained theoretically74. This value is in the range of typical T0 values of QD

lasers emitting around 1.1-1.2 µm.

It should be noted in this context that laser diodes with exceptionally

high threshold currents usually exhibit high characteristic temperatures.

This does not necessarily mean, however, that the effect of thermal excita-

tion of carriers does not occur or that the radiative efficiency in the matrix

6. Quantum-dot lasers

is especially large. On the contrary, this effect is more likely due to leakage

currents that are slowly varying functions of temperature. If leakage cur-

rents are large as compared to the current needed for inversion of the laser

active medium, the threshold current increases much less with temperature.

In order to assess the contribution of non-radiative recombination in the

matrix to the low T0 values, QD lasers with and without Al0.3Ga0.7As carrier

diffusion barriers close to the active region were grown and characterized.

These barriers prevent carriers to escape from the active material to matrix

states. In addition, single quantum well lasers with and without diffusion

barriers were grown as reference. The design of such a double-barrier active

region goes back to an idea of Tsang et al.230 who demonstrated an im-

proved characteristic temperature of double-barrier QW lasers.

Fig. 49 (a) shows a schematic diagram of the center of such a double-

barrier single-QW laser structure, emitting around 1050 nm at room tem-

perature. The barrier below the InGaAs QW is 5 nm thick, the p-side barrier

has a thickness of only 4 nm, taking smaller tunnel rates due to the larger

effective mass of the holes into account. The layer sequences as well as the

temperature ramps of QD and QW laser structures were exactly the same.

In contrast to the design of Tsang, a distance of 5 nm was kept between the

barriers and the active layers since PL experiments have shown that if the

AlGaAs barriers are grown too close to the active layer, the radiative effi-

ciency drops by orders of magnitude. This is ascribed to contaminations at

the GaAs/AlGaAs interfaces. The non-radiative recombination efficiency

related to these contaminants increases strongly with decreasing distance

between InGaAs and AlGaAs.

Fig. 49: (a) Schematic diagram of a double-barrier DHS QW injection laser. (b)

Temperature dependence of the threshold current of a QW laser with and without

double-barrier design.

n-side

5 nm GaAs

5 nm In0.15Ga0.85As

4 nm Al0.3Ga0.7As

5 nm GaAs

n-side

4 nm Al0.3Ga0.7As

5 nm GaAs

5 nm In0.15Ga0.85As

5 nm Al0.3Ga0.7As

5 nm GaAs

(a) p-sidep-side

150 200 250 300 350 400

100

1000

(b)

T0= 160 K

T0= 212 K

T0= 73 K

without barriers

with barriers

Threshold current density (A/cm

)

Temperature (K)

6.1. Edge emitters

Lasers with 75 µm wide ridges and as-cleaved facets were processed

from these 4 laser structures; the cavities were 1.8 mm long. All measure-

ments were performed in pulsed mode operation. T0 measurements of the

single-QW-laser with and without barriers are shown in Fig. 49 (b). The

given temperatures are those of the copper heat sink. One can see that be-

tween 150 K and room temperature, the T0 values of the devices with and

without barriers are 212 K and 160 K, respectively. Above room tempera-

ture, the T0 value of the barrier-free QW laser drops to 73 K whereas the

device with the barriers keeps the large T0 value up to an operation tempera-

ture of 330 K. From the influence of the barriers on the T0 of QW lasers, it

can be concluded that non-radiative recombination in the GaAs matrix does

indeed impact the temperature dependence.

In contrast to the QW laser, the impact of the double barrier on the T0 of

QD lasers is negligible. T0 of the triple-stack QD laser without barriers is

60 K at room temperature. The device with the diffusion barriers showed a

T0 of about 70 K. The threshold current of this device is slightly larger,

however, most likely due to impurities at the Al0.3Ga0.7As/GaAs interfaces.

It can be concluded from these findings that the improvement of the charac-

teristic temperature due to the presence of the barriers is negligible, indicat-

ing that other mechanisms dominate the high temperature sensitivity of QD

lasers. It is likely that this temperature sensitivity is related to the close en-

ergetic hole spacing of the QDs. Long-wavelength QD lasers with larger

carrier confinement and larger inter-sublevel spacing of electron and hole

states show better temperature stabilities231.

Experiments on p-type modulation doping near the QD layers of laser

diodes have shown that T0 can significantly be increased by an excess hole

population of the QDs229, 232. This way, the number of hole states that can

be populated by current injection is strongly reduced. The holes available

for radiative recombination are thus concentrated in a narrower energy win-

dow. With p-doping levels between 5 × 1017 cm-3 and 1.5 × 1018 cm-3, char-

acteristic temperatures above 160 K between 0°C and 80°C were achieved

at reasonably low threshold currents for a twofold QD stack as active me-

dium232. The ground-state gain of this structure could be increased from

9 cm-1 to maximum values of 18 cm-1. A similar problem of closely spaced

hole subbands exists in planar QWs. P-type modulation doping could also

increase the characteristic temperature of QW lasers. However, the level of

p-doping needed for QWs is nearly one order of magnitude larger229. In the

p-doping range required for such QW lasers, doping-related optical loss

mechanisms such as free-carrier absorption would worsen the device per-

formance significantly.

6. Quantum-dot lasers

6.1.2 Edge emitters grown with alternative precursors

MOCVD of QD laser diodes with close-to-ideal characteristics and high

output power has been demonstrated using conventional precursors only12,

156, 157 (cf. also sect. 6.1.1). QD laser diodes with equivalently outstanding

performance exclusively grown using alternative precursors are demon-

strated here. This is all the more remarkable since no in-situ annealing was

applied during growth of the lasers presented in this section. Only during

growth of the 1.24 µm lasers described in section 6.1.3, in-situ annealing

led to a significant improvement of the device characteristics. The demon-

stration of lasers grown using alternative precursors is decisive for the

propagation of MOCVD as laser fabrication technology in regions with

high safety standards where strongly toxic hydrides cannot be used.

The n-i-p type laser-diode structure was grown on an n-GaAs:Si (001)

substrate. The active medium consists of three In0.5Ga0.5As QD layers,

stacked with a distance of 35 nm in the center of a 300 nm thick undoped

GaAs optical-confinement layer. The QDs were deposited at 500°C with a

growth rate of about 0.4 ML/s. All precursor flows were switched off here-

after for 30 s to allow QD formation. The first 5 nm of GaAs on top of the

QDs were also grown at 500°C. The reactor temperature was then ramped

to 600°C. The deposition temperature of layers underneath the QDs was

650°C. All layers on top of the QDs were grown at 600°C. 1 µm thick

Al0.65Ga0.35As:Te and Al0.7Ga0.3As:C layers were used as bottom and top

cladding layers, respectively. 100 µm wide ridge-waveguide lasers were

processed using standard techniques, facets were left uncoated.

Fig. 50 shows room-temperature electroluminescence spectra of such a

ridge-waveguide QD laser as a function of drive current. The PL spectrum

of an undoped PL structure containing one QD layer grown under the same

conditions as the laser structure is also plotted. The wavelength of the PL

peak coincides almost perfectly with the lasing wavelength, indicating that

lasing occurs via the QD ground state. The evaluation of the L-I curves of

four devices with different cavity lengths is shown in Fig. 51. A transpar-

ency current as low as 29.7 ± 0.3 A/cm2 is derived. The internal quantum

efficiency is 91.4 ± 3.0 % and the internal optical loss is 2.2 ± 0.2 cm-1 (cf.

inset). These numbers are close to the best values ever achieved with QD

lasers grown by MOCVD using AsH3

156, although no spacer annealing was

applied here.

The outstanding characteristics of the first QD laser grown with alterna-

tive-precursor MOCVD gives evidence of the high optical quality of QDs

grown using TBAs.

6.1. Edge emitters

6.1.3 Long-wavelength (>1.24 µm) QD lasers

Using alternative-precursor MOCVD, an edge-emitting laser structure

with long-wavelength QDs as active region was grown, processed and char-

acterized15. To achieve long-wavelength emission, ternary InGaAs QDs

were overgrown with Ga-rich InGaAs QWs. Spectroscopic and structural

studies of such long-wavelength QDs were described earlier in sect. 5.3.1

and 5.3.2. In this section, room-temperature lasing activity at wavelengths

beyond 1.24 µm is reported. This is the worldwide first demonstration of

lasing beyond 1.24 µm reported for MOCVD-grown GaAs-based QD la-

sers.

The laser structure was grown on a Si-doped GaAs(001) substrate. The

active region consists of 10 sheets of In0.65Ga0.35As/In0.2Ga0.8As/GaAs long-

wavelength QDs, separated by 40 nm GaAs. The QD stack was centered in

a 480 nm thick undoped GaAs optical confinement layer, sandwiched by n-

and p-doped 1.1 µm Al0.6Ga0.4As cladding layers. A 350 nm thick p++ con-

tact layer was deposited on top. The bottom cladding layer was grown at

650°C; waveguide, top cladding layer and contact layer were grown at only

600°C to avoid any In-Ga intermixing of the QDs. The QDs were grown as

follows: 2.7ML of In0.65Ga0.35As were deposited at 500°C. A GRI of 1 min

0.91.01.11.2

PL intensity (arb. u.)

2.1 x jth

1.1 x jth

0.9 x jth

0.4 x jth

0.2 x jth

EL intensity (arb. u.)

Wavelength (µm)

EL, pulsed mode

T = 19°C

Fig. 50: Electroluminescence spectra of a 100 µm × 2 mm ridge-waveguide laser

with a threshold current density of 60 A/cm2, driven in pulsed mode (500 ns,

5 kHz). The laser structure was grown with alternative precursors only. Laser

spectra were recorded with monochromator slit widths reduced from 1 to 0.1 mm,

leading to a signal decrease by about 100. Peaks at 1.01 µm stem from the labora-

tory lighting. Dotted line: PL of a single QD layer excited with 5 W/cm2 at

514.5 nm. This figure has been published in Ref. 161.

6. Quantum-dot lasers

was introduced to allow QD formation. All precursors were switched off

during the GRI, including TBAs. 5 nm of In0.2Ga0.8As were subsequently

deposited directly on top of the QDs. Overgrowth of In0.65Ga0.35As QDs

with such a QW does not lead to a significant decrease of the integrated PL

efficiency, while the redshift of the QD luminescence is considerable. As

estimated from TEM images, the QD density is about 2.6 × 1010 cm-2. After

deposition of the InGaAs QW, GaAs growth was resumed. First, 2 nm

GaAs were deposited at the QD growth temperature. Then the growth was

again interrupted for in-situ annealing at 600°C during 3 min (cf. also sect.

5.1). The remaining part of the GaAs spacer is grown at 600°C. A compari-

son of PL spectra of test structures with single QD sheets grown with and

without in-situ annealing shows that upon in-situ annealing, the PL intensity

is significantly higher. The spectral width (FWHM) of the PL emission is

reduced by the annealing step from 72 meV to 64 meV.

Ridge-waveguide lasers with 50 µm ridge width were processed using

standard techniques. Laser facets were left uncoated. The devices were

mounted p-side up and were driven in pulsed mode with a pulse width of

800 ns at 1 kHz repetition rate. Fig. 52 shows the dependence of threshold

current on the cavity length at room temperature. The lowest threshold cur-

rent was measured as 220 A/cm2 for the 2.9 mm long device. The extrapola-

tion of a logarithmic fit of the threshold current densities as a function of

cavity length yields a transparency current density of 139 A/cm2. Fig. 52

0.0 0.2 0.4 0.6 0.8 1.0

100

0.0 0.5 1.0 1.5 2.0

1.0

1.5

2.0

(29.7+0.3) A/cm2

jth (A/cm

)

1/L (mm-1)

T = 19 °C

pulsed mode

jtrans =

L (mm)

ηη

ηint = (91.4 + 3.0) %

αα

αint = (2.2 + 0.2) cm-1

Fig. 51: Threshold current densities of 100 µm wide ridge-waveguide lasers as a

function of inverse cavity length. The dashed line shows a linear fit of the decadal

logarithm of jth. Inset: inverse differential quantum efficiency

d as a function of

cavity length. The devices were driven with 500 ns long current pulses at a repeti-

tion frequency of 5 kHz. This figure has been published in Ref. 161.

6.2. Quantum-dot vertical-cavity surface emitters

shows the lasing spectrum of a 2.3 mm long and 50 µm wide laser stripe

with for j = 1.07×jth. The internal efficiency and the internal optical losses

are (47 ± 5) % and (4.0 ± 0.4) cm-1, respectively. A transparency current

density of 139 A/cm2 corresponds to a transparency current density per QD

sheet of about 14 A/cm2. This is roughly twice the current as for the laser

diodes reported in sect. 6.1.1.

The relatively low internal quantum efficiency points to a yet significant

rate of non-radiative recombination. This points to a large density of defects

that are still present in the QD sheets. The internal efficiency is potentially

increased by optimizing the in-situ annealing step. For a further redshifting

of the lasing emission wavelength, the use of binary InAs stressors might be

advantageous. Due to the larger strain of such stressors, the alloy phase

separation can be enhanced. Previous to the use of such InAs stressors,

however, the large inhomogeneous broadening (cf. sect. 5.3.4) must be con-

trolled.

6.2. Quantum-dot vertical-cavity surface emitters

As outlined in sect. 2.1, vertical emitters are very attractive devices, e.g.

because of the possibility to fabricate monolithic arrays for parallel data

transmission; the circular beam profile of VCSELs allows high coupling ef-

ficiencies into optical fibers. The fabrication of VCSELs, however, repre-

sents an enormous challenge to both device epitaxy and device processing:

012345678

100

1240 1245

1 / Cavity length (cm-1)

(A/cm²)

500

jtrans = 139 A/cm2

pulsed mode

(800 ns, 1 kHz)

Wavelength (nm)

EL int. (arb. u.)

j =

1.07 x jth

L = 2.3 mm

Fig. 52: Threshold current vs. inverse cavity length of a laser diodes based on a

10-fold stack of In0.65Ga0.35As/In0.2Ga0.8As/GaAs QD sheets. The stripe width is

50 µm. The inset shows a lasing spectrum of a 2.3 mm long and 50 µm wide de-

vice at room temperature. This figure has been published in Ref. 15.

6. Quantum-dot lasers

100

The resonance wavelength of the cavity must exactly match the gain maxi-

mum of the active medium, and mirror reflectivities must be high at this

wavelength.

Whereas for edge-emitting devices, cavity lengths ranging from several

100 µm to few millimeters can arbitrarily be chosen by ex-situ cleavage, the

cavity length of a VCSEL is fixed during growth by choice of the cavity-

layer thickness. The longitudinal mode spacing of edge emitters is much

smaller than the inhomogeneous broadening of the QD ground-state transi-

tion so that exact matching of the cavity length to longitudinal modes is not

required. The longitudinal mode spacing of VCSELs, however, is much lar-

ger than the typical inhomogeneous broadening of the QD ground state

transition. This is because of the short cavity height of only a few λ. There-

fore, exact matching of the cavity height of VCSELs to the desired longitu-

dinal mode within 1 % is mandatory. If the thickness of a 300 nm thick

GaAs cavity as used for a VCSEL structure emitting at 1.0 µm is changed

by 1 %, for example, the cavity resonance wavelength is altered by about

10 nm. This can have a dramatic impact on the device performance.

The short cavity length of a few λ requires extremely large mirror reflec-

tivities in order to keep total optical losses below the gain maximum of the

active region (cf. Eq. 2.1). Such reflectivities of more than 99 % can only

be provided by DBRs. On GaAs substrates, Al(Ga)As/GaAs DBRs can be

used. GaAs-based optoelectronic devices benefit from the unique incidence

that AlAs is almost lattice-matched to GaAs. Due to the large difference of

optical indices between GaAs and AlAs, such AlAs/GaAs DBRs are highly

efficient. In other material systems, DBRs are much less efficient. The effi-

ciency of InP/InGaAsP DBRs as used for the growth of monolithic

VCSELs on InP substrates233, for example, is much lower so that around 50

InP/InGaAsP pairs are required for each DBR. For GaAs-based VCSELs,

30 pairs of AlAs/GaAs are usually sufficient. The large penetration depth of

the light wave into the Bragg mirrors makes high demands on the layer

quality.

Although AlAs/GaAs DBRs are very efficient, Al(Ga)Ox/GaAs oxide

DBRs were used as mirrors for the VCSELs grown within this work. The

difference of the optical indices of GaAs and Al(Ga)Ox is much larger so

that only 7 and 6 Al(Ga)Ox/GaAs pairs are needed for bottom and top mir-

rors at 1.1 µm emission wavelength, respectively. This reduces growth

times drastically and saves expensive precursor material as compared to all-

semiconductor AlAs/GaAs DBRs. Another great advantage of oxide DBRs

is the wide stop band, enabling comparably large tolerances of the DBR

layer thicknesses. The processing of oxide-DBR VCSELs, however, re-

quires a sophisticated wet-thermal oxidation technique234, 235 as outlined in

the following section.

6.2. Quantum-dot vertical-cavity surface emitters

101

Since QD VCSELs are very sensitive to optical losses and low mirror re-

flectivities, owing to the low modal gain of QDs, the development of a QD

VCSEL took place in two steps. As a first step, a QW VCSEL was designed

which helped to assess the quality of the different device modules, as de-

scribed in the following section. A two step oxidation process was applied

for the fabrication of the QW VCSELs. Previous to the deployment of QDs

as active medium, drawbacks could be eliminated (cf. sect. 6.2.2). The QD

VCSELs were fabricated using a one-step oxidation process, as described in

sect. 6.2.2.

6.2.1 Quantum-well VCSEL

As a first step towards the fabrication of a QD VCSEL, a single-

quantum-well VCSEL structure was grown, processed and characterized.

Detailed information on the quality of the different parts of a VCSEL can

only be accessed by laser parameters such as threshold current densities and

efficiencies. Therefore, it is particularly important to achieve lasing with

such a VCSEL test structure. A single QW was hence chosen as active me-

dium for this device since the modal gain of QDs might have been too low

to obtain lasing, potentially due to non-optimized layer design.

The single-InGaAs-QW VCSEL structure with Al(Ga)As/GaAs layers

for subsequent wet-thermal oxidation of Al(Ga)Ox/GaAs DBRs was grown

on the Aix200 machine using conventional precursors. The InGaAs QW in

the active region was deposited at 490°C. It has a thickness of 5 nm and

shows maximum room-temperature PL at 1050 nm. The bottom DBR struc-

ture was grown at 720°C, the top DBR was deposited at 600°C. All other

layers were deposited at 600°C. In contrast to In-rich QDs, the effect of In-

Ga intermixing in Ga-rich InGaAs QWs is rather low so that a larger depo-

sition temperature could have been chosen for the top DBR. It was the aim

of this growth experiment, however, to optimize a VCSEL structure that

can eventually be used for quantum dots as active material. The deposition

temperature for layers above QDs should not be increased to more than

600°C. To increase the mechanical stability of the oxide DBRs, 10 nm thick

AlxGa1-xAs buffer layers were inserted between the GaAs layers and the

Al0.98Ga0.02As layers of the bottom and top DBR. The aluminium fraction x

was ramped digitally within these buffer layers from 0.3 to 0.9, with the

low Al-fraction sides oriented towards the GaAs layers. Such Ga-rich Al-

GaAs layers oxidize very slowly and prevent the GaAs layers from vertical

oxidation75.

The design of the VCSEL is shown schematically in Fig. 2. The structure

has 7 bottom and 6 top DBR pairs and two oxide apertures on both sides of

the active region. The design was optimized for minimum optical losses,

6. Quantum-dot lasers

102

taking into account low modal gain of QD sheets, which are deployed later

as active region. Low optical losses were realized using the lowest possible

doping levels for the contact layers. n- and p-doping levels of 1 × 1018 cm-3

were chosen for the n- and p-contact layers, respectively. This is close to

the minimum values needed to achieve Ohmic contacts.

Fig. 53 shows a schematic overview of the process steps applied for ox-

ide-DBR VCSELs with intra-cavity contacts. The process starts with Che-

mically-Assisted Ion-Beam Etching (CAIBE) of the top-mirror mesas down

to the top interface of the p-contact layer. The process outlined schemati-

cally in Fig. 53 includes two separate oxidation steps. First, top DBR and

the apertures are selectively oxidized. A λ/4 thick SiNx layer is then depos-

ited onto the structure to protect the top oxide DBR from degradation dur-

ing the second oxidation step. The SiNx is structured using Reactive-Ion

Etching (RIE). The same photoresist mask is used for the subsequent

CAIBE etching of the middle mesa down to the top interface of the n-

contact layer, while a new mask is required previous to CAIBE etching of

the bottom-DBR mesas. After the following simultaneous selective oxida-

tion of the apertures and the bottom DBR, the pillars are partly planarized

using photo-bencocyclobutene. A second λ/4 layer of SiNx is deposited

onto the whole structure to protect the oxide layers during the subsequent

lithography steps for the electrical contacts. This has turned out to be neces-

sary since all developing solutions used for the definition of electrical con-

tacts have shown to strongly etch AlOx. Now the SiNx layer has a thickness

Fig. 53: Schematic overview of the two-step oxidation process used for the fabri-

cation of single-QW VCSELs with oxide aperture and all-oxide DBRs (cf. text).

After Ref. 76.

6.2. Quantum-dot vertical-cavity surface emitters

103

of λ/2 and consequently does not alter the reflectivity of the structure at the

design wavelength. The last step is the metallization of the p- and n-

contacts.

Lasing in pulsed mode was achieved for different mesa sizes. The low

doping levels of the contact layers imply high series resistances that would

lead to extensive heat development if the devices were driven in (a) shows

the L-I-curve for a device with a 25 µm wide circular aperture. The thresh-

old current is 2 mA, corresponding to a threshold current density of

400 A/cm2, and the slope amounts to 0.16 W/A (corresponding to 13.5 %

differential efficiency). A maximum output power of 1.6 mW CW mode.

Immediate thermal rollover and device degradation would be the conse-

quences. Fig. 54could be achieved. Fig. 54 (b) shows spectra closely below

and above threshold. For I = 1.5 × Ith, the electroluminescence peak exhibits

a FWHM of 0.3 nm. A device with an aperture diameter of 12 µm showed

lasing with a threshold current of 1 mA (900 A/cm2) at a bias of 8 V. The

onset voltage of this device is 3.5 V in agreement with previously calcu-

lated values.

Fig. 55 (a) shows a reflection and a lasing spectrum of a device with a

25 µm aperture. A weak cavity dip is visible at the lasing wavelength. It has

a width of 1-2 nm, which is comparable to the resolution of the measure-

Fig. 54: (a) L-I-characteristics and (b) spectra below and above lasing threshold of

a QW VCSEL with an aperture diameter of 25 µm. The threshold current density

is 400 A/cm2.

6. Quantum-dot lasers

104

ment setup. The visibility of the cavity dip indicates a non-perfect top mir-

ror. The image of a QW-VCSEL with a 25 µm aperture as well as its near-

field distribution under pulsed-mode operation are shown in Fig. 55 (b) and

(c), respectively. For an aperture diameter as large as 25 µm, a very inho-

mogeneous current injection with current maxima near the boundary of the

aperture is expected (current crowding). In this case, lasing is expected only

at the boundary regions of the aperture. However, the CCD image also

shows intensity maxima in the center of the aperture. Since the cavity be-

tween the contact layers is undoped, the current probably spreads over the

whole aperture, which leads to sufficient gain also in the center of the opti-

cal mode.

The aperture of the 25 µm VCSEL has the same diameter as the top mir-

ror. In that case, huge optical losses are expected. The threshold current

density of this device is nevertheless rather good. This indicates low optical

losses inside the cavity. The device with the smaller aperture diameter of

12 µm exhibits a significantly larger threshold current density. This effect is

Fig. 55: (a) Reflection spectra and EL spectra of a single-InGaAs-QW oxide-DBR

VCSEL with a 25 µm aperture. The reflection shows a weak cavity dip. (b) Mi-

croscopic image of an oxide-DBR QW VCSEL with a 25 µm aperture. (c) Near-

field distribution of the same device at an averaged output power of 13 µW (duty

cycle: 0.8 %). The nearfield was recorded using a Si-CCD camera having a rather

low sensitivity at 1050 nm.

(a)

(b) (c)

6.2. Quantum-dot vertical-cavity surface emitters

105

ascribed to a lower quality of the top oxide-DBR in the center region: Al-

though all top mirrors were covered with more than 200 nm SiNx to prevent

them from degradation during the oxidation of the bottom DBRs, a slight

degradation could not be avoided. Smaller top DBRs with a larger perime-

ter-to-area ratio are more sensitive to degradation, and their quality and thus

reflectivity are lower, requiring stronger pumping to reach the laser thresh-

old.

The InGaAs-QW oxide-DBR test VCSEL gives evidence of the control

of both VCSEL growth technology and wet-thermal oxidation of DBRs.

This is an important result with regard to the implementation of a QD active

region where mirror reflectivities and doping levels are more critical to the

device performance because of the low modal gain of QD layers. For the

processing of quantum-dot VCSELs, a one-stage oxidation process was

therefore developed which circumvents the problem of mirror degradation

and leads to a larger top-DBR reflectivity.

6.2.2 Quantum-dot VCSEL

A QD full-oxide-DBR VCSEL structure was grown with alternative pre-

cursors using the Aix200/4 machine. The structure was fabricated using a

one-stage oxidation process. This way, the danger of over-oxidizing the top

oxide DBR is eliminated. On the other hand, the oxidation rates for top-

DBR oxide, bottom-DBR oxide and aperture have to be calibrated within

10 % (cf. also sect. 3.3.1.3). Room-temperature lasing was observed around

1100 nm on the QD ground state. This is the worldwide first demonstration

of an MOCVD-grown electrically pumped QD VCSEL.

The refractive-index profile of the VCSEL structure is depicted in

Fig. 56. The graph shows the active region, the contact layers, and also the

intensity distribution of the optical wave. As active medium, 9 In0.5Ga0.5As

QD layers were used in 3 stacks. Details of the QD growth are described in

sect. 5.2. These QDs have a sheet density of 4-5 × 1010 cm-2. QDs deposited

under the same growth conditions were previously used as 3-fold stacks in

edge-emitting lasers, as described in sect. 6.1.2.

As inferred from the characteristics of the QW test VCSEL described in

the previous section, a two-stage oxidation process leads to a poor quality

of the top DBR. Excess oxidation of the top-DBR oxide during the second

oxidation step reduces its optical quality and increases strain in the device75.

Since high mirror reflectivities are highly needed in QD VCSELs, however,

a one-stage oxidation process was developed during which top DBR, bot-

tom DBR and apertures are oxidized simultaneously. Due to the pillar de-

sign of the present VCSELs with three different mesa sizes for top DBR,

cavity with aperture, and bottom DBR, three different oxidation lengths

6. Quantum-dot lasers

106

have to be controlled in one oxidation step. For the larger mesas, higher

oxidation rates are thus required. The oxidation rates of bottom and top

DBR are adjusted by calibrating the Ga fraction of the respective Al(Ga)As

layers, as described in sect. 3.3.1.3. For well-calibrated oxidation rates, top

and bottom DBR oxidation is completed when the target oxidation length of

the aperture is reached.

The grey-shaded areas in Fig. 56 depict the different doping levels of the

contact layers. The 200 nm thick n-contact layer below the active medium

is doped at n = 1 × 1018 cm-3; the 300 nm thick p-contact layer above the

aperture is doped at p = 1 × 1018 cm-3 (medium grey). The topmost 50 nm of

the p-contact layer (dark grey) are doped at 3 × 1018 cm-3, however, to ob-

tain better p contacts. The differential electrical resistance of the VCSELs

critically depends on the doping level of the aperture. The aperture is

strongly doped with p = 3 × 1018 cm-3. For a lower series resistance of the

cavity, the 150 nm GaAs immediately below the aperture are additionally p-

doped at a level of p = 1 × 1017 cm-3.

1.01.52.02.53.03.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

aperture

p-contact

etch stop

3 QD groups

z (µm)

Optical field intensity (arb. u.)

Refractive index

Fig. 56: Vertical refractive-index profile of the oxide-DBR quantum-dot VCSEL.

The active region consists of 3 stacks of 3 QD layers each. The bold line denotes

the refractive index, the thin line depicts the intensity of the optical field. The cav-

ity is cladded by the bottom oxide DBR (left side, 7 pairs) and top oxide DBR

(right side, 6 pairs) and contains the n-contact, the active region, the aperture, the

p contact and an AlGaAs p-contact etch stop. The contact layers are marked by

grey-shaded regions (cf. text).

6.2. Quantum-dot vertical-cavity surface emitters

107

The layer thickness homogeneity across the as-grown wafer was 0.3 %,

excluding 8 mm of the wafer edge. Whereas the layer thickness homogene-

ity of the as-grown VCSEL structure was excellent, large inhomogeneities

were caused during the wet-thermal oxidation. Oxidation rates are very

temperature-sensitive, and the oxidation temperature strongly varied across

the susceptor of the used setup. Nevertheless, all of the 200 characterized

VCSELs with aperture diameters smaller than the top DBRs showed lasing.

The unusually high yield is ascribed to the large tolerance of the wide QD

gain spectrum to variations of the optical cavity thickness. The VCSEL

structure was grown with 7 bottom and 6 top DBR pairs. From some wafer

pieces, one and two top DBR pairs were selectively removed during the

process, so that VCSELs with 6, 5 and 4 top DBR pairs could be character-

ized.

Fig. 57 shows the L-I-curve of a 3.5 µm aperture VCSEL with 4 top

DBR pairs. The maximum output power of this device is 0.68 mW. Lasing

sets on at 280 µA at a bias voltage of 2.5 V. The maximum differential effi-

ciency is 43 %. The inset of Fig. 57 shows a spectrum of a VCSEL with a

3.5-µm aperture and 5 top DBR pairs. Single-mode operation is observed

for this aperture diameter with a side-mode suppression ratio of 35 dB. A

1080 1090 1100 1110

-40

-20

01234

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Output power (mW)

Drive current (mA)

4 mirror pairs

3.5 µm aperture

Ith = 280 µA

I = 4.6 x Ith

5 mirror pairs

3.5 µm aperture

Ith = 207 µ

µµ

µA

SSR = 35

Intensity (dB)

Wavelength (nm)

Fig. 57: L-I curve of a 3.5 µm aperture VCSEL with 4 top-mirror pairs. The

threshold current is 280 µA at a bias voltage of 2.5 V, the maximum slope effi-

ciency is 43 %, and the maximum output power amounts to 0.68 mW. Inset:

Spectrum of a QD VCSEL with 5 top DBR pairs and a 3.5 µm aperture. Since the

aperture is placed in a node of the optical wave, multimode operation can be ob-

served even for such small apertures. The side-mode suppression ratio (SSR) is

35 dB.

6. Quantum-dot lasers

108

lower maximum output power of 0.14 mW was observed for this device,

owing to the larger top mirror reflectivity.

Threshold currents and threshold current densities of QD VCSELs with 5

top DBR pairs are plotted in Fig. 58 as a function of aperture diameter. The

threshold currents are lowest for the smallest apertures. The behavior of the

threshold current densities is opposite, however; threshold current densities

are lowest for the largest apertures and increase significantly with decreas-

ing aperture diameter. The increase of the threshold current with increasing

aperture size is expected since more and more QDs are pumped. If the lat-

eral extension of the pumped gain medium coincided with the aperture area

for all aperture diameters, a threshold current density independent from the

aperture size would result. The deviation from this ideal behavior is on one

side due to current spreading to regions below the aperture oxide, and on

the other side due to current crowding near the aperture edge. Thus the in-

jection current density in the middle of a VCSEL decreases significantly

with increasing aperture size.

The aperture is placed in a node of the optical wave. This means that the

lateral spreading of the optical wave is significant for small apertures. Cur-

rent spreading is likewise considerable for small apertures. It is estimated

that the inverted gain medium of VCSELs with 1.5 µm wide apertures is at

024681012

0.2

0.4

0.6

0.8

1.0 Ith

(mA)

Aperture diameter (µm)

Jth

(kA/cm

)

Fig. 58: Threshold current densities of QD VCSELs with 5 top DBR pairs as a

function of aperture diameter. For small apertures, current spreading is significant.

The threshold current density therefore increases with decreasing aperture diame-

ter.

6.2. Quantum-dot vertical-cavity surface emitters

109

least 3 µm. As a consequence of these two effects, the lateral overlap of the

pumped gain medium with the optical wave is best for the smallest aper-

tures. The differential efficiency is therefore highest for the smallest aper-

tures. For an aperture diameter of 3.5 µm, the maximum differential effi-

ciency amounts to 9.8 %. Due to the increasing importance of current

crowding for larger apertures, VCSELs with 11.5 µm wide apertures show

differential efficiencies below 2 %.

111

7. Summary and outlook

Within the present work, metalorganic chemical vapor deposition

(MOCVD) of novel GaAs-based semiconductor laser structures based on

self-organized InGaAs quantum dots (QDs) grown in the Stranski-

Krastanow mode was advanced with regard to device performance. The

characteristics of such devices had not been close to the earlier predicted

theoretical limits before. The characteristics of laser diodes processed from

QD structures grown within this work have approached these limits consid-

erably.

The previously established QD growth technology was complemented by

an in-situ annealing step that leads to a reduction of defects and to a

smoothening of the growth front which is roughened by the deposition of

three-dimensional islands. Lasers based on such QDs were demonstrated

and exhibit transparency current densities around 6 A/cm2 per QD sheet for

emission wavelengths between 1.1 and 1.2 µm at room temperature. Using

in-situ annealing, the internal quantum efficiency was increased from pre-

viously 50 % to above 90 %. The flattening of the growth front allows to

reduce the distance between vertically stacked QD layers and thus helps to

increase the QD volume density. In the early stages of QD device research,

concern was raised about the reliability of QD lasers due to their highly

strained active regions. However, lasers based on 6-fold stacks of in-situ

annealed QDs as active medium show room-temperature peak output pow-

ers of 11.7 W in quasi-CW mode and 4.7 W under CW operation at emis-

sion wavelengths around 1.14 µm. This was the first demonstration of opti-

cal output powers of more than 10 W for QD lasers. It was found during

lifetime measurements at 50°C and output powers of 1.0 -1.5 W that the

device characteristics remain unchanged for more than 3000 h. This sup-

plies conclusive evidence of the high reliability of QD lasers.

In MOCVD of III-V semiconductors, strongly toxic hydrides such as

AsH3 are typically used as group-V precursors. In the course of this work,

MOCVD was changed over to alternative precursors. The organic-

compound precursor tertiarybutylarsine (TBAs) was used as substitute for

AsH3. Since the physical and chemical properties of TBAs are different

from those of AsH3, epitaxy of all layer structures needed for the growth of

QD lasers was to be recalibrated. The worldwide first QD laser grown using

TBAs was demonstrated. This is a major breakthrough in the development

of a more environment-friendly epitaxy method.

Using self-organized QDs, the emission wavelength of GaAs-based la-

sers can be tuned to the 1.3 µm wavelength regime which is important for

datacom applications. In contrast to 1.3 µm InGaAsN/GaAs quantum well

(QW) lasers, QD lasers exhibit much lower threshold currents and larger

7. Summary and outlook

112

quantum efficiencies. Different techniques to grow 1.3 µm QDs using

MOCVD were experimentally explored and evaluated in this work. These

included the overgrowth of InGaAs stressors with gallium-rich InGaAs

QWs and the admixture of nitrogen and antimony to InGaAs QDs. Un-

doped test structures containing such QD layers were investigated by photo-

luminescence spectra and transmission electron microscopy. Using the

overgrowth concept, lasing at 1.24 µm was achieved. This is the first dem-

onstration of an MOCVD-grown QD laser emitting beyond 1.2 µm at room

temperature. The growth of laser structures for the fabrication of QD verti-

cal-cavity surface-emitting lasers (VCSELs) with all-oxide distributed

Bragg reflectors (DBRs) and nine-fold stack of InGaAs/GaAs QDs as ac-

tive region was implemented. VCSELs with a room-temperature emission

at 1.1 µm were processed from these structures. A VCSEL with a 3.5 µm

aperture and four top DBR pairs exhibits a maximum output power of

0.68 mW, a differential efficiency of 43 % at a threshold current of 280 µA

at 2.5 V.

The demonstration of MOCVD-grown QD-based all-oxide DBR

VCSELs on GaAs substrates as well as edge-emitting QD lasers emitting

close to 1.3 µm are major advances towards the commercialization of

1.3 µm QD VCSELs. MOCVD is an important fabrication standard of op-

toelectronic devices since it is cost-efficient and scalable. It is thus well

suited for the commercial growth of such VCSEL structures. However,

there are still some tasks to fulfill before such devices can be launched:

First of all, the emission wavelength must be extended to the dispersion

minimum of glass fibers at 1.31 µm. The current 1.24 µm laser is based on

InGaAs stressor QDs overgrown with gallium-rich QWs. InGaAs stressor

QDs are preferred in MOCVD since they are more stable with regard to de-

fect formation than binary InAs45, 142. The redshift upon overgrowth with

Ga-rich QWs can be larger, however, if binary InAs stressors are used,

since these exhibit larger strain, favoring the alloy decomposition within the

overgrown QW. Using InAs stressors, probably larger effective QD sizes

will be achieved in such structures49. In contrast to ternary InGaAs QDs, the

InAs QDs grown with alternative precursors presented in this work have

very low aspect ratios and exhibit a large size distribution. The increase of

the aspect ratio of such stressors and the reduction of the QD size distribu-

tion will allow to grow QDs with longer emission wavelengths and lower

defect densities.

Alternatively to gallium-rich InGaAs QWs as used to overgrow InAs or

InGaAs stressor QDs, the use of low-band-gap dilute nitrides such as

GaAsN or InGaAsN will allow to achieve longer emission wavelengths

than with InGaAs QWs. If dilute nitride QWs are used, 1.3 µm room-

temperature lasing might likewise be achieved with ternary InGaAs stress-

ors. QD structures containing dilute nitrides may suffer, however, from ni-

7. Summary and outlook

113

tride-related defects213 which might require special post-growth annealing

techniques210. The extension of the QD lasing wavelength to even larger

values than 1.3 µm is essentially driven by the demand of lasers for long-

haul optical datacom, taking place at the attenuation minimum of conven-

tional glass fibers at 1.55 µm. At present, dilute nitrides are indispensable to

achieve lasing at this wavelength on GaAs substrates. Luminescence studies

of nitride-containing QD heterostructures at 1.55 µm are presently carried

out using MBE236.

Self-organized long-wavelength QDs are very large and highly strained.

In sheets of such large QDs, dislocated defect clusters are often formed.

This requires an appropriate defect reduction technique if the density of

such clusters is high. The annealing step reported in this work has been de-

veloped for defect reduction in QD sheets that are free of dislocated clus-

ters. Defect reduction in long-wavelength QD sheets must additionally be

capable of cluster elimination, similar to annealing procedures reported for

long-wavelength QD growth using MBE237. Here, the thickness of the cap

deposited previous to the annealing step is chosen only to cover coherent

QDs. Clusters are larger than coherent QDs and emerge from the surface of

the thin cap. During a high-temperature growth interruption (GRI), such

clusters can be evaporated237. The holes that are left behind in the thin GaAs

cap are filled with either GaAs or AlAs. In addition to GRIs at high tem-

peratures in MOVCD, in-situ etching using tertiarybutylchloride238, 239

(TBCl) might help to remove clusters and to reestablish smooth surfaces.

If large, long-wavelength QDs are densely stacked, the vertical strain

fields of successive QD sheets can overlap, depending on the thickness of

the spacer layer. The strain field of an underlying QD layer can influence

the formation of QDs in a subsequent layer. This has two disadvantages:

First, vertical correlation can result (cf. Fig. 21), potentially leading to lar-

ger QDs than in the preceding layer; secondly, the danger of cluster forma-

tion might be enhanced due to accumulated strain from buried QD sheets.

To avoid these two effects in highly strained QD stacks, strain-

compensating Ga(PAs) layers might be useful. GaPAs has a smaller lattice

constant than GaAs, compressive strain from indium-containing layers can

be compensated by the tensile strain of GaPAs. GaPAs layers have been

used to for strain compensation in edge-emitting InGaAs(N)/GaAs-QW

edge emitters197, 240 and InGaAs(N)-QW VCSELs241, 242 with comparably

low threshold currents.

115

List of acronyms

Chemical elements

Al aluminium

Ar argon

As arsenic

C carbon

Ga gallium

Ge germanium

H hydrogen

In indium

N nitrogen

O oxygen

P phosphorus

Sb antimony

Si silicon

Te tellurium

Precursors

AsH3 arsine / arsenic hydride / arsenic trihydride

CBr4 carbon tetrabromide

DETe diethyltelluride

DMHy (unsymmetric) dimethylhydrazine

DMZn dimethylzinc

NH3 ammonia

TBAs tertiarybutylarsine

TBCl tertiarybutylchloride

TEAs triethylarsine

TEGa triethylgallium

TESb triethylantimony

TMAl trimethylaluminium

TMAs trimethylarsine

TMGa trimethylgallium

TMIn trimethylindium

UDMHy unsymmetric dimethylhydrazine

Abbreviations

AAPS activated alloy phase separation

AFM atomic force microscopy

AR anti reflection

COMD catastrophic optical mirror damage

List of acronyms

116

CW continuous-wave

DBR distributed Bragg reflector

DHS double heterostructure

GRI growth interruption

HR high reflection

L-I light-current

MBE molecular beam epitaxy

MFC mass-flow controller

ML monolayer

MOCVD metalorganic chemical vapor deposition

PL photoluminescence

PLE photoluminescence excitation (spectroscopy)

QD(s) quantum dot(s)

QW(s) quantum well(s)

RHEED reflection high-energy electron diffraction

RIE reactive-ion etching

SCH separate-confinement heterostructure

SI semi-insulating

SK Stranski-Krastanow

slm standard (T = 0°C, p = 1.013 bar) liters per minute

STM scanning tunneling microscopy

TEM transmission electron microscopy

VCSEL vertical-cavity surface-emitting laser

WL wetting layer

XSTM cross-section scanning tunneling microscopy

Physical quantities

E energy (J)

F free energy (J)

H free enthalpy (J)

j current density (A/cm2)

p pressure (bar, Pa, Torr)xiii

S entropy (J/K)

surface energy (J/cm2)

t time (s)

T temperature (K or °C)

Physical constant

kB Boltzmann´s constant (1.3806503·10-23 J/K)

xiii 1 bar = 105 Pa ≈ 750 Torr

117

List of figures

Fig. 1: Schematic diagram of the cross-sectional view on the laser facet of a

typical, fully processed edge-emitting laser diode. The different layers

are: 1. Ti/Pt/Au top contact. 2. SiNx insulating layer. 3. p++ GaAs

contact layer. 4. p+ AlGaAs top cladding. 5. Undoped optical-

confinement GaAs layer with the active region in the center. 6. n+

AlGaAs bottom cladding. 7. Substrate. 8. Ni/AuGe bottom contact..... 4

Fig. 2: Layer and structure design of a GaAs-based full-oxide-DBR VCSEL

with intracavity p- (top) and n-contacts (bottom). The active region as

shown schematically in the magnification can alternatively consist of a

multi QD layer stack or QWs. After Ref. 14. ........................................ 5

Fig. 3: Lowest threshold current densities reported for double-

heterostructure (DHS) lasers, QW lasers, and QD lasers, respectively,

versus publication year. After Ref. 28. .................................................. 8

Fig. 4: Simplified schematic of the MOCVD setup of both the Aix200 and

Aix200/4 machine. The two kits basically differ by the size of the

quartz-glass reactor. The Aix200 disposes of a hydride lines for arsine

instead of the TBAs-bubbler line. Dopant lines and pressure controllers

are not shown. ...................................................................................... 16

Fig. 5: Electron concentration of AlxGa1-xAs:Te and AlxGa1-xAs:Si as

functions of the aluminium fraction x, determined by Hall

measurements at room temperature. The data for AlxGa1-xAs:Te were

taken from Ref. 87, the data for AlxGa1-xAs:Si are from Ref. 86. ....... 19

Fig. 6: Schematic diagram of the principal cracking mechanisms of arsine

and tertiarybutylarsine (TBAs) during pyrolysis in MOCVD.

Homolytic fission and β-elimination are competing decomposition

mechanisms of TBAs. Homolytic fission produces a reactive AsH2

radical. β-elimination generates an arsine molecule and an inert

isobutene molecule............................................................................... 20

Fig. 7: Comparison of thermal decomposition of TBAs and arsine in an

atmospheric pressure reactor. After Stringfellow et al.96..................... 22

Fig. 8: AFM images (3 × 3 µm2) of GaAs substrate pieces heated up to

700°C and cooled down under different TBAs partial pressures (A:

2 × 10-2 mbar, B: 2 × 10-3 mbar, C: no stabilization)........................... 23

Fig. 9: Hole concentration of 1 µm thick, nominally undoped Al0.4Ga0.6As

layers as a function of V/III ratio. The hole concentration was

measured by the van-der-Pauw method............................................... 23

Fig. 10: GaAs and AlGaAs growth rates as determined from cross-sectional

AFM images, plotted as a function of growth temperature. The sample

consisted of alternating AlxGa1-xAs/GaAs layers grown at different

temperatures, using TMGa, TMAl and AsH3 as precursors. The TMGa

List of figures

118

and TMAl flows were 7.4 and 5.8 µmol/min, respectively; the V/III

ratio was 180 for all layers. Error bars originate from statistical

treatment of 10 thickness measurements of each layer; dashed lines are

guides to the eye. An Al fraction of x = 65.1 % was determined from

an x-ray diffraction spectra of a 1 µm thick AlGaAs epilayer grown at

720°C with the given precursor flows. The Al fraction may slightly

change with decreasing growth temperature........................................ 27

Fig. 11: Reflection spectrum of an unoxidized, as-grown VCSEL structure.

The DBRs consist of six periods of GaAs / Al0.98Ga0.02As.................. 28

Fig. 12: Strongly simplified gas inlet schematic of the Aix200/4 reactor

with susceptor and substrate. Group-V precursor gases are conducted

separately from group-III precursors. In order to avoid prereactions,

they are not mixed before they reach the entrance of the reactor.

Dopants are added to the group-III line. .............................................. 29

Fig. 13: Thickness of the AlAs layers of AlAs/GaAs superlattices,

normalized to 1.0 for a radial distance of 0.5 inch = 12.7 mm. For the

sake of clearness, the values for GaAs were omitted. The behavior of

GaAs layer thicknesses is analogous to those of AlAs. The AlAs layers

were grown with a V/III of 30 at a growth rate of 2 µm/h. ................. 30

Fig. 14: Al1-xGaxAs oxidation rate at 420°C versus composition for 100-nm-

thick layers cladded by 100 nm thick GaAs layers. From Ref. 109. ... 31

Fig. 15: Typical structure of an undoped sample containing one or more QD

layers for PL measurements. AlGaAs diffusion barriers below and

above the QDs suppress carrier diffusion to the substrate and to the

surface, respectively............................................................................. 32

Fig. 16: Free energy gained by the 2D to 3D transition of a

pseudomorphically strained WL as a function of island base length.

Curves for different values of the paramter

(cf. text) are plotted.

From Ref. 119. ..................................................................................... 38

Fig. 17: Strain energy around a spontaneously formed island in a one-

dimensional atomistic model by Barabási126. Es is the strain energy of

an atom placed on the top of the substrate or on the island. Es is largest

at the island edge. From Ref. 126. ....................................................... 40

Fig. 18: Monte-Carlo simulations of the temporal evolution of 2D

submonolayer islands. A coverage of 4 % was deposited randomly on

the surface at a flux of 1 ML/s. Every 0.01s, a histogram of the island

size distribution is recorded. The average island diameter 〈N〉 is

plotted as a function of GRI duration for different temperatures. The

hopping probability p of an adatom is proportional to

exp[-(E0 + nEn - Es) / kBT] where E0 is the adatom diffusion barrier, En

the chemical bond energy to a neighbor atom, n the number of

chemically bound neighbors, and Es the strain energy at the position of

the considered adatom. From Ref. 127. ............................................... 41

List of figures

119

Fig. 19: (a) Reference sample without QDs. A 30 nm thick GaAs layer was

directly deposited at 490°C on 100 nm GaAs grown at 600°C. (b) AFM

image of a 30 nm thick QD-burying GaAs cap deposited at 490°C on

QDs that were also grown at 490°C. The hillock density amounts to

2.5 × 109 cm-2. The picture has been published in Ref. 73 (c) Plan-view

TEM image of a single-sheet QD sample. The QD density is about

3.4 × 1010 cm-2. The underlying QDs were deposited under the same

parameters as those of sample (b)........................................................ 48

Fig. 20: (a) AFM surface image of a sample where the growth temperature

of the 30 nm thick GaAs cap was ramped in-situ to 590°C. The picture

has been published in Ref. 73. (b) AFM surface image of a sample

where the QDs were capped with 7 nm GaAs. Subsequently, the

surface was annealed during 10 min at 600°C under As stabilization. 50

Fig. 21: Cross-sectional dark-field TEM image of a five-fold In0.8Ga0.2As

QD stack grown with in-situ annealing after each QD layer was capped

by 7 nm GaAs. The vertical sheet-to-sheet distance is 18-20 nm........ 51

Fig. 22: Room-temperature PL spectra of annealed QD samples at low

excitation density (a) (5 W/cm2) and at high excitation density (b)

(5 kW/cm2). In (a), recombination occurs essentially on the QD ground

state. Luminescence from the WL and the matrix is not visible. In (b),

radiative recombination mainly takes place in the GaAs matrix. The

stars denote transitions from the excited QD states. This figure has

been published in Ref. 73..................................................................... 53

Fig. 23: PL spectra of In0.67Ga0.33As QDs, deposited at 485°C. “QD”,

“QD*”, and “WL” denote transitions from the QD ground state, the

first excited QD state, and the WL, respectively. Inset: Plan-view dark-

field transmission electron micrograph of the same QD layer. This

figure has been published in Ref. 161.................................................. 55

Fig. 24: PL peak emission wavelengths of the ground-state transitions of

In0.67Ga0.33As QDs as a function of growth temperature and V/III ratio.

QDs of the V/III series were deposited at 485°C, the V/III ratio of the

temperature series was kept at 1.5. The reproducibility of the peak

positions is within 10 nm. This figure has been published in Ref. 161.

.............................................................................................................. 56

Fig. 25: Flow chart of different approaches to achieve lasing emission at

1.3 µm and beyond by overgrowing small In(Ga)As QDs. Structures in

bold scripture were grown and characterized within this work. .......... 58

Fig. 26: Low-temperature PL spectra of single In0.8Ga0.2As QD layers

overgrown with 5 nm InxGa1-xAs QWs of varying indium fraction x.

The spectra were normalized to equal peak intensities........................ 59

Fig. 27: Full squares: PL peak wavelength of In0.8Ga0.2As QDs overgrown

with 5 nm InxGa1-xAs of denoted indium fraction x, recorded at 9 K.

From TEM images, the QDs are estimated to be 3-5 nm high and 20-

List of figures

120

25 nm wide. Hollow squares: Ground-state transition wavelengths of a

2.6 nm high truncated-pyramid-shaped InAs QD overgrown with an

InxGa1-xAs QW of denoted indium fraction x, calculated within an

eight-band k·p framework. ................................................................... 60

Fig. 28: Room-temperature PL spectra of In0.8Ga0.2As QDs overgrown with

In0.25Ga0.75As QWs of varying thickness d. ......................................... 62

Fig. 29: (a)-(c) Fourier-filtered cross-sectional high-resolution TEM images

of In0.8Ga0.2As QDs covered with (a) 1 nm, (b) 2 nm, and (c) 3 nm

In0.25Ga0.75As. The dotted white lines depict the boundary between

InGaAs and GaAs. The boundaries were determined after the reversal

of the In-As and Ga-As image contrast, occurring at an indium fraction

of about 15 %. (d) Heights of the QDs and the QW near the QD bases.

A generous error of 1 nm is assumed for each value, taking

uncertainties of the determination of the InGaAs/GaAs boundary into

account. ................................................................................................ 63

Fig. 30: Schematic diagrams illustrating the separation of a ternary InGaAs

alloy into indium-rich and gallium-rich phases, activated by the strain-

relaxed surface of an underlying In(Ga)As stressor QD. (a) Initial

In(Ga)As stressor QD, assumed to have the shape of a truncated

pyramid. (b) Partial decomposition of the InGaAs alloy into indium-

rich and gallium-rich regions. (c) Final structure, buried by GaAs..... 64

Fig. 31: Cross-sectional dark-field TEM image of QDs formed upon MBE

of 2 ML InAs at 485°C, overgrown with 5 nm In0.15Ga0.85As. From

Ref. 169................................................................................................ 65

Fig. 32: Empty-state XSTM image taken at VS = +2.1 V. The contours of

the dot are indicated by dotted lines, and those of its In-rich zone by

dashed lines. The In distribution was evaluated along the intersection

lines (a) and (b) by evaluation of atom-chain distances (cf. text). After

Lenz et al.179......................................................................................... 66

Fig. 33: Room-temperature PL spectra of InGaAs QDs grown with (a)

AsH3 (from Ref. 142) and (b) TBAs, deposited around 500°C. The

respective V/III ratios are marked in the viewgraphs. ......................... 68

Fig. 34: Flow chart of different approaches to achieve QD emission at

1.3 µm by addition of antimony and nitrogen to In(Ga)As QDs or

InGaAs QWs. ....................................................................................... 69

Fig. 35: Band gap energy versus lattice parameter of GaAsN, InGaAs, and

In0.53Ga0.47As1-yNy. From Ref. 205....................................................... 70

Fig. 36: (a) Free energy F versus solid composition for a hypothetical

semiconductor alloy with a large positive mixing enthalpy H. The

points labelled A and B are the binodal points, the inflection points C

and D are the spinodal points. The miscibility gap is the composition

range between xC and xD. (b) The fabrication of InxGa1-xAsyN1-y can be

achieved by separately mixing Ga and In on the group-III sublattice

List of figures

121

with parameter x, and As and N on the group-V sublattice with

parameter y........................................................................................... 71

Fig. 37: (a) Room-temperature low-excitation spectra of undoped samples

with single InGaAsN QD sheets. (b) PL peak intensities of the QD

ground-state (GS) PL, plotted as a function of pDMHy/pTBAs. A power

density of 5 kW/cm2 was used for excitation. The inset shows the

spectra belonging to the first data point at pDMHy/pTBAs = 0................. 73

Fig. 38: PL spectra of InGaAsN QD samples. The InGaAs QD samples

were nitrided by adjusting the denoted DMHy partial pressures in the

reactor chamber during a GRI of 1 min after deposition of the InGaAs

QD material. (a) Low-excitation room-temperature PL spectra. The

spectra of the nitrided samples are smoothed. (b) High-excitation low-

temperature spectra of the same samples, logarithmically plotted. ..... 75

Fig. 39: Room-temperature PL spectra of (a) InAs QDs and (b) InAsSb QD

sheets. In both cases, the deposition amount was varied over the same

range. Both InAs and InAsSb QDs were grown at Tgr = 485°C. A GRI

of 5 s was introduced after deposition of the QD material to allow QD

formation. The InAsSb QDs were deposited with an additional TESb

flow of 4.4 µmol/min during deposition and the subsequent GRI. The

excitation power density was 5 W/cm2................................................ 77

Fig. 40: Low-temperature PLE contour plot of an InAs sample with a TESb

flow of 8.8 µmol/min during the GRI of 5 s, subsequent to the InAs

deposition (compare to Fig. 43). PL was excited using a white-light

source and a double monochromator. The excitation density is below

5 mW/cm2. The intensity is plotted on a logarithmic scale, the thin

lines are equi-intensity lines. The bold black curve shows a PL

spectrum, excited at 514.5 nm and an excitation power density of

5 W/cm2. The logarithmic scale of the PL intensity is given on the

right. The peaks of the PL spectrum are partially numbered............... 79

Fig. 41: High-resolution cross-section transmission micrograph of a single

InAs QDs sheet grown at 500°C. A GRI of 5 s under TESb flux of

8.8 µmol/min was performed after the deposition of the InAs. The

lower InAs/GaAs interface as well as the top interfaces of the

truncated-pyramidal InAs islands are depicted with thin white lines.

The boundaries are determined according to the reversal of the group-

III/group-V-atom contrast, occurring at an indium fraction x of about

15 % for InxGa1-xAs. ............................................................................ 80

Fig. 42: Transition energies of truncated-pyramidal InAs islands as a

function of island height, calculated using 8-band k·p-theory. The InAs

islands have constant base length of 13.6 nm and 45° facets. With

increasing island height, the shape changes from a truncated pyramid

towards a complete pyramid. The aspect ratio of a complete pyramid

with 45° facet angle is 0.5.................................................................... 81

List of figures

122

Fig. 43: Low-temperature low-excitation PL spectra of undoped test

structures, each containing a single QD sheet. 1.7 ML InAs were

deposited at 485°C with a V/III ratio of 1.5. During the subsequent

GRI of 5s, different precursors (TBAs and TESb) were switched on.

The impact of TBAs supply is shown in (a), the influence of TESb

during the GRI is shown in (b). The rightmost curves in (a) and (b) are

reference InAs samples, grown without any precursor flow during the

GRI. They are slightly different in (a) and (b) since they stem from

different growth series. The nominal deposition parameters are the

same, however...................................................................................... 82

Fig. 44: Output power (two facets) of a QD laser (10 min annealing step)

versus injection current (pulsed). Inset: electroluminescence of a laser-

structure at 0.9 × jth (1) and 1.01 × jth (2). The stripe width of the device

is 200 µm, the cavity length is 1.3 mm. The threshold current density

and slope efficiency are 110 A/cm2 and 85 %, respectively................ 88

Fig. 45: Threshold current density versus optical loss for devices with

10 min and 30 min annealing and stripe-width of 75 µm. Triangles

show data for a laser with QDs annealed during 10 min. Filled and

empty triangles are data from nominally the same laser structure but

from different growth runs. The stripe width is 100 µm for these lasers.

Straight lines are exponential fits. The stated errors are from the

numeric regression fit only, a general error of 10 % for the

measurements can additionally be assumed, however. Inset: L-I-

characteristics of devices grown with annealing durations of tA = 0 min

(jth = 340 A/cm2), 10 min (jth = 130 A/cm2), and 30 min

(jth = 180 A/cm2). Stripe widths are 75 µm, cavity lengths are 1.4 mm

(tA = 0 min) and 1.3 mm (tA = 10, 30 min). This figure has been

published in Ref. 156. .......................................................................... 89

Fig. 46: Inverse differential quantum efficiency as a function of cavity

length for different annealing durations. Dashed lines are linear fits.

Stripe widths are 75 µm. An estimated error of about 10 % can be

assumed. This figure has been published in Ref. 156.......................... 90

Fig. 47: Front facet output power of two different devices, driven in CW

(A) and quasi-CW mode (B), respectively. The stripe geometry is

2 mm × 150 µm. The heat sink temperature was stabilized at 20°C.

Device (A) shows CW operation up to 4.7 W. Device (B) was driven

up to 11.7 W in quasi-CW mode with 50 µs pulses and 50 Hz

repetition frequency. This figure has been published in Ref. 157. ...... 91

Fig. 48: Lifetime measurements of six 2 mm long and 150 µm wide laser

stripes, driven in CW mode at a heat sink temperature of 50°C at two

different drive currents. The measurement was interrupted after 910

hours. The depiction of the temporal offset between both

measurements is arbitrary. The inset shows lasing spectra near

List of figures

123

threshold at 0.6 A, and at 2.5 A. This figure has been published in Ref.

157........................................................................................................ 92

Fig. 49: (a) Schematic diagram of a double-barrier DHS QW injection laser.

(b) Temperature dependence of the threshold current of a QW laser

with and without double-barrier design. .............................................. 94

Fig. 50: Electroluminescence spectra of a 100 µm × 2 mm ridge-waveguide

laser with a threshold current density of 60 A/cm2, driven in pulsed

mode (500 ns, 5 kHz). The laser structure was grown with alternative

precursors only. Laser spectra were recorded with monochromator slit

widths reduced from 1 to 0.1 mm, leading to a signal decrease by about

100. Peaks at 1.01 µm stem from the laboratory lighting. Dotted line:

PL of a single QD layer excited with 5 W/cm2 at 514.5 nm. This figure

has been published in Ref. 161. ........................................................... 97

Fig. 51: Threshold current densities of 100 µm wide ridge-waveguide lasers

as a function of inverse cavity length. The dashed line shows a linear

fit of the decadal logarithm of jth. Inset: inverse differential quantum

efficiency

d as a function of cavity length. The devices were driven

with 500 ns long current pulses at a repetition frequency of 5 kHz. This

figure has been published in Ref. 161.................................................. 98

Fig. 52: Threshold current vs. inverse cavity length of a laser diodes based

on a 10-fold stack of In0.65Ga0.35As/In0.2Ga0.8As/GaAs QD sheets. The

stripe width is 50 µm. The inset shows a lasing spectrum of a 2.3 mm

long and 50 µm wide device at room temperature. This figure has been

published in Ref. 15. ............................................................................ 99

Fig. 53: Schematic overview of the two-step oxidation process used for the

fabrication of single-QW VCSELs with oxide aperture and all-oxide

DBRs (cf. text). After Ref. {Hopfer, #947}...................................... 102

Fig. 54: (a) L-I-characteristics and (b) spectra below and above lasing

threshold of a QW VCSEL with an aperture diameter of 25 µm. The

threshold current density is 400 A/cm2.............................................. 103

Fig. 55: (a) Reflection spectra and EL spectra of a single-InGaAs-QW

oxide-DBR VCSEL with a 25 µm aperture. The reflection shows a

weak cavity dip. (b) Microscopic image of an oxide-DBR QW VCSEL

with a 25 µm aperture. (c) Nearfield distribution of the same device at

an averaged output power of 13 µW (duty cycle: 0.8 %). The nearfield

was recorded using a Si-CCD camera having a rather low sensitivity at

1050 nm.............................................................................................. 104

Fig. 56: Vertical refractive-index profile of the oxide-DBR quantum-dot

VCSEL. The active region consists of 3 stacks of 3 QD layers each.

The bold line denotes the refractive index, the thin line depicts the

intensity of the optical field. The cavity is cladded by the bottom oxide

DBR (left side, 7 pairs) and top oxide DBR (right side, 6 pairs) and

contains the n-contact, the active region, the aperture, the p contact and

List of figures

124

an AlGaAs p-contact etch stop. The contact layers are marked by grey-

shaded regions (cf. text)..................................................................... 106

Fig. 57: L-I curve of a 3.5 µm aperture VCSEL with 4 top-mirror pairs. The

threshold current is 280 µA at a bias voltage of 2.5 V, the maximum

slope efficiency is 43 %, and the maximum output power amounts to

0.68 mW. Inset: Spectrum of a QD VCSEL with 5 top DBR pairs and a

3.5 µm aperture. Since the aperture is placed in a node of the optical

wave, multimode operation can be observed even for such small

apertures. The side-mode suppression ratio (SSR) is 35 dB.............. 107

Fig. 58: Threshold current densities of QD VCSELs with 5 top DBR pairs

as a function of aperture diameter. For small apertures, current

spreading is significant. The threshold current density therefore

increases with decreasing aperture diameter...................................... 108

125

Bibliography

1 R. V. Steele, Semiconductor Laser Market Review and Forecast, Proc.

of the Laser Focus World Marketplace Seminar, San Jose, California,

January 21 (2002).

2 D. Bimberg, M. Grundmann, and N. N. Ledentsov, Quantum Dot Het-

erostructures (John Wiley & Sons, Chichester, 1998).

3 Y. Arakawa and H. Sakaki, Multidimensional quantum well laser and

temperature dependence of its threshold current, Appl. Phys. Lett. 40,

939 (1982).

4 M. Asada, Y. Miyamoto, and Y. Suematsu, Gain and the threshold of

three dimensional quantum-box lasers, J. Quantum El. QE-22, 1915

(1986).

5 T. Takeuchi, Y.-L. Chang, A. Tandon, D. Bour, S. Corzine, R. Twist,

M. Tan, and H.-C. Luan, Low threshold 1.2 µm InGaAs quantum well

lasers grown under low As/III ratio, Appl. Phys. Lett. 80, 2445

(2002).

6 X. W. Liu, A. A. Hopgood, B. F. Usher, H. Wang, and N. S. J.

Braithwaite, Formation of misfit dislocations during growth of InxGa1-

xAs/GaAs strained-layer heterostructures, Semicond. Sci. Technol. 14,

1154 (1999).

7 H. E. Li and K. Iga (Ed.), Vertical-Cavity Surface-Emitting Laser De-

vices (Springer, 2002).

8 N. N. Ledentsov, D. Bimberg, V. M. Ustinov, Z. I. Alferov, and J. A.

Lott, Self-Organized InGaAs Quantum Dots for Advanced Applica-

tions in Optoelectronics, Jpn. J. Appl. Phys. (1) 41, 949 (2002).

9 D. Bimberg, Quantum Dots: Lasers and Amplifiers, Proc. of the 28th

International Symposium on Compound Semiconductors, Tokyo, Ja-

pan, 1-4 Oct (2001), p. 485.

10 D. Bimberg, M. Grundmann, F. Heinrichsdorff, N. N. Ledentsov, V.

M. Ustinov, A. E. Zhukov, A. R. Kovsh, M. V. Maximov, Y. M.

Shernyakov, B. V. Volovik, A. F. Tsatsul'nikov, P. S. Kop'ev, and Z.

I. Alferov, Quantum dot lasers: breakthrough in optoelectronics, Thin

Solid Films 376, 235 (2000).

11 M. Grundmann, The present status of quantum dot lasers, Physica E 5,

167 (2000).

12 F. Heinrichsdorff, C. Ribbat, M. Grundmann, and D. Bimberg, High-

power quantum-dot lasers at 1100 nm, Appl. Phys. Lett. 76, 556

(2000).

13 F. Heinrichsdorff, M.-H. Mao, N. Kirstaedter, A. Krost, D. Bimberg,

A. O. Kosogov, and P. Werner, Room-temperature continuous-wave

Bibliography

126

lasing from stacked InAs/GaAs quantum dots grown by metalorganic

chemical vapor deposition, Appl. Phys. Lett. 71, 22 (1997).

14 N. N. Ledentsov, Long-wavelength quantum-dot lasers on GaAs sub-

strates: from media to device concepts, J. Sel. Top. Quantum El. 8,

1015 (2002).

15 I. N. Kaiander, R. L. Sellin, T. Kettler, N. N. Ledentsov, and D. Bim-

berg, Long Wavelength (<1.24 µm) InGaAs/GaAs Quantum Dot Laser

Grown by Alternative Precursor MOCVD, subm. to Appl. Phys. Lett.

(2003).

16 J. von Neumann, Collected Works (Pergamon Press, Oxford, 1963),

Vol. 5.

17 R. N. Hall, G. E. Fenner, J. D. Kingsley, T. J. Soltys, and R. O. Carl-

son, Coherent Light Emission From GaAs Junctions, Phys. Rev. Lett.

9, 366 (1962).

18 N. Holonyak and S. F. Bevacqua, Coherent (visible) light emission

from Ga(As1-xPx) junctions, Appl. Phys. Lett. 1, 82 (1962).

19 M. I. Nathan, W. P. Dumke, G. Burns, F. H. Dill, and G. I. Lasher,

Stimulated emission of radiation from GaAs p-n junctions, Appl. Phys.

Lett. 1, 62 (1962).

20 Z. I. Alferov and R. F. Kazarinov, Semiconductor laser with electrical

pumping, Patent No. 181737 (1963).

21 H. Kroemer, Semiconductor laser with electrical pumping, Patent No.

3309553 (1967).

22 C. W. Wilmsen, H. Temkin, and L. A. Coldren (Ed.), Vertical-Cavity

Surface-Emitting Lasers (Cambridge University Press, 1999).

23 M. Grundmann, N. N. Ledentsov, O. Stier, J. Böhrer, D. Bimberg, V.

M. Ustinov, P. S. Kop'ev, and Z. I. Alferov, Nature of optical transi-

tions in self-organized InAs/GaAs quantum dots, Phys. Rev. B 53,

R10509 (1996).

24 M. Grundmann, J. Christen, N. N. Ledentsov, J. Böhrer, D. Bimberg,

S. S. Ruvimov, P. Werner, U. Richter, U. Gösele, J. Heydenreich, V.

M. Ustinov, A. Y. Egorov, A. E. Zhukov, P. S. Kop’ev, and Z. I. Al-

ferov, Ultranarrow Luminescence Lines from Single Quantum Dots,

Phys. Rev. Lett. 74, 4043 (1995).

25 O. Stier, Electronic and Optical Properties of Quantum Dots and

Wires in Berlin Studies in Solid State Physics, ed. by C. Thomsen, D.

Bimberg, M. Dähne, and W. Richter (Wissenschaft & Technik Verlag,

Berlin, 2001), Vol. 7.

26 Y. Miyamoto, Light emission from a quantum-box structure by cur-

rent injection, Jpn. J. Appl. Phys. (2) 26, L225 (1987).

27 N. Kirstaedter, N. N. Ledentsov, M. Grundmann, D. Bimberg, V. M.

Ustinov, S. S. Ruvimov, M. V. Maximov, P. S. Kop'ev, Z. I. Alferov,

Bibliography

127

U. Richter, P. Werner, U. Gösele, and J. Heydenreich, Low threshold,

large T0 injection laser emission from (InGa)As quantum dots, El.

Lett. 30, 1416 (1994).

28 N. N. Ledentsov, M. Grundmann, F. Heinrichsdorff, D. Bimberg, V.

M. Ustinov, A. E. Zhukov, M. V. Maximov, Z. I. Alferov, and J. A.

Lott, Quantum-dot heterostructure lasers, J. Sel. Top. Quantum El. 6,

439 (2000).

29 R. C. Miller, R. Dingle, A. C. Gossard, R. A. Logan, W. A. Nordland,

and W. Wiegman, Laser oscillation with optically pumped very thin

GaAs-AlxGa1-xAs multilayer structures and conventional double het-

erostructures, J. Appl. Phys. 47, 4509 (1976).

30 Z. I. Alferov, V. M. Andreev, E. L. Portnoi, and M. K. Trukan, AlAs-

GaAs heterojunction injection lasers with a low room-temperature

threshold, Fiz. Tverd. Tela 9, 279 (1969).

31 R. D. Dupuis, P. D. Dapkus, N. Holonyak, E. A. Rezek, and R. Chin,

Room-temperature laser operation of quantum-well Ga(1-x)AlxAs-GaAs

laser diodes grown by metalorganic chemical vapor deposition, Appl.

Phys. Lett. 32, 295 (1978).

32 N. N. Ledentsov, V. M. Ustinov, A. Y. Egorov, A. E. Zhukov, M. V.

Maximov, I. G. Tabatadze, and P. S. Kop'ev, Optical properties of

heterostructures with InGaAs-GaAs quantum clusters, Fiz. i Tekn. Po-

luprovodn. 28, 1484 (1993).

33 Z. I. Alferov, V. M. Andreev, D. Z. Garbuzov, Y. V. Zhilyaev, E. P.

Morozov, E. L. Portnoi, and V. G. Trofim, Effect of heterostructure

parameters on the laser threshold current and the realization of con-

tinuous generation at room temperature, Sov. Phys. Semicond. 4,

1573 (1970).

34 I. Hayashi, M. B. Panish, P. W. Foy, and S. Sumski, Junction lasers

which operate continuously at room temperature, Appl. Phys. Lett.

17, 109 (1970).

35 W. T. Tsang, Extremely low threshold (Al,Ga)As modified multi-

quantum well heterostructure lasers grown by molecular beam epi-

taxy, Appl. Phys. Lett. 39, 786 (1981).

36 W. T. Tsang, Extremely low threshold (Al,Ga)As graded-index

waveguide separate confinement heterostructure lasers grown by mo-

lecular beam epitaxy, Appl. Phys. Lett. 40, 217 (1982).

37 N. N. Ledentsov, M. Grundmann, N. Kirstaedter, O. Schmidt, R.

Heitz, J. Böhrer, D. Bimberg, V. M. Ustinov, V. A. Shchukin, A. Y.

Egorov, A. E. Zhukov, S. Zaitsev, P. S. Kop'ev, Z. I. Alferov, S. S.

Ruvimov, A. O. Kosogov, P. Werner, U. Gösele, and J. Heydenreich,

Ordered Arrays of Quantum Dots: Formation, Electronic Spectra, Re-

laxation Phenomena, Lasing, Sol. St. Electron. 40, 785 (1996).

Bibliography

128

38 Z. I. Alferov, A. M. Vasiliev, S. V. Ivanov, P. S. Kop'ev, N. N. Le-

dentsov, B. Y. Meltser, and V. M. Ustinov, Reduction of the threshold

current density (52 A/cm2, 300K) in SCH GaAs-AlGaAs lasers by us-

ing a quantum well confined by a variable-step short-period superlat-

tice, Sov. Phys. Techn. Phys. Lett. 14, 782 (1988).

39 Z. I. Alferov, S. V. Ivanov, P. S. Kop'ev, N. N. Ledentsov, M. E. Lut-

senko, M. I. Nemenov, B. Y. Meltser, V. M. Ustinov, and S. V.

Shaposhnikov, Spreading and surface recombination in quantum well

(Al,Ga)As double heterostructure separate confinement lasers with a

broad stripe, Sov. Phys. Semicond. 24, 92 (1990).

40 G. T. Liu, H. Li, K. J. Malloy, and L. F. Lester, Extremely low room-

temperature threshold current density diode lasers using InAs dots in

In0.15Ga0.85As quantum well, El. Lett. 35, 1163 (1999).

41 N. Chand, E. E. Becker, J. P. van der Ziel, S. N. G. Chu, and N. K.

Dutta, Excellent uniformity and very low (<50 A/cm2) threshold cur-

rent density strained InGaAs quantum well diode lasers on GaAs sub-

strate, Appl. Phys. Lett. 58, 1704 (1991).

42 P. G. Eliseev, H. Li, A. Stintz, G. T. Liu, T. C. Newell, K. J. Malloy,

and L. F. Lester, Transition dipole moment of InAs/InGaAs quantum

dots from experiments on ultralow-threshold laser diodes, Appl. Phys.

Lett. 77, 262 (2000).

43 M. Grundmann and D. Bimberg, Gain and Threshold of Quantum Dot

Lasers: Theory and Comparison to Experiments, Jpn. J. Appl. Phys.

(1) 36, 4181 (1997).

44 N. N. Ledentsov, A. E. Zhukov, A. R. Kovsh, N. A. Maleev, S. S.

Mikhrin, A. P. Vasile'ev, E. S. Semenova, M. V. Maximov, Y. M.

Shernyakov, N. V. Kryzhanovskaya, V. M. Ustinov, and D. Bimberg,

High performance quantum dot lasers on GaAs substrates operating

in the 1.5 µm range, El. Lett. 39, 1126 (2003).

45 A. Passaseo, G. Maruccio, M. D. Vittorio, R. Rinaldi, and R. Cingo-

lani, Wavelength control from 1.25 to 1.4 µm in InxGa1-xAs quantum

dot structures grown by metal organic chemical vapor deposition,

Appl. Phys. Lett. 78, 1382 (2001).

46 V. M. Ustinov, A. Y. Egorov, A. E. Zhukov, A. R. Kovsh, N. N. Le-

dentsov, M. V. Maximov, B. V. Volovik, A. F. Tsatsul'nikov, P. S.

Kop'ev, Z. I. Alferov, I. P. Soshnikov, N. Zakharov, P. Werner, and D.

Bimberg, 1.75 µm emission from self-organized InAs quantum dots on

GaAs, J. Crystal Growth 201/202, 1143 (1999).

47 Y. M. Shernyakov, D. A. Bedarev, E. Y. Kondrat'eva, P. S. Kop'ev, A.

R. Kovsh, N. A. Maleev, M. V. Maximov, S. S. Mikhrin, A. F.

Tsatsul'nikov, V. M. Ustinov, B. V. Volovik, A. E. Zhukov, Z. I. Al-

ferov, N. N. Ledentsov, and D. Bimberg, 1.3 µm GaAs-based laser us-

Bibliography

129

ing quantum dots obtained by activated spinodal decomposition, El.

Lett. 35, 898 (1999).

48 A. E. Zhukov, A. R. Kovsh, A. Y. Egorov, N. A. Maleev, V. M. Usti-

nov, B. V. Volovik, M. V. Maksimov, A. F. Tsatsul'nikov, N. N. Le-

dentsov, Y. M. Shernyakov, A. V. Luniv, Y. G. Musikhin, N. A. Bert,

P. S. Kop'ev, and Z. I. Alferov, Photo- and electroluminescence in the

1.3 µm wavelength range from quantum-dot structures grown on

GaAs substrates, Semiconductors 33, 153 (1999).

49 M. V. Maximov, A. F. Tsatsul’nikov, B. V. Volovik, D. S. Sizov, Y.

M. Shernyakov, I. N. Kaiander, A. E. Zhukov, A. R. Kovsh, S. S.

Mikhrin, V. M. Ustinov, and Z. I. Alferov, Tuning quantum dot prop-

erties by activated phase separation of an InGa(Al)As alloy grown on

InAs stressors, Phys. Rev. B 62, 16671 (2000).

50 Y. Arakawa, Progress in Self-Assembled Quantum Dots for Optoelec-

tronic Device Application, IEICE Trans. El. E85-C, 37 (2002).

51 N. A. Maleev, A. R. Kovsh, A. E. Zhukov, S. S. Mikhrin, A. P. Vasi-

l'ev, Y. M. Shernyakov, D. A. Livshits, M. V. Maximov, V. M. Usti-

nov, and Z. I. Alferov, High-performance long-wavelength QD diode

lasers on GaAs substrates, Proc. of the 18th International Conference

on Semiconductor Lasers, Garmisch, Germany, Sept 30 - Oct 3, 2002

(2002), Vol. 1, p. 153.

52 G. Park, O. B. Shchekin, S. Csutak, D. L. Huffaker, and D. G. Deppe,

Room-temperature continuous-wave operation of a single-layered 1.3

µm quantum dot laser, Appl. Phys. Lett. 75, 3267 (1999).

53 J. A. Lott, N. N. Ledentsov, V. M. Ustinov, Z. I. Alferov, and D. Bim-

berg, Continuous wave 1.3 µm InAs-InGaAs quantum dot VCSELs on

GaAs substrates, Proc. of the 13th Int. Conf. Ind. Phosph. Rel. Mat.

(IPRM) (2001), p. 137.

54 H. Y. Liu, M. Hopkinson, C. N. Harrison, M. J. Steer, R. Frith, I. R.

Sellers, D. J. Mowbray, and M. S. Skolnick, Optimizing the growth of

1.3 µm InAs/InGaAs dots-in-a-well structure, J. Appl. Phys. 93, 2931

(2003).

55 K. Nishi, H. Saito, S. Sugou, and J.-S. Lee, A narrow photolumines-

cence linewidth of 21 meV at 1.35 µm from strain-reduced InAs quan-

tum dots covered by In0.2Ga0.8As grown on GaAs substrates, Appl.

Phys. Lett. 74, 1111 (1999).

56 L. F. Lester, A. Stintz, H. Li, T. C. Newell, E. A. Pease, B. A. Fuchs,

and K. J. Malloy, Optical characteristics of 1.24-µm InAs quantum-

dot laser diodes, Phot. Techn. Lett. 11, 931 (1999).

57 V. M. Ustinov, N. A. Maleev, A. E. Zhukov, A. R. Kovsh, A. Y.

Egorov, A. V. Lunev, B. V. Volovik, I. L. Krestnikov, Y. G.

Musikhin, N. A. Bert, P. S. Kop’ev, Z. I. Alferov, N. N. Ledentsov,

Bibliography

130

and D. Bimberg, InAs/InGaAs quantum dot structures on GaAs sub-

strates emitting at 1.3 µm, Appl. Phys. Lett. 74, 2815 (1999).

58 P. G. Eliseev, H. Li, G. T. Liu, A. Stintz, T. C. Newell, L. F. Lester,

and K. J. Malloy, Ground-State Emission and Gain in Ultralow-

Threshold InAs–InGaAs Quantum-Dot Lasers, J. Sel. Top. Quantum

El. 7, 135 (2001).

59 R. Krebs, F. Klopf, J. P. Reithmaier, and A. Forchel, High Perform-

ance 1.3 µm Quantum-Dot Lasers, Jpn. J. Appl. Phys. (1) 41, 1158

(2002).

60 H. Zellmer, S. Buteau, A. Tünnermann, and H. Welling, All fibre laser

system with 0.1 W output power in blue spectral range, El. Lett. 33,

1383 (1997).

61 E. B. Mejía, A. N. Starodumov, and Y. O. Barmenkov, Blue and in-

frared up-conversion in Tm3+-doped fluorozirconate fiber pumped at

1.06, 1.117, and 1.18 µm, Appl. Phys. Lett. 74, 1540 (1999).

62 F. Klopf, J. P. Reithmaier, and A. Forchel, 980 nm Quantum Dot La-

sers with Very Small Threshold Current Densities, phys. stat. sol. (b)

3, 845 (2001).

63 S. S. Mikhrin, A. E. Zhukov, A. R. Kovsh, N. A. Maleev, V. M. Usti-

nov, Y. M. Shernyakov, I. P. Soshnikov, D. A. Livshits, I. S. Tarasov,

D. A. Bedarev, B. V. Volovik, M. V. Maximov, A. F. Tsatsul'nikov,

N. N. Ledentsov, and P. S. Kop'ev, 0.94 µm diode lasers based on

Stranski-Krastanow and sub-monolayer quantum dots, Semicond. Sci.

Technol. 15, 1061 (2000).

64 J. Porsche, M. Ost, F. Scholz, A. Fantini, F. Phillipp, T. Riedl, and A.

Hangleiter, Growth of self-assembled InP quantum islands for red-

light-emitting injection lasers, J. Sel. Top. Quantum El. 6, 482 (2000).

65 T. Riedl, A. Hangleiter, J. Porsche, and F. Scholz, Small-signal modu-

lation response of InP/GaInP quantum-dot lasers, Appl. Phys. Lett.

80, 4015 (2002).

66 P. Borri, S. Schneider, W. Langbein, U. Woggon, A. E. Zhukov, N. N.

Ledentsov, Z. I. Alferov, D. Ouyang, and D. Bimberg, Ultrafast car-

rier dynamics and dephasing in InAs quantum dot amplifers emitting

near 1.3 µm-wavelength at room temperature, Appl. Phys. Lett. 79,

2633 (2001).

67 P. Borri, Ultrafast Optical Properties of Quantum Dot Amplifiers in

Nano-Optoelectronics, ed. by M. Grundmann (Springer, Berlin, Hei-

delberg, 2002), p. 411.

68 P. Bhattacharya and S. Ghosh, Tunnel injection In0.4Ga0.6As/GaAs

quantum dot lasers with 15 GHz modulation bandwidth at room tem-

perature, Appl. Phys. Lett. 80, 3482 (2002).

Bibliography

131

69 C. Ribbat, R. Sellin, I. Kaiander, F. Hopfer, N. N. Ledentsov, D. Bim-

berg, A. R. Kovsh, V. M. Ustinov, A. E. Zhukov, and M. V. Maxi-

mov, Complete suppression of filamentation and superior beam qual-

ity in quantum-dot lasers, Appl. Phys. Lett. 82, 952 (2003).

70 N. Kirstaedter, O. G. Schmidt, N. N. Ledentsov, D. Bimberg, V. M.

Ustinov, A. Y. Egorov, A. E. Zhukov, M. V. Maximov, P. S. Kop’ev,

and Z. I. Alferov, Gain and differential gain of single layer InAs/GaAs

quantum dot injection lasers, Appl. Phys. Lett. 69, 1226 (1996).

71 K. Haberland, Optical in-situ Studies during Metal-Organic Vapor

Phase Epitaxy with respect to III-V-Device Production in Berlin Stud-

ies in Solid State Physics, ed. by C. Thomsen, D. Bimberg, M. Dähne,

and W. Richter (Wissenschaft & Technik Verlag, Berlin, 2002).

72 F. Heinrichsdorff, A. Krost, D. Bimberg, A. O. Kosogov, and P.

Werner, Self organized defect free InAs/GaAs and InAs/InGaAs/GaAs

quantum dots with high lateral density grown by MOCVD, Appl. Surf.

Sci. 123/124, 725 (1998).

73 R. Sellin, F. Heinrichsdorff, C. Ribbat, M. Grundmann, U. W. Pohl,

and D. Bimberg, Surface flattening during MOCVD of thin GaAs lay-

ers covering InGaAs quantum dots, J. Crystal Growth 221, 581

(2000).

74 M. Grundmann, O. Stier, S. Bognár, C. Ribbat, F. Heinrichsdorff, and

D. Bimberg, Optical Properties of Self-Organized Quantum Dots:

Modeling and Experiments, phys. stat. sol. (a) 178, 255 (2000).

75 V. A. Haisler, F. Hopfer, R. L. Sellin, A. Lochmann, K. Fleischer, N.

Esser, W. Richter, N. N. Ledentsov, D. Bimberg, C. Möller, and N.

Grote, Micro-Raman studies of vertical-cavity surface-emitting lasers

with AlxOy/GaAs distributed Bragg reflectors, Appl. Phys. Lett. 81,

2544 (2002).

76 F. Hopfer, Oberflächenemittierende Quantenpunktlaser, Ph.D. thesis

(Technische Universität Berlin), to be published.

77 W. Stolz, Alternative N-, P- and As-precursors for III/V-epitaxy, J.

Crystal Growth 209, 272 (2000).

78 G. B. Stringfellow, Organometallic Vapor-Phase Epitaxy (Academic

Press, 2nd edition, 1999).

79 F. Höhnsdorf, J. Koch, C. Agert, and W. Stolz, Investigations of

(GaIn)(NAs) bulk layers and (GaIn)(NAs)/GaAs multiple quantum

well structures grown using tertiarybutylarsine (TBAs) and 1,1-

dimethylhydrazine (UDMHy), J. Crystal Growth 195, 391 (1998).

80 M. P. Chase, M. D. Deal, and J. D. Plummer, Diffusion modeling of

zinc implanted into GaAs, J. Appl. Phys. 81, 1670 (1997).

81 F. Brunner, T. Bergunde, E. Richter, P. Kurpas, M. Achouche, A.

Maaßdorf, J. Würfl, and M. Weyers, Carbon doping for the GaAs

Bibliography

132

base layer of Heterojunction Bipolar Transistors in a production scale

MOVPE reactor, J. Crystal Growth 221, 53 (2000).

82 K. Tateno, Y. Kohama, and C. Amano, Carbon doping and etching

effects of CBr4 during metalorganic chemical vapor deposition of

GaAs and AlAs, J. Crystal Growth 172, 5 (1997).

83 U. W. Pohl, K. Knorr, C. Möller, U. Gernert, W. Richter, J. Bläsing, J.

Christen, J. Gottfriedsen, and H. Schumann, Low-Temperature Meta-

lorganic Vapor Phase Epitaxy (MOVPE) of GaN using Tertiarybutyl-

hydrazine, Jpn. J. Appl. Phys. (2) 38, L105 (1999).

84 D. K. Maude, DX centres in GaAs in Properties of Gallium Arsenide,

ed. by M. R. Brozel (INSPEC, London, 1996), Vol. XXVII, p. 250.

85 Y.-M. Houng and T. S. Low, Te doping of GaAs and AlxGa1-xAs using

diethyltellurium in low pressure OMVPE, J. Crystal Growth 77, 272

(1986).

86 L. G. Salmon and I. J. D'Haenens, The effect of aluminum composition

on silicon donor behavior in AlxGa1-xAs, J. Vac. Sci. Technol. B 2, 197

(1984).

87 S. Z. Sun, E. A. Armour, K. Zheng, and C. F. Schaus, Zinc and tellu-

rium doping in GaAs and AlxGa1-xAs grown by MOCVD, J. Crystal

Growth 113, 103 (1991).

88 S. Leu, H. Protzmann, F. Höhnsdorf, W. Stolz, J. Steinkirchner, and E.

Hufgard, Si-doping of MOVPE grown InP and GaAs by using the liq-

uid Si source ditertiarybutyl silane, J. Crystal Growth 195, 91 (1998).

89 W. Stolz and T. Whitaker, TBAs and TBP - Improved Group V

Sources for MOVPE, Compound Semiconductors 5, 29 (1999).

90 P. Kurpas, Untersuchungen zur Eignung alternativer Phosphor- und

Arsen-Ausgangsstoffe für die Metallorganische Gasphasenepitaxie

(MOVPE) (Shaker Verlag, Aachen, 1995).

91 R. M. Lum, J. K. Klingert, D. W. Kisker, D. M. Tennant, M. D. Mor-

ris, D. L. Malm, J. Kovalchick, and L. A. Heimbrook, Investigation of

carbon incorporation in GaAs using 13C-enriched trimethylarsenic

and 13CH4, J. Electron. Mat. 17, 101 (1988).

92 R. M. Lum, J. K. Klingert, and M. G. Lamont, Comparison of alter-

nate As-sources to arsine in the MOCVD growth of GaAs, J. Crystal

Growth 89, 137 (1988).

93 T. Kobayashi and N. Inoue, Carbon-doped GaAs grown by metalor-

ganic vapor phase epitaxy using TMAs and TEG, J. Crystal Growth

102, 183 (1990).

94 G. Haacke, S. P. Watkins, and H. Burkhard, Metalorganic chemical

vapor deposition of high-purity GaAs using tertiarybutylarsine, Appl.

Phys. Lett. 54, 2029 (1989).

Bibliography

133

95 R. Beccard, G. Lengeling, D. Schmitz, Y. Gigase, and H. Jürgensen,

Replacements of hydrides by TBAs and TBP for the growth of various

III-V materials in production scale MOVPE reactors, J. Crystal

Growth 170, 97 (1997).

96 G. B. Stringfellow, Novel precursors for organometallic vapor phase

epitaxy, J. Crystal Growth 128, 503 (1993).

97 J. Mimila-Arroyo, A. Lusson, J. Chevallier, M. Barbé, B. Theys, F.

Jomard, and S. W. Bland, Carbon acceptor doping efficiency in GaAs

grown by metalorganic chemical vapor deposition, Appl. Phys. Lett.

79, 3095 (2001).

98 K. Watanabe and H. Yamazaki, Characterization of annealed heavily

C-doped p+-AlGaAs, J. Appl. Phys. 74, 5587 (1993).

99 S. Krishna, S. Raghavan, A. L. Gray, A. Stintz, and K. J. Malloy,

Characterization of rapid-thermal-annealed InAs/In0.15Ga0.85As dots-

in-well heterostructure using double crystal x-ray diffraction and pho-

toluminescence, Appl. Phys. Lett. 80, 3898 (2002).

100 F. Heinrichsdorff, M. Grundmann, O. Stier, A. Krost, and D. Bimberg,

Influence of In/Ga intermixing on the optical properties of In-

GaAs/GaAs quantum dots, J. Crystal Growth 195, 540 (1998).

101 R. Leon, S. Fafard, P. G. Piva, S. Ruvimov, and Z. Liliental-Weber,

Tunable intersublevel transitions in self-forming semiconductor quan-

tum dots, Phys. Rev. B 58, R4262 (1998).

102 H. Kakinuma, M. Mohri, and M. Akiyama, Characterization of Oxy-

gen and Carbon in Undoped AlGaAs Grown by Organometallic Va-

por-Phase Epitaxy, Jpn. J. Appl. Phys. (1) 36, 23 (1997).

103 S. Leu, F. Höhnsdorf, W. Stolz, R. Becker, A. Salzmann, and A.

Greiling, C- and O-incorporation in (AlGa)As epitaxial layers grown

by MOVPE using TBAs, J. Crystal Growth 195, 98 (1998).

104 F. Reinhardt, B. Dwir, and E. Kapon, Oxidation of GaAs/AlGaAs het-

erostructures studied by atomic force microscopy in air, Appl. Phys.

Lett. 68, 3168 (1996).

105 F. Reinhardt, B. Dwir, G. Biasiol, and E. Kapon, Atomic force micros-

copy of III-V nanostructures in air, Appl. Surf. Sci. 104-105, 529

(1996).

106 A. Krost, G. Bauer, and J. Woitok, High Resolution X-Ray Diffraction

in Optical Characterization of Epitaxial Semiconductor Layers, ed. by

G. Bauer and W. Richter (Springer, Berlin, 1996), p. 287.

107 S. A. Furman and A. V. Tikhonravov, Optics of Multilayer Systems

(Editions Frontières, 1992).

108 R. Westphalen, G. Landgren, B. Stalnacke, and R. Beccard, Improved

homogeneity of LP-MOVPE grown InP/GaInAsP heterostructure for

DBR using an optimized liner and susceptor arrangement, Proc. of the

Bibliography

134

11th Int. Conf. Ind. Phosph. Rel. Mat. (IPRM), Davos, Switzerland,

May 16-20 (1999), p. 139.

109 K. D. Choquette, K. M. Geib, C. I. H. Ashby, R. D. Twesten, O.

Blum, H. Q. Hou, D. M. Follstaedt, B. E. Hammons, D. Mathes, and

R. Hull, Advances in Selective Wet Oxidation of AlGaAs Alloys, J. Sel.

Top. Quantum El. 3, 916 (1997).

110 K. D. Choquette, K. M. Geib, H. C. Chui, B. E. Hammons, H. Q. Hou,

T. J. Drummond, and R. Hull, Selective oxidation of buried AlGaAs

versus AlAs layers, Appl. Phys. Lett. 69, 1385 (1996).

111 H. Q. Jia, H. Chen, W. C. Wang, W. X. Wang, W. Li, Q. Huang, J.

Zhou, and Q. K. Xue, Improved thermal stability of wet-oxidized AlAs,

Appl. Phys. Lett. 80, 974 (2002).

112 S. Ruvimov, P. Werner, K. Scheerschmidt, U. Gösele, J. Heydenreich,

U. Richter, N. N. Ledentsov, M. Grundmann, D. Bimberg, V. M.

Ustinov, A. Y. Egorov, P. S. Kop'ev, and Z. I. Alferov, Structural

characterization of (In,Ga)As quantum dots in a GaAs matrix, Phys.

Rev. B 51, 14766 (1995).

113 J. Oshinowo, M. Nishioka, S. Ishida, and Y. Arakawa, Highly uniform

InGaAs/GaAs quantum dots (15 nm) by metalorganic chemical vapor

deposition, Appl. Phys. Lett. 65, 1421 (1994).

114 J. Márquez, L. Geelhaar, and K. Jacobi, Atomically resolved structure

of InAs quantum dots, Appl. Phys. Lett. 78, 2309 (2001).

115 D. Leonard, M. Krishnamurthy, C. M. Reaves, S. P. Denbaars, and P.

M. Petroff, Direct formation of quantum-sized dots from uniform co-

herent islands of InGaAs on GaAs surfaces, Appl. Phys. Lett. 63, 3203

(1993).

116 U. Woggon, W. Langbein, J. M. Hvam, A. Rosenauer, T. Remmele,

and D. Gerthsen, Electron microscopic and optical investigations of

the indium distribution in GaAs capped InxGa1 – xAs islands, Appl.

Phys. Lett. 71, 377 (1997).

117 J. W. Gibbs, Collected Works (Longmans, London, 1928), Vol. 1

(Thermodynamics).

118 C. Herring, Some Theorems on the Free Energies of Crystal Surfaces,

Phys. Rev. 82, 87 (1951).

119 V. A. Shchukin and D. Bimberg, Spontaneous ordering of nanostruc-

tures on crystal surfaces, Rev. Mod. Phys. 71, 1125 (1999).

120 F. C. Frank and J. H. van der Merve, One-dimensional dislocations,

Proc. Roy. Soc. London A 198, 205 (1949).

121 M. Volmer and A. Weber, Keimbildung in übersättigten Gebilden, Z.

Physik. Chem. 119, 277 (1926).

Bibliography

135

122 I. N. Stranski and L. Krastanow, Zur Theorie der orientierten Aus-

scheidung von Ionenkristallen aufeinander, Sitzungsber. Akad. Wiss.

Wien, Math.-Naturwiss. Klasse 146, 797 (1937).

123 T. R. Ramachandran, A. Madhukar, I. Mukhametzhanov, R. Heitz, A.

Kalburge, Q. Xie, and P. Chen, Nature of Stranski–Krastanow growth

of InAs on GaAs(001), J. Vac. Sci. Technol. B 13, 1330 (1998).

124 W. Ostwald, Über die vermeintliche Isomerie des roten und gelben

Quecksilberoxyds und die Oberflächenspannung fester Körper, Z.

Physik. Chem. 34, 495 (1900).

125 L. G. Wang, P. Kratzer, M. Scheffler, and N. Moll, Formation and

Stability of Self-Assembled Coherent Islands in Highly Mismatched

Heteroepitaxy, Phys. Rev. Lett. 82, 4042 (1999).

126 A.-L. Barabási, Self-assembled island formation in heteroepitaxial

growth, Appl. Phys. Lett. 70, 2565 (1997).

127 M. Meixner, E. Schöll, V. A. Shchukin, and D. Bimberg, Self-

Assembled Quantum Dots: Crossover from Kinetically Controlled to

Thermodynamically Limited Growth, Phys. Rev. Lett. 87, 236101

(2001).

128 C. Priester and M. Lannoo, Origin of Self-Assembled Quantum Dots in

Highly Mismatched Heteroepitaxy, Phys. Rev. Lett. 75, 93 (1995).

129 V. A. Shchukin, N. N. Ledentsov, A. Hoffmann, D. Bimberg, I. P.

Soshnikov, B. V. Volovik, V. M. Ustinov, D. Litvinov, and D. Gerth-

sen, Entropy-Driven Effects in Self-Organized Formation of Quantum

Dots, phys. stat. sol. (b) 224, 503 (2001).

130 A. Rosenauer, T. Remmele, and D. Gerthsen, Atomic scale strain

measurements by the digital analysis of transmission electron micro-

scopic lattice images, Optik 105, 99 (1997).

131 R. Leon, C. Lobo, X. Z. Liao, J. Zou, D. J. H. Cockayne, and S.

Fafard, Island shape instabilities and surfactant-like effects in the

growth of InGaAs/GaAs quantum dots, Thin Solid Films 357, 40

(1999).

132 D. Leonard, K. Pond, and P. M. Petroff, Critical layer thickness for

self-assembled InAs islands on GaAs, Phys. Rev. B 50, 11687 (1994).

133 G. S. Solomon, J. A. Trezza, and J. S. Harris, Effects of monolayer

coverage, flux ratio, and growth rate on the island density of InAs is-

lands on GaAs, Appl. Phys. Lett. 66, 3161 (1995).

134 R. Leon, C. Lobo, J. Zou, T. Romeo, and D. J. H. Cockayne, Stable

and Metastable InGaAs/GaAs Island Shapes and Surfactantlike Sup-

pression of the Wetting Transformation, Phys. Rev. Lett. 81, 2486

(1998).

Bibliography

136

135 R. E. Welser and L. J. Guido, Evidence of two-species nucleation of

InAs islands on (100) and (111)B GaAs substrates, Appl. Phys. Lett.

68, 912 (1996).

136 M. Sopanen, H. Lipsanen, and J. Ahopelto, Self-organized InP islands

on (100) GaAs by metalorganic vapor phase epitaxy, Appl. Phys. Lett.

67, 3768 (1995).

137 J. Johansson and W. Seifert, Kinetics of self-assembled island forma-

tion: Part I - Island density, J. Crystal Growth 234, 132 (2002).

138 P. B. Joyce, T. J. Krzyzewski, G. R. Bell, T. S. Jones, S. Malik, D.

Childs, R. Murray, and S. Den Baars, Growth rate effects on the size,

composition and optical properties of InAs/GaAs quantum dots grown

by molecular beam epitaxy, J. Crystal Growth 227-228, 1000 (2001).

139 P. B. Joyce, T. J. Krzyzewski, G. R. Bell, T. S. Jones, E. C. Le Ru,

and R. Murray, Optimizing the growth of 1.3 µm InAs/GaAs quantum

dots, Phys. Rev. B 64, 235317 (2001).

140 I. Mukhametzhanov, Z. Wei, R. Heitz, and A. Madhukar, Punctuated

island growth: An approach to examination and control of quantum

dot density, size, and shape evolution, Appl. Phys. Lett. 75, 85 (1999).

141 E. Steimetz, W. Richter, F. Schienle, D. Fischer, M. Klein, and J.-T.

Zettler, The Effect of Different Group V Precursors on the Evolution

of Quantum Dots Monitored by Optical In Situ Measurements, Jpn. J.

Appl. Phys. (1) 37, 1483 (1998).

142 F. Heinrichsdorff, MOCVD growth and laser applications of

In(Ga)As/GaAs Quantum Dots (Mensch & Buch Verlag, Berlin,

1998).

143 R. Sellin, N. N. Ledentsov, and D. Bimberg, Verfahren zur Verbesse-

rung der Effizienz von epitaktisch hergestellten Quantenpunkt-

Halbleiterbauelementen mit einer oder mehreren Quantenpunkt-

schichten, Patent pending, ref. no. 10044040 (DE) (2000).

144 K. Scheerschmidt and P. Werner, TEM Characterization of Quantum

Dots in Nano-Optoelectronics, ed. by M. Grundmann (Springer, Ber-

lin, Heidelberg, 2002), p. 67.

145 M. Kasu and N. Kobayashi, Equilibrium multiatomic step structure of

GaAs(001) vicinal surfaces grown by metalorganic chemical vapor

deposition, Appl. Phys. Lett. 62, 1262 (1993).

146 N. N. Ledentsov, G. M. Gurianov, G. E. Tsyrlin, V. N. Petrov, Y. B.

Samsonenko, A. O. Golubok, and S. Y. Tipisev, Effect of heat-

treatment conditions on the surface morphology of gallium arsenide

grown on vicinal GaAs (100) substrates by molecular-beam epitaxy,

Semiconductors 28, 526 (1994).

Bibliography

137

147 M. Kasu and N. Kobayashi, Surface kinetics of metalorganic vapor-

phase epitaxy: surface diffusion, nucleus formation, sticking at steps,

J. Crystal Growth 174, 513 (1997).

148 T. Ito and K. Shiraishi, A Theoretical Investigation of Migration Po-

tentials of Ga Adatoms near Step Edges on GaAs(001)-c(4×4) Sur-

face, Jpn. J. Appl. Phys. (2) 35, L1016 (1996).

149 E. J. Heller and M. G. Lagally, In-situ scanning tunneling microscopy

observation of surface morphology of GaAs(001) grown by molecular

beam epitaxy, Appl. Phys. Lett. 60, 2675 (1992).

150 A. Kley, P. Ruggerone, and M. Scheffler, Novel Diffusion Mechanism

on the GaAs(001) Surface: The Role of Adatom-Dimer Interaction,

Phys. Rev. Lett. 79, 5278 (1997).

151 V. Holý, G. Springholz, M. Pinczolits, and G. Bauer, Strain Induced

Vertical and Lateral Correlations in Quantum Dot Superlattices,

Phys. Rev. Lett. 83, 356 (1999).

152 M. M. Sobolev, I. V. Kochnev, V. M. Lantratov, N. A. Bert, N. A.

Cherkashin, N. N. Ledentsov, and D. A. Bedarev, Thermal annealing

of Defects in InGaAs/GaAs Heterostructures with Three-Dimensional

Islands, Semiconductors 34, 195 (2000).

153 P. D. Dapkus, H. M. Manasevit, K. L. Hess, T. S. Low, and G. E.

Stillman, High purity GaAs prepared from trimethylgallium and ar-

sine, J. Crystal Growth 55, 10 (1981).

154 N. N. Ledentsov, M. V. Maximov, D. Bimberg, T. Maka, C. M. S.

Torres, I. V. Kochnev, I. L. Krestnikov, V. M. Lantratov, N. A. Cher-

kashin, Y. M. Musikhin, and Z. I. Alferov, 1.3 µm luminescence and

gain from defect-free InGaAs–GaAs quantum dots grown by metal-

organic chemical vapour deposition, Semicond. Sci. Technol. 15, 604

(2000).

155 I. L. Krestnikov, N. A. Cherkashin, D. S. Sizov, D. A. Bedarev, I. V.

Kochnev, V. M. Lantratov, and N. N. Ledentsov, InGaAs Nanodo-

mains Formed in situ on the Surface of (Al,Ga)As, Tech. Phys. Lett.

27, 233 (2001).

156 R. L. Sellin, C. Ribbat, M. Grundmann, N. N. Ledentsov, and D. Bim-

berg, Close-to-ideal device characteristics of high-power In-

GaAs/GaAs quantum dot lasers, Appl. Phys. Lett. 78, 1207 (2001).

157 R. L. Sellin, C. Ribbat, D. Bimberg, F. Rinner, H. Konstanzer, M. T.

Kelemen, and M. Mikulla, High-reliability MOCVD-grown quantum

dot laser, El. Lett. 38, 883 (2002).

158 R. Leon, C. Lobo, A. Clark, R. Bozek, A. Wysmolek, A. Kurpiewski,

and M. Kaminska, Different paths to tunability in III–V quantum dots,

J. Appl. Phys. 84, 248 (1998).

Bibliography

138

159 N. Moll, A. Kley, E. Pehlke, and M. Scheffler, GaAs equilibrium crys-

tal shape from first principles, Phys. Rev. B 54, 8844 (1996).

160 E. Pehlke, N. Moll, A. Kley, and M. Scheffler, Shape and stability of

quantum dots, Appl. Phys. A 65, 525 (1997).

161 R. L. Sellin, I. Kaiander, D. Ouyang, T. Kettler, and D. Bimberg, Al-

ternative-precursor metalorganic chemical vapor deposition and laser

application of self-organized InGaAs/GaAs quantum dots, Appl. Phys.

Lett. 82, 841 (2003).

162 A. Y. Egorov, D. Bedarev, D. Bernklau, G. Dumitras, and H. Riechert,

Self-Assembled InAs Quantum Dots in an InGaAsN Matrix on GaAs,

phys. stat. sol. (b) 224, 839 (2001).

163 H. Saito, T. Makimoto, and N. Kobayashi, MOVPE growth of strained

InGaAsN/GaAs quantum wells, J. Crystal Growth 195, 416 (1998).

164 W. J. Fan, S. F. Yoon, T. K. Ng, S. Z. Wang, W. K. Loke, R. Liu, and

A. Wee, Comparison of nitrogen compositions in the as-grown

GaNxAs1-x on GaAs measured by high-resolution x-ray diffraction and

secondary-ion mass spectroscopy, Appl. Phys. Lett. 80, 4136 (2002).

165 W. Li, M. Pessa, and J. Likonen, Lattice parameter in GaNAs epilay-

ers on GaAs: Deviation from Vegard’s law, Appl. Phys. Lett. 78, 2864

(2001).

166 G. T. Liu, A. Stintz, H. Li, T. C. Newell, A. L. Gray, P. M. Varangis,

K. J. Malloy, and L. F. Lester, The Influence of Quantum-Well Com-

position on the Performance of Quantum Dot Lasers Using

InAs/InGaAs Dots-in-a-Well (DWELL) Structures, J. Quantum El. 36,

1272 (2000).

167 F. Guffarth, R. Heitz, A. Schliwa, O. Stier, N. N. Ledentsov, A. R.

Kovsh, V. M. Ustinov, and D. Bimberg, Strain engineering of self-

organized InAs quantum dots, Phys. Rev. B 64, 85305 (2001).

168 N.-T. Yeh, T.-E. Nee, and J.-I. Chyia, Matrix dependence of strain-

induced wavelength shift in self-assembled InAs quantum-dot het-

erostructures, Appl. Phys. Lett. 76, 1567 (2000).

169 M. V. Maximov, A. F. Tsatsul'nikov, B. V. Volovik, D. A. Bedarev,

A. E. Zhukov, A. R. Kovsh, N. A. Maleev, V. M. Ustinov, P. S.

Kop'ev, Z. I. Alferov, R. Heitz, N. N. Ledentsov, and D. Bimberg,

Quantum dots formed by activated spinodal decomposition of

InGa(Al)As alloy on InAs stressors, Physica E 7, 326 (2000).

170 H. Y. Liu, X. D. Wang, B. Xu, D. Ding, W. H. Jiang, J. Wu, and Z. G.

Wang, Effect of In-mole-fraction in InGaAs overgrowth layer on self-

assembled InAs/GaAs quantum dots, J. Crystal Growth 213, 193

(2000).

171 R. Heitz, I. Mukhametzhanov, A. Madhukar, A. Hoffmann, and D.

Bimberg, Temperature Dependent Optical Properties of Self-

Bibliography

139

Organized InAs/GaAs Quantum Dots, J. Electron. Mat. 28, 520

(1999).

172 O. Stier, M. Grundmann, and D. Bimberg, Electronic and optical

properties of strained quantum dots modeled by 8-band k·p theory,

Phys. Rev. B 59, 5688 (1999).

173 X. Z. Liao, J. Zou, D. J. H. Cockayne, R. Leon, and C. Lobo, Indium

Segregation and Enrichment in Coherent InxGa1–xAs/GaAs Quantum

Dots, Phys. Rev. Lett. 82, 5148 (1999).

174 P. B. Hirsch, A. Howie, R. B. Nicholson, D. W. Pashley, and M. J.

Whelan, Electron Microscopy of Thin Crystals (Butterworths, Lon-

don, 1965), p. 161.

175 J. A. Ibers and W. C. Hamilton (Ed.), International Tables for X-Ray

Crystallography (Kynoch Press, Birmingham, Vol. IV, 1974), p. 155.

176 A. E. Zhukov, A. R. Kovsh, N. A. Maleev, S. S. Mikhrin, V. M. Usti-

nov, A. F. Tsatsul’nikov, M. V. Maximov, B. V. Volovik, D. A. Be-

darev, Y. M. Shernyakov, P. S. Kop’ev, and Z. I. Alferov, Long-

wavelength lasing from multiply stacked InAs/InGaAs quantum dots

on GaAs substrates, Appl. Phys. Lett. 75, 1926 (1999).

177 Q. Xie, P. Chen, and A. Madhukar, InAs island-induced-strain driven

adatom migration during GaAs overlayer growth, Appl. Phys. Lett.

65, 2051 (1994).

178 M. Grundmann, O. Stier, and D. Bimberg, InAs/GaAs pyramidal

quantum dots: Strain distribution, optical phonons, and electronic

structure, Phys. Rev. B 52, 11969 (1995).

179 A. Lenz, R. Timm, H. Eisele, C. Hennig, S. K. Becker, R. L. Sellin, U.

W. Pohl, D. Bimberg, and M. Dähne, Reversed truncated cone com-

position distribution of In0.8Ga0.2As quantum dots overgrown by an

In0.1Ga0.9As layer in a GaAs matrix, Appl. Phys. Lett. 81, 5150 (2002).

180 W. Wu, J. R. Tucker, G. S. Solomon, and J. S. Harris, Atom-resolved

scanning tunneling microscopy of vertically ordered InAs quantum

dots, Appl. Phys. Lett. 71, 1083 (1997).

181 B. Grandidier, Y. M. Niquet, B. Legrand, J. P. Nys, C. Priester, D.

Stiévenard, J. M. Gérard, and V. Thierry-Mieg, Imaging the Wave-

Function Amplitudes in Cleaved Semiconductor Quantum Boxes,

Phys. Rev. Lett. 85, 1068 (2000).

182 H. Eisele, Cross-Sectional Scanning Tunneling Microscopy of

InAs/GaAs Quantum Dots in Berlin Studies in Solid State Physics, ed.

by C. Thomsen, D. Bimberg, M. Dähne, and W. Richter (Wissenschaft

& Technik Verlag, Berlin, 2002), Vol. 10.

183 W. Barvosa-Carter, M. E. Twigg, M. J. Yang, and L. J. Whitman, Mi-

croscopic characterization of InAs/In0.28GaSb0.72/InAs/AlSb laser

structure interfaces, Phys. Rev. B 63, 245311 (2001).

Bibliography

140

184 R. M. Feenstra, J. A. Stroscio, J. Tersoff, and A. P. Fein, Atom-

selective imaging of the GaAs(110) surface, Phys. Rev. Lett. 58, 1192

(1987).

185 P. Ebert, B. Engels, P. Richard, K. Schroeder, S. Blügel, C. Domke,

M. Heinrich, and K. Urban, Contribution of Surface Resonances to

Scanning Tunneling Microscopy Images: (110) Surfaces of III-V

Semiconductors, Phys. Rev. Lett. 77, 2997 (1996).

186 O. Flebbe, H. Eisele, T. Kalka, F. Heinrichsdorff, A. Krost, D. Bim-

berg, and M. Dähne-Prietsch, Atomic structure of stacked InAs quan-

tum dots grown by metal-organic chemical vapor deposition, J. Vac.

Sci. Technol. B 17, 1639 (1999).

187 N. Grandjean, J. Massies, and O. Tottereau, Surface segregation in

(Ga,In)As/GaAs quantum boxes, Phys. Rev. B 55, R10189 (1997).

188 K. Muraki, S. Fukatsu, Y. Shiraki, and R. Ito, Surface segregation of

In atoms during molecular beam epitaxy and its influence on the en-

ergy levels in InGaAs/GaAs quantum wells, Appl. Phys. Lett. 61, 557

(1992).

189 H. Toyoshima, T. Niwa, J. Yamazaki, and A. Okamoto, In surface

segregation and growth-mode transition during InGaAs growth by

molecular-beam epitaxy, Appl. Phys. Lett. 63, 821 (1993).

190 H. Eisele, O. Flebbe, T. Kalka, F. Heinrichsdorff, A. Krost, D. Bim-

berg, and M. Dähne-Prietsch, The Stoichiometry of InAs Quantum

Dots Determined by Cross-Sectional Scanning Tunneling Microscopy,

phys. stat. sol. (b) 215, 865 (1999).

191 H. Eisele, O. Flebbe, T. Kalka, C. Preinesberger, F. Heinrichsdorff, A.

Krost, D. Bimberg, and M. Dähne-Prietsch, Cross-sectional scanning-

tunneling microscopy of stacked InAs quantum dots, Appl. Phys. Lett.

75, 106 (1999).

192 H. Eisele, O. Flebbe, T. Kalka, and M. Dähne-Prietsch, Cross-

sectional STM Study of InAs Quantum Dots for Laser Devices, Surf.

Interface Anal. 27, 537 (1999).

193 N. Liu, J. Tersoff, O. Baklenov, A. L. Holmes, and C. K. Shih, Non-

uniform Composition Profile in In0.5Ga0.5As Alloy Quantum Dots,

Phys. Rev. Lett. 84, 334 (2000).

194 P. W. Fry, I. E. Itskevich, D. J. Mowbray, M. S. Skolnick, J. J. Finley,

J. A. Barker, E. P. O’Reilly, L. R. Wilson, I. A. Larkin, P. A. Mak-

sym, M. Hopkinson, M. Al-Khafaji, J. P. R. David, A. G. Cullis, G.

Hill, and J. C. Clark, Inverted Electron-Hole Alignment in InAs-GaAs

Self-Assembled Quantum Dots, Phys. Rev. Lett. 74, 733 (2000).

195 W. Sheng and J.-P. Leburton, Electron-hole alignment in InAs/GaAs

self-assembled quantum dots: Effects of chemical composition and dot

shape, Phys. Rev. B 63, R 161301 (2001).

Bibliography

141

196 S. Sato and S. Satoh, Metalorganic chemical vapor deposition of

GaInNAs lattice matched to GaAs for long-wavelength laser diodes, J.

Crystal Growth 192, 381 (1998).

197 N. Tansu and L. J. Mawst, Low-Threshold Strain-Compensated In-

GaAs(N) (

=1.19–1.31 µm) Quantum-Well Lasers, Phot. Techn. Lett.

14, 444 (2002).

198 D. A. Livshits, A. Y. Egorov, and H. Riechert, 8 W continuous wave

operation of InGaAsN lasers at 1.3 µm, El. Lett. 36, 1381 (2000).

199 A. Ramakrishnan, G. Steinle, D. Supper, C. Degen, and G. Ebbing-

haus, Electrically pumped 10 Gbit/s MOVPE-grown monolithic 1.3

µm VCSEL with GaInNAs active region, El. Lett. 38, 322 (2002).

200 P. R. C. Kent and A. Zunger, Evolution of III-V Nitride Alloy Elec-

tronic Structure: The Localized to Delocalized Transition, Phys. Rev.

Lett. 86, 2613 (2001).

201 S. Adachi, GaAs and Related Materials (World Scientific, Singapore,

1994).

202 I. Vurgaftman, J. R. Meyer, and L. R. Ram-Mohan, Band parameters

for III–V compound semiconductors and their alloys, J. Appl. Phys.

89, 5815 (2001).

203 H. P. Hjalmarson, P. Vogel, D. J. Wolford, and J. D. Dow, Theory of

Substitutional Deep Traps in Covalent Semiconductors, Phys. Rev.

Lett. 44, 810 (1980).

204 W. Shan, W. Walukiewicz, J. W. Ager, E. E. Haller, J. F. Geisz, D. J.

Friedman, J. M. Olson, and S. R. Kurtz, Band Anticrossing in GaIn-

NAs Alloys, Phys. Rev. Lett. 82, 1221 (1999).

205 D. Serries, T. Geppert, P. Ganser, M. Maier, K. Köhler, N. Herres, and

J. Wagner, Quaternary GaInAsN with high In content: Dependence of

band gap energy on N content, Appl. Phys. Lett. 80, 2448 (2002).

206 T. Schmidtling, M. Klein, M. Pristovsek, K. Knorr, U. W. Pohl, and

W. Richter, MOVPE of GaAsN on GaAs(001) using tertiarybutylhy-

drazine, Ext. Abstr. EWMOVPE VIII, Praha, 433 (1999).

207 S. Makino, T. Miyamoto, T. Kageyama, N. Nishiyama, F. Koyama,

and K. Iga, GaInNAs/GaAs quantum dots grown by chemical beam

epitaxy, J. Crystal Growth 221, 561 (2000).

208 S. Makino, T. Miyamoto, T. Kageyama, Y. Ikenaga, F. Koyama, and

K. Iga, Nitrogen Composition and Growth Temperature Dependence

of Growth Characteristics for Self-Assembled GaInNAs/GaAs Quan-

tum Dots by Chemical Beam Epitaxy, Jpn. J. Appl. Phys. (1) 41, 953

(2002).

209 M. Sopanen, H. P. Xin, and C. W. Tu, Self-assembled GaInNAs quan-

tum dots for 1.3 and 1.55 µm emission on GaAs, Appl. Phys. Lett. 76,

994 (2000).

Bibliography

142

210 E. V. K. Rao, A. Ougazzaden, Y. L. Bellego, and M. Juhel, Optical

properties of low band gap GaAs1-xNx layers: Influence of post-growth

treatments, Appl. Phys. Lett. 72, 1409 (1998).

211 H. P. Xin and C. W. Tu, GaInNAs/GaAs multiple quantum wells

grown by gas-source molecular beam epitaxy, Appl. Phys. Lett. 72,

2442 (1998).

212 M. R. Gokhale, J. Wei, H. Wang, and S. R. Forrest, Growth and char-

acterization of small band gap (~ 0.6 eV) InGaAsN layers on InP,

Appl. Phys. Lett. 74, 1287 (1999).

213 A. Ougazzaden, S. Bouchoule, A. Mereuta, E. V. K. Rao, and J. De-

cobert, Room temperature laser operation of bulk InGaAsN/GaAs

structures grown by AP-MOVPE using N2 as carrier gas, El. Lett. 35,

474 (1999).

214 E. Tournié, M.-A. Pinault, and A. Guzmán, Mechanisms affecting the

photoluminescence spectra of GaInNAs after post-growth annealing,

Appl. Phys. Lett. 80, 4148 (2002).

215 J. M. Gérard, J. B. Génin, J. Levebre, J. M. Moison, N. Lebouché, and

F. Barthe, Optical investigation of the self-organized growth of

InAs/GaAs quantum boxes, J. Crystal Growth 150, 351 (1995).

216 G. B. Stringfellow, J. K. Shurtle, R. T. Lee, C. M. Fetzer, and S. W.

Jun, Surface processes in OMVPE - the frontiers, J. Crystal Growth

221, 1 (2000).

217 N. N. Ledentsov, J. Böhrer, D. Bimberg, I. V. Kochnev, M. V. Maxi-

mov, P. S. Kop’ev, Z. I. Alferov, A. O. Kosogov, S. S. Ruvimov, P.

Werner, and U. Gösele, Formation of coherent superdots using metal-

organic chemical vapor deposition, Appl. Phys. Lett. 69, 1095 (1996).

218 C. Ribbat, Hochleistungs-Quantenpunkt-Halbleiterlaser (Ph.D. thesis,

Technical University of Berlin, http://edocs.tu-berlin.de/diss/2002/

ribbat_christian.htm, Berlin, 2003).

219 L. V. Asryan, M. Grundmann, N. N. Ledentsov, O. Stier, R. A. Suris,

and D. Bimberg, Effect of excited-state transitions on the threshold

characteristics of a quantum dot laser, J. Quantum El. 37, 418 (2001).

220 C. Ribbat, R. Sellin, M. Grundmann, and D. Bimberg, High Power

Quantum Dot Lasers at 1160 nm, phys. stat. sol. (b) 224, 819 (2001).

221 R. Heitz, H. Born, F. Guffarth, O. Stier, A. Schliwa, A. Hoffmann,

and D. Bimberg, Existence of a phonon bottleneck for excitons in

quantum dots, Phys. Rev. B 64, 241305(R) (2001).

222 R. Heitz, Optical Properties of Self-Organized Quantum Dots in

Nano-Optoelectronics, ed. by M. Grundmann (Springer, Berlin, Hei-

delberg, 2002), p. 239.

223 A. Al-Muhanna, L. J. Mawst, D. Botez, D. Z. Garbuzov, R. U. Marti-

nelli, and J. C. Connolly, High-power (> 10 W) continuous-wave op-

Bibliography

143

eration from 100-µm-aperture 0.97-µm-emitting Al-free diode lasers,

Appl. Phys. Lett. 73, 1182 (1998).

224 X. He, S. Srinivasan, S. Wilson, C. Mitchell, and R. Patel, 10.9 W

continuous wave optical power from 100 µm aperture InGaAs/AlGaAs

(915 nm) laser diodes, El. Lett. 34, 2126 (1998).

225 D. A. Livshits, I. V. Kochnev, V. M. Lantratov, N. N. Ledentsov, T.

A. Nalyot, I. S. Tarasov, and Z. I. Alferov, Improved catastrophic op-

tical mirror damage level in InGaAs/AlGaAs laser diodes, El. Lett. 36,

1848 (2000).

226 O. B. Shchekin, G. Park, D. L. Huffaker, and D. G. Deppe, Discrete

energy level separation and the threshold temperature dependence of

quantum dot lasers, Appl. Phys. Lett. 77, 466 (2000).

227 Y. Q. Wei, S. M. Wang, F. Ferdos, J. Vukusic, A. Larsson, and Q. X.

Zhao, Large ground-to-first-excited-state transition energy separation

for InAs quantum dots emitting at 1.3 µm, Appl. Phys. Lett. 81, 1621

(2002).

228 H. Jiang and J. Singh, Strain distribution and electronic spectra of

InAs/GaAs self-assembled dots: An eight-band study, Phys. Rev. B 56,

4696 (1997).

229 O. B. Shchekin and D. G. Deppe, The role of p-type doping and the

density of states on the modulation response of quantum dot lasers,

Appl. Phys. Lett. 80, 2758 (2002).

230 W. T. Tsang, A new current-injection heterostructure laser: The dou-

ble-barrier double-heterostructure laser, Appl. Phys. Lett. 38, 835

(1981).

231 A. R. Kovsh, N. A. Maleev, A. E. Zhukov, S. S. Mikhrin, A. P. Va-

sil'ev, Y. M. Shernyakov, M. V. Maximov, D. A. Lvshits, V. M. Usti-

nov, Z. I. Alferov, N. N. Ledentsov, and D. Bimberg,

InAs/InGaAs/GaAs quantum dot lasers of 1.3 µm range with high

(88%) differential efficiency, El. Lett. 38, 1104 (2002).

232 O. B. Shchekin and D. G. Deppe, 1.3 µm InAs quantum dot laser with

T0 = 161 K from 0 to 80°C, Appl. Phys. Lett. 80, 3277 (2002).

233 S. Rapp, F. Salomonsson, K. Streubel, S. Mogg, F. Wennekes, J. Ben-

tell, and M. Hammar, All-Epitaxial Single-Fused 1.55 µm Vertical

Cavity Laser Based on an InP Bragg Reflector, Jpn. J. Appl. Phys. (1)

38, 1261 (1999).

234 J. M. Dallesasse, J. N. Holonyak, A. R. Sugg, T. A. Richard, and N.

El-Zein, Hydrolyzation oxidation of AlxGa1-xAs-AlAs-GaAs quantum

well heterostructures and superlattices, Appl. Phys. Lett. 57 (1990).

235 F. Hopfer, Oberflächenemittierende Quantenpunktlaser (Diploma the-

sis, Technische Universität Berlin, 1998).

Bibliography

144

236 V. M. Ustinov, A. Y. Egorov, V. A. Odnoblyudov, N. V. Kryzha-

novskaya, Y. G. Musikhin, A. F. Tsatsul'nikov, and Z. I. Alferov, I-

nAs/InGaAsN quantum dots emitting at 1.55 µm grown by molecular

beam epitaxy, J. Crystal Growth 251, 388 (2003).

237 D. S. Sizov, M. V. Maksimov, A. F. Tsatsul'nikov, N. A. Cherkashin,

N. V. Kryzhanovskaya, A. B. Zhukov, N. A. Maleev, S. S. Mikhrin,

A. P. Vasil'ev, R. Sellin, V. M. Ustinov, N. N. Ledentsov, D. Bim-

berg, and Z. I. Alferov, The Influence of Heat Treatment Conditions

on the Evaporation of Defect Regions in Structures with InGaAs

Quantum Dots in the GaAs Matrix, Semiconductors 36, 1020 (2002).

238 M. Kondow, B. Shi, and C. W. Tu, In situ etching using a novel pre-

cursor of tertiarybutylchloride (TBCl), J. Crystal Growth 209, 263

(2000).

239 P. Wolfram, W. Ebert, J. Kreissl, and N. Grote, MOVPE-based in situ

etching of In(GaAs)P/InP using tertiarybutylchloride, J. Crystal

Growth 221, 177 (2000).

240 N. Tansu and L. J. Mawst, High-Performance Strain-Compensated

InGaAs–GaAsP–GaAs (

= 1.17 µm) Quantum-Well Diode Lasers,

Phot. Techn. Lett. 13, 179 (2001).

241 C. Ellmers, S. Leu, R. Rettig, M. Hofmann, W. W. Rühle, and W.

Stolz, GaAs-based VCSEL-structures with strain-compensated

(GaIn)As/Ga(PAs)-MQWH active regions grown by using TBAs and

TBP, J. Crystal Growth 195, 630 (1998).

242 C. Ellmers, F. Höhnsdorf, J. Koch, C. Agert, S. Leu, D. Karaiskaj, M.

Hofmann, W. Stolz, and W. W. Rühle, Ultrafast (GaIn)(NAs)/GaAs

vertical-cavity surface-emitting laser for the 1.3 µm wavelength re-

gime, Appl. Phys. Lett. 74, 2271 (1999).