(Mitsuru Sugawara, Robert K. Willardson, Eicke R.

Self-Assembled InGaAslGaAs
Quantum Dots
SEMICONDUCTORS
AND SEMIMETALS
Volume 60
Semiconductors and Semimetals
A Treatise
Edited by R. K. Witlardson Eicke R. Weber

DEPARTMENT
CONSULTING PHYSICIST OF MATERIALS
SCIENCE
WASHINGTON AND MINERAL
SPOKANE, ENGINEERING
OF CALIFORNIA
UNIVERSITY AT
BERKELEY
Self-Assembled InGaAslGaAs
Quantum Dots
SEMICONDUCTORS
AND SEMIMETALS
Volume 60
Volume Editor
MITSURU SUGAWARA
OPTICAL SEMICONDUCTOR DEVICE LABORATORY
FUJITSU LABORATORIES LTD.
ATSUGI, JAPAN
ACADEMIC PRESS
San Diego London Boston
New York Sydney Tokyo Toronto
This book is printed on acid-free paper. @
COPYRIGHT 0 1999 BY ACADEMICPRESS
ALL RIGHTS RESERVED.
NO PART OF THIS PUBLICATION MAY BE REPRODUCED OR TRANSMITTED IN ANY FORM OR BY ANY
MEANS, ELECTRONIC OR MECHANICAL, INCLUDING PHOTOCOPY, RECORDING, OR ANY INFORMATION
STORAGE AND RETRIEVAL SYSTEM, WITHOUT PERMISSION IN WRITING FROM THE PUBLISHER.
The appearance of the code at the bottom of the first page of a chapter in this book indicates
the Publisher’s consent that copies of the chapter may be made for personal or internal use of
specific clients. This consent is given on the condition, however, that the copier pay the stated
per-copy fee through the Copyright Clearance Center, Inc. (222 Rosewood Drive, Danvers,
Massachusetts 01923), for copying beyond that permitted by Sections 107 or 108 of the US.
Copyright Law. This consent does not extend to other kinds of copying, such as copying for
general distribution, for advertising or promotional purposes, for creating new collective works,
or for resale. Copy fees for pre-1999 chapters are as shown on the title pages; if no fee code
appears on the title page, the copy fee is the same as for current chapters. 0080-8784/99 $30.00
ACADEMIC PRESS
525 B Street, Suite 1900. San Diego, C A 92101-4495. USA
http://www.apnet.corn
ACADEMIC PRESS
24-28 Oval Road, London NWl 7DX, UK
http://www.hbuk.co.uk/ap/
International Standard Book Number: 0-12-752169-0

International Standard Serial Number: 0080-8784
PRINTED I N THE UNITED STATES OF AMERICA

98 99 00 01 02 QW 9 8 7 6 5 4 3 2 I
Contents
PREFACE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi
LISTOF CONTRIBUTORS . . . . . . . . . . . . . . . . . . . . . . . . . . . xv
Chapter 1 Theoretical Bases of the Optical Properties of Semiconductor

Quantum Nano-Structures
Mitsuru Sugawara
I . INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
I1. ELECTRONIC STATES OF SEMICONDUCTOR QUANTUM NANO-STRUCTURES . . . . . 3
111. INTERBAND OPTICAL TRANSITION . . . . . . . . . . . . . . . . . . . . . 11
1. Linear and Nonlinear Optical Susceptibility . . . . . . . . . . . . . . . 12
2. Spontaneous Emission of Photons . . . . . . . . . . . . . . . . . . . 21
3. Rate Equations for Laser Operutions . . . . . . . . . . . . . . . . . . 21
IV . EXCITON OPTICAL PROPERTIES . . . . . . . . . . . . . . . . . . . . . . 30
1. State Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2. Effecrive-Muss Equations . . . . . . . . . . . . . . . . . . . . . . . 36
3. Esciton- Photon Interactions . . . . . . . . . . . . . . . . . . . . . 43
4. Optirul Absorption Spectra . . . . . . . . . . . . . . . . . . . . . . 41
5 . Spontaneous Emission of Photons in Quantum Wells and Mesoscopir
Quantum Disks . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
6. Spontaneous Etnission of Photons in Quantum Disks Placed in a Planar
Microcavity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
7. The Coulomb Effect on Optiraf Gain Spectra . . . . . . . . . . . . . . . 11
V . QUANTUM-DOT LASERS. . . . . . . . . . . . . . . . . . . . . . . . . 83
1. The Efect of Carrier Relaxation Dynamics on Laser Performance . . . . . . 86
2. Efect of Homogeneous Broadening of Optical Gain on Lusing Emission
Spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
3. Bi-Esciton Spontaneous Emission mil Lasing . . . . . . . . . . . . . . 91
VLSUMMARY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
A P P E N D I X .., . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
REFERENCES. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
V
vi CONTENTS
Chapter 2 Molecular Beam Epitaxial Growth of

Self-Assembled InAslGaAs Quantum Dots
Yoshiaki Nakata. Yoshihiro Sugiyarna. and Mitsuru Sugawara
1. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
I1. THESTRANSKI-KRASTANOW GROWTH MODE . . . . . . . . . . . . . . . . 119
1. Energy-Balmce Model.for Islund Formation . . . . . . . . . . . . . . . 119
2. InAs Island Growth . . . . . . . . . . . . . . . . . . . . . . . . . 121
3. Multiple-Layer Growth and Perpendicular Alignment ofls1and.r. . . . . . . 125
4. In-Plane Alignment of Isbnds . . . . . . . . . . . . . . . . . . . . . 130
111. CLOSELY STACKED InAs/GaAs QUANTUM DOTS. . . . . . . . . . . . . . . 132
1. Close Stacking ofInAs Islands . . . . . . . . . . . . . . . . . . . . . 133
2. Photoluminescence Properties . . . . . . . . . . . . . . . . . . . . . 137
3 . Zero-Dimensional Exciton Confnmwnt Ewluuted by Diumagnetic ShiJis . . . 140
IV. COLUMNAR InAsJGaAs QUANTUM DOTS. . . . . . . . . . . . . . . . . . 143
V . SUMMARY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
ACKNOWLEDGMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . 151
REFERENCES. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
Chapter 3 Metalorganic Vapor Phase Epitaxial Growth

of Self-Assembled InGaAslGaAs Quantum Dots Emitting at 1.3 pm
Kohki Mukai. Mitsuru Sugawara. Mitsuru Egawa. and Nobuyuki Ohtsuka
I. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
I1. ATOMICLAYER EPITAXIAL GROWH. . . . . . . . . . . . . . . . . . . . 157
111. ALTERNATESUPPLY GROWTH OF InGaAs DOTSBY In-As-Ga-As SEQUENCE. . . 160
IV . ALTERNATESUPPLY GROWTH OF InGaAs DOTSBY THE In-Ga-As SEQUENCE . . . 166
1. Two Types o f A L S Dot . . . . . . . . . . . . . . . . . . . . . . . . 168
2. Multil,l e-Layer Growth . . . . . . . . . . . . . . . . . . . . . . . . 172
V . THEGROWTH PROCESS. . . . . . . . . . . . . . . . . . . . . . . . . 176
VLSUMMARY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
REFERENCES. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
Chapter 4 Optical Characterization of Quantum Dots

Kohki Mukai and Mitsuru Sugawara
I. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
11. LIGHTEMlSSlON FROM DISCRETE ENERGY STATES . . . . . . . . . . . . . . 185
.
1. Photoluminescence Photoluminescence Excitation. und Electroluminescence
Spectru . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
2. Wufer Mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
3 . Microprobe Photoluminescence . . . . . . . . . . . . . . . . . . . . 192
111. CONTROLLABILITY OF QUANTUM CONFINEMENT . . . . . . . . . . . . . . . 196
1. Two Methods of Controlling Quuntizud Energies . . . . . . . . . . . . . 196
2. Magneto-Optical Spectroscop!. . . . . . . . . . . . . . . . . . . . . 200
IV. RADIATIVEEMISSION EFFICIENCY . . . . . . . . . . . . . . . . . . . . . 201
V . SUMMARY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
REFERENCES. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
CONTENTS vii
Chapter 5 The Photon Bottleneck Effect in Quantum Dots

I . INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
I1 . A MODELOF THE CARRIER RELAXATION PROCESS
IN QUANTUM DOTS. . . . . .
I11. EXPERIMENTS ON LIGHTEMISSION AND CARRIER RELAXATION I N QUANTUM-DOT
DISCRETE ENERGYLEVELS. . . . . . . . . . . . . . . . . . . . . . . . 214
1. Electrolurninescence Spectra . . . . . . . . . . . . . . . . . . . . . . 215
2. Time-Resolved Photoluminescence . . . . . . . . . . . . . . . . . . . 217
3 . Simulation of Electroluminrscencr Spectra . . . . . . . . . . . . . . . . 225
IV . INFLUENCE OF THERMAL TREATMENT. . . . . . . . . . . . . . . . . . . 229
I . Change in Emission Spectra ufier Annealing . . . . . . . . . . . . . . . 229
2. Competition between Carrier Relasation and Recombination . . . . . . . . 231
V. SIMULATION OF LASER~ R F O R M A N C EINCLUDING
THE AUGER RELAXATION
PROCESS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
VLSUMMARY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
Chapter 6 Self-Assembled Quantum Dot Lasers

Hajime Shoji
I . INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
I1. FUNDAMENTAL PROPERTIES OF QUANTUM-DOT LASERS. . . . . . . . . . . . 243
1. Gain Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . 244
2. Threshold Current . . . . . . . . . . . . . . . . . . . . . . . . . . 246
3 . Dynamic Characteristics . . . . . . . . . . . . . . . . . . . . . . . 248
111. FABRICATIONOFSELF-ASSEMBLEDQUANTUM-DOTLASERS . . . . . . . . . . 249
1. Fabrication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
2. Device Characteristics . . . . . . . . . . . . . . . . . . . . . . . . 255
3. Limiting Factors of laser Performance . . . . . . . . . . . . . . . . . 267
IV . KEYTECHNOLOGIES FOR THE NEXTERA . . . . . . . . . . . . . . . . . . 269
1. Closely Stacked Quantum-Dot Lasers . . . . . . . . . . . . . . . . . . 270
2. Columnar Quantum-Dot Lasers . . . . . . . . . . . . . . . . . . . . 273
3. Lorig- Wavelength Quantum-Dot Losers . . . . . . . . . . . . . . . . . 276
4 . Quantum-Dot Vertical-Cuvit.v Sur:fiice-EmittingLasers . . . . . . . . . . 279
V. CONCLUSION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
ACKNOWLEDGMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . 283
REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
Chapter 7 Applications of Quantum Dot to Optical Devices

Hiroshi Ishikawa
I. INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
I1. PROPERTIES OF QUANTUM DOTS . . . . . . . . . . . . . . . . . . . . . 288
1. The Quantuni Dot us a Two-Level System . . . . . . . . . . . . . . . 288
2. Attractive Features of Quantum Dots.for Device Application . . . . . . . . 294
I11. QUANTUM DOTSFOR VERYHIGH SPEEDLIGHTMODULATION . . . . . . . . . 295
1. The Needfor High-Speed. Low- Wuwlength-Chirp Light Sources . . . . . . 295
viii CONTENTS
2. Direct Modulation of Quantum-Dot Lasers . . . . . . . . . . . . . . . 298

3. The Quantum-Dot Intensity Modulator . . . . . . . . . . . . . . . . . 302
IV . QUANTUM DOTSAS A NONLINEAR MEDIUM . . . . . . . . . . . . . . . . 303
1. The Need for Large Nonlinearity with a Large Bandwidth . . . . . . . . . 303
2. Analysis of xf3 . . . . . . . . . . . . . . . . . . . . . . . . . . . 306
3. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
V . PERSISTENT HOLEBURNING MEMORY. . . . . . . . . . . . . . . . . . . 314
1. Persistent Spectral Hole Burning Memory Using Quantum Dots . . . . . . 314
2. Experimenral Results . . . . . . . . . . . . . . . . . . . . . . . . 316
3. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
VI . SUMMARY AND PERSPECTIVES ON QUANTUM-DOT OPTICAL DEVICES. . . . . . 319
1. Trends in Optoelectronics . . . . . . . . . . . . . . . . . . . . . . . 320
2. Uses for Quantum-Dot Optical Devices . . . . . . . . . . . . . . . . . 321
ACKNOWLEDGMENT. . . . . . . . . . . . . . . . . . . . . . . . . . . 321
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
REFERENCES 321
Chapter 8 The Latest News

Mitsuru Sugawara. Kohki Mukai. Hiroshi Ishikawa. Koji Otsubo.
and Yoshiaki Nakata
I. LASING WITH LOW-THRESHOLD CURRENT AND HIGH-OUTPUT POWER
FROM COLUMNAR-SHAPED QUANTUM DOTS . . . . . . . . . . . . . . . . 325
I1. EFFECTOF HOMOGENEOUS BROADENING OF SINGLE-DOT OPTICAL GAIN
ON LASING SPECTRA. . . . . . . . . . . . . . . . . . . . . . . . . . 328
111. QUANTUM DOTSON INGAASSUBSTRATES . . . . . . . . . . . . . . . . 331
IV . QUANTUM DOTSEMITTING AT 1.3 pm GROWN BY Low GROWTH RATES
AND WITH AN INGAASCAP . . . . . . . . . . . . . . . . . . . . . . . 333
V . REDUCED-TEMPERATURE-INDUCED VARIATIONOF SPONTANENOUS EMISSION
IN ALTERNATESUPPLY (ALS) QUANTUM DOTSCOVERED
BY In,.,Ga, ,As . . . . . . . . . . . . . . . . . . . . . . . . . . . . 336
REFERENCES. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
Preface
This volume is devoted to the crystal growth, optical properties, and

optical-device applications of self-assembled InGaAs quantum dots -one
of the major areas of semiconductor research today. The quantum dots’
atom-like density of states should significantly improve optical performance,
especially in semiconductor lasers, and contribute to the development of
new optical devices. However, because of the difficulty of their fabrication,
quantum-dot optical devices, first posited by Arakawa in 1982, were little
more than a dream until the advent of self-assembled quantum dots grown
via molecular beam epitaxy (MBE) and metal organic vapor phase epitaxy
(MOVPE) in the early nineties. Since then, the field has grown rapidly. The
performance of quantum-dot lasers has now reached that of the well-
established quantum-well lasers, and further progress is expected in the near
future.
The authors, all of whom work for Fujitsu Laboratories in Atsugi, Japan,
are leaders in quantum-dot research for optical device applications. They
have not only achieved room-temperature quantum-dot laser operation
with low threshold current and high efficiency but have also gained wide
knowledge of optical device designs and processes, laser properties, device
simulation, theoretical aspects of optical characteristics, MOVPE and MBE
growth techniques, and optical and crystallographic characterization.
Here, the Fujitsu group provides both general knowledge and state-of-the
art research results. They also discuss additional research needed to realize
high-performance optical devices. For these reasons, this volume not only
represents a milestone in semiconductor technology, treating many topics
not been treated elsewhere; it also contributes considerably to the field’s
further development.
Chapter 1 provides a basic theoretical background for the optical prop-
erties of semiconductor quantum nano-structures in particular, those of
quantum wells and quantum dots and the differences between them. Also
covered are interband optical transition, exciton optical properties, and
quantum-dot laser simulations primarily focusing on carrier relaxation.
ix
X PREFACE
Experimental results and numerical calculations are included to improve

understanding of the derived theoretical expressions. The formulae are
presented step by step, without neglecting numerical coefficients, and em-
ploy the MKSA system of units. This should help readers calculate the
optical response of semiconductor nano-structures on their own.
Chapter 2 focuses on the molecular beam epitaxial growth of self-
assembled InAs islands on GaAs substrates through the Stranski-Kras-
tanow (SK) growth mode. These islands work as quantum dots because they
are smaller than the exciton optical Bohr radius in this material system. The
authors’ method of stacking islands perpendicularly creates closely stacked
and columnar-shaped quantum dots. The perpendicular stacking increases
the dot size because of the electronical coupling of each stacked island in
the growth direction, enabling tuning of the emission wavelength as well as
narrowing of the spectrum line width caused by the island-size fluctuations.
A comparison of the structural optical properties of different types of dots
proves the advantages of the columnar-shaped type. These advantages are
demonstrated in Chapters 6 and 8.
Chapter 3 introduces another original type of self-assembled quantum
dot, grown by MOVPE, whose growth sequence is unique in that the
group-111 and group-V precursors are supplied alternately with an amount
corresponding to one monolayer or less. This explains the dot’s name: ALS,
which stands for “alternate supply.” One of the most striking features of
ALS dots is that they emit at a wavelength of 1.3 pm at room temperature,
which is the zero-dispersive wavelength of silica optical fiber used in the
optical data transmission system. In addition, their emission spectrum
linewidth is very narrow and provides a series of distinct lines peculiar to
three-dimensional quantum confinement. These two features make ALS
dots very attractive for practical application.
Chapter 4 deals with optical characterization of quantum dots, focusing
primarily on the ALS type. Through various diagnostic techniques, the
unique properties of ALS dots are presented, including long emission
wavelength, emission spectra with multiple peaks from discrete energy
states, harmonic-oscillator type confinement potential, large wavelength
tunability between 1.2 and 1.5 pm through size control, and carrier lifetimes
through radiative and nonradiative recombinations.
Chapter 5 presents experimental studies on the carrier dynamics in self-
assembled InGaAs/GaAs quantum dots. The physics of carrier relaxation in
quantum dots has been studied intensively ever since quantum dots were
discovered to have promising features for optoelectronic device application.
One of the primary concerns of many researchers has been whether carrier
relaxation into quantum-dot discrete states is significantly slowed because
of the lack of phonons needed to satisfy the energy conservation rule. This
is known as the phonon bottleneck and at one time was considered merely
PREFACE xi
a theoretical problem; however, the recent advent of self-formed quantum

dots has enabled researchers to evaluate it experimentally. The chapter
provides evidence of retarded carrier relaxation and discusses how the
phonon bottleneck works in quantum dots. In addition, the discussion on
how retarded carrier relaxation affects quantum-dot laser performance given
in Chapter 1 is reopened here, taking into account the Auger relaxation
process.
Chapter 6 describes semiconductor lasers with self-assembled quantum
dots in the active layer. After a brief theoretical discussion of what can be
expected from quantum-dot lasers, the chapter moves on to their present-
stage performance as well as their fabrication processes, lasing characteris-
tics, and problems to be solved. Finally discussed are several key technolo-
gies for future improvement of quantum-dot lasers, including new trends
such as closely-stacked and columnar-shaped structures, long wavelength
emission, and vertical-cavity surface-emitting lasers (VCSELs).
Chapter 7 reviews some of the useful properties of quantum dots and how
they might apply to laser technology. Problems associated with the conven-
tional technologies for long-distance high-bit-rate communication systems
are reviewed, as are potential solutions such as direct modulation of
quantum dot lasers and the feasibility of external modulators using quan-
tum dots. To determine the feasibility of quantum dots as a nonlinear
medium, the authors perform a trial analysis of the dots’ third order
nonlinear susceptibility. They also discuss the use of quantum dots for
high-density optical memory through persistent spectral hole burning, and,
finally, the role of quantum-dot-based devices in future optoelectronics.
Chapter 8 presents the latest developments in quantum-dot research:
lasing with low-threshold current and high-output power from columnar-
shaped quantum dots; the effect of homogeneous broadening of single-dot
optical gain on lasing spectra; quantum dots on InGaAs substrates; quan-
tum dots emitting at 1.3 pm grown by low growth rates and with an InGaAs
cap; and reduced-temperature-induced variation of spontaneous emission
wavelength in ALS quantum dots covered by In,,,Ga,,,As.
The success of this volume, as well as the success of the field, will be
largely due to its outstanding contributors. The authors wish to thank Dr.
Hajime Ishikawa, Dr. Naoki Yokoyama, Dr. Shigenobu Yamakoshi, Dr.
Hajime Imai, and Dr. Kiyohide Wakao of Fujitsu Laboratories Ltd. for
their encouragement and strong support. As volume editor, I wish to thank
Miki Sugawara for her invaluable practical support for this volume and her
continuous encouragement during my research.
MITSURU SUGAWARA
This Page Intentionally Left Blank
List of Contributors
Numbers in parenthesis indicate the pages on which the authors’

contribution begins.
MITSURUEGAWA,(1 55) Optical Semiconductor Devices Laboratory, Fujitsu

Laboratories Limited, Atsugi, Kanagawa, Japan
HIROSHI ISHIKAWA, (287, 325) Electron Devices and Materials Laboratory,
Fujitsu Laboratories Limited, Atsugi, Kanagawa, Japan
KOHKIMUKAI,(155, 183, 209, 325) Optical Semiconductor Devices Labora-
tory, Fujitsu Laboratories Limited, Atsugi, Kanugawa, Japan
YOSHIAKI
NAKATA, (1 17, 325) Quantum Electron Devices Laboratory, Fujitsu
NOBUYUKI OHTSUKA, (1 55) Integrated Materials Laboratory, Fujitsu Laboru-
tories Limited, Atsugi, Kanugawa, Japan
KOJI OTSUBO,(325) Electron Devices and Materials Laboratory, Fujitsu
HAJIMESHOJI,(241) Optical Semiconductor Devices Laboratory, Fujitsu
MITSURUSUGAWARA, (1, 117, 155, 183, 209, 325) Optical Semiconductor
Devices Laboratory, Fujitsu Laboratories Limited, Atsugi, Kanagawa,
Japan
YOSHIHIROSUGIYAMA, (1 17) Quantum Electron Devices Laboratory, Fujitsu
...
Xlll
SEMICONDUCTORS AND SEMIMETALS. VOL. 60
CHAPTER 1
Theoretical Bases of the Optical Properties of

Semiconductor Quanturn Nano-St ructures
Mitsuru Sugawara
OPTICAL SEMICONDUCTOR DEVICES
LABURATORY
FUJITSULABORATORES LTD.
ATSUGI.KANAGAWA,
JAPAN
. . . . . . . . . . .
I. INTRODUCTION . . . . . . . . . . . . . . . . . 1
11. ELECTRONIC . . . .
NANO-STRUCTURES
STATES OF SEMICONDUCTOR QUANTUM 3
111. INTERBAND OPTICAL TRANSITION . . . . . . . , . . . , . , . . , , . . 11
1. Linear and Nonlinear Opticul Susceptibiliiy . . . . . . . . . . . . . . . 12
2. Spontaneous Emission of Photons . . . , . . . , , . , . . , . , . , . 21
3. Rate Equations for Laser Opcrutions . , . . . . , , . . . , . . . , . 21
IV. EXCITON OPTICAL PROPERTIES . . . . . . , . . . , , . . . . . . . . . 30
1. State Vectors . . . . . . . . . . . . , . . . . . . . . . . . . . . 32
2. Effective-Mass Equations . . . . . . . . , , . . . . . . . . . . . . 36
3. Exciton-Photon Interactions . . . . . . . , , . . . . . . . . . . . . 43
4. Opticul Absorption Spectra . . . . . , . , . . . . . . . . . . . . . 47
5. Spontaneous Emission of Photons in Quantum Wells and Mesoscopic
Quantum Disks . . . . . . . . . . . . . . . . . , . . , , . . , . 55
6. Spontaneous Emission of Photons in Quantum Disks Placed in a Planar
Microcavity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
I. The Coulomb Effect on Opticd Guin Specisu . . . . . . , , . . , . . . I7
V. QUANTUM-DOT LASERS . . . . . . . . . . . . . . . . . , . , , , . . 83
1. The Effect of Carrier RdcJXUliOn Dynunnc.s on Luser Per:/Om~nnce . . . . . 86
2. Efect of Homogeneous Brmlrning of Opticul Gain on Lasing Emission
Spcctsu.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
3. Bi-Exciton Spontaneous Emission unrl Lasing . . . . . . . . . . . . . . 97
VI. SUMMARY. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
Appendix.. . . . . . . . . . . . . . . . . . . . . . . . . . , , . . . . . 107
References . . . . . . . . . . . . . . . . . . . . . . . . , . . . . . , . . 112
I. Introduction
Research on synthesized semiconductor quantum nano-structures was

initiated by the discovery of superlattices by Esaki and Tsu (1969, 1970).
These researchers observed quantum effects in one-dimensional periodic
1
Copyright I 1999 by Academic Press
All rights of iepioduLtron in any lorm reserved
ISBN 0-12-752169-0
ISSN 0080-8784!99 $70 00
2 MITSURUSUGAWARA
structures consisting of alternating, differing layers, whose thickness was less

than that of the electron mean free path. They named their periodic
structure a semiconductor superlattice and predicted negative resistance and
Bloch oscillations for electrons moving toward the superlattice potential.
Subsequent experimental proof of the negative resistance (Chang et al.,
1974) and the finding of enhanced exciton resonance in quantum wells
(Dingle, 1975) had a great impact on the development of this research field.
The concept rapidly became the standard for producing and designing
semiconductor multilayers with desirable electronic and optical properties.
Semiconductor quantum wells, in which a thin semiconductor film is
sandwiched between different materials via heterojunctions, confine electron
motion in the two-dimensional thin-film plane. This two-dimensionality
gives rise to new optical properties not observed in bulk materials, such as
optical absorption and gain spectra peculiar to the steplike density of states
(Weisbuch, 1987), strong exciton resonance clearly observable even at room
temperature (Ishibashi et al., 1981; Miller et al., 1982a), large optical
nonlinearity (Miller et al. 1982b, 1983; Chemla et al., 1984) and an electric
field-induced energy shift of the resonance, called the quantum-confined
Stark efect (Chemla et al., 1983; Miller et al., 1984a, 1985a). These properties
led the way to a variety of new optical devices: quantum-well lasers (Tsang,
1981), high-speed optical modulators (Wood et al., 1984), optical switches
(Miller et al., 1984b), optical bistable devices such as self-generated electro-
optic effect devices (SEED’S) (Miller et al., 1985b) and the like. Quantum-
well lasers and modulators, in particular have become conventional, widely
used devices in optical transmission systems and optical data storage.
Multi-dimensional quantum-confinement structures, such as quantum
wires and quantum dots, are expected to further improve quantum-effect
optical devices. Their most valuable application will be quantum-dot and
quantum-wire lasers, which will be far superior to quantum-well lasers, as
predicted by Arakawa and Sakaki (1982). This is primarily because, as the
dimension of materials is reduced, the density of states for electrons changes
to give a narrower optical transition spectrum, making the light-matter
interaction more efficient.
In line with the prediction of Arakawa and Sakaki, great efforts have been
made to form lateral confinement potentials in addition to quantum wells,
mostly using high-resolution photo- or electron-lithography and crystal
growth on masked or etched substrates (see references in Chapter 2).
However, crystals fabricated by these artificial processes have suffered from
structural problems such as being too large to produce observable quantum
effects and having low crystal quality at the interfaces. A breakthrough
occurred in the early 1990s with the formation of self-assembled quantum
dots during the highly strained growth processes of molecular beam epitaxy
PROPERTIES
1 OPTICAL OF SEMICONDUCTOR NANO-STRUCTURES3
QUANTUM
(MBE) and metal-organic vapor phase epitaxy (MOCVD), which created

high-quality quantum dots suitable for optical devices (see Chapters 2 and
3). The atom-like density of states in quantum dots has been observed (see
Chapter 4) and room-temperature quantum-dot laser operations have been
realized (see Chapter 6). Promising applications of quantum dots to other
optical devices are discussed in Chapter 7. We are now entering a new world
of zero-dimensional physics and devices, after almost 25 years of research
on two-dimensional quantum wells.
This chapter provides a basic theoretical background for the optical
properties of semiconductor quantum nano-structures. One of its primary
purposes is to show how the optical properties of quantum dots differ from
those of quantum wells. The chapter is written for students who have
finished a course on solid-state physics as well as professional researchers.
In particular, experimental researchers who are struggling with difficult
papers and textbooks written by theorists should benefit from the informa-
tion provided here. The formulae are presented step by step and without
neglecting numerical coefficients; moreover, the MKSA system of units is
used to help readers calculate the optical response of semiconductor
nano-structures by themselves. Experimental results and numerical calcula-
tions are included to improve understanding of the derived theoretical
expressions.
11. Electronic States of Semiconductor Quantum Nano-Structures
Recent progress in nano-scale growth techniques such as MBE and

MOCVD has enabled us to grow high-quality semiconductor quantum wells
and quantum dots. Figure 1.1 shows cross-sectional transmission electron
microscope (TEM) photographs of (a) In,,,,Ga,~,,As/InP strained quan-
tum wells and (b) self-assembled In,.,Ga,.,As/GaAs quantum dots, both
grown by MOCVD. The quantum well consists of a 9.7-nm In,,,,Ga,~,,As
well layer sandwiched by InP. It is lattice-mismatched to InP with an
in-plane strain of -0.81%. Electrons in the conduction band and holes in
the valence band are spatially confined in the In,.,,Ga,,,,As, which acts as
a potential well; that is, the quantum well has a type-I band lineup. The
self-assembled quantum dots are grown by our original process to alter-
nately supply InAs and GaAs monolayers. The InAs supplied to the GaAs
substrate has about 7% lattice mismatch to the GaAs substrate, leading to
the self-assembling of In,~,Ga,,,As clusters instead of a film to relax strain
energy (see Chapters 3 and 4). These dots differ from conventional self-
assembled In(Ga)As/GaAs quantum dots grown via the Stranski-Krastanov
4 MITSURUSUGAWARA
InP
InGaAs
20 nm
H
FIG. 1.1. Cross-sectional transmission electron microscope (TEM) photograph of (a)

In,,,,G,,,,As/InP strained quantum wells. and (b) self-assembled In, ,Ga,,,As/GaAs quantum
dots, both grown by MOCVD. The self-assembled quantum dots are grown by our original
growth sequence of alternately supplying InAs and GaAs monolayers (see Chapter 3).
(SK) mode. In particular, they emit at a much longer wavelength of 1.3 pm

at room temperature- a suitable wavelength for optical communication
systems- and they are often accompanied by a thicker quantum-well
surrounding layer, whose counterpart in the SK growth mode is a wetting
layer.
Quantum nano-structures and their density of states are illustrated in Fig.
1.2. When the size of the crystal is reduced to the nanometer scale in one
direction and the crystal is surrounded by other crystals acting as potential
barriers, the freedom of electron movement is lost in that direction. The
potential height corresponds to the band offset between the two crystals in
the conduction and valence bands. Electrons in the quantum well move in
the x-y plane; those in the quantum wire move in the x direction; and those
in the quantum dot are completely localized. This confinement results in the
quantization of the electron energy and in the variations of the electron
density of states, which is the most remarkable and significant benefit of
low-dimensional semiconductor technology.
In this section, we briefly survey how the energy, the wave function, and
the density of states depend on the dimension of quantum nanostructures,
using an elementary quantum-mechanical approach.
The effective-mass approximation effectively describes the electronic state
of bulk semiconductors. In semiconductor quantum wells, this approxi-
mation has also succeeded in calculating quantum energy shifts as a function
of the well-layer width, as well as various optical properties arising from the
quantized states. The main assumption of the effective-mass approximation
is that the envelope wave function does not vary a great deal in the unit cell
with a length scale of subnanometers; this assumption holds up in quantum
nano-structures like those in Fig. 1 .l. Assuming a parabolic band dispersion,
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR NANO-STRUCTURES5
QUANTUM
Bulk
r3L V=L3
Quantum wire
N,: Areal density
Energy
=t Energy
X
Quantum well Quantum dot
OW
Well Barrier
D = L2
Energy Energy
FIG. 1.2. Schematic view and graph of quantum nano-structures and their density of states.
band-edge electron states of semiconductors can be described by the

effective-mass equation as (Haken, 1973, for example)
1
V2 + V(r) Fk(r) = EFk(r)
Here, m* is the effective mass; h is the Planck's constant divided by 2.n;

r = (x,y, z ) is the electron position vector; V(r) is the confinement potential,
Fk(r) is the envelope wave function; and E is the electron energy. Using the
periodic part of the band-edge Bloch function normalized in the unit cell,
uo(r), the normalized electron wave function close to the band edge is
written simply as
6 MITSURUSUGAWARA
where s2 is the unit-cell volume. The 111-V and 11-VI semiconductor

materials have s-like conduction-band-edge states and p-like valence-band-
edge states that consist of the heavy-hole state, the light-hole state and the
spin split-off state under spin-orbit interactions. See the appendix to this
chapter for uo in the conduction and valence bands.
According to the k * p perturbation theory, the band mixing of these states
determines the effective mass, that is, the band dispersion of the conduction
and valence bands (Kane, 1956, 1957). In the k - p calculation, the electron
wave function is expanded on the basis of the band-edge state functions, and
the eigenvalue problem is solved by taking into account the k - p terms up
to the second order. In the design of state-of-art quantum-well devices, such
as strained quantum-well lasers, a knowledge of band mixing and resultant
band nonparabolicity is indispensable in engineering optimal band struc-
tures for high performance (Bastard et al., 1991, Chuang, 1991, Sugawara et
al., 1993a, 1994). Readers interested in this topic should see Chuang’s
well-written textbook (1995). This so-called band engineering will also be
done in quantum dots if we can learn how to control the dot’s fabrication.
Equation (1.1) uses a constant and isotropic effective mass, whether the
band is conduction or valence, to uncover the primary role of low-dimen-
sional quantum confinement. Assuming barrier potentials with infinite
height, the eigenenergies, the envelope wave functions, and the density of
states are given as follows.
Bulk Materials
Setting V(r) = 0, Eq. (1.1) gives the energy of
h2k2
E = E(k) = ~
2m*
and the envelope function of
1
Fdr) = -exp( - i k r) - (1.4)
Jv
where V = L3 is the crystal volume. The envelope wave function is normal-
ized in the crystal. The wave vector of k = ( k x ,k,, k,) satisfies the periodic
boundary conditions as k , = (2nn,)/L, k, = (2nn,)/L, and k, = (2nn,)/L;
n x , n y , and n, are integers. The density of states per unit volume, which is
the number of states between the energy of E and E + dE, is given as
1 OPTICAL
PROPERTIES
QUANTUM
D(E) = hxG[E(k) - El
Vk
-
-- x-
v '
(243
x R, [: dkk2G[E(k) - El
Here, R, = 4n is the integration in the angular part. The factor of two

represents the degeneracy due to spin.
Quantum Wells
The confinement potential for the square quantum well is given as V(r) =
V(z), where V ( z ) = co when IzI 3 L,/2; V(z) = 0 when IzI < L,/2; and L, is
the well width (Fig. 1.3(a)). Under the infinite potential height, the wave
function vanishes at the boundaries between the well and barrier layers. The
eigenvalue is
with the corresponding normalized wave function of
1
Fk(r)= -cp,,(z) exp(--ikll*rll)
@
and
Here, D = L2 is the area of the quantum well; k = (k", n,n/L,); the in-plane
wave vector is kll = ( k x ,k y ) ; r = (rll, z); rll = (x,y); and n, = 1,2,3,. . . . The
wave function in the z direction becomes stationary. The minimum energy
and the energy separation between each quantized state increase as the well
width decreases. Figure 1.3(b) illustrates the parabolic dispersion curve in
the conduction and valence bands. The conduction-band electron effective
mass is written as mQ. The superscript 11, indicates that the mass is for the
in-plane carrier movement. Since the valence-band effective mass is negative
in most cases, it is more convenient to introduce a positive particle of a hole
8 M~TSURU
SUGAWARA
'T Conduction
Et
transition
________----____---
(4
I
I
band
(b)
I
FIG. 1.3. (a) Confinement potential for the square quantum well, V ( z ) , and the quantized
energies for an electron in the conduction band. En*-,and for a hole in the valence band, En&=.
E, is the band gap of the well material. (b) The band dispersion of the conduction band and
valence band in the quantum-well plane. mr" is the electron effective mass, mrll is the hole
effective mass, and m:I1 is the reduced mass. Interband optical transition occurs almost
perpendicularly in the dispersion curve.
with the opposite effective mass of m;. Then, Eq. (1.6) holds also for holes
with the energy origin at the valence-band maximum and with the energy
positive in the downward direction.
EneZ(or E,J and En,= (or En,J in Fig. 1.3(a) are quantized energies for
electrons and holes, respectively. The quantized wave functions are de-
scribed as cpne,(z) (or cp,,Jz)) for electrons, and cp,,,(z) (or cp.,,(z)) for holes.
Letting LQW be the quantum-well thickness, that is, the sum of the well and
barrier regime thickness- the density of states per unit volume is
where R, = 2x is the integration in the angular part and O(x) is the

Heaviside step function (O(x) = 1 for x 2 0 and O(x) = 0 for x < 0). As
shown in Fig. 1.2, the density of states is steplike.
1 OPTICAL
PROPERTIES
QUANTUM
Quan rum Wive
The confinement potential for the square quantum wire is given as V(r) =
+
V(y) V(z), where V(y) is the additional confinement potential, with
V ( y ) = GO under lyl 3 L,/2 and V ( y ) = 0 under lyl < LJ2; and L, is the
y-direction length of the wire cross-section. The eigenvalue to the infinite
barrier model is
c = E ( k ) = ,h2k2
,.En,,+En==~[k:+(~)2+~~)2] (1.10)
2m*
where EnYhas the same form as that of En=of Eq. (1.6) and ny,II,= 1,2,3.. . .
The electron energy further increases due to the additional confinement. The
corresponding normalized wave function is
Here, L is the length of the quantum wire, and k = ( k x ,nyn/Ly,n,n/L,).

Letting NWi be the area density of the quantum wires (the number of
quantum wires divided by the quantum-wire region area in the y-z plane),
the density of states per unit volume is
2N .
D(E) G[E(k) - E l
ny,n,.k,
2Nwi
=--X-XQ1
L 2n
L
1
ny,nz
1; dkG[E(k) - El
-
Nwi
~~
J2m* c 1
[eV- K~] (1.12)
n h ny.nz J E- E . -
~ E,,=
where Q, = 2.
Quantum Dots
Assuming an infinitely high potential for all directions, the confinement

+
potential becomes V(r) = V ( x ) V ( y ) + V(z). Thus, we get
E = E(k) = Enx + EnY+ En== __

2m*
h2 [(y)’+ (?>’+ (y)’] (1.13)
10 MITSURUSUGAWARA
where n,, ny,n, = 1,2,3.. . and k = (n,n/L,, nyz/Ly,n,z/L,). The energy

states are completely discrete. The corresponding wave function is
Letting N , be the volume density of quantum dots, the density of states is

a series of delta functions as
The interband optical transition between the conduction band and the
valence band reflects the density of states derived above. Figure 1.4(a) shows
the optical absorption spectra of the quantum well of Fig. l.l(a) (Sugawara,
1993a). The absorbance is measured by
(1.16)
where li, is the intensity of the incident light beam; I,, is that of the
transmitted light beam; and N , is the number of well layers-10 in this
case. Clearly observable is the steplike optical absorption spectrum due to
the two-dimensionality of quantum wells. (Strictly speaking, this clear
steplike absorption continuum arises from the parabolic nature of the
ground-state heavy-hole valence band, which is caused by enhanced split-
ting against the light-hole valence subband under the in-plane compressive
strain.) In addition, we see peaky spectra due to the electron-heavy-hole
exciton resonance (le-hh) at the absorption edge and the electron-light-
hole resonance (le-lh) on the absorption continuum at a shorter
wavelength. We also observe higher-order electron-heavy-hole exciton
resonance, as (2e-hh). The numbers of 1 and 2 represent the z-direction
quantum number, nz, which should be equal between the conduction and
valence bands, for the allowed optical transition (see the appendix).
Figure 1.4(b) shows the photoluminescence spectra of the quantum dots
of Fig. l.l(b) (Mukai and Sugawara, 1999). As the excitation laser power
increases, peaky emissions appear one by one, caused by the interband
transition between the discrete energy levels of the conduction and valence
bands. This is experimental proof of the zero-dimensionality of electrons and
holes in quantum dots.
In the following sections, formulae to describe the optical response of
quantum nano-structures are derived.
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 11
NANO-STRUCTURES
Pe-hh 1e-hh
0
t le-lh
- 0
1.2 1.4 1.6
Wavelength (pm)
(4
1 1.2 1.4 1.6

Energy (eV)
(b)
FIG. 1.4. (a) Optical absorption spectrum at 77K of the In,,,Ga,,,,As/InP strained
quantum wells of Fig. l,l(a). The steplike optical absorption spectrum due to the two-
dimensional character of the density of states is clearly observed. Resonant spectra are due to
the ground-state electron-heavy-hole exciton (le-hh) at the absorption edge, the electron-
light-hole exciton (le-lh) on the absorption continuum at the shorter wavelength, and the
higher-order electron-heavy-hole exciton (2e-hh) (from Sugawara et al., 1993a. Copyright
1993 by The American Physical Society). (b) Photoluminescence spectra at 5 K of the
In,,,Ga,,,As/GaAs quantum dots of Fig. l.l(b) for various excitation level. Emissions from the
discrete energy levels due to three-dimensional quantum confinement are clearly observed
(from Chapter 4).
111. Interband Optical Transition
The interband transition of an electron between the conduction and

valence bands occurs by the transfer of the interband energy between the
electron and the photons. When semiconductors have a direct band gap, the
optical transition between the band edges occurs at a highly efficient rate,
12 MITSURUSUGAWARA
making the materials quite useful for optical devices. “Direct” means that
both the conduction-band minimum and the valence-band maximum are at
the same electron wave vector position in the band dispersion diagram,
making possible the electron-hole recombination by the emission of
photons, with a wave vector selection rule satisfied. This is also the case in
quantum nano-structures consisting of direct band-gap materials and hav-
ing type-I band lineup, where both electrons and holes are confined in the
materials. A variety of material systems fall into this category, such as
GaAs/AlGaAs, InGaAsP/InP, InGaP/AlGaInP, and InGaN/GaN.
The optical responses of semiconductors are categorized as optical
absorption, stimulated light emission, spontaneous light emission, and
various nonlinear optical processes. All are applied to various optical
devices-for example, optical absorption to photodetectors, modulators,
and solar batteries; stimulated emission to lasers and optical amplifiers;
spontaneous emission to light-emitting diodes; nonlinear processes to
switches and wavelength conversion devices. Depending on the required
photon energy in practical application fields such as optical transmission
systems and optical storage, we choose appropriate semiconductor ma-
terials, from infrared to red to blue to violet. As seen in Fig. 1.4, the optical
response varies, depending on the confinement dimension.
This section presents the theoretical bases of the interband optical
transitions in quantum nano-structures. In addition, a prescription to
simulate lasing operation using rate equations is briefly summarized. The
excition effect due to the electron-hole Coulomb interaction is not taken
into account here, but is treated in Section IV. By comparing the results in
this section with those in Section IV, readers will see how the exciton effect
influences semiconductor optical response.
1. LINEARAND NONLINEAR
OPTICALSUSCEPTIBILITY
Let us derive linear and nonlinear optical susceptibilities using the

density-matrix theory, which plays an important role in any discussion of
the optical properties of materials in quantum electronics. The optical
absorption and gain coefficient can be derived from the susceptibilities. A
great merit of using the density-matrix approach is that the dephasing of
light-induced polarization can be incorporated, giving us the spectrum
profile of the linear and nonlinear response. Chapter 7 presents the calcula-
tion of the dephasing rate due to electron-electron and hole-hole scattering
in quantum dots. Readers unfamiliar with the density-matrix approach
should refer to the quantum optics textbook of Meystre and Sargent 111
(1990) and other similar works.
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 13
NANO-STRUCTURES
By taking )I as the electron wave function of the material system and

expanding it using the complete sets of the conduction-band and the
valence-band states, the density matrix is given as
Here, I represents the band index, k represents the electron wave vector, and
12, k) represents the Dirac state vectors to describe the electron states in the
conduction and valence bands. The dimension of materials need not be of
concern for the moment. The state vector satisfies the orthogonality relation
Of (A‘, k’II, k) = 6;.;,6kk. and the complete relation of Cj.,k 12, k)(A, kl = 1.
By taking the time differential of Eq. (1.17) and using the time-dependent
Schrodinger equation, the density operator satisfies the equation of motion
as
i
aP(t)
-=
at
--[H,+H,,p(t)]
h
+( - (1.18)
where H , is the electron unperturbed Hamiltonian, and H r is the Hamil-

tonian representing the interaction between electrons and photons. The
second term of the righthand side added phenomenologically represents
relaxation due to the incoherent scattering processes. This relaxation is
characterized by the longitudinal relaxation time constant for the diagonal
density-matrix components and the transverse relaxation time constant for
the nondiagonal density matrix components.
Using the Dirac representation, the Hamiltonian for the kinetic energy of
electrons is given as
(1.19)
The interaction Hamiltonian with the electromagnetic field is given as
where e is the electron charge, m, is the electron mass, and p is the

momentum operator. A,&, t ) is the vector potential for the optical mode
with the polarization unit vector, e,, the wave vector, q, and the frequency,
14 MlTSURU SUGAWARA
w. It is written in a classical manner as
A&, + C.C.]
t ) = fe,[Aq,u(w)e''~"-"') (1.21)
The electric field is related to the vector potential as
Using the linear susceptibility, x ( l ) , and the third-order susceptibility, x(~),

the polarization of the isotropic material is defined as
P(r, t ) = Eofl)E(r, t ) + E ~ X ( ~ t)E(r,

) E ( ~t)E(r,
, t) (1.23)
where E~ is the permittivity of a vacuum and subscripts of q and u are

omitted for simplicity. The expressions for the polarization, fl) and are
given by the density matrix that satisfies Eq. (1.18), as shown below.
In the matrix element of Eq. (1.20), the k-vector selection rule between the
conduction-band electron vector, k, , and the valence-band electron vector
k,, holds as (see the appendix)
k, z k, = k (1.24)
Thus, the interband optical transition occurs perpendicularly in the energy

band dispersion diagram, as shown in Fig. 1.3(b).
Linear Susceptibility
The linear susceptibility and the linear optical absorption and gain
spectra are derived here. Let us use a two-band model, that is, we consider
the interband transition between one conduction band, denoted by c, and
one valence band, denoted by v. The contributions from other bands can be
added up later. One optical mode with a fixed wave vector, polarization, and
frequency is taken into account because this is a situation in which we
observe the linear optical absorption and gain of semiconductors. Thus, Eq.
(1.19) becomes
where u&k is the frequency of the conduction-band electron with k, and wsk
is the frequency of the valence-band electron with k. Equation (1.20)
becomes
H I = I*cr,k(t)lcrk)(v, kl + &,k(L)lv? k)<c, kl (1.26)
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 15
NANO-STRUCTURES
with
where
(1.28)
(1.29)
and
See the appendix for the interband transition matrix element of P:L,.k.The
density matrix of Eq. (1.17) becomes
The nondiagonal terms, pCDand pVc,are related to the polarization, and the
diagonal terms, p,, and p,,, represent the population of the conduction and
valence bands, respectively.
From Eqs. (1.18), (1.25), (1.26), and (1.32), we get the motion of the
nondiagonal and the diagonal terms as
and
16 MITSURUSUGAWARA
Here, o,,,k = w,,k - uL..k; r,, is the transverse relaxation rate due to scat-
tering; rTIis the longitudinal relaxation rate of the conduction-band elec-
tron; and r[ is the longitudinal relaxation rate of the valence-band electron.
Due to intraband scattering. the relationship of jc,(k, t ) = -jJk, t ) does
not always hold. These equations of motion can be solved by the conven-
+ + +
tional perturbation expansion of p = p(O) p ( l ) p(') . . . , where the
number in parentheses represents how many times the electron has interac-
ted with the electromagnetic field. The zero-th order term, p(O), represents
the initial state of materials. Equation (1.18) can be separated according to
the perturbation order. By expanding the density operator into the Fourier
series of p(t) = X u ,p(o')e-iu'*,we obtain the first-order perturbation term as
(1.36)
The diagonal terms, pi:) and p!:,', are zero.

The linear polarization of the system is given by tracing the diagonal
terms of the p ( ' ) ( o ) - polarization product as
(P:f:(o)/Z) = = -
Tr[p(’)(o)P] p$)(o)( jlee;rli)Qd
i.j=c.r
where P is the polarization per unit volume in the e, direction, and the
relationship of
(1.38)
is used. Qd (d = 0,1,2,3) is the inverse of the crystal volume; Q3 = V-' for

bulk materials with a volume of V ; Qz = ( D L Q W ) - ' for quantum wells with
an area of D and a width of LQw;Q1 = N w i / L for quantum wires with an
areal density of N w i and a length of L; and Qo = N , for quantum dots with
a volume density of N , (see Fig. 1.2). The term p::) is usually replaced with
the Fermi-Dirac distribution function of the valence band, f l , k , and pi:) with
that of the conduction-band, J , k , as
1
( i = c,v) (1.39)
= exp[(hw,., - Pi)/kBT] +1
1 OPTICAL
PROPERTIES QUANTUM
OF SEMICONDUCTOR 17
NANO-STRUCTURES
where p i is the quasi-Fermi level; k , is the Boltzmann's constant; and T is

temperature. The value of p i is determined for 2Ckf;.k to give the total
number of electrons for each band.
The Fermi-Dirac distribution holds when electron-electron and electron-
phonon scattering toward thermal equilibrium occurs much faster than
electron-optical interaction processes or other nonradiative recombination
processes. This is not the case in semiconductor lasers operating at high-
output power, where the carrier-photon interaction is so strong that a large
number of cavity photons extract carriers from the bands through the
stimulated emission process with the cavity photon lifetime, and carrier
relaxation cannot catch up with the extraction. This effect becomes remark-
able especially in quantum-dot lasers, where the energy levels are discrete
and the carrier relaxation lifetime is significantly slowed because of insuffi-
cient phonons to satisfy the energy conservation rule. This is known as the
phonon bottleneck. The experimentally measured relaxation lifetimes are
several tens to hundreds of picoseconds at a low excitation regime (see
Chap. 5)-definitely lower than the typical cavity photon lifetime of ten
picoseconds. Quantum-dot laser operation is simulated in Section V.
From Eq. (1.23), the relationship between the Fourier components of the
linear polarization and the electric field is
Thus, we get the first-order susceptibility as
(1.41)
Degeneracy due to spin is taken into account. The contribution of other k

states and bands is included. Taking only the resonant term, the linear
optical absorption coefficient, sc"'(w), or gain, g(’)(co), is given from the
imaginary part of Eq. (1.41) as
18 MITSURUSUGAWARA
where
( 1.43)
is the normalized broadening function and nr is the background refractive

index. The broadening obeys a Lorentzian distribution with a full width at
half maximum (FWHM) of 2r,,, which generally increases with the number
of carriers as a result of electron-electron scattering, and with temperature
as a result of electron-phonon scattering. The refractive index, including the
contribution of interband transition, is given from the real part as
(1.44)
Let us perform the summation on the wave vector in Eq. (1.42) for the
parabolic band with the dispersion of
(1.45)
where E , is the band gap;
is the reduced effective masses; ELd) is the quantized energy in the conduction
band; ELd’ is that of the valence band (see Fig. 1.3, depicting for quantum
wells); and k(d) is the electron wave number in the free direction. By
substituting the delta function for the broadening function of Eq. (1.43), we
get an analytical expression for bulk (d = 3), quantum wells (d = 2), and
quantum wires (d = 1) as (Haug and Koch, 1994)
x @(hw - E , - E$) - Eid))[.f,(w)

- f,(o)] (1.47)
where L3 = I/; L2 = D; the integrations in the angular part are R, = 471,

Q, = 2n, and Q, = 2; O(x)is the Heaviside step function; and
(i = c,v) (1.48)
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 19
NANO-STRUCTURES
For quantum dots, by omitting the wave vector from Eq. (1.42), we get
where indexes c and v represent discrete energy states in the conduction and
valence bands. By comparing the results in Section I1 with Eqs. (1.47) and
(1.49), we see that the interband optical absorption spectrum directly reflects
the density of states or, more correctly, the joint density of states between
the conduction band and the valence band. The steplike absorption continu-
um in Fig. 1.4(a) in quantum wells fits well with Eq. (1.47) for d = 2, .f, = 0,
and S, = 1. The sign of the absorption coefficient turns from positive to
negative when f, - f, < 0; in other words, the optical gain appears.
Nonlinear Susceptibility
The linear susceptibility obtained above describes the response of semi-

conductors when one electron interacts with one photon at one time. As the
power of the electromagnetic field increases, higher-order effects become
remarkable, where more than two photons participate in the electron-
photon interaction within a coherence time. The way materials respond to
the higher-order process is described by the nonlinear susceptibilities. This
optical nonlinearity has a wide variety of applications, such as high-speed
all-optical space and time domain switches, wavelength converters for
wavelength division multiplexing systems, and conjugate wave generation
for dispersion compensation. For the practical application of nonlinear
devices, both large and broad-band nonlinear susceptibility is required.
(Chapter 7 discusses the possibilities of quantum-dot optical nonlinear
devices.) In semiconductor lasers, however, nonlinearity usually degrades
laser performance, since the nonlinear optical gain causes gain saturation at
a high cavity-photon number, leading to output-power saturation and
limiting the highest modulation speed.
Nonlinear susceptibilities can be derived by further continuing the pertur-
bation expansion of the density-matrix equation. The two-band model used
above describes the system in this case as well as long as the wavelengths of
all the electromagnetic fields participating in the electron-photon interac-
tion are close to the resonant wavelength of the two bands, that is, w ~ , ,In~ .
this two-band model, even-order nonlinearity does not appear because the
even-order interaction erases the polarization. To describe the second-order
nonlinearity, a three-level system, as well as the crystal geometry without an
inversion symmetry is required. The calculation of the second-order non-
20 MITSURUSUGAWARA
linear susceptibility is biased quantum wells with a tilted potential profile is

presented by Kuwatsuka (1994).
Let us derive the formulae for the third-order susceptibility. From Eqs.
(1.33) through (1.35), the second-order components of the density matrix are
and
p1.t;'(w) = 0 (1.52)
The second-order polarization vanishes, as seen from the nondiagonal term

of Eq. (1.52). The third-order component of the nondiagonal term, taking
only the resonant terms, is
The third-order polarization of the system is given by the diagonal trace as
(P$(w = 2w2 - w J 2 ) = T r [ ~ ' ~ ) ( w )= +

P l-[pL~)(w)(ulreo.rlc) h.c.]Q,
(1.54)
and is related to the electromagnetic field as
( P g ( 0 = 2w, - cu1)/2)
= (&)%X(3Y2W2 - (111; (023 w2, -ol)E~,,,,(w,)Eqi.02(02)2cc/m21
(1.55)
In the numerical coefficient in Eq. (1.55), 3! represents the order of the
interaction and 2! indicates that two of the three electromagnetic fields
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 21
NANO-STRUCTURES
cannot be discriminated. Thus, the third-order nonlinear susceptibility is

given as
1
w2- w,, + ir,,
-
1
]
- TCt,
(ol - o,,~
[m2/V2] (1.56)
Taking the light “2” with the frequency of w 2 as the pump beam, and the
light “1” with w 1 as a probe beam, Eq. (1.55) describes the polarization to
generate a signal beam with a frequency of w, = 2w2 - wl. If the two pump
beams proceed in the opposite directions, the signal is the phase conjugate
wave of the probe beam. See Chapter 7 for experiments on nondegenerate
four-wave mixing for wavelength conversion and conjugate wave generation
in semiconductor lasers. The bandwidth of is dominated by the dephas-
ing rate of the polarization To,, which determines the response speed of the
nonlinear process (see Chapter 7 ) .
Using f 3 ’ of Eq. (1.56) under the degenerate case of o2= ol, the optical
gain, including the nonlinear effect, is given as
where the power density of the electromagnetic field is
(1.58)
has) the sign opposite to that of g(’)(w), the optical gain

Since ~ ‘ ~ ’ ( w
decreases as the power of the electromagnetic field in the cavity increases.
2. SPONTANEOUS EMISSION
OF PHOTONS
Let us consider a spontaneous emission process where an electron in the

conduction band recombines with a hole in the valence band by emitting a
photon. It is now well known that the time evolution, the spatial pattern,
22 MITSURUSUGAWARA
and the spectrum of the spontaneous emission depend on the electromag-

netic environment of a material -that is, the optical modes -and can be
modified by tailoring the modes using a cavity (for example, Yokoyama and
Ujihara, 1995). The purpose here is to derive a formula for the spontaneous
emission rate, or lifetime, in semiconductor nano-structures in the three-
dimensionally symmetric optical modes. The quantized expression for the
electromagnetic field is used because the concept of spontaneous emission
naturally comes out of the treatment. The remarkable difference from the
derivation of the optical susceptibility in Section 111.1 is that all possible
optical modes into which photons are emitted are taken into account.
Using the quantized expression for the electromagnetic field, the vector
potential of the quantized photons is written as
+
AqJr, t ) = A4,~(o,)e,[efq"h,(t)e-iq%,'(t)] (1.59)
where
(1.60)
2: and h,, are respectively, the generation and annihilation operators of

photons; v is the optical mode; oqis the frequency of the optical mode with
the wave vector, q; and V , = L,3 is the quantization volume of photons. The
quantization is done in the crystal as L = L,. To define the density of states
for the electromagnetic fields, we use the periodic boundary conditions,
where the wave function should be periodic in the x, y, and z directions with
a period, L,. This gives q, = 2d/L,, q,, = 2nrn/L,, and q, = 2nn/L,, where I,
m, and n are integers. Thus, the electron-photon interaction Hamiltonian
for the interband transition is
(1.61)
Let us describe the temporal variation between the excited state with the
electron in the conduction band and no photon Ic,k;O) and the ground
state with the electron in the valence band and one emitted photon
Iu, k; lq,u)using a time-dependent Schrodinger equation. The wave function
for the system is written under the rotating wave approximation with only
the resonant terms considered as
Let us assume the initial condition where one electron is in the conduction
PROPERTIES
1 OPTICAL OF SEMICONDUCTOR
QUANTUM 23
NANO-STRUCTURES
band and the field is in the vacuum state with a(0) = 1 and b,,(O) = 0. The
electron drops into the valence band by spontaneously emitting a photon
with an energy of oqand a momentum of q. By substituting Eq. (1.62) into
the time-dependent Schrodinger equation of
[H, + HI]’€’ = ih-aaty (1.63)
we get the following equations of motion for the probability amplitude as
and
where the coupling constant is
( 1.66)
The decay rate, y d , includes all the dissipation processes besides spontaneous
emission, such as nonradiative recombination at crystal defects or impurities
and scattering into other electronic states. Integrating Eq. (1.65) and then
substituting it into Eq. (1.64), according to Weisskopf and Wigner (1930),
we obtain
b(t) = -1Iyq.012f: k)(tpt’)l dt’ - ‘id a(t)

a([’)e[-i(cf)qpco<t
(1.67)
q.0
2
Let us transform the summation on q into the integration on the frequency,

using the three-dimensional continuous optical mode density p(w). (Note
that p ( o ) here is not a density matrix.) Then, Eq. (1.67) becomes
(1.68)
24 MITSURUSUGAWARA
where gw,ois the coupling constant at the frequency of o.This result shows
that the coupling to the continuous optical mode always causes an exponen-
tial irreversible spontaneous emission. It can be visualized as an electron
coupling to various optical modes with different frequencies and losing its
memory as soon as it emits a photon due to the interference of the different
optical modes. From Eq. (1.68), we get the spontaneous emission rate as
This is Fermi's golden rule.

Let us calculate the spontaneous emission rate of Eq. (1.69) for quantum
wells. As seen in Fig. 1.5(a), both the S-polarized mode (its electric vector
normal to the plane defined by q and the z-axis) and the P-polarized mode
(its electric vector in the plane defined by q and the z-axis) are taken into
account. Thus, the coupling constant is given as
Z
4
FIG. 1.5. (a) S-polarized mode with its electric vector normal to the plane defined by q. and
the z-axis and P-polarized mode with its electric vector in the plane defined by q and the z-axis.
The UJ plane corresponds to the quantum-well plane. (b) Wave vector of the electromagnetic
field, q = (q,,.4,). The xy plane corresponds to the quantum-well plane. The wave number of
the electromagnetic field emitted by the interband transition is n,q,,k/c, where w , , , ~is the
frequency dillerence between the k states of the conduction band and the valence band, In
exciton emission, q = n,tu,/c, where w.,' IS the exciton energy.
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 25
NANO-STRUCTURES
where q = (qI1,qJ, q" = ( q x , 4), is the in-plane (quantum-well plane) compo-

nent of the photon wave vector; q, is its perpendicular component; M is the
matrix element between the conduction-band and valence-band edge states;
Zncz,,t.Z is the overlap integral between the z-direction envelope wave func-
tions for the electron and the hole quantized states; and M,, is the factor
dependent on the electromagnetic field polarization direction of e l land e,
and on the electron wave vector. Reflecting the nature of base fuctions, M,,
differs between the electron-heavy-hole and electron-light-hole transitions
(see the appendix). So, Eqs. (1.69) and (1.70) give
The next step is the summation on the possible optical modes that satisfy
the energy conservation rule. Let us transfer the summation only on 9, into
the integration, using the mode number of (LC/27c)dq,between q, + dqz and
dq,. This is done to clarify the difference between the electron-hole
transition and the exciton spontaneous emission treated in Section IV.5.
Thus, we obtain
where
( I .74)
and q, = n,w,,k/c. Here, D = Lf = L2. The summation on the in-plane wave

26 M~TSURU
SUGAWARA
vector is taken within qll < qc to satisfy the energy conservation rule (see
Fig. lS(b)). Then, by integrating Eq. (1.72) into the quantum-well plane, we
get
(1.75)
The polarization dependence arising from the asymmetric characters of the

base function of valence bands disappears from the final result because the
optical mode distribution is symmetric. In Section IV.5, we will see how Eq.
(1.75) is modified for the exciton spontaneous emission process.
When there are N electrons in the conduction band and N holes in the
valence band, the average spontaneous emission rate is given by summing
Eq. (1.75) in k space as
The summation in k can be transformed into the integration in the energy

via the same procedure used to derive the absorption coefficient in Eq.
(1.47).
Note that Eq. (1.75) holds not only in quantum wells but also in bulk
materials, quantum wires, and quantum dots, simply by changing the matrix
element. This is so because the three-dimensionally symmetric photon mode
is considered and because the photon wave vector does not appear in the
wave vector selection rule. For example, by omitting Zncz.,,zz in Eq. (1.75), the
expression for bulk materials becomes
( 1.77)
Let us use material parameters of bulk In,~,,Ga,,,,As as me = 0.044m0; the

spin-orbit splitting energy A = 0.35 eV; E, = 0.75 eV; M 2 = 2.12m,E, with
D’ = 0 in Eq. (A.21), and n, = 3.5. Thus Eq. (1.77) gives rsD= 2.8 ns.
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 27
NANO-STRUCTURES
3. RATE EQUATIONS
FOR LASEROPERATIONS
Semiconductor lasers are the most important optoelectronic devices in

modern optical technologies such as optical transmission systems, optical
data storage, biomedical optoelectronics. The principle of laser operation is
that light, first produced by spontaneous emission, going back and forth
through the gain medium in a cavity is amplified as a result of the stimulated
emission process, leading to high-power coherent light output if the gain
exceeds the cavity loss. The optical gain appears in semiconductors when
electrons and holes are injected into the conduction band and the valence
band, respectively, to the extent that the absorption coefficient of Eq. (1.42)
becomes negative. This situation is called populution inuersion, since in
thermal equilibrium most electronic states in the conduction band are empty
and those in the valence band are full. The semiconductor also acts as an
optical wave guide to form a cavity using both side cleaved facets or, in the
case of vertical cavity surface emitting laser diodes, distributed Bragg
reflection mirrors. Laser performance has improved steadily over nearly
three decades since room-temperature continuous operation of double
heterostructure lasers was first achieved (Hayashi et al., 1970), leading to
high-performance strained quantum-well lasers (Thijs et al., 1990). See
Chapter 6 for an overview of recent developments in quantum-dot lasers
and their significance and future prospects.
Laser operation is defined by a set of rate equations to describe the
time-dependent variation of carriers and cavity photons. Using rate equa-
tions, various laser properties such as lasing threshold currents, threshold
carrier density, current-power relationship, external differential quantum
efficiency, and small-signal modulation bandwidth can be simulated. In this
section, we summarize the rate equations that will be used in the discussion
of quantum-dot lasers in Section V.l.
The most common laser is an edge-emitting type with cleaved facets (Fig.
1.6(a)). Currents are injected into the active region through the p-n junction,
and the optical wave guide confines the light to the active region's vicinity.
The smooth cleaved facets with the reflectivity of R , and R , act as both a
reflector and an output coupler with the back facet usually coated to
increase the reflectivity. The typical dimensions are a cavity length, L,,, of
between 200 and 1500 pm and a lateral width of 2- 10 pm. Several require-
ments must be satisfied by the materials, such as band gap, refractive index,
and lattice constant.
The rate equations to describe the time-dependent variation of the active
region carrier number, N , and the cavity photon number, S, are
(1.78)
28 MITSURUSUGAWARA
FIG. 1.6. (a) Cavity laser. (b) Schematic plot output power and carrier number versus
injected current in a cavity laser. Above the threshold of If,,.the injected current is transformed
into photons, keeping the carrier number clumped. The nonlinear gain causes gain saturation
at high output power, leading to a decrease in the quantum efficiency and an increase in the
carrier number.
and
(1.79)
where I is the injected current; r is the optical confinement factor; C , is the

spontaneous emission factor representing the fraction of spontaneous
emission entering the lasing mode; E, is the nonlinear gain coefficient
derived from Eq. (1.57); and V , is the active-layer volume. (See Section V.1
for an explanation of 8, in quantum-dot layers.)
The photon lifetime in the cavity, z p , is written as
where ai is the internal loss. Carriers injected into the active region
recombine radiatively with the lifetime of rsp or nonradiatively with the
lifetime of z,, giving the total lifetime as
In long-wavelength infrared lasers, nonradiative recombination includes

Auger as well as impurity-related recombination. The optical gain for the
lasing mode, g:), is usually the maximum value of g(l)(a) in conventional
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 29
NANO-STRUCTURES
Fabry-Perot lasers, since the separation of the cavity mode is usually less
than 1 nm.
The output power is proportional to the cavity-photon number and is
given as
Po,, = hwcS In(1/R)/(2Lcanr) (1.82)
where R is R , or R , and Aw is the emitted photon energy. The temporal

variation of S and N , including the rise time of the response, the relaxation
oscillation, and steady-state values, etc., can be obtained by numerically
solving Eqs. (1.78) and (1.79). The steady-state solution of the rate equations
is given by setting the lefthand time-derivative term at zero in Eqs. (1.78)
and (1.79). For Eq. (1.79) to have a solution with S > 0, the lasing threshold
condition is given as T; = (c/n,)Tg!& with the threshold gain of g:,',,, and
thus,
The contribution of C, is disregarded for a conventional cavity laser.

Let us check the solution simply by setting an approximate formula for
the optical gain as
where ga)' is the differential optical gain and N , , is the carrier number for
the transparency where the gain is zero. This formula is often used to express
the optical gain in bulk semiconductor lasers. For quantum-well lasers, an
empirical formula using a logarithmic dependence of carrier is usually
employed (Chuang, 1993). The solutions are
N = z,I/e (1.85)
and
s=o (1.86)
below the lasing threshold, and
(1.87)
30 MITSURUSUGAWARA
and
S = (t,/e)(f - eN/z,) (1.88)
above the threshold. The threshold current is given from Eqs. (1.87) and
(1.88) with S = 0 as
(1.89)
Figure 1.6(b) is a schematic graph of the output power and carrier numbers
as functions of the injected current. The carrier number is clumped above
the threshold, since the injected electrons are transformed to photons
through stimulated emission. An increase in the injection current above the
threshold of Eq. (1.89) leads to an almost linear increase in output laser
power until its saturation occurs. The saturation of power and the increase
in carrier number are due to the nonlinear gain saturation effect through the
third-order nonlinear gain coefficient.
A procedure to design strained quantum-well lasers with low-threshold
currents is as follows. First, the band dispersion curves for the conduction
and valence bands are calculated by means of the k * p perturbation method,
the curves are quite sensitive to the amount of strain and the well width.
Then the spontaneous emission lifetime (z), and the optical gain are
calculated as a function of carrier density through Eqs. (1.42) and (1.76).
Next, the strain and well-width dependence of the threshold currents are
obtained. In strained InGaAs/InP quantum-well lasers, it has been shown
that either direction of strain -that is, bi-axial compressive or tensile -
lowers the threshold currents (Sugawara and Yamazaki, 1994).
In this simple description of laser operation, it is assumed that all of the
carriers are injected into the active region without a time delay. This is not
a good assumption, especially in quantum-dot lasers, since the carrier
relaxation is retarded due to the complete discrete level formed in quantum
dots. The laser simulation including retardation of carrier relaxation is
presented in Section V.l.
IV. Exciton Optical Properties
In Section 111, optical susceptibility (absorption and gain) and sponta-

neous emission due to interband electron- hole transition were discussed.
The absorption spectra have the same energy dependence as that of the
interband joint density of states, and they vary with the dimensions of the
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 31
NANO-STRUCTURES
nano-structures. As seen in the experimental optical absorption spectrum of

Fig. 1.4(a), however, strong resonant peaky spectra are superimposed on the
steplike absorption continuum peculiar to quantum wells. This is due to the
creation of electon-hole bound states, called excitons, through the Coulomb
interaction.
Since the discovery of strong exciton resonance in the absorption spectra
(Dingle, 1975), excitons in quantum wells have revealed various unique
properties, which are not only physically attractive but also useful for
optical device applications.
First, exciton resonance is observed even at room temperature, owing to
a variety of factors: the enhanced oscillator strength and binding energy
caused by wave-function compression under the potential barriers; the
relatively low scattering rate (several hundred femtoseconds at room tem-
perature, as will be seen in the experiments depicted in Fig. 1.9 on exciton
spectrum broadening); and high-quality crystals with smooth interfaces and
homogeneous compositions.
Second, remarkable optical nonlinear effects have been found under both
real and virtual excitation. In particular, the a.c. Stark effect, which shows
a blue shift of exciton resonance and its bleaching under virtual excitation,
has attracted much attention as a possible mechanism for sub-picosecond
switching and as one of the most fundamental phenomena in light-matter
coupling (Schmitt-Rink et al., 1989, for a review).
Third, when the electric field is applied perpendicularly to the quantum-
well layers, exciton resonance exhibits an electric-field-induced shift toward
the lower energy region (Chemla et al., 1983). What is unique is that the
resonance clearly remains up to an electric field of about lo5V/cm because
the potential barriers prevent the ionization of excitons. This so-called
quantum-confined Stark effect shows a significant change in the absorption
coefficient and is now being used in high-speed optical modulators up to
10 GHz. The bistable devices- SEEDS-operate on the basis of this
phenomenon (Miller et al., 1985b).
Fourth, spontaneous emission of excitons in semiconductor nano-struc-
tures has led to various interesting phenomena, such as rapid spontaneous
emission on the order of picoseconds in quantum wells (Hanamura, 1988),
mesoscopic enhancement of the spontaneous emission rate in microcrystals
(Hanamura, 1988), and the cavity-polariton- that is, the exciton-photon-
coupled mode- in semiconductor microcavities (Houdre et al., 1994, Cao
et al., 1995). These phenomena come from a unique selection rule to preserve
the wave vectors between the exciton center-of-mass motion and the photon.
Fifth, the possibility of exciton lasing in wide-gap semiconductor mater-
ials such as the II-VI and GaN systems, which has been frequently discussed
in recent years (Ding et al., 1992, 1993). We should discount the simple
32 MITSURU
SUGAWARA
explanation that the high exciton-binding energy of these wide-gap mater-

ials stabilizes exciton states, resulting in exciton lasing. Excitons as bound
states between an electron and a hole cannot exist under the population
invertion (Uenoyama, 1995). Thus, in order for bound-state excitons to
generate a gain, the exciton emission energy should be shifted from the
resonant exciton absorption energy through some mechanism, such as scat-
tering (Galbraith, 1995, for example), exciton localization, and bi-exciton
formation (Sugawara, 1997a, 1998), etc.
This section reviews the optical properties of excitons. In Section IV.l
through 3, using the second quantization, formulae are introduced for
exciton state vectors, effective-mass equations, and the interaction Hamil-
tonian with photons. In Section IV.4, a formulae for the optical absorption
spectra is derived and compared with the experiments. In Section IV.5, the
spontaneous emission of excitons is discussed. Also, a quantum disk is
treated to show how the emission properties vary from quantum wells to
quantum dots by a change in the disk radius. In Section IV.6, the disk is
placed in a planar microcavity so that the spontaneous emission behaviors
of cavity-polaritons can be observed. In Section IV.7, the effect of the
electron-hole Coulomb interaction on optical gain is briefly reviewed. For
details on the quantum-confined Stark effect, refer to Chuang’s textbook
(1995).
1. STATEVECTORS
An exciton is composed of an electron in the conduction band and a hole

in the valence band under the Coulomb interaction. A hole is a positive
particle representing an empty state in the valence band. Its effective mass
is defined as mf = -mz, and its sign is positive in most cases, as stated
before. The hole energy is at its minimum at the top of the valence band
and increases along the valence-band dispersion curve. The creation oper-
ator of a hole with a wave vector in the valence band, k, is defined as
(1.90)
where ak,, is the electron annihilation operator in the valence band, and
k h --ku
- (1.91)
The conduction-band effective mass and the wave vector are renamed for
the electrons as m,*and k,. The position vectors of an electron and a hole
are re and r,, respectively.
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 33
NANO-STRUCTURES
The wave function of the exciton state-an electron-hole pair state-is

given by the electron creation operator, a:, and the hole creation operator,
b;, on the ground state (filled valence band and empty conduction
band) and by summing the pair function over all possible states as
(1.92)
To determine the expansion coefficient, Ck,kh, in Eq. (1.92), let us replace the
creation operators of an electron and a hole with the field operator, using
the relationship below (Haken, 1973, for example):
(1.93)
and
Here, Fk,(re) and Fkh(rh)are solutions to a one-particle effective-mass

equation for an electron and a hole, respectively. By substituting Eqs. (1.93)
and (1.94) into Eq. (1.92), we get
r r 1
(1.95)
Here, t,hex(re,rh) satisfies the effective-mass equation for excitons, as will be

shown below. The exciton state laex)is given by the overlap of the created
electron-hole pair with the probability amplitude of +ex(re, rh).The exciton
state is described in the wave vector space by expanding the field operator
in Eq. (1.95) as
(1.96)
34 MITSURUSUGAWARA
and
Then, by substituting Eqs. (1.96) and (1.97) into Eq. (1.95), we get the
exciton state vector as
(1.98)
where A(k,,k,) is the Fourier transform of the exciton envelope wave

function in real space. When the Coulomb interaction is neglected and the
envelope wave function is given by the product of the electron plane-wave
with k, and the hole plane-wave with k,, Eq. (1.98) becomes
Comparing Eqs. (1.98) and (1.99), it is clear that the Coulomb interaction
makes up the exciton state by summing the electron and hole pair state with
the amplitude of A(k,, k,) around the band edge.
The Coulomb potential between an electron and a hole is given by
(1.100)
where E, is the static dielectric constant. The potential depends only on the
difference between the electron and hole position vectors. Therefore, let us
change the coordinate vectors into the relative motion vector as
r = re - rh (1.101)
and the center-of-mass motion vector as
R=
+
m$re mtrh
(1.102)
M*
where
M* = m: + mt (1.103)
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 35
NANO-STRUCTURES
Transforming the wave vectors as
K = k, + k,, (1.104)
and
(1.105)
we get
A(k,, kh) = A(K, keX)= V d3rd3R$,,(r, R)e-'K.Re-'kex'r( 1.106)
The above equations can be applied to exciton states regardless of the

dimension of materials. However, it is more convenient to use the quantized-
state wave function explicitly in expanding the field operator. This is done
for both quantum wells and quantum dots below.
Quantum Wells
When the quantum-well confinement potential in the z direction is much

greater than the electron-hole Coulomb potential energy, the z-direction
wave function can be separated from the wave function for the in-plane
motion. Thus, the field operator is expanded as
(1.107)
and
where ri = (r\I,zi);ki = (k\I, niJ; and i = e, h. Then, from Eqs. (1.95), (1.107),
and (1.108), we get
(1.109)
36 MITSURUSUCAWARA
where $ex(rll, R") is the in-plane envelope wave function separated from the
z-direction component. By transforming the variables, the Fourier compo-
nent is written as
The vectors related to the relative motion and center of motion are all
two-dimensional.
Quantum Dots
Let us expand the field operator as
(1.111)
and
where k , = (nex,ney,nez) and kh = (nhx,nhy,nhz). Thus, the state vector is

given as
The integration in Eq. (1.1 13) represents the mixing of different quantized
states via the Coulomb interaction. Without the Coulomb interaction, Eq.
(1.1 13) is reduced to Eq. (1.99).
2. EFFECTIVE-MASS
EQUATIONS
The effective-mass equation for the electron-hole system is given as

37
NANO-STRUCTURES
1 OPTICAL PROPERTIES OF SEMICONDUCTOR QUANTUM
Here, the Hamiltonian is as follows (Haken, 1973, for example):
where H ; is the electron kinetic energy; H i is the hole kinetic energy; V(r, r’)
is the Coulomb potential energy of Eq. (1,100) between two particles; and
the field operator working on both the conduction and valence bands is
By substituting Eq. (1.116) into Eq. (1.1 19, the Hamiltonian becomes
where
for the electron, with the energy origin taken at the valence-band top,
(1.119)
for the hole, and
( 1.120)
38 MITSURUSUGAWARA
The commutation relation of the field operators are used (Haken, 1973, for
example). V,, is the electron-electron interactions; V,, is the hole-hole
interaction; V,, is the electron-hole interaction; and V,, is the exchange
interaction.
By substituting the exciton state vector of Eq. (1.95) into the effective-
mass equation of Eq. (1.1 14), we get the exciton effective-mass equation as
where the exchange interaction term is neglected. Analytical solution to Eq.

(1.121) are derived for both the three-dimensional (Elliot, 1957) and the
two-dimensional cases (Shinada and Sugano, 1966). In Chuang’s textbook
(1995), the results are instructively summarized. Equation (1.121) has
bound-state solutions for E,, < E , with a series of discrete energies, and
unbound-state solutions for E,, > E , with continuum energies. In bulk
materials, the bound-state energy is given as
m:e4
E,, = E, -
32rc2h2&;n2
+-k2M*
2K2
(1.122)
where n = 1,2,3. .. . The wave function is separated as
(1.123)
where the ground state for the relative motion (n = 1) is
4(r)= rc -112 aB-312 exp(-r/aB) (1.124)
with the Bohr radius a, = 4ne0~,h2/(m~e2),

and the wave function for the
center-of-mass motion is
(1.125)
Here, several other cases are briefly described: quantum wells, quantum
wells with in-plane confinement potential; and quantum wells under a
magnetic field.
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 39
NANO-STRUCTURES
Quantum Wells
It is assumed that quantum wells are type I; that is, both an electron and
a hole are confined in the same well layer. Though the asymmetry of the
effective mass is taken into account for the direction perpendicular to and
parallel to the well layer, band nonparabolicities are neglected. The super-
scripts of 11 and I are attached to the effective masses and position vectors
to describe the parallel and perpendicular directions, respectively. By adding
the quantum-well confinement potential to Eq. (1.121), an effective-mass
equation results:
where peh = [r"' + (z, - z , ) ~ ] " ~is the distance between an electron and a
hole, and ((ze) and V,(z,) are the z-direction quantum-well confinement
potentials for an electron and a hole, respectively. When the confinement
potential in the z direction is much greater than Coulomb potential energy,
the wave function can be separated as
This is a good approximation for actual quantum wells with several tens to
hundreds meV potential barriers. The in-plane relative motion and the
center-of-mass motion can be separated. The exciton resonance energy is
given as
(1.128)
where Enpl and En,,=are the electron and hole quantized energies in the
quantum well. The energy of ELm for the relative motion is determined by
( 1.129)
and
40 MlTSURU SUGAWARA
+
where n = 1,2,3,. . . and rn = 0, 1,. . . , +(n - 1). The optically active
exitons are s excitons with rn = 0. For the ground-state optically active 1s
excitons (n = 1 and rn = 0), a variational wave function of
(1.131)
is often used. The variational parameter, ,Iex represents the two-dimensional

exciton radius, since ,Iex= ( ~ ( r ~ ~ ) ~ r lThe
l ~ ~energy
( r ~ ~of)E) R, for the center-
of-mass motion is
( 1.132)
and the wave function is
(1.133)
The in-plane wave vector is given by the periodic boundary condition as

KII = 27c(n,/L, n,/L), where D = Lz and n, are integers (i = x and y).
Quantum Wells with In-Plane Harmonic-Oscillator-Type Parabolic Potential
Since actual quantum wells inevitably have local structural imperfections,

such as interface roughness and composition fluctuations, excitons are
localized at the potential minima caused by the imperfections, especially at
low temperatures. Their extreme cases are the quantum dots in Fig. l.l(b),
where the potential depth is large enough to prevent exciton thermal escape
even at room temperature. By adding the in-plane confinement effect to the
quantum-well potential, we get the potential as
O(x) is the Heaviside step function (O(x) = 1 for x 3 0 and O(x) = 0 for
x < 0). If the z-direction confinement potential is high enough to confine
excitons in the in-plane two-dimensional regions, Eq. (1.134) is reduced to
approximately
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 41
NANO-STRUCTURES
These are exact when the z-direction potential height is infinite. Thus, the
wave function can be separated as
using the z-direction quantized electron and hole wave functions. The
+ +
exciton resonance energy is given as E,, = Eg Enez En,,= E,. +
The binding energy E , is determined by
= EB$e,(rll, RII) ( 1.137)
Equation (1.137) can be separated into the relative-motion and center-

of-motion parts by using the following harmonic-oscillator-type parabolic
potentials to describe lateral confinement for both an electron and a hole:
Then,
and
(1.140)
For relative motion, we obtain
where $, is given by Eq. (1.130). For the ground-state s-state optically active
excitons (n = 1 and rn = 01, we solve Eq. (1.141) by a variational method
using the trial wave function of
where a', b', and c' are variational parameters, two of which are independent
because of the normalizing condition. For center-of-mass motion, the
42 MITSURUSUCAWARA
effective-mass equation is
Equation (1.143) describes a two-dimensional harmonic oscillator and can

be solved analytically. The wave function is given as
(1.144)
where
p = ,,hRII/B, B = d m , and Llfl(p) are associated Laguerre poly-

nomials; k = 0,1,2,3,.. . ; and I = 0, ? 1 , . . . , k. The energy is given by
E;, = (2k - 111 + l)tiu, (1.146)
The ground-state wave function ( k = I = 0) is the Gaussian function.
Quantum Wells Under the Mugnetic Field
Let us consider a situation where a magnetic field is applied perpendicular

to the layer (z direction): B = (O,O, B). Taking the vector potential that
works on an electron and a hole as ( i = e and h)
the effective-mass Hamiltonian for excitons is written as
where p, and ph are the momentum operators. First, we perform a trans-

formation to the center-of-mass motion and relative motion coordinates.
PROPERTIES
QUANTUM 43
NANO-STRUCTURES
Second, we choose the envelope wave function as (Knox, 1963)
where Ar(rll) = ( - yB/2, xB/2). Thus, we obtain
(1.150)
with the exciton resonance energy of
(1.151)
The expectation value of the third term (the Zeeman term) and the fifth term
of Eq. (1.150) is zero for optically active S-state excitons. The fourth term,
the diamagnetic energy term, dominates the magnetic-field-induced shifts of
exciton resonance in quantum wells.
INTERACTIONS
3. EXCITON-PHOTON
The Hamiltonian describing the interaction between the electromagnetic

field and the material is written in the second quantization expression as
(1.152)
The vector potential of the quantized photons is given by Eq. (1.59). By

substituting the field operator of Eq. (1.116) into Eq. (1.152), and selecting
only the terms for the interband optical transition between the conduction
band and the valence band, we get
44 Mirsuuu SUCAWARA
By expanding the field operator as
(1.154)
and
(1.155)
Eq. (1.153) becomes
+
The wave vector selection rule is q = k, ke and is already applied for the
interband matrix element of PgSk. Note that the band-edge Bloch functions
of u,(r) for the conduction band and ul,(r) for the valence band are added
in Eqs. (1.154) and (1.155), but were dropped in the effective-mass equations.
Let us derive the exciton-photon interaction Hamiltonian under the
Dirac representation. The exciton state with n photons is
and the ground state with n + 1 photons is

Is> = I q J ~ + 1) (1.15 8 )
Coupled modes between excitons and photons are not taken into account
here. The Dirac representation for the optical transition between the exciton
state and the ground state is given as
where, using Eqs. (1.98) and (1.106)
(1.160)
1 OPTICAL
PROPERTIES QUANTUM
OF SEMICONDUCTOR 45
NANO-STRUCTURES
and
In Eq. (1.161), the relationship of ~k,,e-'k'x'r = V6(r) is used. Care must be

taken not to confuse the Fourier transform of the exciton wave function and
the vector potential of the electomagnetic field. The band-edge value of Pzv,k
is used for excitons, since exciton states are composed of the k-states around
the band edge. The wave vector selection rule for the center-of-mass motion
is K = q, and so the excitons emit photons in the direction of K. This
concept is a key to understanding the exciton-polariton and fast sponta-
neous emission in quantum wells.
The above equations can be applied to exciton states, regardless of the
dimension of materials. However, it is more convenient to use quantized-
state wave functions explicitly in expanding the field operator. This is done
in the following equations for both quantum wells and quantum dots:
Quantum W e h
(1.162)
and
(1.163)
Eq. (1.153) becomes
In the Dirac representation of Eq. (1.159),
(1.165)
46 MITSURUSUGAWARA
and
Quuntum Dots
(1.167)
and
(1.168)
Eq. (1.153) becomes

e
H, =- + h.~.]
C 1 Aq,(~~)[P",l,a,a~b,',h (1.169)
mO q,o k,.kh
In the Dirac representation of Eq. (1.159)
(1.170)
and
If the general formulae of Eqs. (1.98) and (1.06) are used,
P:; =
kex
P&A(q,kex) % P&
s d3R$,,(0, R)e-'q'R2 P;"
s d3R$,,(0, R)
(1.172)
The last approximation holds when the dot volume is smaller than the
exciton resonant wavelength and the electric dipole approximation holds -
that is, eciq'Rz 1. Under this approximation, the square of the matrix
1 OPTICAL PROPERTIES OF SEMICONDUCTOR QUANTUM 47
NANO-STRUCTURES
element is proportional to the volume covered by the center-of-mass motion.

As will be seen in Section IV.4, the spontaneous emission rate of excitons
increases in proportion to the dot volume. This is often referred to as the
mesoscopic enhancement of the exciton spontaneous emission rate. In con-
trast, the spontaneous emission rate of the electron-hole pair is independent
of the crystal volume and dimension, as seen in Section 111.2.
ABSORPTION
4. OPTICAL SPECTRA
Let us use a density-matrix equation to derive the formula for the optical
absorption caused by excitons. The procedure is almost same as that used
in deriving linear susceptibility without the Coulomb interaction in Section
111.1. Slight differences are the base functions and the initial condition. The
base functions are taken as the exciton state, le), and the ground state, Is),
where
le> = laex) and Is) = Pg) (1.173)
The system is initially in the ground state. Note that, as the number of
excitons increases, the system is transferred to the electron-hole plasma
state, making it impossible to use the exciton base functions. The classical
expression for the electromagnetic field of Eq. (1.21) is used. The kinetic
energy term is
H, +
= ~cJ~ele)(el ho,Ig)(gl (1.174)
The perturbation term for the exciton-photon interaction is
with
+
,ueg(t) = +[peg(o)e-iw' peg(-o)e'"'] (1.176)
where
(1.177)
and
(1.178)
48 MITSURUSUGAWARA
The density matrix for the exciton state is
The initial condition is p::) = 0 and p::) = 1. By solving the density-matrix

equation as in Section 111.1, we obtain the nondiagonal term under the
first-order perturbation as
(1.180)
where we, = oj, - 01,. The polarization of the system is given by the
diagonal trace as
Degeneracy due to spin is taken into account and the relationship of
(1.182)
is used. The summation over the exciton states is also taken into account.
From its imaginary part, the optical absorption coefficient is given as
(1.18 3)
Comparing Eq. (1.142) and Eq. (1.183), the exciton effect on the optical
absorption spectra appears in the transition matrix of P:; and the transition
energy of a,,.These are determined by the exciton effective-mass equations,
The optical absorption spectrum consists of a series of resonance, the
continuum spectrum whose intensity is affected by the Coulomb interaction,
and the smoothly connecting part in between. The enhancement ratio of the
continuum spectrum strength and the strength of the noninteracting elec-
tron-hole pair is known as the Somrnerfeld factor. The Coulomb effect on
the optical absorption spectra is schematically summarized for one- two,
and three-dimensional cases by Ogawa (1995).
For quantum wells, using Qd = (DLQw)-', Eq. (1.1 85) becomes
ELQW= SP*BPl(hC1)- hoe,) (1.184)

1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 49
NANO-STRUCTURES
with the integrated intensity of
(1.185)
the oscillator strength per unit area of
( 1.186)
and the spectrum-broadening function of
B,(hw - hw,,) =
hLg/n (1.1 87)
( h -~ +
When probing light is applied perpendicularly to the quantum-well layer,
ql1= 0, and thus KII = 0 from the wave vector selection rule (see Eq. (1.166).
The exciton optical absorption strength is proportional to Id(rl1= 0)l2,
which represents the probability of finding the electron and the hole at the
same position. For the ground-state excitons in quantum wells, Ig5(rl1= 0)l2
is inversely proportional to the square of the two-dimensional exciton
radius-that is, to as seen in Eq. (1.131). Equation (1.187) shows that
the exciton optical absorption spectrum has a Lorentzian profile, with an
FWHM of 2hTeg, which increases with an increase in temperature due to
scattering by phonons.
Figure 1.7 shows the calculation of the ground-state exciton resonance
characteristics for various 111-V and 11-VI semiconductor quantum wells
(Sugawara, 1992): the well-width dependence of (a) radius, ,Iex; (b) binding
energy, E,; calculated by Eq. (1.126); and (c) integrated intensity, Sex,
calculated by Eq. (1.185). The exciton radius decreases as the well width
decreases due to the quantum-confinement effect, and increases in wells
narrower than 2 nm due to the breakdown of the confinement effect -that
is, the spread of the z-direction quantized-state wave function to the barriers.
In conjunction with this, the binding energy increases as the well width
decreases to about 2nm, and then goes down. The variation in the
integrated intensity shows almost the same tendency as the variation in the
binding energy. The decrease in narrow quantum wells is due to both the
increase in the radius and the decrease in the electron-hole overlapping
integral. Figure 1.7(d) shows the band-gap dependence of the integrated
intensity in 5-nm quantum wells. The dashed line serves as a visual guide.
Wider-gap semiconductors have a smaller radius, a greater binding energy,
50 M ITSURu SUGAWARA
'ZnMnSsRnSe
10 20 O;' " "10 " " 20

" '
Well width (nm) Well width (nm)
(a) (b)
-
210 , , , , s , , , . r . 2
I
8
s ! . s! ZnSelZnMn%sQ
X
Y ZnSaRnMnSe
z 6 -
.-
1:-
tn
5 4- CdZnTmTm /
.-
CI
C
9'
C'
J&aPIIIGaInP-
cn
C InGaAsilnP
c
'0
C
-
Y
0
C
I
2.
.
d""
*.g
GaAa(AIGaAs .
d%GaAallnP
FIG. 1.7. Characteristics of 1s exciton resonance calculated for various 111-V and 11-VI
semiconductor quantum wells (from Sugawara. 1992). (a) Well-width dependence of the radius,
&. (b) Well-width dependence of the binding energy, E,. (c) Well-width dependence of the
integrated intensity of the optical absorption spectra. (d) Band-gap dependence of the
integrated intensity in 5-nm quantum wells. The dashed line serves as a visual guide.
and stronger resonance. This is due to their larger effective masses and
smaller dielectric constants.
Experiment on Exciton Optical Absorption Intensity
Though the quantum wells used in the calculation in Fig. 1.7 have
actually been grown, so far there are no systematic experimental analyses of
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 51
NANO-STRUCTURES
how the exciton resonance intensity depends on materials. Here, instead of

changing materials, the exciton in-plane wave function is experimentally
compressed under a magnetic field perpendicular to the quantum-well layers
to observe how the exciton resonance intensity depends on the exciton
radius.
Figure 1.8(a) shows the magneto-optical absorption spectra for the
Ino.6sGao,,sAS/InP quantum well, whose cross-sectional TEM photograph
is shown in Fig. l.l(a). The ground-state electron-heavy-hole (le-hh)
exciton resonance at the absorption edge shows diamagnetic shifts, and its
intensity increases considerably. The absorption continuum separates into
discrete spectra, which are assigned to 2s states of le-hh excitons.
Figure 1.8(b) shows the diamagnetic shifts as a function of the square of
the magnetic field. The results for another quantum well with a composition
of In,,,,Ga,,,,As/InP are also plotted. The curves are fitted by Eq. (1.150)
for the S-state optically active excitons using a variational wave function for
the relative motion of Eq. (1.142). The third and the fifth terms are zero.
Taking into account the nonparabolicity of the conduction and valence
bands, the energy is calculated in the wave vector space using the Fourier-
transformed exciton wave function. The conduction-band dispersion is
calculated by the first-order 8 x 8 matrix (Sugawara et al., 1993a), and the
valence-band dispersion is calculated by the Luttinger-Kohn 6 x 6 k * p
matrix (Luttinger and Kohn, 1955). The fitting parameters are valence-band
Luttinger-Kohn parameters. The band-edge masses are rnL1 = 0.047rnOand
mil = O.lrn, for In,,6,Ga,,,,As/InP quantum wells, and rn:’ = 0.05rn0 and
m)ll = 0.12m0 for 1n0,,,Ga,~,,As/1nP quantum wells. If the magnetic field is
treated as a perturbation for exciton states in the low-field limit and uses
the hydrogenic wave functions of Eq. (1.131), Eq. (1.150) gives a diamagnetic
shift of
AEr = 3e2&B2/16rn,*II ( 1.188)
which is proportional to the square of the magnetic field (dashed line). The
approximation of the low-field limit holds at most up to 1 or 2 Tesla. Under
higher magnetic fields, diamagnetic shifts are no longer proportional to the
square of the magnetic field and deviate from the straight line, indicating
that the in-plane exciton wave function shrinks (Aex decreases) under the
magnetic field.
The integrated intensity of the le-hh exciton resonance is plotted as a
function of the magnetic field in Fig. 1.8(c). Here, the curves are the
calculation of Eq. (1.185). Actually, the calculation was done in wave vector
space and band nonparabolicities were taken into account. The second-
order k * p term in Eq. (A. 21) is set at D' = - 6 as a fitting parameter, giving
52 MITSURU SUGAWARA
I' ' '

In,pa,AdlnP IQWS
6 - on (001) InP I
3- ----4.2Tesla .\
4l '
8 .- -7.OTesla ' \
--.x=0.35, Lz=9.7nm
(a) Wavelength ( p) Magnetic field (T*)
0.5' " . . ' . 1

0 2 4 6 8
(c) Magnetic field (Tesla)
FIG. 1.8. Experiments on exciton diamagnetic shifts (from Sugawara, 1993a. Copyright
1993 by The American Physical Society). (a) Magnetooptical absorption spectra of the
In,~,,Ga,,,,As/lnP quantum well, shown in Fig. l.l(a). (b) Diamagnetic shifts of the le-hh
exciton resonance as a function of the square of the magnetic field. The results for another
quantum well, with a composition of ln,,,,Ga,,,,As/InP, are also plotted. The curves are fitted
by Eq. (1.150) for the IS-state optically active excitons using the variational wave function of
Eq. (1.142). The dashed straight line is for the Iow-field limit by Eq. (1.1823). The downward
bending of the diamagnetic shifts shows that the in-plane exciton wave function shrinks (,?e.r
decreases) under the magnetic field. (c) Integrated intensity of the le-hh exciton resonance as
a function of the magnetic field. The curves are the calculation of Eq. (l.lX5). The increase in
intensity is due to exciton shrinkage.
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 53
NANO-STRUCTURES
about a 1.2-times larger value of M 2 than that given by neglecting D’ in Eq.

(A.21). The excellent agreement between the calculation and the experiments
shows that the enhancement of integrated intensity is due to the shrinkage
of the exciton wave function, and it proves the derived theoretical express-
ions for the exciton optical absorption strength.
Experiment on the Resonance Spectrum ProJile
In addition to the broadening due to scattering by Eq. (1.187), in actual

quantum wells we must take into account inhomogeneous broadening due
to spatial fluctuations of the resonance energy. The excition resonance
energy depends on the structural imperfections of sites in quantum wells,
and each level shows thermal broadening caused by scattering. Therefore,
the spectrum-broadening function should be written using the convolution
integral as
B,,(hw - EeX)= BO(ho - E,, + E)BT(E)d E (1.189)
where B , is the inhomogeneous broadening function. It is assumed that the

exciton scattering lifetime is independent of the slight variation in exciton
energy. This is reasonable, since structural imperfections primarily change
the band-gap energy or quantum-confinement energy, and the change in the
exciton-binding energy is negligible.
Figure 1.9 shows the measured (solid lines) and calculated (dashed lines)
optical-absorption spectra at (a) 4.2 K and at (b) 295 K of In,,,,Ga,,,,As/
InP quantum wells used in the magneto-absorption measurements of Fig.
1.8. At 4.2 K, the electron-heavy-hole exciton resonance spectrum has a
Gaussian distribution of
with an FWHM of To = 2.35(, = 4.7 meV. The calculated curve at 295 K

fits the profile of the resonance very well, using Eq. (1.189) with re;' =
150 fs. Figure 1.9(c) shows the FWHM of the electron-heavy-hole exciton
resonance spectrum as a function of temperature for different samples with
almost the same composition and well width (10-nm Ino~,,Ga,,,,As/InP
quantum wells) but with different inhomogeneous broadening. The solid
54 MITSURUSUGAWARA
"
1.40 1.& 1.!ill
(a) Wavelength (pn) (b) Wavelength (pm)
2 0 h . I - , . , . A
E
2
2.
i
u)
u
OO 100 200 300 400
(c) Temperature (K)
FIG. 1.9. Experiments on exciton optical absorption spectra (from Sugawara et al., 1990.
Copyright 1990 by The American Physical Society). (a) Measured (solid line) optical absorp-
tion spectrum of In,,,,Ga,~,,As/InP quantum wells at 4.2 K. The dashed line is Eq. (1.190)
with an FWHM of 4.7meV. (b) Measured (solid line) optical absorption spectrum of
In,,,,Ga,,,,As/lnP quantum wells at 295 K. The dashed line is calculated by Eq. (1.189) with
r,' = 150 fs. (c) FWHM of the ground-state electron-heavy-hole exciton resonance spectrum
as a function of temperature for three samples with almost the same composition and well
width (10-nm In,~,,Gao~,,As/lnP quantum wells). The solid curves are calculated by Eq.
(1.189),assuming the temperature-dependent phonon scattering rate of Eq. (1.191). The dashed
line is the thermal broadening due to Eq. (1.187).
curves are calculated assuming the temperature-dependent phonon scatter-

ing rate of
reg= rph[exp(ho,,/kT) - 13-l (1.191)
with the LO phonon energy of hw,, = 30 meV and r;,,' = 65 fs. The dashed
line is the FWHM of Eq. (1.187). Calculated values for the three samples
are in good agreement with measurements, supporting the assumption
1 OPTICAL PROPERTIES OF SEMICONDUCTOR NANO-STRUCTURES55
QUANTUM
concerning the origin of scattering. Since the LO phonon energy of 30 meV

is much larger than the exciton binding energy of about 6 meV, excitons are
scattered into the band-continuum state with a rate of reg reaching down to
(15Ofs)-' at room temperature. In almost all studies of the temperature
dependence of the exciton spectrum, the spectrum width is given by adding
+
up the FWHM of each broadening factor, that is, To 2hT,,. This is partly
because the treatment is much simpler and partly because the origin of the
inhomogeneous broadening is attributed to alloy or interface scattering.
Since actual quantum wells have structural inhomogeneity on a much larger
scale than that of the exciton radius, not all the broadening factors can have
a scattering origin. Thus, convoluted integration of Eq. (1.189) is much more
realistic.
5. SPONTANEOUS EMISSION
OF PHOTONS IN QUANTUM
WELLS
AND MESOSCOPIC
QUANTUM DISKS
In Section 111.2, we derived a formula for the lifetime of the electron-hole

recombination through the spontaneous emission process. The resultant
formula was independent of the dimension of semiconductor nano-struc-
tures except in the transition matrix, and gave nanosecond-order lifetime.
As will be seen in this section, the exciton spontaneous emission property
depends on the dimension and volume of the crystal and differs much from
that of the electron-hole pair. This is primarily due to the selection rule to
preserve the wave vector between the exciton center-of-mass motion and a
photon. In bulk materials, this characteristic forms an exciton-polariton that
propagates in the crystal. In quantum nano-structures, it is altered because
the translational symmetry is lost in the quantum-confinement direction.
Theories of exciton spontaneous emission lifetime in semiconductor
nano-structures have been presented for quantum wells and quantum dots
(Hanamura, 1988; Feldman et al., 1987; Andreani et al., 1991; Citrin, 1993;
Sugawara, 1995). In quantum wells, the wave vector selection rule holds for
the in-plane component, since the exciton motion is free in the two-
dimensional plane, and the lateral length of the quantum-well plane is much
greater than the exciton resonant wavelength. As long as the emitted photon
is not reabsorbed by the quantum well (this occurs in microcavities, as
discussed in Section IV.6), the two-dimensional coherent nature causes rapid
spontaneous emission on the order of picoseconds at low temperatures. In
quantum dots, the spontaneous emission rate increases in proportion to the
crystal volume. This occurs because, quantum dots have a size similar to the
exciton Bohr radius and are much smaller than the resonant wavelength;
thus, their electric dipole approximation of exp(iq.R) z 1 holds for the
56 MirsuRu SUGAWARA
center-of-mass motion, giving a transition matrix that is proportional to the

volume covered by the exciton wave function extent (Eq. (1.172)). The
expected crystal volume dependence is experimentally proved in CuCl
microcrystals and is thought to be due to the coherent nature of excitons
throughout entire microcrystals (Itoh, 1992).
In this section, theoretical formulae for the spontaneous emission lifetime
of excitons are derived in quantum wells and in mesoscopic semiconductor
quantum disks. The mesoscopic disks approach macroscopic quantum wells
as the disk diameter increases above the exciton resonant wavelength; they
approach microscopic quantum dots as the radius decreases below the
exciton Bohr radius. The lifetime formula of the disk indicates how the
spontaneous emission lifetime varies from quantum wells to quantum dots.
It is found that the lifetime increases two orders of magnitude as the disk
radius decreases.
Excitons in Quantum Wells
Let us consider a quantum a well with well width, L,, and an exciton
resonance wavelength, Leg, that satisfies the following relation:
Lz < 2u, < Leg << L (1.192)
Here, uB is the exciton Bohr radius in bulk materials and L2 = D is the area
of quantum-well crystals in the x-y plane. Under this condition, electron and
hole energies are quantized in the z direction, and excitons move freely in
the x-y plane. We also assume that the length of the crystal in the z
direction, L, is much larger than A,,. The electromagnetic field is quantized
in the crystal with the volume of V, = L3 and with the refractive index of n,.
The basic rules to derive a formula of exciton spontaneous emission
lifetime in quantum wells are as follows: (1) The selection rule of the in-plane
wave vector is taken into account between the electromagnetic field, 411, and
the exciton center-of-mass motion, KII, that is, KII = qll (Eq. (1.166)); (2) only
excitons with the wave vector within a critical value, qc, can radiate
spontaneously because of the energy conservation rule. The emission is in
the direction of (q”, q,), with q, = ,,/(n,o,g/c)2 - q1Iz.This is illustrated in
Fig. 1.5(b).
The procedure to derive the formula is almost identical to the one used
for the electron-hole pair in Section 111.2. The wave function for the
exciton-photon system is written as
1 OPTICAL
PROPERTIES NANO-STRUCTURES57
QUANTUM
OF SEMICONDUCTOR
where I@,,,O) is the exciton state with no photon and lag,lq,b)is the
ground steate with one photon. We assume the initial condition where one
exciton with the resonant frequency of we, is in a quantum well and the field
is in the vacuum state with a(0) = 1 and bq.b(0)= 0. The exciton drops into
the ground state by spontaneously emitting a photon with an energy of wq
and a wave vector of q. For the exciton-photon interaction Hamiltonian,
Eqs. (1.159) and (1.165) with n = 0 are used. By substituting Eq. (1.193) into
the Schrodinger equation of Eq. (1.63) and combining the equations of
motion for a(t) and b,,(t), we get
where the coupling constant is
(1.195)
The rate, yd, includes all the dissipation processes besides the spontaneous
emission, such as acoustic phonon scattering into nonradiative states in K
space, ionization due to longitudinal optical phonons, and nonradiative
recombination processes at crystal defects or impurities.
As long as the exciton interacts with the continuous optical mode, Eq.
(1.194) is reduced to Fermi's golden rule, as seen in Eq. (1.68). This is also
the case for excitons in quantum wells, which interact with a definite optical
mode in the in-plane direction and with continuous optical modes in the z
direction. So, we again get Fermi's rule as
4t)
b(t) = - 'z,[ + yd] 2 (1.196)
with
'si1 = 2'1 Igq,o12'(wq - w e g ) (1.197)

4."
By adding the two independent polarization components, the square of the

coupling constant becomes
58 MITSURUSUGAWARA
Since exciton states are formed mainly from the states around the inverse of
the relative motion radius, kll is replaced by’,1 in the polarization factor.
Considering the optical-mode density in the z direction, as in Section 111.2,
we obtain the spontaneous emission rate as
where ylI and q 2 are given in Eqs. (1.73) and (1.74). Comparing Eq. (1.72)
for the electron-hole pair and Eq. (1.199) for the exciton, the most
remarkable difference is the wave-vector selection rule.
The spontaneous emission spectrum is given as
( 1.200)
where y = z,pl + y d . The spectrum is a Lorentzian function with an FWHM

of y.
Let us check numerically the spontaneous emission lifetime of the
electron- heavy-hole exciton with K1l = 0 for the In,~,,Ga,,,,As/InP quan-
tum well with a well width of L, = 10nm. Material parameters are m,*" =
0 . 0 5 ~ V~I $; =0.12m0; M:" =0.035m,,; M*" =0.17~,; M,*' =0.05mn; m t L =
0 . 5 ~and
~ ; n, = 3.5. The electron masses are calculated from the first-order
k . p matrix and the hole masses from the Luttinger-Kohn parameters
obtained by analyzing the diamagnetic shifts (Sugawara et al., 1993a,b).
Equation (1.199) gives the lifetime of 23 ps, about two orders of magnitude
shorter than that of the electron-hole pair transition. This fast emission is
due to the selection rule to convey the oscillator strength of excitons into
one optical mode with the in-plane wave vector fixed.
However, the fast emission occurs only at low temperatures the exciton
thermalization above the critical wave number of q, in the center-of-mass
motion wave vector space. Since the energy of ti2qF/(2M*ll) is around 1 K,
at which low temperature excitons are localized at some potential fluctu-
ations, it seems almost impossible to observe a fast decay with the usual
photoluminescence measurements. The effect of exciton localization on
exciton's spontaneous emission lifetime can be modeled on the spontaneous
emission in quantum disks, as discussed below.
NANO-STRUCTURES
Excitons in Mesoscopic Quantum Disks
Let us discuss the spontaneous emission lifetime in quantum disks with a

width of L,, which is less than 2u,, and a radius of R , (Fig. l.lO(a)). The
disk consists of a direct band-gap semiconductor, and both an electron and
a hole are confined in the disk by the potential barriers of surrounding
regions. When R , is much greater than the exciton resonance wavelength in
the crystal, Ae,/n, (that is, R , >> &Jn,), the disk is a quantum well with a
well width of t,.When 2R, is approximately equal to L,, it is a quantum
dot. The disk is mesoscopic when ,Ie, < R , < A,,/2nr. For example, A,, =
16.0 nm, A,, E 1.6 pm, and n, = 3.5 in a 10-nm In,,,,Ga,,,,As/InP quantum
well. The origin of the coordinate axis is the center of the disk, and the z-axis
is perpendicular to the layer. In the z direction, the disk has the same carrier
confinement potential as that of quantum wells. The additional in-plane
potential gives three-dimensional quantum confinement. The disk is situated
inside the crystal with the size of L, which is much larger than the exciton
resonant wavelength. The radiation field is quantized in the crystal with a
volume of V , = L3 and a refractive index of n,.
Disk radius. R," (nm)

. .
,< 200 100 50 20 10 6
(a) (b) Confinement potential, w, (s-')
FIG. 1.10. Calculation of the exciton spontaneous emission lifetime (from Sugawara, 1995.
Copyright 1995 by The American Physical Society). (a) Quantum disk with a width L,, which
is less than 2u,, and a radius R,. The origin of the coordinate axis is the center of the disk,
and the z-axis is perpendicular to the layer. (b) Calculated spontaneous emission lifetime (solid
line) of the optically active ground-state excitons as a function of w, and R, by Eq. (1.208). The
disk is In,~,,Ga, 4 7 A with
~ a width of Lz = IOnm, and the surrounding barrier is InP. As R,
exceeds L,,/2n,, the lifetime gradually increases from a constant value. Under a strong
confinement of R , < A,,, the lifetime increases by two orders of magnitude. The horizontal
dashed line is the lifetime of free excitons with K" = 0 in quantum wells (23 ps), calculated from
Eq. (1.199), and the dotted line is that calculated from the electric dipole approximation of Eq.
(1.209).The lifetime calculated from Eq. (1.208) smoothly connects the two limiting cases.
60 MITSURUSUGAWARA
The disk is assumed to have the harmonic-oscillator-type in-plane para-

bolic potential described in Section IV.2. Let us expand the localized
center-of-mass motion wave function of Eq. (1.144) into a Fourier series to
apply the selection rule with the radiation field as
(The subscripts for the quantum numbers are added for clarity.) The
expansion is done in the whole crystal. From the periodic boundary
condition, KII = 27c(n,/L,n,/L), where ni is an integer (i = x and y). Note
that Eq. (1.201) is the superimposition of free-exciton wave functions with
various wave vectors. For the optically active excitons with 1 = 0 (see Eq.
(1.204) below), the Fourier coefficient in Eq. (1.201) is given as
p exp( - fl2K1I2/4)Lk(p2;Il2)
~
(1.202)
k!
where L, is the Laguerre polynomial. Equation (1.202) shows that, as fl

decreases, excitons become composed of the states with larger wave vectors.
For the ground state of k = 0, Eq. (1.202) gives
C,,(K~~)= & p exp( - p2K1I2/4) (1.203)
Using the relationship of
(1.204)
the disk radius is defined as
R, = *fl= 2,/- (1.205)
since 2np2 is the area covered by the center-of-mass motion.

The square of the coupling constant is
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 61
NANO-STRUCTURES
Note that in & I , ~ , I Eq. (1.198), for free excitons is replaced by lCk,(qll)lz.
Using the same procedure as before, the spontaneous emission rate on the
basis of Fermi's golden rule is given as
Using Eq. (1.202) for optically active excitons with 1 = 0, Eq. (1.207)
becomes
For the ground-state excitons (k = 1 = 0), Lk(P2q2/2)/k!= 1.

Equation (1.208) overlaps with the formula for quantum dots as fi
decreases; that of quantum wells, as p increases. When &,<< 1, that is,
j << Aex/nr, exp( -p2q2/2) 2 1 - pzq2/2 z 1, and L,(P2q2/2)/k ! 1. Then,
Eq. (1.208) reduces to
The spontaneous emission rate is proportional to the area covered by the

center-of-mass motion. We can derive Eq. (1.209) using the three-dimen-
sional mode density of the electromagnetic field quantized in the crystal with
the refractive index of n,, and using the electric dipole approximation of
exp(iq-R) = 1, which is valid for quantum dots much smaller than the
resonant wavelength. When the variable separation is prohibited, we have
only to replace the wave function as
(1.210)
where $ex(r, R) is the three-dimensionally confined exciton wave function.

For large p, we can substitute tl,(O), qz(0), and Lk(0)in the integral, so that
62 MlTSURU SUGAWARA
Eq. (1.208) reduces to
where z(,O) represents the spontaneous emission lifetime of free excitons

with K" = 0 in quantum wells (Eq. 1.199). Thus, as ,f? increases, Eq. (1.208)
approaches that for quantum wells. As a result, Eq. (1.208) for the exciton
spontaneous emission lifetime continuously connects the microscopic and
macroscopic limits via mesoscopic regions.
In Fig. l.lO(b), the solid line is the calculated spontaneous emission lifetime
of the optically active ground-state excitons as a function of w, and R , using
Eq. (1.208). The disk is In,,,,Ga,,,,As; the width is L, = lOnm, and the
surrounding barrier is InP. Since the topmost valence band is the heavy-hole
band in this sytem, an electron-heavy-hole exciton is considered. As R ,
exceeds IZeg/2nr,the lifetime gradually increases from a constant value. This is
because the center-of-mass motion wave function in K space spreads outside q,,
and the components that never contribute to spontaneous emission increase.
Under a strong confinement of R , < A,,, the lifetime increases by two orders of
magnitude, The horizontal dashed line is the lifetime of free excitons with
K = 0 in quantum wells (1 1.5 ps) calculated from Eq. (1.199), and the dotted
line is that calculated from the electric dipole approximation of Eq. (1.209).The
lifetime calculated from Eq. (1.208) smoothly connects the two limiting cases.
Temperature Dependence of the Exciton Lifetime
The spontaneous emission lifetime varies with temperature due to the

thermal distribution of excitons in K space. The way the lifetime depends on
temperature varies from quantum wells to quantum dots. In quantum wells,
the distribution of excitons outside q, increases the average exciton lifetime.
This principle for the lifetime increase is similar to the case of an exciton in
a quantum disk. Assuming that the thermal distribution in Kll space obeys
the Boltzmann distribution, and averaging the lifetime over K" space, the
temperature dependence of the exciton lifetime is given as
ts;'(T) z z,pl(0)[l - exp(- T,/T)]

2 Tsi'(o)(Tc/T)(T >> T) (1.212)
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 63
NANO-STRUCTURES
where T, = h2q2/2M*llk,. Since T, = 0.6 K in In,~,,Ga,,,,As, the lifetime

increases almost linearly with temperature. In quantum disks, using the
discrete energy levels of Eq. (1.140), the temperature dependence of the
lifetime for the ground-state excitons is written as
where AEf, = E i - Et0 and AEL", = E*,, - E\o. The denominator repre-
sents all the carrier distribution, and the numerator represents the carrier
distribution only to the optically active states. The lifetime increases as the
temperature increases due to the distribution of excitons to the inactive
states. As the quantum-disk radius decreases, the lifetime becomes insensi-
tive t o temperature due to the increase in the interlevel separation. In
quantum dots with an energy separation much larger than k,T = 25 meV,
the lifetime is kept almost constant up to room temperature.
E.uperirnents on the Relationship Between Exciton Localization and

Spontaneous Emission
Theoretical consideration of spontaneous emission in quantum disks

indicates that, if excitons are localized in quantum wells at low tempera-
tures, their spontaneous emission lifetime drastically increases. Experimental
evidence is provided here by comparing the spontaneous emission proper-
ties between free and localized states.
Figure 1.1l(a) shows the photoluminescence spectra of Ino,,,Ga,,,,As/
InP quantum wells, which can be attributed to the spontaneous emission of
the ground-state electron-heavy-hole excitons. Note that the peak wave-
length and intensity of the exciton spectra depend on temperature in an
unusual manner. While the peak wavelengths are almost the same at 2 K
and 4.2K, the spectrum at 20K is found at a shorter wavelength. The
intensity at 4.2K is slightly higher than that at 2 K . As temperature
increases above 20 K, the spectrum gradually shifts toward longer
wavelengths and its intensity decreases. The photoluminescence spectrum at
2 K appears in the lower energy of the exciton absorption resonance by a
Stokes shift of 4.5 meV.
Figure I.ll(b) shows the diamagnetic shifts of exciton spectra up to 12
Tesla for photoluminescence at 2,4.2,40, and 100 K, and optical absorption
64 MITSURUSUGAWARA
10
2.
.-
c
u)
2
c
.-
8
5 05
0c
.g-
0
I
0
E o
1 42 146 150
(4 Wavelength (pn)
Free excitm
(c)
0; A 40 M,
Temperature (K)
' eb ’A
(b) Magnetic field (Ta)
FIG. 1.11. Experiments on exciton spontaneous emission (from Sugawara, 1995. Copy-
right 1995 by The American Physical Society). (a) Photoluminescence spectra of 10-nm
In,,,,Gao,,,As/InP quantum wells resulting from the spontaneous emission of the ground-state
electron-heavy-hole excitons. (b) Diamagnetic shifts of exciton spectra up to 12 Tesla for
photoluminescence at 2,4.2,40, and 100 K and optical absorption at 4.2 K. The magnetic field
is applied perpendicular to the quantum-well layers. The error bars represent the shifts of
self-assembled In,~,,Ga,,,As quantum dots in Fig. l.l(b). Values of w, are indicated in Fig.
l.lO(b). (c) Peak intensity of photoluminescence as a function of temperature. The intensity of
localized excitons (below 20 K) is almost constant; the intensity of free excitons decreases as
temperature increases. The exciton localization significantly enhances the spontaneous emission
lifetime.
at 4.2 K. The magnetic field is applied perpendicularly to the quantum-well

layers. While the shifts of photoluminescence at 40 K and 100 K almost
agree with those of the optical absorption, they are clearly smaller at 2 and
4.2 K. Those at 20 K are in between (not shown). The 2- and 4.2-K shifts
can be attributed to the presence of additional in-plane confinement
potential. The error bars represent the shifts of the self-assembled
In,,,Ga,,,As quantum dots of Fig. l.l(b). Very small diamagnetic shifts
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 65
NANO-STRUCTURES
clearly demonstrate their strong three-dimensional quantum confinement.

The lateral confinement of excitons at low temperatures in the quantum well
is weaker than that of excitons in the quantum dots. The solid lines show
the calculation of the diamagnetic shifts by
V,II -
e2 ( e 2 B 2 I rn:lk.o~)
+-
4 7 - c ~ 8m,*11
~ ~ ~ ~2 ~ ~
]
rli2 Yr = E,Y, (1.214)
with m,*" = 0.035m0 and E = 13.9~,,using o,as a parameter. The harmonic-

oscillator potential is added to Eq. (1.150). Comparing the calculation with
the measured shifts, we find that excitons at 2 and 4.2K are laterally
confined with the potential of w, = 1 x 1013s-'. We obtain = 11.6 nm for
w, = 1 x 1013s-' using M*" = 0.17m0. The radius of the relative motion
wave function at a zero field decreases from Aex = 16.0 nm to 11.3 nm due
to lateral Confinement. At 40K and 100K, the confinement potential is
below the detection limit of 1 x 10" s-' and the excitons are almost free.
The diamagnetic shifts in Fig. 1.1 l(b) indicate that localized excitons below
20 K are subject to intermediate lateral confinement between free excitons
in quantum wells and three-dimensionally confined excitons in quantum
dots.
Figure 1.1l(c) plots the peak intensity of photoluminescence as a function
of temperature. Note that while the intensity of the localized excitons (below
20K) is almost constant, the intensity of the free excitons decreases as the
temperature increases. We analyze exciton photoluminescence intensity
using
(1.215)
where G is the generation rate of excitons by laser excitation, T~ is the

radiative spontaneous emission lifetime of free or bound excitons, and znr is
the nonradiative recombination lifetime. The solid line in Fig. l.ll(c) is
calculated from Eqs. (1.212) and (1.215) to fit the data for free excitons
/ T0.1
above 20 K ( T ~ ~ ( O ) = ~ ~ and G = 12 in arbitrary units). It is assumed
that the nonradiative recombination lifetime is independent of temperature.
The calculation indicates that, if the excitons were free even below 20K,
their intensity would continue to increase and become much stronger than
that of the localized excitons. The localized exciton emission intensity (the
horizontal dashed line) indicates that the localized exciton lifetime is
zB = 32tS,(0); that is, exciton localization significantly enhances sponta-
neous emission lifetime. The solid circle in Fig. 1.10(b) represents the
66 MITSURUSUGAWARA
localized exciton lifetime in In,,,,Ga,,,,As/InP quantum wells, plotted

using o,= 1 x 1013s-", z B = 32rS,(0),and z,(O) = 1 1 . 5 ~The
~ . estim-
ated lifetime agrees quite well with the calculation, supporting the theoreti-
cal prediction.
6. SPONTANEOUS EMISSION DISKS

OF PHOTONS IN QUANTUM
PLACED IN A PLANARMICROCAVITY
The exciton-photon interaction in semiconductor microcavities has at-

tracted much attention since the discovery of coupled-mode splitting be-
tween cavity photons and resonant excitons in quantum wells (Weisbuch et
al., 1992). The coupled mode is called a cavity polariton, from the analogy
of exciton polaritons in bulk semiconductors (Houdre et al., 1994), or
dressed excitons, from the analogy of dressed atoms in microwave cavities
(Cao et al., 1995). Many experiments in the spectral region, such as the
reflection, transmission, and photoluminescence spectra, not only revealed
how energy splitting depends on exciton-cavity-photon mode detuning,
oscillator strength, and quantum-well position in the cavity, but also
clarified exciton-polariton dispersion relations (Houdre et al., 1994; Zhang
et al., 1994). Also found was the oscillatory behavior of spontaneous
emission in the temporal region, which is analogous to Rabi oscillations in
atomic physics and is evidence for the coherent exciton-photon energy
exchange (Cao et al., 1995; Norris et al., 1994; Jacobson et al., 1995). These
phenomena contribute greatly to semiconductor physics in that we are now
able to control exciton-polariton behavior by artificially designing micro-
cavity structures. This might lead to new types of semiconductor devices,
such as THz electromagnetic field generators (Kadoya et al., 1995).
As we saw in Section IV.5, exciton spontaneous emission occurs by
preserving the wave vector between the exciton center-of-mass motion and
the emitted photon. In quantum wells, the wave vector selection rule for the
quantum-well plane component gives rise to fast decay with a spontaneous
emission lifetime of about lops. The decay is exponential due to the
exciton's coupling to the z-direction continuous optical mode. If the z-
direction continuous photon mode is changed into discrete modes using a
planar cavity, one-to-one correspondence is realized between the exciton
and photon modes for all directions, leading to strong coupling where the
coherent exciton-photon energy exchange occurs.
Here, the mesoscopic semiconductor disk in Section IV.5 is placed in a
planar cavity to observe the cavity effect on the exciton spontaneous emis-
sion. Excitons resonant with a cavity-photon mode show cavity-enhanced
spontaneous emission. The enhancement is more remarkable as the disk
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 67
NANO-STRUCTURES
radius increases and the exciton's in-plane coupling to the photon mode
becomes stronger. When the disk radius is much larger than the exciton
resonant wavelength and the cavity photon decay rate is smaller than the
exciton-photon coupling constant, the exciton-photon system transits from
the weak coupling to the strong-coupling regime, leading to oscillating
spontaneous emission decay. The temporal evolution of the emission spec-
trum is evaluated to show coupled-mode spectrum splitting.
Figure 1.12(a) shows the coordinate system, where the x- and y-axes lie
in the mirror plane and the z-axis is normal to the mirrors. The mirrors are
assumed to be spaced a distance, L , apart, and have an area, D = L2, where
the length of mirrors in the x and y directions is infinite compared to the
exciton resonant wavelength, ;leg. The planar cavity is assumed to be ideal
in the sense that it has a reflectivity, R,, that is constant not only as a
function of wavelength but also as a function of the incident angle. The
radiation field is quantized in the cavity with a volume of V , = L L 2 and a
refractive index of n,. The quantum disk is placed in the middle of the cavity
with its center at z = 0.
Figure 12(b) represents the wave number space of the cavity-disk system.
Due to the resonant condition, only excitons with an in-plane wave number
less than q, = n,o,,/c can radiate spontaneously. When the wave number of
the cavity mode in the z direction is fixed at q, and the resonant condition
of weg= c(ql12 + qZ)'i2/n, is satisfied, the excitons emit photons in the
direction of (K1l,qz). For real Bragg mirrors with alternate high and low
refractive indices, the photon emitted in the direction outside a critical angle
connects with the electromagnetic-field background in free space. This
means that excitons with the in-plane wave vector below a critical value, qs,
are connected into the cavity mode and excitons with qs < K d y, are
connected into the open-side mode.
It is assumed that the photons repeatedly reflected between the mirrors
interact with the disk every time they pass the middle of the cavity. This
assumption is particularly important when the cavity photon proceeds in
the declined direction and the disk radius is smaller than the resonant
wavelength. Here, to satisfy this condition, it is assumed that disks are
packed in the z = 0 plane to the extent that the distance between neighbor-
ing disks is significantly smaller than the exciton resonant wavelength.
However, the overlap of exciton wave functions between neighboring wells
is still assumed to be negligible. The photons emitted in the declined
direction due to the nonzero in-plane wave vector component depart from
the initial disk and interact with another one. This is the exciton transfer
between the disks in the in-plane direction.
The inclusion of the cavity effect in the exciton decay problem is not
straightforward, although the basic strategy to expand the exciton wave
68 MITSURUSUGAWARA
(a Disk radius, Ro (nm)

FIG. 1.12. Calculation of exciton spontaneous emission in the microcavity (from Sugawara,
1997h). (a) Quantum disk placed in a planar cavity. The x- and y- axes lie in the mirror plane,
and the z-axis is normal to the mirrors. The quantum disk is placed in the middle of the cavity
with its center at z = 0. The mirrors are assumed to be spaced a distance, L', apart and have
an area of D = Lz, where the length of the mirrors in the x and y directions is infinite compared
to the exciton resonant wavelength, I,,,. (b) Spontaneous emission into the cavity mode with
the wave vector q = (qll, qz),where q" is equal to the exciton center-of-mass motion wave vector
KI1. When 4. < K" i4,. the emission is into the open-side mode. (c) Spontaneous emission
lifetime of excitons in the 10-nm In,,,,Ga,,,,As/InP system as a function of the disk radius,
R, =b , /3. The longitudinal dotted lines indicate half of the exciton resonant wavelength,
ieX/(2nr)P 200 nm, and the in-plane exciton radius for quantum wells, lex = 16.0 nm. The
dashed line is for free space without a cavity. The solid lines are for the cavity with a variety
of photon lifetimes, K-’.
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 69
NANO-STRUCTURES
function in K" space need not be changed. First, the z-direction optical-
mode density should be changed from the continuous one in a free space to
a definite one in a cavity. Second, the formula for the coupling constant
should be modified to include the standing-wave electric field in the cavity.
Finally, instead of the simple Fermi's rule, a general expression that is valid
regardless of coupling strength, should be used, since the coupling strength
varies continuously from the strong-coupling regime to the weak-coupling
regime depending on the disk radius. All of these processes bring consider-
able changes in the resultant formulae, as shown below.
The exciton decay in a planar microcavity is typically characterized by
the following coupling constants: the dipole coupling constant between the
exciton and the cavity-mode photon, y; the decay rate of the cavity-mode
photon via mirror losses, K ; the spontaneous emission rate into the open-
side modes due to the electromagnetic field background, yo; and the exciton
dissipation rate besides spontaneous emission, yd. The photon decay rate is
given as
C
K = ---lnR, (1.216)
n, L’
As shown in Fig. I.l2(b), yo depends on the critical in-plane wave vector, qs,
above which excitons are connected to the open-side mode.
The vector potential in a planar cavity is modified from Eq. (1.59) by
internal reflection and interference. For the S-polarized mode (its electric
vector is normal to the plane defined by q and the z-axis), we obtain
For the P-polarized mode (its electric vector is in the plane defined by q and
the z-axis), we obtain
x [&4+ a,(t) + e-'q"i?:(t)] (1.2 18)

The z-direction wave vector is given as
qz = mn/L' ( m = 0,1,2,. . .) (1.219)

and the frequency of the cavity-photon mode is
(1.220)
70 MITSURUSUGAWARA
The wave function for the exciton-photon system is the same as in Eq.
(1.193). Again, the initial condition is that one exciton is in a quantum disk
and the field is in the vacuum state with a(0) = 1 and b,,(O) = 0. The
exciton drops into the ground state by spontaneously emitting a photon
with an energy of ho, and a momentum of hq. Using the rotating wave
approximation, the equations of motion for the probability amplitude of a(t)
and b,,(t) are found as
and
(1.222)
Taking into account both P and S polarization, the coupling constant is

given as
where
The upper term in parentheses corresponds to the free excitons in quantum

wells, and the lower term corresponds to the localized excitons in disks.
From Eqs. (1.221) and (1.222), as before, we obtain
Depending on the way the exciton interacts with the electromagnetic field,
Eq. (1.225) gives the following types of solutions:
Decay into the open-side mode in the cavity: The components with the
wave vector in the range of q, < q d q, contribute to this decay with a
rate of yo.
PROPERTIES QUANTUM 71
NANO-STRUCTURES
Weak coupling with the cavity mode: For K >> g, the emitted photon
escapes from the cavity and is barely reabsorbed by the quantum disks.
In this weak coupling regime, a ( t ) can be removed from the integral.
Thus, the remaining integral gives, for t >> C1,
(1.226)
where
and
( 1.228)
is the normalized Lorentzian function representing the cavity-mode

broadening with the FWHM of K. Equation (1.226) gives the exponen-
tial decay as in the free space. The validity of the weak-coupling
formula is ensured if the calculated result of zspl is sufficiently smaller
than K . This corresponds to the situation in which the cavity-mode
broadening of Eq. (1.228) is greater than the homogeneous broadening
of exciton emission given by Eq. (1.200) and the optical mode is
considered to be still continuous. If K is nearly equal to or less than z5G1,
we should resort to the general solution below.
Strong coupling with the cavity mode (a general case in the cavity):
When the coupling between the exciton and the cavity is so strong that
a photon emitted into the cavity is likely to be absorbed before it
escapes, the approach in the weak-coupling regime ceases to be justified
(Goldstein and Meystre, 1995). In this case, since a(t') changes more
rapidly than the exponential dissipation factor, we should numerically
evaluate Eq. (1.225). By transforming the variable of the integral, we
obtain
a(t - s)f(s)eCKSI2
ds - (t+ $) a(t) (1.229)
72 MITSURUSUGAWARA
where
Analytical expressions for the strong-coupling limit can be obtained only

for free excitons in quantum wells that interact with a definite cavity mode.
By omitting the q summation in Eq. (1.221), the second derivative equation
on a(c) can be derived. The solution to the equation is written as
a(t) = clen" + c2en2' (1.231)
where
c(1,2= (y,
- - -+-++
2 2
IC( j ,
2
) +' [(h---
-2 2
ti
2
-
i(jqy ~cP]'" (1.232)
Q2 = 2 1 lgq,aI2 (1.233)
o=P.S
6, = oq- ueg, q = (Kll, qz), and cl.z are numerical coefficients determined
by the initial conditions of a(0) = 1 and h(0) = -yd/2. The factor of two in
Eq. (1.233) represents the two z directions, that is, the front and back sides
of the cavity. For resonant excitons with 6, = 0, Eq. (1.231) becomes
cos(@t/2)+ -sin(Qt/2)]
20 (1.234)
with
(1.235)
and with the strong-coupling condition of
4 0 - ( y d - ti)2/4 > 0 (1.236)
The frequency of Q is known as the Rabi frequency in atomic physics.

1 OPTICAL PROPERTIES OF SEMICONDUCTOR NANO-STRUCTURES73
QUANTUM
From Eqs. (1.221) and (1.222), the coefficient for each optical mode is
given as
The emitted photon power at a time, f, and at a frequency between w and

w + dw is given as
_- 1
(1.238)
Equation (1.238) gives the temporal evolution of the emission spectra.
Numerical Calculations
Let us use the half-wavelength cavity with the cavity space as L' =
Aeg</2nr.By changing ( around unity, the slight detuning from the exact
<
half-wavelength cavity of = 1 is expressed. From Eqs. (1.219) and (1.220),
the cavity mode matching the exciton resonance energy is
2nmn,
q; = (rn = 0,l) (1.239)
i
~
&g
Only the m = 1 mode is taken into account since the rn = 0 mode is coupled
to the exciton with q = q, (>q,) and should be treated as the open-side
decay in a real planar microcavity. For simplicity, the dissipation rate into
the open-side mode is not taken into account here. The spontaneous
emission of the optically active ground-state excitons is calculated for the
In,,,,Ga,,,,As/InP disk with a thickness of L, = 10nm. Free excitons in
quantum wells have a resonant wavelength of Leg = 1.45 pm, a resonant
frequency of weg= 1.30 x l 0 l 5 s-' at 4.2K, and an in-plane radius of
16 nm. The exciton resonant wavelength in the cavity is approximately given
as Aeg/nr2 400 nm. The dissipation rate of yd is neglected so that the cavity
effect can be clearly seen.
14 MITSURUSUGAWARA
For K1I= 0 excitons in the exact half-wavelength cavity of [ = 1, Eq.

(1.227) becomes
( 1.240)
The rate is 4oeg/lc71times larger than that of K = 0 excitons in a free space

given by Eq. (1.199). This effect is well known as cauity-enhanced sponta-
neous emission. The solid line in Fig. 1.12(c) represents the spontaneous
emission lifetime of Eq. (1.227) into the m = 1 mode as a function of the disk
radius for the exact half-wavelength cavity (i= 1) with K - ~= 240fs, 51 fs,
23 fs, and 11 fs ( R P= 0.99, 0.95, 0.9, and 0.8, respectively), and the dashed
line is the lifetime in free space. In quantum wells-disks with a radius
much larger than the exciton resonant wavelength- the emission rate is
enhanced by about two or three orders of magnitude from that in a free
space as the photon lifetime increases. In the cavity with K - = 240 fs, the
exciton lifetime is reduced below the photon lifetime, and the transition from
the weak-coupling regime to the strong-coupling regime occurs. Needless to
say, Eq. (1.240) can not be applied in this strong-coupling regime. It is clear
from Eq. (1.227) that the emission of excitons with KI1 > 0 is suppressed
since no optical mode is available. This condition is known as inhibited
spontaneous emission.
The lifetime increases from a constant value with the decrease in R,
because localized excitons in quantum disks spread KII-space wave functions
and connect only weakly to the cavity mode. Note that, while the emission
rate of free excitons in quantum wells is enhanced by two or three orders of
magnitude due to the cavity effect, the rate in the dots is hardly changed by
the planar cavity. Only about a 30% increase in the emission rate is expected
even under a very small photon decay rate. This weak coupling to the cavity
mode comes from the extended exciton wave function in K" space. To realize
the strong coupling between excitons in quantum dots and the photon
mode, we need to control the cavity photon mode three-dimensionally.
Figure 1.13(a) represents la(t)12 for K" = 0 excitons in the exact half-
wavelength cavity as a function of t for K-’ = 240 fs, 80 fs, 51 fs, 23 fs, and
l l f s ( R P = 0.99, 0.97, 0.95, 0.9, and 0.8, respectively), with yd = 0. The
minimum point of the oscillation should approach zero, and its finite value
merely comes from the finite discrete time interval used in the calculation.
While the almost exponential decay occurs at K - = ~ 23 fs and 11 fs, the
oscillatory behavior appears at K-’ = 240 fs and 80 fs. In the intermediate
case of K - = ~ 51 fs, the curve is not completely exponential at the initial
decay of 0 to 200fs. The oscillation approaches la(t)12= cos2(Rt) as the
NANO-STRUCTURES
Time (ps)
(a)
(c) Time (ps)
FIG. 1.13. Calculation of exciton spontaneous emission in the microcavity (from Sugawara,
1997). ( a ) Probability of ln(t)12 for K = 0 excitons in the 10-nm In, ,,Ga, ,,As/InP quantum
wells as a function of r. The photon lifetimes in the half-wavelength cavity (i = I ) are
ti-’ = 240fs, 8Ofs, 51 fs, 23fs, and 11 fs. (b) Probability of la(f)I2 for K = 0 excitons in the
10-nm In,,,,Ga,,,,As/InP quantum wells when I < - ' = 240fs at a variety of detunings i. (c)
Probability of la(t)12 for the ground-state excitons in the In, ,,Ga,,,,As/InP quantum disks in
'
the half-wavelength cavity (( = 1) when ti- = 2.4 p.
photon lifetime increases. For K = 0 resonant excitons in the half-

wavelength cavity, Eq. (1.233) gives
(1.241)
The cavity-exciton detuning dependence of exciton emission decay is

shown in Fig. l.l3(b), which describes the temporal evolution of lu(t)12 for
KII = 0 excitons when IC-’ = 240 fs at a variety of detunings, i.Note that
76 MITSURUSUGAWARA
the m = 1 mode frequency is weg/(.As increases, the contrast of oscil-

latory decay becomes unclear and the oscillation frequency increases. Even
at 0.5% detuning, the oscillation disappears after a few repetitions.
Figure 1.13(c) represents la(t)12 in the exact half-wavelength cavity as a
function o f t for the disk radius of R , = 7400, 2300, 1200, 400, and 200nm
when K - = 2.4 ps. As R , decreases, the exciton wave function spreads in K
apace and the exciton-cavity-mode coupling is weakened, resulting in a
decrease in the oscillation frequency and the oscillation amplitude. At
around 400 to 1200 nm, the oscillation gradually disappears, as expected,
from the weak-coupling calculation in Fig. 1.12(c). Considering that the
exciton resonant wavelength is Leg/nr z 400 nm, the radius should be larger
than the resonant wavelength to exhibit the cavity polariton, that is,
Ro > L e g l n r .
Figure 1.14 shows the temporal evolution of the emission spectra for (a)
R, = 200 nm and (b) R , = 7400 nm, where IF(w - oeg, t)I2 in Eq. (1.238) is
plotted as a function of w - we, for 4' = 1 and K - ' = 2 . 4 ~ For ~ . R, =
200nm, where the almost exponential decay is observed in Fig. l.l3(c), we
obtain emission spectra at w = w,, - 0.3 [THz] with a fullwidth at half
maximum proportional to the inverse of t . Considering that the exciton
resonance frequency is weg= 1.30 x 1015s-', the detuning of the emis-
sion corresponds to about 2.3 x 10-40,g and is almost impossible to
detect in experiments. For R , = 7400 nm, we observe double peaks around
o = weg separated by 2!2 = 8.4THz, where SZ is the oscillation frequency
calculated from Eq. (1.241). Note that almost one oscillation period (0.75 ps)
is necessary to observe the double peaks, showing that the coupled state-
101 ' . ' ' ! , ' ' ' I , ' ' . I ' ' ' ' 1
(a) Detuning, o--oeg (THz) (b) Detunlng, o--ww(THz)
FIG. 1.14. (F(w- weg,t)I2 in Eq. (1.238) as a function of Q - w , for

~ (a) R , = 200nm and
(b) R , = 7400nm, representing the temporal evolution of the emission spectra calculation on
exciton spontaneous emission in the microcavity (from Sugawara, 1997b).
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 77
NANO-STRUCTURES
the cavity-polariton -is formed through one period of energy transfer

between an exciton and a photon.
The condition for excitons to show reversible spontaneous emission is
summarized as follows. First, excitons should interact with a definite optical
mode. This condition is satisfied for free excitons with a definite wave vector,
K", in quantum wells, since they interact with only one photon mode with
the wave vector (K1I,qz).This is due to the in-plane selection rule and the
cavity-induced z-direction discrete optical mode. As the disk radius de-
creases below the exciton resonant wavelength, excitons begin to interact
with continuous photon modes due to the spread of the wave function in K
space, resulting in the destructive interaction between different modes
leading to an exponential decay. Second, the photon decay rate should be
smaller than the exciton emission rate; otherwise, the optical mode of Eq.
(1.228) is broader than the homogeneous broadening of exciton resonance,
which is equivalent to the situation in which excitons couple with continu-
ous optical modes. Third, the exciton resonant frequency should closely
match the cavity-mode frequency within 2R, since the detuning causes
destructive interference in the time domain.
To summarize, the conditions are (1) R , > Leg/nr; (2) zs;' > rc; and (3)
Iw, - wegl/oeg < 2Q/w,,, where w, is the cavity-matching frequency. Both
the homogeneous broadening due to yd and the inhomogneous broadening
due to structural imperfections smear the oscillation, since they enhance the
detuning from the cavity-photon mode.
7. THECOULOMB
EFFECTON OPTICALGAINSPECTRA
That excitons take part in lasing has long been discussed in wide band
gap materials like 11-VI semiconductors with large binding energies (Koch
et al., 1978, Haug and Koch, 1977, Klingshirn and Haug, 1981). This topic
is once again of broad interest, given that green to blue semiconductor
quantum-well lasers have been produced primarily owing to the success of
high-density p-type doping. Room-temperature operation of ZnSe-based
(Haase, 1991) and GaN-based (Nakamura et al., 1996), lasers has had a
great impact on semiconductor optical technology, since short-wavelength
lasers are the key to next-generation high-density optical data storage
systems. Studies of the lasing mechanism in wide-gap quantum-well lasers
have become critical in the creation of practical lasers and in their improved
performance.
Exciton-related lasing in 11-VI wide-gap semiconductor quantum wells
was first suggested by experiments showing that lasing in ZnCdSe/ZnSe
multiple quantum wells occurred around the tail of the free-exciton optical
78 MITSURUSUGAWARA
absorption resonance (Ding et al., 1992, 1993). Ding and his colleagues
proposed a mechanism by which excitons created by photo excitation diffuse
into the low-energy tail of inhomogeneous broadening and cause population
inversion, leading to lasing. Experimental data have since suggested various
kinds of exciton-related lasing mechanisms in quantum wells (for example,
Kuroda et al., 1992; Jen et al., 1993; Kawakami et al., 1994; Masumoto et
al., 1993; Kreller et al., 1995; Kozlov et al., 1996).
We should discount the simple explanation that the large exciton-binding
energy of wide-gap materials stabilizes exciton states, resulting in exciton
lasing. This is because excitons as bound states between an electron and a
hole cannot exist under the population inversion, as the semiconductor
Bloch equation predicts (Haug and Koch, 1994; Uenoyama, 1995). Thus, in
order for bound-state excitons t o generate a gain, the exciton emission
energy should be shifted from the resonant exciton absorption energy
through some mechanism. Possible emission processes are: (1) radiative
recombination assisted by exciton-exciton, exciton-optical phonon, and
exciton-electron (hole) scattering, (2) exciton localization; and (3) bi-ex-
citon-exciton transition. All of these processes need the participation of a
third partner besides the photon and excitons: scattering particles; local
potential minima; bi-excitons. In this sense, exciton lasing occurs in very
special circumstances. See Galbraith, 1995 on the calculation of the scatter-
ing-related mechanism. He showed that, in the ZnSe bulk active layer, the
exciton-electron scattering process can cause a room-temper-
ature optical gain of up to 100cm-’ at a lower carrier density than the
Mott density of about 5 x 1 0 ” ~ m - ~This . calculation suggests that
exciton-related lasing is possible if the cavity loss is well restricted with a
high-quality long cavity and high-reflectivity mirrors. Lasing due to bi-
exciton-exciton transition is discussed in Section V.2.
Though exciton lasing will occur only in special cases, the Coulomb
interaction itself considerably affects optical gain. Here, the semiconductor
Bloch equation is summarized to see how the optical gain spectrum in
Section 111.1 is changed by the Coulomb interaction. The advantage of
Coulomb enhancement of gain for wide-gap laser performance is discussed.
Outline of the Semiconductor Bloch Equation
In Section 111.1, the optical susceptibility of semiconductors was derived

using a density-matrix equation where no electron-hole Coulomb interac-
tion was taken into account. As the carrier density increases, the population
factor of .f;,.k- fc,k becomes negative, generating the optical gain (Eq. 1.42)).
In Section IV.4, the exciton optical absorption spectrum was derived using
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 19
NANO-STRUCTURES
a similar density-matrix equation with the exciton state as a base function.

However, this treatment holds only for low or intermediate excitation where
the interaction among excitons is so weak that the exciton base function can
be used safely. Even so, in the high-excitation regime where the system
transits into the electron-hole plasma state, the treatment breaks down. So,
in order to describe the Coulomb effect on optical gain, a more general
treatment is necessary where the base function of the system is not assumed
a priori.
The semiconductor Bloch equation describes the continuous variation of
optical susceptibility from the low-excitation exciton regime to the high-
excitation electron-hole plasma regime (Haug and Koch, 1994). The strat-
egy to obtain a gain formula is to solve the Heisenberg equation of
(1.242)
where He, is the unperturbed electron and hole Hamiltonian of Eq. (1.115)
and H I is the Hamiltonian resulting from the interaction with the photons
of Eq. (1.152). The polarization of the system is given as
4
P(t) = - Qd d3r($+(r, t)ee; r$(r, t))
(1.243)
where
( 1.244)
is called a pair function and Qd is the inverse of the crystal volume as defined
before. Here, the two-band model and the k-selection rule are adopted, as
in Section 111.1. The problem is to determine the pair function of Eq. (1.244)
by solving the Heisenberg equation of Eq. (1.242). The four-operator terms
coming from the Coulomb interaction term in Eq. (1.242) are approximately
factorized into the products of the two operators: one two-operator term
describing the carrier density and the other describing the interband pair
function. The product of two operators with different wave vectors vanishes
by taking the time average since they have different rotation frequencies.
The semiconductor Bloch equation thus derived represents the kinetic
motion of the pair function under the quasi-equilibrium condition. Its
80 MITSURUSUGAWARA
Fourier-transformed equation is
(1.245)
where
represents the electron (i = ej and hole (i = h) frequencies red-shifted from

the band-edge frequency of w,,k due to the screened exchange, C,,, and
Coulomb-hole self-energies, ZcH; 6 represent the relaxation rate of the pair
function; and Y k - k l is the Fourier-transformed screened Coulomb potential
with a positive value. The Fourier transformation of the polarization is
given as
Equation (1.245) is the simultaneous equation of Qcrrk, and can be solved by

numerically inverting a matrix. It is clear that the Coulomb interaction in
the nondiagonal terms mixes different wave vector states, giving rise to
variation in the polarization.
Figure 1.15(a) shows the calculated optical gain spectra for 6-nm GaN-
Ale. 14Ga0,86Nquantum wells for various carrier densities (Chow et al.,
1996). The exciton optical absorption resonance appears at a low excitation
of 1 x 10’2cm-2, and the optical gain appears in the tail of the exciton
resonance at high excitation. The gain peak is red-shifted with respect to the
first absorption resonance. The shift is the net result of energy shifts, both
due to band-filling, giving a blue shift that depends on the band structure
and, due to the band-gap renormalization, giving a large red shift due to the
term of C,, + &,. Figure 1.15(b) shows the gain spectra at the carrier
density of 7 x 1012cm-’ with (solid curve) and without (dashed curvej
Coulomb effects, showing that the gain is enhanced as a result of the
Coulomb interaction. This is called the Coulomb, or excitonic, enhancement
of gain.
In order to know what the Bloch equation means, let us rewrite Eq.
(1.245) by setting the k’ summation of the product of the Coulomb potential
and the pair function as
(1.248)
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 81
NANO-STRUCTURES
I I I I
'E -1 E
0 0
U
0
0 -12
m
r F
v w
.-c -3
d
-5 -28
-0.1 0.0 0.1 0.2 -0.1 0.1 0.3
(a) fi%-Ego (ev) (b) h0-Ea (ev)
FIG. 1.15. (a) Computed TE gain spectra at T = 300 K for 6-nm GaN-AI, ,,Ga,,s6N at
various carrier densities. (Reprinted with permission from Chow et al., 1996. Copyright 1996,
American Institute of Physics.) (b) Gain spectra at the carrier density of 7 x l O " c r ~ - ~
computed with (solid curve) and without (dashed curve) Coulomb enhancement. E,, is the
unexcited bulk material band-gap energy. (Reprinted with permission from Chow et al., 1996.
Copyright 1996, American Institute of Physics.)
The reason for Eq. (1.248) is that, the mixing of different wave-vector
states under the Coulomb interaction, causes QL.,ck to spread in k space
and there is a finite range around k, where Qu,ck can be regarded as almost
constant. The term V, represents the summation of t $ - k / in the area
and has a positive real value. The term V, is the rest of the summa-
tion of Yk-k,Qc,ck(u)on k'. When the Coulomb interaction is zero, that
is, Yk-k'l = 0, only a resonant term of Qc.,ck responds to the electro-
magnetic field. Due to the mixing of different k states through the
Coulomb interaction, the nonresonant k state takes part in the polar-
ization. The k-state range that gives V, depends on the occupation
factor of fr,k - fl,k. In the case of j C . k- h , k = 0, the range is approxi-
mately Ik - k'l < l/Aex, where Aex is the exciton radius. As the occupa-
tion factor increases, the range decreases. When the population inversion
is achieved with L..k - fL.k > 0, the polarization concentrates on the res-
onant k state and V, approaches zero, since the Coulomb interaction
changes the sign and becomes repulsive, as seen in Eq. (1.245). However, V'
still survive, under this condition due to the states with A , k - f;.,k < 0 in the
higher energy region. This is the origin of the Coulomb, or excitonic,
enhancement of gain. Then, from Eqs. (1.245) and (1.248), the pair function
82 MITSURUSUGAWARA
is written as
Using the imaginary part of the polarization derived from Eqs. (1.247) and
(1.249), the optical gain becomes
where the broadening function is given as
and Sk(w) represents the variation of amplitude due to the Coulomb

+
interaction. When fE,k fh,k = 0, Sk(w)at w = ee,k+ ejt,k is the Sommerfeld
factor. Equation (1.250)is reduced to Eq. (1.42) by setting V, = 0, Sk(w) = 1,
and = w , , ~ Comparing
. Eqs. (1.251) and (1.42), the influence of the
Coulomb effect on optical gain is to change the absolute value above the
band edge and to cause the red shift due to the screened exchange and
Coulomb-hole effects. Homogeneous broadening of optical gain also occurs
due to carrier-carrier Coulomb scattering (see Chapter 7 and the latest
news chapter).
The above discussion is schematically depicted in Fig. 1.16. Whether the
system absorbs light or generates gain is simply determined by the popula-
tion term of , h , k - f i , k . Equations (1.250) and (1.251) tell us that, when -
fL.k < 0, an optical absorption resonance corresponding to the excitons
appears at an energy of around ( h , k - f i , k ) V l below the band edge. The term
of - ( L . k - A , k ) v l corresponds to the exciton binding energy, which de-
creases as the carrier population increases and vanishes at fi.k - f i , k = 0.
At the same time, the band-gap renormalization proceeds, and gain appears
in the exciton tail when the population is positive. No gain appears below
the band edge under the population inversion of . f r , k - L,,k > 0, since
(fc,k - J,k)T/1 > 0. In this sense, bound-state excitons are not formed under
the population inversion and never contribute to lasing.
What is the advantage of Coulomb gain enhancement on wide-gap laser
performance? The most remarkable is that the enhancement might reduce
the number of multiple quantum-well layers necessary for lasing. Wide-gap
materials like InGaN have large-hole effective mass and thus low-hole
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 83
NANO-STRUCTURES
Binding energy = ( f v - fC),EB

't I 4- '
/
t: I
E
- " Boundkate"
.-0 0 exciton
-
CI
Q
=I
Renormalization
r...\i
a
0
n .E
d Unbound stale
-’4 Energy
FIG. 1.16. Schematic of optical absorption and gain spectra under the Coulomb interaction.
mobility. This material character gives rise to inhomogeneous hole distribu-

tion between quantum wells, such as high-hole concentration in quantum
wells in the vicinity of the p-n junction and low-hole concentration in
quantum wells far from the p-n junction. As a result, the total material gain
is lowered at a given injected current, since quantum wells in the far position
work as light absorbers in an extreme case due to low-hole density. If the
Coulomb enhancement works well and only one or two quantum-well layers
are enough for the lasing threshold, high performance of wide-gap lasers will
result.
V. Quantum-Dot Lasers
Quantum dots have long been expected to drastically improve the

performance of semiconductor lasers because their delta-function-like den-
sity of states due to three-dimensional carrier confinement enhances gain
and differential gain, suppresses the thermal distribution of carriers, gives a
nearly zero alpha parameter at the peak gain, and so forth. However, since
their potential was first predicted theoretically (Arakawa and Sakaki, 1982)
more than ten years ago, quantum-dot lasers have been merely a dream
primarily because their fabrication is so difficult. This problem has now been
84 MITSURUSUCAWARA
overcome by the self-assembling process during highly strained epitaxy. In

addition to the InGaAs quantum dots treated in this book, a wide variety
of others consisting of various semiconductor materials have been fabricated
through self-assembling processes, such as InGaP (Carlsson et al., 1995;
Reaves et al., 1995), CdSe (Arita et al., 1997), and G a N (Tanaka et al., 1996,
1997). Now, room-temperature lasing from quantum-dot quantized levels is
viable, and the threshold current has decreased comparably to that of
strained quantum-well lasers (see Chapters 6 and 8). In the near future, this
new material category is expected to improve semiconductor laser perform-
ance in many aspects, such as threshold current, external quantum efficiency,
temperature stability, modulation speed and spectrum linewidth, and in
many wavelength regions ranging from infrared to red to blue.
Three physically and practically important aspects of quantum dots were
not noted in the earlier discussion on quantum-dot laser performance. They
are the speed of carrier relaxation into quantum dots, homogeneous
broadening of the optical gain spectra of single dots, and the Coulomb effect
acting on electrons and holes localized in quantum dots.
Retarded carrier relaxation appears to be an inherent problem with the
discrete energy levels of quantum dots, since the transition between the
discrete levels is significantly slowed due the lack of phonons needed to
satisfy the energy conservation rule. This is the so-called phonon bottleneck
problem (see Chapter 5). Though this was at one time merely a theoretical
problem (Bockelmann and Bastard, 1990 Benisty et al., 1991; Inoshita and
Sakaki, 1992), the recent advent of self-assembled quantum dots has enabled
us to evaluate carrier relaxation lifetime experimentally. Retarded carrier
relaxation, with a lifetime ranging from several tens to hundreds of
picoseconds, has been observed in many experiments using time-resolved
photoluminescence, some of which confirm the importance of phonon
resonant excitation conditions to accelerate relaxation (Mukai et al., 1996a
and b; Raymond et al., 1995, 1996). Recent improved theoretical calculations
are rather optimistic (Bockelmann and Egeler, 1992; Efros et al., 1995;
Gerard, 1995; Nakayama and Arakawa, 1994). For example, an Auger-like
process has been suggested for possible fast relaxation channels with a
lifetime of ten picoseconds or less. Such fast relaxation is reported to
manifest itself in the quick rise of time-resolved photoluminescence and in
the initial stage of carrier relaxation, where a large number of carriers
surround quantum dots (Ohnesorge et al., 1996; Grosse et al., 1997;
Bockelmann et al., 1997). Although more detailed theoretical and experi-
mental work is needed to clarify the relaxation mechanisms, we are sure that
the photon bottleneck problem does exist in the sense that the relaxation in
quantum dots is slower than in quantum wells, with a typical intraband
relaxation lifetime of 0.1 to 1 ps (Asada, 1989).
1 OPTICAL OF SEMICOND~JCTOR
NANO-STRUCTURES
A study of single dot optical properties has been one of the most exciting
subjects in the field of semiconductor quantum-dot research since the advent
of nanocrystals embedded in glass dielectric matrices (Ekimov et al., 1985)
and self-assembled quantum dots on semiconductor substrates. A key to
observing single dot light emission is to reduce the spatial resolution to
about 1 pm or less using, for example, a microscope objective and also to
reduce the number of dots in the observation area to less than, say, a
hundred. Various unique nature of semiconductor quantum dots has been
revealed; a very narrow natural emission linewidth less than 0.1 meV at low
temperatures (Zrenner et al., 1994; Brunner et al., 1994; Grundmann et al.,
1995; Gammon et al., 1996; Empedocles et al., 1996; Marzin et al., 1997), the
broadening and subsequent extinction of the emission line as temperature
increases due to homogeneous broadening (Ohta et al., 1997), an intermit-
tent photoluminescence like random telegram signal explained as a result of
carrier transfer between a single dot and a deep trap state (Nirmal et al.,
1996; Efros et al., 1997). Do these single-dot properties have any correlation
with the performance of quantum-dot lasers? If the area density of self-
assembled dots is 5 x 10" cm in the active layer of a cavity laser with a
~
cavity length of 300pm and a stripe width of 3pm, the number of dots
amounts to 450,000. This dot ensemble have the fluctuation of size and/or
semiconductor compositions, and thus, the fluctuation in the resonance
energy, which is called inhomogeneous broadening. Because of this large
number of hundreds of thousands in the cavity and the energy fluctuation,
people would think that single dot optical properties are masked in the
operation of conventional cavity lasers. It is pointed out here and in the
latest news chapter of this volume that lasing emission spectra are greatly
influenced by the magnitude of homogeneous broadening of optical gain of
single dots, and that the laser performance is never a mere ensemble of
optical properties of single dots.
Also unique in quantum dots is the Coulomb interaction among a limited
number of carriers confined in quantum dots. The quantum-dot discrete
state contains up to two electrons and two holes, and each quantum dot in
the ensemble has a different photo-emission energy depending on the
number of carriers it contains. The ground state of the filled two-electron
and two-hole state with opposite spin directions has lower energy than the
one-electron and one-hole states have, as shown both theoretically (Hu,
1990, 1996) and experimentally (Kamada et al., 1997), and can be regarded
as a bi-exciton state. Bi-exciton lasing was observed in CuCl quantum dots
(Masumoto et al., 1993).
This section describes how the carrier relaxation lifetime, homogeneous
broadening of the optical gain of single dots, and the bi-exciton effect
manifest themselves in quantum-dot laser operation. First, carrier-photon
86 MITSURUSUGAWARA
rate equations for quantum-dot lasers are derived to simulate the relation-
ship of output power versus injected current and the small-signal modula-
tion response. Criteria for the carrier relaxation lifetime, as well as
inhomogeneous broadening linewidth, dot density and crystal quality, are
clarified in how they achieve low-t hreshold, high-efficiency, high-power, and
high-speed operation. It is shown that the carrier relaxation lifetime forms
a hierarchy in quantum-dot laser performance. Then, the optical gain
formula taking into account both inhomogeneous broadening and homo-
geneous broadening is derived. How lasing emission spectrum varies de-
pending on the magnitude of homogeneous broadening is discussed. After
discussing a framework of bi-exciton lasing, we develop an optical gain
formula for the bi-exciton-exciton transition and modify the rate equations
to describe bi-exciton lasing in quantum dots. Also, we simulate the
spontaneous emission and lasing properties of bi-excitons and we discuss
the effect of carrier relaxation rate and oscillator strength enhancement on
laser operations.
1. THE EFFECT RELAXATION DYNAMICS

OF CARRIER ON LASER
PERFORMANCE
Figure 1.17 is an energy diagram of the laser active region, including the
self-assembled quantum dots of Fig. l.l(b), and the relaxation process of
carriers into the quantum-dot ground state. It is assumed that only a single,
discrete electron and hole ground state is formed inside a quantum dot, and
that a charge neutrality always holds in each dot. The energy axis represents
the energy difference between the conduction-band and valence-band edges.
Holes and electrons are treated together, which is enough to check how the
retardation of carrier relaxation affects laser performance. The injected
carriers diffuse through the separate confinement heterostructure (SCH)
layer, relax into the quantum well, and then relax into the dots. Some
carriers recombine radiatively or nonradiatively both outside (in the quan-
tum-well region) and inside the dots. Above the lasing threshold, carriers in
the ground state emit photons into the lasing mode primarily due to the
stimulated emission process. The associated time constants are diffusion in
the SCH region (TJ, carrier recombination in the SCH region (TJ, carrier
emission from the quantum well to the SCH region (T+J, carrier emission
from the quantum dot to the quantum well (re),carrier recombination in the
quantum well (T& carrier relaxation into the quantum dot (T~), and
recombination in the quantum dot (TJ Needless to say, the discussion
below also holds for SK dots accompanied by much thinner wetting layers.
1 OPTICAL
PROPERTIES OF SEMICONDUCTOR QUANTUM
NANO-STRUCTURES 87
FIG. 1.17. Energy diagram of the laser active region, including the self-assembled quantum
dots of Fig. l.l(b), and the relaxation process of carriers into the quantum-dot ground state
(from Sugawara et al., 1997~).
The rate equations for the carrier-photon system are
dN,ldt =Ile - NJ?, - N,/?,, + NqlTqe (1.252)
dN,/dt = N , / T , -k N I T , - N,/T,, - N,/T,, - N,/T, (1.253)
and
- SIT, (1.255)
where N , is the carrier number in the SCH layer, N , is that in the

quantum-well layer, N is that in the quantum dots, V , is the quantum-well
layer volume or active-layer volume, and other parameters are in common
with those in Section 111.3. The spontaneous emission factor is neglected
here. From Eq. (1.49), assuming that Gaussian inhomogeneous broadening
is dominant due to the size fluctuation of quantum dots, and substituting
the Gaussian function for the Lorentzian function of Eq. (1.43), the
88 MITSURUSUGAWARA
maximum optical gain at the center of the broadening function is given as
(1.256)
where To is the FWHM of the Gaussian function. In Eq. (1.256), the

coverage of dots, 5, is related to the dot density, N , , and dot volume, V,, as
< = N,VD (1.257)
Under the charge neutrality in each single quantum dot, we set f, = P and
f, = 1 - P. The carrier occupation probability, P , is determined by the
balance among the rates of carrier relaxation, carrier emission, and photon
emission. According to Pauli's exclusion principle, P is related to N as
P = N/(2NDI/,) (1.258)
Since carrier relaxation is prevented by the state filling according to the

exclusion principle, the relaxation rate is written as
's; = (1 - P)ZO1 (1.259)
where r o 1 is the relaxation rate when the ground state is unoccupied, that
is, P = 0. From Eqs. (1.56) through (1.58), the third-order optical gain in
quantum dots is written as
where
(1.261)
is the normalized broadening function. Substituting a Gaussian function for

Eq. (1.261), the maximum optical gain, including nonlinear susceptibility, is
written as
(1.262)
NANO-STRUCTURES
This form is often used in laser simulation to avoid negative gain. The
third-order nonlinear coefficient is
(1.263)
where rlT = ryl- + l-1- I. The relationship of
( 1.264)
is used.
Using a set of the above formulae and setting the time-dependent terms
of Eqs. (1.252) through (1.255) to be zero, we can calculate steady-state laser
operation as a function of the injected current. For simplicity, carrier
emission from the quantum dots is ignored; that is, Z/r >> N/T,. This
assumption will hold when the quantum-dot energy state is deep enough
compared to the thermal energy state. The parameters used are as follows.
The dots have a cylindrical shape with a radius of 7 nm and an equal height,
ho,,. = 1 eV, considering InGaAs quantum dots; three dot layers are grown;
the total optical confinement factor is r = 4%, R , = 30%, R , = 90%0,
M = 5 cm-', and L,, = 300ym; and the stripe width is 2pm. The photon
'
lifetime of the cavity is z P = 4.4 ps. The threshold gain is 670 cm from Eq.
(1.83). The spontaneous emission lifetime in quantum dots is calculated to
be z S p= 2.8 ns using Eq. (1.210) with the exciton effect neglected. The time
constants for nonlinear gain are re;' = 100 fs and rli'= z P .
Figure 1.18(a-c) shows the output power as a function of the injected
current using T~ as a parameter. The spectrum broadening is To = 20 meV,
and the coverage is 4 = 10%. which gives P = 0.84 at the lasing threshold
and T~ = 6.37, from Eq. (1.259). The recombination lifetimes are (a) T, =
T~~ = 2.8 ns and T,, = 3 ns; (b) T , = 2.8 ns and zqr = 0.5 ns; and (c) T , = 0.5 ns
and T ~ = , 0.5 ns. Case (a) corresponds to high quality both in the quantum
dots and the surrounding quantum well; case (b) corresponds to high-
quality quantum dots and a low-quality quantum well; and case (c) cor-
responds to low quality both in the quantum well and in the quantum dots.
The lowest threshold current is 110 pA in case (a) with t o = 1 ps. An
increase in injected current increases the carrier number in the quantum
well, N , , and carriers that relax into the quantum-dot ground state at a rate
of N,/rd are transformed into lasing-mode photons through the stimulated
emission process. The calculation is stopped when the quantum-well carrier
density exceeds 1 x 1012cm-2, because lasing from the quantum well
90 MITSURUSUGAWARA
F
.53
z8 2
CI
P,
0
300 Ps fg,= 3.0ns
0
(a) Injected current (mA) (b) Injected current (mA)
2 4 6
Injected current (mA)
FIG.1.18. (a)-(c) Output power as a function of the injected current using T~ as a

parameter. The spectrum broadening is r, = 20 meV, and the coverage is 5- = 10%. which gives
P = 0.84 at the lasing threshold and 7d = 6.35, from Eq. (1.259). The recombination lifetimes
are (a) T, = T~~~~ = 2.8 ns and z g r = 3 ns; (b) T~ = 2.8 ns and T~~ = 0.5 ns; and (c) T , = 0.5 ns and
tq,= 0.5 ns (from Sugawara et al.. 1998). ( d ) Maximum output power as a function of 70 for
various quantum-dot coverages at rt,= 20 meV (from Sugawara et al., 1998).
occurs around this point and this clumps the carrier number in the quantum
well, N , , leading to saturation of the output power from the quantum-dot
ground state. (We have found experimentally that lasing from the ground
state saturates at high output power and then gradually disappears as the
lasing from the upper levels starts and its output power increases. The
quenching of lasing might be due to the elevated temperature of quantum
dots under high current injection and the resultant increase in the emission
rate, T ~ . )What is remarkable is as follows:
1. As the relaxation lifetime, to.increases, the threshold current increases

1 PROPERTIES
OPTICAL OF SEMICONDUCTOR 91
NANO-STRUCTURES
QUANTUM
and the external quantum efficiency, represented by the slope of the

curve, decreases. This is because many injected carriers are consumed
in the quantum-well region and thus d o not contribute to lasing
oscillation.
2. The zo dependence becomes more remarkable when the recombination
lifetime in the quantum well, zqrr decreases, as shown in Fig. l.l8(b).
The zqr dependence, and thus the quantum-well crystal-quality depend-
ence, of the threshold current and the external quantum efficiency is
due to the retarded carrier relaxation, which increases the opportunity
for carriers to recombine through the nonradiative process outside the
dots.
3. Possible maximum output power decreases as the relaxation lifetime
increases. For example, when zo = 1OOps in Fig. 1.18(a), the quantum-
well carrier density reaches up to 1 x 1 0 ' 2 c m ~ 2at the power of
3.4mW, causing lasing emission from the quantum well. This is
because, as the cavity photon number increases, the extraction of
carriers from the quantum-dot ground state accelerates via the stimu-
lated emission process with a rate of S/z,, requiring the increase in N ,
to supply carriers to the ground state. Since z, is given by Eq. (1.259),
the relaxation lifetime can be reduced by lowering P at lasing. Figure
1.18(d) shows maximum output power as a function of T,, for various
levels of quantum-dot coverage at To = 20meV. The increase in
coverage leads to a lower P , resulting in high output power. If we need
a power of 40 mW, we must have t o= 22 ps for 4 = 40% and r0 = 8 ps
for 5 = 10%. Note that reducing the linewidth, To, has the same effect
on maximum output power as the coverage does, since it increases the
optical gain, resulting in lower P at lasing, shorter relaxation lifetime,
and higher maximum power.
4. As the recombination lifetime in quantum dots decreases, the threshold
current increases correspondingly, as seen in Fig. 1.18(c) due to the
consumption of carriers inside the quantum dots. This occurs even for
the rapid relaxation case.
For a small-signal current injection with a modulation angular frequency

of w,, that is, I = I , + I(o,) exp(iw,t), let us check the response of laser
output. The response function is defined as M(w,) = eS(wn,)/1(com),where
S = So + S(c0,) exp(iqnt). Though the function has a rather complex form,
it can be simplified by the following assumptions: (1) z,, << zS; ( 2 ) z, << rqr,
zqe; and (3) z, << z,. Each means that (1) carriers survive to reach the
quantum well; (2) all carriers in the quantum well relax into quantum dots;
and (3) emission of carriers from quantum dots is negligible. These assump-
tions correspond to high-quality quantum dots with a deep confinement
92 MITSURUSUCAWARA
potential. Also, these are enough for our purpose which is to clarify how the
modulation bandwidth is limited by the relaxation lifetime. The resultant
response function is given as
1 Q (1.265)
M(wm)= (1 + o;tj)l’2(1 + c1&rf)”2 - of)’
[(oz, + y om]l / 2
2 2
(1.266)
(1.267)
(1.268)
and
(1.269)
where 0,is the relaxation oscillation frequency, y is the damping factor, and
g‘ is the differential gain. The major factors that limit modulation bandwidth
are relaxation oscillation frequency and rolloff due to retarded carrier
relaxation. The parameter that dominates relaxation oscillation frequency is
the differential gain and thus the spectrum width. The 3-dB bandwidth
limited by the relaxation lifetime of sd is given as
(1.270)
Figure 1.19(a) shows the frequency dependence of the modulation re-

sponse function with parameters of zo = 10 ps, To = 20 meV, and 5 = 10%.
The response function shows the resonance at the relaxation oscillation
frequency, which increases with the output power. The rolloff of the response
due to the retarded relaxation is also seen.
In Fig. 1.19(b), the 3-dB bandwidth is plotted as a function of t o for
various levels of coverage with the broadening of To = 20 meV. The increase
in coverage leads to lower P. For 10-GHz modulation, we need t o = 7ps
for = 40% and t o = 3 ps for 5 = 10%. For 40-GHz modulation (not
shown in the figure), z, should be close to 1 ps. Requirements for the
relaxation lifetime are much more severe in high-speed modulation than in
static performance.
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 93
NANO-STRUCTURES
2, m
1 5 = 10%
1.4 mw\ ( P = 0.84)
h
v3 1.5-. ro=20meV -
13.8 mW
(a) Frequency, f = d2n: (GHz)
20
15
P
9
10
c
OD
'0 10 20 30 40 50
(b) Relaxation lifetime, ro (ps) (c) Output power (mw)
FIG. 1.l9. ( a ) Frequency dependency of the modulation response function with parameters
of T" = lops, To = 20meV, and 5 = 10%. (b) 3dB bandwidth as a function of T" for various
coverages. (c) Relaxation oscillation frequency ( 1;.= <ov/277)as a function of output power using
To as a parameter (from Sugawara et al., 1998).
Figure 1.19(c) shows the relaxation oscillation frequency (.f, = wr/2.rr) as

a function of output power using To as a parameter. The dot coverage is
5 = 10% for To = 5 , 10, and 20meV; 5 = 20% for To = 40meV; and
i; = 40% for To = 80 meV. The relaxation oscillation frequency is a good
measure of modulation bandwidth as long as the limit imposed by the
retardation of carrier relaxation is not taken into account. The linewidth
narrowing gives rise to the increase in J.; for example, f , reaches 40GHz
when To = 10 meV, and 57 GHz when To = 5 meV at the power of 20 mW.
At these narrow linewidths, the alpha parameter comes close to zero, giving
an almost negligible wavelength chirp. This gives birth to a directly
94 MITSURUSUGAWARA
modulated high-speed quantum-dot laser as a practical device. Note again

that in order to take full advantage of this potential; we must overcome the
rolloff of the response arising from retarded carrier relaxation into quantum
dots.
From what we have seen from the rate equation analyses, there is a
hierarchy in quantum-dot lasers that depends on relaxation lifetime (Fig.
1.20). First of all, for sub-mA low-threshold operation, quantum dots should
have a narrow spectrum linewidth of 20-30 meV and be of high quality with
the carrier lifetime close to the intrinsic spontaneous emission lifetime. The
relaxation should be accelerated as we go from sub-mA low-threshold to
highly efficient, high-power, high-speed operation. If the carrier relaxation
lifetime is reduced to around 10 ps, quantum-dot lasers will be able to show
high output power, high external quantum efficiency, and low threshold
currents. The slope efficiency of quantum-dot lasers will be superior to that
of quantum-well lasers because lower carrier density at lasing leads to low
internal loss. Actually, the measured efficiency of our columnar-shaped
quantum-dot lasers is excellent (see Chapter 6).
If 1.3-pm lasing is achieved in InGaAs/GaAs quantum-dot lasers (see
Chapter 3), the temperature characteristics of the threshold current will be
greatly improved for the following reason. Conventional strained quantum-
Hierarchy of quantum-dot lasers

A 7
Q
u Direct modulation lasers
2z
i! - High quality I
V
QD lasers
I I 11111111 I 11111111 I 11111111 I i l l l d
lo-’ 1 10 102 1o3

Relaxation lifetime (ps)
FIG. 1.20. Hierarchy in quantum-dot lasers, depending on the relaxation lifetime. For
sub-mA low-threshold operation, quantum dots should have a narrow spectrum linewidth of
20-30meV and be of high quality with the carrier lifetime close to the intrinsic spontaneous
emission lifetime. The relaxation should be accelerated as we go from sub-mA low-threshold
to highly efficient, high-power, and high-speed operation. If the carrier relaxation lifetime is
reduced to 1 ps and, at the same time. the spectrum linewidth is narrowed to less than IOmeV,
quantum-dot lasers will work as high-speed direct-modulation lasers with almost zero-chirp.
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 95
NANO-STRUCTURES
well lasers emitting in the infrared (1.3 and 1.55 pm grown on InP substrates
have suffered from the low temperature stability of threshold currents,
primarily due to carrier overflow from the active region to the SCH region
with relatively low potential height (Ishikawa, 1993; Ishikawa and Suemune,
1994). Carrier overflow into SCH regions with high density of states
increases the carrier number at lasing and thus accelerates the Auger
recombination process, the rate of which is proportional to the cube of the
carrier density, resulting in the consumption of carriers. At the same time,
large carrier density at lasing greatly increases the internal cavity loss,
resulting in low external quantum efficiency. In self-assembled InGaAs
quantum dots grown on GaAs substrates, carrier overflow is greatly
reduced, since high-potential barrier (SCH) and clad materials like Al-
GaInAs and AlGaInP can be grown on GaAs substrates to prevent carrier
overflow. As a result, quantum-dot lasers on GaAs substrates, especially
infrared lasers for optical transmission systems, will, in many respects, be
superior to strained quantum-well lasers on InP substrates.
If the carrier relaxation lifetime is reduced to 1 ps and, at the same time,
the spectrum linewidth is narrowed to less than 10 meV, quantum-dot lasers
can work as high-speed direct-modulation lasers with almost zero chirp.
Note that the GHz modulation of strained quantum-well lasers is done by
external quantum-well modulators using the quantum-confined Stark effect.
This is because, in strained quantum-well lasers. wavelength chirp is an
inevitable problem for high-speed modulation. A high-speed direct-modula-
tion laser, which was predicted by Arakawa and Sakaki in 1982, might be
one of the goals in quantum-dot laser research.
How long is the relaxation lifetime? To what extent is the relaxation
accelerated? As stated in the introduction to this section, much more work
should be done to find solutions to this problem. In addition to improved
theoretical and experimental studies on carrier relaxation itself, detailed
experimental analyses of quantum-dot laser performance will give a clear
picture of carrier relaxation dynamics. The thinking presented in this section
will help analyze laser operations. Finding a way to accelerate relaxation
through the design of a quantum-dot active layer might be required.
Experimental evaluation of the phonon bottleneck problem is presented in
Chapter 5.
2. EFFECTOF HOMOGENEOUSBROADENING
OF OPTICAL GAIN
SPECTRA
ON LASINGEMISSION
The formula for the optical gain spectrum is presented here, taking into
account both inhomogeneous and homogeneous broadening. Using the
96 MITSURUSUGAWARA
formula, we discuss the lasing emission spectra of quantum-dot lasers. (See

the latest news chapter for the experimental results.)
The optical gain formula of Eq. (1.49) for quantum dots has homogeneous
broadening of Lorentz shape with a width that is proportional to the
transverse relaxation rate due to scattering (Bcs in Eq. (1.43)). In the actual
quantum-dot lasers, we need to take into account inhomogeneous broaden-
ing due to dot size fluctuation and their spatial isolation as
where Ec,, is the average energy, and the FWHM is given as To = 2.35&.
The optical gain in the laser cavity is then given by the convolution integral
of the two broadening factors as
(1.272)
Equation (1.272) indicates that the optical gain at a given energy, E , is

created by collecting the individual contribution of all dots within the
homogeneous broadening of single dots around that energy.
If the homogeneous broadening is a delta function, that is, B,,(E) = 6(E),
Eq. ( 1.272) becomes
In this case, dots with different energies have no correlation to each other
and gives optical gain independently since they are spatially isolated from
each other. As a result, all dots with an optical gain above the lasing
threshold start lasing emission independently, which leads to broad band
lasing emission as depicted in Fig. 1.21(a). The quantum-dot laser at this
situation behaves in the same way as if it included independent lasing media
in the same cavity.
When homogeneous broadening of the optical gain is comparable to
inhomogeneous broadening, lasing-mode photons receive gain not only
from energetically resonant dots but also from other nonresonant dots that
lie within the amount of homogeneous broadening. Since carriers of non-
resonant dots are extracted and put into the central lasing modes by the
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 97
NANO-STRUCTURES
(b) B(E) E G(E)
FIG. 1.21. Relationship between lasing spectra and homogeneous broadening of single-dot
optical gain. (a) If the homogeneous broadening is a delta function, all dots with an optical
gain above the lasing threshold start lasing emission independently, which leads to broad-band
lasing spectrum. (b) When the homogeneous broadening of the optical gain is comparable to
the inhomogeneous broadening, carriers of non-resonant dots are extracted and put into the
central lasing modes by the stimulated emission. leading to lasing spectrum with one central
line.
stimulated emission, lasing emission with one line occurs as depicted in Fig.
1.21(b). Note that the calculation in Sec. V.1 implicitly assumed this
collective lasing.
See our recent experiments on the lasing emission spectra of our orig-
inally-designed columnar shaped quantum dots in the latest news chapter in
this volume, where experimental evidence for the above discussion is
presented and the origin of homogeneous broadening of optical gain is
discussed.
3. BI-EXCITONSPONTANEOUS EMISSIONAND LASING
Excitons, in the sense of electron-hole bound states, are not formed under
the population inversion bet ween the conduction and valence bands, and
cannot generate optical gain, as shown in Section IV.7. The Coulomb
interaction affects the optical gain such that the gain’s absolute value is
enhanced, a phenomenon introduced earlier Coulomb, or excitonic, enhan-
cement, and such that the gain spectrum is red-shifted due to screened
exchange energies and Coulomb-hole self-energies. For bound-state excitons
to generate gain, the exciton emission should be shifted toward lower
energies from the exciton absorption resonance through some mechanism.
This becomes possible by the assistance of third partners besides photons
and excitons, such as scattering particles, bi-excitons, or localized states.
The bi-exciton-exciton transition process is one of the possible candi-
dates for exciton-related lasing (Shaklee et al., 1971; Leheny and Nahory,
98 MITSURU
SUGAWARA
1971; Masumoto et al., 1993; Shionoya, 1995; Kreller et al., 1995; Sugawara,
1996, 1997b, 1998; Kozlov et al., 1996). A bi-exciton, analogous to a
hydrogen molecule, is formed through the coupling of a pair of excitons. The
ground-state bi-exciton is composed of two excitons with opposite spin
directions and has a binding energy of B,,, as seen in Figure 1.22(a). The
bi-exciton binding energies increase as the ratio of effective mass of m,*/mt
increases. For example, the binding energy is calculated to be about 0.07EB
at m,*/m: = 0.2 (Akimoto and Hanamura, 1972). The oscillator strength of
Bi-exicton dispersion
4M
2M
Wave number, K
(b)
Two excitons Formation

._I.
Exciton
-
Ground state
(c)
FIG.1.22. (a) Three-level model of a bi-exciton. an exciton and a ground state. B,, is the
bi-exciton binding energy. (b) Energy-wave vector dispersion relationships of excitons with a
wave vector of K, and bi-excitons with wave vector of K,. (c) Bi-exciton formation process.
Excitons excited by pumping, whether electronically or optically, collide with each other and
form bi-excitons. A bi-exciton annihilates through spontaneous emission and the stimulated
emission process, leaving one exciton behind.
NANO-STRUCTURES
the bi-exciton-exciton transition is proportional to the wave function extent

of the bi-exciton and is enhanced from that of the exciton transition.
The energy-wave-vector dispersion relationships of excitons with a wave
vector of K, and bi-excitons with a wave vector of K, are shown in Fig.
1.22(b), assuming that the dispersions are isotropic with the translational
mass of M,* for excitons and M:, = 2M,* for bi-excitons. A bi-exciton
annihilates by emitting a photon, leaving one exciton behind. The optical
transition occurs almost perpendicularly in the energy-wave-vector space
+
with the selection rule of K, = K, q 3 K, = K. This rule is similar to the
ordinary electron-hole interband optical transition. The emitted photon has
less energy than does the exciton optical absorption resonance by B,, +
h2K2/(4Mz).In the extreme, a combination of one bi-exciton and no exciton
generates optical gain for the electromagnetic field with the energy of hwbe.
This is in contrast to the optical gain due to ordinary electron-hole
transition, where many carriers are needed to achieve population inversion.
In this sense, a bi-exciton is analogous to a single quantum dot filled with
two electrons and two holes in the ground states.
Thus, a key to bi-exciton lasing is whether bi-excitons outnumber excitons
to generate gain equal to the threshold gain of a cavity, and whether this
occurs at lower carrier densities than the exciton and bi-exciton Mott
densities. For these conditions to be achieved, bi-excitons should be more
stable than excitons, and the bi-exciton formation should occur faster than
the bi-exciton and exciton annihilation through the spontaneous and
stimulated emission processes. The expected large oscillator strength of
bi-exciton-exciton transitions help to lower the threshold carrier density.
Figure 1.22(c) illustrates the bi-exciton formation process. Excitons excit-
ed by pumping, whether electronically or optically, collide with each other
and form bi-excitons. A bi-exciton annihilates through the spontaneous
emission and stimulated emission processes, leaving one exciton behind. The
corresponding rate equations for the bi-exciton number, N,,, and the
exciton number, N , , are
for bi-excitons, and
dN,ldt = G - N,/z, + 2N,,/z,, + 2(N,, - N:,)/r, + (c/n,)g,TS (1.275)
for excitons, where z, is the bi-exciton spontaneous emission lifetime. r , is

the exciton spontaneous emission lifetime, z, is the interconversion time
between bi-excitons and excitons, and G is the pumping rate. Nonradiative
recombination is not considered here. The photon rate equation in a cavity
100 MITSURt! SUGAWARA
is similar to Eq. (1.255), with the nonlinear gain effect omitted for simplicity.
The bi-exciton gain is given as
where gb is the differential gain. The bi-exciton number under the thermal
equilibrium, N:!, is given by the Saha equation (Kim et al., 1994) as
N2.y = Ni/N* (1.277)
with
(1.278)
for quantum wells.

The spontaneous emission can be modeled by setting S = 0 in Eqs. (1.274)
and (1.275). For the steady-state solution, the spontaneous emission inten-
sity is given as
Ixx =-2Nxx = GJ(1 + G/G,)”’ - 13’ (1.279)

‘xx
for bi-excitons and
(1.280)
for excitons, where
(1.281)
When G << Go, the bi-exciton emission intensity is proportional to G 2 , while

the exciton emission intensity is proportional to G. As G increases, the bi-
exciton intensity exceeds the exciton intensity at G = 8Go. The system
reaches transparency at G = GJ(1 + 4tx/zx,)’ - 11’. Figure 1.23 shows the
total exciton density-2Nx, + N,-needed to reach transparency ( N , = N x x )
as a function of temperature. The lifetimes are set as z,, = 45 ps and
z, = 60 ps. At low temperatures, where the thermal energy is about one
order of magnitude smaller than the bi-excion binding energy, bi-excitons
I OPTICAL OF SEMICONDUCTOR
NANO-STRUCTURES
c
.-00
CI
lO O
X
-Q 12
c1
lo8
10’
0 50 100 150 200 250 300
Temperature (K)
FIG. 1.23. Total exciton density. ( N , + 2N,,)/D, needed to reach transparency as a
function of temperature, using B,, and T~ as parameters. The horizontal dashed line represents
the exciton saturation density due to the phase-space-filling efiect. (Reprinted from Sugawara,
<Q 1998, with permission from Elsevier Science)
are stable and population inversion is realized at quite low densities. The
temperature range where the electron-hole system stabilizes as bi-excitons
is 10-20 K for B,, = 10 meV and below 30 to 50 K for 30 meV.
The above scenario of bi-exciton lasing seems unlikely to occur in
wide-gap semiconductors. One reason is that the bi-exciton binding energy
is around lOmeV, or 30meV at most, in 11-VI and InGaN materials,
requiring low temperatures for bi-exciton stability. The other is that spatial
energy fluctuations due to composition fluctuations lower the optical gain
at a low injection density, since only a small fraction of the active region
volume can contribute to the gain. This leads to electron-hole plasma lasing
with the local potential minima fully filled with a large number of carriers.
Similarly, quantum dots with large-size inhomogeneity cannot generate
enough optical gain for lasing, so lasing caused by electron-hole plasma
occurs from the quantum well or wetting layers. As long as spatial energy
fluctuations are not controlled to give homogeneous samples, bi-exciton
lasing is not observed.
102 MlTSURU SUGAWARA
If we fabricate well-controlled homogeneous local potential minima as

quantum dots, we have a chance. Each discrete level of quantum dots
incorporates up to two electrons and two holes with opposite spin directions
according to the Pauli exclusion principle. The filled state of two electrons
and two holes state in the quantum-dot ground state has a positive binding
energy close to the exciton binding energy in bulk materials (Hu, 1990; Hu
et al., 1996), and thus this two-pair state can be called the bi-exciton ground
state. The emission spectrum caused by the bi-exciton-exciton transition
appears red-shifted from the emission of a one-electron-hole state, that is,
the exciton state. In blue-regime semiconductor quantum dots, the bi-
exciton binding energy should reach several tens of meV, and we expect the
bi-exciton emission to be well resolved from the exciton emission in spite of
inhomogeneous broadening.
What is interesting about bi-excitons in quantum dots is that we can
study various aspects of bi-exciton lasing by artificially designing quantum-
dot structures. Let us examine a one-to-one correspondence between quan-
tum wells and quantum dots:
1. In quantum wells, the bi-exciton and exciton excited states and many
exciton states appear along with the ground-state excitons. By control-
ling the size and strains of quantum dots and the band-gap discontinu-
ity against surrounding materials, we can design the number of discrete
states in the dots. If there is only one in both the conduction and
valence bands, a pure bi-exciton state is formed when the state is fully
filled with two electrons and two holes. If quantum dots contain more
than two levels, the bi-exciton and exciton-excited states and many
exciton states appear, depending on the number of carriers in the dot.
2. For bi-exciton lasing, bi-excitons should be more stable than excitons,
and the bi-exciton formation should occur faster than the bi-exciton
and exciton annihilation. The bi-exciton stability in quantum dots can
be improved by increasing the quantum-dot potential barrier height.
The counterpart of the bi-exciton formation rate in quantum dots is
the relaxation rate discussed in Section V.1. The rate is controlled to
some extent by designing the quantum-dot energy levels and by
devising injection methods.
3. The Mott density of bi-excitons is a matter to be investigated. We can
experimentally study how the carrier number around quantum dots (in
the quantum well in the model of Section V.1) affects the dots’ binding
energy and oscillator strength, by changing the dot density, the laser
output power, etc.
To take into account the bi-exciton effect in quantum dots, let us use the
1 OPTICAL
PROPERTIES QUANTUM
OF SEMICONDUCTOR 103
NANO-STRUCTURES
three-level model, where the bi-exciton and exciton states are taken as base
functions. The bi-exciton and exciton system is described by the Hamil-
tonian of (Hu, 1990; Hu et al., 1996)
where hwy is the ground-state energy, h o e is the exciton energy, hw, is the
bi-exciton energy, lg) is the ground state, le) is the exciton state, and Ib) is
the bi-exciton state. The transition between the eigenstates of the system
occurs as a result of the interaction of the electromagnetic field with the
wave vector, q, and the polarization, 0. The interaction with a field is
described as
where p,,(t) is given by Eq. (1.176) and p h e ( t ) is given similarly. The

procedure to derive the optical gain is the same as in Section 111; that is, the
density-matrix equation is solved with the initial condition of pg’+ C, pi:’+
C,&) = 1. The polarization of the system is given from the diagonal
summation of the density matrix-polarization product. As a result, the
linear optical gain is given as
where weY= me - cog, whe= w b - w, = weg- B J h . The broadening func-

tions, Be, and Bhe, are the Lorentz functions with a full width at half
maximum of dephasing rate 2Feg and 2rb,, respectively. The first term of
Eq. (1.284) is the result of the exciton-ground-state transition, where the
oscillator strength is given by Eq. (1.172) as
(1.285)
Here, $c(r, R) is the exciton envelope wave function. The second term of Eq.
(1.284) represents the bi-exciton-exciton transition, where the oscillator
104 MITSURUSUGAWARA
strength is given as
where $be(rl, R,, rz, R,) is the bi-exciton envelope wave function and the
subscripts of 1 and 2 represent the first and second excitons. Qualitatively,
Eq. (1.286) shows that the oscillator strength of the bi-exciton-exciton
transition is proportional to the product of the absolute square of the
first-exciton overlap integral, the probability of finding an electron and a
hole at the same position in the second exciton, and the volume covered by
the second exciton. The oscillator strengths of Eqs. (1.285)and (1.286)can
be measured as a ratio against the oscillator strength for the electron-hole
transition of
(1.287)
where a,, is the interband optical transition frequency at the band edge. The
oscillator strength, f: and ,j":, was calculated by Takagahara (1989) as a
function of the radius of the spherical quantum dot. It increases as the dot
radius increases, a phenomenon known as mesoscopic enhancement. Accord-
ing to Takagahara's calculation, j;,/fo2 4, when the radius is equal to the
bulk exciton Bohr radius.
To calculate the population balance between ground, exciton, and bi-
exciton states, let us use the occupation probability, P, defined by Eq.
(1.258). The probability of finding a bi-exciton in quantum dots is P 2 , the
probability for a spin-up exciton is P(l - P), the probability for a spin-
down exciton is P(l - P), and the probability of not finding any excitons is
(1 - P ) 2 . Thus, we obtain
pi:) - p::,' = P2 - P(l - P)= 2p2 - p (1.288)
and
pL0,' - p g ) = P(l - P ) - (1 - P), = -1 + 3P - 2Pz (1.289)
The population inversion appears when P > lj2 for both transitions.
Bi-exciton lasing from quantum dots can be described using rate equa-
tions only slightly different from those in Section V.l in the following points.
Considering quantum dots that include only the ground-state exciton and
PROPERTIES
QUANTUM 105
NANO-STRUCTURES
bi-exciton levels, Eq. (1.284) gives the maximum optical gain at the bi-
exciton-exciton transition as
(1.290)
Again, the Gaussian broadening function arising from the size inhomogene-
ity of quantum dots is assumed to be dominant. The spontaneous emission
rate is given as
where
N
-
rr
=2N,Vu
LXX
-P+2
z, 1
2P(1 - P)
(1.291)
( 1.292)
for bi-excitons ( i = xx) and excitons ( i = x), and the emitted photon fre-
quency is (0. The relaxation rate of Eq. (1.259) is unchanged, since
~d-' =(1 - +
P ) 2 ~ i 12P(1 - P)(tG1/2) = (1 - P ) T ~ ' (1.293)
Setting S = 0 for the spontaneous emission process and omitting the

time-dependent terms in Eqs. (1.252) through (1.254), we obtain the current
dependence of P and the spontaneous emission intensity as
2NDP(1 - P )
I, = (1.294)
5,
for excitons and
(1.295)
for bi-excitons.
The spontaneous emission and steady-state laser output power as a
function of current can be calculated by the same procedure as in Section
V.l. The parameters considering wide-gap InGaN quantum dots are as
follows. The dots have a cylindrical shape with a radius of 3.6 nm (equal to
the exciton Bohr radius in bulk materials) and an equal height; 5 = 5%,
hw,,. = 3.0 eV, To = 20 meV, and L,, = 300 pm; three dot layers are grown;
the total optical confinement factor is = 4%; R , = 30%, R , = 90%, and
a = 5 cm-'; and the stripe width is 2pm. Since the bi-exciton binding
106 MITSURUSUGAWARA
energy is estimated to be 30meV in a quantum dot of this size, the

bi-exciton emission is well separated frrom the exciton emission.
Figure 1.24(a) shows the spontaneous emission intensity, I , and I,,, as a
function of the injected currents. The oscillator strength is set at f,,/fo = 4,
and the spontaneous emission lifetime is calculated to be T,, = 0.34 ns by
Eq. (1.292). The relaxation rate is set at lops. (We have no idea whether
this is an realistic value. To realize this fast relaxation for quantum dots, the
tuning of energy levels to LO phonon energy might be needed.) The
bi-exciton emission intensity increases in proportion to the square of the
current, while the exciton emission increases in proportion to the current.
The bi-exciton emission exceeds the exciton emission intensity at the current
of 1.1mA. As seen in the inset of the figure, the lasing threshold is 1.4mA.
Figure 1.24(b) shows the maximum output power and the occupation, P , at
the lasing threshold as a function of oscillator strength. Again, the “maxi-
mum” is defined as the output power when the carrier density around the
dots reaches 1 x 1012cm-2. As the oscillator strength increases, P de-
creases, increasing the maximum output power for the same physical reason
as in Section V . l . This increase is the most beneficial point of the bi-exciton
effect, or Coulomb enhancement, of the optical gain in quantum dots.
Note that the oscillator strength formulae of Eqs. (1.285) and (1.286)
correspond to the case in which there are no carrier number in the quantum
well surrounding the quantum dots, N , , increases as the laser output power
increases. This decreases the bi-exciton binding energy and the oscillator
strength enhancement. At the same time, the homogeneous broadening of
B
P
I
2 3 4
(a) Current (mA) Oscillator strength, fxx/fo
Fiti. 1.24. (a) Spontaneous emission intensity due to excitons, I , . and bi-excitons, I,,, in
the quantum dot laser as a function of injected currents. The bi-exciton emission intensity
exceeds the exciton emission at a current of I.lmA. The lasing threshold is 1.4mA (from
Sugawara et al., 1998). (b) Maximum output power and the occupation, P , at the lasing
threshold as a function of oscillator strength (from Sugawara et al., 1998).
1 OPTICAL
PROPERTIES
OF SEM~CONDUCTOR
QUANTUM 107
NANO-STRUCTURES
the optical gain will increase because of scattering, resulting in smaller

optical gain, When the carrier density reaches the level at which population
inversion occurs in the quantum-well regime, the bi-exciton binding energy
and the oscillator strength enhancement will have almost disappeared, and
lasing will occur from the quantum-well regime. From this, we see that
bi-exciton lasing properties, such as emission energy, optical gain, and
output power, are dominated by the carrier relaxation and emission rate,
and the energy level structures of quantum dots.
VI. Summary
This chapter provided a theoretical background of the optical properties

of semiconductor quantum nano-structures, emphasizing how those proper-
ties vary from quantum wells to quantum dots and how they are influenced
by the exciton effect. We saw that optical gain, optical absorption, and
spontaneous emission depend strongly on the dimension of quantum
nano-structures and on the exciton effect. The importance of accelerating
carrier relaxation into quantum dots in the realization of high-performance
quantum-dot lasers was stressed. A framework of bi-exciton lasing was
discussed.
The following physical properties of quantum dots remain to be clarified:
(1) the relaxation rate into quantum dots by electron-electron and elec-
tron-phonon scattering and by tunneling injection; (2) the laser structure
design to realize fast relaxation, ( 3 ) the dephasing time of the carrier-photon
interaction via carrier-carrier and carrier-phonon scattering to determine
the linear and nonlinear susceptibility spectra, that is, homogeneous
broadening, (4) the many-body effect of a limited number of electrons and
holes in quantum dots to clarify the Coulomb effect on the optical gain
spectrum; and (5) the band engineering of quantum dots and surrounding
materials for fast carrier relaxation and strong carrier-photon interaction.
These issues, which are now being studied by many researchers, will help us
to create new high-performance quantum-dot optical devices.
Appendix
To determine optical transition strength, the matrix element of

108 Mr-rsuuu SUGAWARA
in Eq. (1.20) should be calculated. For the interband transition between the
conduction and valence bands, using Eq. (1.21), we get
The matrix element of (c, k,le'q''e, - p1u, k,) is evaluated as follows.

Bulk Muteriuls
From Eqs. (1.2) and (1.4), the electron wave function in the conduction
band is
-
ik, r)u,(r)
and that in the valence band is
where u,(r) is for the conduction band and u,(r) is for valence bands. Thus,
the matrix element is given as
(c, k,le'q"e,~ pJu,k,)
+ q)
=
J
exp[i(k, - k,, * r]d3r
Jn u,*(r)e, pu,(r)d3r
+ $1, exp[i(k, + q) rle, * p exp[ik, r]d3r - u,*(r)u,,(r)d3r (A.4)
The integration is separated into the product of the integration over the unit
cell and over the entire crystal because the envelope function varies slowly
over the unit cell. Since the Bloch states in the conduction band and the
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 109
NANO-STRUCTURES
valence band are orthogonal,
where the wave vector selection rule is
Thus, Eq. (1.31) for the interband transition is written as
PPr,k= (c, kle, * plu, k ) = u,*(r)e, * pu,,(r)d3r (A.7)
Quantum Wells
From Eqs. (1.2) and (1.7), the electron wave function in the conduction
band is
and the valence-band state is
The matrix element is
where
I
(c, k,le'q''e,- plu, k,) = d k ~ l , ~ ~ l ~ q l I ~ n , , . f l ,u:(r)e;pu,,(r)d3r
, (A.lO)
(A.ll)
is the overlap integral between the quantized envelope wave function of the
conduction and valence bands. Here, since the well-layer thickness is much
smaller than the wavelength of the electromagnetic field, the electric dipole
approximation of e'Yz"= 1 is used. For the infinite potential well, the
110 Mr~suuuSUCAWARA
selection rule for the quantum number in the z direction is
In actual quantum wells with a finite potential height, Z,lcz,na,2 is less than 1
even when ncz = nrz. The wave vector selection rule is
kII = kII - qII - kII
3
(A.13)
Thus, Eq. (1.31) for the interband transition is written as

P
Quantum Dots
Since quantum dots are much smaller than the photon wavelength, the
electric dipole approximation of e'q" = 1 holds. Thus,
(A.15)
The next problem is to calculate jnu;(r)e; pu,(r)d3r. According to Kane's

k * p method of calculating the band dispersion around the band edges, the
periodic part of the band-edge Bloch functions is taken as (Kane, 1956,
1957)
for uc, where r

and J are spin functions and S has atomic s-orbital
symmetry. The functions for the valence band, u , , are
for the heavy-hole valence band,
(A.18)
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 111
NANO-STRUCTUKES
for the light-hole valence band, and
(A.19)
for the split-off band resulting from the spin-orbit interaction, where X , X
and Z have p-orbital symmetry. The quantization axis of angular momen-
tum is taken in the z direction. The P-state bases of Eqs. (A.17) through
(A.19), written in ( J , M , ) representation, are taken to diagonalize the
spin-orbit interaction. The electron wave vector is assumed in the z
direction, and the base functions rotate with k vectors. We should first
calculate the matrix element with the z-axis at the k direction and then
transform the element into the xyz coordinate system. Finally, we average
the matrix elements over degenerate k vectors depending on the symmetry
of the crystal. If the system has structural asymmetry, as do quantum wells
and quantum wires, the matrix element has polarization and k-vector
dependence, according to the valence-band type related to the optical
transition.
Bulk Materials
The momentum matrix element is obtained by averaging Eq. (A.7) with

respect to the solid angle. Then.
(P$k12= I(Slpli)12/6 s M 2 (i =X , X and 2 ) (A.20)
The k *p perturbation gives the transition matrix in bulk materials as
(A.21)
where mo is the free electron mass, 15, is the band gap, A is the spin-orbit
splitting energy, rn: is the conduction-band-edge effective mass, and D'
is the second-order perturbation term (Sugawara, 1993b).
Quantum Wells
The wave vector is (k",n,n/L,). Let 6' be the angle between the z axis and
the wave vector. The character gives the polarization dependence to the
112 MITSURUSUGAWARA
matrix element as (Yamanishi and Suemune, 1984)
(A.22)
where
M&4ell, k) = 3(1 + cos2H)/4 (A.23)
and
M&(eL,k) = 3 sin2H/2 (A.24)
for conduction to heavy-hole band transition,
M&(ell, k) = (1 + cos28)/4 + sin2@ (A.25)
and
M&(el, k) = sin28/2 + 2 cos2H (A.26)
for conduction to light-hole band transition, and
(A.27)
for conduction to split-off-band transition.
Quuntum Dots
The averaging procedure of matrix elements is done over degenerate k

states, depending on the symmetry. If a quantum dot has a spherical shape,
the averaging gives
IP&12 = M2 (A.28)
REFERENCES
Akimoto, 0.- and Hanamura, E. (1972). J. Phys. Sot.. Jpr7. 33, 1537.
Andreani, L. C., Tassone, F., and Bassani, F. (1991). Solid Sfcite Commun. 77, 641.
Arakawa, Y., and Sakaki, H. (1982). Appl. Phys. Left. 40, 939.
Arita, M., Avramescu, A,, Uesugi, K., Suemune, I.. Numai, T.. Machida, H., and Shimoyama,
N. (1997). J p n . J . Appl. P h ~ x36. 4097.
Asada, M. (1989). IEEE J . Quantum Electron. 25, 2019.
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 113
NANO-STRUCTURES
Bastard, G., Brum, J. A,, and Ferreira. R. (1991). “Electronic States in Semiconductor
Heterostructures,” in Solid State Physics 44, Academic Press, New York.
Benisty, H., Sotomayor-Torres, C. M.. and Weisbuch, C. (1991). Phvs. Reu. B. 44, 10945.
Bockelmann, U., and Bastard, G. (1990). Phys. Rev. B. 42, 8947.
Bockelmann, U., and Egeler, T. (1992). Phys. Reu. B. 46, 15574.
Bockelmann, U., Heller, W., Filoramo, A., and Roussignol, P. (1997). Phys. Rev. B. 55, 4456.
Brunner, K., Abstreiter, G., Bohm, G . , Trinkle, G., and Weimann, G. (1994). Phys. Rev. Lett.
73, 1138.
Cao, H., Jacobson, J., Bjork, G., Pau, S., and Yamamoto, Y. (1995). Appl. Phys. Lett. 66. 1107.
Carlsson, N., Georgsson, K., Montelius. L.. Samuelson, L., Seifert, W., and Wallenberg, R.
(1995). J . Cryst. Growth 156, 23.
Chang, L. L., Esaki, L., and Tsu, R. (1974). A p p l . Phys. Lett. 24, 593.
Chemla, D. S., Damen, T. C., Miller, D. A. B., Gossard, A. C., and Wiegmann, W. (1983). Appl.
Phys. Lett. 42, 864.
Chemla, D. S., Miller, D. A. B., Smith. P. W., Gossard, A. C., and Wiegmann, W. (1984). I E E E
J . Quantum Electron. QE-20, 265.
Chow, W. W., Wright, A. F., and Nelson, J. S. (1996). Appl. Pliys. Lett. 68, 296.
Chuang, S. L. (1991). P ~ J J Reo.
S . B. 43, 9649.
Chuang, S. L. (1995). Physics ufOptoelec,fronirsDeoices. John Wiley & Sons, New York.
Citrin, D . S. (1993). Phys. R ~ I JB.. 47, 3832.
Ding, J., Jeon, H., Ishihara, T., Hagerott, M., Nurmiko, A. V., Luo, H., Samarth, N., and
Furdyna, J. (1992). Phj’s. Reu. Left. 1707.
Ding, J., Hagerott, M.. Ishihara, T., Jeon, H., and Nurmiko, A. V. (1993). Phjx Rev. B. 47.
10528.
Dingle, R. (1975). Festkorperprobleme 15, 21.
Efros, A. L., and Rosen, M. (1997). Phys. Rer. Lett. 78, 1110.
Efros. A. L., Kharchenko, V. A,, and Rosen. M. (1995). Solid State Cornmun. 93, 281.
Ekimov, A. L., Efros, A. L., and Onushchenko, A. A. (1985). Solid State Commun. 56. 921.
Elliot, R. J. (1957). Phys. Rev. B. 108, 1384.
Empedocles, S. A., Norris, D. J., and Bawendi. M. G. (1996). Phys. Rec. Lett. 77. 3873.
Esaki, L., and Tsu, R. (1969). IBM Resetrrch Note. RC-2418.
Esaki, L., and TSU,R. (1970). I B M J . Res. Deiwlop. 14, 61.
Feldmann, J.. Peter, C.. Gobel, E. 0..Dawson. P., Moore, K., Foxon, C., and Elliott, R. (1987).
J . P h w . Reo. Lett. 20, 2337.
Galbraith, I. ( 1995). “Excitonic gain processes in ZnSe,” in Microscopic Theorj of Semiconduc-
tors: Qumtum Kinetics. Confinemen/ rind Ltrsers, World Scientific Publishing Co. Pte. Ltd.,
Singapore.
Gammon, D., Snow, E. S., Shanabrook, B. V., Katzer, D. S.. and Park, D. (1996). Phys. Rev.
Lett. 76, 3005.
Gerard, J.-M. (1995). “Prospect of high efficiency quantum boxes obtained by direct epitaxial
growth,” in Confined Electrons mid Photons: New Physics and Applicutions. Plenum. New
York.
Goldstein, E. V.. and Meystre, P. (1995). ”Spontaneous Emission and Laser Oscillation in
Microcavities.” CRC Press, Boca Raton. FL.
Grosse, S.. Sandmann. J. H. H. (1997). Phys. Ret:. B. 55, 4473.
Grundmann, M., Christen, J., Ledentsov. N. N., Bohrer, J.. Bimberg. D.. Ruvimov. S. S.,
Werner, P.. Richter, U.. Gosele. U., Heydenreich, J., Ustinov, V. M.. Egorov, A. Yu,
Zhukov, A. E., Kop’ev, P. S.. and Alferov, Zh. I. (1995). PIiys. Reu. Lett. 74, 4043.
Haase, M. A. (1991). Appl. Phys. Lett. 59. 1272.
Haken, H. (1973). Quuntenfeldtheorie drs Fesfkurpers. B.G. Teubner, Stuttgart.
114 MITSURUSUCAWARA
Hanamura, E. (1988). Phys. Rec. B. 38, 1228.

Haug, H., and Koch, S. W. (1977). Phys. Stilt. S O / . 82, 531.
Haug, H., and Koch, S. (1994). Quuntum Theory of the Optical und Electronic Properties qf
Semiconductors, 3rd ed. World Scientific, Singapore.
Hayashi, I., Panish, M. B., Foy, P. W., and Sumski, S. (1970). Appl. Phys. Lett. 17, 109.
Houdre, R., Stanley, R. P., Oesterle. U., Ilegems, M., and Weisbuch, C. (1994). Phys. Rev. B.
49, 16761.
Hu, Y. Z., Lindberg, M., and Koch. S. W. (1990). Phys. Rev. B. 42, 1713.
Hu, Y. Z., Gieben. H., Peyghambarian, N., and Koch, S . W. (1996). Phys. Rev. B. 53, 4814.
Inoshita, T., and Sakaki. H. (1992). Phys. Rev. 5. 46, 7260.
Ishibashi, T., Tarucha, S., and Okamoto. H. (1981). Proc. lnr. Symp. GaAs und Related
Compounds, Oslo.
Ishikawa, H. (1993). Appl. Pkys. Leu. 63. 712.
Ishikawa, H., and Suemune, I. (1994). Photon Tech. Lett. 6, 344.
Itoh, T. (1992). Nonlineur Optics 1, 61.
Jacobson. J. et al. (1995). Phys. Reu. A. 51. 2542.
Jen, Y., Tsutsumi. T., Souma. I., Oka. Y., and Fujiya. H. (1993). Jpn. J . Appl. Phys. 32, L1542.
Kadoya, Y., Kameda, K., Yamanishi. M., Nishikawa, T., Kannari, T., Ishihara, T., and Ogura,
1. (1995). Appl. Phys. Lett. 68, 281.
Kamada, H., Ando, H.. Temmyo, J., and Tamamuira, T. (1997). Exr. Abs. 58th Autumn Meet.
Jupun Society of Applied Physics 0, Akita.
Kane, E. 0. (1956). J . P h j ~Chern. Solids 1. 82.
Kane, E. 0. (1957). J . Phys. Chem Solids 1. 249.
Kawakami, Y., Hauksson, I., Simpson, J., Stewart, H., Galbraith, I., Prior, K. A,, and Cavenett,
B. C. (1994). J . Cryst. Growth 138. 759.
Kim, J. C., Wake, D. R., and Wolfe, J. P. (1994). Phys. Rev. B. 50, 15099.
Klingshirn, C., and Haug, H. (1981). Phys. Reu. 5. 70, 315.
Knox, R. S. (1963). Theory ofExcitons. Academic Press, New York.
Koch, S. W., Haug, H.. Schmieder, G.. Bohnert, W., and Klingshirn, C. (1978). Phps. Srut. Sol.
89, 431.
Kozlov, V. et al. (1996). Phys. Rev. 5. 53, 10837.
Kuroda, Y., Suemune, I., Fuji, Y., and Fujirnoto, M. (1992). Appl. Phys. Lett. 61, 1182.
Kreller, F.. Lowisch, M., Puls, J., and Henneberger, F. (1995). Phys. Rev. Lett. 75, 2420.
Leheny, R. F., and Nahory, R. E. (1971). Phjs. Rec. Lett. 26, 888.
Luttinger, J. M., and Kohn, W. (1955). Phys. Rev. B. 97, 869.
Kuwatsuka, H., and Ishikawa. H. (1994). P h w . Rev. B. 50. 5323.
Marzin, J.-Y., Gerard, J.-M., Izrael, A.. Barrier. D., and Bastard, G. (1997). Plij~.Rec. Lett. 73,
716.
Masumoto, Y., Kawamura, T., and Era, K. (1993). Appl. Phys. Lett. 62, 225.
Meystre, P., and Sargent, M. (1990). Elements of’ Quuntuni Optics. Springer-Verlag, Berlin.
Miller, D. A. B., Chemla, D. S., Smith, P. W.. Gossard, A. C., and Tsang, W. T. (1982a). Appl.
Phys. Rev. B. 28, 96.
Miller, D. A. B., Chemla, D. S., Eilenberger. D. J.. Smith. P. W., Gossard, A. C., and Tsang,
W.T. (1982b). Appl. Phys. Leu. 41, 679.
Miller, D. A. B., Chemla, D. S., Eilenberger, D. J., Smith, P. W., Gossard, A. C., and Wiegmann,
W. (1983). Appl. Phys. Lett. 42, 925.
Miller. D. A. B., Chemla, D. S., Damen. T. C.. Gossard, A. C., Wiegmann, W., Wood, T. H.,
and Burrus. C. A. (1984a). Phys. Rec. Lett. 53, 2173.
Miller, D. A. B.. Chemla, D. S., Damen, 7.C.. Gossard, A. C., Wiegmann, W., Wood, T. H.,
and Burrus, C. A. (l984b). Appl. Phps. Lerr. 45, 163.
1 OPTICAL
PROPERTIES
OF SEMICONDUCTOR
QUANTUM 115
NANO-STRUCTURES
Miller, D. A. B., and Burrus, C. A. (1985a). Phys. Rev. B. 32, 1043.

Miller, D. A. B., Chemla, D. S., Damen, T. C., Wood, T. H., Burrus, C. A., Gossard, A. C., and
Wiegmann, W. (1985b). IEEE J . Quuntutn Electron QE-21, 1462.
Mukai, K., Ohtsuka, N., Shoji, H., and Sugawara, M. (1996a). Appi. Pkys. Lerr. 68, 3013.
Mukai, K., Ohtsuka, N., Shoji, H., and Sugawara, M. (1996b). Phys. Re[).B. 54, R5243.
Nakamura, S., Senoh, M., Nagahama, S., Iwasa, N., Yamada, T., Matsushita. T., Kiyoku, H.,
and Sugimoto, Y. (1996). Jpn. J . Appl. Phys. 35, L74.
Nakayama, H., and Arakawa, Y. ( 1 994). 7th Int. Conf Superluttices, Microstructures, and
Microdevices, Banff, Canada.
Nirmal, M., Dabbousi, B. O., Bawendi, M. G., Macklin, J. J., Trautman, J. K., Harris, T. D.,
and Brus, L. E. (1996). Nature 383, 802.
Norris, T. B., Rhee, J. K., Sung, C. Y., Arakawa, Y., Nishioka, M., and Weisbuch, C. (1994).
Phys. Rev. B. 50, 14663.
Ogawa, T. (1995). “Dimensionality and Optical Responses of Materials,” in Opricul Properties
of Low-Dimensional Muteriuls. World Scientific, Singapore.
Ohnesorge, B., Albrecht, M., Oshinowo, J., Forchel, A,, and Arakawa. Y. (1996). Phys. Rei). B.
54, 11532.
Ohta et al. (1989). Physicu, E2, 573.
Raymond, S., Fafard, S., Charbonneau, S., Leon, R., Leonard, D., Petroff, P. M., and Merz, J. L.
(1995). Phys. Rev. B. 52, 17238.
Raymond, S., Fafard, S., Poole, P. J., Wojs, A,, Hawrylak, P., Charbonneau, S., Leonard, D.,
Leon, R., Petroff, P. M., and Merz. J. L. (1996). Phys. Rev. B. 54, 11548.
Reaves, C. M., Bressler-Hill, V., Varma. S., Weinberg, W. H., and DenBaars, S. P. (1995). Surf:
Sci. 326, 209.
Schmitt-Rink, S., Chemla, D. S., and Miller, D. A. B. (1989). Adv. Phys. 38, 89.
Shaklee, K. L., Leheny, R. F., and Nahory, R. E. (1971). Phys. Rev. Lett. 26, 888.
Shinada, M., and Sugano, S. (1966). J . Phys. Soc. J p n . 21, 1936.
Shionoya, S. (1995). Kotai-Butsuri 5, 438 [in Japanese].
Sugawara, M., Fuji, T., Yamazaki, S., and Nakajima, K. (1990). Phys. Rev. B. 48. 8102.
Sugawara, M. (1992). J . Appl. Phys. 71, 277.
Sugawara, M., Okazaki, N., Fuji, T., and Yamazaki, S. (1993a). PAys. Rev. B. 48, 8102.
Sugawara, M., Okazaki, N., Fujii, T.. and Yamazaki, S. (1993b). Pkys. Reu. B. 48, 8848.
Sugawara, M., and Yamazaki, S. (1994). Microiouve und Optic. Tech. Lett., 7, 107 (special issue).
Sugawara, M. (1995). Phys. Rev. B. 51, 10743.
Sugawara, M. (1996). Jpn. J . Appl. Phy.s. 35, 124.
Sugawara, M. ( 1997a). Proc. 2nd Int. Cnnf. Nitride Semiconductors, Tokushima.
Sugawara, M. (1997b).Jpn. J . Appl. P h w . 36, 2151.
Sugawara, M., Mukai, K., and Shoji, H. (1997~).Appl. P h j ~ sLett.
. 71, 2791.
Sugawara, M. (1998). Proc. Optoelectronics Y8,Photonic West, San Jose, CA.
Takagahara, T. (1989). Phys. Rev. B. 39, 10206.
Tanaka, S., Iwai, S., and Aoyagi, Y. (1996). Appl. Phys. Lett. 69. 4096.
Tanaka, S., Hirdyama, H., Ramvall, P.. and Aoyagi, Y. (1997). Proc. 2nd Int. Con/. Nitride
Senziconrluctors, Tokushima.
Thijs, P. J. A., and van Dongen, T. (1990). E.ut. Ahs. 22nd Int. Cur$ Solid Siute Devices mid
Muteriuls, Sendai, Japan.
Tsang, W. T. (1981). Appi. Phys. Lett. 39. 786.
Uenoyama, T. (1995). Phys. Rev. B. 51, 10228.
Weisbuch, C. (1987). Semiconductors trnd Srmirnetals, 24, Academic Press, New York.
Weisbuch, C., Nishioka, M., Ishikawa, A., and Arakawa, Y. (1992). Phys. Rer. Lett. 69, 3314.
Weisskopf, V., and Wigner, E. P. (1930). 2. Phys. 63, 54.
116 MITSURUSUGAWARA
Wood, T. H., Burrus, C. A,, Miller, D. A. B.. Chemla, D. S., Damen, T. C., Gossard, A. C., and
Wiegmann, W. (1984). Appl. Phys. Lett. 44, 16.
Yamanishi, M., and Suemune, I. (1984). J p n . J . Appl. Phys. 23, L35.
Yokoyama, H., and Ujihara, K. (1995). "Spontaneous Emission and Laser Oscillation in
Microcavities." CRC Press, Boca Raton, FL.
Zhang, 2. L., Nishioka, M., Weisbuch. C., and Arakawa, Y . (1994). Appl. Phys. Lett. 64, 1068.
Zrenner, A,, Butov, L. V., Hagn, M. Abstreiter, G.. Bohm, G., and Weimann, G . (1994). Phys.
Rev. Lett. 12, 3382.
SEMICONDUCTORS A N D SEMIMETALS. VOL. 60
CHAPTER 2
Molecular Beam Epitaxial Growth of

Self-Assembled InAs/GaAs Quantum Dots
Yoshiaki Nakata and Yoshihiro Sugiyama
ELECTRON
QUANTUM DEVICES
LABORATllRY
FUIITSULABORATORIES LTD
ATSUGI.KANACAWA.
JAPAN
Mitsuvu Sugawara
OPTIC'AL SEMICONOUCTOR DEVICES LABOKATORY
FUIITSULABORATORIES
LID
JAPAN
ATSUGI,KANAGAWA,
I. INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . 117
11. THESTRANSKI-KRASTANOW GROWTH MODE. . . . . . . . . . . . . . . . 119
1. Energy- Balance Model,for Islcind Formation . . . . . . . . . . . . . . 119
2. InAs Island Growfh . . . . . . . . . . . . . . . . . . . . . . . . . 121
3. Multiple-Layer Growth anti Perpendicular Alignment of Islands . . . . . . 125
4. In-Plane Alignment of Ishncis . . . . . . . . . . . . . . . . . . . . 130
111. CLOSELY STACKED InAsJGaAs QUANTUM DOTS. . . . . . . . . . _ . _ _ 132
1. Close Stacking of InAs Islunt6 . . . . . . . . . . . . . . . . . . . . 133
2. Photoluminescence Properties . . . . . . . . . . . . . . . . . . . . 137
3. Zero-Dimensional Exciton Corlfinenwnt Evaluated by Diamagnetic Shifts. 1 140
I V . COLUMNAR InAsJGaAs QUANTUM DOTS. . . . . . . . . . . . . . . . . 143
v. S U M M A R Y . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
Acknowledgments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I52
I. Introduction
The atom-like density of states in quantum dots should drastically

improve the performance of optical devices, especially semiconductor lasers,
and should also be instrumental in the development of novel optoelectronic
and single-electron devices. Since high-performance lasers, including quan-
tum-wire and quantum-dot in active regions, were first proposed theoreti-
cally by Arakawa and Sakaki (1982), fabricating these low-dimensional
117
Copyright 1 1999 by Academic Press
All rights of reproduction in any form reserved.
ISBN 0-12-752169-0
ISSN 0080-8784/99 $3000
118 NAKATA
YOSHIAKI AND YOSHIHIROSUGIYAMA
quantum nano-structures has been a major interest in semiconductor

research. In particular, 111-V nano-structures such as GaAs/AlGaAs, In-
GaAs/GaAs, and InGaAs/InP are major targets in optical and electronic
device applications.
Numerous challenges in quantum-wire and quantum-dot fabrication have
been reported during the last two decades. The most straightforward
technique is to laterally pattern the quantum-well structures through a
combination of high-resolution electron beam lithography and dry or wet
etching (Petroff et al., 1982; Reed et al., 1986; Miyamoto et al., 1987;
Gershoni et al., 1988). Other techniques exploit regrowth of epitaxial layers,
such as fractional layer growth on a vicinal substrate (Petroff et al., 1984;
Fukui and Saito, 1987), selective growth on a patterned substrate (Kapon
et al., 1989, 1992), and cleaved-edge overgrowth (Goni et al., 1992; Pfeiffer
et al., 1993; Wegscheider et al., 1995). However, artificial structures fab-
ricated in these ways did not take full advantage of engineered energy states,
and some had drawbacks. For example, lithography- and etching-based
technologies caused damage to the crystals, such as impurity contamination,
defect formation, and poor interface quality. Further, there were serious
problems in the fabricated structures themselves, such as large size, low
density, and size irregularity. Significant quantum-size effects advantageous
to, say, optical devices appear when the size is less than the exciton Bohr
radius. Moreover, millions of the structures should be densely packed, with
size uniformity on the atomic scale, to obtain the desired optical signal.
Unfortunately, no technique could succeed in achieving all of these require-
ments.
Self-assembling, a novel way to fabricate quantum dots, is now being
welcomed as the most promising approach in overcoming the various
problems with previous techniques. This process exploits the three-dimen-
sional island growth of highly lattice-mismatched semiconductors. The
growth of InAs on a GaAs substrate is a typical example, where the lattice
mismatch between InAs and GaAs is about 7%. Dislocation-free high-
density coherent islands of InAs are self-assembled on the GaAs substrate,
accompanied by a wetting layer. Typical InAs self-assembled islands have a
dome or pyramid shape with a base length of about 20 nm and a height of
a few nm. Since the exciton Bohr radius is 10 to 20nm in an InAs-GaAs
system, the island size is small enough to exhibit the three-dimensional
quantum confinement effect. Actually, well-separated photo-emission spectra
from discrete energy states have been observed (Mukai et al., 1994; Grund-
mann et al., 1996). Room-temperature lasing from the quantum-dot quan-
tized levels has been achieved, and the threshold current has decreased and
is now close to that of strained quantum-well lasers (see Chapter 6).
Though self-assembling is a new way to form quantum dots, the growth
2 MOLECULAR GROWTH
BEAMEPITAXIAL 119
process itself is not new but is simply a Stranski-Krastanow (SK) mode

growth (Stranski and Krastanow, 1937). InGaAs/GaAs islands grown via
the SK mode were observed and evaluated by several groups in the
mid-1980s (Schaffer et al., 1983; Lewis et al., 1984; Glas et al., 1987; Houzay
et al., 1987). However, the research did not attract broad interest, at least as
it applied to quantum dots. That SK InAs islands on a GaAs substrate
might work as quantum dots was proposed by Tabuchi et al. (1992). Since
then many works have identified the islands as quantum dots with three-
dimensional quantum confinement, primarily on the basis of their optical
emission properties (Leonard et al., 1993; Mukai et al., 1994; Marzin et al.,
1994; Oshinowo et al., 1994; Moison et al., 1994). Three-dimensional island
growth, now the center of interest in semiconductor material and device
research, is known as self-assembled growth.
In addition to the InGaAs quantum dots treated in this volume, a variety
of dots of other semiconductor materials have been fabricated through
self-assembling processes. These include InGaP (Carlsson et al., 1995;
Reaves et al., 1995), CdSe (Arita et al., 1997), and GaN (Tanaka et al., 1996,
1997). In the near future, this new material category will improve various
optical devices like semiconductor lasers in many aspects (see Chapter 7)
and in wavelength regions ranging from infrared to red to blue.
This chapter focuses on the molecular beam epitaxial (MBE) growth of
InAs islands on a GaAs substrate. Section 11, briefly reviews the three
well-known growth modes and a simple energy balance model for island
formation, and then describes the growth process of InAs three-dimensional
islands with various diagnostic results. Then, Section I11 introduces the
perpendicular stacking of InAs islands, called closely stacked quantum dots,
and Section IV introduces columnar-shaped quantum dots. Perpendicular
stacking increases the island size primarily in the growth direction and
enables us not only to tune the emission wavelength but to narrow the
spectrum linewidth caused by the island-size fluctuations. By comparing the
structural and optical properties of different types of dots, we can see the
perpendicularly stacked columnar shaped quantum dots’ advantages.
11. The Stranski-Krastanow Growth Mode
1. ENERGY-BALANCE
MODELFOR ISLAND FORMATION
The crystal growth on a bulk substrate occurs in one of three distinct

modes, schematically shown in Fig. 2.1; Frank-van der Merwe (FM),
Volmer-Weber (VW), and Stranski-Krastanow (SK). While growth pro-
ceeds layer by layer in the FM mode, VW growth causes three-dimensional
120 YOSHIAKINAKATA SUGIYAMA
AND YOSHIHIRO
Frank-van der Merwe mode Volmer-We&r mode

Epitaxial layer Island
FIG. 2.1. Three types of growth mode on a substrate; Frank-van der Merwe (FM);
Volmer-Weber (VW); and Stranski-Krastanow (SK). In the F M mode growth proceeds layer
by layer. In the VW mode growth occurs as three-dimensional islands on the substrate. The
SK mode is a combination of the FM and VW modes, where the growth of a several-
monolayer-thin film, called a wetting layer. is followed by cluster nucleation, leading to island
formation.
islands on the substrate, if the film has a higher surface energy than that of
the substrate. SK growth is a combination of the FM and VW modes, where
the growth of a several-monolayer thin film, called a wetting layer, is
followed by cluster nucleation and then to island formation. Which growth
mode occurs depends primarily on the difference in the surface energy
between the substrate and the grown material and on the strain energy
accumulated in the grown materials as a result of lattice mismatch. In the
strained-layer growth of semiconductors discussed here, the SK growth
mode has been found to dominate.
Why islands form in the lattice-mismatched epitaxy can be understood by
a simple model of energy balance (Tu et al., 1992). Let us compare the total
energy between a coherently strained film on a substrate and strain-free
islands made by the same number of atoms as in the film. The total energy
is supposed to consist of the strain energy due to lattice mismatch and the
surface energy. The strain-free islands are, for simplicity, assumed to be
cubic shaped with a side length of X . Instead of releasing the strain energy,
the cubic islands have greater surface energy due to the increased surface
area. The island energy is lower than the film energy, then, when the island
is larger than a critical value determined by the surface energy, y , and the
in-plane strain, E, as
x > x,K y/&2

2 MOLECULAR
BEAMEPITAXIAL
GROWTH 121
When E # 0, there is always a value of X above which the island is in the

lower energy state; that is, when the side length of the islands is greater than
X,, the island configuration will be favored. This indicates that, given
sufficient growth temperature and time, islands are always formed in the
strained layer epitaxy. In contrast, the present strained-layer epitaxial
growth using metalorganic chemical vapor deposition (MOCVD) and MBE
enables us to obtain coherent strained layers over 1% lattice mismatch
without dislocations, which tells us that strained-layer epitaxy occurs under
metastable conditions, and that surface kinetics also play an important role.
Though the present energy balance model looks too simple, in Eq. (2.1)
we see that the critical size is inversely proportional to the square of the
strain and that we will have smaller islands as the lattice mismatch increases.
In the following sections, we will see that this simple rule explains actual
island growth rather well.
2. InAs ISLAND GROWTH
When InAs is supplied to a GaAs substrate, three-dimensional island

growth occurs followed by the two-dimensional growth of a wetting layer.
Before we move on to the islands’ perpendicular stacking, let us see in detail
how the growth of individual islands proceeds. With MBE growth, the
reflection high-energy electron diffraction (RHEED) pattern gives us im-
portant information on the surface state. When the layer-by-layer growth
proceeds, RHEED shows streak patterns. The RHEED pattern transits to
spots when three-dimensional islands start to grow. For the post-growth
surface morphology evaluation technique, we use atomic force microscopy
(AFM), from which we know the shape, size, density, alignment, and the
distance between neighboring islands. We can get similar information from
plan-view transmission electron microscopy (TEM) even when islands are
covered by overgrown layers, we can also evaluate defects, such as disloca-
tions and stacking faults. Cross-sectional TEM is quite useful for observing
cross-sectional shape and height of the islands, and how the islands are
stacked perpendicularly and interact with each other. We will see this in
Sections I11 and IV.
For the growth of InAs islands, a conventional MBE was used with
metallic In, Ga, Al, and As, as source materials. The substrates were
(001)-oriented GaAs. Before growth, they were thermally cleaned at about
680°C for one minute under an arsenic pressure of 1.2 x lo-’ Torr. InAs
islands were grown on a GaAs (100 nm)/AlGaAs (50 nm)/GaAs (400 nm)
buffer layer and covered with a GaAs (50 nm)/AlGaAs (50 nm)/GaAs
(100 nm) cap layer. The substrate temperature for the growth of InAs was
122 YOSHIAKI AND YOSHIHIRO
NAKATA SUGIYAMA
varied between 470 and 560°C, and was 650°C for the growth of the buffer
layer. The substrate temperature was monitored by a pyrometer calibrated
by the melting temperature of In-A1 alloys.
The growth rates of InAs and GaAs were approximately 0.1 and 0.75
pm/h. In other words, it takes about 10 seconds to grow one monolayer
(ML) InAs and about 1.3 seconds to grow one ML GaAs. The arsenic
pressure used for InAs island growth was 6 x Torr and was 1.2 x
Torr for buffer-layer growth. These values were measured by an ion gauge
at the substrate position. Two arsenic cells were used, both set at 6 x lop6
Torr. The arsenic pressure was changed abruptly at the interface by
switching one of these cells on and off. After 60-second annealing of the InAs
islands, GaAs layers were overgrown.
The variation of the RHEED pattern and the diffraction intensity are
shown in Fig. 2.2. During the growth of the buffer layer, the surface was
Growth time (s) 2 0 ~ 1 div.

1
FIG.2.2. Variation of the RHEED pattern and the reflection intensity for the various stages
of growth of lnAs islands. The pattern transition from streaks to spots began as the growth of
lnAs reached around 1.6 monolayers at 16 seconds, indicating that two-dimensional growth
transited to three-dimensional island growth. The reflection intensity reached an almost
constant value at the 1.8-ML InAs supply. The pattern went back to streaky at the 6-ML GaAs
growth, showing that 6-ML GaAs completely covers the InAs islands and almost flattens the
surface.
2 MOLECULAR
BEAMEPITAXIAL
GROWTH 123
As-stabilized (2 x 4) and then transited to c(4 x 4) reconstruction while the

substrate temperature dropped to 510°C. The leftmost photograph in Fig.
2.2 is the [ l l O ] azimuthal RHEED pattern obtained before InAs growth.
The lp-order fractional diffractions were clearly observed. Immediately
after the InAs growth started, the fractional-order reflections disappeared,
and the transition from streaks to spots started as the growth of InAs
reached around 1.6 M L (at 16 seconds), indicating that a two-dimensional
layer growth transited to a three-dimensional island growth. When the
GaAs supply started to overgrow the InAs islands as a cap layer, the spot
intensity rapidly decreased. The pattern went back to streaky at the 6-ML
GaAs growth, showing that the 6-ML GaAs supply fully covers the InAs
islands and almost flattens the surface.
The island size and density were evaluated by ex-situ AFM operated in
the air. Figure 2.3 shows AFM images of the island surfaces grown with the
InAs supply of 1.3, 1.6, 2.1, and 2.6 ML. These images refer to different
epilayers grown under the same conditions. The scanned area is 0.5 x
0.5 pm2. The island growth started at an InAs supply of about 1.6 ML, as
seen in the change of RHEED pattern. As the amount of InAs supply
(c) 2.1 ML (d) 2.6 ML

0.5 x 0.5 pm
2
FIG. 2.3. AFM images of the islanding surfaces when the InAs supply was 1.3, 1.6, 2.1, and
2.6 MLs. The islanding growth starts at the 1.6-ML supply, as seen in the change in RHEED
pattern. As the supply amount increases, the dot density rapidly increases and the dots coalesce.
124 YOSHIAKI
NAKATA
AND YOSHIHIRO
SUGIYAMA
increased, the island density rapidly increased, and the islands began to
coalesce when the amount of InAs exceeded approximately 2.3 ML.
Figure 2.4 shows the in-plane base diameter and the density of the islands
as a function of the amount of InAs supply (InAs nominal thickness (ML)).
As depicted in the figure, growth proceeds from the two-dimensional growth
of a wetting layer to the nucleation of InAs islands at about a 1.6-ML
supply, to the increase in the island density, and finally to the coalescence
of neighboring islands at about a 2.3-ML supply.
Figure 2.5 shows the base diameter and density of islands grown with the
2.1-ML InAs supply as a function of the growth temperature. The AFM
images of islands grown at each temperature are also shown. As temperature
increases from 470 to 56OoC,the island diameter increases from 15 to 45 nm
and the density decreases correspondingly. This temperature dependence is
due to the surface migration lengh of TnAs, which increases as the tempera-
ture increases.
Figure 2.6 shows the plan view (a) and the (1 10) cross section (b) of TEM
images for InAs islands grown with a 2.5-ML InAs supply. Observable were
lnAs nominal thickness (ML)

FIG. 2.4. In-plane diameter and density of islands as a function of the InAs nominal
thickness. The growth proceeds from the two-dimensional growth of a wetting layer, to the
nucleation of InAs islands at about 1.6 ML supply. to the increase in the density, and finally,
to the coalescence of neighboring islands at 2.2-2.3-ML supply.
60 - BEAMEPITAXIAL
2 MOLECULAR GROWTH
lo'*
Ts =
510°C
540°C
125
0
450 500 550 600
Growth temperature ( "C) 560°C
0.5 x 0.5 pm2

FIG. 2.5. Diameter and density of islands at 2.1-ML InAs supply as a function of growth
temperature, and the AFM images of islands at each temperature. As the temperature increases
from 470 to 560'C. the island diameter increased from 15 to 45 nm and the density decreases
correspondingly.
two types of islands-small, with a diameter of 15-20 nm, and large, with
about twice the diameter -which were formed probably by coalescence.
While the height of most islands is 3 to 5 nm, that of the large-size islands
exceeds 10nm. Contrasts due to strain were observed both in the GaAs
buffer layer and in the GaAs overlayer, and they extended as the dot size
increased. There are two types of defects: V-shaped dislocations and multiple
stacking faults. In the upper middele part of Fig. 2.6(a), a large island
accompanied by line-shaped defects on both sides could be seen. These
defects are a V-shaped dislocation pair lying on two equivalent { 11l } planes.
Note that both types of defect are generated from large islands and almost
no defects are observed around the small islands (Ueda et al., 1998).
3. MULTIPLE-LAYER
GROWTH
AND PERPENDICULAR
ALIGNMENT
OF ISLANDS
The InAs islands buried in GaAs work as quantum dots because they are
smaller than the exciton wave function both in the in-plane and perpendicu-
lar directions. When we apply the TnAs island layers to the active region of
126 YOSHIAKI AND YOSHIHIROSUGIYAMA
NAKATA
FIG. 2.6. (a) the (001) plan-view and (b) the (110) cross-sectional TEM images of InAs
islands (2.5-ML InAs coverage). Observed are small islands with a diameter of about 20nm
and large islands with about twice that diameter formed by the coalescence of two islands.
While the height of most islands is 3 to 5 nm, that of the large islands exceeds 10 nm. Contrasts
due to strain are observed in both the GaAs buffer layer and the GaAs overlayer, and they
extend as the dot size increases. There are two types of defect to be observed: V-shaped
dislocations and multiple-stacking faults.
2 MOLECULAR
BEAMEPITAXIAL
GROWTH 127
optical devices, multiple-layer stacking of the island layer is often required

to obtained enough interaction between confined electrons and the electro-
magnetic field. This is immediately understood if we think of multiple
quantum-well lasers.
Figure 2.7 shows the cross-sectional TEM images of the stacked island
structures grown with the intermediate layer thickness of lOnm (a) and
20nm (b). For the multiple-layer stacking, the growth sequence shown in
Fig. 2.2 is simply repeated. These intermediate-layer thicknesses are much
greater than the island height of about 3-5 nm. In the 20-nm structure (b),
the islands form independently of the lower-layer islands. AFM images show
that the mean island size and the density are almost constant in all island
layers. When the intermediate layer is reduced to 10 nm, the islands align in
the perpendicular direction and the size increases as we move to the upper
layers.
This perpendicular alignment of islands is due to the strain fields induced
by the lower layer islands. Xie (1995) provided an analytical description of
correlated island formation in the growth direction under strain fields. His
explanation follows: Islands on the first layer produce tensile stress in the
GaAs intermediate layer above the islands. Let 1, be the effective length of
the lateral strain field extent at the GaAs surface. The InAs that forms the
next layer impinging inside I, around the lower island preferentially migrates
from regions of higher to lower lattice mismatch and accumulates just on
top of the lower islands. Thus, the InAs can achieve an energetically lower
thermodynamic state due to lower lattice mismatch against the GaAs in
tension. The strain fields induced by the islands provide the driving force for
(a) Intermediate layer: 10 nm (b) Intermediate layer: 20 nm

FIG. 2.7. Cross-sectional TEM images taken when the intermediate GaAs layer thickness
was (a) 10 nm and (b) 20 nrn. These intermediate layer thicknesses are greater than the height
of the InAs islands, which is 3-5 nm. In the 20-nm structure, the island formation occurs
independently between layers. When the intermediate layer was reduced to 10 nm, the islands
aligned in the perpendicular direction and the size increased as growth moved on to the upper
layers.
128 AND YOSHIHIROSUCIKAMA
YOSHIAKINAKATA
perpendicularly aligned self-assembling island formation. In the region

outside l,, where there is negligible stress, some of the impinging In atoms
initiate formation of islands. Thus, in order to achieve effective perpendicu-
lar alignment, the island separation, I, should satisfy the condition of
21, > 1. This condition is achieved by decreasing the intermediate GaAs
thickness below a critical intermediate layer thickness.
The size and density of islands vary as the multiple-layer stacking
proceeds, as seen in Fig. 2.8. When we grew 1.8-ML InAs islands using
10-nm thick intermediate GaAs layers, the average diameter of the 10th-
layer islands was about 45 nm, 90% larger than the first-layer islands, which
were 24 nm in diameter. The island density decreased from 1 x 10" cm-2
for the first-layer islands to 3 x 10'ocm-2 for the 10th-layer islands.
These phenomena were also reported by Solomon et al. (1996).
The increase in the island size and the decrease in the density can be
understood by the above-mentioned strain field model as follows. The size
distribution of self-assembled islands leads to spatial fluctuation of strain
fields at the surface of GaAs. Supplied InAs for the growth of the next island
layer preferentially accumulates on the sites with larger strain fields. Due to
smaller lattice mismatch between InAs and the surface, the island becomes
large. As a result, large islands are formed preferentially on the large-strain-
field region, preventing island formation on the small-strain-field region.
This explains the increase in the island size and the decrease in density in
multiple-island growth with about a 10-nm intermediate layer.
As seen in Fig. 2.8, as we further reduce the thickness of the intermediate
layer below 10 nm, the increase in size and the decrease in density become
less remarkable. For example, when we grew multiple layers with 3-nm thick
intermediate layers, the average diameter of the 10th-layer islands was about
33 nm-40% larger than the 1st-layer islands, with 24-nm diameter- while
the islands' density decreased from 1 x 10' cm2 for the 1st-layer islands to
8 x lOl0cm2 for the 10th-layer islands. This is because the thinner inter-
mediate layer causes the strain field over small islands to become large
enough to accumulate supplied InAs.
Photoluminescence spectra from single-layer and multiple-layer samples
with 20-nm intermediate layers at 77 K are shown in Fig. 2.9. The sample
was excited by an Ar' ion laser at a power of 1 mW, and the luminescence
was dispersed by a monochrometer and detected by a cooled Ge detector.
The laser spot was about 100pm in diameter. The emission spectrum
appeared at around 1.2 eV with a full width at half maximum (FWHM) of
90meV for both samples. Such large spectrum width is typically observed
in self-assembled SK-mode islands so far reported. This is inhomogeneous
broadening, caused by the fluctuation of the quantized energies among
islands included in the measured area (lo6 to lo7 islands). By means of
2 MOLECULAR
BEAMEPITAXIAL
GROWTH 129
Intermediate layer thickness, L (nm)
FIG. 2.8. In-plane diameter and density normalized against the first-island-layer density as
a function of the intermediate-layer thickness. When 1.8-ML InAs islands were grown using
LO-nm thick intermediate GaAs layers, the average diameter of the 10th-layer islands was about
45m, 90% larger than the first-layer islands, with a 24-nm diameter. The island density
decreased from 1 x 10" em-' for the first-layer islands to 3 x 10" cm-' for the 10th-layer
islands. As the thickness of the intermediate layer was further reduced below 10 nm, the increase
in size and the decrease in density became less remarkable.
microprobe photoluminescence to access a limited number of islands, a

sharp emission spectrum with around 100peV is observed (Marzin et al.,
1994; Fafard et al., 1994; Grundmann et al., 1995a; Leon et al., 1995;
Hessman et al., 1996).
When we apply SK-mode islands as the quantum dots to the laser active
region, the large spectrum broadening lowers the optical (differential) gain
and prevents us from achieving the high performance predicted in quantum-
dot lasers. For example, low gain leads to lasing from excited levels with a
high density of states, increasing the threshold current. Low differential gain
lowers the relaxation oscillation frequency and limits the modulation
bandwidth. One exceptional device that prefers large broadening is multi-
YOSHIAKINAKATA
AND YOSHIHIROSUGIYAMA
PL
77 K
Stacked
island layers
FWHM:
90 meV
0.9 1.0 1.1 1.2 1.3 1.4

Energy (eV)
FIG. 2.9. Photoluminescence spectra from single-layer and multiple-layer samples with
20-nm intermediate layers at 77 K. The emission spectrum appeared at around 1.2 eV with the
full width at half maximum (FWHM) of 90meV for both samples. This inhomogeneous
broadening was caused by the fluctuation of the quantized energies among islands included in
the measured area (lo6 to lo’ islands).
wavelength optical memory, where the spectrum hole burning is exploited

for data storage and the large spectrum broadening increases the memory
size (Muto, 1995).
The broadening of luminescence should be controlled by the structural
fluctuations, especially by the island height, since the SK-mode islands have
a height of 3-5nm, which is much smaller than the diameter of about
20nm. This is understood by a well-known concept that the quantized
energy changes by a constant size fluctuation as the size of the confinement
region decreases. Thus, if the height can be increased or more accurately
controlled, emission spectrum broadening will be greatly reduced. Increase
or control is achieved by means of close stacking of islands in the
perpendicular direction, as will be seen in Sections I11 and IV.
ALIGNMENTOF ISLANDS
4. IN-PLANE
The islands become technologically more interesting if we can manipulate

their arrangement laterally as well as vertically to achieve three-dimensional
arrays. There are several reports on lateral ordering, in-plane alignment, and
2 MOLECULAR GROWTH
BEAMEPITAXIAL 131
control of islanding sites in specific regions, which we summarize in the

following paragraphs.
The short-range ordering of InAs SK-mode islands grown on (001) GaAs
by MBE was reported by Grundmann et al. (1995b), who showed by
plan-view TEM observation that InAs islands align in the quasi-periodic
square lattice with an axis of (100). Shchukin et al. (1995) theoretically
explained this spontaneous ordering by showing that a periodic array of
strained islands arranged in a two-dimensional square lattice has a mini-
mum total energy with main axes along the [loo] and [OlO] directions.
Tersoff et al. (1996) presented a model showing that island size and spacing
grow progressively more uniform as successive island layers are stacked.
Nishi et al. (1997) observed spontaneous lateral alignment of InGaAs
islands grown by gas-source MBE aligned in a direction inclined about 60"
from the [Oll] direction on (311)B surfaces.
Surface steps play an important role in determining strained-island
nucleation. Leonard et al. (1994) and Ikoma and Ohkouchi (1995) showed
by AFM and ultra-high-vacuum scanning tunneling microscopy (UHV-
STM) that InAs islands form in alignment along the monolayer step edges
on (001) GaAs. Kitmura et al. (1995) demonstrated that InGaAs islands
grown by MOCVD preferentially form on the bunched steps on the
misoriented (001) GaAs substrates, and that selective island formation on
the bunched step is possible.
Growth on prepatterned substrates has been extensively studied as a way
to control directly the in-plane alignment or position of islands. Preferential
island formation has been found either on top of the ridges or along the
sidewalls of the mesa stripes and at the bottom of the V-grooves, trenches,
holes, and pits. Mui et al. (1995) demonstrated the self-alignment of InAs
islands on etched GaAs ridges running along the [ O l l ] and [OlT] directions
on the (100)-oriented substrates. InAs islands were formed on the sidewalls
of ridges running along the [llO] direction, while for the ridges running
along the [OlT] direction, islands were formed on the (100) plane on and at
the foot of the mesa stripes. Moreover, as the grating pitch was reduced to
0.28 mm, islands were located either on the sidewalls or at the bottoms, with
none on the tops. Seifert et al. (1996) reported InP islands on InGaP/GaAs
overgrown stripes lithographically defined by metal stripes 30" off from the
[TlO] direction on the (001) GaAs substrates. They demonstrated that the
islands aligned either on top of the ridges, at the sidewall near the mesa
edges, or at the bottom of the trenches, depending on the geometry of the
InGaP/GaAs overgrown mesa stripes. Similarly results were obtained by
Jeppensen et al. (1996) in an InP islands/InGaP/(OOl) GaAs system. They
found that InAs islands grown by chemical-beam epitaxy form in chains
with a minimum period of 33 nm along the trenches, and that single or a
132 YOSHIAKINAKATA
AND Yos~mwoSUGIYAMA
few islands that grow in an array in electron-beam lithographic holes on a

(100) GaAs surface. Tsui et al. (1997) and Konkar et al. (1998) tried to form
InAs islands selectively on top of the mesa stripes fabricated on (001) GaAs
substrates. They demonstrated selective positioning of InAs islands on the
top mesas depending on the geometry. The selective island formation on
non-(001) prepatterned substrates were studied. Saitoh et al. (1996) used
lithographically fabricated (wet-etched) tetrahedral-shaped recesses and
V-grooves on (11 l)B GaAs substrates. They found that InAs islands
selectively form at the bottom of those recesses.
One of the most important purposes of the study of in-plane alignment of
islands is improved structural uniformity. However, this purpose has yet to
be achieved. We must further improve size and composition uniformity of
islands grown on those stepped surfaces and grown by selective area growth.
111. Closely Stacked InAslGaAs Quantum Dots
The alignment of islands in the perpendicular direction will make it

possible to couple islands electrically in the vertical direction as we reduce
the intermediate GaAs thickness to the extent that the electron wave
functions of neighboring wells are overlapped. This is the same situation as
multiple quantum-well transfer to a superlattice when the barrier layer
thickness is reduced. Vertical coupling enables electron tunneling between
quantum dots, which lead to such novel electronic applications as a
single-electron tunneling device.
When we think of optical-device application of quantum dots, the most
promising and practical advantage of the stacking technique will be size
control. By choosing the intermediate-layer thickness and the repetition
number of island layers, we will be able to tune the island size (height) and
thus the quantized energies to meet device requirements. In addition,
according to the discussion in Section 11, the size increase in the perpendicu-
lar direction will lead to a narrowing of the emission spectrum which will
be quite beneficial for lasers.
This section describes the growth process, the crystal structures, and the
optical properties of perpendicularly stacked islands when the intermediate
GaAs layer is reduced to a few nanometers close to or comparable to the
height of InAs islands. Even under this condition, InAs islands are repeatedly
grown. Optical diagnostics show that the stacking of the InAs islands
increases the effective size of quantum dots in the perpendicular direction due
t o electrical coupling, resulting in the narrowing of the spectrum FWHM to
25 meV. We call the stacked structure closely stacked quantum dots (Nakata
et al., 1997).
2 MOLECULAR GROWTH
BEAMEPITAXIAL 133
1. CLOSESTACKING
OF InAs ISLANDS
Figure 2.10 shows the growth sequence for the close stacking of InAs
islands. The amount of InAs supply for island formation was fixed at about
1.8 ML, and the nominal thickness of the GaAs intermediate layers was set
at 2 and 3 nm. Prior to and following InAs island growth, the sample was
annealed for 2 minutes and 1 minute, respectively. The growth rates, arsenic
pressure, and growth temperature were the same as in Section 11.1.
Figure 2.1 1 shows the RHEED-pattern intensity transition, observed at
the area indicated by an arrow in the inset patterns, during the growth of
InAs islands. The RHEED shows the streak pattern for the two-dimensional
growth in the early stage, and the change to the spot pattern for the
three-dimensional growth at above a critical amount of InAs supply. The
SK-mode islands also grow for the 3rd and 5th layer. Note that the 3rd- and
5th-layer islanding started when the growth of InAs reached about 1 ML-
about 63% of the 1st-layer islanding (approximated 1.6 ML). The reason for
the smaller critical amount for the islanding is thought to be that the strain
induced by the lower-layer islands accumulates InAs preferentially or that
segregation of InAs from lower islands. The transition of the growth mode
from two- to three-dimensional and the existence of wetting layers, as will
be shown in the TEM image (Fig. 2.14(a)), both indicate that SK-growth
islands were formed even on such thin GaAs intermediate layers.
FIG. 2.10. Growth sequence for the close stacking of InAs islands. The amount of InAs
supply for island formation was fixed at about 1.8 ML, and the nominal thickness of the GaAs
intermediate layers was set at 2 and 3nm. Prior to and following InAs island growth, the
sample was annealed for 2 minutes and 1 minute, respectively.
134 YOSHIAKINAKATA
0 10 20 30 40
Time (s)
FIG. 2.1 1. RHEED pattern intensity transition observed at the area indicated by an arrow
in the inset patterns during the growth of InAs islands. The RHEED pattern shows that the
SK island growth occurred also for the 3rd and 5th layer. Note that the 3rd- and 5th-layer
island started when the growth of InAs reached about IOML, which was about 63% of the
1st-layer islanding.
The island size and the density at each layer were evaluated by ex-situ
AFM. Figure 2.12 shows AFM images of the islanding surfaces at the (a)
lst, (b) 3rd, (c) 5th, and (d) 10th layers stacked with 3-nm thick intermediate
layers. These images refer to different epilayers grown under the same
conditions. The scanned area is 250 x 250 nm'. The upper layer islands
expanded slightly as the number of stacked layers increased. Figure 2.13
shows the dependence of island size on the number (diameter) and density
of stacked layers. The average diameter of the 10th-layer islands was about
33 nm-40% larger than the 1st-layer islands, which were about 24 nm in
diameter-while the island density decreased from 1 x loi1 cm-' for the
1st-layer islands to 8 x 10'ocm-2 for the 10th-layer islands. The increase in
the diameter and decrease in the density were caused by the strain field
formed by the lower islands as in Section 1.3.
TEM photography shows the overall structural features. Figure 2.14(a) is
a (110) cross-sectional TEM image of a 5-island stacked structure grown
with 2-nm thick intermediate layers. Each island layer was accompanied by
2 MOLECULAR
BEAMEPITAXIAL
GROWTH 135
250 x 250 nm2
FIG. 2.12. AFM images of the islanding surfaces at the (a) Ist, (b) 3rd, (c) Sth, and (d) 10th
layers stacked with 3-nm thick intermediate layers. The upper-layer islands expanded slightly
as the number of stacked layers increased.
40 1.5
n
30
EC
U
10
0 0
0 2 4 6 8 10 12
Layer number
FIG. 2.13. The dependence of the island size (diameter) and density on the number of
stacked layers. The average diameter of the 10th-layer islands was about 33 nm-40"/0 larger
than the diameter of the 1st-layer islands (about 24 nm), while the island density decreased from
1 x 10" cm-' for the 1st-layer islands to 8 x IOL0cm-*for the 10th-layer islands.
136 YOSHIAKI NAKATA
AND YOSHIHIRO SUGIYAMA
(a)
FIG.2.14. (a) ( I 10) cross-sectional TEM image of a 5-layer stacked island structure grown
with 2-nm thick intermediate layers. Each island layer was accompanied by a wetting layer.
The islands as a whole were about 22 nm in diameter and 13 nm in height, shown in the image
as dark megaphone-like strained regions. (b) Plan view of a TEM image obtained from a
5-layer stacked island structure. Island density was 8 x 10'" cm-'. Size uniformity and
lateral ordering were improved compared to a single-island layer.
a wetting layer, indicating that the upper-layer islands formed via SK

growth, as did the 1st-layer islands. The upper-layer islands grew just on the
lower-layer islands, aligning vertically. These islands as a whole had a
structure of about 22 nm in diameter and 13 nm in height, shown in the
image as dark megaphone-like strained regions. Figure 2.14(b) is a plan view
of a TEM image obtained from a 5-layer stacked island structure grown
with 2-nm thick intermediate layers. The density of the islands was
8 x 10'0cm-2, which agreed with the AFM results. What is surprising is
that tht size uniformity and lateral ordering improved compared to the
ordinary SK-mode islands (a single island layer). Although we attempted
2 MOLECULAR GROWTH
BEAMEPITAXIAL 137
GaAs overgrowth Annealing 2nd lnAs growth

on lnAs islands (growth interruption)
FIG. 2.15. Growth process of stacked islands. The height of the InAs island after GaAs
overgrowth decreased to the thickness of the intermediate layer. Then SK growth occurred on
top of the lower island. This phenomenon enables us to precisely control the thickness of
islands in the growth direction through the intermediate-layer thickness.
control only in the vertical dimension of the islands, uniform lateral

dimension was also achieved in these closely stacked structures.
Islands in each layer are seen to be spatially isolated in the vertical
direction, with a 3- to 4-ML distance between the bottom of the upper-layer
islands and the top of the lower-layer islands. The individual island was
smaller in height than the intermediate GaAs layer thickness, and looked as
if it were being buried in GaAs. The RHEED showed a streak pattern after
overgrowth of GaAs and growth interruption, indicating that the interface
between the overgrown GaAs and the upper InAs wetting layer is almost
flat. A model to explain the formation process of this structure is illustrated
in Fig. 2.15. First, the intermediate layers of GaAs overgrow away from the
InAs islands, as confirmed by Xie et a]. (1995). During annealing (growth
interruption after the GaAs overgrowth), the InAs of the upper part of the
islands above the GaAs overlayers leave the islands and regrow to form
parts of a wetting layer on GaAs. Then some of the InAs at the top regions
in the remaining islands replaced with GaAs to reduce the total energy. This
leads to a decrease in the island height. Similar features were also shown by
Bimberg et al. (1996) and Ledentsov et al. (1996). This phenomenon enables
us to precisely control the island height in the growth direction through the
intermediate-layer thickness.
The island-to-island distance of 3 to 4 M L is so thin that the electron
wave functions of each island can be overlapped along the vertical direction.
This suggests that the islands are electronically coupled and behave as a
single quantum dot, which we will confirm by the optical diagnostics as
follows.
2. PHOTOLUMINESCENCE
PROPERTIES
Light emission properties of these structures were evaluated using a

photoluminescence technique with the same measurement conditions given
138 YOSHIAKI NAKATA
I I I I I
0.9 1.0 1.1 1.2 1.3 1.4

Energy (eV)
0.9 1 1.1 1.2 1.3 1.4

Energy (eV)
FIG.2.16. (a) Emission spectra at 77 K of a single island layer and closely stacked 3-, 5-,
and 10-layer island structures grown with 3-nm thick intermediate layers. The peak energy
shifted to a lower energy as the number of stacked layers increased (about 90 meV in the 5-layer
islands). At the same time, the broad emission spectrum of the single-island layer drastically
narrowed at the 5-stacked layer to an FWHM of 27 meV-about one-third of that obtained
for a single-layer island. (b) Interval layer thickness dependence on emission spectra with the
five stacked layers. As the intermediate layer thickness decreased, the emission spectrum shifted
toward lower energies and narrowed to an FWHM of 27 meV a t the 3-nm interval.
2 MOLECULAR GROWTH
BEAMEPITAXIAL 139
in Section 111. Figure 2.16(a) shows the emission spectra at 77 K of a single

island layer ( N = 1) and closely stacked 3-, 5-, and 10-layer island structures
grown with 3-nm thick intermediate layers ( N = 3, 5, lo). The peak energy
shifted to a lower energy as the number of stacked layers increased (about
90 meV in the 5-layer stacked islands). At the same time, the broad emission
spectrum of the single-island layer drastically narrowed at the 5-layer-
stacked structure to an FWHM of 27 meV, which is about one-third of that
obtained for a single-layer island. For a stack of 10 layers, the FWHM
increased and integrated photoluminescence intensity decreased. This could
be due to the increase in the total amount of strain around the stacked
island structures, which induces dislocations and structural modulation.
Figure 2.16(b) shows the dependence of the intermediate-layer thickness on
emission spectra with the five stacked layers. As the intermediate layer
thickness decreased, the emisson spectrum shifted toward lower energies and
narrowed to an FWHM of 27 meV at the 3-nm thick intermediate layer.
Figure 2.17 shows the excitation power dependence of the photolumines-
cence spectra for the 5-layer stacked islands with 2-nm (a) and 3-nm
intermediate layers (b). As the excitation power increased, both samples
exhibited peaks in the higher energy regime, which can be attributed to the
higher-order quantized states. Note that energy separation between adjacent
states was almost constant for both samples. The energy separation between
the ground-state and the first excited-state emission was about 52 meV for
the sample grown with 2-nm-thick intermediate layers and 44 meV for the
sample grown with 3-nm-thick intermediate layers. The energy separa-
tion was smaller than that of the ordinary SK-mode islands (single-island
layer) of about 70 meV, as seen in the electroluminescence spectrum (see
Chapter 6).
All the results observed above indicate that the electron wavefunctions
between neighboring islands in the perpendicular direction overlap each
other and that the stacked structure, as a whole, can work as a single
quantum dot with a larger size. The proofs are (1) the red shift of the
emission spectra with an increase in the repetition number; (2) the red shift
of the emission spectra with a decrease in the intermediate-layer thickness;
and (3) the decrease in the energy separation between the discrete energy
states. As a result, we can control the quantum-dot size in the perpendicular
direction by this close-stacking technique through the change in the stacking
repetition number and in the intermediate layer thickness.
The drastic improvement in spectrum broadening was just as expected.
One direct reason for this success is that island size (height) is effectively
increased, as seen above, decreasing the influence of dot size (height)
fluctuation. Also, the island height is controlled more by the intermediate-
layer thickness, as seen in Fig. 2.15. Finally, lateral dot size is more uniform,
as seen in Fig. 2.14(b).
140 YOSHIAKI NAKATA SUGIYAMA
AND YOSHIHIRO
0.9 1 1.1 1.2 1.3 1.4

Energy (eV)
(a)
0.9 1 1.1 1.2 1.3 1.4

Energy (eV)
(b)
FIG.2.17. Excitation power dependence of photoluminescence spectra for the 5-layer
islands with (a) 2-nm and (b) 3-nm thick intermediate layers. As the excitation power increased,
both samples exhibited peaks in the higher energy regime, which can be attributed to the
higher-order quantized states.
EXCITONCONFINEMENT EVALUATED
3. ZERO-DIMENSIONAL BY
DIAMAGNETIC
SHIFTS
A question that often arises when we observe luminescence spectra from

quantum-dot samples is whether the emission really originated from the
dots and reflects the characteristics peculiar to their three-dimensional
confinement. With closely stacked dots in particular, the identification is not
so straightforward due to their structural complexity -namely, multiple
stacked structure accompanied by wetting layers. The emission spectra in
2 MOLECULAR
BEAMEPITAXIAL
GROWTH 141
Section 111.2 indicated that the exciton states of the stacked dot form
through the wave function overlap of each SK dot component, and that the
perpendicularly stacked structure should be considered as one quantum dot.
Magnetophotoluminescence, treated here, will give us not only the evidence
of the three-dimensional confinement of excitons but also the exciton wave
function extent in quantum dots. The principle can be summarized this way.
The magnetic field confines excitons in the plane perpendicular to the field
and increases their energy. We can determine the magnitude of other
competing confinement potentials and thus the extent of the wave function
by evaluating the number of energy shifts.
The samples and the magnetic field direction are schematically shown
in Fig. 2.18(a). The samples were five-layer-stacked InAs SK islands with
3-nm GaAs intermediate layers and, for comparison, an In,,,,Ga,~,,As/
GaAs quantum-well sample grown on a GaAs substrate by MOCVD.
The In,,,,Ga,,,,As well-layer thickness was 8 nm and subjected to - 1.3%
biaxial compressive strain. Figure 2.18(b) shows the photoluminescence
spectra of the two samples with no magnetic field applied.
Figure 2.18(c) shows the diamagnetic shifts for both samples under
magnetic fields perpendicular and parallel to the sample plane at 4.2 K. In
the quantum well, while the shift under a perpendicular field reached up to
7.3 meV at a maximum field of 11.8 T, the shift under the parallel field was
much smaller (only 1.6meV at 11.8 T). The shift of the dots was almost
independent of the field direction, and the maximum shift was 2.4meV in
either direction. These results immediately give us the following insights into
the difference in the quantum-confinement characteristics between the two
samples. The asymmetrical shift in the quantum well is due to its one-
direction quantum confinement. When the field is perpendicular to the
plane, it works as a two-dimensional confinement potential for excitons in
the well layer, causing large diamagnetic shifts. When the field is parallel to
the plane, it affects only one direction in the quantum-well plane, since the
other direction is already confined by strong potential barriers. The sym-
metrical, small diamagnetic shifts in the closely stacked dots show that
excitons are three-dimensionally confined by an almost symmetrical confine-
ment potential. The extent of the exciton wave function is estimated to be
almost equal to the quantum-well thickness of 8 nm, since the shifts for the
closely stacked dots and the quantum well under the parallel configuration
are almost the same.
Detailed quantitative analyses support this simple qualitative estimation
(Sugawara et al., 1997). The extent of the wave function, approximately
8 nm, is smaller than the dot size observed by plan-view TEM (Fig. 2.14(a))
and definitely larger than the height of the single SK dot of about 3 nm. This
indicates that the wave function of the ground-state exciton extends over
142 YOSHIAKINAKATA
AND YOSHIHIRO
SUGIYAMA
Stacked islands
(A = 1.088 (pm))
B
QW
(A = 0.936 (pm))
i;-
I I I I I
InGaAdGaAs QW 4.2 K
lnAslGaAs dot
*5.*meA
0.9 0.95 1.o 1.05 1.1 1.15

Wavelength (pm)
FIG.2.18. (a) Samples for magneto-optical measurements. Five-layer stacked InAs SK
islands with 3-nm GaAs intermediate layers and, for comparison, an In,~,,Ga,,,,As/GaAs
quantum-well sample grown on a GaAs substrate by MOCVD. The In,,,,Ga,,,,As well layer
was 8 nm thick and subjected to - 1.3% biaxial compressive strain. (b) Photoluminescence
spectra of the two samples with no magnetic-field applied (from Sugawara et al., 1997.
Copyright 1993 by The American Physical Society). (c) Diamagnetic shifts of emission spectra
for both samples under magnetic fields perpendicular and parallel to the sample plane at 4.2 K
(from Sugawara et al., 1997. Copyright 1993 by The American Physical Society).
GROWTH
BEAMEPITAXIAL
2 MOLECULAR 143
W
W
0 20 40 60
A80 100 120 140
Square of magnetic field ( T 2 )
FIG. 2.18. (c)
two or three SK dots out of five. Such exciton localization in the whole
stacked structure can be attributed to the composition and/or size in-
homogeneity of each SK dot components; that is, the potential minimum is
formed by the coupling of the adjacent two or three SK dots with low
band-edge energies.
IV. Columnar InAslGaAs Quantum Dots
The advantages of the stacking technique for device applications can be

summarized as follows. First, the spectrum width can be made much
narrower than in ordinary SK dots. Second, the size (height) and the
symmetry vary with the number of stacked layers, which makes it possible
to artificially control the energy separation between the discrete quantum
levels, the emission wavelength, the degeneracy of the quantum levels, the
overlap integral of the electron-hole wave functions that determines the
oscillator strength of optical transitions, and so forth.
Though we succeeded in narrowing the spectrum width by close stacking
of SK-mode islands and growth interruption (annealing) as explained in
Section 111, we found that the emission efficiency was greatly damaged. The
photoluminescence from the closely stacked islands rapidly weakened as the
temperature rose and was barely observed at room temperature. Among
144 YOSHIAKINAKATAAND YOSHIHIROSUCIYAMA
various growth conditions, annealing after the growth of the GaAs inter-
mediate layers (see the growth conditions in Fig. 2.10) was found to cause
the most serious degradation. A large number of defects and/or impurities
should have been introduced during annealing, working as nonradiative
recombination centers. On the contrary, though the emission efficiency was
fairly good even at room temperature without annealing, the spectrum
width was reduced to 50-60 meV at most, presumably due to fluctuations
in the island height and in the thickness of the spacer between the bottom
of the upper-layer islands and the top of the lower-layer islands (Endoh et
al., 1998).
To overcome this problem, we stacked SK mode islands using even
thinner intermediate layers -3 ML thick (less than 1 nm) -so that islands
would physically contact each other in the perpendicular direction. This
thickness is much less than the island height. As seen in Fig. 2.2, about 6 ML
of GaAs supply was needed to fully cover the islands.
The SK-mode islands could be grown even on such thin GaAs intermedi-
ate layers. Figure 2.19 shows the RHEED-pattern intensity transition during
the stacked island growth (the GaAs intermediate layer on the first InAs
island layers and the second InAs island layers). In this experiment, the first
island layer was grown with the InAs supply of 1.8 ML. The intermediate-
layer thickness was reduced to 3 ML from 10 ML, which is the same as that
used in the closely stacked structures in Section 111. Immediately after the
GaAs supply, the spot intensity decreased, showing that the growth surface
was becoming flat by the growth of the intermediate layers. The spot
intensity recovered to that of the first islands when the InAs supply started,
showing that the second island layers grew. It should be noted that even on
3- and 5-ML thick intermediate layers, which were much thinner than the
island height, second island layers grew as well as the first island layer.
However, the critical amount of lnAs supply for the transition from
two-dimensional layer growth to three-dimensional island growth decreased
as the intermediate layer thickness has reduced: 0.84ML for the 10-ML
intermediate layers, 0.76 ML for the 7-ML intermediate layers, 0.60 ML for
the 5-ML intermediate layers, and 0.26ML for the 3-ML intermediate
layers. These were much smaller than that for the first island layer, shown
at the bottom of Fig. 2.19.
Figure 2.20 shows the RHEED-pattern intensity transition when the InAs
islands were repeatedly stacked using the 3-ML thick intermediate layers.
The inset photographs are the AFM images of the topmost layer after 8 to
9 repetitions. The amount of InAs supply for the stacked-island formation
was changed from about 0.5 ML to 0.8 ML. When the stacked islands were
grown with an InAs supply of about 0.7ML, the RHEED intensity
oscillated with the InAs island growth (RHEED intensity increased) and
BEAMEPITAXIAL
2 MOLECULAR GROWTH 145
Growth time (s) 20 s I 1 div.

FIG.2.19. RHEED pattern intensity transition during the stacked island growth (GaAs
intermediate layer o n the first InAs islands and the second InAs island). The critical thickness
for the transition from two-dimensional growth t o island growth decreases as the intermediate
GaAs layer thickness reduces.
GaAs intermediate-layer growth (RHEED intensity decreased). This sug-

gests that the islands were formed successfully. The AFM image of the
topmost island layer grown with the 0.7-ML InAs supply shows that stacked
islands were formed with the same size and density as in the first islands.
On the other hand, when stacked with an InAs supply of about 0.5 ML and
0.8 ML, RHEED intensity oscillation stopped. When stacked with the InAs
supply of about 0.5 ML, RHEED intensity damped at the third repetition,
and the AFM image showed no island formation. This is possibly because
the InAs supply was not enough for island formation. When stacked with
an InAs supply of about 0.8 ML, RHEED intensity did not recover
completely at 8 to 9 repetitions and islands like the first one were not seen
in the AFM image. The supply of 0.8 ML is excessive. The optimum amount
of about 0.7 ML is smaller than in the case of ordinary SK-mode islands.
When stacked islands using such thin intermediate layers, the amount of
InAs supply has to be reduced and optimized precisely.
146 YOSHIAKINAKATA
a
YI
U
Growth time ( 5 ) 40 8 I 1 div.

FIG.2.20. RHEED pattern intensity transition when the InAs islands were repeatedly
grown with the intermediate layer of 3 ML. The inset photograph shows the AFM images of
the topmost layer. The above results suggest that there is an optimum lnAs supply amount.
Figure 2.21(a) is a cross-sectional TEM image of the structure shown

schematically in Fig. 2.21(b). InAs island layers formed with an InAs supply
of 0.7 M L and GaAs intermediate layers of 3 M L were grown alternately (8
repetitions) on the first island layer formed with an 1.8-ML InAs supply.
The stacked upper islands were grown on the lower-layer islands aligning
perpendicularly on the first-layer islands. The islands on each layer were in
contact with each other physically, and the stacked islands as a whole had
a columnar shape with a diameter and a height of about 17 nm and 13 nm,
respectively. All island layers were accompanied by wetting layers, seen as
horizontal line-shaped contrasts in the TEM image, indicating that the
stacked islands were formed by the SK mode.
Figure 2.21(a) is a low-magnification image. The columnar-shaped islands
were seen clearly in the quantum-well-like dark contrast region composed
of lnAs multiple wetting layers and GaAs intermediate layers.
In Chapter 3, we will introduce InGaAs/GaAs quantum dots with a light
emission wavelength of 1.3 pm grown by the alternate growth of elementary
In-Ga-As in MOCVD. The present structures look very similar to the ALS
dots. Although the large islands were formed, the contrasts corresponding
to dislocations and stacking faults, which observed at the coalesced large
islands, as shown in Fig. 2.6, were barely observed, suggesting that the
composition of lnAs was smaller than that in the ordinary SK islands.
2 MOLECULAR
BEAMEPITAXIAL
GROWTH 147
FIG.2.21. (a) Cross-sectional TEM image of columnar-shaped quantum dots. (b) Sche-
matic structure. (c) Low magnification image.
148 YOSHIAKI NAKATA
Figure 2.22 shows the photoluminescence spectrum of these columnar-

shaped quantum dots, measured at 300K. The peak wavelength of the
ground level was 1.17pm, and that of the excited level was 1.10pm. An
emission peak related to the multiple wetting layers was observed at
1.01 pm. An FWHM of the spectrum was about 42meV. This is much
smaller than that for the ordinary SK-mode quantum dots of about 80 meV,
indicating that structural uniformity was fairly improved. The photo-
luminescence intensity was also better than the SK-mode quantum dots and
was over 1000 times stronger than in the closely stacked quantum dots
described above, suggesting that the introduction of defects and impurities
working as the nonradiative recombination centers was remarkably sup-
pressed. The emission from the wetting layers red-shifted compared with the
ordinary SK mode quantum dots. This is because multiple wetting layers
were coupled electrically and formed a quantum level as a whole. The energy
separation between the ground state and the wetting layer was 168meV,
which is much smaller than 230meV of the ordinary SK-mode quantum
dots. An excellent performance of columnar-shaped quantum dots is dem-
onstrated in Chapter 6.
Figure 2.23 plots the photoluminescence wavelength at room temperature
and 7 7 K as a function of the stacked-layer number. As the layer number
0.9 1.o 1.1 1.2 1.3 1.4

Wavelength (pm)
FIG.2.22. Photoluminescence spectrum of the columnar-shaped quantum dots shown in
Fig. 2.21 measured at 300 K. The peak wavelength of the ground level was 1.17 pm and that
of the excited level was 1.10 pm. An emission peak related to the multiple wetting layers was
observed at 1.01 p m A full width at half maximum (FWHM) of the spectrum was about
42 meV.
2 MOLECULAR
BEAMEPITAXIAL
GROWTH 149
0.9’
t I ’ I I ”
i
0 10 20 30 40
Stacked-layer number
FIG.2.23. Photoluminescence wavelength at room temperature and 77 K as a function of
the stacked layer number. As the layer number increases, the emission wavelength shifts to
longer wavelength due to the size increase.
increased, the emission wavelength becomes longer due to the size increase.
It is surprising that the longest wavelength at room temperature was
1.24 ym at the 23rd stacking- very close to the practically applicable
1.3 pm.
Structural features of the stacked islands are summarized in Fig. 2.24.
Stacked-island structures strongly depend on the intermediate layer thick-
ness, t (nm). When t > 20 nm, stacked islands form without any correlation
to the lower-layer islands. The mean island size and density are constant in
all island layers, which is useful for increasing island density per unit
volume. When t < 20 nm, stacked upper islands form in correlation to the
lower-layer islands due to strain fields induced by the lower-layer islands.
The upper-layer islands form in alignment with the first-layer islands in the
perpendicular direction, and expand with an increase of the stacked-layer
number. As the intermediate-layer thickness decreases almost to the island
height, the size expansion of the stacked islands is suppressed, and almost
equal size islands are stacked closely. In this case, the electron wave
functions between neighboring islands in the perpendicular direction over-
lap each other, and the perpendicularIy stacked structure works effectively
as a single large quantum dot. Drastic improvement of the spectrum
linewidth is observed. When the intermediate-layer thickness further de-
150 NAKATAAND YOSHIHIROSUGIYAMA
YOSHIAKI
0 10 20 30
Intermediatelayer thickness, t (nm)
FIG. 2.24. Structural features of stacked islands. Stacked island structures strongly depend
on the intermediate-layer thickness, t (nm). When f z 20nm, stacked islands were formed
without any correlation to the lower-layer islands. When t < 20 nm, stacked upper islands were
formed in correlation to the lower-layer islands due to strain field induced by the lower-layer
islands.
creases to a few monolayers of less than 1 nm, which is much smaller than
the island height, stacked islands are formed. This success is achieved by
optimizing the amount of InAs supply for the stacked-island formation. If
we use 3-ML-thick intermediate layers, the amount of InAs supply has to
be reduced to about 0.7ML. The islands in the stacked structure are in
contact with each other physically in the perpendicular direction and the
stacked structure has a columnar shape as a whole. Even when we use these
thin intermediate layers, all stacked islands are accompanied by wetting
layers, indicating that the islands grow in the SK mode.
V. Summary
The structural features and optical properties of quantum dots presented

in this chapter are summarized in Fig. 2.25. Ordinary InAs SK-mode
quantum dots on GaAs substrates show broad spectra with a typical
FWHM of 80meV because of their large structural fluctuation. Since the
shape of ordinary SK-mode quantum dots is rather flat, their spectrum
broadening depends mainly on the fluctuation of the island height. The
stacking techniques to grow the closely stacked and columnar-shaped dots
2 MOLECULAR GROWTH
BEAMEPITAXIAL 151
SK islands Close1 stacked Columnar-shaped

isinds islands
Schematic
structure
PL 90 meV 25 meV 40 meV

FWHM
PL Good Poor Good

efficiency
In-plane 40% 40% 40%

coverage
(Density) > 1 x 10'1 cm-2 > 1 x 1011 cm-* >1 x loll cm-2
Wavelength Clm 1.2 pm > 1.2 pm

(300K)
Frc. 2.25. Structural and optical properties of different types of dots
are very useful in suppressing island height fluctuation because the effective
height can be controlled artificially by the number of stacked island layers.
The luminescence spectrum was remarkably improved from 80 meV to
25 meV by the close stacking method. In columnar-shaped quantum dots,
where both narrow spectrum width and high emission efficiency are ob-
tained, we have successfully achieved low-threshold and highly efficient
operation of quantum-dot lasers (Chapter 6). According to the simulation
of quantum-dot laser performance in Chapter 1, the narrow spectrum width
we have achieved is very promising for high-performance quantum-dot
lasers that are superior to strained quantum-well lasers. If we aim at
high-speed direct-modulation quantum-dot lasers, as predicted by Arakawa
and Sakaki (1982), we need to further reduce the spectrum width to 10 meV
(Chapter 1).
The in-plane size fluctuation of our columnar-shaped dots was evaluated
by AFM and found to give a broadening of about 20meV (Endoh et al.,
1998). For that reason, we are now concentrating on improving in-plane size
homogeneity.
Acknowledgments
The authors would like to thank Dr. Hiroshi Ishikawa, Dr. Hajime Shoji,
Dr. Kohki Mukai, Dr. Osamu Ueda, Dr. Akira Endoh, and Dr. Toshiro
Futatsugi, of Fujitsu Laboratories Ltd., and Professor Shunichi Muto of
152 YOSHIAKINAKATA
AND YOSHIHIRO
SUGIYAMA
Hokkaido University for their fruitful input and support. We also are
grateful to Dr. Hajime Ishikawa and Dr. Naoki Yokoyama of Fujitsu
Laboratories Ltd., for their encouragement and strong support throughout
this work.
REFERENCES
Arakawa, Y., and Sakaki, H. (1982). Appl. Phys. Lert. 40, 939.
Arita, M., Avramescu, A., Uesugi, K., Suemune, I., Numai, T., Machida, H., and Shimoyama,
N. (1997). Jpn. J . Appl. Phys. 36. 4097.
Bimberg, D., Ledentsov, N. N., Grundmann, M., Kirstaedter, N., Schmidt, 0.G., Mao, M. H.,
Ustinov, V. M., Egorov, A. Y., Zhukov, A. E., Kop’ev, P. S., Alferov, Z . I., Ruvimov, S. S.,
Gosele, U., and Heydenreich, J. (1996). J p n . J. Appl. Phys. 35, 1311.
Carlsson, N., Georgsson, K., Montelius. L.. Sarnuelson, L., Seifert, W., and Wallenberg, R.
(1995). J . Cryst. Growth 156, 23.
Endoh, A., Nakata, Y., Suguyama, Y., Takatsu, M., and Yokoyama, N. (1998). Pmc. 10th Inr.
Conf: InP and Related Materials.
Fafard, S., Leon, R., Leonard, D., Martz, J. L., and Petroff, P. M. (1994). Phys. Reu. B. 50,8086.
Fukui, T., and Saito, H. (1987). Appl. Phys. L m . 50, 824.
Gershoni, D., Ternkin, H., Dolan, G. J., Dunsmuir, J., Chu, S. N. G., and Panish, M. B. (1998).
Appl. Phys. Lett. 53, 995.
Goni, A. R., Pfeiffer, L. N., West, K. W.. Pinzuk, A., Baranger. H. U., and Storrner, H. L. (1992).
Grundrnann, M., Christen, J., Ledentsov. N. N., Bohrer, J., Bimberg, D., Ruvirnov, S. S.,
Werner, P., Richter, V., Gosele, V., Heydenreich, J., Ustinov, V. M., Egorov, A. Y., Zhukov,
A. E., Kopev, P. S., and Alferov, 2 . 1. (1995a). Phys. Rev. Lett. 20, 464.
Grundmann, N.. Ledentsov, N., Heitz, R.. Eckey, L., Christen, J., Bohrer, J., Bimberg, D.,
Ruvimov, S. S., Werner, P., Richter, V., Heydenreich, J., Ustinov, V. M., Egorov, A. Y . ,
Zhukov, A. E., Kopev, P. S., and Alferov, Z. 1. (1995b). Phys. Stat. Sol. 188, 249.
Glas, F., Guill, C., Henoc, P., and Haussy, F. (1987). Inst. Phys. Can$ Ser., No. 87; Section 2,
p. 87.
Grundmann, M., Ledentsov. N. N., Steir, 0..Bimberg, D., Ustinov, V. M. Kop’ev, P. S., and
Alferov, Z. I. (1996). Appl. Phys. Lert. 68. 979.
Hessman, D.. Castrillo, P., Pistol, M.-E.. Pryor. C., and Samuelsson, L. (1996). Appl. Phys. Lett.
69, 749.
Houzay, F., Guille, C., Moisson, J.-M., Henoc, P., and Barth, F. (1987). J . Cryst. Growth 81,67.
Ikoma, N., and Ohkouchi, S. (1995). J p n . J . Appl. Phys. 34, L724.
Jeppensen, S.. Miller, M. S., Hessrnan. D., Kowalski, B., Maximov, I., and Sarnuelson, L. (1996).
Kapon, E.. Kash, K., Olausen, E. M.. Hwang, D. M., and Colas, E. (1992). Appl. Phys. Lett.
60, 477.
Kapon, E., Simhony, S., Bhat, R., and Hwang. D. M. (1989). Appl. Phys. Lett. 55, 2715.
Kitamura, M., Nishioka, M., Oshinowo, J., and Arakawa, Y. (1995). A p p l . Phys. Lett. 66, 3663.
Konkar, A. Madhukar, A,, and Chen, P. (1998). Appl. Phys. Lett. 72, 220.
Ledentsov, N. N., Bohrer, J., Bimberg. D., Kochnev, 1. V., Maximov, M. V., Kop’ev, P. S.,
Alferov, 2 . I., Kosogov, A. O., Ruvimov. S. S., Werner, P., and Gosele, U. (1996). Appl.
Phys. Lett. 69, 1095.
2 MOLECULAR
BEAMEPITAXIAL
GROWTH 153
Leon, R., Petroff, P. M., Leonard, D., and Fafard, S. (1995). Science 267, 1966.
Leonard, D., Krishnamurthy, M., Reaves, C. M., Denbaars, S. P., and Petroff, P. M. (1993).
Leonard, D., Pong, K., and Petroff, P. M. (1994). Phys. Rev. B. 50, 11687.
Lewis, B. F., Lee, T. C., Grunthaner. F. J.. Madhukar, A,, Fernandez, R., and Masanian, F. F.
(1984). J . Vac. Sci. Techno/. B2, 419.
Marzin, J.-Y., Gerard, J.-M., Izrael, A,, and Barrier, D. (1994). Phys. Rec. Lett. 73, 716.
Marzin, J.-Y., Gerard. J.-M., Israel, A., Barrier, D., and Bastard, G. (1994). Phys. Rer. Lett. 73,
716.
Miyamoto, Y., Cao, M., Shingai, Y., Furuya, K., Suematsu, Y., Ravikumar, K. G., and Arai. S.
(1987). Jpn. J . Appl. Phys. 26, L225.
Moison, J. M., Houzay, F., Barthe, F., Leprince, L.. Andre, E., and Vatel, 0. (1994). A p p l . Phys.
Lett. 64, 196.
Mui, D. S., Leonard, D., Coldren. L. A,. and Petroff, P. M. (1995). Appl. Phys. Lett. 66. 1620.
Mukai, K., Ohtsuka, N., Sugawara, M., and Yamazaki, S. (1994). J p n . J . Appl. Phys. 33. L1710.
Muto, S. (1995). Jpn. J . Appl. Ptiyr. 34. L210.
Nakata, Y.. Sugiyama, Y., Futatsugi. T,. and Yokoyama, N. (1997). J . Cryst. Growth 175/176,
173.
Nishi, K., Anan, T., Gomyo, A,. Kohmoto, S.,and Sugou, S. (1997). Appl. Phys. Lett. 70. 3579.
Oshinowo, J., Nishioka, M., Ishida, S.. and Arakawa, Y. (1994). Appl. Phys. Lett. 65, 1421.
Petroff, P. M., Gossard, A. C., Logan, R. A., and Wiedmann, W. (1982). Appl. Phys. Lett. 41,
635.
Petroff, P. M., Gossard, A. C., and Wiegmann, W. (1984). Appl. Phys. Lett. 45. 620.
Pfeiffer, L. N., Stormer, H. L., Baldwin. K. W., West, K. W., Goni, A. R., Pinczuk, A. Ashoori,
R. C., Dignarn, M. M., and Wegscheider, W. (1993). J . Cryst. Growrh 127. 849.
Reaves, C. M., Bressler-Hill, V., Varma, S.. Weinberg, W. H., and DenBaars, S. P. (1995). Surf
Sci. 326. 209.
Reed, M. A., Bate, R. T., Bradshow. K.. Duncan, W. M., Frenseley, W. R., Lee, J. W., and Shih,
H. D. (1986). J . Vuc. Sci. Echnol. B4. 358.
Saitoh, T., Taniniura, A., and Yoh. K. (1996). J p n . J . Appl. Phys. 35, 1370.
Schaffer. W. J., Lind, M. D., Kowalczyk, S. P., and Grant, R. W. (1983). 1. Vac. Sci. Trchnot.
B1, 688.
Seifert, W., Carlsson, N., Peterson, A.. Wernersson, L.-E., and Samuelson, L. (1996). Appl. Phjs.
Lett. 68, 1684.
Shchukin, V. A,, Ledentsov, N. N., Kop’ev, P. S., and Bimberg, D. (1995). Appl. Phys. Lett. 75,
2968.
Solomon, G. S., Komarov, S. K.. Harris. J. S., Yamamoto, Y. (1997). J . Cryst. G r o w h 175/176.
707.
Solomon, G. S., Trezza, J. A., Marshall, A. F., and Harris, J. S. (1996). Pkjs. Reu. Lett. 76. 952.
Stranski, 1. N., and Krastanow, L. (1937). Aktid. Wiss. Wien Math.-Natur. ilh 146. 797.
Tackeuchi, A., Nakata, Y.. Muto, S., Sugiyama, Y., Inata, T., and Yokoyama. N. (1995). J p n .
J . Appl. Phys. 34, L405.
Tabuchi, M., Noda, S., and Sasaki, A. (1992). Sci. & Tech. Mesoscopic Structures 379.
Tanaka, S.. Hirayama, H., Ramvall, P.. and Aoyagi, Y. (1997). Piocec~cli,zgs of the 2nd
Inrernationul Corzference on Nitride Seniiconductors, 48.
Tanaka, S., Iwai, S., and Aoyagi, Y . (1996). Appl. Plzys. Lett. 69, 4096.
Tersoff, J., Tieichert, C., and Lagally, M. G. (1996). Phys. Rev. Lett. 76, 1675.
Tsui, R., Zhang, R., Shiralagi, K., and Golonkin. H. (1997). Appl. Phys. Lett. 71, 3254.
Tu,, K.-N., Mayer, 3. W., and Feldnian. C. (1992). “Electronic Thin Film Science,” Chapter 7.
Macmillan Publishing Company.
154 YOSHIAKINAKATA
Ueda, O., Nakata, Y., and Muto, S. (1998). Prcir. 10th Int. ConJ InP and Related Muterial.?, p.
113.
Wegscheider, W., Pfeiffer, L. N., Pinczuk. A., West, K. W., Dignam, M. M., Hull, R., and
Leibenguth, R. E. (1995). f.Cryst. Growth 150, 285.
Xie, Q., Madhukar, A,, Chen, P., and Kobayashi, N. P. (1995). Phys. Rev. Lett. 75, 2542.
SEMICONDUCTORS AND StMIMETALS. VOL 60
CHAPTER 3
Metalorganic Vapor Phase Epitaxial Growth of

Self-Assembled InGaAs/GaAs Quantum Dots
Emitting at 1.3 ,urn
Kohki Mukai, Mitsuru Sugawara, and Mitsuru Egawa
OPTICAL SEMICONVVCTOR D E V I C H LAB<>RATORY
FUJITSU
LABORATORIFS
LTD.
ATSUGI, JAPAN
KANAGAWA.
Nobuyuki Ohtsuku
[NTGRATtU MATERIALS
LABVRATORY
LTU.
FUJITSULABORATORIES
A T ~ ~ J KANAGAWA.
G~. JAPAN
1. VTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . I
11. ATOMICLAYEREPITAXIAL GROWTH. . . . . . . . . . . . . . . . . . . 157

111. ALTERNATE SUPPLY GROWTH OF InGaAs DOTS BY In-As-Ga-As SEQUENCE . . 160
IV. ALTERNATE SUPPLY GROWTH OF InGaAs DOTSBY THE In-Ga-As SEQUENCE . . 166
1. Tiiw Types of ALS Do/ . . . . . . . . . . . . . . . . . . . . . . . 168
2. Multiple-Loyer Growfh . . . . . . . . . . . . . . . . . . . . . , , 172
V . THEGROWTH PROCESS. . . . . . . . . . . . . . . . . . . . . . . . 176
VI. S U M M A R Y. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
I. Introduction
The molecular beam epitaxial (MBE) growth of InAs self-assembled

quantum dots was described in Chapter 2. This growth process occurs in
the well-known Stranski-Krastanov (SK) mode, which produces three-
dimensional islands accompanied by a thin wetting layer to release the
strain energy accumulated by the large lattice mismatch between the GaAs
substrate and the InAs. Typical InAs self-assembled islands have a flat shape
with a base length of about 20nm and a height of a few nm. Since the
exciton Bohr radius is 10 to 20nm in the InAs-GaAs system, the island is
155
Copyright i :1999 by Academic Press
All rights of reproductioii in any form reserved.
ISBN 0-12-752169-0
ISSN 0080-8784 99 $3000
156 KOHKIMUKAI,
MITSURU
SUGAWARA, AND MITSURU
EGAWA
small enough to observe the three-dimensional quantum confinement effect;

that is, the islands work as quantum dots. The perpendicularly stacked InAs
islands, named closely stacked quantum dots and columnar quantum dots in
Chapter 2, have enabled us not only to tune the emission wavelength by
increasing the dot size primarily in the perpendicular direction, but also to
narrow the spectrum linewidth by reducing the fluctuation of island size. In
particular, columnar quantum dots have had great success in achieving
low-threshold and highly efficient quantum-dot lasers, as described in
Chapter 6 and late news, due to their high crystal quality.
A new type of self-assembled quantum dots, grown by metalorganic vapor
phase epitaxy (MOVPE), is introduced in this chapter (Mukai et al., 1994;
Ohtsuka et al., 1995). Their growth sequence is unique in that the group-111
and group-V precursors are supplied alternately with an amount corre-
sponding to one or less than one monolayer. One of the most striking
features of our original quantum dots is that they emit at the wavelength of
1.3 pm at room temperature, which is the zero-dispersive wavelength of the
silica optical fiber used in optical data transmission systems. Considering
that the emission wavelength of ordinary SK dots is typically 1.1-1.2pm,
the longer emission wavelength suggests larger incorporation of In atoms
into quantum dots and/or larger dot size. In addition, the emission spectrum
linewidth is very narrow and gives a series of distinct lines peculiar to
three-dimensional quantum confinement, as seen in Fig. 4.1 l(a) in Chapter
4. The minimum full width at half maximum (FWHM) is 28meV-much
smaller than the 80-120meV in the SK-growth islands. These two fea-
tures- 1.3-pm emission and narrow spectrum linewidth -make quantum
dots very attractive for practical applications. If we succeed in 1.3-pm lasing
from quantum dots, strained quantum-well lasers -currently the standard
device in optical transmission systems -might be replaced by high-perform-
ance quantum-dot lasers. Vertical-cavity surface-emitting lasers (VCSELs)
with 1.3-pm quantum-dot active regions are also attractive applications,
since high-reflectivity distributed Bragg reflector mirrors can be fabricated
on GaAs substrates.
We named our original materials ALS quuntum dots, for Alternate Supply.
We found these new quantum dots when we tried to fabricate (GaAs),/
(InAs), short-period superlattices on GaAs substrates by atomic layer
epitaxy (ALE). The purpose of our research was to realize materials that
emit at 1.3pm on GaAs substrates using the ALE technique, with the
expectation that lasers on GaAs substrates would have high temperature
stability since high-potential barriers like AlGaAs and AlGaInP can prevent
carrier leakage from the active region. Simply growing InGaAs quantum
wells on GaAs substrates does not work because the large lattice mismatch
causes misfit dislocations, severely damaging the crystal quality. Our belief
3 METALORGANIC
VAPOR PHASE
EPITAXIAL
GROWTH 157
was that short-period superlattices can break through the limit set by the
critical thickness to reach a highly efficient 1.3-pm emission. Actually, Roan
and Cheng (1991) had already succeeded in growing (GaAs),/(InAs),
short-period superlattices on GaAs substrates by MBE that showed 1.3-pm
emission at room temperature. Using the ALE approach, we also succeeded
in realizing 1.3-pm emission and in making the emission spectrum linewidth
much narrower (Ohtsuka and Ueda, 1994). However, we found that the
grown materials were not short-period superlattices but had bizarre struc-
tures, as seen in the transmission electron microscope (TEM) photograph of
Fig. l.l(b) in Chapter 1, where spherical dark regions are buried in the
quantum wells. Having done various diagnostics of the materials (see
Chapter 4), we finally concluded that the structures are quantum dots.
This chapter reviews ALS quantum-dot growth. We start by describing
the self-limiting one-monolayer growth of InAs and GaAs and the growth
of (InAs),/(GaAs), superlattices. Then we present two kinds of growth
sequences for the quantum dots: one, an In-As-Ga-As sequence based on the
concept of making monolayer superlattices by ALE; the other, a newly
developed In-Ga-As sequence. Quantum dots are characterized by their
shape, size, composition, and emission wavelength. In particular, we illus-
trate how the dots vary with alternate supply cycle, growth temperature, and
composition of the buffer layers on which the dots are grown. Finally, we
compare the growth process of ALS quantum dots with that of SK dots.
11. Atomic Layer Epitaxial Growth
As stated in the introduction, our original purpose was to grow short-

period superlattices using the ALE technique. Because we found that the
grown materials were quantum dots, we tried to refine various growth
conditions to improve crystal quality. As we pursued dots with higher
emission efficiency, we changed the growth sequence from In-As-Ga-As,
according to the concept of superlattice atomic layer epitaxy, to In-Ga-As,
and we reduced the supply amount of sources to less than one monolayer.
However, the alternate supply similar to ALE growth is definitely a key to
producing this type of quantum dot. Another inheritance from ALE growth
is trimethylindium-dimethylethylamineadduct (TMIDMEA), which we
developed as a source material for indium. Here, we will briefly describe the
self-limiting growth of GaAs, InAs, and short-period superlattices.
For ALE growth, we used pulse jet epitaxy (PJE) with a chimney reactor,
in which precursors are supplied in a fast pulsed stream, and self-limited
111-V compounds can be grown under a wide range of conditions (Ozeki et
158 AND MITSURU
KOHKIMUKAI,MITSURUSUGAWARA, EGAWA
al., 1988, 1989, 1991; Sakuma et al., 1988). The structure of the growth
system is shown in Fig. 3.1. The reactor was designed through a simulation
to realize smooth gas flow streamlines without vortices. We adopted a
low-pressure atmosphere for precursors to decompose only on the heated
substrate surface. The substrate was positioned in the fast gas stream
emitted from a jet nozzle to prevent thermally decomposed source molecules
from building up in stagnant layers. Source materials were trimethylindium
(TMIn), TMIDMEA, trimethylgallium (TMGa), triethylgallium (TEGa),
and arsine (ASH,) diluted with hydrogen (HJ. Experiments on self-limiting
ALE growth were done on GaAs substrates for GaAs growth and on InAs
substrates for InAs growth.
The most severe problem we encountered in fabricating InAs/GaAs
superlattices using ALE was that the temperature range for self-limiting
growth was different for the two binary compounds-450 to 550°C for
GaAs using TMGa and 300 to 400’C for InAs using TMIn. Since there is
no overlap in ALE growth temperatures for the two materials, it appeared
difficult to fabricate the superlattices under precise self-limiting growth
(Mori and Usui, 1992). A key factor in determining the temperature range
for self-limiting growth is the decomposition temperature of the group-I11
precursors. We expected that the thermal stability of TMIn, which is a
Substrate [InAs (IOO)]
FIG. 3.1. Growth system for pulse-jet epitaxy with a chimney reactor, in which precursors
are supplied in a fast, pulsed stream and self-limited 111-V compounds can be grown under a
wide range of conditions.
3 METALORGANIC
VAPOR PHASE GROWTH
EPITAXIAL 159
conventional In source, might be sustained at higher temperatures by the

addition of some adduct, and so developed TMIDMEA as a new In source
for ALE.
The ALE growth of InAs using TMJDMEA and TMIn for indium
sources is shown in Fig. 3.2, where the relationship between the group-I11
pulse duration and the growth thickness per cycle is plotted at 460°C. When
we used TMIn, the thickness per cycle increased monotonously with the
TMIn pulse duration, because TMIn decomposes in the boundary layer at
this high temperature and a self-limiting mechanism does not work. When
we used TMIDMEA, complete self-limiting growth could be achieved -the
growth thickness per cycle saturated at one monolayer of InAs.
The temperature dependence of the growth rate is shown in Fig. 3.3. The
self-limiting growth of InAs using TMIDMEA was observed over a tem-
perature range of 350 to 500°C. The upper limit of this temperature region
is 100°C higher than that using TMIn. As a result, we can grow InAs and
GaAs with a self-limiting mechanism in the common temperature region,
which makes it possible to grow InAs/GaAs heterostructures at a constant
growth temperature.
ALE-In As I
I
Tg = 460;C
0
2 4 6 8
Group-I11 pulse duration (s)

FIG. 3.2. InAs thickness per cycle of ALE growth a s a function of indium-source pulse
duration at 460°C. When T M l n is used, the thickness per cycle increases with the TMIn pulse
duration. When TMIDMEA is used, complete self-limiting growth can be achieved; that is, the
growth thickness per cycle is saturated at one monolayer of InAs (from Ohtsuka and Ueda,
19941.
160 KOHKIMUKAI,
MITSURU
SUGAWARA, A N D MITSURUEGAWA
300 400 500 600
Growth temperature (“C)

FIG.3.3. InAs and GaAs thickness per cycle of ALE growth as a function of growth
temperature. We can grow InAs and GaAs with a self-limiting mechanism in the common
temperature region. This makes it possible to grow InAs/GaAs heterostructures at a constant
growth temperature (from Ohtsuka and Ueda. 1994).
Using this technique, we grew (InAs),/(GaAs), superlattices on a GaAs

substrates. The grown sample consisted of a 500-nm-thick undoped GaAs
buffer layer, a 12-period (InAs),/(GaAs), layer with a total thickness of
7nm, and a 30-nm thick undoped GaAs cap layer. Both buffer and cap
layers were grown by a conventional MOVPE at 460°C using TEGa. The
room-temperature photoluminescence spectrum of the sample grown using
TMIDMEA showed a peak at 1.34pm with an FWHM of 28 meV (Fig. 3.4).
The emission intensity was much stronger than the sample grown using
TMIn. The transmission microscope (TEM) photograph of the “superlat-
tice” surprised us. We discovered quantum dots in the alternately supplied
layer, as shown below.
111. Alternate Supply Growth of InGaAs Dots by In-As-Ga-As Sequence
In this section, we illustrate the growth of ALS dots by In-As-Ga-As

alternate supply using the same reactor for PJE and the same sources as
VAPOR PHASE
3 METALORGANIC GROWTH
EPITAXIAL 161
I-
I
v1
c,
.3 TMIDMEA
c
s = I TMIn
h
c,
.M
- FWHM
8c
w
= 28 meV
.&
1 .o 1.2 1.4 1.6

Wavelength (pm)
FIG. 3.4. Room-temperature photoluminescence spectra of (InAs),/(GaAs), superlattices
on (100) GaAs substrates. The room-temperature photoluminescence spectrum of the sample
grown using TMIDMEA shows a peak at 1.343pm with a full width at half maximum
(FWHM) of 28 meV. The emission intensity is much stronger than that of the sample grown
using TMIn (from Ohtsuka and Ueda, 1994).
used in Section 11. Each supply is separated by H, purging pulses. The dots
were grown on an In,Ga, -,As buffer layer on a (001)-GaAs substrate and
were covered by an In,Ga, -,As layer having the same composition as the
buffer layer. The supply amount of TMIDMEA and TMGa was set at less
than one monolayer to effect higher emission efficiency.
Figure 3 4 a ) and (b) show typical plan-view and cross-sectional TEM
images of ALS dots, grown by 12 cycles of (InAs),/(GaAs), periodic supply
on a 300-nm GaAs buffer layer on a (001)-GaAs substrate, covered by a
30-nm-thick GaAs cap layer. The growth temperature was 460°C which
provides distinct self-limiting growth for both InAs and GaAs (see Fig. 3.3).
The growth pressure was 15 Torr. A plan-view image shows uniform
dot-like microstructures about 20 nm in diameter and an area coverage of
5-10%. The dots diameters follow a Gaussian distribution with a standard
deviation of 2.9 nrn. A cross-sectional dark-field image indicates that sphere-
shaped dots were self-assembled within the InAs/GaAs short-period layer
(dot layer) sandwiched by GaAs. They were about 10nm in height and
surrounded laterally by a quantum-well layer having almost the same
thickness as that of the dots. The thickness of the quantum-well layer was
definitely greater than that of the so-called wetting layer of the SK islands
(Krost et al., 1996; Ramachandran et al., 1997a, and b. Kitamura et al.,
162 KOHKIMUKAI.MITSURUSUGAWARA, EGAWA
AND MITSURU
FIG.3.5. (a) Plan-view and (b) cross-sectional TEM images of ALS dots, grown by 12
cycles of (InAs),/(GaAs), periodic supply on a 250-nm GaAs buffer layer on a (001) GaAs
substrate and covered by a 30-nm-thick GaAs cap layer.
1997). We observed two clear upper and lower borders between the GaAs
sandwiching layers and the dot layer. We also observed a high-resolution
TEM image around the dot in a cross-sectional sample and did not find any
dislocations or defects in the lattice image (Fig. 3.6). A weak contrast at the
border of the dots suggests that the composition and lattice distortion do
not abruptly change around the dot.
Figure 3.7 shows the spatial composition distribution of a sample cross
section evaluated by energy-dispersive X-ray microanalysis (EDX) with an
excitation electron beam focused to a diameter of about 1 nm. The sample
was thinned down to lOnm, which is less than the in-plane diameter of the
dots, to obtain a signal only from the dots. Signal intensity was calibrated
to estimate a precise indium composition using the signal intensity of a
lattice-matched In,,,,Ga,~,,As layer grown on an InP substrate as a
reference. The indium composition was 0.5 at the center of the dots (i.e.,
GROWTH
VAPORPHASEEPITAXIAL
3 METALORGANIC 163
FIG. 3.6. High-resolution T E M image around the dot in a cross-sectional sample. No

dislocation or defects are found in the lattice image.
Ino~,Gao,,As)and 0.1 in the quantum-well layer surrounding the dots (i.e.,

Ino.,Gao.!Js).
Our self-assembled quantum dots were found to have a unique structure,
different from that of SK islands. For one thing, ALS dots are almost buried
in the quantum-well layer with flat upper and lower interfaces, and they are
not accompanied by a so-called wetting layer. For another, the proportion
of height and diameter is different from that of the SK island; that is, the
height of the SK dots is generally less than one-fourth of the in-plane
0.27
0.53(dot)
0 0.5 1.o
Indium composition
FIG.3.7. Spatial indium composition distribution of quantum-dot cross sections evaluated
by energy dispersive X-ray microanalysis (EDX) with an excitation electron beam focused to
a diameter of about 1 nm (from Mukai et al., 1994).
164 KOHKIMUKAI,
MITSURU
SUGAWARA,
AND MITSURU
EGAWA
diameter, while the height of ALS dots is about half of the in-plane diameter.
These characteristics suggest that our dots were not self-assembled by the
conventional SK mode.
The outstanding characteristic of our quantum dots is a wide-range
wavelength tunability of between 1.2 and 1.5 pm via 1.3 pm, which is
achieved by changing the cycle of short-period growth and the indium
composition of the In,Ga,-xAs buffer layer. As shown in Fig. 3.4, emission
from the ground state of the ALS dots was observed at 1.34pm, and that
from the excited state at 1.24 pm. The FWHM of 28 meV shows a high
degree of uniformity in terms of dot size, strain, and composition.
Figure 3.8 shows (a) the photoluminescence wavelength and (b) the
in-plane diameter of quantum dots evaluated by TEM photography as a
function of the number of cycles. The buffer layer was In,,,,Ga,~,,As. By
increasing the alternate growth cycle number from 7 to 24, the emission
peak wavelength was increased from 1.29 to 1.41 pm. Figure 3.9 shows the
photoluminescence wavelength as a function of the buffer-layer indium
composition. By increasing the In composition of the buffer layer from 0
(GaAs) to 0.09 (Ino,09Gao,91As), the emission peak wavelength was in-
creased from 1.3 to 1.46 pm at the 12 cycles of InAs/GaAs. Since there was
no distinct differences among the In composition in dots measured by EDX,
we know that the wavelength variation is due to the change in dot size, as
seen in Fig. 3.8(b). Combining the above results concerning the number of
supply cycles and the indium composition of the buffer layer, we can control
the emission wavelength continuously from 1.2 to 1.5 pm.
In order to improve the quantum-dot crystal quality, we optimized
various growth conditions such as temperature, sequence, and source supply
amount, by checking the photoluminescence emission efficiency. Figure 3.10
shows the photoluminescence intensity versus growth temperature for dots
grown at 12 cycles of (InAs),/(GaAs),. The optimized growth temperature
was between 440 and 460"C, above and below which the emission intensity
rapidly decreased. We also found that the source supply amount should be
less than one monolayer. In addition, we adopted a new sequence of
In-Ga-As, which provided an emission efficiency more than two times higher
than that of the In-As-Ga-As sequence.
Under these optimized growth conditions, we made cavity lasers and
succeeded in lasing at 80 K, though the emission was from the higher-order
excited states (see Chapter 6). We could not achieve lasing from the ground
state, because of insufficient optical gain for lasing due to the low area
density of the ALS dots. Since the PJE system was not equipped with a gas
supply sufficient to fabricate laser devices, we were obliged to grow parts of
the laser structure other than the quantum-dot active region using other
growth systems. We believe that this process severely damaged the crystal,
3 METALORGANIC
VAPORPHASEEPITAXIAL
GROWTH 165
1.25
5 10 15 20 25
Number of InAdGaAs cycles
Number of InAs/GaAs cycles

FIG. 3.8. (a) Photoluminescence wavelength at 300 K and (b) in-plane diameter of quantum
dots evaluated by a TEM photograph as a function of the number of cycles. The wavelength
variation is due to changes in dot size. ((a) Reprinted from Appl. Surface Science, vol. 112,
Mukai et a!., "Growth and optical evaluation of InGaAs/GaAs quantum dots self-formed
during alternate supply of precursors," pp. 102-109, 1997, with kind permission of Elsevier
Science-NL, Sara Burgerhartstraat 25, 1055 KV Amsterdam, The Netherlands; (b) from Mukai
et al., 1996.)
166 KOHKIMUKAI,MITSURUSUGAWARA,
AND MITSURUEGAWA
0 0.05 0.1
Indium composition of buffer layer

FIG.3.9. Photoluminescence wavelength at 300 K as a function of the indium composition
of the In,Ga,-,As buffer layer. When supplying 12 cycles of alternate supply, the emission
wavelength increased from 1.3 to 1.46pm as the indium composition increased from 0 (it.,
GaAs) to 0.09. By evaluating the reciprocal lattice using X-ray diffraction, it was found that an
increase in the indium composition of the buffer layer corresponded to an increase in strain
relaxation. (Reprinted from Appl. Sur/ucr Science, vol. 112, Mukai et al., “Growth and optical
evaluation of InGaAs/GaAs quantum dots self-formed during alternate supply of precursors,”
pp. 102- 109, 1997, with kind permission of Elsevier Science-NL, Sara Burgerhartstraat 25, 1055
KV Amsterdam, The Netherlands.)
resulting in lasing capability only at low temperature. Unfortunately, we

were unable to improve on this, since our research on ALS quantum dots
was suspended for almost two years until we acquired another next reactor.
IV. Alternate Supply Growth of InGaAs Dots by In-Ga-As Sequence
Supply growth thereafter was performed in a state-of-the-art MOVPE

system, a standard type designed by Fujitsu Laboratories and used to
produce commercial optical semiconductor devices. The reactor has a
vertical configuration with a wide-area inlet consisting of multiple gas
injectors (Kondo et al., 1992). The injectors are arranged in a threefold
honeycomb configuration so that the inlet gas impinges perpendicularly in
the entire area of the substrate. Organometallics, hydrides, and dopants
were mixed with the main hydrogen carrier gas in a compact gas-switching
3 METALORGANIC
VAPORPHASE
EPITAXIAL
GROWTH 167
380 420 460 500

Growth temperature (“C)
FIG. 3.10. Photoluminescence intensity of ALS dots at 300K as a function of growth
temperature. The ALS dots were grown by 12 cycles of (InAs),/(GaAs), alternate supply. Note
that this figure is a semi-log plot. (Reprinted from Appl. Surfuee Science, vol. 112, Mukai et al.,
“Growth and optical evaluation of InGaAs/GaAs quantum dots self-formed during alternate
supply of precursors,” pp. 102-109, 1997, with kind permission of Elsevier Science-NL, Sara
Burgerhartstraat 25, 1055 KV Amsterdam, The Netherlands.)
manifold, then divided into subflows and introduced to the injectors. For
the gas-handling system, we used a fast switching manifold with a pressure-
balanced vent-and-run configuration.
Self-limiting growth was observed for InAs and GaAs at common
temperatures using TMIDMEA and TMGa. Note that the growth tempera-
ture indicated in this section does not agree with the temperature of the PJE
reactor in Sections I1 and 111. Comparing temperatures for self-limiting
growth in each reactor, a growth temperature of 460°C for the PJE reactor
in Sections I1 and 111 corresponds to 520°C for the MOVPE reactor in this
section.
The most striking change from Sections I1 and 111 was that we adopted
the growth sequence In-Ga-As from the beginning. Optimized source supply
amounts were 2.0 seconds for TMIDMEA and 0.3 seconds for TMGa,
which correspond to 0.5-monolayer In and 0.1-monolayer Ga. The duration
of the ASH, supply was fixed at 7 seconds. The H, purge was inserted after
the TMGa and ASH, supply for 3 seconds and 0.5 seconds, respectively. The
supply rates of TMIDMEA, TMGa, and ASH, sources were 0.04, 4.6, and
40 cc per minute, respectively. A single cycle time of 12.8 seconds provided
a very low growth rate of about 0.04 ML per second.
168 KOHKIMUKAI, SUGAWARA,
MITSURU AND MITSURUEGAWA
In this section, we further explore the growth conditions and processes of

ALS quantum dots. We first show that there are two types of ALS quantum
dots, depending on the growth temperature and the alternate supply cycle.
We then show the multiple-layer growth of quantum dots.
1. Two TYPES
OF ALS DOT
The growth procedure was as follows. A 0.25-pm GaAs buffer layer was
grown on a (100)-GaAs substrate at 630°C. The growth was then interrup-
ted for 10 minutes and the substrate temperature was lowered under an
ASH, atmosphere. Then, after an additional 0.05-pm GaAs buffer layer
growth, the In-Ga-As layer was grown by the alternate source supply,
followed by the 30-nm GaAs cap layer. The GaAs buffer and cap layers were
grown by conventional MOVPE using TEGa and ASH,. The growth
pressure was 15 Torr throughout the structure.
Figure 3.1 l(a) shows the wavelength of a photoluminescence spectrum
peak as a function of the growth temperature for the 18-cycle In-Ga-As
supply. The emission wavelength varied considerably between 1.2 and
1.3pm depending on the growth temperature, with a 1.3-pm emission at
around 500°C. Figure 3.11(b) shows the peak wavelength of the photo-
luminescence spectra as a function of the cycle number at various growth
temperatures. The relationship between the cycle number and the emission
wavelength also varied considerably between the two temperature ranges,
that is, the dots grown at 435 and 460°C showed an almost constant
emission wavelength, while the emission wavelength of dots grown at
490-510°C increased as the cycle number increased. This suggests that there
are two types of ALS dots grown according to different growth processes.
The dots with an emission wavelength of 1.2pm are similar to SK dots in
that the emission wavelength is independent of the source supply amount.
Chapter 2 states that the diameter of SK dots saturates above a critical InAs
layer thickness and that the dots are controlled by growth temperature only.
Figure 3.12 shows how the photoluminescence spectra vary with cycle
number and growth temperature. They were measured at 300 K using a Krf
laser with an excitation power density of 60 W/cm2. Using both the peak
wavelength and intensity as standards, we classified these photolumines-
cence spectra into five types. With a small cycle number, a very weak and
broad emission was observed, which can be attributed to an emission from
the two-dimensional layer prior to three-dimensional nucleation. No dots
were observed in the TEM photograph. As the cycle number increased,
single peaked and intense emissions appeared at around 1.2 pm (we call this
emission type A), followed by double-peaked, intense emissions at around
VAPORPHASE
3 METALORGANIC EPITAXIAL
GROWTH 169
E
1
W
I-!
a
Growth temperature ("C)
Cycle number
FIG. 3.11. (a) Wavelength of a photoluminescence spectrum peak as a function of growth
temperature for the 18 cycles of In-Ga-As sequence. The emission wavelength varied consider-
ably between 1.2 and 1.3pm, depending on the growth temperature, with a 1.3-pm emission at
around 500°C. (b) Peak wavelength of photoluminescence spectra as a function of the In-Ga-As
cycle number grown at various temperatures. The relationship between the cycle number and
the emission wavelength varies considerably between the two temperature ranges; that is, the
dots grown at 435°C and 460°C showed an almost constant emission wavelength, while the
emission wavelength of dots grown at 490-510°C increased as the cycle number increased.
170 SUGAWARA,
KOHKIMUKAI,MITSKJRU AND MITSURU
EGAWA
J - - - f c . os g Jp- r/ 0.9 1 5 0.9
rn
0.9 12 I5 0.9 11 15 0.9 I2 15 0.9 12 15 1.2 12 15
Mo*cfIl -
LJ LJLid
...f$$jq 0.9
b/ 12 15 0.9 12 15 0 9 13 I5 0.9 12 IS 0.9 12 15 0.9 12 15
0.9 12 I5 0.9 1 2 I5 0.9 12 15 0.9 12 1 5 0.9 12 I5 0.9 1 2 I5
No*ck-lt@ Lid u
-
0.9 12 15
K/K/~1 a~1
0.9 1.2 1.5 0.9 1.2 15 0.9 12 15 0.9 1.2 1 5 0.9 1.2 15
4 9 -e0 . c ~ ~ /
0.9 1.2 1.5 0.9 12 15 0.0 12 15 0.9 12 1.5 0.9 1.2 1 5 0.9 12 15
N = 12 14 18 18 20 22
FIG.3.12. Room-temperature photoluminescence matrix of ALS InGaAs quantum dots as

a function of cycle number and growth temperature. There are five types of photoluminescence
spectra characterized by peak wavelength and intensity: wetting layer (negligible emission), type
A (broad 1.2-pm emission), type B (sharp 1.3-pn emission), type A (weak 1.2-pm emission),
and type B' (weak 1.3-pm emission).
1.3 pm (type B). At temperatures above 560"C, a single peaked emission at

1.2 pm weakened (type A'). A double-peaked, intense 1.3-pm emission
appeared in the narrow temperature region of 500-550°C. Emissions
weakened above 560°C and below 490°C (type B').
The TEM observation revealed the existence of quantum dots in the type
A, A', B, and B' regions. Hereafter, the quantum dots characterized by the
3 METALORGANIC GROWTH
VAPORPHASEEPITAXIAL 171
type A and type B photoluminescence spectra are called type A and type B
quantum dots respectively.
The photoluminescence matrix of Fig. 3.12 tells us that source supply
amount and growth temperature govern the formation of ALS quantum
dots. For a growth temperature of 5O0-55O0C, type-A quantum dots with a
density of about 6 x 10'' c m p 2 (determined from plan-view TEM) h i -
tially form after two-dimensional growth. The subsequent source supply
transforms type-A quantum dots to type-B quantum dots with an identical
density. The shorter wavelength peak in the type-B photoluminescence
spectra is the second sublevel emission of type-B quantum dots because, as
the excitation laser power increases, several peaks appear in the higher
energy regime of the second peak (see Fig. 4.11(a) in Chapter 4). The
photoluminescence linewidth for type-B quantum dots is narrow, with an
FWHM of 30-50meV, while the linewidth for type-A quantum dots is
broad, with an FWHM of 80-120 meV. At higher growth temperatures,
type-A' quantum dots are transformed into type-B' quantum dots with an
increase in source supply. The weak photoluminescence in type-A' and
type-B' quantum dots above 560°C can be attributed to their low density,
as confirmed by the TEM photograph; the density of type-A' quantum dots
at 560°C and 18 cycles was about 2 x 10" cm-2-one-third that of type-A
and type-B quantum dots. For growth temperatures below 490"C, type-A
quantum dots transform into type-B' quantum dots with weak photo-
luminescence intensity. An introduction of dislocations reduced the photo-
luminescence intensity of the type-B' dots below 490°C, as detected by the
TEM photograph, although their density was high ( - 1 x 10l1cm-2).
Figure 3.13 shows high-resolution cross-sectional TEM images of typical
type-A (14-cycle) and type-B ( 1 8-cycle) quantum dots grown at 520°C.
Type-A dots have a diameter of 15nm and a height of 5nm. The In
composition detected by EDX is 0.30 (i.e., Ino.30Gao,70As). For type B the
diameter, height, and In composition were 25 nm, 8 nm, and 0.23, respect-
ively (i.e., Ino,23Gao,77As).Judging from the smaller In composition for
type-B dots, the longer photoluminescence wavelength is due to larger dot
sizes. We notice that the quantum-well layer surrounding quantum dots in
Fig. 3 3 b ) has become thinner like the wetting layer of an SK-mode dot as
seen in Fig. 3.13(b). This is because the supply of gallium is reduced from
an early amount of almost 1 monolayer to about 0.1 monolayer.
The structure of Fig. 3.5(b) can be reproduced in this MOVPE chamber
by an In-As-Ga-As sequence (Fig. 3.14(a)) or with an InGaAs layer to cover
the dots grown by an In-Ga-As sequence (Fig. 3.14(b)). In the TEM image
of Fig. 3.14(a)), the upper and lower borders of the quantum well are
not flat, since growth conditions were not yet optimized. In Fig. 3.14(b),
type-B quantum dots grown by an In-Ga-As sequence are covered by
172 KOHKIMUKAI,
MITSURUSUGAWARA,
AND MITSURU
EGAWA
FIG.3.13. Cross-sectlonal TEM images of quantum dots: (a) Type-A quantum dots grown
at 520°C with 14 cycles, and (b) Type-B quantum dots grown at 520°C with 18 cycles. The
diameter and height of type-A quantum dots are 15 nm and 5 nm, respectively, and those of
type-B quantum dots are 25 nm and 8 nm, respectively.
subsequent growth of In,,,Ga,,,As quantum wells. The dots in Figs. 3.14(a)

and (b) emit at around 1.3 um (see late news).
2. MULTIPLE-LAYER
GROWTH
A great deal of effort has been devoted to fabricating quantum-dot arrays

that are vertically stacked, vertically aligned, and electronically coupled in
the growth direction (Kuan and Iyer, 1991; Yao et al., 1991; Xie et al., 1995;
Solomon et al., 1996; Ledentsov et al., 1996a, and b). Vertically aligned
arrays can be realized by closely stacking the quantum dots with a
3 VAPORPHASE
METALORGANIC GROWTH
EPITAXIAL 173
FIG. 3.14. TEM image of (a) ALS dots grown by an In-As-Ga-As sequence, and (b) ALS
dots grown by an In-Ga-As sequence embedded in an In,,,Ga,,,As layer.
spacer-layer thickness comparable to or less than their height. With a simple

model showing that new quantum dots tend to nucleate directly above
buried quantum dots. Tersoff et al. (1996) demonstrated that the sizes and
in-plane spacings of the dots become more uniform with the growth of
successive layers. See Chapter 2 for the MBE growth of our closely stacked
and columnar SK quantum dots.
Here, let us see how multiple stacking proceeds in type-A and type-B
quantum dots. We examined triple-layered structures with GaAs spacer
layers whose thicknesses varied from 10 nm to 20 nm. The growth tempera-
ture was set at 52OoC, which was a favorable temperature for type-B
quantum-dot formation. The cycle number of alternate supply was set to 14
and 18 in each layer for both types, respectively.
Figures 3.15 and 3.16 show cross-sectional TEM images of triple-layered
structures of type-A and type-B quantum dots, respectively; a TEM image
174 KOHKIMUKAI,MITSURU
SUGAWARA,
AND MITSURU
EGAWA
FIG.3.15. Cross-sectional TEM images of triple-layered structures of type-A quantum dots

grown at 520°C with 14 cycles. The GaAs spacers are (a) 20 nm, (b) 15 nm, and (c) 10nm. The
single-layer structure of type-A quantum dots is shown in (d).
of a single-layer structure is also shown for comparison. In both types, the

well-known vertical alignment of multistacked quantum dots occurred by
thinning the GaAs spacer layer, originated by the effect of strain field
induced by buried quantum dots. The spacer, at which the vertical align-
ment occurs, was thicker for type B because the strain field around the
buried dots is larger than that for type A due to larger quantum-dot volume.
3 METALORGANIC
VAPORPHASE GROWTH
EPITAXIAL 175
FIG. 3.16. Cross-sectional TEM images of triple-layer structures of type-B quantum dots
grown at 520'C with 18 cycles. The GaAs spacers are (a) 20nm. (b) 15 nm, and (c) 10nm. The
single-layer structure of type-B quantum dots is shown in (d).
Vertically aligned quantum dots became larger toward the upper layers in
the triple-layered structure, a phenomenon also observed in SK islands in
Chapter 2. The diameter of the dots at the top layer was about 35 nm, while
the diameter of those at the bottom layer was about 25 nm.
We discovered a unique phenomenon in the multiple-layer growth of
type-B quantum dots separated with a thick GaAs spacer. As shown in Fig.
176 SUGAWARA,
KOHKIMUKAI,MITSURU EGAWA
AND MITSURU
3.16, in the case of a 20-nm GaAs spacer, the quantum-dot density in the
bottom layer increased to about twice that of the single-layer structure, and
few dots were formed on the upper layers. For type-A quantum dots, a
drastic increase in density did not occur and similar structures were simply
repeated. Since the photoluminescence spectrum of the type-B quantum-dot
triple-layer structure with a 20-nm spacer was similar to that of the
single-layer structure, the newly formed quantum dots in the bottom layer
were of type B. At the present stage, we have no idea why the two types of
dot display different multiple stacking behavior.
V. The Growth Process
We have seen in this chapter that ALS dots are very different from SK
islands. Let us assign type-A quantum dots to SK islands because, with what
we saw in Chapter 2, they have many common characteristics. The type B
ALS dots emit photons with a wavelength of 1.3pm or longer, because of
thicker structure than that of the SK islands. The FWHM of the emission
spectra is 30-50 meV, which is much smaller than the 80- 120 meV of the
SK islands. The ALS dots are buried in a quantum well with clear upper
and lower borders against the GaAs. The thickness of the quantum well
decreases and becomes like that of a wetting layer in SK-mode growth as
the supply amount of group-111 sources is reduced. The area coverage of the
dots is 5-lo%, which is definitely smaller than the 20-40% of the SK
islands. As the alternate supply cycle increases, dot size and emission
wavelength increase, whereas SK islands are almost insensitive to the supply
amount of InAs. As seen in Chapter 2, further indium supply after the
formation of SK islands increases the islands density and then causes them
to coalesce.
The photoluminescence matrix of Fig. 3.12 gives us a hint as to how the
alternate supply works. We have seen that, as the supply cycle increases the
wetting layer and then the SK islands (type A) form, leading to the
appearance of ALS dots (type B). The SK islands are self-assembled
primarily by the thermodynamical balance between strain energy and
surface energy. If we supply only indium after the SK islands (type A) form,
dot size will hardly change, as in the case of usual SK growth, since the
islands have already reached an equilibrium point. A supply of gallium to
the islands decreases the total strain energy, making it possible to add
further indium to the islands and thus increase dot size in a perpendicular
direction. This process repeats itself until the dots reach another equilibrium
point.
3 VAPORPHASEEPITAXIAL
METALORGANIC GROWTH 177
We believe that the growth process is not exactly the same for the early
ALS dots grown by supply sequence of In-As-Ga-As and for the recently
adopted ALS dots grown by In-Ga-As sequence, with both sequences
providing a unique 1.3-pm emission. The plan-view TEM images in Figs.
3.17 and 3.18 indicate how the In source gas contents in a supply cycle
affects the structure of the two types of ALS dots. By the increase of source
(b)
FIG. 3.17. Plan-view TEM images of the ALS dots grown by 18 cycle of the In-Ga-As
supply sequence: (a) the standard case and (b) the case when In source gas contents in a supply
cycle was raised by 20%.
178 KOHKIMUKAI,MITSURUSUGAWARA,
AND MITSURUEGAWA
(bi
FIG. 3.18. Plan-view TEM images of the ALS dots grown by 12 cycle of the In-As-Ga-As
supply sequence: (a) the standard case and (b) the case when In source gas contents in a supply
cycle was raised by 100%.
gas contents, the numerical density increases but the diameter does not
change for the ALS dots grown by In-Ga-As sequence (Fig. 3.18), while the
numerical density does not change but the diameter increases for the ALS
dots grown by In-As-GaAs sequence (Fig. 3.17). We don’t know the details
in the growth processes, but we suppose that the supply of ASH, after the
supply of TMIDMEA makes metal indium on growth surface stable by
VAPOR PHASE
3 METALORGANIC EPITAXIAL
GROWTH 179
forming semiconductor InAs, affecting the successive process of the ALS

dots' formation.
We notice that the ALS dots are very similar to the columnar SK dots in
Chapter 2 in terms of alternate supply of In and G a sources. The distinct
difference is the supply amount, that is, the ALS dots in Section IV supply
0.5 monolayer indium and 0.1 monolayer gallium, while the columnar dots
formed by an alternate supply of 3 monolayer GaAs and 0.7 monolayer
InAs on SK islands. The role of alternate supply in both types of quantum
dots is common, that is, inserting gallium into the structures reduces the
total strain energy, leading to larger and more uniform dots. However, we
should note that the growth process is not exactly the same for both types
of dots since multiple wetting layers are clearly observed in the columnar
dot structure but not in the ALS dot structure. We conjecture that
reconstruction on surface and/or in bulk may occur during the formation of
the ALS dots, in considering with the curious multiple stacking behavior
(see Section IV.2).
The problem of low density in ALS dots arises from the low source supply
amount. If we first form SK islands intentionally by supplying only indium
of more than one monolayer (the InAs layer deforms to islands over
one-monolayer deposition) and then start an alternate supply, we can
increase the dot density, as in the columnar dots grown by MBE. Figure
3.19 shows a trial of the above, photoluminescence spectra at room
0.9 1.0 1.1 1.2 1.3 1.4 1.5

Wavelength (pm)
FIG.3.19. Photoluminescence spectrum of a sample grown by 10 cycles of the In-Ga-As
sequence after supply of 2-monolayer In and As. A 1.3-pm emission was obtained at room
temperature.
180 KOHKIMUKAI,
MITSURU
SUGAWARA,
AND MITSURU
EGAWA
FIG.3.20. Plan-view TEM image of a sample grown by 10 cycles of an In-Ga-As sequence

after supply of 2-monolayer In and As. Quantum dots having an in-plane diameter of 20 nm
were observed.
temperature. A sample was grown by 10 cycles of In-Ga-As sequence after

a supply of 2-monolayer In and As. A 1.3-pm emission with high emission
efficiency was obtained by this hybrid growth method. A plan-view TEM
image of the sample indicated the formation of quantum dots, with a
diameter of about 20 nm (Fig. 3.20). The numerical density increased slightly
(see Fig. 3.5(a)), suggesting that this growth process will solve dot density
problems in the future. At the same time, the hybrid method may be able to
rectify the fact that multiple layers cannot be formed in ALS dots at present
while they can in SK dots.
As we progress from closely stacked to columnar dots by continuously
thinning the spacer, we must also explore the intermediate regime between
ALS dots and columnar dots to expand controllability of size, emission
wavelength, numerical density, and emission efficiency. The alternate supply
method, which provides great controllability to self-assembling growth, will
be a key approach to creating long-wavelength optical devices, such as
1.3-pm quantum-dot lasers.
VI. Summary
We reviewed the growth characteristics of our original ALS quantum

dots, which are grown by an alternate supply of a monolayer or less with a
sequence of In-As-Ga-As and In-Ga-As. The dots have a 1.3-pm emission
3 METALORGANIC
VAPORPHASE
EPITAXIAL
GROWTH 181
and a narrow spectrum linewidth of 28 meV. The quantum dots also acquire
great controllability in size and emission wavelength by an adjustment in
the cycle number and the composition of buffer layers on which the dots are
grown. By comparing the ALS quantum dots with SK islands, especially
their spectrum variation with the supply cycle number, we have achieved a
picture of the growth process of ALS dots. Optical characterization of ALS
dots is presented in Chapter 4.
REFERENCES
Kondo, M., Kuramata, A., Fuji, T., Anayama, C., Okazaki, J., Sekiguchi, H.. Tanahashi. T.,
Yamazaki, S., and Nakajima, K. (1992). J . Cryst. Growth 124, 265.
Krost, A., Heinrichsdorff, F., Bimberg, D., Darhuber, A,, and Bauer, G . (1996). Appl. Phys. Lett.
68, 785.
Kuan, S. T., and Iyer, S. S. (1991). Appl. Phys. Lett. 59, 2242.
Kitamura. M., Nishioka, M., Schur. R.. and Arakawa, Y. (1997). J . Cryst. Growth 170. 563.
Ledentsov, N. N., Grundmann, M., Kirstaedter. N., Schmidt, O., Heitz, R., Bohrer, J.. Bimberg,
D., Ustinov, M. V., Shchukin, A. V.. Kop'ev, S. P., Alferov, I. Z., Ruvimov, S. S., Kosogov,
0. A., Werner, P., Richter, U., Gosele. U., and Heydenreich, J. (1996a). Solid-State
Electron. 40, 785.
Ledentsov, N. N., Bohrer, J., Bimberg, D., Kochnev, V. I., Maximov, V. M., Kop'rv, S. P.,
Alferov, I. Z., Kosogov, 0. A., Ruvimov, S. S., Werner, P., and Gosele, U. (1996b). A p p l .
Phys. Lett. 69, 1095.
Mukai, K., Ohtsuka, N., Sugawara, M., and Yamazaki, S. (1994). Jpn. J . Appl. Phps. 33. L1710.
Mukai, K., Ohtsuka, N., and Sugawara, M. (1996). Jpn. J . Appl. Phys. 35, L262.
Mnkai, K., Ohtsuka, N., Shoji, H., and Sugawara, M. (1997). Appl. Surf: Sci. 112, 102.
Mon, K., and Usui, S. (1992). Appl. Phys. Lett. 60, 1717.
Ohtsuka, N., and Ueda, 0. (1994). In Gus-Phase und Surface Clieinistry in Electronic Materials
Processing Vol. 334, 225-22.
Ohtsuka, N., and Mukai, K. (1995). Proc,. I n t . Conf 1nP and Related Materials, Hokkaido, 303.
Ozeki, M., Mochizuki, K., Ohtsuka, N., and Kodama, K. (1988). Appl. Phys. Lett. 53. 1509.
Ozeki, M., Mochizuki, K., Ohtsuka, N..and Kodama, K. (1989). Thin Solid Films 174. 63.
Ozeki, M., Ohtsuka, N., Sakuma, Y.. and Kodama, K. (1991). J . Cryst. Growth 107. 102.
Ramachandran, R. T., Heitz, R., Chen, P.. and Madhukar, A. (1997a). Appl. Phys. Lett. 70, 640.
Ramachandran, R. T., Heitz, R., Kobayashi, P. N., Kalburge, A,, Yu, W., Chen, P., and
Madhukar, A. (1997b). J . Cryst. Growth 175/176, 216.
Roan, E. J., and Cheng, K. Y. (1991). Appl. Phys. Lett. 59, 2688.
Sakuma, Y., Kodama, K., and Ozeki, M. (1988). Jpn. J . Appl. Phys. Lett. 27, L2189.
Solomon, S. G., Trezza. A. J., Marshall. F. A.. and Harris. S. J. (1996). Pkys. Rev. Lett. 76, 952.
TersoK, J., Teichert. C., and Lagally, G. M. (1996). Phys. Reo. Lett. 76, 1675.
Xie, Q., Madhukar, A., Chen, P., and Kobayashi, P. N. (1995). Phys. Reii. Lett. 75, 2542.
Yao, Y. J., Andersson, G. T., and Dunlop, L. G. (1991). J . Appl. Phys. 69, 2224.
SEMICONDUCTORS AND SEMIMETALS, VOL. 60
CHAPTER4
Optical Characterization of Quantum Dots

OPTICAL SEMICONDUCTOR DEVlCtS LABORATORY
FUIITSULABORATORIES
LTV
ATSUGI.KANAGAWA.
JAPAN
I. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
11. LIGHTEMISSION FROM DISCRETE ENERGYSTATES . . . . . . . . . . . . . . 185
I, Photoluminescence, Photoluminescence Excitation, and Electroluininescence
Spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
2. Wafer Mapping . . . . . . . . . . . . . . . . . . . . . . . . . . 190
3. Microprobe Photoluminescence . . . . . . . . . . . . . . . . . . . . 192
111. CONTROLLABILITY OF QUANTUM CONFINEMENT . . . . . . . . . . . . . . 196
1. Two Methods of Controlling Quantized Energies . . . . . . . . . . . . 196
2. Magneto-Optical Spectroscopy . . . . . . . . . . . . . . . . . . . . 200
IV. RADIATIVE EMISSION EFFICIENCY. . . . . . . . . . . . . . . . . . . . 20 1
V. S U M M A R Y . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
1. Introduction
Optical diagnostics have played a central role in characterizing semicon-

ductor nano-structures. This is because, by probing photons that interact
with electrons in nano-structures and by analyzing the optical spectra in
time and wavelength regimes, we obtain ample information on the nano-
structures’ electronic states, such as energy levels, density of states, band
structures, and wave functions, as well as their carrier dynamics, such as
relaxation and scattering. Information on strength and resonant wavelength
of the photon-electron interaction process itself, such as spontaneous
emission, stimulated emission, optical absorption, and various optical non-
linear interactions, is also important in the fabrication of optical devices.
Another advantage of optical diagnostics is that, in most cases, we are free
from time-consuming sample preparation. In the two decades of quantum-
well research, many useful optical diagnostic techniques have been estab-
183
Copyright i: 1999 by Academic Press
All rights of reproduction in any form reserved.
ISBN 0-12-752169-0
ISSN 0080-R784!99 $30.00
184 KOHKIMUKAI
AND MITSURUSUGAWARA
lished. Photoluminescence is one of the most conventional and useful,

providing much information on energy states as well as crystal quality.
Among other optical methods are photoluminescence excitation spectro-
scopy, time-resolved photoluminescence, microprobe optical spectroscopy
performed with a microscope or sharpened optical fiber, and magneto-
optical measurements.
This chapter deals with optical characterization of quantum dots, focusing
on the original alternate supply (ALS) quantum dots introduced in Chapter
3 (see also Mukai et al., 1994). The ALS dot is one of four types of quantum
dot treated in this volume; the other three are Stranski-Krastanov (SK) dots
(Tabuchi et al., 1992), closely stacked dots (Nakata et al., 1997a), and
columnar dots (Nakata et al. 1997b; also see Chapter 2).
Transmission electron microscope (TEM) photographs of ALS and SK
dots are shown in Fig. 4.1. The SK dots are dome-shaped (in another case,
Cross-sectional TEM
SK dot
ALS dot
FIG.4.1. Two types of self-assembled dots. The InAs SK dots are grown by MBE or
MOVPE via the Stranski-Krastanov (SK) mode under highly strained epitaxy. The InGaAs
ALS dots are grown by alternately supplying InAs and GaAs precursors with one or fewer than
one monolayer. While SK dots are dome- or pyramid-shaped islands on a growth surface, ALS
dots are spherical and surrounded laterally by a quantum-well layer having almost the same
thickness as that of the dots.
4 OPTICAL CHARACTERIZATION OF QUANTUM
DOTS 185
pyramid-shaped; Daruka and Barabasi, 1997) islands on a highly lattice-

mismatched growth surface; the ALS dots are spherical and surrounded
laterally by a quantum-well layer having almost the same thickness as that
of the dots. One of the unique features of ALS dots is that they emit at
1.3 pm-a suitable wavelength for an optical data transmission system.
Using diagnostic techniques, we will show the unique properties of ALS
dots, such as long emission wavelength, emission spectra with multiple
peaks from discrete energy states, harmonic-oscillator type confinement
potential, large wavelength tunability between 1.2 and 1.5 pm through size
control, and carrier lifetimes through radiative and nonradiative recombina-
tions. We will compare ALS and SK dots, showing the advantages of the
alternate supply method.
11. Light Emission from Discrete Energy States
1. PHOTOLUMINESCENCE,
PHOTOLUMINESCENCE
EXCITATION
AND
ELECTROLUMINESCENCE
SPECTRA
Photoluminescence spectra of ALS and SK self-assembled quantum dots

at 300K are shown in Fig. 4.2, measured with a Kr’ laser at a low
excitation power of 60 W/cm2. The differences between the two are remark-
able. The spectrum peak wavelength of the ALS dots is about 1.35pm,
which is longer than the 1.12pm of the SK dots. We also see that the
spectrum broadening in the ALS dots is less than half that of the SK dots.
Figure 4.3 shows the size fluctuation of ALS dots evaluated by plan-view
TEM images (see Chapter 3). The diameter distribution follows a Gaussian
curve and is comparable to or smaller than that of the SK dots (Leonard et
al., 1993; Moison et al., 1993). Since the spectrum consists of the ensemble
of an emission line from each quantum dot, the narrower spectrum in ALS
dots indicates higher uniformity of dot emission energies and thus higher
uniformity of dot size, composition, and strain.
Figure 4.4 shows the full width at half maximum (FWHM) of the
emission spectra of ALS dots as a function of temperature from 60 to 300 K.
Spectra at 4.2, 77, and 300 K are shown in the inset. Temperature-indepen-
dent spectrum width indicates that the broadening is dominated neither by
the thermal carrier distribution nor by homogeneous broadening from
phonon scattering, but by structural inhomogeneity. The homogeneous
broadening of a single SK dot has been shown by microprobe photo-
luminescence to be less than 0.1 meV, which is much smaller than the
measured spectrum width in Fig. 4.2 (Marzin et al., 1994). Theoretically, the
186 KOHKIMUKAIAND MITSURUSUGAWARA
0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6

Wavelength (pm)
FIG. 4.2. Photoluminescence spectra at 300 K of the ALS and SK dots. The ALS dots were
grown by 18 cycles of In-Ga-As alternate supply on an In,,,,Ga,,,As buffer layer and capped
by an In,,,,Ga,,,,As layer. The emission wavelength of the ALS dots was above 1.3 pm,longer
than that of the SK dots. The full width at half maximum (FWHM) of the spectrum was
30 meV in the ALS dots, showing high uniformity. (Reprinted from Appl. Surface Science, vol.
112, Mukai et al., “Growth and optical evaluation of InGaAs/GaAs quantum dots self-formed
during alternate supply of precursors,” pp. 102-109, 1997, with kind permission of Elsevier
Science-NL, Sara Burgerhartstraat 25, 1055 KV Amsterdam, The Netherlands.)
homogeneous broadening of each quantum dot, referred to as the natural

spectrum linewidth, is controlled by the spontaneous emission rate and
carrier scattering rate, as seen in Eq. (1.200) in Chapter 1. (See Eq. (1.210)
in Chapter 1 for the spontaneous emission rate and Chapter 7 for the
calculation of the scattering rate as a function of the carrier density around
dots.) It is yet to be investigated how large the inhomogeneous broadening
of a single dot becomes at room temperature and when there are numerous
carriers both inside and outside the dots, as is the case with the lasing
operation. Note that, in quantum wells, photoluminescence spectra become
broad toward high energies as temperature increases, primarily due to the
thermal distribution of carriers over continuous energy states. The tempera-
ture-independent spectrum width in Fig. 4.4 is one of the factors proving
that photoluminescence is caused by the interband transitions between the
discrete states in the conduction and valence bands.
4 OPTICAL
CHARACTERIZATION
OF QUANTUM
DOTS 187
Diameter observed by TEM (nm)

FIG. 4.3. Diameter of 106 ALS dots observed in a plan-view TEM image. The distribution
follows a Gaussian curve with a standard deviation of 2.9 nm.
The inset of Fig. 4.4 shows additional shoulders at the shorter wavelength
regime of the spectra. These were the first sign of excited-level photo-
emission from self-assembled dots (Mukai et al., 1994). Multiple peak
emissions from discrete states can be detected owing to a narrower spectrum
linewidth than that of the interlevel separation. Figure 4.5 shows the
excitation power dependence of the photoluminescence spectra of ALS dots
at 77 K. As the excitation power increased, the second peak appeared, and
its intensity increased to exceed the first peak. At the highest power of
306 W/cm2, a shoulder assigned to the third peak appeared. Figure 4.6
compares the photoluminescence spectrum at a rather high excitation
intensity of 100 W/cm2, and the photoluminescence excitation spectrum,
monitored at the energy of 1.0 eV (the lowest-energy first emission). It was
possible to detect two clear resonances exactly at the same energy as those
in the photoluminescence spectra. This represents direct proof that the
multiple peaks in the photoluminescence spectra are due to the discrete
states under three-dimensional quantum confinement. Let us call the first
peak in the photoluminescence spectra the ground-state emission and the
second peak the first excited-state emission.
Figure 4.7 shows the electroluminescence spectra of the ALS dots at 77 K
from the p-n junction diode with the dot active region. The area of the
k 0 1. . , * I , . .
* I ' ' * ' I a .' ' I ' ' ' ' ". a '
50 100 150 200 250 300 350

Temperature (K)
FIG.4.4. Photoluminescence spectrum width (FWHM) of the ALS dots as a function of
temperature. The dots were grown using 12 cycles of In-As-Ga-As alternate supply on an
In,o,Ga,,,,As buffer and capped with an In,,,,Ga,,,As layer. Photoluminescence spectra at
4.2, 77, and 300 K are shown. The width of 30 meV is independent of temperature.
electrode was 20 x 900 pm'. Since the electric carrier injection can provide
many more carriers into dots than the optic excitation can, peaks up to the
5th appeared at the current injection of 400mA. As the injected current
increased, the emission intensity of the first and second peaks were almost
saturated. Since the emission rate is given by NITsp, where N is the carrier
number occupying the state and t S p is the spontaneous emission lifetime. the
saturation of the emission intensity is due to the saturation of the carrier
number in the corresponding energy state. As low energy states come to be
almost occupied by carriers, the relaxation rate into the states decreases due
to Pauli blocking, and the emission starts to rise from higher energy states.
However, Pauli blocking cannot completely explain the excitation inten-
sity dependence of emission spectra in Fig. 4.7. We should note that, at
lOmA, the second and third emissions had appeared though the first
emission intensity was less than its maximum. The carrier number, N , of
each state, governing the emission intensity, is determined by the balance
among the rates of various processes, such as spontaneous emission at the
state, relaxation into the state, and thermal emission from the state. In the
ALS dots with an energy separation of 50-80 meV-much higher than the
thermal energy at 77 K-the thermal emission rate is negligible and the
4 OPTICAL
CHARACTERIZATION
OF QUANTUM
DOTS 189
I I I I I I I I
I
I I I I I I I
1 .o 1.1 1.2 1.3 1.4
Wavelength (pm)
FIG.4.5. Excitation power dependence of photoluminescence spectra of ALS dots grown
using 9 cycles of alternate supply. The sample was excited by a 647.1-nm Kr’-ion laser. As the
excitation power increased, the second peak appeared and its intensity increased to exceed the
first peak. At the highest power of 306 W/cm2, a shoulder assigned to the third peak appeared.
(Reprinted from Appl. Surface Science, vol. 112, Mukai et al., “Growth and optical evaluation
of InGaAs/GaAs quantum dots self-formed during alternate supply of precursors,” pp.
102- 109, 1997, with kind permission of Elsevier Science-NL, Sara Burgerhartstraat 25, 1055
KV Amsterdam, The Netherlands.)
experimental emission spectrum with multiple peaks even at low excitation

indicates that the relaxation rate is comparable to the spontaneous emission
rate. This is a phonon bottleneck (discussed in detail in Chapter 5). One
notices immediately about this bottleneck that the emission spectrum at
IOmA in Fig.4.7 cannot be reproduced by a calculation that simply
combines the Fermi-Dirac distribution at 77 K and the density of states of
the quantum dots. Due to the phonon bottleneck effect, N does not
completely reach 2 N , to give maximum intensity.
Three features of the emission spectra of ALS dots are observed in Fig.
4.7. First, the higher the peak order, the larger the maximum peak intensity.
Second, the peak energies were constant during the increase of injected
current, supporting our belief that the observed peaks correspond to
sublevels in the dots. Third, the peak intervals between the neighboring
emission lines are almost uniform -a characteristic of harmonic-oscillator-
type quantum confinement. The confinement potential is determined by the
magneto-optical measurements described in Section 111.
I I I I 1
n
3
-a
3
3 -
.3
LQ
G -
I
0.9
I
1.o
Y I
1.1
I
1.2
1I
1.3
Energy (eV)
FIG.4.6. Photoluminescence excitation spatrum at 4.2 K of the sample grown by 18 cycles
of In-Ga-As supply. The sample was excited by the monochromated light of a Halogen lamp,
and a Ge detector was used to detect luminescence from the sample surface at the energy of
1.0 eV, that is, the lowest-energy first emission. Two clear resonances could be detected exactly
at the same energy as that in the photoluminescence spectra (from Mukai et al., 1996c.
Copyright 1996 by the American Physical Society).
2. WAFERMAPPING
Figure 4.8 shows the microprobe photoluminescence mapping of an ALS

quantum-dot wafer at room temperature: (a) the energy separation between
the ground-state (first) and the first excited-state (second) emission peaks;
(b) the ground-state emission energy; and (c) the FWHM of the ground-
state emission spectrum. The correction of luminescence from the sample
surface was restricted to the area measuring 2 x 2 pm2 using a crossing slit.
The sample was grown by 14 cycles of In-Ga-As supply on a 2-inch (100)
GaAs wafer in a state-of-the-art MOVPE chamber (see Chapter 3 for details
on growth). During growth, the wafer was rotated at the rate of 30rpm,
which caused the distribution of the three diagnostic items to have a roughly
rotational symmetry around the wafer center.
By comparing the three maps, we can deduce what caused the observed
patterns. The energy separation of Fig. 4.8(a) is in the 85- to 95-meV range
and is almost uniform except at the wafer edge. The ground-state emission
energy in Fig. 4.8(b) was the largest at about 1 inch from the wafer center.
4 OPTICAL
CHARACTERIZATION
OF QUANTUM
DOTS 191
0.9 1.o 1.1 1.2 1.3
Wavelength (pm)
FIG. 4.7. Electroluminescence spectra of ALS dots grown by 18 cycles of In-Ga-As
alternate supply. The luminescence parallel to the surface was dispersed and detected by an
InGaAs photo-multi detector quenched to - 70°C using a conventional lock-in technique.
Electrode size was 20 x 900prn'. Up to the fifth emission peaks appeared as the injected current
increased (from Mukai et al., 1996b).
The FWHM in Fig. 4.8(c) had a distribution similar to that of the

ground-state emission energy of (b) -the higher the ground-state energy,
the narrower the spectrum width. To understand the patterns, we note two
factors causing the observed distributions: the dot size and the indium
composition of the dots. As the quantum-dot size decreases, all three items
of Fig. 4.8- the energy separation, the ground-state emission energy, and
the spectrum width -increase. As the indium composition decreases, the
ground-state emission energy increases but the energy separation hardly
increases.
The observed maps are explained as follows. First, energy separation in
Fig. 4.8(a) suggests the good uniformity of dot size all over the wafer size.
Second, the high-energy region off the wafer center in Fig. 4.8(b) has the
smallest indium composition but does not have the smallest dot size. Third,
the narrowest spectrum broadening off the wafer center in Fig. 4.8(c) is due
to the uniform indium composition. This indicates that spectrum broaden-
ing is caused not only by size fluctuation but also by compositional
fluctuation of indium among each quantum dot.
192 KOHKIMUKAI
AND MITSURUSUGAWARA
0 90-95meV 0 095-O96eV 0 40-50me~

85 - 90 0 94 - 0 95 a 50-60
80 - 85 093-094 unmeasurable
70-80
Energy separation Ground-level energy FWHM of ground level
FIG.4.8. Area mapping of the photoluminescence of an ALS quantum-dot wafer at room

t e w r a t u r e : (a) energy separation between the ground-state (first) and the first excited-state
(second) emission peaks; (b) ground-state emission energy: and (c) FWHM of the ground-state
emission spectrum. The probe size was 2 p m . The scan step was 2mm in both x and y
directions. The sample was excited by an Ar+ laser at 514.5 nm, and luminescence from the
sample surface was monochromed by a 50-cm rnonochrorneter and detected by a multichannel
photodetector.
3. MICROPROBE
PHOTOLUMINESCENCE
Increasing the spatial resolution of the optical probe is a major target in

the development of optical diagnostic techniques. It enables us to obtain
signals from a small number of dots, leading to an understanding of
quantum-dot optical properties that was heretofore unobtainable through
averaged signals over a large number of dots by macroscopic measurements
(Brunner et al., 1992). Microscope photoluminescence that focuses the
excitation laser beam via an objective lens to about 1-pm diameter on a
sample surface is already a popular technique, and its experimental setup is
commercially available. Higher spatial resolution is obtained via simple
sample processing that opens a window in a metal mask covering the surface
through which the photoluminescence can be measured. Sharp emission
lines caused by a single dot or a small number of dots can be detected, and
the linewidth -homogeneous broadening of a single-dot emission line -
has been evaluated by several researchers (see Marzin et al., 1994; Hessman
et al., 1996).
4 OPTICAL
CHARACTERIZATION DOTS
OF QUANTUM 193
FIG. 4.9. Near-field optical microscope. Samples were set in the cryostat and cooled to 5 K.
With a wavelength of 633nm, the excitation beam went from the laser diode through a
sharpened optical fiber to the sample surface. Luminescence from the surface was collected by
the fiber probe. The tip was controlled in close proximity to the sample by the shear-force
feedback technique. The luminescence was dispersed by a monochrometer, and the photon
signal was counted with a n InGaAs photo-multi detector kept at -8O'C.
The other approach to microprobe spectroscopy is near-field optical

microscopy (fiber probe microscopy) (Durig et al., 1986; Betzig et al., 1991),
which involves collecting near-field light. Its advantage is that spatial
resolution is not limited by the wavelength of the excitation laser; its
drawback is weak signal intensity. With the shear-force feedback technique,
high resolution optical imaging can be obtained by tracing the surface
structure of samples (Betzig et al., 1993; Ambrose et al., 1994; Grober et al.,
1994; Ghoemi et al., 1995).
We performed microprobe photoluminescence spectroscopy of ALS dots
using the near-field scanning optical microscopy system (NSOM). Our
experimental equipment is illustrated in Fig. 4.9. The excitation beam of a
wavelength of 633 nm from the laser diode went through an optical fiber to
the sample surface, and luminescence from the surface was collected by the
fiber probe. The luminescence was monochromed by a monochrometer with
a resolution of 1 nm. The tip was kept in close proximity to the sample
(- 10 nm) by applying shear-force feedback (Betzig et al., 1992). The fiber
-
probe resolution was restricted to 1 pm, because of the tradeoff between
spatial resolution and detection sensitivity (Saiki et al., 1996). The number
of the dots was estimated to be about 200 in a 1-pm spot.
Figure 4.10 shows the luminescence spectrum of the ALS dots by the
NSOM system with an excitation power of 90 W/cm2. Spectra of inacro-
scopic photoluminescence with a 300-,um$excitation laser spot are superim-
c
0
Y
0
d
a
. 90 W/cm2
NSOM :
r . . . . l . . . . l . . . . l . . . . i
0.95 1.oo I .05 1.10 1.15
Energy (eV)
FIG.4.10. Photoluminescence spectra of the ALS dots by the near-field optical microscope.
The excitation power was estimated to be 90 W/cm2. The spectra of macro-photoluminescence
is shown by a dashed line. The two sets of multiple emission peaks in the NSOM spectra
correspond to the ground-state and the first excited-state emission.
posed by a dashed line. The two sets of emission spectra with fine structures
denoted as S, and S,, correspond to the quantum dots' ground state and
first excited state, respectively. Note that S, and S, are well separated and
that almost no emission was observed in between. This is because the
smaller probing area reduced the inhomogeneous spectrum broadening.
However, the emission spectrum of S, and S, did not become as narrow as
we expected, indicating that inhomogeneity of dots had already occurred in
the area smaller than 1 pm. Inhomogeneity in such a small area cannot be
caused by nonuniformity in growth conditions, such as wafer temperature
distribution and source gas streamline. The fine structures on S, and S, are
reproduced by repeated measurements and can be attributed to emission
from one or a few dots. Sharp peaks, denoted by the common small letter
with the subscripts 1 and 2, might be the ground-state and the excited-state
emission of the same dots.
Figure 4.1 l(a) shows the excitation-power dependence of the NSOM
spectra. Comparing that with the macroscopic photoluminescence spectra
of Fig. 4.5, we see that the emission from each state is sharper and up to the
seventh emission is resolved, due to the higher spatial resolution. We found
that the higher the sublevel order, the wider the peak width. Thus, the
4 OPTICAL
CHARACTERIZATION
OF QUANTUM
DOTS 195
0.9 1 1.1 1.2 1.3 1.4 1.5 1.6

Energy (eV)
0.95 1.oo 1.05

Energy (eV)
FIG.4.1 1. (a) Photoluminescence spectra of the ALS dots by a near-field optical micro-
scope at various excitation-power densities. As the excitation power increased, a higher-energy
emission peak was clearly discernable. (b) NSOM spectra around the ground-state emission.
An additional emission appeared at the lower-energy side of the peak at high excitation (from
Mukai et al., 399713).
higher-ordered peaks were not well isolated, perhaps because the energy
interval was not uniform in dot ensembles in the area; the discrepancy in
the interval enhances the peak broadening at the excited sublevels. Another
reason might be that an additional emission appeared at the lower-energy
side in each peak. The spectrum around the ground-state emission is
magnified in Fig. 4.11(b). As the excitation power was increased from
90W/cm2 to 9kW/cmz, the peak emission intensity of the ground state
decreased and the lower-energy emission rose. This red shift of emission
might be due to the many-body effect in quantum dots (Sugawara, 1996).
111. Controllability of Quantum Confinement
The energy states (e.g., quantized energies and density of states) and
optical characteristics of quantum nano-structures can be artificially con-
trolled by changing the size of the structures. In quantum wells, well width
and strain are designed to match target devices using well-established k . p
band calculation methods (for example, see Chuang, 1995) and the designed
structures are precisely fabricated in the atomic scale by state-of-the-art
epitaxial growth techniques like MOVPE and MBE. The ability to control
quantum-dot structures is not yet as strong as it could be in spite of its
significance for practical device applications. In self-assembled InAs SK dots
on GaAs substrates, the size and emission wavelength can be controlled only
within a narrow range by changing the growth temperature (Chapter 2).
Changing the supply amount of sources has a negligible effect because the
assembling process is driven toward a thermodynamical equilibrium point.
In this section, we will introduce two methods of controlling the size and
the emission wavelength of InGaAs ALS quantum dots over a wide range
of 1.2 to 1.5pm. Then we will evaluate the quantum-dot confinement
potential by magneto-optical measurements.
1. TWO METHODS
OF CONTROLLING QUANTIZED
ENERGIES
The first method of controlling the emission wavelength of quantum dots

is to change the number of supply cycles in the alternate supply growth (See
Chapter 3 for the ALS growth process). Figure 4.12 shows the 4.2-K
photoluminescence spectra of ALS quantum dots at various cycle number
of the In-As-Ga-As supply cycle. At cycle 7, the dots were assembled to
release the strain energy accumulated by the lattice mismatch between InAs
and GaAs. As the number of cycles increased from 7 to 30, the peak
4 OPTICAL
OF QUANTUM 197
4.2 K
1.o 1.1 1.2 I .3 I .4
Wavelength (pm)
FIG. 4.12. Photoluminescence spectra of the ALS dots grown by various cycles of In-As-
Ga-As alternate supply. As the number of cycles increased, the peak wavelength shifted from
1.17 to 1.3 pm. The second peaks were observed at the left side of the first peak. The maximum
emission intensity was obtained by the sample grown by 9 cycles (from Mukai et al., 1996a).
wavelength shifted from 1.17 to 1.3pm. We observed weak peaks or

shoulders (indicated by arrows) that had higher emission energies than the
first peak had. In the 7-cycle sample, the second peak was hardly observed,
probably due to the low optical quality. As the number of cycles increased,
the second peak moved toward the first peak, and in the 30-cycle sample
the two merged. Plan-view TEM images of the samples in Fig. 4.13 show
that, as the supply cycles increased, so did the diameter of the dots. As
summarized in Fig. 4.14, the cycle increase enlarges the diameter and
reduces the emission energy and the energy separation between the two
peaks. If the first peak is assigned to the ground-state emission and the
second peak to the exited-state emission, as in Section IT, the results in Fig.
198 KOHKIMUKAI
AND MITSURUSUGAWARA
(C)
-
40nm
FIG.4.13. Plan-view TEM images of ALS dots grown with (a) 9 cycles, (b) 12 cycles, and
(c) 24 cycles. The dot diameter increased as the number of cycles increased.
4.14 go well with the quantum-confinement concept of energy increase due

to confinement. The maximum emission intensity and the narrowest spec-
trum width were achieved at the same time at 9 cycles, showing that the
alternate supply method controls not only the size but also the crystal
quality.
The second way to control emission wavelength is to change the lattice
constant of the layer on which the dots are grown. This method is based on
the idea that formation of the self-assembled three-dimensional islands is
motivated by the release of accumulated strain energy, and that the size of
the islands increases as the lattice mismatch between the substrate and the
supplied materials decreases. We grew 500-nm In,Ga, -,As buffer layers
with the composition of x = 0 to 0.09 on GaAs substrates. A reciprocal
space map of an X-ray diffraction showed that up to 50% strain relaxation
occurred in the In,Ga, - ,As buffer layer and that the lattice constant of the
buffer layer increased with an increase of x due to the relaxation of strain.
Figure 4.15 shows that, as the indium composition of the buffer layer
increased from 0 (GaAs) to 0.09 (Ino,09Gao,91As),the emission wavelength
increased from 1.3 to 1.46 pm when supplying 12 cycles. TEM measurements
4 OPTICAL
CHARACTERIZATION
OF QUANTUM
DOTS 199
- 35
n 1.05 - x
- 30
%
W
1st
0
1.00 ~
- 25
-
4.2 K
- 20
0.95l' '
lI
"
' I ' ' I " " I ' " * I " " ' I " 'I
5 10 15 20 25 30 35
(InAs)/(GaAs) supply cycle number

FIG. 4.14. Energies of the first and second peaks in the photoluminescence spectra and dot
size observed by TEM as a function of the supply cycle number during ALS quantum-dot
growth (from Mukai et al., 1996b).
E
1.45 F----A IntAsfGdAs x 12 cycles
3.
W
l1!
1.35
1.3 i/
c
0 0.05
--I
0.1
Indium composition of buffer layer
FIG. 4.15. Emission wavelength as a function of the indium composition of an InGaAs
buffer layer. With 12 cycles of In-As-Cia-As alternate supply, the emission wavelength increased
from 1.3 to 1.46pm as the indium composition increased from 0 (i.e., GaAs) to 0.09 (from
Ohtsuka and Mukai, 1995).
Indium composition of InGaAs buffer layer

FIG.4.16. In-plane dot diameter observed by plan-view TEM and dot composition
measured by EDX as a function of the indium composition of the buffer layer. Composition
was determined by calibrating the signal intensity, using the intensity of the lattice-matched
In,.,,Ga, ,,As/InP as a reference. The indium composition was between 0.4 and 0.6 in these
samples and did not obviously depend on the buffer layer's composition. The diameter
increased as the indium composition of the buffer layer increased.
show that the diameter of the ALS dots increased from 20 nm to 35 nm as

the indium composition of the buffer layer increased (Fig. 4.16). Energy
dispersive X-ray microanalysis (EDX) determined that the composition of
the dots was between 0.4 and 0.6 and did not obviously depend on the
composition of the buffer layer.
By combining two control methods just described, we can control the
emission wavelength to the ALS dots from 1.2 to 1.5pm. The buffer layer
can be also applied to tune the size and emission wavelength of ordinary
SK dots.
2. MAGNETO-OPTICAL SPECTROSCOPY
Exciton diamagnetic shifts are powerful tools for evaluating not only the
reduced effectivemass but also the magnitude of potential affecting excitons.
The magnetic field confines excitons in the plane perpendicular to the field
and increases their energy. We can determine the magnitude of other
4 OPTICAL
CHARACTERIZATION
OF QUANTUM
DOTS 201
competing confinement potentials by evaluating the diamagnetic energy

shifts. In studying the diamagnetic shifts of luminescence from 1.3-pm-
emitting InGaAsP/InP quantum wells and the ALS dots, we detected the
confinement potentials for localized excitons in the quantum wells and
quantum dots. Magnetic fields were applied perpendicularly to the wafer
using a superconducting magnet immersed in liquid helium. The sample was
placed near the center of this magnet, which can generate a magnetic field
of up to 14 Tesla. A multiline Ar' laser beam was fed to the sample through
an optical fiber bundle. The photoluminescence from the sample surface was
led to a 32-cm monochrometer through the optical fiber and detected by a
cooled InGaAs photomultiplier using a conventional lock-in technique.
Figure 4.17(a) shows measured magnetic-field-induced energy shifts of the
emission spectra of dot samples grown by various alternate supply cycles.
The diamagnetic shifts of the exciton emission at 1.3 pm in InGaAsP/InP
quantum wells are superimposed as a reference. The shifts in dots were
smaller than those in quantum wells due to the in-plane confinement
potential, and they increased with the cycle number. In 9-cycle samples the
shift was negligible. By simulating the energy shifts using the harmonic
oscillator-type confinement potential introduced in Chapter 1, we evaluated
the in-plane confinement potential of quantum dots. The effective-mass
equation for this problem is Eq. (1.214) in Chapter 1 with the confinement
potential of rn,*l1w,2/2.Solid lines in Fig. 4.18 show the calculated diamag-
netic shifts using the confinement frequency of 0 , as a parameter. The
reduced effective mass of m,*" = 0.04~1,and the static dielectric constant of
c = 1 3 . 9 ~were
~ used (Sugawara et al., 1993). Comparing the calculated
shifts with the measured ones, we obtained w, = 7 x IOl3 s - l for 24 cycles,
1 x 1014s-' for 18 cycles, 1.3 x 1014 s-' for 12 cycles, and 4 x l O I 4 s - l for
9 cycles.
As the cycle number increases, the confinement potential spreads. Figure
4.17(b) compares the confinement potential estimated by TEM and EDX
(dashed line) with that estimated by magneto-optical evaluation (solid line)
for the sample grown by 12 cycles. The origin of the x-axis was set at the
center of the dot. The two estimations agree.
IV. Radiative Emission Efficiency
Quantum-dot laser performance -for example, threshold currents and

external quantum efficiencies-are quite sensitive to the crystal quality of
dots, as seen in the simulation in Chapter 1. This is because nonradiative
centers, if produced during the self-assembling process, consume carriers
202 KOHKIMUKAIAND MITSURU
SUGAWARA
0 50 100 150
Square of magnetic field (T2)
In-plane distance (nm)

FIG.4.17. (a) Measured and calculated magnetic-field-induced energy shifts as a function of
the square of the magnetic field. Diamagnetic shifts of exciton resonance in InGaAsP quantum
wells emitting at 1.3pm are superimposed as a reference. The shifts in quantum dots were
smaller than those in quantum wells and increased with the number of supply cycles. Lateral
quantum-confinement potentials were determined by fitting the calculation (solid lines) to the
measured shifts. (b) Confinement potential estimated by TEM and EDX (dashed line) and that
estimated by magneto-optical evaluation (solid line) for the sample grown using 12 cycles. The
origin of the x-axis was set at the center of the dot. The two results agree well. ((a) Reprinted
from Appl. Surfure Science, vol. 112, Mukai et a]., “Growth and optical evaluation of
InGaAs/GaAs quantum dots self-formed during alternate supply of precursors,” pp. 102-109.
1997, with kind permission of Elsevier Science-NL, Sara Burgerhartstraat 25, 1055 KV
Amsterdam, The Netherlands; (b) from Mukai et al.. 1994.)
4 OPTICAL
OF QUANTUM 203
injected into the laser active region. We must pay attention to the quality
of both inside dots and outside dots, since the retarded carrier relaxation
peculiar to quantum dots enhances the opportunity for carriers to be killed
before they relax. Here, we evaluate the nonradiative channel in quantum-
dot crystals by measuring time-resolved photoluminescence decay and
photoluminescence intensity as a function of temperature.
Recombination lifetimes of the discrete emission spectra from ALS dots
are shown in Fig. 4.18. (See Chapter 5 for details on time-resolved photo-
luminescence measurements.) The measured lifetimes, T " , are almost inde-
pendent of the emission level and are about 0.8 to 1.5 nanoseconds, which
we judge to be long enough for laser application. The lifetime is slightly
shorter than the calculated spontaneous emission lifetime of z S p =2.8 ns
using Eq. (1.210) of Chapter 1, with the exciton effect (the Coulomb
enhancement effect) neglected. We are not sure whether the slightly shorter
experimental lifetime is due to a simple error caused by missing factors in
the theory or to uncertain parameters in the calculation or to nonradiative
process in the measurements. If we estimate the nonradiative lifetime, z,,,
using the relationship of T,- = t,;’ + T,;' and z, = 0.8 - 1.5 ns, we get
T,, = 0.6 - 1.0ns. The measured lifetimes are almost independent of tem-
perature up to 300 K because of the discrete quantized levels that prevent
0
+0 0
4 54 'f .$
10-'O r 0 2nd +5th
A 3rd
0 50 100 150 200 250 300

Temperature (K)
FIG. 4.18. Recombination lifetimes of the five emission peaks in the ALS dots measured by
time-resolved photoluminescence. The lifetimes were about one nanosecond and almost
independent of temperature up to 300 K .
exciton thermal distribution, which are peculiar to quantum dots (see

Chapter 1). Since T,, normally depends on temperature, we assume that the
nonradiative process did not influence the recombination lifetime.
Figure 4.19 shows the photoluminescence intensity of the ground-state
emission of ALS dots as a function of temperature. The intensity was
normalized by the value at 4 K. Reported photoluminescence intensities of
the SK dots grown by MBE are also shown as references. From 4 K to
300 K, the photoluminescence intensity of the ALS dots decreased mono-
tonously by about one order of magnitude. This decrease is almost equal to,
or even smaller than, that of the quantum wells (Mirin et al., 1995). The
photoluminescence intensity of the SK dots decreased by almost three
orders of magnitude. For the SK dots, there seems to be a critical
temperature between 150 and 200 K, where the slope inclination changes,
- Calculation 4
1.1 1.3 1.5
FIG.4.19. Photoluminescence intensity of the ALS dots as a function of temperature. The

intensity was normalized by the value at a low temperature. Excitation was done by a 647.1-nm
Kr’ ion laser. Reported intensities of the SK dots grown by MBE are also shown as references.
The decrease in intensity of the ALS dots following a temperature increase was monotonous
and about one order of magnitude at 300K. Photoluminescence intensity of the SK dots
decreased by almost three orders of magnitude, and there seemed to be a critical temperature
between 150 and 200 K, where the slope inclination changed. Solid and dashed lines indicate
the results of fitting. (Reprinted from Appl. Surfhce Science, vol. 112, Mukai et al., “Growth and
optical evaluation of InGaAs/GaAs quantum dots self-formed during alternate supply of
precursors,” pp. 102- 109, 1997, with kind permission of Elsevier Science-NL, Sara Burgerhar-
tstraat 25, 1055 KV Amsterdam, The Netherlands.)
4 OPTICAL
CHARACTERIZATION
OF QUANTUM
DOTS 205
suggesting that there are at least two sources for the reduction in photo-
luminescence intensity.
The temperature independence of the recombination lifetimes of Fig. 4.19
tells us where the nonradiative centers are located: The centers outside the
dots degrade emission efficiency as temperature increases. Since the excita-
tion wavelength was 647.1 nm, carriers were initially excited outside the dots
and diffused into the dots to emit inside them. Assuming nonradiative
centers outside the dots, we calculated the temperature dependence of
photoluminescence intensity. We neglected the radiative recombination
outside the dots, since no emission was observed from the GaAs substrate
and the In,,, ,Ga,,.,As layer surrounding the dots. The photoluminescence
intensity can thus be described using
where G is the carrier generation rate and zd is the carrier relaxation time
from outside to the ground state of the dots. Since only the emission from
the ground state was observed, we included the intersubband carrier
relaxation time in z d . The lifetime of zz:'(T) is the nonradiative recombina-
tion lifetime outside the dots and is written as
where N is the trap density, V(T) is the carrier thermal velocity

and oout(
(Jm),
T ) is the capture cross-section. We can describe the cross-section
using the activation energy of Q as
(T~"'(T) = CT x exp( - Q/kT) (4.3)
By a change of two parameters, N x o and Q, in Eqs. (4.1) through (4.3),

the measured emission intensity was simulated by the calculation, as shown
by the solid line for the ALS dots and the dashed line for the SK dots in
Fig. 4.19. The carrier relaxation time was set at zd = 1 ns from the results
of time-resolved measurements (see Chapter 5). The parameter set of
[ N x o,Q] = [102cm-', 35meVl explains the results of the ALS dots,
whereas the combination of [lo2 cm- 35 meV] and [lo8 cm-', 280meVl
describes the results of the MBE-grown dots. The former set is common
with the ALS dots; the latter set explains the large bending over 150K. If
we use zd = 100 ps, the values of N x o decrease by one order of magnitude.
Table 4.1 lists the known parameters of nonradiative centers in the
InGaAs system. The parameter set of the ALS dots in calculation is close to
TABLE
4.1
REPORTEDPARAMETERSOF NONRADIATIVECENTERS IN InGaAs MATERIALS. E IS THE ACTIVATION
ENERGYFOR ne THERMAL RATE,B IS THE CAPTURE
EMISSION CROSS-SECTION, Q IS THE
ACTIVATION
ENERGY FOR THE CAPTURECROSS-SECTION, AND N IS THE TRAP DENSITY. “ V P E IS
AN ABBREVIATIONFOR VAPORPHASEEPITAXY. (From Mukai et al., 1997a.)
Growth
E (W B (cm2) Q (meW N(~m-~) Method
a’ 0.82 1.6E-16 2.6E + 13 MBE

b2 0.54 3.1E-16 1.OE + 13 MBE
C‘ 0.21 5.9E-14 7E + 12 MBE
d’ 0.74 1.6E-13 2.1E + 14 MBE
el.2
0.53 5E-15 1.4E + 15 MBE, VPE
f2.3
0.17 2E-12 - 3E + 15 MBE, VPE
g4 0.83 1E-14 80 VPE
h4 0.76 1E-15 80 VPE
i4 0.68 1E-16 40 VPE
‘Bhattacharya, P. K., and Dhar, S. (1988). Semicond. and Semimeta. 26.

20gawa, M., Hongo, S., Watanabe, Y., Sano, N., Katoh, H., Nakayama, M., Ishida, K., and
Shirafuji, J. Sixth Record of Alloy Semicond. Phys. and Electro. Symposium, Hakone, Japan 1987.
30gura, M., Mizuta, M., Onaka, K., and Kukimoto, H. (1983). J p n . J . Appl. Phys. 22, 1502.
4Mircea, A,, Mitonneau, A,, Hallais, J., and Jaros, M. (1997). Phys. Reu. B. B16, 3665.
the sets in Table 4.1. Remarkably the parameter set of the MBE-grown dots
is much different from the sets in Table 4.1, which suggests that the point
defects outside the dots cannot explain the MBE-grown dots’ near-three-
order decrease. Other origins, such as the network of dislocation lines or the
nonradiative center inside the dots should be assumed to explain the large
parameters.
The high emission efficiency of ALS dots can be attributed to their unique
growth process. One possible reason for their better optical quality is that
the quality of their interface is higher. The ALS dots are surrounded
laterally by a quantum-well layer, and the interface around them is ambigu-
ous (see Fig. 4.1). This may be because the atoms around the dots
reconstruct the interface during the growth, given that ALS dots are
supposedly formed through the repetition of an equilibrium point and a
non-equilibrium point of thermodynamical balance between strain energy
and surface energy (see Chapter 3). On the other hand, the interface of the
SK dots is abrupt, as it is exactly the border between the growth islands and
the overgrown layer. Another possible reason for the better optical quality
of the ALS dots is that they are in-situ annealed due to the low growth rate
of the alternate supply of precursors. The growth rate was about 1
4 OPTICAL
CHARACTERIZATION
OF QUANTUM
DOTS 207
monolayer per 10 seconds for the alternate supply -significantly smaller

than that of MBE. The growth temperature of ALS dots is lower than the
temperature where arsenic desorption occurs (Sakuma et al., 1993), and
some defects may have time enough to go out from the surface during the
slow growth.
V. Summary
The optical characteristics of ALS dots were compared with those of SK

dots. Emission from discrete levels caused by three-dimensional quantum
confinement was observed using photoluminescence, electroluminescence,
and photoluminescence excitation. The photoluminescence mapping of a
wafer revealed the compositional fluctuation of indium among quantum
dots. Microprobe photoluminescence revealed the fine structure of emission.
Quantum confinement potential was controlled by changing the number of
supply cycles and the composition of the buffer layer, and was evaluated by
diamagnetic energy shifts. Time-resolved photoluminescence and photo-
luminescence intensity as a function of temperature were studied to evaluate
the nonradiative recombination channel inside and outside the dots.
In comparing ALS dots with SK dots, the 1.3-pm emission wavelength is
the most remarkable advantage of the ALS dots, as it means that they can
be applied to lasers for optical telecommunication systems. Good controlla-
bility of the wavelength of ALS dots is also significant in practical device
manufacturing. The ALS dots achieved a spectral width of 30 meV -less
than half that of the SK dot -and have better emission efficiency.
ALS dots have the drawback of low density and insufficient multiple
stacking (Chapter 3) which prevent the realization of high-performance
1.3-pm lasers. Research on eliminating these drawbacks is not being carried
out.
Optical evaluation suggests that self-assembled dots achieve discrete-state
density and results in unique characteristics not available before. We are
confident that quantum-dot devices will have a significant impact on
optoelectronics applications in the near future. Besides its intrinsic import-
ance, optical characterization is indispensable for the development of
practical devices.
One of the issues we should pursue is the carrier relaxation problem
discussed in Chapter 1 as it relates to quantum-dot laser performance. This
topic will be treated in Chapter 5.
REFERENCES
Ambrose, P. W.. Goodwin, M. P., Martin, C . J., and Keller, A. R. (1994). Phys. Reu. Lett. 72,
160.
Bimberg, D., Ledentsov, N. N., Kirstaedter, N., Schmidt, O., Grundmann, M., Ustinov, M., V.,
Egorov, Y. A., Zhukov, E. A., Maximov, V. M., Kop’ev, S. P., and Alferov, I., 2 . (1995).
Ext. Abs. 1995 Int. Con$ Solid Slate Devices and Materials, Osaka.
Brunner, K., Bockelmann, U., Abstreiter, G., Walther, M., Bohm, G., Trankle, G., and
Weimann, G. (1992). Pliys. Rev. Leu. 69, 3218.
Betzig, E., Trautman, K. J., Harris, D. T., Weiner, S. J., and Kostelak, L. R. (1991). Science 251,
1468.
Betzig, E., Finn, L. P., and Weiner, S. J. (1992). Appl. Phys. Lett. 60, 2484.
Betzig, E., and Chicheser, J. R. (1993). Science 262, 1422.
Chuang, S. L. (1995. Physics ofOptoelectronics Devices, John Wiley & Sons, New York.
Daruka, I., and Barabasi, A.-L. (1997). Phys. Rev. Lett. 79, 3708.
Diirig, U., Pohl, W. D., and Rohner, F. (1986) J . Appl. Phys. 59, 3318.
Grober, D. R., Harris, D. T., Trautrnann. K. J., Betzig, E., Wegscheider, W., Pfeiffer, L., and
West, K. (1994). Appl. Phys. Lett. 64. 1421.
Ghaemi, F. H., Goldberg, B. B., Cates, C.. Wang, D. P., Sotomayor-Torres, M. C., Fritze, M.,
and Nurmikko, A. (1995). Superlattices and Microstructures 17, 15.
Hessman, D., Castrillo, P., Pistol, E. M., Pryor, C., and Samuelson, L. (1996). Appl. Phys. Lett.
69, 749.
Leonard, D., Kishnamurthy, M., Reaves, M. C., Denbaars, P. S., and Petroff, M. P. (1993). Appl.
Phys. Lett. 63, 3203.
Marzin, Y. J., Gerard, M. J., Izrael, A,. Barrier, D., Bastard, G. (1994). Phys. Rev. Lett. 73, 716.
Moison, M. J., Houzay, F., Barthe, F., and Leprince, L. (1993). Appl. Phys. Lett. 64, 196.
Mukai, K., Ohtsuka, N., Sugawara, M., and Yarnazaki, S. (1994). Jpn. J . Appl. Phys. 33, L1710.
Mukai, K., Ohtsuka, N., and Sugawara, M. (1996a). Jpn. J . Appl. Phys. 35, L262.
Mukai, K., Shoji, H., Ohtsuka, N., and Sugawara, M. (1996b). Appl. Phys. Lett. 68, 3013.
Mukai, K., Ohtsuka, N., Shoji, H., and Sugawara, M. (1996~).Phys. Rev. B. 54, R5243.
Mukai, K., Ohtsuka, N., Shoji, H., and Sugawara. M. (1997a). Appl. Surf Sci. 112, 102.
Mukai, K., Ohtsuka, N., and Sugawara. M. (1997b). Ext. Abs. 58th Autumn Meet. Japan
Society of Applied Physics, 4pS-8 (in Japanese).
Mirin, R. P., Ibbetson, P. J., Nishi, K., Gossard, C. A., and Bowers, E. J. (1995). Appl. Phys.
Lett. 67, 3795.
Nakata, Y., Sugiyama, Y., Futatsugi, T.. and Yokoyama, N. (1997a). J . Crystal Growth 175/176,
713.
Nakata, Y., Sugiyama, Y., Futatsugi, T., Mukai, K., Shoji, H., Sugawara, M., Ishikawa, H., and
Yokoyama, N. (1997b). Ext. Abs. 58th Autumn Meet Japan Society of Applied Physics,
2pM-4 (in Japanese).
Ohtsuka, N., and Mukai, K. (1995). Proc. Int. Con$ InP and Related Materials, Hokkaido.
Sakuma, Y., Ozeki, M., and Nakajima, K. (1993). J . Cryst. Growth 130, 147.
Saiki, T., Mononobe, S., Ohtsu, M., Saito. N., and Kusano, J. (1996). Appl. Phys. Lett. 68,2612.
Sugawara, M. (1996). Jpn. J . Appl. Phys. 35, 124.
Sugawara, M., Okazaki, N., Fuji, T., and Yamazaki, S. (1993). Phys. Rev. 8. 48, 8102.
Tabuchi, M., Noda, S., and Sasaki, A. (1992). Science and Technology of Mesoscopic Structure,
Springer-Verlag, Tokyo, p. 379.
SEMICONDUCTORS AND SEMIMETALS, VOL 60
CHAPTER5
The Photon Bottleneck Effect

in Quantum Dots
Kohki Mukai und Mitsuru Suguwara
OPTICAL SCMlCONDLlrTOll &VICES LABORAT(1RY
LABORATORIES
FUJITSU LTI,.
ATSUGI,KAHAGAWA.
JAPAN
I. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
11. A MODELOF THE CARRIER RELAXATION PROCESS IN QUANTUM DOTS . . . . . 21 1
111. EXPERIMENTS ON LIGHTEMISSION AND CARRIER RELAXATION I N QUANTUM-DOT
DISCRETE ENERGY LEVELS . . . . . . . . . . . . . . . . . . . . . . . 2 14
1. Elertroluminescence Spectra . . . . . . . . . . . . . . . . . . . . . 2 15
2. Time-Resolved Photoluminescence . . . . . . . . . . . . . . . . . . 2I I
3. Simulation of Electroluminescence Spectra . . . . . . . . . . . . . . . 225
IV. INFLUENCEOF THERMAL TREATMENT . . . . . . . . . . . . . . . . . . . 229
1. Change in Emission Spectra after Annealing . . . . . . . . . . . . . . 229
2. Competition between Carrier Relr.rution and Recombination . . . . . . . 231
V. SIMULATION OF LASERPERFORMANCE. INCLUDING THE AUGERRELAXATION
PROCESS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
VI. SUMMARY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
I. Introduction
The physics of carrier relaxation in quantum dots has been intensively

studied ever since it was discovered that quantum dots held promise for
optoelectronic device applications (Arakawa and Sakaki, 1982, 1984; Brus,
1986; Chemla and Miller, 1986; Asada, 1986). Nevertheless, many re-
searchers feared that carrier relaxation into the quantum-dot discrete levels
would be significantly slowed due to a lack of phonons to satisfy the energy
conservation rule. This is the so-called phonon bottleneck problem. If we form
well-separated discrete energy levels by quantum dots so as to increase laser
performance, carrier relaxation into the levels may be much retarded, and
therefore laser performance may not be improved. As we saw in Chapter 1,
209
Cop)righl 1 1Y99 hy AcddcmiL Preas
All ri$it\ of reproductioii in m y Corm re\erved
ISBN o - I z - ~ Q I ~ ~ . ~
ISSN ooxn-x7x4 99 m o o
the carrier relaxation lifetime dominates the performance that can be

expected for quantum-dot lasers.
Theoretical studies on carrier relaxation due to phonon scattering show
that the relaxation lifetime of the dots reaches nanoseconds or longer unless
the requirements of energy conservation are strictly satisfied (Bockelmann
and Bastard, 1990). This long relaxation lifetime, which is comparable to the
radiative or nonradiative carrier recombination lifetimes in semiconductors,
was thought to significantly degrade radiative efficiency (Benisty, 1991;
Benisty et al., 1995; Bockelmann, 1993), because carriers are consumed
through radiative and nonradiative channels on their way to the ground
level (Fig. 5.1).
Yet there are also more optimistic theoretical results. An auger process
has been suggested for a possible fast relaxation channel with a lifetime of
ten fewer picoseconds (Bockelmann and Egeler, 1992; Efros et al., 1995;
Gerard, 1995). Inoshita and Sakaki (1992) showed that two phonon
processes, including LO and LA phonons, reduce the requirements of energy
conservation. And Nakayama and Arakawa (1994) solved the time-depend-
ent Schrodinger equation including an electron-LO-phonon interaction
Hamiltonian for the relaxation problem. They noted a margin of a few tens
meV for the energy conservation rule due to the time-energy uncertainty,
and predicted a relaxation time of less than one picosecond.
Energy
.........................
I
I Nonradiative
I channel
I
I
I
...
(Defect level)
I
-Y.--
Radiative
channel
FIG.5.1. Schematic of the phonon bottleneck. Discrete sublevels in quantum dots may
hinder carrier relaxation toward the ground level. In this case. carriers will recombine to emit
at higher levels or go to the nonradiative channel during a carrier cascade toward the ground
level.
5 PHOTON EFFECTIN QUANTUM
BOTTLENECK DOTS 211
The recent discovery of self-assembled quantum dots composed of various

semiconductor materials has enabled us to evaluate the carrier relaxation
lifetime experimentally (Eaglemann and Cerullo, 1990; Tabuchi et al., 1992;
Ahopelto et al., 1993; Leonard et al., 1993; Mukai et al., 1994; Oshinowo et
al., 1994; Moison et al., 1994; Apetz et al., 1995; Utzmeier et al., 1996).
Retarded carrier relaxation, with a lifetime of several tens to hundreds of
picoseonds, has been observed in many experiments using time-resolved
photoluminescence (Mukai et al., 1996b; Yu et al., 1996; Adler et al., 1996).
Some of these experiments confirmed the importance of phonon resonant
excitation conditions in accelerating relaxation (Raymond et al., 1995;
Vollmer et al., 1996). Fast relaxation through the Auger process reportedly
manifests itself in the fast rise of time-resolved photoluminescence and
during the initial stage of carrier relaxation, when a large number of carriers
surround the quantum dots (Ohnesorge et al., 1996, Bockelmann et al.,
1997). While more detailed theoretical and experimental work is needed to
reach precise conclusions about relaxation mechanisms, we are convinced
that the phonon bottleneck does exist, in the sense that relaxation is slower
in quantum dots than in quantum wells, where a typical intersubband
relaxation lifetime is 0.1 to 1 ps (Asada, 1989).
In this chapter, we present our findings on carrier dynamics in our self-
assembled InGaAs/GaAs quantum dots grown by the alternate supply
(ALS) process (Mukai et a]., 1994; see Chapter 3). The dot’s high crystal
quality, which features narrow spectrum broadening and high emission
efficiency, enabled us to pursue the problem. After a brief discussion of
possible carrier relaxation processes into quantum dots, we analyze electro-
luminescence and time-resolved photoluminescence data to provide recom-
bination and relaxation lifetimes. Next we introduce our experiments on
annealed quantum-dot samples to demonstrate the influence of retarded
carrier relaxation on emission spectra. Finally, we simulate laser per-
formance by adding an Auger relaxation process to the rate equations in
Chapter 1. The results in this chapter will clarify how the phonon bottleneck
effect works in quantum dots.
11. A Model of the Carrier Relaxation Process in Quantum Dots
The carrier relaxation process into quantum dots is actually two pro-
cesses, as shown in Fig. 5.2. One is the carrier relaxation from continuous
energy levels into quantum-dot discrete levels (A). The other is the relax-
ation between the discrete levels inside dots (B). In many optical experi-
ments as well as in quantum-dot lasers, carriers undergo these two processes
212 KOHKIMUKAI
AND MITSURUSUGAWARA
FIG. 5.2. Two cases of carrier relaxation in a quantum dot: (A) the relaxation from
continuous levels and (B) the relaxation between discrete levels. It is expected that case (A) will
occur faster than case (B) since the energy conservation rule is more severe for the latter case.
except when they are brought directly into the discrete levels by resonant
excitation or tunneling (Kamath et al., 1997a). Since the energy conservation
rule must be satisfied for carrier relaxation, relaxing carriers transfer a
corresponding energy to other particles such as phonons and to other
carriers. Thus, the relaxation rate strongly depends on the density of final
levels and on the number of particles other than the transition matrix
elements.
Figure 5.3 illustrates the two representative processes: (a) a single- or
multiphonon process, and (b) an Auger process. Inoshita and Sakaki (1992)
showed that two-phonon processes, including LO and LA phonons, reduce
the requirements of energy conservation, thus drastically decreasing the
lifetime to subpicosecond order. Even so, this relaxation process cannot send
carriers deep inside the dots, since the transition matrix for the relaxation
processes that include more than three phonons will be greatly reduced.
Therefore, carriers once trapped in the upper levels of the dots relax into
lower energy levels by emitting phonons step by step. The Auger process of
Fig. 5.3(b) works effectively when there are many carriers, since an electron
can find another electron or a hole into which to transfer its energy and thus
fall into the dot energy levels. This process works to further relax carriers
trapped in the discrete levels into lower energy regions, as long as there is
still a larger number of carriers outside the dots. For that reason the Auger
process is more effective for carrier relaxation into deep sublevels.
Figure 5.4 shows an experimental study of the rising characteristics of
photoluminescence for SK dots as a function of excitation power (Ohnes-
orge et al., 1996). The rising time depends on the optical excitation density.
For an excitation density between lo-' and 4, rising time stays almost
5 PHOTON BOTTLENECKEFFECT
IN QUANTUM
DOTS 213
(a) Multiphonon process
Shallow sublevel Deep sublevel
High carrier density Low carrier density
FIG. 5.3. Schematic of two processes of carrier relaxation from continuous levels: (a) the
multi-phonon process and (b) the Auger process. Usually, the energy separation between the
continuous level and the sublevel in a dot does not exactly match the energy of a single phonon.
There can be a combination of LO and LA phonons during relaxation from the continuous
energy level, but this is unlikely for deep sublevels. The Auger process is effective when carrier
density is high.
constant at 90ps. An increase in the excitation density to lo2 W/cm2

achieves a rising time of 40 ps. The authors claimed that the multi-phonon
process is dominant while the rising time is constant; the decrease of rising
time is due to the Auger process.
In time-resolved photoluminescence, a large number of carriers, many
more than the number of dots, are excited by a short intense laser pulse.
Then, within a t least several tens of picoseconds, most carriers are captured
to occupy the quantum-dot levels, primarily via the Auger process. This is
followed by a carrier relaxation process in individual dots toward lower
energy levels via phonon emission. In photoluminescence and electro-
luminescence, where there are fewer carriers than quantum dots, carrier
injection and relaxation into dots is dominated by phonon emission. As a
result, carriers are first captured by upper energy levels and then relax as if
they were going down a ladder of discrete energy levels. The capture process
occurs at a relatively high rate, since the transition is from continuous to
discrete levels and the energy conservation requirements are easily satisfied
by the multi-phonon process. However, the intersubband relaxation is
SUGAWARA
120 4
I
T=5K
100 I
e 80
.-
a 40
0
.. 0
2oL I ' ’’mns-s’ ' ' ' 1 4 1 1 6 1 ' ' ' s m m a m ' 8 1
lo-’ 100 10' 102

Excitation density (W/cm')
FIG.5.4. Experimental study of rising characteristics by Ohnesorge et al. (1996). The
authors showed that the rising time depended on the excitation density in the SK dots. They
reported that about 4 W/cmZwas the critical excitation density for the Auger carrier relaxation
process (from Ohnesorge et al., 1996. Copyright 1996 by the American Physical Society).
significantly slowed as long as the energy separation closely matches the

phonon energies. During laser operation, at a low injection level even above
the threshold, carrier relaxation relies on the ladder process. As the injection
level increases, the number of carriers around the dots increases, making the
Auger process effective for carrier relaxation.
111. Experiments on Light Emission and Carrier Relaxation in

Quantum-Dot Discrete Energy Levels
An experimental clue to the phonon bottleneck problem was given in the

electroluminescence spectra of Fig. 4.7 in Chapter 4, where well-resolved
multiple-peak light emissions from discrete energy levels were observed in
the ALS dots. We increased the current injection and found that well before
the emission from the ground level is saturated, a higher energy emission
occurs that exceeds the ground-level emission. The carrier distribution
between the discrete levels in no way follows the Fermi-Dirac distribution
for the measured temperature of 77 K. The most probable explanation for
this is that the carrier relaxation lifetime is comparable to the carrier
recombination lifetime.
5 PHOTON BOTTLENECKEFFECTIN QUANTUM
DOTS 215
After discussing the electroluminescence spectra in a little more detail, we

will evaluate the carrier relaxation and recombination lifetimes in ALS
quantum dots by time-resolved photoluminescence. The emission decay
curve from the excited levels is found to be double exponential, while the
decay from the ground level follows a single exponential curve. The decay
curves are analyzed by a model that takes into account a rapid initial carrier
capture process and the independence of each quantum dot. This model
enables us to obtain the relaxation lifetime as well as the recombination
lifetime. The unique electroluminescence spectra are well simulated by the
determined lifetimes.
1. ELECTROLUMINESCENCE
SPECTRA
Figure 5.5 shows the electroluminescence spectra for (a) 300 K and (b) 77 K.
To fabricate diode structure, a 0.5-pm n-GaAs layer, a 1-pm n-InGaP layer,
an ALS-dot layer sandwiched by 100-nm GaAs layers, a l-pm p-AlGaAs
layer, and a 0.5-pm p-GaAs layer were grown on a (001) GaAs substrate.
The luminescence parallel to the sample surface was dispersed and detected
with a lock-in technique using an InGaAs photo-multi detector that was
kept at -70°C. The electrode size was 20 x 900pm’. The current-
injected area was estimated by the near field pattern, and according to this
estimation, it was spread out over 150 to 200 x 900pm2 in the quantum-dot
layer. Three emission peaks corresponding to the interband transition
between the discrete levels of the conduction and valence bands appeared at
300 K when the injected current was increased, five peaks appeared at 77 K.
In Chapter 4, we pointed out the phonon bottleneck from the emission
spectra at 77 K. We should note that, at lOmA, the 2nd and 3rd peak
emissions exceeded the first peak emission much earlier than the first
emission intensity saturated, indicating that the relaxation rate into the
ground level is comparable to the spontaneous emission rate of the excited
levels.
The electroluminescence spectra of Fig. 5.5 varied between the two
temperatures, which indicates that some features of carrier dynamics depend
on temperature. For example, when the emission intensity of the ground
level reached half maximum, the intensity of the second level was stronger
than that of the ground level at 77 K (10 mA), but the two were almost equal
at 300 K (50 m A).
Random carrier capturing into the dots was proposed by Grundmann
and Bimberg (1997) as an alternative interpretation for the multiple
emission peaks (Mukai et al., 1996b; Lipsanen et al., 1995). They calculated
the emission spectra using a Poisson distribution to count the number of
SUGAWARA
0.8 0.9 1.0 1.1 1.2 1.3

Energy (eV)
(a>
l""l""l""l""
- Measurement
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 * 1 1 1
0.9 1.0 1.1 1.2 1.3 1.4

Energy (eV)
(b)
FIG. 5.5. Electroluminescence spectra at (a) 300 K and (b) 77 K. As the injected current was
increased, five discrete levels appeared. Higher-level emissions appeared before the emission
intensity of the lowest level reached its maximum value. Considering that the kT value is much
narrower than the broadening of the spectra, these figures suggest the existence of a phonon
bottleneck that impedes carrier relaxation (from Mukai et al., 1996b. Copyright 1996 by the
American Physical Society).
5 PHOTON
BOTTLENECK
EFFECTI N QUANTUM DOTS 217
carriers that were trapped in quantum dots. As a result, they claimed that,
even if the retarded carrier relaxation is not taken into account, the emission
from high energy levels appears even before the lower energy level emission
is saturated. However, their calculation of the emission spectra cannot
explain the emission spectra we measured, where the third-level emission
exceeded the ground-level emission before the ground-level intensity
saturated. Above all, the random capture model offers no explanations for
the difference in the emission spectra between 77 K and 300 K.
PHOTOLUMINESCENCE
2. TIME-RESOLVED
To evaluate carrier relaxation and recombination dynamics, we measured

the time-resolved photoluminescence of ALS quantum dots. The dots were
grown by 18 cycles of In-Ga-As alternate supply onto a 500-nm GaAs buffer
layer on a (001) GaAs substrate and capped b y a 100-nm GaAs layer. The
growth temperature was 460°C. Samples set in a temperature-controlled
cryostat were excited by a mode-locked 532-nm Nd:YAG laser beam with
a pulse width of 200 ps and a repetition rate of 82 MHz. Luminescence from
the sample surface was dispersed by a monochrometer and time-resolved
with a streak camera. Figure 5.6 shows typical experimental results of
luminescence intensity decay for up to the fifth emission peak at 20 and
250K. The order of emission peaks corresponds to that in the electro-
luminescence spectra of Fig. 5.5(b).
Various features can be noted in the decay curves. First, the luminescence
decays faster as the order of emission peaks gets higher. Similar results were
obtained by Raymond et al. (1996). Since the spontaneous emission lifetime
due to the interband transition in quantum dots is common for each energy
level if we assume a constant oscillator strength (Chapter l), the observed
variation of the decay curves suggests that the carrier relaxation lifetime
decreases with increasing order. Second, the emission decay is nonexponen-
tial, which is probably due to the fact that the decay is a combination of
relaxation and recombination components. Third, the temporal evolution of
the decay is very fast during the initial stage, but becomes eventually very
slow at the end (Mukai et al., 1996b; Adler et al., 1996; Yu et al., 1996). This
is in contradiction to the calculated curves, which are moderate at the
beginning but become fast toward the end (Grundmann and Bimberg, 1997;
Grosse et al., 1997). Fourth, the decay is faster at 250 K than at 20 K, which
indicates that relaxation and/or recombination lifetime is shorter at higher
temperatures.
To analyze the decay curves, we introduce a new model that we call
random initial occupation (RIO), which is appropriate for analyzing time-
I " " " " " " " ' 1
0 2 4 6 8
Time (ns)
FIG. 5.6. Example of luminescence decay of the five levels at 250 and 20K. The higher the
temperature, the faster the decay; the lower a level, the slower the decay (from Mukai et al.,
1996b. Copyright 1996 by the American Physical Society).
resolved photoluminescence data and is based on the following two points.

First, assuming that all carriers relax into the quantum-dot discrete levels
during pulse excitation, we start the calculation of the decay curve at the
point where levels inside the dots are occupied by carriers. This is reasonable
because there are a large number of carriers within 10" to 1013cm-2
around the dots during the short pulse excitation, and carrier-carrier
scattering causes rapid relaxation into dots through the Auger process.
Second, we take into account that the carrier capture process occurs
randomly and that the number of carriers and their distribution in discrete
levels differs from dot to dot.
Taking into account only the two electron levels of ground and excited
for reasons of simplicity, the initial carrier distribution in dots can be
categorized into the four types shown in Fig. 5.7(a): A, both levels are filled;
B, only the excited level is filled; C , only the ground level is filled; and D,
both levels are empty. The temporal evolution of this system is described in
the figure. Case D is the final situation of every carrier process. When a dot
is initially in case A, three paths-A,, A,, and A,-are possible for
BOTTLENECK DOTS 219
A B C D
[E]I?[I I
a I
DI
I I I I
t l I I I I j
3
0
> NBO
3
0
NAO
c
.3
D
k
E
1
c
I I I
0 1.0 2.0 3.0 4.0 5.0

(b) Time (ns)
FIG. 5.7. (a) Model for the carrier dynamics between two levels in an uncoupled dot. When
the carrier injection from continuous levels into sublevels is sufficiently fast, all possible initial
situations can be classified as four cases, A, B, C, or D. The decay in the carrier number in the
excited level is described by considering paths A , , A,, A,, B,, and B,. ( b ) Calculated number
of carriers in the excited levels as a function of time. The solid line shows N(r), the dashed lines
show N,(t) and N,(t). The figure shows a double exponential decay line, which matches the
measured results (Mukai et al., 1996; Adler et al., 1996). The faster decay indicates a carrier
process that begins at case B, and the slower decay indicates one that begins at case A (from
Mukai et al., 1998).
reaching the final destination of D. Moreover, two paths, B, and B,, are
possible for case B, and a single path, C,, is possible for case C. Radiative
decay is allowed in both levels on the assumption that hole quantization
energies are less than k,T due to the comparative heavy mass of holes
(Benisty et al., 1991).
The carrier number in the excited level is counted by evaluating the paths
beginning from case A and the paths beginning from case B. Let NAA(t)be
the number of A at time t; and NAB@),the number of B formed from A as
a result of the recombination of the ground-level carrier. The carrier number
in the excited level in dots when the initial situation is A is described by
Here, the rate equation for N A A ( t )is
where z, is the carrier recombination lifetime, which is assumed to be

common between the two levels, Considering both paths A , and A,, the rate
equation for NAB(t)is given by
where zreLis the carrier relaxation lifetime from the excited to the ground
level. Solving Eqs. (5.2) and (5.3), we obtain
(5.4)
where NAo is the initial number of case-A dots. When zrel<< z,, NAB(t)
becomes zero and the decay lifetime of N A ( t ) reduces to 2 , / 2 . When zrel is
not negligible compared with z, the decay of NAB(t)is not negligible and
the decay lifetime of NA(t) is larger than r r / 2 .
For the carrier number in the excited level when the initial situation is B,
BOTTLENECK DOTS 221
NB(t), the rate equation is
and we obtain
where N,, is the initial number of case B. The decay lifetime of NB(t) is
(T~;: +
~ ~ - ‘ ) -and
l gets closer to sret when z~, << T ~ .As a result, the total
number of carriers in the excited level is given by
Ne.x(t) = N A ( t ) + NB(t) (5.8)

The emission intensity is given by Ne,(t)/z,.
The measured double-exponential decay curves in Fig. 5.6 are reproduc-
ible by this type of calculation. The solid curve in Fig. 5.7(b) shows the
decay in the number of carriers in the excited levels as calculated by Eq. (5.8)
with T~ = 3 ns, t r e=
l 200 ps, and NAo = N,,/8 as parameters. The resultant
solid curve consists of two dashed lines: steep, reflecting the decay starting
from case B (NB(t));and moderate, reflecting the decay starting from case A
(NA(t)).The critical point when the inclination of the line changes depends
on the ratio of NAo/NB,,.For N,, > NAo, we get a decay curve with fast and
slow components.
The carrier number in the ground level is also calculated by considering
the paths A,, A,, B,, and C,. As for the excited level, let us define the carrier
numbers NABC(t),NAC(t),NBc(t),and N,(t) as shown in Figs. 5.7(a). The rate
equations for these are
(5.10)
(5.11)
and
(5.12)
222 KOHKIMUKAI
AND MITSURUSUGAWARA
Solving Eqs. (5.9) through (5.12), we obtain
(5.13)
(5.14)
(5.15)
f
Nc(r) = Nc,ept (5.16)
where N,, is the initial number of carriers in case C. Finally, the total
number of carriers in the ground level is given by
When treI << t,, Eqs. (5.13) through (5.15) become zero and the decay time
of Ng,(t) reduces to t,. In the same case, we saw that the decay lifetime of
NA(t) reduces to 2,/2. However, in Fig. 5.6, the later moderate decay in the
second level is not half of the decay in the ground level. This indicates that
the relaxation lifetime is not negligible compared with the recombination
lifetime and/or that the recombination lifetime may not be strictly uniform
among the sublevels.
We analyzed the measured decay curves of Fig. 5.6 by the double-
exponential formulation of
(5.18)
after deconvoluting the decay curves from the excitation pulse. The second
term of Eq. (5.18) is of course neglected in considering the decay of the
ground level. Here, from the estimation by the RIO model we assumed the
initial steep decay component to be dominated by the carrier relaxation and
the subsequent slow decay component to be by the recombination. In strict
treatments, the degeneracy and the selection rule in relaxation for the
multiple sublevels must be taken into account in solving the rate equations.
However, Eq. (5.18) is a good approximation because the initial situation of
the dots is roughly categorized into two cases: when the lower level is empty
and when the lower level is filled.
5 PHOTONBOTTLENECK
EFFECTIN QUANTUM
DOTS 223
The results of the analysis are plotted in Fig. 5.8 as a function of

temperature. The recombination lifetime in Fig. 5.8(a) is independent of level
order and temperature. The relaxation lifetime of Fig. 5.8(b) varies among
the levels and is temperature-dependent. The nano-second recombination
lifetimes and their temperature independence match a theory for the
spontaneous emission of electron-hole pairs in quantum dots that was
discussed in Chapter 1. Similar experimental results were obtained for
Y
.3
4
;10
0
Y
.C
m
c
._
0 50 100 150 200 250 300

Temperature (K)
(4
10p-
0 SO 100 150 200 250 300
Temperature (K)
(b)
Re 8 ,
FIG. 5.8. Carrier lifetimes of sublevels as a function of temperature: (a) recombination
lifetime and (b) relaxation lifetime. The recombination lifetime is independent of temperature
and about 1 ns in all sublevels. The higher the temperature and the higher the order of a level,
the faster the carrier relaxation (from Mukai et al., 1996b. Copyright 1996 by the American
Physical Society).
InGaAs/GaAs SK dots by Wang et al. (1994) and Kamath et al. (1997b)

and for InP/InGaP SK dots by Kurtenbach et al. (1995). As the order of
levels rose from the second to the fifth, the relaxation lifetime decreased by
one order of magnitude. This is probably because the density of levels
increases with the level order, as seen in the electroluminescence spectra,
which increases the probability that carriers will relax into the lower empty
levels. The decrease in relaxation lifetimes following a temperature decrease
can be attributed to the increase in the number of phonons following the
Bose distribution function.
Let us check what the decay curves look like if we assume that carriers
relax from the continuous level to the topmost fifth level and then relax step
by step. The rate equations for this case are
(5.19)
(5.20)
and
(5.22)
where T,, is the recombination lifetime in the continuous level, z, is the

relaxation lifetime from the continuous level to the topmost fifth level; and
7i denotes the relaxation lifetime from the ith sublevel to the (i - 1)th
sublevel. Using the occupation factor of fi = Ni/Di with the level density of
D i , we get
where zi0 is the intrinsic carrier relaxation lifetime at the ith level. Using the
lifetimes in Fig. 5.8, we calculated the temporal evolution of the carrier
number in the sublevels (Fig. 5.9). To estimate the fastest rise under
step-by-step relaxation, we assumed ,h = 0 in the calculation and found that
the delay of rising at the lower-ordered level cannot be neglected. The delay
is about 2 ns between the fifth and the ground level, and is approximately
equal to the simple sum of the relaxation lifetimes of the sublevels.
5 PHOTON
BOTTLENECK
EFFECT
IN QUANTUM
DOTS 225
I"
0 0.5 1 1.5 2
Time (ns)
FIG. 5.9. Calculated rise of the carrier number when the direct carrier relaxation from the
continuous level is negligible. The rise slowed as the level order decreased. The delay is
approximately equal to the simple sum of the relaxation lifetimes of the sublevels.
Comparing the experimental results of Fig. 5.6 and the calculated results of
Fig. 5.9, it is clear that the assumption of the step-by-step relaxation process
cannot explain the fast rising of photoluminescence in the time-resolved
measurements. The RIO model, which assumes fast carrier setup in quan-
tum dots resulting from carrier-carrier scattering and the subsequent slow
decay inside the dots, successfully explains the decay curves of time-resolved
photoluminescence.
3. SIMULATION OF ELECTROLUMINESCENCE
SPECTRA
Let us simulate the electroluminescence spectra of Fig. 5.5 using the

evaluated relaxation and recombination lifetimes. Note that the carrier
number around the quantum dots is estimated to be about two to three
orders of magnitude smaller than that in time-resolved photoluminescence
measurements. Thus, we employ the step-by-step relaxation model, where
the carriers first relax into the topmost level and then relax into the low
energy levels. We ignore the thermal excitation of electrons, since the level
spacing is much larger than the value of kBT
The rate equation for the carrier number in sublevels during electro-
luminescence measurement is obtained from Eqs. (5.20) through (5.22) by
assuming that the differential of time is zero,
N5
- +-
N5 - G = O , (5.24)
55 ‘r
Ni N . N , , ,
-+>--- - 0 (i = 2, 3,4), (5.25)
Ti zr Ti+ 1
and
(5.26)
where G is the carrier injection rate from the continuous level into dots.
Here, we consider that the recombination lifetime, zr, is common to all
levels, as seen in Fig. 5.8(a), and disregard nonradiative recombination in
our high-quality samples (Chapter 4).Assuming a Gaussian distribution for
the broadening of the emission spectrum, we calculated the luminescence
spectra by
(5.27)
where E is the energy of the photon, E , is the energy at the ith level, and r
is an inhomogeneous broadening factor. Figures 5.10(a) and 5.10(b) show
the results of the calculation for 300K and 77K, respectively. During
calculation, we used an in-plane dot density of 5%, a broadening factor of
50meV, T~ = 1 ns, and D i= 2 . i ( i = 1 to 5), taking the spin into consider-
ation. The intrinsic relaxation lifetimes used were zo2 = 400, zo3 = 200,
T~~ = 50, and z05 = 20ps for 300K; and T~~ = 1000, T~~ = 500, T~~ = 200,
and z 0 5 = 90 ps for 77 K.
The calculations of Fig. 5.10 explain the measured electroluminescence
spectra of Fig. 5.5. The relaxation lifetime is short enough that the ground-
level emission is dominant for small currents. As the number of injected
carriers increases, the carrier relaxation rate decreases and emission from
higher levels occurs. Since the relaxation lifetime is closer to the radiative
recombination lifetime at 77K than at 300K, the emission from higher
levels at 77 K occurs for a smaller current. At 77 K, the calculated current
values match the measured values. At 300K, the trend of relative peak
intensities is well explained, but the values for the current by the calculation
are much smaller than the measured values. This discrepancy grows larger
as the current increases- a tendency that can be attributed to the consump-
5 PHOTON BOTTLENECK DOTS
EFFECTIN QUANTUM 221
0.9 I0 1.1 1.2 1.3 1.4
Energy (eV)
(b)
FIG. 5.10. Calculated electroluminescence spectra at (a) 300 and (b) 77 K using the rate
equations of the five levels, The results of the calculation correlate with the measured spectra
in 5(a) and (b) (from Mukai et al., 1996b. Copyright 1996 by the American Physical Society).
tion of carriers due to nonradiative recombination via defects outside the

dots (see Chapter 4), and to a rise in local temperature due to the non-
radiative process -leading to further acceleration of nonradiative recom-
bination.
Figure 5.11 shows the values of f;. as a function of the current for (a)
300K and (b) 77K, respectively. Note that the ground, second, and third
0 50 100 150 200 250 300

(a> Current (mA)
Current (mA)
(b)
FIG. 5.1 1. Value of f; as a function of injected current for (a) 300 K and (b) 77 K. The
ground, second, and third levels were not completely filled, and the maximum occupation factor
of these levels depended on the temperature.
levels are not completely filled and that the characteristic features depend
on temperature. For example, the maximums of f i for the ground and the
second levels are about 80% at 300 K and about 60% at 77 K. For the third
level, these maximum are 90% at 300 K and 80% at 77 K. While the
emission from the ground level is dominant, the maximum occupation factor
at the ground level is 60% for 300 K. In quantum-dot lasers, the suppression
of carrier filling may limit gains needed for lasing at the ground level (Shoji
et al., 1995).
BOTTLENECK DOTS 229
IV. Influence of Thermal Treatment
The electroluminescence spectra with a series of emission peaks are

dominated by the balance between carrier relaxation and recombination
rates, as explained in the previous section. Here, we introduce an interesting
example of this effect found in annealed quantum dots.
1. CHANGE
IN EMISSION
SPECTRA
AFTER ANNEALING
Annealing of ALS quantum-dot samples was performed in a reactor tube

of a liquid-phase epitaxy system between 500 and 68OoC, in a pure N, gas
atmosphere so as to protect the samples from oxidation. The quantum dots
were originally grown at 460°C using the growth sequence of 18 cycles of
In-Ga-As (see Chapter 3). A GaAs plate was placed on a sample surface to
raise the arsenic vapor pressure and thus prevent arsenic desorption from
the sample surface.
Figure 5.12 compares the photoluminescence spectrum of an as-grown
sample with the spectra of samples after annealing at 615°C for 30 and 60
""'""'"""""""'1
h
Y
.r(
2
Y
E
H
1.o 1.1 1.2 1.3 1.4 1.5

Wavelength (pm)
FIG. 5.12. Photoluminescence spectra before and after annealing at 615°C for 30 and 60
minutes. The emission from the sublevels is observed. The integrated photoluminescence
intensity decreased after annealing. It is remarkable that the emission intensity of the ground
level decreased several times over that of the second level (from Mukai et a!., 1998).
minutes. In the as-grown sample, three peaks can be observed that are due
to the interband transition between the discrete levels of the conduction and
valence bands. The excitation power is very low, as the emission intensity of
the ground level in Fig. 5.12 is less than one-fifth of the maximum value
reached at a higher excitation. Annealing greatly reduces the integrated
photoluminescence intensity, broadens the spectra, and slightly moves the
emission peak toward higher energies. What is most remarkable is that the
relative peak intensity between the ground- and second-level emission is
greatly changed after annealing: While the ground-level emission is about
three times stronger than the second-level emission in the as-grown sample,
the two peaks are almost comparable in their intensity after 30-minute
annealing. This is not due to drastic structural changes and the resultant
reduction of the transition oscillator strength, since the emission energy shift
is only about 5 meV and the TEM image of the 30-minute-annealed samples
shows no significant changes in dot structure, size, and numerical density.
One possible explanation for this phenomenon is that the carrier dynamic
process in quantum dots is influenced by the thermal treatment, which leads
to the change in the relative intensity.
Figure 5.13 shows the energy of the ground- and second-level emission
peaks, as well as the relative emission intensity of the ground level against
Annealing temperature (iC)

FIG. 5.13. Emission intensity of the ground level divided by that of the second level
and emission energy of the ground and second levels as a function of annealing
(Iground/12nd)
temperature. At over 600"C, the energy shift occurred, and the relative intensity of the ground
level was greatly reduced after annealing. The energy shift was larger in the second level up to
650°C (from Mukai et al., 1998).
5 PHOTONBOTTLENECK
EFFECT DOTS
IN QUANTUM 231
the second level, as a function of the annealing temperature. The energy shift
occurs at over 600°C because of interdiffusion, and simultaneously the
relative intensity of the ground-level emission is greatly reduced. The figure
also shows that the energy shift is larger in the second peak and that the
level separation is increased by annealing. This contradicts our experiences
in the study of interdiffusion in quantum wells, where the level separation
always decreases after interdiffusion, which spreads the confinement poten-
tial in the growth direction. This unique feature can be attributed to
structural asymmetry between the in-plane and perpendicular directions in
ALS quantum dots with a disk-like shape because of the asymmetry, the
interdiffusion in the early stage occurs faster at the edge of the dot than in
the central region (Sallese et al., 1995). Since this corresponds virtually to
the decrease in the in-plane diameter of the dots, the energy separation
increases.
2. COMPETITION BETWEEN CARRIER RELAXATION

AND RECOMBINATION
Figure 5.14 shows examples of time-resolved photoluminescence of the

ground- and second-level emission after annealing at 615°C for 60 minutes.
The decay time of the ground-level emission decreases as the temperature
rises. This is in contrast to the decay times of as-grown samples, which are
insensitive to temperature (Fig. 5.8(a)). We see that the decay curves of
the second level are double-exponential, a result of the competition between
the relaxation into the ground level and the recombination at the second
level, as seen in Section 111. As the temperature increases, the relaxation
accelerates, but the recombination lifetime is insensitive to temperature.
Figure 5.15 shows (a) recombination and (b) relaxation lifetimes up to the
third level before and after annealing. The solid and open symbols indicate
the lifetimes of as-grown and annealed samples, respectively. The third level
in the annealed sample is not observed clearly because of emission broaden-
ing. The recombination lifetimes at the ground level become temperature
-dependent after annealing and decrease to a few hundred picoseconds near
room temperature. This suggests that the nonradiative channel affects only
the ground level. The nonradiative lifetime for the ground level is estimated
to be 250 ps if we use the recombination lifetime of 200 ps and the radiative
lifetime of 1 ns. The relaxation lifetimes decrease to about half, and it is not
clear whether this decrease is significant since, following the RIO model we
used in the analysis, the faster decay time is not a pure carrier relaxation
lifetime. However, we should note the possibility that the nonradiative levels
assist carrier relaxation, which was predicted by Sercel (1995).
232 KOHKIMUKAl AND MITSURUSUGAWARA
6 ' I I " I " ' ' 1 ' I ' ' I " ' " " I ' I ' ' ' 1 ' ' I " 'g
h
3
'a1
d
W
2
5
.H
Ei
B
E
Y
0 2 4 6 8 10 12 14 16 18 20
Time (ns)
FIG.5.14. Examples of decay curves of the ground and second levels after annealing at
615°C for 60 minutes. The decay time of the ground level decreased as the temperature rose,
but was insensitive to temperature in as-grown samples. The decay curves of the second level
were double-exponential(from Mukai el al.. 1998).
We consider several possibilities to explain why the nonradiative channel

affects only the ground level. One is that carrier trapping may require some
phonon scattering and the probability of scattering is low for the higher
levels. Another is that defect levels outside the dots, resonant in energy with
the ground level, are the origin of the nonradiative channel (such resonant
carrier trapping was reported by Tsang (1978)).
The photoluminescence spectra can be simulated by the same procedure
as in Section 111. The rate equation for the ground-level population should
be changed to include the nonradiative process as
(5.28)
where znlis the nonradiative lifetime at the ground level. Figure 5.16 shows
the photoluminescence spectra as calculated using the nonradiative lifetime
as a parameter. We used a carrier injection rate of G = s - l per indi-
vidual dot, a broadening factor with a Gaussian distribution of 50 meV,
z, = 1 ns, Di = 2 . i (i = 1,2,. . . , 5 ) , T~~ = 200, zo3 = 100, T~~ = 50, and zo5 =
BOTTLENECK
DOTS 233
j
G-
v
10-8 . . , . .. . , . . . . , . , . . , . I I . , .., . i
't
0
300
Temperature (K)
o so 100
Temperature (K)
is0
FIG. 5.15. Lifetimes up to the third level as a function of temperature: (a) recombination
200
A
-
250
:t300
lifetimes and (b) relaxation lifetimes. The solid and open symbols indicate lifetimes of as-grown
and annealed samples, respectively. The recombination lifetimes in the ground level became
temperature-dependent after annealing and decreased to a few hundred picoseconds near room
temperature. The relaxation lifetimes decreased slightly after annealing (from Mukai et al.,
1998).
20 ps. Unknown relaxation lifetimes of T~~ were assumed to be half of that

used in Section 111, since the present experiments show that the relaxation
lifetimes of T~~ decrease by about half after annealing. Unknown relaxation
lifetimes of T~~ and zO5 do not affect the results of Fig. 5.16 as long as the
emission from the fourth and fifth levels does not appear.
The injection rate was chosen to give a low injection condition, where the
fourth and fifth excited levels do not apper in the calculated spectrum. The
234 KOHKIMUKAIAND MITSURUSUCAWARA
1 .o 1.1 1.2 1.3 1.4 1.5

Wavelength (pm)
FIG. 5.16. Photoluminescence spectra calculated by assuming values for the nonradiative
lifetime (tnl). As the nonradiative lifetime decreased from infinity to a few hundred picoseconds,
the emission intensity of the ground level decreased and became comparable to that of the
second level (from Mukai et al.. 1998).
occupation at the ground level, fi, is about 8%. As the nonradiative lifetime
decreases from infinity to a few hundred picoseconds, the emission intensity
of the ground level decreases and becomes comparable to that of the second
level. The value of the nonradiative lifetime in the simulation matches the
value estimated in Fig. 5.15, using the recombination lifetime of 200 ps and
the radiative lifetime of 1 ns.
The carrier dynamics in the ALS dots are summarized in Fig. 5.17. The
relative emission intensity of the ground and second levels is determined by
the equilibrium between the radiative recombination process (rl, r2), the
relaxation process (rel), and the nonradiative recombination process
(nl, n2). In an as-grown sample, the nonradiative lifetimes are negligible,
which can be expected from the temperature-independent radiative lifetimes
(see Section 111). After annealing, the recombination lifetime at the ground
level decreases due to the nonradiative channel, while the nonradiative
process of the second level is still negligible. The drastic change in relative
emission intensity demonstrates that the balance of radiative and non-
radiative recombination and relaxation rates determines the shape of the
emission spectrum. This contradicts our experience with quantum wells,
whose emission spectra generally become narrow toward a lower energy
5 PHOTON BOTTLENECK
EFFECT DOTS
IN QUANTUM 235
Energy
t
I
I
I
I
r2
channel
channel
>Radiative
channel
FIG. 5.17. Schematic of carrier dynamics in a quantum dot before and after annealing. The
relative emission intensity of the ground and second levels was determined by the equilibrium
between the radiative recombination process ( r l . r2), the relaxation process ( r e / ) , and the
nonrddiative recombination process (rrl, n2). The nonradiative lifetimes, T,, and rn2, were
negligible for the as-grown sample. After annealing, the nonradiative lifetime T , , ~ .became
effective. When carriers were not sufficiently supplied to the ground level, the emission intensity
of the ground level decreased compared to that of the second level.
regime as the emission intensity decreases. In quantum wells, an intersub-

band carrier relaxation is much faster than the recombination process.
V. Simulation of Laser Performance, Including the Auger

Relaxation Process
From what we have seen in the experiments and their respective analyses,
we can construct a framework of the carrier relaxation problem as follows.
First, when there are a large number of carriers of 10'2-10'3 cm-2 around
quantum dots, as in the initial stage of time-resolved photoluminescence, the
carrier capture into dots occurs with a considerably short lifetime. Also,
carriers are relaxed into deep low-order levels through the Auger processes.
When there are only a few carriers of less than 10" cmP2, they are first
trapped at the high energy levels and relax toward the lower energy levels
later. During quantum-dot laser operation, the actual relaxation process
varies between the above mentioned two cases as the current injection
increases.
Let us here introduce the Auger effect into the simulation presented in
Section V.l of Chapter 1. The same rate equations are used, and we take
into account the dependence of the carrier injection time into dots, zo, on
the carrier density in the quantum well Nq as
where C is the Auger coefficient. Figure 5.18(a) shows the output power as
a function of the injected current for C = 1.0 x cm4/s and C = 0
with zC0= 100 ps. The inhomogeneous broadening is To = 20 meV, and the
coverage is = 10%. The recombination lifetimes we used are zr = zSpon=
2.7 ns and z~~= 3 ns (solid lines), and 7 r = 0.5 ns and z~~ = 0.5 ns (dashed
lines). It is clear that the Auger process increases efficiency, lowers threshold
current, and enables higher output power (again, calculation stops when the
quantum-well carrier density reaches 1.0 x 10l2cm2). Figure 5.18(b) shows
the quantum-well carrier density as a function of injected current. For C =
cm4/s, zo is superimposed. We see that z0 decreases close to 10ps as
the injected current increases and that the increase in carrier density is
-- l r =0.5 ns, T ~ ~ ns
0.5 =
(a) Injected current (mA) (b) Injected current (mA)
... . .me
witnout . rr .
Auger enecr.-. . . . -- .-
FIG.5.18. (a) Output power as a function of the injected current calculated with and
. * ..
I ne spectrum oroaaening was ro= LU mev, ana me coverage was
I .
5 = lo%, which gave P = 0.84 at the lasing threshold (from Sugawara, 1998). (b) Carrier
density in a quantum well located around the quantum dots as a function of injected current.
The carrier relaxation lifetime into dots decreased close to l o p s as the injected current
increased when C = 1 x lo-'' cm"/s (from Sugawara, 1998).
5 PHOTON BOTTLENECK
EFFECT DOTS
IN QUANTUM 231
, , 1 1 1 1 1
10-
Auger coefficient (cm4/s)
FIG. 5.19. Maximum output power as a function of the Auger coefficient for various
quantum-dot coverages at Po = 20 meV (from Sugawara, 1998).
considerably restricted. Figure 5.19 shows the maximum output power as a

function of the Auger coefficient. For a power of 30 mW, we need C = 2.1 x
lo-'" cm"/s at 5 = 40% and C = 6.0 x lo-'" cm"/s for 5 = 10%. Uskov et
al. (1998) calculated the Auger coefficients as a function of the quantum-dot
diameter and showed that the C value reached 10-12cm4/s for both the
electron and the hole capture processes via electron scattering. Whle this
result looks quite promising with respect to static performance values such
as external quantum efficiency and maximum power, further acceleration is
required for high-speed modulation (see Chapter 1). To check the Auger
coefficient values suggested by these theoretical calculations, and to design
quantum-dot active layers to enable fast relaxations, we need experimental
studies that concentrate on carrier relaxation dynamics under various
carrier densities around quantum dots. The relaxation lifetime that we
achieve determines the performance of quantum-dot lasers and the applica-
tion fields in which they might be used.
VI. Summary
We presented our experimental studies on carrier dynamics in self-

assembled quantum dots and discussed their influence on lasing perform-
ance. Electroluminescence and time-resolved photoluminescence data were
analyzed to provide recombination and relaxation lifetimes as a function of

temperature. The RIO model in the analysis explained the double-exponen-
tial decay of excited levels, and, using the obtained lifetimes, electrolumines-
cence and photoluminescence spectra were successfully simulated. One thing
we should stress is that the phonon bottleneck problem does exist, as we
experimentally found that carrier relaxation between discrete energy levels
was very slow with lifetime above 100 ps. Also, carrier-carrier scattering can
send carriers deep inside the dots, as seen in the initial stage of time-resolved
photoluminescence decay. These two findings, we believe, help to clear up
the confusion in this research field. As shown in the laser simulation, high
static performance of quantum-dot lasers can be expected if the Auger
process accelerates the relaxation up to about 10 ps.
REFERENCES
Adler, F., Geiger, M., Bauknecht, A., Scholz, F., Schweizer, H., Pilkuhn, H. M., Ohnesorge, B.,
and Forchel, A. (1996). J. A p p l . Phys. 80, 4019.
Anopelto, J., Yamaguchi, A. A,, Nishi, K., Usui, A., and Sakaki, H. (1993). Jpn. J . Appl. Phys.
32, L32.
Apetz, R.,Vescan, L., Hartmann, A,, Dieker, C., and Luth, H. (1995).A p p l . Phys. Lett. 66,445.
Arakawa, Y., and Sakaki, H. (1982). Appl. Phys. Lett. 40, 939.
Arakawa, Y., Vahala, K., and Yariv, A. (1994). Appl. Phys. Lett. 45, 950.
Asada, M., Miyamoto, Y.,and Suematsu, ?. (1986). IEEE J . Quantum Electron. QE-22, 1915.
Asada, M. (1989). IEEE J . Quantum Electron. 25, 2019.
Benisty, H., Sotomayor-Torres, M. C., and Weisbuch, C. (1991). Phys. Rev. B. 44, 10945.
Benisty, H. (1995). Phys. Rev. B. 51, 13281.
Bockelmann, U., and Bastard, G. (1990). Phys. Rev. B. 42, 8947.
Bockelmann, U., and Egeler, T. (1992). Phys. Rev. B. 46, 15574.
Bockelmann, U. (1993). Phys. Rev. B. 48, 17637.
Bockelmann, U., Heller, W., Filoramo, A,, and Roussignol, P. (1997). Phys. Rev. B. 55, 4456.
Brus, L. (1986). IEEE J. Quantum Electron. 22, 1909.
Chemla, D. S., and Miller, B. A. D. (1986). O p t . Lett. 11, 522.
Eaglemann, D. J., and Cerullo, M. (1990). Phys. Reti. Lett. 64, 1943.
Efros, A. L., Khachenko, A. V., and Rosen, M. (1995). Solid Stute Cornmun. 93, 281.
Gerard, J.-M. (1 995). Con$ned Electrons and Photons: New Physics und Applicutions. Plenum,
New York, p. 357.
Grundmann, M., and Bimberg, D. (1997). Phys. Rer. B. 55,9740.
Grosse, S., Sandmann, H. H. J., Plessen, V. G., Feldmann, J., Lipsanen, H., Sopanen, M.,
Tulkki, J., and Ahopelto, J. (1997). Phys. Rev. B. 55, 4473.
Inoshita, T., and Sakaki, H. (1992). Phys. Rev. B. 46, 7260.
Kamath, K., Klotzkin, D., and Bhattacharya. P. (1997a). Proc. Confi IEEE Lasers and
Electro-Optics Society, p. 498.
Kamath, K., Chervela, N., Linder, K. K., Sosnowski, T., Jiang, H.-T., Norris, T., Singh, J., and
Battacharya, P. (1997b). Appl. Phys. Len. 71, 927.
Kurtenbach, A., Ruhle, W. W., and Eberl. K. (1995). Solid State Corn. 96, 265.
5 PHOTON BOTTLENECK DOTS
EFFECTIN QUANTUM 239
Leonard, D., Krishnamurthy, M., Reaves, M. C., DenBaars, P. S., and Petroff. M. P. (1993).
Lipsanen, H., and Sopanen, M. (1995). Phys. Rev. B. 51, 13868.
Moison, J. M., Houzay, F., Barthe, F., Leprince, L., Andre, E., and Vatel, 0. (1994). Appl. Phys.
Lett. 64, 196.
Mukai, K., Ohtsuka, N., Sugawara, M., and Yamazaki, S. (1994). Jpn. J . Appl. Phys. 33, L1710.
Mukai, K., Shoji, H., Ohtsuka, N., and Sugawara, M. (1996a). Appl. Phys. Lett. 68, 3013.
Mukai, K., Ohtsuka, N., Hajirne, S., and Sugawara, M. (1996b). Phys. Rev. B. 54, R5243.
Mukai, K., Sugawara, M. (1998). Jpn. 1. Appl. Phys. 37, 5451.
Nakayama, H., and Arakawa, Y. (1994). 7th Int. Con$ Superlattices, Microstructures, and
Microdevices, Banff, Canada.
Ohnesorge, B., Albrecht, M., Oshinowo, J., Forchel, A., and Arakawa, Y . (1996). Phys. Rev. B.
54, 11532.
Oshinowo, J., Nishioka, M., Ishida, S., and Arakawa, Y. (1994). Appl. Phys. Lett. 65, 1421.
Raymond, S., Fafard, S., Charbonneau, S., Leon, R., Leonard, D., Petroff, M. P., and Merz, L. J.
(1995). Phys. Rev. B. 52, 17238.
Raymond, S., Fafard, S., Poole, J. P., Wojs, A., Hawrylak, P., Charbonneau, S., Leonard, D.,
Leon, R., Petroff, M. P., and Merz, L. J. (1996). Phys. Rev. B. 54, 12548.
Sallese, J. M., Carlin, F. J., Gaihanou, M., and Grunberg, P. (1995). Appl. Phys. Lett. 67, 2633.
Sercel, P. C. (1995). Phys. Rev. B. 51, 14532.
Shoji, H., Mukai, K., Ohtsuka, N., Sugawara, M., Uchida, T., and Ishikawa, H. (1995). I E E E
Photo. Tech. Lett. 7, 1385.
Sugawara, M. (1998). Prnc. of Optoelectronics '98, Photonic West. San Jose.
Tabuchi, M., Noda, S., and Sasaki, A. (1992). Science and Technology of Mesoscopic Siructures
(Springer-Verlag. Tokyo, p. 379.
Tsang, W. T. (1978). Appl. Phys. Lett. 33, 245.
Uskov, A. V., Mcinerney, J., Adler, F., Schweizer, H., and Pilkuhn, M. H. (1998). Appl. Phys.
Lett. 72, 58.
Utzmeier, T., Postigo, A. P., Tamayo, J., Garcia, R., and Briones, F. (1996). Appl. Phys. Lett.
69, 2674.
Vollrner, M., Mayer, J. E., Riinle, W. W., Kurtenbach, A,, and Eberl, K. (1996). Phys. Reu. B.
54, R 17292.
Wang, G., Fafard, S., Leonard, D., Bowers, E. J., Merz, L. J., and Petroff, M. P. (1994). Appl.
Phys. Lett. 64. 2815.
Yu, H., Lycett, S., Roberts, C., and Murray, R . (1996). Appl. Phys. Lett. 69, 4087.
CHAPTER6
Self-Assembled Quantum Dot Lasers

Hajime Shoji
OPTICAL
SEMICONDUCTOR
DEV~CES
LABORATORY
FUJ~TSU
LABOKATQRIES
LTD.
ATSUGI.
JAPAN
I. INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . 24 1
11. FUNDAMENTAL PROPERTIES OF Q U A N T U M - D O T LASERS . . . . . . . . . . . 243
1. Gain Characteristics . . . . . . . . . . . . . . . . . . . . . . . . 244
2. Threshold Current . . . . . . . . . . . . . . . . . . . . . . . . . 246
3. Dynamic Characteristics . . . . . . . . . . . . . . . . . . . . . . 248
111. FABRICATION OF SELF-ASSEMBLED QUANTUM-DOT LASERS. . . . . . . . . . 249
1. Fabrication . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
2. Device Characteristics . . . . . . . . . . . . . . . . . . . . . . . 255
3. Limiting Factors of’Laser Performance . . . . . . . . . . . . . . . . 261
IV. KEYTECHNOLOGIES FOR THE NEXT ERA . . . . . . . . . . . . . . . . . 269
I . Close1.v Slacked Quantum-Dot Lusers . . . . . . . . . . . . . . . . . 270
2. Columnur Quantum-Dot La~ers . . . . . . . . . . . . . . . . . . . 213
3. Long- Wavelength Quantum-Doi Lasers . . . . . . . . . . . . . . . . 216
4. Quantum-Dot Vertical-Cavity Surface-Emitting Lasers . . . . , . . , , . 219
V. CONCLUSION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
Acknowledgments . . . . . . . . . . . . . . . . . . , , . . , . . , . 283
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
I. Introduction
The quantum-dot laser is one of the most representative candidates for

the practical application of quantum-dot structures. Originally called the
quantum-box laser, it was first proposed by Arakawa and Sakaki in 1982.
Three-dimensional quantum confinement of electrons and holes in a quan-
tum dot with a size equal to or below the exciton Bohr radius causes the
atom-like density of states, or delta function, which is believed to be essential
to improved semiconductor laser performance. The effect of three-dimen-
sional quantum confinement was shown in an experiment using a strong
magnetic field normal in quantum wells (Arakawa et al., 1983). Semiconduc-
24 1
Copyright I I Y Y 9 by AcJdemlc Press
All ~ i p h l rof reprodurtmn In m y lorm reserved
ISBN 0-12-752169-0
ISSN M I R O - X ~ X ~ ,~~3~ 0 0 0
242 HAJIME
SHOJI
tor lasers equipped with quantum dots as the active region are expected to
exhibit ultralow threshold current, temperature-insensitive operation, nar-
row spectral linewidth, and large modulation bandwidth (Arakawa and
Sakaki, 1982; Asada et al., 1986). All of these features are attributed to the
quantum dots’ enhanced gain properties, especially the extremely high
differential gain, induced by their three-dimensionally discrete density of
states.
Despite its theoretically predicted superiority, no quantum dot laser has
been actually realized. This is because quantum structures fabricated so far
have not produced a strong carrier confinement, especially in a lateral
direction. In the last decade, two approaches have been tried for realizing
quantum dot structures. The first one is a precise patterning technique
(Hirayama et al., 1994), originally developed for quantum-wire structures. It
involves patterning of quantum wells by wet or dry etching and overgrowth
on the patterned structure, or selective growth on a patterned substrate
(Kash et al., 1986; Fukui et al., 1991). Because of recent progress in electron
beam lithography and etching techniques, InGaAs-InGaAsP quantum-dot
lasers became a reality in 1994 (Hirayama et al., 1994). However, the
threshold current density was as high as 7.6 kA/cmZ even at 77 K. In this
approach, it is generally very difficult to obain a high-quality structure
without damage, since nonradiative recombination, especially at the surface
of etched sidewalls where the active region is exposed to air, severely limits
laser performance.
The second approach is self-organization of quantum dots during highly
strained heteroepitaxy, and it has been very successful. Uniform dots with
small size, high density, and high emission efficiency, unachievable with
conventional techniques, have been easily realized in conventional molecu-
lar beam epitaxy (MBE) and metalorganic vapor phase epitaxy (MOVPE).
The self-organization technique has lead to three methods for fabricating
quantum-dot structures: (1) spontanous formation of In(Ga)As islands on
GaAs in the Stranski-Krastanow (SK) mode (Leonard et al., 1993; Kirstaed-
ter et al., 1994; Oshinowo et al.. 1994; Tackeuchi et al., 1995; Mirin et al.,
1994); (2) self-organization of dot-like disk structures on (311) B GaAs
substrates (Notzel et al., 1994); and (3) self-formation of InGaAs dots on
GaAs by alternative supply of InAs and GaAs in atomic layer epitaxy (ALE)
(Mukai et al., 1994 and 1996a). Furthermore, the self-organization of
quantum dots has been successful in many material systems, such as InAs
on InP (Fafard et al., 1996; Nishi et al., 1998); InP on InGaP (Moritz et al.,
1996); and GaSb (Hatami et al., 1995) or InSb, GaSb, and AlSb on GaAs
(Glaser et al., 1996). Attempts to create a new quantum-dot structure,
including development of new material systems and new growth mechan-
isms, are ongoing, bringing with them a deeper understanding of the growth
6 SELF-ASSEMBLED
QUANTUM
DOTLASERS 243
mechanism of self-assembled quantum-dot structures.

Recent attempts to fabricate quantum-dot injection lasers by the methods
mentioned above have been reported by several groups. Kirstaedter et al.
(1994) reported the first demonstration of InAs-GaAs quantum-dot laser
grown in the SK mode in 1994. Also in 1994, Temmyo et al. reported a
continuous-wave (CW) lasing in an InGaAs-AlGaAs quantum-disk struc-
ture on (31 l)B GaAs although the lateral size of the disk was 60 nm, which
was rather large for sufficient quantum confinement. In 1995, Shoji et al.
reported a laser oscillation at 80 K in InGaAs-GaAs quantum-dot lasers
grown by ALE. Besides these researchers, several other groups have re-
ported successful laser oscillation (Moritz et al., 1995b; Mirin et al., 1996;
Kamath et al., 1996; Shoji et al., 1996a). Clearly, research on the quantum
dot structures is creating a new field covering growth techniques and device
applications.
In this chapter we describe semiconductor lasers with self-assembled
quantum dots in the active layer. Section I1 briefly summarizes the funda-
mental properties of quantum dot lasers from theoretical points of view as
well as what can be expected from them. Section I11 introduces actually
fabricated quantum-dot lasers, mainly focusing on the results of our work
and taking InAs-GaAs quantum-dot lasers -which are most studied as ~
examples. The fabrication process, lasing characteristics and problems to be

solved are discussed in detail. Section IV treats several key technologies for
improving quantum-dot lasers, including closely stacked quantum-dot
structures, columnar quantum-dot structures, and quantum dots emitting in
long wavelength region. The chapter ends with a discussion of how these
technologies can be applied in quantum dot vertical cavity surface emitting
laser (VCSEL).
11. Fundamental Properties of Quantum Dot Lasers
Most of the performance expected from quantum-dot lasers originates in

their atom-like density of states. Figure 6.1 shows how active regions in
semiconductor lasers and the density of states of electrons vary with an
increase in the dimensions of quantum confinement. By the use of multi-
dimensional quantum-confined structures, freedom of carrier motion is
reduced to a limited direction. Ideally, the carriers are completely confined
in the quantum dot. For that reason, the dot’s density of states is expressed
by the delta function. The width-of-gain spectrum is determined only by
homogeneous broadening due to the intraband relaxation in the quantum
dot. These gain properties are the basis of those features that give the
244 HAJIMESHOJI
Bulk Quantum Quantum Quantum

well wire dot
‘li D(E) D(E) W) WE)

FIG. 6.1. Variations in active regions in semiconductor lasers and the density of states for
bulk, quantum well, quantum wire, and quantum dot.
quantum-dot laser its advantages over conventional lasers, and we discuss

them here. The theoretical treatment of quantum dots is provided in
Chapter 1. Basic theories of quantum-dot lasers and conventional lasers are
summarized in several papers (for example, Asada et al., 1986; Arakawa and
Yariv, 1986).
1. GAINCHARACTERISTICS
The basic theory of the gain characteristics of quantum dots was derived
by Asada et al. (1986). This density-matrix theory gives the linear gain of
quantum dots without inhomogeneous broadening as follows:
where w is the angular frequency of light; e0 and p0 are, respectively, the

dielectric constant and permeability of the vacuum; n, is the refractive index
of the quantum dot; f, and J , are the Fermi functions; and M is the dipole
moment of the quantum-dot structure. L(E, Q) corresponds to the relaxation
6 SELF-ASSEMBLED
QUANTUM
DOTLASERS 245
1o4
ox looxlooxloo A3
-
6
v
.-
C
103
2x102
1.2 1.3 1.4 1.5 1.6 1.7
Wavelength, h (pm)
FIG. 6.2. Calculated gain spectra at 300K for a quantum dot (box), quantum wire,
quantum well (film), and bulk. (Reprinted with permission from Asada et al., 1986. $'I 1986
IEEE.)
broadening given by the Lorentzian function, where T~,,is the intraband

relaxation time. D(E) is the density of states of electron-hole pairs with the
same energy levels, where d,, d,, and d, are the dot sizes assuming a box
shape, respectively, and 6(E) is the delta function. The density of states in
quantum dots is illustrated in Fig. 6.1. Densities of electrons and holes in
quantum dots are determined by the neutral condition.
Figure 6.2 shows calculated gain spectra at 300K for quantum dot
(box), quantum wire, quantum well, and bulk structures at a constant
carrier density, N , of 3 x 10l8~ m - Except
~ . for the bulk structure, the
thickness in the quantum-confined direction was set at 100A.We clearly see
that enhancement of peak material gain and narrower gain bandwidth in
the quantum-dot structure are the results of an increase in the dimension of
quantum confinement from bulk to quantum dot. The peak gain of the
quantum dot increases by one order of magnitude compared with that of
the quantum well. This result also indicates that the differential gain, dg/dN,
is drastically improved in the quantum dot due to the large peak gain and
the narrow gain bandwidth.
246 HAJIMESHOJI
2. THRESHOLD
CURRENT
The large optical gain and differential gain result in a low threshold
current operation. Assuming that the current is injected only through the
top surfaces of cubic quantum dots, as described by a schematic image
shown in Fig. 6.3, the relation between the injection current and the carrier
density is expressed by a simple formula as follows:
J = td,eN/z, (6.4)
where J is the averaged current density defined by dividing the total current
by the whole area of the active region ( J = I/WL in Fig. 6.3); e is the
electron charge; N is the carrier density; z, is the carrier lifetime; and 5 is
the filling factor of quantum dots in the active region. Here, the active region
is assumed to consist of a single dot layer. Referring to Fig. 6.3, 5 is defined
as 5 = d,d,/WL, where W is width of active region and L is the cavity
length. The effect of interaction among dots and finite potential depth is
ignored. Generally, the carrier lifetime z, consists of radiative and non-
radiative components of z, and znr, respectively, as
(6.5)
Current, I
FIG. 6.3. Schematic of a quantum-dot laser.

6 SELF-ASSEMBLED DOTLASERS
QUANTUM 241
The radiative lifetime is given approximately by the bimolecular lifetime as
zr =N /[
c [: I M I ~ E ~ Lmw)
3/2Eolj2n , ~ n h - 4
imn
-U 4 (6.6)
Using Eqs. (6.1) through (6.6), the optical gain can be related to the
current density. Figure 6.4 shows calculated peak gains as a function of
current density for four types of lasers with active regions of quantum dots,
quantum wires, quantum wells and bulk (Asada et al., 1986). The structures
assumed here are the same as in Fig. 6.2, which shows the increase of peak
gain with the increased dimension of quantum confinement. The threshold
gain, g t h , is denoted as follows by considering the optical confinement
factors in the given structures:
Current density, J (A/cm2)
FIG. 6.4. Calculated peak gain at 300 K for a quantum dot (box), quantum wire, quantum
well (film), and bulk. The dashed line on each curve indicates the threshold gain. The
corresponding threshold current density is also shown. (Reprinted with permission from Asada
et al., 1986. 1Ci 1986 IEEE.)
248 HAJIME
SHOJI
where u,, and a,, are the absorption losses in the active region and other
regions, respectively; tx,F,, and 5, are the optical confinement factors in the
x,y , and z directions, respectively; and R is the facet reflectivity. 5, and 5,
can be approximately defined as the filling factors in the x and z directions,
respectively, because the optical field in the cavity is much larger than that
in the quantum dot. t, is determined by the waveguide structure. For
example, assuming that tx= 5, = 0.5 for a quantum-dot structure, 4, = 1
and t, = 0.5 for a quantum-wire structure, 5, = tz= 1 for quantum-well
and bulk structures, and assuming reasonable values for other parameters
(Asada et al., 1986), the threshold gains can be calculated by Eq. (6.7). Then,
if the results are put into Fig. 6.4, the threshold current densities for the
given structures are derived. As noted in Fig. 6.4, lower threshold current
density should be achieved in quantum-dot lasers-about one tenth that of
quantum well lasers. Of course. this value strongly depends on the par-
ameters assumed, and they should be much improved by optimizing the
structure. The low threshold current in the quantum-dot laser also brings
about a lesser temperature dependence of the threshold current due to
strong confinement of the electrons and holes.
Regarding threshold gain, there is another attractive feature of the
quantum-dot laser. As described above, the internal loss expressed by the
first and second terms of Eq. (6.7) becomes smaller in the quantum-dot
laser, therefore, larger differential quantum efficiency, given by
r
In-
1
Rf
1
is expected, where qd is the differential quantum efficiency, qi is the internal
quantum efficiency, and R, and R , are the reflectivities of the front and rear
facets, respectively.
3. DYNAMIC
CHARACTERISTICS
Large differential gain results in improvement of dynamic characteristics.

The relaxation oscillation frequency, L, which determines the intrinsic
modulation bandwidth, is given by
f ' =L/ii$g
2a (6.9)
6 SELF-ASSEMBLED
QUANTUM
DOTLASERS 249
where neff is the effective refractive index of the waveguide, c is the speed of
light in the vacuum, Po is the photon density in the cavity, and zp is the
photon lifetime. It is quite clear that greater differential gain results in a
larger relaxation oscillation frequency, which potentially enhances the
modulation bandwidth of the quantum-dot laser.
Narrower spectral linewidth can be also expected in the quantum-dot
laser. The spectral linewidth, Av, is expressed by
(6.10)
(6.11)
where M , R,, vg, hw, r, nsp, n, and Po are, respectively, the linewidth
enhancement factor, the mirror loss, the group velocity of light, the photon
energy, the optical confinement factor, the spontaneous emission factor, the
effective refractive index, and the output power. A large differential gain
results in a linewidth enhancement factor of nearly zero, leading to smaller
spectral linewidth. These parameters are strongly related to the direct
modulation characteristics, and chirping under high-speed modulation
should be suppressed because of the small a, leading to long transmission
distance in fiber communication applications.
111. Fabrication of Self-Assembled Quantum-Dot Lasers
In the realization of self-assembled quantum-dot lasers, quantum-dot

structures grown in the SK mode are the most successful to date for device
performance. Recent intensive work has clarified the properties of uctuul
quantum-dot lasers grown in the SK mode. Most of this work focuses on
InAs or InGaAs quantum dots on GaAs because they are suitable for laser
- -
application, given their small diameter of 20 nm, high density of > 1 x
10'' cm ',typical size distribution of lo%, high emission efficiency, and
~
so on. Room-temperature CW lasing has been demonstrated in In(Ga)As-

GaAs quantum dots.
This section describes the fabrication and lasing characteristics of self-
assembled In(Ga)As-GaAs quantum-dot lasers, focusing mainly on the
results of our work. Also discussed are several factors that limit the
performance of quantum-dot lasers and what can be done about them.
250 HAJIMESHOJI
1. FABRICATION
Formation of Quantum Dots in Stranski-Krastanow Mode
The SK growth mode is used mostly in the formation of quantum-dot

structures. Typically, formation of In(Ga)As-GaAs quantum dots in this
mode occurs at temperatures between 460°C and 550°C. The formation is
reported both in molecular beam epitaxy (MBE) (Leonard et al., 1993;
Oshinowo et al., 1994; Tackeuchi et al., 1995; Mirin et al., 1995; Sugiyama
et al., 1996) and in metalorganic vapor phase epitaxy (MOVPE) (Ledentsov
et al., 1996; Heinrichsdorff et al., 1996 and 1997). A few-monolayer supply
of In(Ga)As on GaAs result in the formation of three-dimensonal islands on
an extremely thin film called a wetting layer. The typical lateral diameter
obtained by MBE and MOVE ranges between 12 nm and 20 nm, and the
dot density is typically between 2 x 10'' cm-2 and 1 x 10" cm-2. De-
tails on the SK growth mode are given in Chapter 3.
Optimization of dot structure is important in semiconductor lasers. Too
small a size results in the absence of a quantum-confined level or in a small
energy difference between the barrier and the quantized energy level of dots.
Too large a size leads to a weaker quantum size effect, which means that the
atom-like nature of the dots will not be extracted. Dot density is another
important factor. For example, although it depends on the waveguide
structure, if there is no fluctuation in dot size, the optimum areal coverage
in a single active plane will be about a few percent to minimize the threshold
current (Asada et al., 1986). However, in actual quantum dots, there is a
large size fluctuation. Higher-density dots are required to obtain large
modal gain; low-density dots lead to gain saturation due to small optical
confinement to the active layer, which degrades performance (Schmidt et al.,
1996). Therefore, in addition to increasing in-plane density, vertically
stacked dot structures are often required.
Figure 6.5(a) and (b) shows, respectively, a plan view and a cross-sectional
view, taken with a transmission electron microscope (TEM), of a triple-
stacked quantum-dot structure (Shoji et al., 1997a). The grown quantum
dots were 20 nm in diameter and 5 nm in height. The areal coverage was
about 10%. Each dot layer was separated by GaAs barriers that were 20 nm
thick. In this case, the neighboring dot layers in the vertical direction could
be optically coupled through the waveguide structure, but they were not
electronically coupled because the barrier layer between the dot layers was
thick enough. Stacked dot structures with 10 or more layers have also been
reported (Saito et al., 1996; Pan et al., 1997). Electronically coupled
quantum-dot structures with thinner barriers are discussed in Sections IV. 1
and IV.2.
6 SELF-ASSEMBLED
QUANTUM
DOTLASERS 251
FIG.6.5. (a) Plan-view and (b) cross-sectional TEM images of a triple-stacked quantum-
dot structure. The grown quantum dots were 20 nm in diameter and 5 nm in height. The areal
coverage was about 10%. Each dot layer was separated by 20-nm-thick GaAs barriers. (From
c,
Shoji et al., 1997a. 1997 IEEE.)
Optical Proper ties
An example of a photoluminescence (PL) spectrum at 300K for a

triple-stacked quantum dot structure is shown in Fig. 6.6 (Shoji et al.,
1997a). An emission peak at -
1.13 pm was clearly observed at room
temperature, and the full width at half maximum (FWHM) of the peak was
about 80 meV. In most cases, the emission peak of In(Ga)As-GaAs quantum
dots is between 0.95pm and 1.2pm. The FWHM of the emission peak
represents the uniformity of dot size, and is typically 40-100meV. The
252 HAJIMESHOJI
Energy (eV)
lo-'
ground
Wetting 2nd
1 o'6
800 900 1000 1100 1200 1300
Wavelength (nm)
FIG. 6.6. Photoluminescence spectrum at 300 K for a triple-stacked quantum-dot structure.
The emission peak at around 1.13pm is from the ground state, and the full width at half
maximum of the peak is about 80meV. (From Shoji et al., 1997a. 0 1997 IEEE.)
narrowest linewidth is 15meV (Oshinowo et al., 1994). This value is

determined by inhomogeneous broadening, not by homogeneous broaden-
ing. Emission linewidth of a single quantum dot is much less than 1 meV at
low temperature (Grundmann et al., 1995), which corresponds to the phase
coherence time in a quantum dot. Consequently, if the inhomogeneous
broadening dominates the spectral linewidth, almost constant FWHM is
obtained over a wide temperature range. This fact is in great contrast to
cases of bulk material and quantum wells.
In Fig. 6.6, the asymmetric PL spectrum suggests the existence of several
excited states. By the Gaussian fitting to the PL spectrum, the second state
of electrons is found at around 1.04- 1.05pm. The energy difference between
the ground state and the second state is about 90meV, which agrees well
with a theory that considers the grown dot size. Although the third state is
not identified clearly in the PL spectrum, assuming that the energy level
between the neighboring states is equally spaced in the parabolic potential
(Mukai et al., 1994 and 1996b; Ahopelto et al., 1995), the third level should
be at around 0.97 pm. Note that a peak at 0.92 pm is from the underlying
wetting layer, and a peak at 0.88 pm is from the GaAs barrier layer.
The number of excited states is strongly affected by dot size. Figure 6.7
-
illustrates this relationship. If the dot is small enough ( < 15 nm), only the
transition between the ground states of electrons and holes become a
dominant recombination process. Although emission peaks from the excited
QUANTUM DOTLASERS
6 SELF-ASSEMBLED 253
Dot size Small dot Large dot
Energy El
El E2
E3
Allowed Ground state Several

transition only excited states
Electro-
luminescence
kE Ei
Ei Ez E3
E
FIG. 6.7. Effect of dot size on the excited states of electrons and holes.
states are also generally observed together with those from the ground state
because of nonorthogonality due to anisotropy of dot shape, the emission
from the ground state always becomes dominant even with increasing
excitation. On the other hand, in larger dots, there are several excited states
of electrons in the conduction band, which leads to the transitions from the
excited states of electrons and holes. The result in Fig. 6.6 corresponds to
the latter case.
Device Fabrication
A schematic view of a typical In(Ga)As-GaAs quantum dot laser is given

in Fig. 6.8. Profiles of the conduction-band energy and the corresponding
refractive index are depicted. The active layer of a quantum-dot laser
consists of a two-dimensional or three-dimensional array of quantum dots
surrounded by a pair of higher bandgap barriers to effectively confine
injected carriers. The active layer is sandwiched by lower and upper
cladding layers with a higher band gap energy (smaller refractive index),
resulting in the formation of an optical waveguide. Thus, the structure is
basically the same as that of conventional double heterostructure lasers.
Ridge-waveguide structure is often employed to evaluate the performance
of quantum-dot lasers in addition to broad-area lasers. A schematic struc-
ture of a fabricated ridge-waveguide quantum-dot laser is shown in Fig. 6.9
254 HAJIMESHOJI
c-GaAs mAlGaAs GaAs pAlGaAs pGaAs

substrate cladding layer barrier cladding layer contact
layer
Refractive
t index
FIG. 6.8. Schematic of a typical In(Ga)As-GaAs quantum-dot laser, showing profiles of the
conduction-band energy and the corresponding refractive index. The dashed curve shows the
optical field in the lateral direction.
FIG. 6.9. Schematic of a fabricated ridge-wave guide quantum-dot laser. The laser contains
a triple-stacked quantum-dot structure. The width of the ridge-waveguide is about 2.5 pm.
6 SELF-ASSEMBLED
QUANTUMDOTLASERS 255
(Shoji et al., 1997a). The laser structure was grown on a (001) n-GaAs
substrate followed by a 0.9-pm-thick n-Al,,,Ga,.,As cladding layer, a
quantum-dot active layer, a 1.2-pm-thick p-Al,,,Ga,,,As classing layer, and
a 0.3-pm-thick p-GaAs contact layer. As the active layer, a triple-stacked
quantum-dot structure was employed to obtain larger optical confinement
to the active layer. In this case, the optical confinement factor to the
active-layer plane was about 6% (actual optical confinement to quantum
dots is much less than this value because of low areal coverage of quantum
dots in the active layer). A narrow ridge-waveguide structure is usually
formed by chemical or dry etching of the p-GaAs contact layer and a part
of the p-AlGaAs cladding layer. In Fig. 6.9, the width of the ridge-
waveguide was about 2.5 pm, which enabled stable control of the lateral
mode. After growing the entire structure, several laser chips with different
device parameters were prepared for measurement. Typically, cavity length
and facet reflectivity varied case by case. In addition, the number of dot
layers often varied wafer to wafer.
2. DEVICE
CHARACTERISTICS
Electroluminescence
Electroluminescence (EL) from quantum dots under current injection

below the threshold gives us much information on the subbands of quantum
dots as well as the PL spectrum. Figure 6.10 shows measured EL spectra for
several injection currents (Shoji et al., 1997a). Clear discrete peaks were
found in the EL spectra which were attributed to subbands of quantum dots
(Ahopelto et al., 1995; Mukai et al., 1996a; Grundmann et al., 1996; Xie et
al., 1996). The three peaks in Fig. 6.10 corresponded to the transitions
between the ground states, the second states, and the third states of electrons
and holes, respectively, in descending order of wavelength. Note that a
shoulder peak at around 0.9 pm originated from the wetting layer just below
the quantum dots.
In Fig. 6.10, with an increasing injection current, the emission peak shifts
from the longer wavelength side to the shorter wavelength side. This is
mainly due to the band-filling effect. The stronger emission at higher-order
subbands directly reflects the larger degeneracy of eigenstates at them. There
are more states to be filled with carriers, at these subbands, which leads to
a stronger emission. Slight shifts of the peaks to the shorter wavelength side
with an increasing injection current are mainly due to the overlap of the
emission from the higher-order subbands. Because the fluctuation in dot size
and composition causes the large broadening of each peak, the peak
256 HAJIMESHOJI
L = 300 ;m
as-cleaved
3rd 2nd
50?* I 1
800 900 1000 1100 1200

Wavelength (nm)
FIG.6.10. Electroluminescence spectra for several injection currents, showing clear discrete
peaks corresponding to the subband of quantum dots. The peak in the longest wavelength side
is from the ground state. (From Shoji et al., 1997a. C 1997 IEEE.)
wavelength of the lower-order subbands is shifted to the shorter wavelength

side by the increased emission from the higher-order subbands. If a
quantum-dot structure with a smaller size fluctuation can be fabricated, this
kind of peak shift will vanish.
Lasing Churucteristics
Laser oscillation at room temperature has been achieved in self-assembled

In(Ga)As-GaAs quantum-dot lasers by several groups (Kirstaedter et al.,
1994; Mirin et al., 1996; Kamath et al., 1996; Shoji et al., 1996a and b). An
example of cavity-length dependencies of threshold current density for lasers
with different numbers of dot layers and different facet reflectivities is shown
in Fig. 6.11 (Shoji et al., 1997a). Here, the threshold current density was
normalized by a value for one dot layer for comparison. The open circles
represent as-cleaved devices, and the closed circles represent devices with
high-reflection (HR) coating on both facets. As shown in the figure, the
reduced cavity loss by the use of long cavity structures and the enlarged
optical gain by multiple stacking of the dot layers led to the reduction of
the threshold current density. A 95%-HR coating also greatly reduced the
threshold current density, especially in short cavity lasers. In 600-pm-long
and 900-pm-long lasers with triple-stacked dot layers and HR-coated facets,
QUANTUM 257
300 K, pulsed
I
I I
300 600 900 1200
Cavity length (pm)
FIG. 6.11. Cavity-length dependence of threshold current density for quantum-dot lasers
with different numbers of dot layers and different facet reflectivities, Niayeris the number of
quantum-dot layers in the active layer. The threshold current densities are normalized for one
dot layer for comparison. (From Shoji et al., 1997a.)
the threshold current densities for a dot layer were smaller than 1 kA/cm2.
In these lasers, room-temperature CW operation was achieved (Shoji et al.,
1996b).
Figure 6.12 shows light output versus current characteristics for a
CW-lasing quantum-dot laser. In this case, the threshold current was about
140 mA at 20"C, which corresponded to the threshold current density of
3.1 kA/cm2. Much lower threshold current density, below 400 A/cm2 at
room temperature, has also been reported (Kamath et al., 1997b; Hein-
richsdorff et al., 1997). The difference in the threshold current density is
mainly due to the difference in the emission efficiency of quantum dots.
Reduction of nonradiative recombination as a result of improved crystal
quality leads to the improved emission efficiency, especially at room tem-
perature, which results in lower threshold current density.
Various parameters are derived from the mirror loss dependence shown
in Fig. 6.11. The internal loss, m i , the differential quantum efficiency, yld, and
the internal quantum efficiency, q i are important factors in laser quality.
Although these parameters strongly depend on the waveguide structure and
the quality of the active region, those values can be derived from a
relationship expressed by Eq. (6.8). Experimentally, the internal loss, ( x i , was
estimated to be 3-5 cm-' (Shoji et al., 1997a). Of course, this value might
depend on the density of quantum dots in the active layer. The internal
258 HAJIME
SHOJI
1.o
N,aye, = 3,L = 900 pm
HR-coated
0.8 20°C
z
v
cw
c
0.6
3
Q
c
3
0.4
E0)
7
0.2
a
100 200 300 400
Current (mA)
FIG. 6.12. Light output versus current characteristics for a quantum-dot laser under the
CW condition. Triple-stacked quantum dots are in the active layer (PIloyer = 3). and the
cavity length is 900pm ( L = 900pm). The threshold current is about 140mA at 20’C, which
corresponds to the threshold current density of 3.1 kA/cm2.
quantum efficiency is still lower than that of well-established quantum-well

lasers, which have an internal quantum efficiency of nearly 100%. The value
of quantum-dot lasers is still between 40% (Bimberg et al., 1997) and 70%
(Ustinov et al.., 1996). A large amount of nonradiative recombination at
room temperature might be the cause of the small value. The phonon
bottleneck, discussed in Chapter 5, might also degrade the internal quantum
efficiency.
The gain properties of quantum dots are influenced by spectral charac-
teristics dependent on device parameters such as number of dot layers and
cavity loss. Figure 6.13 shows a systematic comparison of wavelength
behaviors of quantum-dot lasers with different device parameters (Shoji et
al., 1996a, 1996b, and 1997a). It indicates clearly that increased optical gain
in the multi-stacked dot layer and decreased cavity loss suppressed the
emission from higher-order subbands, resulting in lasing at lower-order
subbands. In a comparison using lasers with a single dot layer, when the
cavity length was 300 pm long, lasing occurred at the wetting layer, while in
a 900-pm-long laser with smaller cavity loss, lasing occurred at the third-
order subband. In the same manner, in a comparison using 900-pm-long
lasers, the increase in dot layers from one to three led to second-state lasing
from third-state lasing. Further reduction of the cavity loss by the HR
coating on both facets of the laser finally enabled CW lasing at the ground
QUANTUM 259
rayer =3 175 mA (CW)

= 900 pm ground state
WHR) 1
50
20
10
5
550 rnA (pulsed

layer = 3 2nd state
100
20
layer'1 360 mA (pulsed)

= 900 pn 3rd state
/aver = 1
-300 vm
800 900 1000 1100 1200

Wavelength (nm)
FIG.6.13. Wavelength behaviors of quantum-dot lasers with different device parameters.
The increased optical gain in the multi-stacked dot layer and the decreased cavity loss
suppressed the emission from higher-order subbands, resulting in lasing at the lower-order
subbands. In a laser with triple-stacked dot layers (Nlayer
= 3) and a 900-pm-long cavity. CW
lasing at the ground state was observed. (From Shoji et a]., 1997a. $? 1997 IEEE.)
state. A key point here is that the lasing wavelength did not change
continuously from the higher-order subbands to the lower-order subbands,
but changed discontinuously, reflecting the existence of discrete quantum-
confined subbands. A relationship between the lasing wavelength and the
threshold gain for one dot layer, as shown in Fig. 6.14, is evidence that the
lasing wavelength jumps from the second state to the ground state with a
lowering of the threshold gain for one dot layer. The critical threshold gain
260 HAJIME
SHOJI
. ' - I 1 .
0 2 4 6 8 1 0
Threshold gain for one dot layer (cm-I)
FIG 6 14. Relationship between the lasing wavelength and the threshold gain for one dot
layer. The lasing wavelength jumps from the second state to the ground state when the
threshold gain for one dot layer is lowered by the use of a long cavity structure and a multiply
stacked dot layer (From Shoji et al. 1997a ( 1997 IEEE)
for one dot layer was found to be 3.5 cm - in this case, as marked by a
dashed line (Shoji et al., 1997a).
A relationship between modal gain and current density can be also
derived from the cavity-loss dependence of threshold current density, as
shown in Fig. 6.15 (Shoji et al., 1997a), where both values were normalized
for one dot layer. It was found that the differential gain for the current
density depended on the subband at which lasing occurred. The different
differential gains can be attributed to the variation in degeneracy between
the ground state and the second state-the larger degeneracy at the higher
state leads to the larger differential gain. This means that lasing at the
ground state reduces the threshold gain for one dot layer by multiply
stacking the dot layer, increases the optical gain for one dot layer by
increasing the in-plane dot density, or reduces the cavity loss by the use of
a long cavity structure and HR coating. At around the cross-point of the
two gain curves, which coincides with the critical threshold gain of 3.5 cm-
in Fig. 6.14, lasing occurred at the two subbands at the same time because
of comparable modal gains at the two subbands at a given current density.
-
Measurement of material gain was also carried out (Kirstaedter et al., 1996).
A large gain of 7 x lo4cm- was reported at a current injection level
of only 80 A/cm2.
Another important feature observed in actual quantum-dot lasers is gain
saturation (Schmidt et al., 1996). For example, as shown in Fig. 6.13,
QUANTUM 261
0.1 1 10
Current density for one dot layer (kA/crn2)
FIG. 6.15. Relationship between the modal gain and the current density derived from the
cavity-loss dependence of the threshold current density, where both values were normalized for
one dot layer. It was found that the differential gain for the current density depended on the
subband at which lasing occurred. (From Shoji et al., 1997a. c? 1997 IEEE.)
although CW lasing at the ground state was achieved in a 900-pm-long

HR-coated laser, the lasing wavelength of 1105 nm was about 25 nm shorter
than the ground-state PL peak wavelength of 1130nm. This is the gain
saturation effect in quantum dots. In quantum-dot structures with large size
and composition fluctuations, only the dots with relatively high density can
contribute to lasing because gain from a single quantum dot is almost
constant. Consequently, carrier density in low-density dots is not clamped
even at the threshold of the high-density dots, and the carriers begin to
distribute to the second state after the ground state has been filled with
carriers, resulting in the generation of optical gain at the tail of the second
state. This situation is essentially different from that of conventional bulk
and quantum-well lasers. The total optical gain takes the maximum value
at a wavelength shorter than the PL peak. In fact, a lower threshold gain,
as shown in Fig. 6.14, led to a longer lasing wavelength because carrier
distribution to the higher-order subbands was suppressed.
The gain saturation effect sometimes appears in light-current character-
istics. Even once lasing starts at the ground state just above the thres-
hold current, if the optical gain at the second state is close to the gain at
the ground state, current injection induces the increase in optical gain at the
second state due to the gain saturation at the ground state, resulting in the
beginning of lasing at the second subband. This behavior is clearly shown
in Fig. 6.16. At the threshold current, only a sharp peak was observed in the
lasing spectrum. However, the spectrum was broadened because many dots
262 HAJIMESHOJI
2nd j Ground
state \ state
20 C
Pulsed
1 A,:
' A
ncn
1000 1050
..-.?length (nm)
Current (A)
FIG. 6.16. Large kink observed in their light-current characteristics of a quantum-dot laser.
Because of gain saturation at the ground state, the lasing spectrum broadened with an increase
in the injection current, and lasing at the second state began. At that point a large kink in the
light-current curve was observed.
began to lase with increased current injection; lasing finally began at the
second state due to the gain saturation at the ground state. For that reason,
a large kink was observed in the light-current curve at which lasing at the
second state started.
Temperature Depetzdence
Temperature dependence of lasing characteristics is of great interest both

practically and physically. Temperature-dependent threshold characteristics
and spectral characteristics are popular targets of research.
In Fig. 6.17, the temperature dependency of lasing wavelength for two
lasers are compared, together with that of PL peak wavelength (Shoji et al.,
1997a and b). The two lasers lased at different subbands at room tempera-
ture. One oscillated at the ground state (type-A; 900-pm-long HR-coated
laser); the other oscillated at the second state (type-B; 900-pm-long as-
cleaved laser). With a lower temperature, interesting changes were observed
in the laser's spectral characteristics. Type A's lasing wavelength gradually
QUANTUM 263
Type B, as-cleaved
1000 . ' ' .' .. .'. .. '. . ..' ...
' ' '
50 100 150 200 250 300

Temperature (K)
FIG. 6.17. Temperature dependencies of lasing wavelength for two lasers together with that
of the PL peak wavelength. The two lasers lased at two different subbands at room
temperature; one oscillated at the ground state (type A, 900-pm-long HR-coated laser), and the
other oscillated at the second state (type B, 900-pm-long as-cleaved laser). With a lowering
temperature, interesting spectral behaviors were observed. (From Shoji et al., 1997a. (" 1997
IEEE.)
shifted to the shorter wavelength side in proportion to the temperature

dependence of the band-gap energy. At the same time, the wavelength
difference between the PL peak and the lasing wavelength became smaller,
converging at the PL peak when the temperature went below 100K. In
contrast, type-B's lasing wavelength jumped to the longer wavelength side
at around 260 K. Lowering the temperature increased the emission efficiency
and reduced the thermal excitation of carriers to the second state. This
allowed lasing at the ground state at the longer side of the second state.
Further lowering of the temperature caused the lasing wavelength to
converge at the PL peak as in type A. This result reveals again that the two
lasers were oscilating at different states at room temperature. A similar result
was reported by others (Bimberg et al., 1996, Heinrichsdorff et al., 1997).
This comparison of temperature dependence indicates that smaller thresh-
old gain and less distribution of carriers to higher-order states lead to a
smaller difference between the lasing and PL wavelengths. This connection
fits well with the relationship between lasing wavelength and threshold gain
shown in Fig. 6.14.
264 HAJIMESHOJI
Figure 6.18 shows the EL and lasing spectra of the type-A laser in Fig.
6.17 for three temperatures: 60, 200, and 300 K (Shoji et a]., 1997a and b).
Distinct differences were observed in the spectra. At 300 K, the threshold
current was large and the EL spectra below the threshold were asymmetri-
cally shaped. This indicates that the injected carriers distribute to the second
state because of high threshold carrier density, which is caused by poor
emission efficiency of quantum dots and large thermal excitation of carriers
from the ground to the second state at room temperature. Consequently, the
EL peak, which coincided with the PL peak at a low current-injection level,
exhibited a large shift to the shorter wavelength side with an increase in
current. On the other hand, with a lowering of temperature, the FWHMs of
J/aver = 3,L = 900 pm

iR-coated
300 K
150.
100.
900 1000 1100 1200

Wavelength (nm)
FIG.6.18. EL and lasing spectra of the type-A laser from Fig. 6.17 for three temperatures
of 60, 200, and 300 K. A change in the carrier distribution to higher-order states was seen in
the temperature-dependent spectra. Multi-wavelength of operation peculiar to the quantum-
dot laser was observed at 60 K. (From Shoji et al., 1997a. (c) 1997 IEEE.)
QUANTUM 265
the EL spectra decreased as the threshold current decreased, and the

emission from the second state was well suppressed. The EL spectra at 60 K
had almost symmetrical shapes, and the peak wavelength of the lasing
spectrum almost coincided with the PL peak. Because of the large decrease
in the threshold carrier density and the thermal excitation of the carriers,
the effect of higher-order states became negligible, and only the gain at the
ground state contributed significantly to the lasing.
Interestingly, the lasing at many wavelengths was sometimes observed at
60K, as shown in Fig. 6.18. This was not the multi-mode oscillation
observed in ordinary Fabry-Perot lasers. Many quantum dots with different
sizes and compositions were lasing at the ground state at the same time
because of the reduced threshold carrier density and the increased emission
efficiency at 60 K. In Fig. 6.18, the wavelength span of lasing at 60 K was as
wide as 50 nm, which was much wider than that of the Fabry-Perot lasers.
However, the span decreased as in the lasing spectrum at 200K which is
shown in Fig. 6.18. This result clearly shows that the quantum dots are
spatially separated in the active layer but optically coupled through the
waveguide structure. It also suggests that improved dot uniformity will
result in lasing in the narrow wavelength span, leading to an extremely low
threshold current, as theoretically predicted.
Figure 6.19 shows an example of the temperature dependency of threshold
.- 7'
t
h
.mm rn a - PL .-cn
z-
E
-I
C
3
-
v R
\
0
41K
Type-8 (as-cleaved) A
x
.-c
cn
C
a,
c
50 100 150 200 250 300

Temperature (K)
FIG.6.19. Temperature dependence of the threshold current for type-A and type-B lasers
from Fig. 6.17 together with that of the PL intensity. (From Shoji et al., 1997a. CI 1997 IEEE.)
266 HAJIMESHOJI
currents for type-A and type-B lasers, as already shown in Fig. 6.17 (Shoji
et al., 1997a and b). Temperature dependence of P L intensity is also shown
for comparison. The type-A laser exhibited a high characteristic tempera-
ture, To, defined by
);(
I , , x exp (6.12)
of over 300K in the temperature range below 200K, and the threshold
current density was less than 600 A/cm2. In particular, the characteristic
temperature was as high as 477 K in the temperature range of 60- 110 K,
and the threshold current density was less than 450 A/cm2. However, in the
temperature region over 200 K, the characteristic temperature rapidly de-
creased to a value below 100 K . It was as small as 48 K at room tempera-
ture.
As for the type-B laser, although it also exhibited a high characteristic
temperature of over 400K in a temperature region below 100K, the
characteristic temperature started to decrease at 170 K, which was a lower
temperature compared with that in the type-A laser. The reason for this
unusual behavior of the threshold current at temperatures of 120-130 K ha5
not been determined. Another unusual behavior in the temperature range of
220-260 K -the wavelength shift between the ground state and the second
state observed at around 260K, shown in Fig. 6.17-might be attributed
to this phenomenon. As for the value of the characteristic temperature, a
much higher value of 530 K over a temperature range of 80-220 K has been
reported by Bimberg et al. (1997).
It can be assumed that the temperature dependence of the threshold
current is strongly related to that of the P L intensity; that is, the decrease
in the PL intensity with increasing temperature is due to the increased
nonradiative recombination in the quantum dots, which results in the
proportional increase in the threshold current. However, the results shown
in Fig. 6.19 do not necessarily support this assumption. The PL intensity
begins to decrease at 150 K, whie the rapid increase in the threshold current
starts at 170 K in the as-cleaved laser and at 200 K in the HR-coated laser.
Although it is qualitatively correct that a decrease in the P L intensity leads
to an increase in the threshold current, the relationship between the two
values is not so simple. Further investigation should be done on temperature
characteristics, because the energy separation between the subbands is also
related to them; nevertheless, the energy separation between the ground
state and the second state is much larger than kT of 26 meV, even at room
temperature. On the other hand, the comparison between the HR-coated
laser and the as-cleaved laser shows that a lower threshold current leads to
better temperature characteristics in a higher temperature region. Reduction
6 DOTLASERS
QUANTUM
SELF-ASSEMBLED 267
of the threshold current is essential in improving temperature character-

istics. Also critical is optimizing the energy separation between the subbands
of quantum dots as well as improving emission efficiency.
Modulation Characteristics
As described in Section 11, enhanced differential gain due to the atom-like

density of states in quantum dots is expected to bring about excellent
modulation in quantum-dot lasers. There have been several studies on the
frequency response of quantum-dot lasers through the measurement of
direct modulation bandwidth by current injection and the measurement of
relative intensity noise (RIN).
In the small signal modulation, a 3-dB bandwidth, f3dBr of 7.5 GHz and
8.5GHz has been measured using lasers lasing at the ground state of
quantum dots. This modulation bandwidth corresponds to the differential
gain of 2-6 x 10- l4 cm2, which is higher than the best values reported for
multiple quantum-well lasers (Kamath et al., 1997a and b).
Measurement of RIN also gives much information on the dynamic
properties of quantum-dot lasers because this measurement is free from the
C R constant dependent on device structures. In this case, lasers to be
measured are operated under the CW condition, and an observed peak
frequency in the noise spectrum corresponds to the relaxation oscillation
frequency, f ; . The 3-dB bandwidth of the modulation response is related to
the relaxation oscillation frequency by fSdB = 1.55f,. With this method, a
relaxation oscillation frequency of 5.3 GHz corresponding to ,f3,, of
8.2 GHz has been reported (Ma0 et al., 1997).
As described above, modulation bandwidth or resonant oscillation fre-
quency of over several tens of GHz has been theoretically predicted for an
ideal quantum-dot laser. The possible reason for this is twofold: (1)
insufficient uniformity of dots might fail to fully extract the potential wide
bandwidth, and (2) a relatively long relaxation time from the upper states
to the lower states in quantum dots -the so-called phonon bottleneck ~
might limit the bandwidth. Although both factors should contribute to the
bandwidth, it is not clear at present which is the dominant factor. Further
investigation of carrier dynamics in quantum dots is needed.
3. LIMITING
FACTORS
OF LASERPERFORMANCE
Room-temperature CW operation at the ground state of quantum dots

can now be realized through use of self-assembled structure. The lasing
268 HAJIMESHOJI
characteristics are reaching those of well-established quantum-well lasers,

especially in threshold current density, However, the obtained performances
are still far from theoretical predictions (Arakawa and Sakaki, 1982; Asada
et al., 1986). In order to realize the predicted performance in quantum-dot
lasers, drastic improvement in fabrication technologies and deep insight into
the physics of quantum dots are required.
The key improvements necessary are as follows:
0 Dot uniformity, density, and size

0 Crystal quality
0 Phonon bottleneck
0 Emission wavelength
Precise control of dot size and composition is critical for proportional

enhancement of optical gain in the active layer. Nonuniformity measured
through FWHMs of the PL spectra should greatly reduce. A PL spectrum
with an FWHM below lOmeV, which is comparable to or narrower than
that of a typical quantum well, is preferable for extracting the atom-like
nature of quantum dots.
Optimization of dot density and size is also required. Optimum density
should be based on dot uniformity. If the uniformity is not sufficient, we
need to increase the density, whereas a low density is sufficient for high
uniformity. Dot size is critical for temperature-insensitive operation in a
high temperature range. If the dot is large, it contains several excited states.
A small energy difference between the excited states leads to carrier
distribution to the upper states, resulting in an increase in threshold current
and poor temperature characteristics. In contrast, if the dot is small, only
the ground state can exist in it. In this case, however, the energy difference
between the energy levels of the barriers and the quantized states becomes
a key parameter. A wider bandgap barrier might also be required. As
discussed in the following section, vertically coupled quantum-dot structures
work well in controlling dot uniformity and size.
For improved device performance, crystal quality also needs to be
improved. As shown in Section 111.2, the internal quantum efficiency of
quantum-dot lasers is typically less than 80%, which is inferior to that
of quantum-well lasers. At present, one of the major causes of poor internal
quantum efficiency is insufficient crystal quality. As shown in Fig. 6.19,
threshold current tends to increase when PL intensity decreases. It seems
certain that the existence of nonradiative centers inside or outside of
quantum dots degrades emission efficiency, especially in the room tempera-
ture region. Although further investigation is required to identify the origin
of the nonradiative center, an improved growth method would lead to
QUANTUM 269
enhanced performance. In fact, excellent threshold characteristics are

achieved in lasers with defect-free InAs-GaAs quantum dots grown by
MOVPE (Heinrichsdorff et al., 1997). Note that, in addition to improving
the crystal quality of the dots themselves, as discussed in Chapter 1, the
crystal quality of the barrier layer should also improve because, if there is a
large number of defects in the barrier, the injection efficiency of carriers into
quantum dots decreases markedly.
It was predicted that the discrete levels in quantum dots might severely
disturb carrier relaxation from the excited states to the ground states
(Bockelmann and Bastard, 1990; Benisty et al., 1991). This is the so-called
phonon bottleneck effect. Whether this effect creates serious problems for
quantum-dot laser operation is hotly debated. Some theoretical studies say
that multiphonon relaxation enhances carrier relaxation (Inoshita and
Sakaki, 1992; Heitz et al., 1996), while some experimental results indicate a
carrier relaxation time of 10-loops (Mukai et al., 1996b and c). As
described in Chapters 1 and 5, lasing characteristics are greatly influenced
by carrier relaxation time. In particular, modulation characteristics, differ-
ential quantum efficiency, and maximum output power are severely limited
by the photon bottleneck, erasing the advantages of quantum-dot lasers
over conventional quantum-well lasers (Sugawara et al., 1997b). We need to
clarify the effect of the phonon bottleneck. If it is serious, a new structure
for accelerating carrier relaxation must be designed and fabricated to avoid
the bottleneck. In the design of dot size as well, the phonon bottleneck
should be taken into account.
The above issues have been discussed from a physical point of view. From
a practical point of view, emission wavelength from quantum dots becomes
important. For example, in InAs-GaAs quantum dots, the emission
wavelength is typically between 0.9 pm and 1.2 pm, for which unfortunately,
it is not easy to find practical applications. For that reason, a new dot
structure must be realized that emits at practical wavelengths such as 1.3 pm
and 1.55 pm, which are widely used in optical communications.
IV. Key Technologies for the Next Era
Quantum-dot structures grown in the SK mode have been a break-

through in fabricating quantum-dot lasers. However, their performance is
still unsatisfactory, and the problems discussed in the previous section must
be overcome before we see any improvement. In this section, we will look
at several emerging technologies that will be critical in the next era of
optoelectronics.
270 HAJIME
SHOJI
1. CLOSELY QUANTIJM-DOT
STACKED LASERS
Multiply stacked quantum-dot structures increase optical confinement to

the active layer. Unlike in-plane density, dot density can be easily controlled,
and optical confinement can be more widely tuned only by changing the
number of stacking layers. For this purpose, the first approach, shown in
Fig. 6.5, was to grow multistacked dot structures with thick ( > 7-8 nm)
barriers. In this case, the dot layers are not electronically coupled in the
vertical direction due to the thick barriers. Each dot layer acts as an
independent active layer. In this approach, the increased number of dot
layers leads almost proportionally to enhancement of total modal gain, but
modal gain from a single dot layer is unchanged.
Closely stacked dot structures, that is, multistacked dot structures with
thinner barriers, have been investigated by many (Solomon et al., 1996;
Ledentsov et al., 1996; Nakata et al., 1997). Figure 6.20 shows (a) a
cross-sectional TEM photograph of a InAs/GaAs closely stacked quantum-
dot structure and (b) the corresponding schematic illustration (Shoji et al.,
1997~).By use of the thin GaAs barriers, the upper-layer dots were
FIG.6.20. Closely stacked quantum-dot structure. (a) a cross-sectional TEM photograph.

(from Shoji et al., 1997c), and (b) the corresponding schematic.
QUANTUM 271
self-aligned just on the lower-layer dots, forming a columnar structure,

where the GaAs barriers separating the dot layers were as thin as 3 nm, and
fivefold InAs quantum dot layers were grown (Nakata et al., 1997; Shoji et
al., 1997~).Each dot layer of the fivefold structure was grown in the SK
mode. It was found that the vertically aligned dots equivalently act as a
single dot due to electronic coupling through tunneling in the vertical
direction. Measurement of the exciton diamagnetic shift confirmed that the
vertically aligned dot has an almost spherical quantum-confined potential
and that the luminescence is due to three-dimensionally confined exci-
tons (Sugawara et al., 1997a). The SK growth technique is described in
Chapter 2.
The closely stacked quantum-dot structure has features that make it
attractive for laser application. The first one is the narrow emission
spectrum obtained. Figure 6.21 shows a PL spectrum of the closely stacked
quantum dots at 4.2K (solid curve) together with that of single-layer
quantum dots without electronic coupling (dashed curve). The FWHM is as
small as 25 meV in the closely stacked dot structure, which is much smaller
than that of the conventional structures (Sugiyama et al., 1997; Shoji et al.,
1997~).We can expect enhancement of optical gain due to small in-
homogeneous broadening. One reason for the narrowed emission spectrum
is the effect of size averaging during the stacking process. The upper-layer
dots are grown on seed potentials formed by lower-layer dots, which results
1
.e I- ' I I I t
8
8
'I
' I
0.8 0.9 1.0 1.1 1.2 1.3

Wavelength (vm)
FIG.6.21. PL spectrum of a closely stacked quantum dot at 4.2 K (solid curve). together
with that of a conventional quantum dot without electronic coupling (dashed curve). The
FWHM is as small as 25 meV in the closely stacked dot structure. (From Shoji et al., 1997c.)
272 HAJIME SHOJI
in size averaging in the lateral direction. The other reason is the larger
volume of quantum dots. Because the effective vertical height becomes
greater than that of single-layer dots, the emission wavelength becomes less
sensitive to fluctuations in dot size. The red shift of the PL peak in Fig. 6.21
is further evidence of the larger volume of quantum dots due to the
electronic coupling of stacked dot layers.
The second attractive feature is enhancement of modal gain from a single
dot layer. In the closely stacked structure, optical confinement to the active
layer is enhanced by the number of stacked layers. Because the closely
stacked dots act as a single dots, the modal gain obtained from a single dot
layer becomes larger. Larger optical gain is expected in the closely stacked
structure at an injection level that compares with that of the structure
without vertical coupling. Effective dot height can be precisely controlled by
changing the number of stacking cycles.
Laser oscillation has been achieved by use of the closely stacked quan-
tum-dot structure. Shoji et al. (1997~)reported pulsed lasing at room
temperature and CW lasing at 50 K. The threshold current density at 50 K
was 520 A/cm2. As shown in Fig. 6.22, lasing occurred at the second
subband at 50 K. Bimberg et al. (1996) demonstrated a laser oscillation
using a different closely stacked structure, where the thickness of GaAs
barriers separating InAs dot layers was 6nm. Even for this rather thick
barrier, a decrease in the FWHM of the quantum-dot emission down to
L = 900 pm
as-cleaved
>
.-
c
u)
c
0)
-
c
C
900 1000 1100 1200

Wavelength (nm)
FIG. 6.22. EL and lasing spectra of closely-stacked quantum dot lasers for various injection
currents at 50 K. Lasing occurs at the second state. (From Shoji et al., 1997c.)
6 SELF-ASSEMBLED
30-50meV was observed. As for lasing, a threshold current density of

270 A/cm2 was achieved at 77 K, and improvement in the characteristic
temperature was confirmed.
The problem with this structure was its large nonradiative recombination
rate, which led to the drastic increase in the threshold current at room
temperature. This might be related not only to the growth conditions, which
was not yet optimized, but also to the increased number of growth interfaces
in the closely stacked structure. In the result shown in Fig. 6.22, the
temperature dependence of the PL intensity showed that the peak intensity
at room temperature was less than one-thousandth of that at 77K. In
addition, the emission efficiency at 77 K was lower than that of conventional
single-layer dots grown in the SK mode. These properties suggest that the
closely stacked dots still contained many defects or nonradiative centers
introduced during the growth, which degraded the optical properties of the
quantum dots even at low temperatures. By improving the growth tech-
nique, we will see lower threshold current density operation at the ground
state of quantum dots and improved gain properties due to the reduced
fluctuation of size and composition in quantum-dot lasers with a highly
uniform closely stacked structures.
2. COLUMNAR
QUANTUM-DOT
LASERS
More recently, improved lasing characteristics have been demonstrated

by Mukai et al. (1998a) using a columnar quantum-dot structure shown in
a cross-sectional TEM image in Fig. 6.23. It is similar to the closely stacked
structure with a columnar shape, but the GaAs barrier layers separating the
InAs islands are much thinner. Because only three monolayers of GaAs were
supplied on the InAs islands in each cycle of the stacking process, the
existence of GaAs barriers was not clearly observed between the upper and
lower islands. InAs islands produced during a cycle of InAs supply seemed
to be contacting each other, forming equivalent single dots with strong
electronic coupling in the vertical direction. The diameter and the height
were about 15 nm, and the ground-state emission peak was observed at
1.17 pm.
In the columnar quantum dots, the PL emission intensity was improved
by three orders of magnitude compared with that of the closely stacked dots.
Although the FWHM of the PL spectrum was about 40meV, which was
larger than that of the closely stacked dots, the value was almost half that
of ordinary quantum dots without electronic coupling. Figure 6.24 shows
light output versus current characteristics of a 900-pm-long device with
as-cleaved facets for various temperatures up to 70°C. The threshold current
274 HAJIMESHOJI
I I
10 nm
FIG. 6.23. Cross-sectional T E M image of a columnar quantum dot.
was 31 mA at 2 5 T , which corresponds to the threshold current density of

500 A/cm2. This value was improved by two orders of magnitude compared
with as-cleaved devices with closely stacked dot structures. As the tempera-
ture increased from 25°C to 70°C the threshold current increased up to
54 mA, which corresponds to the characteristic temperature of 81 K. Fur-
thermore, a large output power without saturation was maintained up to
high temperature, and the temperature dependence of the slope efficiency
was considerably smaller, as shown in Fig. 6.24. In the low temperature
range, the temperature characteristics were much better than those at room
temperature. The threshold current density at 100 K was as low as 80 A/cm2,
and the characteristic temperature defined between 100 K and 160 K was as
high as 487 K. Lasing occurred at the second state at room temperature and
at the ground state below 220 K. In devices with smaller mirror loss-for
example through high-reflection coating applied to the facets- room-tem-
perature lasing at the ground state with much lower threshold current, much
better temperature characteristics. and greater output power is expected
(Mukai et al., 1998b).
Greater output power without any saturation and greatly reduced thresh-
old current are attributed not only to improved emission efficiency but also
to a possible enhancement of the carrier relaxation rate. Figure 6.25 shows
the PL spectrum measured at 300 K together with that of the single-layer
dots grown in the SK mode. An important point here is the position of the
emission peak from the wetting layer. The peak wavelength was at around
1.0 pm, which was 0.1 pm longer than that of single layer dots (Shoji et al.,
6 SELF-ASSEMBLED
QUANTUM
DOTLASERS 275
F
E
v
0 20 40 60 80 100 120
Current (mA)
FIG. 6.24. Light output versus current characteristics of a columnar quantum-dot laser.
showing the result of a 900-pm-long device with as-cleaved facets for various temperatures up
to 70°C. The threshold current is 31 mA at 25"C, which corresponds to the threshold current
density of 500 A/cm2. (Reprinted with permission from Mukai et al., 1998. :p 1998 IEEE.)
- 20 meV
I
-
300K
Columnar dots
H I
ui
.-c
c
3
$ -
v
Wetting layer
a Wetting layer
0.8 0.9 1.0 1.1 1.2 1.3 1.4

Wavelength (Vm)
FIG. 6.25. PL spectrum of columnar quantum dots at 300K together with that of a
single-layer dot grown in the SK mode.
276 HAJIMESHOJI
1997a). The emission from the wetting layer at a longer wavelength was due
to electronic coupling of thin wetting layers associated with extremely thin
barrier layers separating the dot layers, which resulted in the formation of
an effectively thick wetting layer, as in a quantum well. As discussed in
Chapter 5, the lasing characteristics of quantum-dot lasers are subject to the
carrier relaxation rate from the continuous level to the quantum confined
states (Sugawara et al., 1997b). If the energy difference between the continu-
ous state in the wetting layer and the discrete levels of quantum dots is
smaller, the carrier relaxation rate should be enhanced by the assistance of
the multiphonon effect (Inoshita and Sakaki, 1992) because the number of
required multi-phonons is smaller for the shallow potential.
3. LONG-WAVELENGTH
QUANTUM-DOT LASERS
Quantum dots that emit in the long wavelength region such as 1.3 pm and
1.55pm are of great practical interest. They can be used in a variety of
practical devices, such as optical communication and optical interconnects.
Emission in the 1.3-pm region has been reported by several groups in
InAs/GaAs or in InGaAs/GaAs quantum-dot structures on GaAs substrates
(Mukai et al., 1994; Tackeuchi et al., 1995; Mirin et al., 1995). On InP
substrates, emission in the 1.4- 1.7-pm spectral region has been demon-
strated in InAs/InAlAs or InAsPnGaAsP quantum-dot structures (Fafard et
al., 1996; Nishi et al., 1998).
As described in Chapter 3, the ALS quantum-dot structure, which is
self-organized by alternately supplying monolayers of InAs and GaAs in the
atomic layer epitaxy (ALE) mode, holds promise for realizing long
wavelength emission in quantum dots. ALS dots are In,,,Ga,,,As quantum
dots typically with a diameter of 20 nm and a height of 10 nm, and they are
surrounded by a 10-nm-thick Ino,,Ga,,,As barrier in the lateral direction
and by a GaAs barrier in the growth direction (Mukai et al., 1994). The
emission wavelength of the ALS dots was about 1.3 pm at room tempera-
ture, as shown in Fig. 6.26, and the wavelength can be controlled in a range
of 1.1-1.5pm by changing the growth conditions, such as the number of
cycles for alternate supply of InAs and GaAs monolayers and the composi-
tion of the InGaAs buffer layer on which quantum dots are grown (Ohtsuka
et al., 1995). The FWHM of the ground-state PL peak was as narrow as
35 meV. These features are quite attractive for laser application.
Using the ALS dots, quantum-dot lasers were fabricated by Shoji et al.
(1995). One is schematically shown in Fig. 6.27. In the active layer, InGaAs
quantum dots were self-organized by performing 12 cycles of (lnAs),/
(GaAs), short-period growth. Figure 6.28 shows the emission spectra for
QUANTUM 271
Wavelength (vm)
FIG. 6.26. PL spectrum of ALS quantum dots at 300K. The emission peak is at 1.3 jtm,
and the FWHM is as small as 35meV. (From Shoji et al., 1995. Ci 1995 JEEE.)
various injection currents at 80 K. Although individual peaks were not as

clear as those in the PL spectrum shown in Fig. 6.26, discrete quantum levels
in the EL spectra were found under current injection as well as under
photo-excitation (Mukai et al,, 1996b). Band filling with the current injec-
tion was clearly observed in this measurement. Several high-order levels
appeared in the shorter wavelength region with increasing injected current
because of a higher excitation level compared with the PL measurement.
Laser oscillation was achieved at the threshold current of 1.1A under a
Ino,Gao ,As
p+-GaAs contact (0.5 pm) Quantum dot
- - - - - - - - -'\
*n-In, 47Ga053Pclad (1 pm)
n-GaAs buffer (0.5 pm) \
*I1: 1
In, ,Ga, ,As barrier
n-GaAs sub.
FIG. 6.27. Schematic of a n ALS quantum-dot laser. The active region consists of single-
layer ALS dots formed by twelve cycles of (InAs),/(GaAs), short-period growth. (From Shoji
et a]., 1995. (3 199.5 IEEE).
278 SHOJI
HAJIME
1.1 A (lasing) 80 K
-
.-v)
4-
40
C
3
.@ 30
Y
m
.220
v)
5
W
4-
C
- 10
0
800 1000 1200 1400
Wavelength (nm)
FIG. 6.28. Emission spectra for various injection currents at 80 K. Although the peaks are
not as clear as in the PL spectrum shown in Fig. 6.26, discrete quantum levels in the EL spectra
are found under current injection. Due to band filling with current injection, several high-order
levels appear in the shorter wavelength region because of a higher excitation level compared
with the PL measurement. Laser oscillation is achieved at the threshold current of 1.1 A under
the pulsed condition. (From Shoji et al., 1995. 8 1995 IEEE.)
pulsed condition, and the corresponding current density was 815 A/cm2. The
lasing wavelength was 91 1 nm.
Considering the fact that the compositional wavelength of the 10-nm-
thick In,,,Ga,,,As barrier layer at 80 K was shorter than 850 nm, the laser
oscillation was from a high-order sublevel of the quantum dots. Note that
the wetting layer, which is formed in the growth of quantum dots in the SK
mode is not observed in the ALS quantum-dot structure. To confirm that
the obtained laser oscillation was really from a high-order sublevel of
quantum dots, the diamagnetic shifts of the lasing peak were also carried
out. A strong magnetic field up to 13T was applied perpendicular to the
lasers, and the experimental results showed that the energy shift of the lasing
peak was much smaller than that of the quantum well, whose compositional
wavelength was close to that of the InGaAs barrier layer surrounding the
quantum dots. This indicates that the obtained laser oscillation was from
the quantum-confined level of the quantum dots. In this experiment,
although lasing was observed up to 140 K, room-temperature operation was
not achieved because the dot density was found to be quite low (a few
percent areal coverage) and only a single-sheet dot layer was grown in the
active layer. In addition, because of the limitation in growth equipment, the
QUANTUM 279
active layer was exposed to the air once before growing the p-type cladding
layer, so that nonradiative centers might have been introduced. Considering
the emission intensity at room temperature, as increase in the dot density
and improved growth would lead to 1.3-pm operation and a lower threshold
current at room temperature.
Mirin et al. (1995) reported a sharp emission at 1.3 pm from In,.,Ga,,,As
quantum dots grown in the SK mode. Although these dots were larger than
the ALS quantum dots, the FWHM of the emission was as narrow as
28meV. These researchers also succeeded in laser oscillation at room
temperature (Mirin et al., 1996). Because the lasing occurred at an excited
state, the lasing wavelength was 1.2 pm. However, this is so far the longest
lasing wavelength ever reported in self-assembled quantum-dot lasers grown
on GaAs substrates.
Nishi et al. (1998) reported a possibility of long wavelength quantum-dot
lasers on InP substrates. InAs quantum dots were grown on (311)B InP
substrates in the SK mode. Although lasing was still at low temperatures,
the lasing wavelength of 1.2- 1.4 pm and the threshold current of 540 A/cmZ
were demonstrated at 77 K. If room temperature lasing can be achieved, a
much longer lasing wavelength is expected in this structure. Note that in the
case of InAs dots on InP substrates the dot diameter is typically 30 nm or
more (Fafard et al., 1996; Nishi et al., 1998), which is larger than that of
InAs dots on GaAs substrates at present. This may originate from the small
lattice mismatch between the lnAs and the substrate. Extracting the atom-
like nature from quantum dots requires a smaller size.
4. QUANTUM-DOT
VERTICAL-CAVITY
SURFACE-EMITTING
LASERS
Recent remarkable progress in the development of vertical cavity surface

emitting lasers (VCSELs) has enabled the threshold current of semiconduc-
tor lasers to be drastically reduced. Ultra-low threshold currents below
100pA have been demonstrated in VCSELs in which strained quantum-well
structures were used in the active regions (Huffaker et al., 1994; Hayashi et
al., 1995). For much more advanced lasers, quantum-dot structures are an
attractive approach. Ultimately low threshold current operation can be
expected in quantum-dot VCSELs because of the enhanced gain character-
istics of the quantum-dot’s active region and the reduction of active volume
in the microcavity structure. Control of the electronic states and the photon
modes in the microcavity structure is another interesting scheme to be
studied from both practical and physical points of view (Baba et al., 1991).
As described in Fig. 6.29, if the cavity mode is completely matched with the
narrow gain bandwidth of quantum dots, a much lower threshold current
280 HAJIME
SHOJI
E
Mirror
Quantum dots
Mirror
Optical field
Photon mode
t -c
(n-1)th
nth
3rd 2nd Ground Wavelength
FIG. 6.29. Photon modes and electronic states in a quantum-dot microcavity, where both
the photon mode and the electronic state become discrete. Although lasing occurs only when
the photon mode and the electronic state match, an extremely low threshold current is
expected.
should be achieved because only a limited number of resonant modes can

exist in the microcavity.
Along with their merits, there is the potential that ideal quantum-dot
VCSELs might work only at a fixed operating condition. The narrow gain
bandwidth would make it difficult to match the gain peak with a resonant
wavelength of the cavity. A difference in the temperature dependence
between the gain peak and the resonant wavelength would also limit the
operating temperature or the operating current. On the other hand, in the
actual quantum-dot structure with large inhomogeneous broadening, a
small active region would effectively reduce the inhomogeneous broadening,
leading to the enhancement of gain properties. In this case, demonstrating
quantum-dot VCSELs is rather easier.
The first demonstration of quantum-dot VCSELs by current injection
was reported by Saito et al. (1996). Figure 6.30 shows a schematic structure
of the quantum-dot VCSEL. The active layer consisted of 10 periods of
In,.,Ga,.,As quantum-dot layer grown in the SK mode and a 10-nm-thick
Al,,,,Ga,,,,As barrier. The top and bottom mirrors consisted of 14.5-
QUANTUM 281
FIG. 6.30. Quantum-dot VCSEL with 10-period InGaAs dots in the active region. (Re-
printed with permission from Saito et al., 1996. 6 1996, American Institute of Physics.)
period and 18-period AIAs/GaAs distributed Bragg reflectors (DBRs),

respectively. The active layer was set at the center of the cavity, and the
cavity size was adjusted to one wave. The resonant wavelength was around
960 nm. CW lasing at room temperature was achieved in the fabricated
devices with 25-pm x 25-pm apertures, and the threshold current was
32 mA. From the measurement of emission spectra at various injection
currents, the lasing was found to be associated with a higher-order transi-
tion in the quantum dots.
It is well known that polarization of VCSELs on (100) substrates is not
easily controlled because the optical gain is isotropic in the (100) plane. To
control the polarization of VCSELs, Saito et al. (1997) proposed an
interesting application of quantum dots. They used the structural anisotropy
of the InGaAs quantum dot structure on (100) GaAs, rather than the change
in the density of states, for polarization control of VCSELs. The shape of
the grown dots was longer in the [ O i l ] direction on the (100) surface, which
resulted in a polarization dependence of PL intensity along the [ O i l ]
direction that was 1.37 times stronger than that along the orthogonal [011]
direction. This anisotropy resulted in the asymmetric gain characteristics in
the active layer plane of VCSELs. As shown in Fig. 6.31, in the VCSELs
with the anisotropic quantum dots, lasing occurred only in the [ O i l ]
polarization state, while the orthogonal [01 13 polarization was suppressed
by 18 dB. Saito et al. also achieved ground-state pulsed lasing in these
devices by increasing the periods of DBRs compared with their previous
work (Saito et al., 1996).
Sub-mA operation at room temperature has been achieved by Huffaker
et al. (1997). They introduced a dielectric aperture by the use of selective
oxidation of AlAs for three-dimensional optical confinement in the cavity
282 HAJIME
SHOJI
1.6
Pulse
1.4
1.2
2
E 1.0
v
4-.
I 2 0.8
2
0
E 0.6
rn
._
1
0.4
0.2
0
0 20 40 60 80 100 120
Current (mA)
FIG. 6.31. Light output versus current for two polarization states of a quantum-dot
VCSEL at room temperature. The orthogonal polarization suppression ratio is 18 dB at 1-mW
output power of a [OTl] emission. (Reprinted with permission from Saito et al., 1997. 0 1997
American Institute of Physics.)
(Dallesasse et al., 1990; Hayashi et al., 1995). The fabricated quantum-dot

VCSEL with a 7-pm-square active region yielded a low threshold current of
560pA at room temperature under a pulsed condition, and the lasing was
nearly from the ground states. A small aperture of less than 10pm asso-
ciated with the index guiding effect resulted in the sub-mA operation of the
quantum-dot VCSEL.
V. Conclusion
As described in this chapter, self-assembled quantum-dot structures

represent a breakthrough in quantum-dot lasers, which until that point had
not been easily realized despite much effort since they were first proposed in
1982. Room-temperature operation through the introduction of self-assem-
bled structures was a milestone in the history of quantum-dot laser research.
Measurement of the fabricated quantum-dot lasers has given us much
quantitative information. On the other hand, there is a big gap between
ideal and actual performance. Many problems must be solved if the
performance of quantum dots is to match what has been theoretically
QUANTUM 283
predicted. Although the self-assembled structure easily produces quantum

dots, the growth process is based primarily on spontaneous mechanisms,
and it is not yet fully controlled. We have to find a way to artificially control
the uniformity, size, density, and position as well as the crystal quality of
grown dots. In this sense, use of vertically coupled quantum-dots such as
closely stacked and columnar quantum dots is of interest. In addition, we
need to find an application for quantum-dot lasers. As described in Section
IV, lasing wavelength becomes a key issue in finding practical uses for them.
Realization of quantum-dot lasers that operate in practical wavelength
ranges, such as 1.3 pm or 1.55 pm, would be advantageous. Exploitation of
a new function for quantum dots would be also required.
Acknowledgments
The author would like to thank many colleagues, K. Mukai, Y. Nakata,

N. Ohtsuka, Y. Sugiyama, T. Futatsugi, Dr. M. Sugawara, Dr. S. Yamazaki,
Dr. N. Yokoyama, and Dr. H. Ishikawa who contributed to the work at
Fujitsu Laboratories Ltd. The author also would like to thank K. Otsubo,
Dr. H. Kuwatsuka, T. Fujii, Dr. S. Yamakoshi, Dr. K. Wakao, and Dr. H.
Imai for their encouragement and fruitful discussion.
REFERENCES
Ahopelto, J., Lipsanen, H., and Sopangen. M. (1995). Proc. 7th Con/: Indium Phosphide rind
Relutrd Materiuls ( I P R M '9s) FB2-4, 19-20.
Arakawa, Y., and Sakaki, H. (1982). Appl. Phys. Lett. 40, 939-941.
Arakawa, Y., Sakaki, H., Nishioka, M., Okamoto, H., and Miura, N. (1983). Jpn. J . A p p l . Phy.~.
22, L804-L806.
Arakawa, Y., and Yariv, A. (1986). IEEE J. Qimritton Electron. QE-22, 1887-1899.
Asada, M., Miyamoto, Y., and Suematsu, Y. (1986). ZEEE J. Quantuni Electron. QE-22,
1915-1921.
Baba, T., Hamano, T., Koyama, F., and Iga. K. (1991). IEEE J. Qumturn Electrori. QE-27.
1347 1 354.
-
Benisty, H.. Sotomayor-Torres, C. M., and Weisbuch, C. (1991). Phys. ReG. B.44, 10945- 10948.
Bimberg, D., Ledentsov, N. N., Grundniann, M., Kirstaedter, N., Schmidt, 0. G., Mao. M. H.,
Ustinov, V. M., Egorov, A. Y., Zhukov, A. E., Kop'ev, P. S., Alferov, 2. I., Ruvimov, S. S.,
Gosele, U., and Heydenreich, J. (1996). J p n . J . Appl. Phys. 35, 1311-1319.
Bimberg, D., Kirstaedter, N., Ledentsov. N. N., Alferov, Z. I., Kop'ev, P. S., and Ustinov, V. M.
( I 997). IEEE J. Selected Topics in Qurintuni Electron. 3, 196-205.
Bockelmann, U., and Bastard, G. (1990). Phys. Reu. B. 42, 8947-8953.
Dallesasse, J. M., Holonyak, N.. Sugg. A. R., Richard, T. A., and El-Zein, N. (1990). Appl. Phys.
Lett. 57. 2844-2846.
284 HAJIMESHOJI
Fafard, S., Wasilewski, Z., McCaffrey, J., Raymond, S., and Charbonneau, S. (1996). Appl. Phys.
Lett. 68, 991-993.
Fukui, T., Ando, S., Tokura, Y., and Toriyama, T. (1991). Appl. Phys. Lett. 58, 2018-2020.
Glaser, E. R., Benett, B. R., Shanabrook, B. V., and Magno, R. (1996). Appl. Phys. Lett. 68,
36 14-3616.
Grundmann, M., Christen, J., Ledentsov, N. N., Bohrer, J., Bimberg, D., Ruvimov, S. S.,
Werner, P., Richter, U., Gosele, U., Heydenreich, J., Ustinov, V. M., Egorov, A. Y.,
Zhukov, A. E., Ustinov, V. M., and Alferov, Z. I. (1995). Phys. Reu. Lett. 74, 4043-4046.
Grundmann, M., Ledentsov, N. N., Stier, 0..Bimberg, D., Ustinov, V. M., and Alferov, Z. I.
(1996). Appl. Phys. Lett. 68, 979-981.
Hatami, F., Ledentsov, N. N. Grundmann, M., Bohrer, J., Heinrichsdorff, F., Beer, M., Bimberg,
D., Ruvimov, S. S., Werner, P., Gosele. U., Heydenreich, J., Ivanov, S. V., Meltser, B. Y.,
Kop’ev, P. S., and Alferov, Z. I. (1995). Appl. Phys. Lett. 67, 656-658.
Hayashi, Y., Mukaihara, T., Hattori, N., Ohnoki, N., Matsutani, A,, Koyama, F., and Iga, K.
(1995). Electron. Lett. 31, 560-561.
Heinrichsdorf, F., Krost, A., Grundmann, M., Bimberg, D., Kosogov, A. O., and Werner, P.
(1996). Appl. Phys. Lett. 68, 3284-3286.
Heinrichsdorff,F., Mao, M. H., Kirstaedter, N., Krost, A,, Bimberg, D., Kosogov, A. O., and
Werner, P. (1997). Appl. Phys. Letr. 71, 22-24.
Heitz, R., Grundmann, M., Ledentsov, N. N.. Eckey, L., Veit, M., Bimberg, D., Ustinov, V. M.,
Egorov, A. Y., Zhukov, A. E.. Kop’ev, P. S., and Alferov, Z. I. (1996). Appl. Phys. Lett. 68,
361-363.
Hiraydma, H., Matsunaga, K., Asada, M., and Suematsu, Y. (1994). Electron. Lett. 30, 142-143.
Huffaker, D. L., Shin, J., and Depper, D. G. (1994). Electron. Lett. 30, 1946-1947.
Huffaker, D. L., Baklenov, O., Graham, L. A,, Streetman, B. G., and Depper. D. G. (1997a).
Appl. Phys. Lett. 70, 2356-2358.
Huffaker, D. L., and Deppe, D. G. (1997b). .4ppl. Phys. Lett. 71, 1449-1451.
Inoshita, T., and Sakaki, H. (1992). Phq’s. Ret:. B. 46, 7260-7263.
Kamath, K., Battacharya, P., Sosnowski, T., and Phillips, J. (1996). Electron. Lett. 30,
1374-1375.
Kamath, K., Phillips, J., Jiang, H., Singh, J.. and Bhattacharya, P. (1997a). Appl. Phys. Lett. 70,
2952-2953.
Kamath, K., Klotzkin, D., and Bhattacharya, P. (1997b). Proc. 10th Ann. Meer. LEOS (LEOS
’97), ThY4.
Kash, K., Scherer, A,, Worlock, J. M., Craighead, H. G., and Tarnargo, M. C. (1986). Appl.
Phys. Lett. 49, 1043-1045.
Kirstaedter, N., Ledentsov, N. N., Grundmann, M., Bimberg. D., Ustinov, V. M., Ruvimov,
S. S., Maximov, M. V., Kop’ev, P. S., Alferov, Z. I., Richter, U., Werner, P., Gosele, U., and
Heydenreich, J. (1994). Electron. Lett. 30, 1416-1417.
Kirstaedter, N., Schmid, 0. G., Ledentsov, N. N., Bimberg, D., Ustinov, V. M., Egorov, A. Y.,
Zhukov, A. E. Maximov, M. V., Kop’ev. P. S., Alferov, Z. I. (1996). Appl. Phys. Lett. 69,
1226-1228.
Ledentsov, N. N., Bohrer, H., Bimberg, D., Kochnev, I. V., Maximov, M. V., Kop’ev, P. S.,
Alferov, Z. I., Kosogov. A. O., Ruvimov. S. S., Werner, P., and Gosele, U. (1996). Appl.
Phys. Lett. 69, 1095-1097.
Leonard, D., Kishnamurthy, M., Reaves. C. M.. Denbars, S. P., and Petroff, P. M. (1993). Appl.
Phys. Lett. 63, 3203-3205.
Mao, M. H., Heinrichsdorff, F., Kirstaedter, N., Krost, A., Ledentsov, N. N., Bimberg, D.,
Ustinov, V. M., Egorov, A. Y., Zhukov, A. E., Kop’ev, P. S., and Alfreov, Z. I. (1997). Proc.
10th Ann. Meet. LEOS (LEOS ’97).ThY5.
6 SELF-ASSEMBLED
Mirin, R. P., Ibbetson, J. P., Nishi, K., Gossard, A. C., and Bowers, J. E. (1995). Appl. Phps.
Lett. 67, 3795-3191.
Mirin, R. P., Gossard, A,, and Bowers, J. E. (1996). Electron. Lett. 32, 1732-1733.
Moritz, A., Wirth, R., Hangleiter, A., Kurtenbach, A,, and Ebert, K. (1996). Appl. Phys. L e t f .
69, 212-214.
Mukai, K., Ohtsuka, N., Sugawara. M., and Yamazaki, S. (1994). Jpn. J . Appl. Phys. 33,
L1710-Ll712.
Mukai, K., Ohtsuka, N., and Sugawara, M. (1996a). Jpn. J . Appl. Phys. 35, L262-L265.
Mukai, K., Ohtsuka, N., Shoji, H., and Sugawara, M. (1996b). Appl. Phys. Lett. 68, 3013-3015.
Mukai, K., Ohtsuka, N., Shoji, H., and Sugawara, M. (1996~).Phys. Rev. B. 54, R5243-R5246.
Mukai, K.. Shoji, H., and Sugawara. M. (1998a). Proc. lO!h In!. Conj: Indium Phosphide and
Related Muteriuls ( I P R M '98), pap. WBI-4.
Mukai, K. (1988b) (unpublished).
Nakata, Y., Sugiyama, Y., Futatsugi, T., Yokoyama, N. (1997). J . Cryst. Growth 175/176,
71 3-719.
Nishi, K., Yamada, M., Anan, T., Gomyo. A.. and Sugou, S. (1998). Ext. Ahs. Jupunese Sociery
of Applied Physics, 3 (in Japanese).
Notzel, R., Temmyo, J., Kamada, H., Furuta, T., and Tamamura, T. (1994). Appl. Phys. Lett.
65, 457-459.
Ohtsuka, N., and Mukai, K. (1995). Proc. 7th Int. Conj: Indium Phosphide und Related
Muterials ( I P R M '95), WP59.
Oshinowo, J., Nishioka, M., Ishida, S., and Arakawa, Y. (1994). Appl. Phys. Lett. 65, 1421-1423.
Pan, D., Zeng, Y. P., Wu, J., Wang, H. M., Chang, C. H., Li, J. M., and Kong, M. Y. (1997).
Appl. Phys. Lett. 70, 2440-2442.
Saito, H., Nishi, K., Ogura, I., Sugou, S.. and Sugimoto, Y. (1996). Appl. Phys. Lett. 69,
3 140-3 142.
Saito, H., Nishi, K., Sugou, S., and Sugimoto, Y. (1997). Appi. Phjs. Lett. 71, 590-592.
Schmidt, 0. G., Kirstaedter, N., Ledentsov, N. N., Mao, M. H., Bimberg, D., Ustinov. V. M..
Egorov, A. Y., Zhukov, A. E., Maximov, M. V., Kop'ev, P. S., and Alferov, Z. I. (1996).
Electron. Lett. 32, 1302-1 304.
Shoji, H., Mukai, K., Ohtsuka, N.. Sugawara, M., Uchida, T., and Ishikawa, H. (1995). IEEE
Photon. Technol. Lett. 7, 1385- 1387.
Shoji, H., Nakata, Y., Mukai, K., Sugiyama, Y., Sugawara. M., Yokoyama, N., and Ishikawa.
H . (1996a). Jpn. J. Appl. Phys. 35, L903-L905.
Shoji, H., Nakata, Y., Mukai, K., Sugiyama, Y., Sugawara, M., Yokoyama. N.. and Ishikawa,
H. (1996b). Electron. Lett. 32, 2023-2024.
Shoji, H., Nakata, Y., Mukai, K., Sugiyama, Y., Sugawara, M., Yokoyama, N., and Ishikawa,
H. (1997a). IEEE J. Selected Topics in Quantuni Electron. 3, 188-195.
Shoji, H., Nakata, Y., Mukai, K., Sugiyama. Y., Sugawara, M., Yokoyama. N., and Ishikawa,
H. (1997b). Appl. Phys. Lett. 71, 193-195.
Shoji, H., Nakata, Y., Mukai, K., Sugiyama, Y., Sugawara, M., Yokoyama, N., and Ishikawa,
H. (1997~).Proc. 2nd ConJ Optoelutronics rind Communicution (OECC' '97). 9C2-5.
Shoji, H.. Nakata, Y., Mukai, K., Sugiyama, Y., Sugawara, M., Yokoyama. N., and Ishikawa,
H. (1997d). Proc. 10th Ann. MCCI.L E O S ( L E O S '97), pap. TbY2.
Soda, H., Iga, K., Kitamura, C., and Suematsu, Y. (1979). Jprz. J . Appl. Phys. 18, 2329-2330.
Solomon, G. S., Trezza, J. A., Marshall, A. F., and Harris, J. S. (1996). Phys. Rer. Lett. 76,
952-955.
Solomon, G. S., Larson, M. C., and Harris, J. S. (1996). Appl. Phys. Lett. 69, 1897-1899.
Sugawara, M., Nakata, Y., Mukai, K., and Shoji, H. (1997a). Phys. Rev. B. 55, 13155-13160.
Sugawara, M., Mukai, K., and Shoji, H. (1997b). Appl. Phys. Lett. 71, 2791-2793.
286 HAJIMESHOJI
Sugiyama, Y., Nakata, Y., Imamura, K.. Muto, S., and Yokoyama, N. (1996). Jpn. J . Appl. Phys.
35, 1320-1324.
Sugiyama, Y., Nakata, Y., Futatsugi, T., Sugawara, M.. Awano, Y., and Yokoyama, N. (1997).
Jpn. J . A p p l . Phys. 36, L158-LI61.
Tackeuchi, A., Nakata, Y., Muto, S.. Sugiyama, Y., Inata, T., and Yokoyama, N. (1995). Jpn.
J . Appl. Phys. 34, L405-L407.
Temmyo, J. Kuramochi, E., Sugo, M., Nishiya, T., Kamada, H., Notzel, R., and Tamamura, T.
(1994). Proc. 14th Int. Semiconductor Luser Cot$ (ISLC '94). PD3.
Uskov, A. Y., McInerney, J., Adler, F., Schweizer. H., and Pilkuhn, M. H. (1998). Appl. Phys.
Leu. 72, 58-60.
Ustinov, V. M., Egorov, A. Y., Kovsh. A. R.. Zhukov, A. E., Maximov, M. V., Tsatsulnikov,
A. F., Gordeev, N. Y., Zaitsev, S. V.,Shernyakov, Y. M., Bert, N. A,, Kop'ev, P. S., Alferov,
Z. I., Ledentsov, N. N., Bohrer, J., Bimberg, D., Kosogov, A. O., Werner, P., and Gosele,
U. (1996). Proc. 9th Int. Cortf: MBE.
Xie, Q., Kalburger, A., Chen, A., and Madhukar, A. (1997). J . Cryst. Growth, 175/176,689-695.
SEMICONDUCTORS AND SEMIMETALS . VOL. 60
CHAPTER 7
Applications of Quantum Dot to Optical Devices

Hiroshi Ishikawa
DFVICES
ELECTRON A N D MATERIALS
LARS
FUJlTSU LABORATORIES LTU
. .
JAPAN
ATSUGI KANAC~AWA
1. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
11. PROPERTIES OF QUANTUM DOTS. . . . . . . . . . . . . . . . . . . . 288
1 . The Quantum Dot us a T M V - L ~ WS.ystern
I . . . . . . . . . . . . . . . 288
2 . Attractive Features of Quantum Dots.for Device Application . . . . . . . 294
111. QUANTUM DOTSFOR VERYHIGHSPEEDLIGHT MODULATION . . . . . . . . 295
1. The Need.for High-speed. Low- Wavelength-Chirp Light Sources . . . . . 295
2 . Direct Modulation of Quantuni-Dot Lasers . . . . . . . . . . . . . . 298
3 . The Quuntuin-Dot Intensity Modulator . . . . . . . . . . . . . . . . 302
IV . QUANTUM DOTSAS A NONLINEAR MEDIUM. . . . . . . . . . . . . . . . 303
1. The Needfor Large Nonlinearity with u Large Bandwidth . . . . . . . . 303
2. Ana!ysis of zl3’ . . . . . . . . . . . . . . . . . . . . . . . . . . 306
3 . Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
V. PERSISTENT HOLE BURNINGMEMORY. . . . . . . . . . . . . . . . . . 314
1 . Persistent Spectrul Hole Burning Memory Using Quantum Dots . . . . . 314
2 . E.xperiinentu1 Results . . . . . . . . . . . . . . . . . . . . . . . . 316
3. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
VI . SUMMARY A N D PERSPECTIVES ON QUANTUM-DOT OPTICAL DEVICES . . . . . 319
1. Trends in Optoelectronics . . . . . . . . . . . . . . . . . . . . . . 320
2 . Usesfor Quantum Dot Opticul Devices . . . . . . . . . . . . . . . . 321
Acknowledyment . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
I . Introduction
The past several years have seen remarkable progress in the fabrication of
quantum dots and quantum-dot lasers. Thanks to new self-organized
growth technologies for example. as described in Chapter 6. ground-state
CW lasing can now be achieved at room temperature (Kirstaedter et al.,
1994; Shoji et al., 1996). Although the performance of quantum-dot lasers is
as yet far from ideal. recent gains have enlarged our expectation of applying
287
Copyright ( 1999 by Academic Press
All rights of reproduction in any form recerved
ISBN 0-1?-752169-0
ISSN 0080-8784 99 $3000
288 HIROSHI
ISHIKAWA
quantum dots to various optical devices. One of the reasons that we are so
interested in quantum dots is that the performance and functions of optical
devices based on current technologies, such as the quantum well or the
strained quantum well, have reached their limit. In this respect, quantum
dots are highly attractive because of their unique three-dimensional carrier
confinement which may provide us new optical properties. At present,
application research is concentrating mainly on quantum-dot laser, and
even here there is much to be done. We do not know what is ultimately
achievable with quantum-dot lasers because our understanding of the
carrier dynamics in quantum dots is incomplete. Nevertheless, we can start
experimental and theoretical studies to explore the advantages of quantum
dots for device applications.
In this chapter, I first review some of the attractive properties of quantum
dots and then discuss several feasible applications. In Section 11, I look at
the quantum dot as a two-level system, and summarize some of its
advantages. In Section 111, I discuss the use of quantum dots for high-speed
modulation of light for long-distance high-bit-rate communication systems.
I review the problems with conventional technologies and then discuss the
direct modulation of quantum-dot lasers and the feasibility of an external
modulator using quantum dots. In Section IV, I discuss the use of quantum
dots as a nonlinear medium. Realization of large nonlinear susceptibility
using quantum structures has long been a dream. To see if such a dream is
feasible, I perform a trial analysis of the third-order nonlinear susceptibility
of quantum dots. In Section V, I discuss the use of quantum dots for
high-density optical memories using persistent spectral hole burning. In
Section VI, I summarize and discuss the roles of quantum-dot-based devices
in the future of optoelectronics.
Because of our so-far limited knowledge of the quantum dots, my
discussions in this chapter are not concrete. I simplify my analysis to show
the feasibility of quantum dots. From an engineering point of view, such an
approach is sometimes more efficient that waiting for the establishment of a
full-blown and rigid theory and complete experimental evidence.
11. Properties of Quantum Dots
1. THE QUANTUM
DOTAS A TWO-LEVEL
SYSTEM
What makes quantum dots attractive is that their carriers are strongly
confined three-dimensionally in a very narrow region; this gives discrete
energy levels. There are also many unresolved problems with quantum dots:
7 APPLICATIONS
OF QUANTUM DEVICES
DOTTO OPTICAL 289
how the carriers relax to the ground state (Bockelmann and Bastard, 1990;
Bockelmann, 1993; Efros et al., 1995), the dephasing time of the ground-state
wave function, the effect of the Coulomb interaction on the emission spectra
(Hu et al., 1990), and so on. However, here I put aside such problems and
focus on some of the fundamental properties of quantum dot, assuming it
to be a simple discrete two-level system. Most of these properties can be
illustrated by this assumption.
Linear Gain
Following a conventional density-matrix analysis, linear gain can be

calculated from the linear susceptibility, fl), of the two-level system as
where c is the light velocity; y~is the refractive index; R is the crystal volume;
n is the dot number; e is the electron charge; E, is the vacuum dielectric
constant; Y : and 4':' are the wave function (including the Bloch part) of the
ground-state electron and hole, respectively; f;. and f , are the Fermi-Dirac
distribution function; EL and E:, are the ground-state energy of the electron
and hole, respectively; and y,, is the inverse of the dephasing time. The suffix
i denotes the ith dot, and the summation is for all dots in a volume, R. The
square of the matrix element in Eq. (7.1) can be expressed by a matrix
element for bulk (Kane, 1957) as
where, $: and Ic/i are the envelope wave functions of the ground state; m, is
the electron mass, mr6 is the electron effective mass at r6;E , is the band
gap of the quantum-dot material; and A is the spin-orbit splitting energy.
From Eq. (7.1) we see that, at resonance, where Ek - El = ho,the linear
gain is inversely proportional to hy,,, which is a broadening determined by
the dephasing time. We obtain a large gain for a small broadening of Wy,,,,
which is highly attractive for semiconductor lasers because a high gain
results in a very low threshold current.
What is interesting about the linear gain is the refractive index dispersion.
Figure 7.l(a) shows schematically the refractive index dispersion of a
quantum well. Refractive index, 17, is governed by dispersions due to various
oscillators (Adachi, 1982). Among them, the carrier-density-dependent re-
290 HIROSHIISHIKAWA
High carrier
density
Gain,
change
.. ..
,* * .
Quantum well Quantum dot

(a) (b)
FIG. 7.1. Schematic of gain (dotted line) and the refractive index (solid line) spectra for (a)
a quantum well, and (b) a quantum dot.
fractive index change results from the change in the optical gain spectrum
caused by band filling, band-gap renormalization, and the effect of plasma
resonance (see, for example, Bennett et al., 1990). Any change in the gain
spectrum changes the refractive index through the Kramers-Kronig relation.
In a typical quantum well with a high-density carrier injection, an increase
in carrier density causes a reduction in the refractive index a t the wavelength
of gain region. This reduction, Aq, amounts to -4 x 10-20n, where n is
the carrier density per cubic centimeter (Lee et al., 1986). Thus, a change of
5 x 10'7cm-3 in carrier density gives a refractive index change of about
0.02.
There is no plasma dispersion associated with carriers in the quantum
dots. Moreover, the gain is symmetric around the resonant wavelength.
Figure 4.l(b) shows schematically the gain and refractive index dispersion
of a quantum dot. If the shift of the ground-state transition wavelength for
a change in carrier number is very small, we obtain an almost zero index
change for a change in carrier number at the resonant wavelength. In the
actual case, a red shift of the resonant wavelength may take place when a
carrier is added to the ground state to form a bi-exciton (Hu et al., 1990),
and the addition of carriers to upper sublevels may also cause some change
in the resonant wavelength through the Coulomb interaction. These shifts
are yet to be analyzed quantitatively but they are presumed to be within a
few meV. In actual quantum dots there is an inhomogeneous broadening,
to be described later, which cancels out the change in refractive index near
the gain peak. If we can reduce the plasma dispersion effect of carriers at the
7 APPLICATIONS DOTTO OPTICAL
OF QUANTUM DEVICES 291
barrier layer and the wetting layer, we might obtain an almost zero
refractive index change at the gain peak wavelength for the change in carrier
number in the quantum dot. As will be discussed in Section 111, there is a
possibility of obtaining no wavelength chirp under high-speed modulation
in quantum-dot lasers.
Third Order Nonlinear Susceptibility
Because the structure of unbiased quantum dots is an inversion symmetry,

we can expect them to have odd-order nonlinearity. The third-order
nonlinear susceptibilities of the two-level system can be calculated using the
density-matrix analysis as
f3'=
4e4
--z
E& =
<
I y:l Iy I)I4( L - .t.)
(hw,- EL + EL $- ihycC)(hwp- ho,+ ihy,,)(hw - Ei
J"
+ E:;- ihy,,)
where we assume a four-wave mixing process and only the resonant terms
are shown. The x’~’is doubled taking account of the spin. In Eq. (7.2), w,,
ws, and w, are angular frequencies of the pump wave, signal wave, and
conjugate wave, respectively. ycc and y,,,.are the dephasing rate of each level.
In the degenerate case, that is, Q,, = oiS,this reduces to
f3'(wp= w s )=
We can expect a large f3) for the smaller dephasing rate y,,,, as in the case
292 HIROSHIISHIKAWA
of linear gain. Of course, this also holds for the nondegenerate case. The
large x'j) encourages us to use quantum dots as a nonlinear medium for
various device applications. The response speed, or a bandwidth of ~ ( j ) ,
depends on the dephasing rate yes. The bandwidth reduces for small y,,, as
can be read from Eq. (7.4).
Rate Equation for Photons
We can derive the rate equation for photons using x"’ and f 3 ' under the
degenerate on-resonant case, that is, w = cop = o,= (Ef - EI)/A. We assume
a single polarization optical field with an angular frequency of w. We start
from the equation for the optical electric field E (Kuwatsuka et al., 1997) as
where r is the optical confinement factor to the dot region, p O is the

magnetic permeability and factor 3/4 is the degeneracy factor of f3).
Substituting dz to be the product of the group velocity ug and d t , transform-
ing electric field to photon density, s, and assuming that all dots have the
same size and composition (then omitting the suffix i), Eq. (7.5) reduces to,
dS S
-= t.,T(G - ES)NS - -
dt 'ph
where, the term of photon lifetime z,, is added assuming a laser cavity and
i n
(7.7)
(7.9)
N represents the inversion carrier density in the ground state of quantum

dots, G is the linear gain, and 2 is the gain compression factor, which
corresponds to the one often used to express the gain saturation in
semiconductor lasers (Nagarajian et al., 1992). The gain compression factor
is larger for small y,,. The gain saturation takes place at high optical power
for a large E.
7 APPLICATIONS DEVICES
DOTTO OPTICAL
OF QUANTUM 293
Mod$cation for Actual Quantum Dots
Inhomogeneous broadening in quantum dots is due to the nonuniformity

of dot size and composition. To take account of this inhomogeneous
broadening, we may carry out the summation over the index, i; giving a
different size and composition for each dot in Eqs. (7.1) and (7.3). Here, to
show the effect of this fluctuation, we assume that the inhomogeneity gives
the Gaussian distribution of the ground-state transition-angular frequency,
w,,, whereas other parameters are the same. We also assume a Gaussian
distribution as
(7.10)
where 2Aw is the eC1 width of the inhomogeneous broadening. The

corresponding full width at half-maximum (FWHM) is 1 . 6 7 8 ~Quantum
.
dots available with present technology give a broadening of 25-80 meV
(Mukai et al., 1996a and b; Nakata et al., 1997), which is quite large when
compared with the homogeneous broadening try,,, which may be on the
order of a few meV or below. In this case, we may presume the homogene-
ous Lorenzian broadening function in Eq. (7.1) to be a 6 function. Thus, the
peak gain is
(7.11)
The linear gain is inversely proportional to the inhomogeneous broadening.

To obtain large linear optical gain, or a large f 3 ) , we must reduce the
broadening.
In Eqs. (7.1) and (7.2), we did not take into account the carrier energy
relaxation process within the dots, which is not well understood quantitat-
ively. If the relaxation process can be properly incorporated, making the
systems a three-level one, these equations can be modified. Recent experi-
mental and theoretical studies reveal that the carrier relaxation to the
ground state is 10- 100 ps, which is fairly slow when compared with that of
the quantum well (Mukai et al., 1995; Mukai et al., 1996b; Uskov et al.,
1997; Uskov et al., 1998). The slow carrier energy relaxation time gives an
additional term in f 3 ) to increase the gain compression factor. This results
in the gain saturation in the quantum-dot lasers at high optical output
power.
294 HIROSHI ISHIKAWA
2. ATTRACTIVEFEATURES
OF QUANTUM
DOTSFOR DEVICE
APPLICATION
On the basis of their basic properties, and from the two-level model just
described, the attractive features of quantum dots for device application can
be enumerated as follows:
1. Their discrete energy levels may open up new applications for quantum
dots. In the quantum well, light with an energy greater than the band-
edge energy is absorbed. In quantum dots, there are transparent
regions between the discrete levels when the inhomogeneous broaden-
ing is not large. In Section 111, we discuss an optical intensity
modulator that makes use of this feature.
2. There is a possibility of obtaining a very small refractive index change
at the resonant wavelength for the change in carrier number in the
quantum dot. This will give a very small wavelength chirp under
high-speed direct modulation in quantum-dot lasers, as will be dis-
cussed in Section 111.
3. Discrete energy levels also affect carrier dynamics. When the LO-
phonon energy does not coincide with the level separation, carrier
energy relaxation is slow, causing the so-called phonon bottleneck
(Bockelmann and Bastard, 1990). However, the slow rate of carrier
energy relaxation means the slow excitation rate of carriers from the
dot. This will be advantageous in an infrared photodetector using the
subband transition because it gives a small dark current. Readers
interested in the quantum-dot infrared photodetector should refer to
Ryzhii (1996).
4. Large x'l) and f 3 ) will be obtained in quantum dots if we can reduce
the inhomogeneous broadening. A large x(l) gives a large linear gain,
which means that we can expect very low threshold current lasers. A
large x(3) will enable us to realize optical nonlinear devices. The
response speed of fl) and x’~’depends largely on the dephasing rate,
y,,, which is determined by the carrier dynamics. The quantum dots'
unique carrier dynamics may give a different dephasing rate from that
of conventional quantum wells. Section IV discusses x(3’ together with
carrier dynamics in optical nonlinear devices.
5. The surface density of a quantum dot grown in the typical SK mode
is about 10"/cm2. If we have both large dot density and large non-
uniformity, leading to many different transition energy levels (equival-
ent to large inhomogeneous broadening), we may be able to use this
as a huge-capacity optical memory. The application of quantum dots
in Tb/cm2 memory is discussed in Section VI.
111. Quantum Dots for Very High Speed Light Modulation
1. THENEEDFOR HIGH-SPEED,
LOW-WAVELENGTH-CHIRP
LIGHTSOURCES
Very high speed light modulation for optical communication systems

has been one of the major targets of semiconductor laser research. For
example the frequency response of Fabry-Perot lasers has reached to
25GHz (Bowers et al., 1986), and distributed feedback (DFB) lasers can
perform single-spectrum lasing for long-haul transmission systems (see, for
example, Ishikawa et al., 1987). However, direct modulation in long-haul
optical communication systems is limited to 2.5 Gb/s, not because of the
limited modulation bandwidth but because of the wavelength chirping
under direct modulation. The chirping of a single-mode DFB laser limits the
transmission distance in high-bit-rate systems because of the chromatic
dispersion of optical fiber. The chromatic dispersion is typically - 17 ps/km/
nm in nondispersion shifted fiber at the wavelength of 1.55pm, where the
fiber loss is the smallest. The goal is to realize high-speed modulation of
laser light with the smallest, or at least a controlled, wavelength chirp.
Here, I first review the problems in direct modulation of semiconductor
lasers and those in the external modulation scheme based on present
technologies. I then discuss two possibilities; the chirpless high-speed direct
modulation of quantum-dot lasers and the use of quantum dots as an
external optical intensity modulator with controlled wavelength chirping.
The Direct Modulation Scheme
Wavelength chirping under direct modulation of semiconductor lasers is

due to a change in the refractive index caused by a change in carrier
concentration. It can be evaluated by (Koch and Linke, 1986; Wiesenfeld et
al., 1987)
(7.12)
where P is the light power, r is the optical confinement factor, 0 is the gain
compression factor (as discussed in Eq. (7.6)), V,,, is the active layer volume,
yld is the differential quantum efficiency, and hv is the photon energy. The
parameter u is (Henry, 1982)
(7.13)
296 HIROSHIISHIKAWA
which is proportional to the ratio of the refractive index change to the gain
change for the change in carrier concentration N . The first term in Eq. (7.12)
gives the change related to the time change of the optical power originating
from the phase modulation of light. The second term gives the offset
wavelength shift, which depends on the optical power. The value of tl in
multiple quantum well (MQW) lasers is around 3 to 4 (Green et al., 1987).
In semiconductor lasers under very fast modulation, the second term is
mostly responsible for the chirp (Wiesenfeld et al., 1987). Its coefficient
depends largely on the device structure. Wavelength chirping is often
evaluated by the time-averaged spectral width. In quantum-well DFB lasers,
a broadened spectral width due to chirping of 0.28 nm (-20 dB full
width) was reported under 2 Gb/s NRZ (non-return zero) modulation
(Kakimoto et al., 1990), and one of 0.34nm was reported under 10Gb/s
NRZ modulation (Uomi et al., 1991). The chirp at 10Gb/s modulation is
large enough to limit the transmission distance in 10Gb/s systems. For a
fiber dispersion of - 17 ps/km/nm, 0.34-nm chirp causes a pulse broadening
of 578ps for a 100-km transmission. This value exceeds the time slot of
100ps for 10 Gb/s NRZ modulation.
Review of the External Modulation Scheme
External modulation schemes have been studied extensively as a way to

overcome the chirping problem. For example, the LiNbO, Mach-Zender-
type modulator gives, in principle, zero-frequency chirp (Seino, 1996). It is
used for lO-Gb/s long-distance systems. Semiconductor intensity modula-
tors also are attractive because of their small size and easy monolithic
integration with a DFB laser. They use either the Franz-Keldysh effect in
the bulk absorption layer (Soda et al., 1988) or a Stark shift of the exciton
peak wavelength in the MQW (Devaux et al., 1992) under an electric field
application. There is still a wavelength chirp, due to phase modulation in
the intensity modulator, however (Koyama and Iga, 1985), which is charac-
terized by the tl value defined as
(7.14)
where the derivatives are by the applied voltage, I/: The wavelength chirp
for the external modulation is given by substituting this a value in Eq. (7.12)
and eliminating the second term, which is specific to semiconductor lasers.
Figure 7.2(a) illustrates how the tl parameter is determined in the MQW
modulator. As shown in the figure, application of the electric field causes a
A
Exciton peak A
Signal
a. +a.
lv&o>a.]+o v = v, v=o v = v,
a.
v=o
(a) Normal operation (]I >0 (b) Prebiased operation aC 0

FIG. 7.2. Parameter in a quantum well under bias-voltage application. In case (a) the usual
operating condition of an MQW modulator i( parameter, defined as i(= -2k(?q/dI/)/dg/dI/;
is positive. Case (b) is the prebiased operating condition. We obtain a negative r value.
shift of the exciton absorption peak. In association with this peak shift, the
refractive index dispersion also shifts, as shown. The refractive index change
is positive, and the gain change is negative; thus, the CI parameter is positive.
This refractive index change gives the blue chirp at the turn-on and the red
chirp at the turn-off, as can be read from Eq. (7.12). The blue-chirped
leading edge of the light pulse is faster than the red-chirped pulse end
because of a fiber dispersion. Therefore, this chirping widens the pulse width
when transmitted through an optical fiber.
In the MQW modulator, however, the CI value is typically 0.2- 1.4
(Kataoka et al., 1994), which is smaller than in the MQW laser. In addition,
there is no second term of Eq. (7.12). Therefore, the wavelength chirp in
external modulators is smaller than that of the direct modulation of
semiconductor lasers. At lO-Gb/s NRZ modulation, a wavelength chirp of
0.15 nm (-20 dB full width) was reported, but this limits the transmission
distnce to about 25 km (Haisch et al., 1994). If we can get a negative CI
parameter, we can get pulse compression in the transmission system. When
the chirp is controlled to give a narrow pulse width for a given transmission
298 HIROSHIISHIKAWA
distance, the signal-to-noise ratio improves. The negative a parameter can

be realized in the MQW modulator by prebiasing the modulator or by
setting the signal wavelength close to the exciton peak, as shown in Fig.
7.2(b). Recently this was done in the modulator-integrated DFB laser, and
transmission over 100 km at 10 Gb/s was achieved without any penalty in
the error rate (Morito et al., 1996). There is a disadvantage to this method,
however. Because of the prebiasing, there is an absorption at light-on state,
which causes the extinction ratio to deteriorate when compared with the
unbiased operation. It is very difficult to get an extinction ratio larger than
15 dB. If we could make a modulator with a zero or negative a value without
such loss, the performance of the semiconductor-based intensity modulator
for high-bit-rate long-haul transmission systems would be much improved.
2. DIRECT
MODULATION
OF QUANTUM-DOT
LASERS
Benefits of Quuntum-Dot Lasers
The most attractive feature of quantum-dot lasers is that they make

possible a very small refractive index change at the resonant wavelength, as
explained in Section 11. This is something that conventional MQW lasers
cannot do. If we could ignore the plasma effect of the carriers in the barrier
layer or in the wetting layer, we might obtain an almost zero a value, which
would lead to no wavelength chirp under high-speed direct modulation. The
problem is how fast we can modulate quantum-dot lasers, which relates to
the yet unresolved problem of carrier dynamics in quantum dots- the
phonon bottleneck (Bockelmann and Bastard, 1990).
Modulation Response and Carrier Dynamics
The modulation response of the quantum-dot lasers can be understood

through the well-developed modulation theory of quantum-well lasers
(Tucker, 1985; Koch and Linke, 1986; Olshanski et al., 1987; Nagarajian,
1992). The intrinsic maximum modulation bandwidth, which is the relax-
ation oscillation frequency, can be given by
1
(7.15)
where z p is the photon lifetime, E is the gain compression factor, vg is the

7 APPLICATIONS
OF QUANTUM
DOTTO OPTICAL
DEVICES 299
group velocity of light, and x represents the ratio of carriers in the upper
subband and the barrier to those in the ground state (Nagarajian et al.,
1992; Ishikawa and Suemune, 1994). At first sight, one benefit of the
quantum-dot laser is that the differential gain, G’ can be made very large if
we can reduce the inhomogeneous broadening, because G’ is inversely
proportional to the inhomogeneous broadening width. This gives a high
f,,,. However, factors specific to quantum-dot lasers prevent a large fmax:
factor and the gain compression factor E. Recent experimental and
theoretical studies suggest a large carrier relaxation time to the ground state
in quantum dots of 10-1OOps (Mukai et al., 1995; Mukai et al., 1996b;
Uskov et al., 1997; Uskov et al., 1998). This is much slower than that in the
MQW laser, which is around 10- 100 fs (Seki and Yokoyama, 1994). As the
factor x is the ratio of total carriers to those in the ground state, it becomes
large because of the slow energy relaxation. This slow energy relaxation also
causes a large value of E resulting in gain saturation at high power levels.
There is a possibility that the product of Zx will be a large value which would
cancel out the large differential gain.
There also is a frequency rolloff associated with carrier transport
(Nagarajian, 1992). In SK-mode quantum dots, carriers are injected into the
barrier layer and then diffuse to a quantum dot and relax to the ground
state. These carrier transport time constants give the rolloff in modulation
response, S(o,), given by
(7.16)
where omis the modulation angular frequency, zd is the diffusion time, and
T ~ is the
, carrier
~ ~ energy
~ relaxation time. If we assume a diffusion distance
of 50 nm and that a diffusion constant of 7 cm2/s, zd is 1.8 ps, and if we
assume a Zrefax of 20 ps, Eq. (7.16) gives a rolloff of -2.1 dB at a modulation
frequency of 10GHz. This roleoff is mainly due to the energy relaxation
time.
Experimental Reports
Kamath et al. (1997) prepared two types of In,,,Ga,.,As/GaAs quantum-

dot laser. One type is the laser with GaAs SCH layers. They made lasers
with different stack of dot layer, The other is a tunnel-injection type in which
an Al,.,Ga,.,As thin tunneling barrier is placed between the dot layer and
the SCH layer at the n-side, carrier injection takes place by tunneling direct
to the dot layer. Kamath and his colleagues examined the optical gain and
300 HIROSHIISHIKAWA
the small signal modulation response in both types. In the SCH type, the
optical gain measured by the Hakki-Paoli method in the 4-layer stacked
quantum-dot laser was 2.5 x 10-'4cmz, which was larger than in the
single-layer laser. Nevertheless, there was no significant increase in the
modulation bandwidth, which was limited to 6.2 GHz. The tunnel injection
laser showed a larger differential gain of 6 x 10-'4cm2, however, the
bandwidth was 8.5 GHz. Also, the increase in bandwidth did not correspond
to the increase in differential gain. The researchers concluded that the
bandwidth was limited by the carrier relaxation, or the phonon bottleneck.
This result fits with our discussion that slow carrier relaxation gives a large
value of E x to limit the modulation bandwidth.
A similar result was reported by Mao et al. (1997). They measured the
relaxation oscillation frequency of InGaAs/GaAs self-organized quantum-
dot lasers, observing a saturation of the relaxation oscillation frequency
despite an increase in optical power. The relaxation oscillation frequency
saturated at 5.3 GHz at the optical power of 9.1 mW, which also suggests a
large value for the gain compression factor, 2, and the existence of the
phonon bottleneck.
Wang et al. (1997) reported a comparison of wavelength chirping in
quantum-well, quantum-wire, and quantum-dot lasers under an ultra-short
optical pulse excitation. They fabricated InGaAs/InGaAsP DFB-type lasers
emitting at 1.4pm using the etching and regrowth technology. Their
quantum-dot diameter was 60 nm, which is a little too-large for sufficient
three-dimensional carrier confinement. The researchers reported that the
wavelength chirp was smaller in the higher-dimensional carrier confinement,
as shown in Fig. 7.3. In quantum-well and quantum-wire lasers, lasing under
pulsed excitation starts at a carrier density higher than the threshold level.
Then the lasing starts at a shorter wavelength. As a depletion of carriers
refractive index increases to give longer wavelength lasing, that is, the red
chirp takes place. What is interesting is that Wang et al. reported almost
zero-frequency chirp in the quantum-dot laser except for a slight blue chirp
at the onset. Their explanation was that the chirping behavior in the
quantum-dot laser is determined predominantly by the carriers in the barrier
layer because of a very small optical confinement to the dot layer of 0.5%.
The carrier relaxation process, which is due mostly to the Auger scattering of
carriers, generates high-energy carriers in the barrier layer. This causes
carrier heating, and which might cause a chirping different from that in the
quantum-well or the quantum-wire laser. Wang et al.'s interpretation did not
refer to the c( parameter of the quantum dot, and it considered the effect of
carriers in the barrier layer to be the cause of anomalous chirp behavior. This
finding is somewhat different from what I have discussed in this section.
However, their result is encouraging in that almost zero chirp was obtained.
Time delay (ps)
FIG. 7.3. Wavelength chirp of quantum-well, quantum-wire, and quantum-dot lasers. Three
solid lines represent lasing optical output under optical pulse excitation. The dotted lines
represent the change in the lasing wavelength (Wang et al., 1997, 0 1997, Optical Society of
America).
Summary
Reported experimental results suggest the existence of the phonon bottle-

neck, which limits modulation bandwidth through the @ term in Eq. (7.15)
and through the rolloff given by Eq. (7.16). This is the source of the
unresolved problems in high-speed modulation of quantum-dot lasers. To
realize the potential of the quantum-dot laser as a high-speed light source,
it is necessary first of all to reduce the inhomogeneous broadening to obtain
a large differential gain. Also, it is extremely important to clarify the carrier
dynamics. We must know the values of x and 2, and we must explore the
design of quantum dots to obtain a small value of the product, 2 ~ High . dot
density would help reduce the distance between dots to obtain a small
diffusion time and a large differential gain. Quantum-dot lasers are worth
the efforts to solve these problems because of the potential for zero-chirp
high-speed modulation and their potential for very low threshold current
operation.
302 HIROSHIISHIKAWA
3. THEQUANTUM-DOT
INTENSITY
MODULATOR
Use of the quantum dots as an intensity modulator was proposed by

Sahara et al. (1996) to realize a modulator with a negative a value. In
quantum dots grown in the SK growth, we can obtain a Stark shift by
applying an electric field parallel to the growth direction. Shown in Fig. 7.4
are the absorption spectra of the quantum dot assumed by Sahara. He
assumed a Gaussian-shaped spectrum with a broadening of 34 meV in C 2
full width. This broadening is due to inhomogeneity in the size and
composition of the dots. Between two absorption peaks, is a region with a
small absorption coefficient due to the discrete energy levels. Adjusting the
signal wavelength to the maximum of the absorption peak without applica-
tion of the field (1.55 pm in Fig. 7.4), we can obtain transmission when the
electric field is applied. In this case, as the transmission wavelength is on the
shorter wavelength side of the absorption peak, the change in the refractive
index gives the negative a value. We can obtain the wavelength chirp to
compress the optical pulse in the fiber transmission.
Figure 7.5 shows the calculated ci parameter and the change in the
absorption coefficient as functions of the applied electric field. In the
calculation, the Stark shift caused by the application of the electric field was
assumed to be the same as that of the quantum well. The a parameter is
negative for a high electric field of 130 to 80 kV/cm, and it shifts to positive
at lower fields. In the operation of this modulator, we apply the electric field
to the modulator for light-on state and reduce the field for the light-off
state. When we turn off the light, electric field is reduced from the high field
to the lower field. At this high field region CI is negative and give the blue
1.45 1-50 1.55 1.60 1.65 1.70

Wavelength (pm)
FIG. 7.4. Assumed absorption spectra of quantum dots. An inhomogeneous broadening of
34 meV (e-’ full width) is assumed. The dotted line represents the absence of the electric field
(Sahara et al., 1996, 5Q1996 I E E E ) .
c
A
1280-
s
c
0
960-
O IlO 1w t30 th 4b io 0
Fleld (kVlcm)
FIG. 7.5. Calculated absorption coefficient and the a parameter, after Sahara et al.. 1996, 8
1996 I E E E .
chirping to compress the light pulse. The positive a parameter at low field
region is not at all a problem, because the light is almost absorbed by
quantum dots under the low field. It was deduced that for a 100-pm-long
device with an optical confinement factor of 0.25, we could obtain trans-
mission changes from 0.57 to 0.017, corresponding to a 15-dB extinction
ratio.
The analysis by Sahara et al. is a preliminary one. The researchers
assumed a Stark shift the same as that of the quantum well, but in actual
quantum dots a Coulomb interaction forms the exciton or bi-exciton, which
might have some effect on the Stark shift. The assumed inhomogeneous
broadening of 34meV is based on the SK-growth mode. The progress of
growth technology will largely reduce it, and a narrower broadening will
largely reduce the absorption loss between discrete levels to give a much
larger modulation extinction. At present, practical modulators that make
use of the quantum dot are not a realized application, given the problems
to be solved. However, they are worth more detailed investigation because
of the need for controlled-chirp high-speed modulation for longer-span
transmission systems.
IV. Quantum Dots as a Nonlinear Medium
1. THENEEDFOR LARGENONLINEARITY
WITH A LARGE
BANDWIDTH
It has long been a dream to achieve large optical nonlinearity with large
bandwidth using semiconductor structures for optical nonlinear devices.
There are many applications waiting for this dream to become a reality:
304 HIROSHIISHIKAWA
high-speed all-optical spatial and time domain switches, wavelength conver-

ters for WDM (wavelength division multiplexing) systems, conjugate wave
generation for dispersion compensation of optical fiber (Watanabe, 1993),
and so on. That there are as yet no practical semiconductor-based optical
nonlinear devices is due to small nonlinear susceptibility or, even with large
susceptibility, small bandwidth. Here, I will explain this situation, taking
four-wave mixing as an example referring to the results of Kawatsuka and
his colleagues (1995, 1997).
The third-order nonlinearity of the semiconductor gain medium, like
semiconductor optical amplifiers, can be used for wavelength conversion by
four-wave mixing. Usually, four-wave mixing is done by coupling a pump
beam and a signal beam to a semiconductor optical amplifier, as shown in
Fig. 7.6(a) (for example, Newkirk and Miller, 1993). A new method, illus-
trated in Fig. 7.6(b), was demonstrated by Kuwatsuka et al. (1995) in which
the lasing power of a A/Cshifted DFB laser itself is used as a pump beam,
allowing us to get a wavelength-converted conjugate wave just by coupling
a signal beam.
Figure 7.7 shows an example of measured wavelength conversion effi-
ciency using this method. The measurement employs a A/Cshifted long-
cavity DFB laser with an MQW active layer. The conversion efficiency is
asymmetric with respect to the pump beam wavelength. The conversion
efficiency is higher for the positive detuning; that is, the input signal
%
0,
-
-
Semiconductor optical amplifier
Single-mode semiconductor laser
(b)
FIG.7.6. Four-wave mixing in a semiconductor gain medium: (a) an optical amplifier to
which both the pump wave and the signal wave are coupled: (b) a single-mode semiconductor
laser, in which the lasing beam itself is used as a pump wave.
7 APPLICATIONSOF QUANTUM DEVICES
DOTTO OPTICAL 305
s
= 0 ' " " " " ; ' " " " " ~
>r
g -10
.-
aa
.-0
=aa -20
E
.-02 -30
aa
>
5 -40
1500 1520 1540 1560 1580 1600
Signal wavelength ( nm )
FIG. 7.7. Example of a conversion efficiency from a signal to the conjugate wave.
wavelength is on the longer wavelength side of the lasing wavelength and is

smaller for the negative detuning.
Shown in Fig. 7.8 is the third-order nonlinear susceptibility, x’~’, evaluated
from the conversion efficiency for the positive detuning. The horizontal axis
is the detuning between the pump beam and the signal beam. There are two
components in xt3).One is the large x’~’ at low-frequency regions that is due
to the carrier density beating whose time constant is a recombination
lifetime, and that has rather low response speed. The other is the f 3 ) that is
the spectral hole burning effect. It has a very large bandwidth of 6THz,
which corresponds to 50fs of dephasing time. This time constant is due to
carrier-to-carrier scattering. The asymmetry in the conversion efficiency
arises from the difference in the phase of x ' ~ between
) the carrier-density-
beating component and the hole burning component. For positive detuning
these two components are in phase to give the larger f 3 ) , while for negative
detuning these components are out of phase to give the smaller total x(3'
(Ogasawara and Ito, 1988).
The fast component of f 3 ' is attractive for very-high-speed nonlinear
optical device applications; however, the value is not large enough. Gen-
erally in nonlinear mediums, the larger the bandwidth, the smaller the
nonlinear susceptibility. The wavelength conversion using four-wave mixing
with a sufficient signal-to-noise ratio was realized only for a small wave-
length detuning of several nms. The fiber dispersion compensation by phase
conjugation using the four-wave mixing in the E./it-shifted DFB laser was
demonstrated for a small detuning of 2.5nm (Watanabe et al., 1997).
!
306 HIROSHIISHIKAWA
10-14, r
10'2 1013 1014

Frequency detuning, Aw (radls)
FIG.7.8. Third-order nonlinear susceptibility of MQW for positive detuning estimated
from the conversion efficiency. The dotted lines represent the experimental result. f 3 ' has two
components: the carrier density beating effect and the hole-burning effect.
Asymmetry in the conversion efficiency is another problem for practical

application.
Using quantum dots to obtain a large x’~)with a sufficiently fast response
speed and no asymmetry will be a breakthrough for practical nonlinear
optical devices. Such applications will not be limited to four-wave mixing.
In Section 11 we saw that we can expect an almost zero refractive index
change at the resonant wavelength in quantum dots. However, at the
off-resonant wavelength or at the shoulder regions of the inhomogeneously
broadened emission line, we can expect a large change in the refractive index
by the change in the pump power owing to the large x’~). We can use this
refractive index change to make all optical switches. Here, I try to calculate
the third order nonlinear susceptibility of quantum dots to see if using them
as a nonlinear medium is feasible.
2. ANALYSISOF x ( ~ )
Framework of the Analysis
In Section 11, I discussed the third-order nonlinear susceptibility of

quantum dots. The key determinant of the amplitude and the response speed
7 APPLICATIONS DOTTO OPTICAL DEVICES
OF QUANTUM 307
of f 3 ) is the dephasing rate, yo,. We want a dephasing rate that gives a

sufficiently fast response-say an order of THz-but not so fast as to
reduce the amplitude of x(~), as in the case of the spectral-hole-burning
component in a quantum well.
So far discussions on carrier dynamics in quantum dots have centered on
carrier energy relaxation, which is related to the phonon bottleneck problem
(Bockelmann and Bastard, 1990; Bockelmann, 1993; Efros et al., 1995;
Uskov et al., 1997; Uskov et al., 1998). However, the fast component of
is determined not by the energy relaxation but by the dephasing rate, y,,.
Figure 7.9 shows several dephasing processes for a carrier in the ground
state of a quantum dot. Dephasing takes place due to thermal excitation,
recombination, electron excitation through Auger like process, and carrier
exchange scattering process between carriers in the ground state and in the
barrier layer. Among these four processes, the thermal excitation rate is not
very high when the sublevel separation energy does not coincide with the
LO-phonon energy. The recombination rate is on the order of lo9- lO'O/s,
which is also not very high. The Auger like carrier excitation rate depends
on the carrier density at the high energy region and does not seem high,
although the inverse of this process is important in considering carrier
energy relaxation. Through the Auger like process carriers are supplied to
the ground state from the barrier region (Efros et al., 1995; Uskov et al.,
1997; Uskov et al., 1998). The highest dephasing rate may be obtained for
i-fr
l---i-r
Thermal excitation
---*--
Auger-like excitation
--om-
Recombination Auger-like carrier exchange
FIG. 7.9. Dephasing processes in a quantum dot.

308 HIROSHIISHIKAWA
the carrier exchange process provided when the carrier concentration in the
barrier layer is high. The e-e or h-h scattering through the Coulomb
interaction causes the exchange of a bound electron/hole and a free
electon/hole.
To explore the feasibility of third-order nonlinear susceptibility of quan-
tum dots, I perform a trial calculation, taking the dephasing rate due to the
carrier exchange as a major cause of dephasing. I introduce some rough
approximations and estimate the third-order nonlinear susceptibility using
Eq. (7.2) of Section 11.
CalcuIation of the Dephasing Rate due to Carrier Exchange
For the sake of simplicity, I make the following assumptions and

approximations.
1. Spherical quantum dot is embedded in a bulk large-band-gap material.

2. The Carrier wave function in the barrier bulk material is a plane wave
and remains so even at the quantum dot.
3. As an interaction Hamiltonian, we use screened Coulomb potential
with a Debye-limit screening parameter (Sinha and DiDomenico,
1970).
4. Dielectric constants are constant, not dependent on the wave number
and energy.
For the spherical quantum dot, the ground-state wave function is
Yd(rl)= A sin(ctr,)/r, for Ir,l <a (7.17)
Yd(rl)= ACexp(-Br,)/r, for Itl\ 2 u (7.18)
where A and C are normalization constants. We assume that a plane wave

is incident on the quantum dot. Thus, the initial-state wave function is the
product of the incident plane wave and the dot wave function,
1
Yi(rl,r2) = exp(ik*r,)Y’,(r ,) (7.19)
Jn
~
where, r l and r2 are the respective position coordinates of the incident

carrier and the ground-state carrier. The final state wave function is
1
YJ(rl, rz) = exp(ik'-rl)Yd(rz) (7.20)
Jn
~
This means that the incident electron is captured at the ground state of the
quantum dot and the bound electron in the dot is excited, becoming a free
electron with a wave number of k'. The interaction Hamiltonian for the
process is a screened Coulomb potential, as
where e is the electron charge, E is the dielectric constant, and 9 is a

screening parameter, which is given as a function of carrier density in the
barrier layer as
9= (qz (7.22)
where k is the Boltzmann constant and T is the absolute temperature. The

matrix element, M , for the process is
This matrix element can be evaluated by the Fourier transform, Fd(k), of the
ground-state wave function as
(7.24)
Then
- d3k" (7.25)
e-~k'.r,~;(kt!)e-Ik".r,Herk.r, Fd(k"')e'k"'~'~d3r
310 HIROSHIISHIKAWA
Assuming that the band is isotropic in K space, the scattering rate is given
by Fermi’s golden rule as
w =2 . 2
h
jk
d3k I, ($r
\MI2 f(k)(l - f(k))G(E - Ef)d3k’ (7.26)
where the initial-state energy, E i , and the final-state energy, E,, are given by
(7.27)
and f(k) is the Fermi-Dirac distribution function. Factor 2 in the scattering

rate is due to spin.
Numerical Calculation and Results
The numerical calculation assumed an Ino,,3Gao,,7As spherical dot

embedded in bulk InP. Numerical parameters are as follows:
Potential depth for the electron: 270 meV

Effective mass of the electron: 0.044m0
Potential depth for the hole: 360 meV
Effective mass of the hole 0.54m0
Diameter of the spherical quantum dot: 20 nm
Dielectric constant: 3.22 x to
Potential depths were calculated following the method of Sugawara (1993).

We assumed for simplicity that the carrier masses were equal within the well
and the barrier.
Figure 7.10 shows the scattering rate at 300 K. Scattering due to e-e and
h-h and the total rate are shown as functions of carrier density in the bulk
barrier of InP. The scattering rate of the hole is larger at a higher carrier
density because of the larger density of states in the barrier layer for holes.
As can be seen from Fig. 7.10, we can obtain a total scattering rate on the
order of l O I 3 for the carrier density of the barrier layer larger than
5 1017/~~3.
The third-order susceptibility was calculated using Eq. (7.3) in Section 11,
where ycc is the e-e scattering rate, yI., is the h-h scattering rate, and y,, is
its total. We assumed the inhomogeneous broadening of 6 meV and 30 meV.
Inhomogeneous broadening was incorporated by calculating the ground-
state transition energies for the Gaussian distribution in the diameter of
DOTTO OPTICAL
7 APPLICATIONSOF QUANTUM DEVICES 311
1oi4 1 I I I
300 K
109' I I I I I
lo** 1015 1016 10'7 10'8
Carrier density in barrier ( ~ r n - ~ )
FIG. 7.10. Calculated exchange scattering rate as a function of carrier density in the barrier
layer.
quantum dots, and by summing for all dots. The dot density was at
2 x 1016/cm3,which is equivalent to the volume of 8%, The result is shown
in Fig. 7.11 as a function of the angular detuning frequency between the
pump and the signal wave. Results are shown for carrier densities in the
barrier layer of 1.85 x 10"/cm3 and 1.94 x 10"/cm3. The ycc, yo,, and y,,
for the carrier concentration of 1.85 x 1017/cm3 are 1.1 x lo", 4.08 x
lo", and 5.18 x 101's-', respectively, and, for 1.94 x 101'/cm3, 5.03 x
lo", 1.107 x and 1.159 x 10'3sp1,respectively.
3. DISCLJSS~ON
A rough estimate of the third-order nonlinear susceptibility was made for

a simplified model, taking the carrier exchange scattering process as a cause
of dephasing. Our model is, of course, incomplete, but we can still obtain
perspectives for the third-order nonlinearity in quantum dots.
Dephasing Rate
There is a possibility of realizing a dephasing rate of 1 x 1013s - l by use

of exchange scattering. Of course, other scattering mechanisms may further
312 ISHIKAWA
HIROSHI
10-13 I
- 1 . k x 1017 crn-a
FWHM=6meV --.1.94 x 1018m-a
-N
>
N
-
Y
E
I
L
X
Detuning, Aw (mas)
FIG.7.11. Calculated x ‘ ~ as
’ a function of detuning between the pump wave and the signal
wave.
shorten dephasing time, however, from the measurement and analysis of the
real carrier energy relaxation, tlelaris estimated to be 10-loops (Mukai,
1995; Mukai et al., 1996b; Uskov et al., 1997; Uskov et al., 1998). The
modulation bandwidth is limited to 5-8 GHz (Kamath et al., 1997; Mao et
al., 1997; Wang et al., 1997). These estimates indicate that dephasings
associated with the energy relaxation is not so fast. Exchange scattering
could be an important way to quicken the response of quantum dots when
the carrier density in the barrier layer is high. Also interesting is that our
dephasing rate depends on the quantum dot’s geometry. This can be seen
from the functional form of the matrix element, which is
where k”’ is the wave number of the ground-state carrier, and k is the wave
number of an incident carrier. This equation means that the maximum is
obtained when the K-space distribution of the ground-state electron wave
function reaches the wave number of the incident electron. Because the
K-space distribution of IFd(k”’)ldepends on the geometry of the quantum
DOTTO OPTICAL
7 APPLICATIONSOF QUANTUM DEVICES 313
dot, the dephasing rate may depend on the geometry. Our calculation is for
a spherical dot, so it will be interesting to extend the calculation to SK dots
whose geometry is not spherical.
Amplitude and Bandwidth of x(3)

When compared with the f 3 ' for 1.85 x 1017/cm3 and 1.94 x 1018/cm3,
the bandwidth is narrower and the amplitude is smaller for higher concen-
tration of 1.94 x lOl8/cm3 as can be seen from Fig. 7.11. Narrower band-
width for higher concentration is because of the smaller dephasing rate of
electrons at very high concentration regions. The smaller amplitude of x(3)
is because of the larger total dephasing rate. At a lower carrier density of
1.85 x lOl7/crn3, both electrons and holes have the appropriate dephasing
rate to produce a wider bandwidth close to 1 THz. When compared with
the hole-burning component of x’~’ of the quantum well in Fig. 7.8, the
bandwidth is smaller because of the smaller dephasing rate. However, a
much larger x(3) was obtained even for the inhomogeneous broadening of
30meV. For the broadening of 6meV, x’~)is still one order of magnitude
larger than that of the 30-meV broadening. There are two reasons for this.
One is the discrete energy levels of the quantum dot, which concentrates
oscillator strength at the resonant wavelength. The other is the compara-
tively longer dephasing time than that of the MQW. The longer the
dephasing time, the larger the xt3). What is important is to obtain an
appropriate dephasing time to give a sufficiently fast response for device
application without decreasing the amplitude of x ' ~ )Our
. calculation shows
that it is possible to achieve such a condition. Of course, as x’~) is
proportional to dot density, an increase of dot density is essential.
Asymmetry in f3)
The asymmetry in x’~’with respect to the pump wavelength is a problem

in the f 3 ' of carrier-injected MQW. It arises from the phase difference
between the carrier-density-beating effect and the hole-burning effect. In
quantum dots, we can expect a very small change in the refractive index for
the change in carrier concentration near the inhomogeneously broadened
resonant wavelength, as discussed in Section 11. We can expect an in-phase
response of the carrier-density-beating effect and the hole-burning effect
near the resonance. Although there may be some effect from the carriers in
the barrier layer, we should see an almost symmetric x’~’with respect to the
pump wavelength.
314 HIROSHIISHIKAWA
Thus, with a simplified model and analysis, I have calculated the x(3) of
quantum dots, and I have demonstrated the possibility of obtaining fast
responses of around 1 THz, and also a much larger value than that of
quantum wells. In addition, I have shown the possibility of obtaining a
symmetric f 3 ) with respect to pump wavelength. Of course, development of
more precise and rigid theory is required to draw decisive conclusions,
especially for the dephasing process.
Our discussion is for a four-wave mixing process, but prospective appli-
cations are not limited to that. I emphasized the small refractive index
change close to the center of the inhomogeneously broadened spectral line,
which should eliminate asymmetry in f 3 ’ . However, at the shoulder regions
of the inhomogeneously broadened line, we can expect a large change in the
refractive index due to the large x(~).Possibly we can control the refractive
index at the shoulder region by the pump power. There will be applications
for all optical switches.
So far, no concrete measurement of third-order susceptibility has been
reported. Experimental verification of the usefulness of the quantum-dot’s
nonlinearity is a very important task, together with development of a more
rigid and precise theory of nonlinearity.
V. Persistent Hole Burning Memory
The inhomogeneous broadening of quantum dots is an obstacle to

obtaining large linear gain or large non-linear susceptibility. However, Muto
(1995) proposed a use of it as a persistent hole burning (PHB) memory.
Here I review Muto’s idea and then introduce recent experimental results.
1. PERSISTENT
SPECTRAL
HOLEBURNING
MEMORY
USING
QUANTUMDOTS
Persistent spectral hole burning memory is highly attractive because it

makes feasible the storing of information with a very high multiplication
factor in a small area by use of the wavelength domain. At a low
temperature of 4 K , high multiplication factors (MPFs) of lo3-lo4 are
expected, using polymers, alloys and glasses. The problem with such
materials is that the M P F reduces at higher temperatures, and it has been
very difficult to make practical PHB memories that operate at room
temperature (Ao et al., 1993). Muto noticed that inhomogeneous broaden-
ing of the ground-state transition can be used for PHB if we can realize
7 APPLICATIONS
OF QUANTUM DOTTO OPTICAL
DEVICES 315
persistent hole burning in quantum dots. The inhomogeneous broadening

so far reported is around 25-80 meV (for example, Ishikawa et al., 1998) and
can be increased to 100meV by adjusting the growth condition. If the
homogeneous linewidth of the ground-state transition is below 0.1 meV, the
M P F can be on the order of lo3.
Muto proposed a method of obtaining persistent hole burning by extract-
ing electrons from the potential well of the conduction band. When a
quantum dot is irradiated by a ground-state electron-hole transition
wavelength, the electron is excited to the ground state of the conduction
band. If it is left as is, recombination takes place and no persistent hole
burning occurs, but if we can extract electrons from the ground state-for
example, by tunneling to the X band of the barrier material-the hole
remains in the ground state and there is no additional absorption. In other
words, persistent hole burning takes place.
Muto gives us a more detailed estimation of the performance of the PHB
memory using quantum dots. The memory duration is determined by the
product of the probability of electrons going back to the ground state and
the recombination rate. Muto assumed the probability to be the Fermi-
Dirac distribution and further assumed 20 ns of recombination lifetime.
Thus, he obtained a duration time of 0.2ms at room temperature. To get
much longer memory time, some refreshing procedure would be necessary.
Muto also made some estimation for the reading. He assumed the reading
by photo-induced current by applying the reading photon flux. The condi-
tion for reading is a sufficient signal-to-noise ratio (S/N) without the reading
photon flux burning the “virgin”, or unburned, wavelength. The signal in
reading is given by the difference between the currents transmitted through
the virgin area and those transmitted through the marked area. The noise
is defined as a sum of the root mean square of fluctuation of shot noise due
to the currents at the virgin region and the burned region. Muto required
an S/N larger than 20, giving the minimum of the reading photon flux, F,.
Another condition that photon flux does not burn the virgin region is that
the product of D F ~ be T ~smaller than unity, where D is the absorption
cross-section and z R is the reading time. This gives the maximum of the
cross-section, D. The maximum MPF is defined to be N,,,/N,, where N,,, is
the total density of the dot and N, is the density of dots within a
homogeneous linewidth at o.Note that the total absorption coefficient,
is DN,,
(7.29)
where L is the absorption length. Assuming a realistic homogeneous broad-

’
ening of 1 meV, an D of 1 x 10- cm2, and a n area of reading laser beam
316 HIROSHIISHIKAWA
of 1pm’, Muto obtained
MPFIL < 6.7 x 10/pm (7.30)
to satisfy the condition aFRzR< 1. For an N,,, of 1 x 1018/cm3,an homog-

eneous broadening of 1meV and an inhomogeneous broadening of
200meV, M P F is 100 for a 1.5-pm-thick absorption layer. In this case,
1.5 x lo4 quantum dots contribute one bit of memory.
RESULTS
2. EXPERIMENTAL
Several experiments confirm the persistent hole burning effect in quantum

dots. One used a Shottkey contact structure, shown in Fig. 7.12 (Imamura
et al., 1995). The i-GaAs region contained self-organized InAs quantum dots
with an inhomogeneous broadening of 73 meV. When the structure was
irradiated by the writing optical pulse, an electron-hole pair was generated
at quantum dots whose transition wavelength coincided with the write
pulse. Because of the built-in field caused by the Shottkey contact, electrons
excited out to the conduction band tunnels from the dot, while the holes
remained at the ground state because of the high barrier. This bleached the
quantum dots. Imamura and his colleagues measured the difference of the
photo-induced current for the write and read pulses, which was At after the
write pulse at a room temperature of 300 K. They used the same wavelength
for both write and read.
The result is shown in Fig. 7.13. For the small time difference, the
difference in the current was large, which means that persistent hole burning
was taking place. The difference reduced for a larger At, which Imamura
FIG. 7.12. Band lineup of a sample used in the persistent hole-burning experiment by
Imamura et al. (1995, 0 1995 J p n . J . Appl. Phys.).
7 APPLICATIONS
OF QUANTUM
DOTTO OPTICAL
DEVICES 317
lo-’ I I I I I I l i
lnAs QDs, 300 K
-
Q)
c
L
.
- I I I t I I I
10-0
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
At (ms)
FIG 7 13 Difference in the photo-induced current between the write pulse and the read
pulse as a function of a time interval between the two pulses (Imamura et al , 1995, P 1995
J p n J Appl P h y s )
attributed to the recombination of residual holes with thermally excited

electrons from the n-GaAs. From the experimental result, Imamura found
the retention time of the hole burning to be 0.27 ms at room temperature.
Sugiyama and his colleagues (1997) observed the spectral hole at a low
temperature of 5 K. Their sample, shown in Fig. 7.14, had a p-i-n structure
n+-GaAs,300 nm 1
I (001) n+-GaAssubstrate I
FIG. 7.14. Schematic of a sample used in the persistent hole-burning experiment by

Sugiyama (1997, Q 1997 I E E E ) .
318 HlROSHl ISHlKAWA
1 x lo-”
9
U
e 0
2
-1 x lo-”
1063 1084 106S 1066
Wavelength (nm)
FIG. 7.15. Comparison of experimental and calculated results of spectral hole burning at
5 K: (a) the experimental result, (b) a calculated fit. The writing wavelength of YAG consists
of three wavelengths, as indicated by the arrows (Sugiyama et al., 1997, 0 1997 IEEE).
with five-layer stacked quantum dots at the i region. Each layer contained
dots with a density of 5 x 10’o/cm2. The sample was irradiated by a CW
YAG laser at 1064 nm, and the reading was taken using a tunable CW TiS
laser. The YAG laser consists of several emission lines 0.1 or 0.2 nm apart
with a linewidth of 27 peV. The linewidth of the TiS laser is below 165 peV.
Measurements were performed at 5 K. After irradiation by the YAG laser
at 8-mW output, the photoresponse current was measured irradiating by the
TiS laser at 30pW. When the reverse-bias voltage to the sample was below
4 V, the researchers observed no spectral hole in the response current.
However, when they increased the inverse bias up to 5 V , an apparent
spectral hole became visible. Trace (a) in Fig. 7.15 is the measured spectral
hole under 6-V reverse bias. At a low bias voltage, the generated electron
and hole remained in the dots and recombined, producing no spectral hole.
When the bias voltage was increased, the electron tunneled out from the dot,
initiating persistent hole burning. This took place for a bias voltages larger
than 4 V.
Sugiyama and his colleagues simulated the shapes of the spectral hole.
Assuming a Gaussian-shaped TiS read light with an FWHM of 165peV,
and assuming a Lorentzian-shaped spectrum for both the YAG read light
and the quantum dot, they performed the fit with a measured spectral hole,
taking convolutions. Trace (b) in Fig. 7.15 shows the calculated result. Note
that the YAG emission line consists of three lines, as indicated by arrows in
the figure. The best fit was obtained for a homogeneous broadening smaller
than 80 peV.
3. DISCUSSION
Based on Muto’s proposal, experiments were performed to verify the

hole-burning effect in quantum dots. The existence of spectral holes was
confirmed at 300 K, with a retention time of 0.27 ms, and the spectral-hole
shape was measured at a low temperature of 5 K. These results encourage
us to believe that persistent hole burning memory using quantum dots is
feasible.
The maximum multiplication factor depends largely on the parameters of
quantum dots, such as homogeneous broadening width, dot density, and
dot-layer thickness. If we can achieve an M P F of 100 for a 1-pm2 area, we
will have 10Tb/cmz of memory. This is a huge capacity. For this reason,
persistent hole burning memory is a highly attractive application of quan-
tum dots.
VI. Summary and Perspectives on Quantum-Dot Optical Devices
In this chapter, I reviewed the basic properties of quantum dots and

discussed several possible device applications. I also discussed chirpless
high-speed modulation of quantum-dot lasers, a blue-chirp intensity modu-
lator, the feasibility of quantum dots as a nonlinear medium, and state-of-
the-art research of Tb/cm2 persistent hole burning memory. Of course, there
are many other applications for quantum dots. Among them is an infrared
photodetector which, although not discussed here, is considered an attrac-
tive alternative to the quantum-well infrared photodetector (QWIP) (Ryzhii,
1996).
Our discussion in this chapter was not necessarily based on concrete
theory. My treatment of the quantum dot as a two-level system is insuffi-
cient under high optical power, and my analysis of the dephasing rate is
merely a rough sketch of the process that may require the inclusion of
phonon scattering to be more comprehensive. Many more detailed and
concrete theoretical studies, together with experimental verification, are
needed. Nevertheless, we have obtained some perspective on quantum dots,
both positive and negative, for device application.
320 HIROSHIISHIKAWA
Finally, I sketch the trends of optoelectronics and illustrate some uses for
quantum-dot optical devices.
1. TRENDSIN OPTOELECTRONICS
Research on optoelectronics has been fueled by a strong need for optical

communication systems. At present, long-haul, local-area, and undersea
cable systems are in practical use. Moreover, applications are expanding to
systems not only for communication but for information processing as well.
Still larger capacity and/or longer-distance trunk line and undersea cable
systems are being developed. Local-area and access network systems are
also urgently required. The optical interconnection between boards and
chips, and within chips, is becoming a research target. Other than communi-
cation and data transmission, optical data storage and optical image
processing are also important research areas.
Two approaches are being taken to achieve larger-capacity transmission.
One is the WDM system; the other is the much higher data rate TDM (time
division multiplexing) system. WDM systems with a multiplication factor of
30-80 wavelengths (with spacing of 0.4-0.8 nm), each carrying a 2.5-10-
Gb/s signal, are being developed. They are attractive not only for their large
multiplication but also for their flexible routing of signals by wavelength.
However, these systems require the development of new devices such as
fixed and/or tunable wavelength filters, multi-wavelength and/or tunable
light sources, and wavelength conversion devices and those with a very high
bit rate of above 10 Gb/s demand very low chirp, high-speed light sources.
40 Gb/s systems are being planned, and Tb/s TDM systems are envisioned.
In both WDM and TDM systems, dispersion compensation of a fiber to
enable long-distance transmission is essential. Access networks systems, such
as fiber to the home (FTTH) and fiber to the curb (FTTC), require low
power consuming lasers that provide temperature-robust performance. For
optical interconnections, we must develop very low power consuming,
temperature-robust light sources for parallel optical links between com-
puters, boards, chips, and, hopefully within chips. The problems of wiring
and pin bottlenecks should be solved by the introduction of optics in
interconnections. However, very cheap, high-performance devices are re-
quired in access networks and interconnections, especially optical intercon-
nections. These devices must be cheaper than copper-wire interconnections
and must have sufficiently higher capacity data throughputs.
In all of the above systems, space, time, and wavelength domain optical
switches will be essential to achieve flexibility and smartness. Data storage
also will be important for small-size and huge-capacity throughput.
DOTOPTICAL
2. USESFOR QUANTUM DEVICES
Quantum dots will play an important role in the future of optoelectronics.

For access networks and interconnections, the quantum-dot laser could be
a key component because of its potential for low threshold current oper-
ation and temperature-robust operation (see Chapter 6), but keeping them
low-cost will take extensive effort. For high-speed TDM systems above
10 Gb/s, the very small wavelength chirp under high-speed modulation in
quantum-dots is attractive, as is the blue-chirp quantum-dot intensity
modulator. For WDM systems, the nonlinearity of quantum dots looks
promising as a wavelength conversion device. Generation of a phase
conjugate wave by a nonlinear process such as four-wave mixing can be
used for dispersion compensation of optical fibers to attain high bit rate
long-distance transmission. The quantum dot’s nonlinearity may also be
used in space and time domain switches to increase their flexibility and
smartness. Finally, quantum dots have the capacity for large data storage
because of the persistent hole burning effect, and the potential for low-dark
current infrared photodetection and imaging.
Thus, we see many important roles for quantum dots in optoelectronics.
As I said before, however, we must establish low-cost, high-quality fabrica-
tion technologies for their practical application. It is my hope that some of
the devices discussed here will come into practical use within the next five
years.
Acknowledgment
The author would like to give hearty thanks to his collaborator, Dr. H.
Kuwatsuka, for his comments and advice. The study of Section IV quantum
dots as a nonlinear medium has been done under the management of
Femtosecond Technology Research Association (FESTA), supported by
the New Energy and Industrial Technology Development Organization
(NEDO) Japan.
REFERENCES
Adachi, S. (1982). 1. Appl. Phys. 53, 5863.

Ao, R., Kummerl, L., Haarer, D. (1993). J p n . J . Appl. Phys. 32, 5248.
Bennett, B. R., Soref, R. A., Del Alamo, J . A. (1990). IEEE J . Quantum Electron. 26, 113
Bockelmann, U., and Bastard, G . (1990). Phys. Rev. B. 42, 8947.
Bockelmann, U. (1993). Phys. Rev. B. 48, 17637.
322 HIROSHIISHIKAWA
Bowers, J. E., Hemenway, B. R., Gunack. A. H., and Wilt, D. P. (1986). IEEE J . Quantum
Electron. QE-22, 833.
Devaux, F., Bigan, E., Ougazzaden, A., Huet, F., Carre, M., and Carenco, A. (1992). Electron
Lett. 28, 2157.
Efros, A. L. Kharchenko, V. A., and Rosen, M. (1995). Solid State Comm. 93, 281.
Green, C . A., Dutta, N. K., and Watson, W. (1987). Appl. Phys. Lett. 50, 1409.
Haisch, H., Baumert, W., Hache, C., Kiihn, E., Klenk, M., Satzke, K., Schilling, M., Weber, J.,
and Zielinski, E. (1994). Proc. 201h Eur. Con$ Optical Communication 2.
Henry, C. H. (1982). IEEE J . Quantum Electron. QE-18, 259.
Hu, Y. Z., Lindberg, M., and Koch, S. W. (1990. Phys. Rev. B. 42 1713.
Imamura, K., Sugiyama, Y., Nakata, Y., Muto. S., and Yokoyama, N. (1995). Jpn. J . Appl. Phys.
34, L1445.
Ishikawa, H., Soda, H., Wakao, K., Kihara, K., Kamite, K., Kotaki, Y., Matsuda, M., Sudo,
H., Yamakoshi, S., Isozumi, S., and Imai, H. (1987). IEEE J. Lightwave Technol. LT-5,848.
Ishikawa, H., and Suemune, I. (1994). IEEE Photonic Technol. Lett. 6, 1315.
Ishikawa, H., Shoji, H., Nakata, Y., Mukai, K., Sugawara, M., Egawa, M., Otsuka, N.,
Sugiyama, Y., Futatsugi, T., and Yokoyama, N. (1998). J . Vacuum Sci. & Technol. A16,794.
Kakimoto, S., Nakajima, Y., Sakakibara, Y., Watanabe, H., Takemoto, A,, and Yoshida, N.
(1990). IEEE J . Quantum Electron. 26, 1460.
Kamath, K., Klotzkin, D., and Bhattacharya. P. (1997). IEEE Lasers and Electo-Optics SOC.
Ann. Meet. ThY4, San Francisco.
Kane, E. 0. (1957). J . Phys. Chem. Solids 1. 249.
Kataoka, T., Miyamoto, Y., Hagimoto, K.. Sato, K., Kotaka, I., and Wakita, K. (1994). Electron
Lett. 30, 872.
Kirstaedter, N., Ledentsov, N. N., Grundmann, M., Bimberg, D., Ustinov, V. M., Ruvimov,
S. S., Maximov, M. V., Kop’ev, P. S., Alferov, Z. I., Richter, U., Werner, P., Gosele, U., and
Heydenreich, J. (1994). Electron. Lerr. 30. 1416.
Koch, T. L., and Linke, R. A. (1986). A p p l . Pkys. Lett. 48. 613.
Koyama, F., and Iga, K. (1985). Elerrron. Lerr. 21, 1065.
Kuwatsuka, H., Shoji, H., Matsuda, M.. and Ishikawa, H. (1995). Electron Lerr. 31, 2108.
Kuwatsuka, H., Shoji, H., Matsuda, M., and Ishikawa, H. (1997). I E E E J . Quantum E/ectron.
33, 2002.
Lee, Y. H., Chavez-Pirson, A,, Koch, S. W.. Gibbs, H. M., Park, S. H., Morhange, J., Jeffery,
A,, and Peyghambaridn, N. (1986). Phys. Rev. Lett. 57,2446.
Mao, M.-H., Heinrichsdorff, F., Kitstaedter, N., Krost, A,, Ledentsov, N. N., and Bimberg, D.
(1997). IEEE Lasers and Electro-Optics Sor. Ann. Meet. ThY5, San Francisco.
Morito, K., Sahara, R., Sato, K., Kotaki, Y . (1996). IEEE Phofonic Technol Lett. 8. 431.
Mukai, K., Otsuka, N., Shoji, H., and Sugawara, M. (1995). Phys. Rev. B. 54, 5243.
Mukai, K., Ohtsuka, N., and Sugawara, M. (1996a). J p n . J . Appl. Phys. 35, L262.
Mukai, K., Ohtsuka, N., Shoji, H., and Sugawara, M. (1996b). Appl. Phys. Lett. 68, 3013.
Muto, S . (1995). Jpn. J . Appl. Phys. 34, L210.
Nagarajian, R., Ishikawa, M., Fukushima. T.. Geel, R. S., and Bowers, J. E. (1992). I E E E J.
Quantum Electron. 28, 1990.
Nakata, Y., Sugiyama, Y., Futatsugi. T.. and Yokoyama, N. (1997). J . Cryst. Growth 175/176,
713.
Newkirk, M. A,, and Miller, B. I. (1993). Appl. Phys. Lett. 63, 1179.
Olshansky, R., Hill, P., Lanzisera, V., and Powazinik, W. (1987). I E E E J . Quantum Electron.
23, 1410.
Ogasawara, N., and Ito, R. (1988). Jpn. J . Appl. Phys. Lett. 27, 615.
Ryzhii, V. (1996). Semirond. Sci. Technol. 11. 759.
Sahara, R., Matsuda, M., Shoji, H., Morito, K., and Soda, H. (1996). IEEE Photonic Tecliriol.
Lett. 8, 1477.
Seki, S., and Yokoyama, K. (1994). Phys. Ren B. 50, 1663.
Seino, M. (1996). 1st Optoelectronics and Cornmunicutions Conj.' (Technical Digest) 17D-2, 182.
Shoji, H., Nakata, Y., Mukai, K., Sugiyama, Y., Sugawara, M., Yokoyama. N.. and Ishikawa,
H. (1996). Electron. Lett. 32, 2023.
Sinha, K. P., DiDomenico Jr., M. (1970). Phys. Rev. B. 1, 2623.
Soda, H., Sato, K., Nakai, K., Ishikawa, H., Imai, H. (1988). Electron Lett. 24, 1194.
Sugiyama, Y., Nakata, Y., Muto, S., Horiguchi, N., Futatsugi, T., Awano, Y., Yokoyama, N.
(1997). Electron Lett. 33, 1655.
Tucker, R. S. (1985). IEEE J. Lightn~nvr~ Technol. LT-3, 1180.
Uomi, K., Tsuchiya, T., Nakano, H., Aoki, M., Suzuki, M., Chinone, N. (1991). I E E E J .
Quantum Electron. 27, 1705.
Uskov, A. V., Adler, F., Schweizer, H., and Pilkuhn, M. H. (1997). J . Appl. Phys. 81. 7895.
Uskov, A. V., Mcinerney, J., Adler, F., Schweizer, H., and Pilkuhn, M. H. (1998). Appl. Phys.
Lett. 72, 58.
Wang, J., Griesingter, U. A., and Schweizer, H. (1997). Con$ Lusers nd Electro-Optics CWF19,
Baltimore.
Watanabe, S., Naito, T., Chikama, T. (1993). IEEE Pliotonic Technol. Lett. 5. 92.
Watanabe, S., Kuwatsuka, H., Takeda. S., and Ishikawa, H. (1997). Electron. Leti. 33. 316.
Wiesenfeld, J. M., Tucker, R. S., and Downey, P. M. (1987). Appl. Phys. Lett. 51, 1307.
CHAPTER 8
The Latest News

Mitsuru Sugawara, Kohki Mukai
OPTICAL SEMICONDUCTOR DEVICES LAHORATORl
FUJITSULABORATORIES
Lm.
ATSLIGI,
KANAGAWA, JAPAN
Hiroshi Ishikawa, Koji Otsubo

ELECTRON
DOVICOS AND MATERIALS
LARoRATORl
FLIJITSLI
LABORAToRltS LTU.
ATSUGI. KANAGAWA.
JAPAN
Yoshiaki Nakata
QUANTUM
ELOCTRON
DEVICES LABORATORY
F U J I T SLABORATORIES
~I LTD
ATSUUI,KANAGAWA, JAPAN
I. LASING WITH LOW-THRESHOLD CURRENT AND HIGH-OUTPUT POWER

FROM COLUMNAR-SHAPED QUANTUM DOTS. . . . . . . . . . . . . . . . 325
11. EFFECT OF HOMOGENEOUS BROADENING OF SINGLE-DOT OPTICAL GAIN
ON LASING SPECTRA . . . . . . . . . . . . . . . . . . . . . . . . . 328
111. QUANTUM DOTSON INGAASSUBSTRATES . . . . . . . . . . . . . . . . 331
IV. QUANTUM DOTSEMITTING AT 1.3Aim GROWN BY Low GROWTH RATES
AND WITH A N INGAASCAP . . . . . . . . . . . . . . . . . . . . . . 333
v. REDUCED-TEMPERATURE-INDUCED VARIATION OF SPONTANENOUS EMWON
I N ALTERNATESUPPLY (ALS) QUANTUM DOTSCOVERED
BY In,,,Ga,,,As . . . . . . . . . . . . . . . . . . . . . . . . . . . 336
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
I. Lasing with Low-Threshold Current and High-Output Power from

Columnar-Shaped Quantum Dots
In the columnar-shaped self-assembled InGaAs quantum dot lasers de-

scribed in Chapters 2 and 6, we achieved a threshold current of 5.4mA, a
325
Copyright ( 1999 by Academic Press
All rightr of reproduction in any form reserved
ISBN 0-12-752169-0
ISSN 0080-8784/99 $3000
326 MITSURUSUGAWARA
ET AL.
current density of 160A/cm2, and an output power of llOmW at room

temperature (Mukai et al., 1998).
Columnar-shaped self-assembled quantum dots are grown by stacking
InAs self-assembled dots with a few-monolayer-thick GaAs intermediate
layer. The structure as a whole can be considered one physically coupled
large dot. This new dot type has high uniformity in size and high emission
efficiency, which are desirable properties for semiconductor lasers. Using
one columnar-dot layer (n = 1) or two sheets of the layers separated by a
30-nm GaAs layer ( n = 2), we fabricated double heterostructure lasers. The
columnar-dot layers were sandwiched between GaAs separate-confinement-
structure (SCH) layers. The active region was grown on a 1.4-pm n-
Al,.,Ga,,,As cladding layer, followed by a 1.4-pm p-Al,.,Ga,,,As cladding
layer and a 0.4-pm p-GaAs contact layer. We formed 3-pm-wide and
1.2-pm-high ridge structures by chemical etching.The cavity lengths ( L )were
300 and 900pm.
For a laser with n = 2, L = 300pm,and a high-reflective (HR) coating on
both facets, lasing oscillation occurred for a threshold current of 5.4mA at
25°C (Fig. 8.1). With L = 900pm and an HR coating on both facets, the
threshold current density was reduced to as low as 160 A/cm2 (i.e., 80 A/cm2
10
FE
W
6
E
0
% 4
1
,a
8 2
n = 2, HFUHR, CW
n
"
I J/, I , , , I , , , I , , , , , , , I , , ,I , , ,
0 20 40 60 80 100 120 140

Current (mA)
FIG. 8.1. Output power versus current characteristics of a columnar-dot laser with n = 2,
L = 300pm,and a high-reflective (HR) coating on both facets. (Reprinted from Mukai et al.,
(3 1998 IEEE).
8 THELATESTNEWS 327
per single sheet of dot layer). In both lasers, lasing occurred from the top of
spontaneous emission of the ground state at 1.15pm. For a laser with
L = 900pm and an HR coating on one side of the facets, an output power
of 110mW was achieved (Fig. 8.2). Even at 70°C, an output power of 75 mW
was obtained. To our knowledge, such low threshold current and high
output power have never before been reported for edge-emitting quantum-
dot lasers.
Figure 8.3 shows the lasing emission spectra of the laser with n = 1,
L = 900pm, and an HR coating on both facets up to 300mA. The lasing
started at the ground state with a threshold current of 11.5mA. As the
current increased to 60mA, the lasing spectra became broad with spikes;
then, at 80mA, lasing from the excited state appeared at around 1.1 pm.
What is surprising is that the lasing at the ground state weakened at 130mA
and completely disappeared at 200mA. The spectrum at 300mA shows
broad-band lasing only from the excited state. This jump of lasing
wavelength from the ground to the excited state can be explained by gain
saturation at the ground state, which is due to the phonon bottleneck
and/or third-order nonlinearity, and a subsequent increase in the carrier
number in the excited state (see Chapter I). The quenching of lasing at the
120, I , I I I I I , , , ,,,,, ,,,, ,,

, , ,, I , ,, , ,
100
80
E
W
60
Y
3 40
B
0' 20
0
0 50 100 150 200 250 300 350
Current fmA)
FIG. 8.2. Output power versus current characteristics of a columnar-dot laser with ti = 2,
L = 900pn, and a high-reflective (HR) coating on one side of the facets. (Reprinted from
Mukai et at., 1998 IEEE).
328 MITSURUSUGAWARA
ET AL.
0.9 1.0 1.1 1.2 3

Wavelength (pm)
FIG.8.3. Lasing emission spectra of the laser with n = 1, t = 900prn, and an HR coating
on both facets up to 300mA.
ground state might be due to the elevated temperature of quantum dots

under high current injection and the resultant increase in the carrier
emission rate from quantum dots to SCH layers.
Let us discuss the reasons that we achieved high lasing performance in
columnar-dot lasers. First, high uniformity in size, large density, and high
emission efficiency of the columnar dots are probably the decisive factors.
Second, closely stacked wetting layers around the dots capture carriers more
efficiently than a single wetting layer does, because the wetting layers in
columnar dots are electrically coupled like a superlattice and thus have a
deep potential (proved by the fact that the emission wavelength is longer
than that of SK dots, as shown in Fig. 2.22. Third, the closely stacked
wetting layers increase the refractive index around the dots and enhance the
optical confinement factor. Fourth, columnar dots with a rather symmetric
shape have a a larger electron-hole overlap integral than do ordinary SK
dots with a lens-like shape and thus have a higher optical gain.
11. Effect of Homogeneous Broadening of Single-Dot Optical Gain

on Lasing Spectra
We found that dots with different energies start lasing independently at

low temperatures due to their spatial localization, while at room tempera-
8 THELATESTNEWS 329
ture the dots contribute to one-line lasing coilectively via homogeneous

broadening of optical gain. The homogeneous broadening at room tempera-
ture is estimated to be more than 10meV and can be attributed to the
scattering process to exchange electrons (holes) between the ground state
and the excited states, including the wetting layers.
In Chapter 1, Section V.B, we derived a formula of quantum-dot optical
gain that takes into account both inhomogeneous and homogeneous
broadening. We pointed out that, if homogeneous broadening is a delta
function, each single quantum dot contributes to lasing independently,
leading to broad lasing spectra. Also, when homogeneous broadening is
comparable to inhomogeneous broadening, the dot ensemble in the cavity
contributes to lasing collectively, leading to lasing from the top of the optical
gain spectrum, This is confirmed in the experiments on light emission
spectra of the columnar-shaped, self-assembled InGaAs quantum dot lasers
with a 300-pm cavity and with a high refectivity (95%) coating on both
facets.
Figure 8.4 compares the light emission spectra between 80 K and 298 K,
up to well above the lasing threshold. At 80K, as the current increases, the
spontaneous emission rises from the top of the spontaneous emission
spectrum, leading to broad lasing emission over the range of 5OmeV at 5 mA
and 60meV at 20mA This can be explained as follows: At this low
temperature, homogeneous broadening is negligible, and thus dots with
different energies have no correlation to each other because they are
spatially isolated from each other. As a result, all dots with an optical gain
above the lasing threshold start lasing emission independently. The quan-
tum-dot laser in this situation behaves as if it included independent lasing
media in the same cavity.
At 298 K, the spontaneous emission at 3 mA -about half the threshold
current -showed the emission due to the ground-state, band-to-band
transition between the discrete quantized states at 1.15pm, the second-state
transition at 1.08pm, and the transition at the wetting layers at 1.0pm. At
lOmA, the lasing emission line appeared from around the top of the
spontaneous emission spectrum. The line at 10 mA has the full width at half
maximum of 1.5 nm, and the mode separation of this cavity laser is 0.62 nm,
indicating that the line includes 2 or 3 longitudinal modes. At higher
injection currents, an additional line appeared at 20 mA with a wavelength
of 1.16pm separated by 19meV from the central line at l.l4pm, and then
at 40mA with 1.123pm separated by 16meV from the central line. This can
be explained as follows: Homogeneous broadening of the optical gain is
comparable to inhomogeneous broadening at 298 K, and thus lasing-mode
photons come not only from energetically resonant dots but also from other,
nonresonant dots that lie within the amount of homogeneous broadening.
330 MITSURUSUGAWARA
ET AL.
.=
v)
C
10
Y
$ 1
a
W
.-E
v)
10.'
C
Q)
E
- 10"
10"
1 .o 1.1
(a) Wavelength (pm)
1.o 1.1 1.2
(b) Wavelength (pm)
FIG. 8.4. Light emission spectra of columnar-shaped self-assembled InGaAs quantum-dot

lasers with I I = 2, L = 900pm. and an HR coating on both facets at (a) 8 0 K and (b) 298K.
Since carriers of nonresonant dots are extracted and put into the central
lasing modes by the stimulated emission, lasing emission with one line
occurs. As the injection current increases, the gain saturation occurs due to
the phonon bottleneck and/or third-order optical nonlinearity, which results
in new lasing lines in addition to the central line. The energy separation
between the lasing lines in Fig. 8.4b- 19 and 16meV-corresponds to the
homogeneous broadening of this laser chip.
The origin of the homogeneous broadening can be attributed to the
8 THELATESTNEWS 331
scattering process to exchange electrons (holes) between the ground state

and the excited state, including the wetting layers. According to the
theoretical calculation of the scattering rate in Chapter 7, the total scattering
rate, this is, the sum of the rates of hole-hole and electron-electron
scattering, reaches up to 1 x 1013s-' when carrier concentration is
1 x 10l8cm-3 in the wetting layers, which causes the broadening of optical
gain with the full width at half maximum of 13 meV. When carrier concen-
tration is 1 x 1015cm-3, the total scattering rate is 8 x l O ' O s - ' , giving
0.1-meV broadening. From the spontaneous emission spectra of Fig. 8.4, we
see that, while carriers are almost in the ground state at 80 K, there are also
carriers in the excited state and in the wetting layer at 298 K. Thus, carriers
in the excited state and/or in the wetting layer work as scattering particles
against electrons and holes in quantum dots, resulting in large homogeneous
broadening of optical gain and one-line lasing at 298K, as seen in Fig.
8.4(b).
From the viewpoint of physics, it is interesting that homogeneous
broadening leads to an interaction of spatially and energetically isolated
quantum dots through photons, and that collective lasing is achieved. With
respect to technology, it is important that the interaction leads to a single
lasing line at room temperature. This is the first demonstration of the effect
of homogeneous broadening on quantum-dot laser characteristics.
By further increasing currents, broad-band lasing spectra are also ob-
served at room temperature, as seen in Fig. 8.3. This is because of the gain
saturation that is due to the phonon bottleneck and/or third-order non-
linearity at high current injection, leading to an increase of the carrier
population and thus lasing in the dots nonresonant to the central lasing line.
For this reason, the phonon bottleneck and third-order nonlinearity domi-
nate the purity of lasing spectra.
111. Quantum Dots on InGaAs Substrates
The growth of quantum dots on ternary InGaAs substrates is an alterna-

tive way to realize quantum-dot lasers emitting at 1.3pm. We show our
preliminary results on the growth of quantum dots on InGaAs substrates.
Research is underway to fabricate ternary In, -,Ga,As as a new category
of substrates for optical devices besides GaAs and InP (Nakajima and
Kusunoki, 1996; Kusunoki et al., 1996; Nakajima et al., 1997). We recently
achieved a low threshold of 176 A/cm2 and a high characteristic temperature
of 140K, a record high operating temperature up to 210"C, and tempera-
ture-insensitive slope efficiency in 1.2-pm strained quantum-well lasers on
332 SUGAWARA
MITSURU ET AL.
In,,,,Ga,,,,As substrates (Shoji et al., 1994; Shoji et al., 1996; Otsubo et al.,
1997). Our aim is to fabricate 1.3-pm lasers on the ternary substrates with
low lasing threshold and excellent temperature characteristics for future
optical interconnection and optical subscriber systems.
The merits of ternary substrates are twofold. The first is that their
quantum wells have deeper potential height than the quantum wells on InP,
which prevents carrier overflow into the SCH layers and results in larger
optical gain and improvement of temperature characteristics (Ishikawa,
1993; Ishikawa and Suemune, 1994). Conventional strained quantum-well
lasers emitting in the infrared-that is, 1.3 and 1.55pm-grown on InP
substrates suffer from low temperature stability of threshold currents,
primarily due to the carrier overflow from the active region to the SCH
region with relatively low potential height. The carrier overflow into SCH
regions with a high density of states increases the carrier number at lasing
and thus accelerates the Auger recombination process, the rate of which is
proportional to the cube of the carrier density, resulting in the consumption
of carriers. At the same time, large carrier density at lasing greatly increases
internal cavity loss, resulting in low external quantum efficiency.
The second merit of ternary substrates is that high-reflectivity distributed
Bragg reflector mirrors can be grown for vertical cavity surface emitting
lasers. The mirrors consisting of In,,,,Ga,~,,As and Ino,25Alo~7,As on
In,,,,Ga,,,,As substrates give a refractive index difference of 0.4-about
twice that of the combination of InP and InGaAsP (A = 1.2pm) on InP
substrates (Shoji et al., 1997). This opens the way to vertical cavity surface
emitting lasers with a lasing wavelength at 1.3pm, which are the key devices
in future optical interconnections.
The combination of ternary substrate and self-assembled dots holds
promise for 1.3-pm lasers with high-temperature stability, along with alter-
nate-supply growth in metal organic vapor phase epitaxy (Chapter 3)
and the low-growth rate method plus InGaAs covering in molecular
beam epitaxy (MBE). The reason for this is that the self-assembling of
quantum dots is a strain relaxation growth process, and the larger the In
content in substrates, the larger the amount of In introduced in self-
assembled dots.
Figure 8.5(a) is an atomic field microscope image of InAs quantum dots
on the (001) ,Ga,,,,As substrate by MBE. The 450-nm In,,,,Ga,,,,As
buffer layer and then the dots were grown. The substrate temperature was
51OoC, the growth rate was 0.1 monolayer (ML)/sec, and 1.8-ML InAs was
supplied. The dots are slightly smaller and have a larger density than dots
grown on GaAs substrates. Figure 8.5(b) shows the photoluminescence
spectra at 4.2K from dots grown on the (001) In,~,,Ga,,,,As substrate
covered by lOOnm In,,,,Ga,~,,As, and dots grown on the (001) GaAs
8 THELATESTNEWS 333
0.8 0.9 1.o 1.1 1.2 1.3

Wavelength (pm)
FIG. 8.5. (a) Atomic field microscope image of InAs quantum dots on a (001)
In, ,,Ga, 8 3 A on
~ the In,, ,,Ga,,,,,As substrate. (b) Photoluminescence spectra from dots on
GaAs and In,~,,Ga,,,,As substrates at 4.2K. The dots are covered by GaAs on the GaAs
substrate and by In,,,,Ga,,,,,As on the In, ,,Ga,,,,As substrate.
substrate covered by 100-nm GaAs. The spectrum shifted from 1.0pm

on GaAs to 1.14pm on In,,,,Ga,,,,As, which can be attributed to larger In
content in the dots. The spectrum width of the dots on In,,,,Ga,,,,As is
56 meV, which is narrower than that on GaAs.
Increasing the growth temperature and/or reducing the growth rate
334 MITSURUSUGAWARA
ET AL.
enhances the migration of growth atoms, resulting in larger size and longer
wavelength emission. Also, by applying the close-stacking technique de-
scribed in Chapter 2, room-temperature 1.3-pm emission from quantum
dots can be achieved. We are now working on it.
IV. Quantum Dots Emitting at 1.3pm Grown by Low Growth Rates

and with an InGaAs Cap
Using MBE, we grew self-assembled quantum dots emitting at 1.3 pm at

room temperature. This was achieved by reducing the supply rate of InAs
to 0.007 ML/sec, about 1/15 that in regular self-assembled dots (Chapter 2),
and, at the same time, by capping the dots with 8-nm In,,,,Gao,,,As.
An alternative technique for growing 1.3-pm quantum dots is to reduce
the growth rate. Murray and his colleagues (1998) achieved 1.3-pm-emission
quantum dots at room temperature with a very narrow spectrum width of
24meV by MBE. They explained that the thermodynamic equilibrium size
at a given growth temperature is realized by the enhancement of migration
length under low growth rates, The InGaAs cap of dots is also useful for
increasing emission wavelength, since the cap with a lattice constant larger
than GaAs releases the compressive strain of dots more than the GaAs cap
does.
Figure 8.6 is a sample illustration. The growth temperature was 510"C,
the growth rate was 0.007 ML/sec, and the As pressure was 1 x lo-' Torr.
Figure 8.7 shows the planview (a) and the cross-sectional-view (b) trans-
mission electron microscope image. The dots have a diameter of 15 nm with
an area density of 4 x 10" crnp2. The cross-sectional picture shows a dark
region over Stranski-Krastanov islands, which may be due to strain field.
Figure 8.8 shows the room temperature photoluminescence spectrum with
FIG. 8.6. InAs quantum dots covered with 8-nm In,,,,Ga,,,As.

8 THELATESTNEWS 335
FIG.8.7. (a) Plan-view and (b) cross-sectional view transmission electron microscope image
of lnAs quantum dots grown at an extremely slow rate of 0.007 ML/sec and covered with 8-nm
1% ,,GaLl 8 3 A S .
an emission wavelength of 1.3 ym from the ground state. Surprisingly, the

low-density problem with the 1.3-pm dots grown by ALS is solved, showing
a great possibility of lasers emitting at 1.3pm at room temperature.
V. Reduced-Temperature-inducedVariation of Spontaneous Emission

in Alternate Supply (ALS) Quantum Dots Covered by In,,,Gao.,As
We found experimentally that the temperature-induced variation of the

spontaneous emission wavelength is greatly reduced by covering InGaAs
quantum dots with In,,,Ga,.,As instead of GaAs.
336 MITSURUSUCAWARA
ET AL.
1.o 1.1 1.2 1.3 1.4 1.5

Wavelength (pm)
FIG. 8.8. Room-temperature photoluminescence spectrum of InAs quantum dots with an
emission wavelength of 1.3pm from the ground state. The dots were grown at an extremely
slow rate of 0.007ML/sec and covered with 8-nm In,,,,Ga,.,,As.
Figure 8.9 shows the photoluminescence emission energy from the quan-
tum-dot ground state as a function of temperature. The samples are InGaAs
quantum dots grown by ALS on GaAs substrates (Chapter 3). The dots are
covered by GaAs or In,,,Ga,,,As. The inset is the emission spectrum of the
dots covered by In,,,Ga,~,As at 20 K, which shows clear emission lines from
the ground and second states. Unfortunately, we could not detect the
emission from the Ino,,Gao,,As-covered dots above 200 K due to their low
crystal quality. The emission energy decreased as temperature increased in
both samples, as always observed in semiconductors. However, the amount
of energy shift in the In,,,Ga,,,As-covered dots was almost half that of the
GaAs-covered dots. We posit the reason for this unique phenomenon as
follows. First, since the thermal expansion coefficient of In,,,Ga,.,As is
about 94% that of GaAs, the temperature-induced release of compressive
strain, which decreases the interband transition energy, is restricted in dots
covered by In,.,Ga,.,As. Second, the asymmetric structure surrounding the
dots when covered by In,.,Ga,.,As gives rise to a remarkable change in the
8 THELATESTNEWS 337
’-0 *I
*--a- 'I
1 Covered by In,.,G%.,As --a. -c -

-U
0 50 100 150 200 250 300 350

Temperature (K)
FIG. 8.9. Photoluminescence emission energy from the quantum-dot ground state as a
function of temperature. The samples are InGaAs quantum dots grown by ALS on GaAs
substrates and covered with In,,,Ga,,,As and G a s h The inset is the emission spectrum of the
dots covered by InGaAs at 20 K, which shows clear emission lines from the ground and second
states
dot shape as temperature increases. These two factors work to cancel the
temperature-induced reduction in the bandgap of quantum-dot materials.
The results tell us that the temperature-induced shift of emission energy can
be artificially reduced in quantum dots, which is quite attractive for lasers
with their lasing wavelength insensitive to temperature.
References
Ishikawa, H. (1993). A p p l . Phys. Lett. 63. 712.

Ishikawa, H., and Suemune, I. (1994). I E E E Photon. Technol. Lett. 6, 344.
Kusunoki, T., Nakajima, K., Shoji, H., and Suzuki, T. (1996). Mar. Res. Soc. Sj~rnp.Proc. 417,
315.
338 MITSURU
SUCAWARA
ET AL.
Mukai, K., Nakata, Y.,Shoji, H., Sugawara, M., Otsubo, K., Yokoyama, N., and Ishlkawa, H.
(1998). Electron. Lett. 34, 1588.
Murray, R.,Childs, D., Malik, S., Siverns, P., Roberts, C., Hartmann, J-M., Stavrinou, P. (1998).
Nakajima, K., and Kusunoki, T. (1996). J . Crysr. Growth 169, 217.
Nakajima, K., Kusunoki, T., and Otsubo, K. (1997). J . Cryst. Growth 173, 42.
Otsubo, K., Shoji, H., Kusunoki, T., Suzuki, T., Uchida, T., Nishijima, Y., Nakajima, K., and
Ishikawa, H. (1997). Electron. Lett. 33. 1795.
Shoji, H., Otsubo, K., Fujii, T., and Ishikawa, H. (1997). IEEE J . Quanfum Electron. 33, 238.
Shoji, H., Otsubo, K., Kusunoki, T.. Suzuki. T., Uchida, T., and Ishikawa, H. (1996). Jpn. J .
Appl. Phys. 35, L778.
Shoji, H., Uchida, T., Klusunoki, T., Matsuda, M., Kurakake, H., Yamazaki, S., Nakajima, K..
and Ishikawa. H. (1994). IEEE Photon. Technol. Letr. 6, 1170.
Index
Numbers followed by the letter f indicate figures; numbers followed by

the letter t indicate tables.
A Atomic force microscopy (AFM)

closely stacked InAs islands, 134, 135f
a.c. Stark effect, 31
InAs island growth, 121, 123-124, 123f
Access network systems, 320
stacked islands, 145
Active region, 87f
Atomic layer epitaxial growth, 157- 160
in equantum-dot vertical-cavity surface-
Atomic layer epitaxy (ALE), 156
emitting lasers, 279
Auger coefficient. 236-237
in quantum dot lasers, 242, 253
Auger recombination, 28, 84, 95, 332
Alignment of islands
and dephasing, 307
in-plane, 130-132
Auger relaxation process, 210-21 1, 213f
perpendicular, 125- 130
and phonon bottleneck, 235-237
ALS quantum dots, 156
Auger scattering, 300
comparison to columnar SK dots, 179
crystal quality, 164, 198
diameter distribution, 187f
fabrication, 156- 180 B
growth by InAsGaAs sequence, 160-166 Band engineering in quantum well lasers, 6
growth by InGaAs sequence, 166- 176 Bandwidth
growth process, 176-180 large, and quantum dots, 303-306
luminescence spectrum using NSOM, of third-order linear susceptibility ( x ' ~ ) ' ) ,
193-194, 193f 313
solution to low-density problem for Bi-exciton lasing, 85, 98- 106
1.3 pm emission, 334 and quantum dot structure design,
types, 168- 172 102-103
Anisotropy Bi-exciton state, 85
in quantum-dot vertical-cavity surface- Bi-excitons
emitting lasers, 281 binding energy vs. effective mass, 98
of quantum dot shape, 253 formation, 100
Annealing, and photoluminescence spectra, formation and Stark shift, 302-303
229-23 1 model, 97f
Arsenic desorption, 207 and population Inversion, 101
Arsine (ASH,), 158 spontaneous emission and lasing, 98-106
Asymmetric third-order linear susceptibility Binding energy, 49
(x’~’),305, 313-314 Bloch equation, 78-83
339
INDEX
Bloch function, 5 effect on quantum dot lasers, 84, 86-95,

Blue chirp, 297 91
Blue chirp intensity modulator, 296-298 in quantum dots, 84
Bohr radius, 55, 56 recombination lifetime relationship, 214
and density of states, 241 Carrier relaxation/recombination
quantum dot radius relationship, 104 competition between, 23 1-235
Boltzman distribution, 62 vs. temperature, 222-224
Bound-state excitons Carrier trapping, 232
gain generation mechanisms for, 98 Cavity-enhanced spontaneous emission, 74
and population inversion, 78 Cavity-exciton detuning, 75-76
Bragg reflector mirrors, 332 Cavity laser, 28f, 29
Broadening with ALS quantum dots, 164
and indium composition, 191 Cavity loss, 258
and quantum dot size, 191 Cavity polariton, 66, 76-77
Broadening functions, 18, 85, 95-98 Cavity-polariton, 66, 77
Buffer layer composition and quantum Center-of-mass motion, 34, 41-43, 45, 55,
confinement control, 198-201, 199f 62
Chemical etching for narrow ridge
C waveguide structure, 255
Chirping. see Wavelength chirp
Capture cross section, 205 Chirpless high-speed modulation, 295-298
Carrier capture, 213-214 Cladding layer in quantum dot lasers, 255
Carrier distribution characterization, 21 8 Closely stacked islands, 132-139
221, 219f growth process, 137, 137f
Carrier dynamics and size control, 132
in ALS dots, 234-235, 235f size vs. number of layers, 134
effect of thermal treatment, 229-231 Closely stacked quantum dot lasers, 270-
and frequency rolloff, 299 273
and modulation response, 298-299 Closely stacked quantum dots, 132, 156
and quantum dot device application, 294 AnAs, 133-137
Carrier exchange scattering and dephasing, InAs/GaAs, 132-139
307 photoluminescence properties, 137- 139
Carrier number structure, 270f
based on path of distribution, 220-221 Columnar quantum dot lasers, 273-276
from injection vs. optic excitation, 188 with low-threshold current, high-output
Carrier overflow to SCH region, 95 power, 325-328
Carrier relaxation performance, 328
in discrete energy levels, 214-228 Columnar quantum dots, 143-150, 156
and laser performance, 86-95 Confinement, 4
model, 211-214 controllability, 196-201
step-by-step process and potential for quantum wells, Sf, 35, 39
electroluminescence simulation. quantum wells vs. closely stacked dots,
225-227 141
step-by-step process and zero-dimensional in excitons, 140-143
photoluminescence, 224-225 Confinement potential
study by photo/electroluminescence, 21 5- quantum dots, 9
218 quantum wells, 7
in two-level quantum dots, 293 quantum wire, 9
and wavelength chirp, 300 Conjugate wave function, 21
Carrier relaxation lifetime Coulomb effect
INDEX 34 1
interaction among quantum dots, 85 in long-wavelength quantum dot lasers,

on optical gain spectra, 77-83 278
Coulomb enhancement of gain, 80 and luminescence of closely stacked
Coulomb-hole self energy, 35 quantum dots, 271
Coulomb potential and quantum confinement control
between electron and hole, 34 evaluation, 200-201
Coupling constant, 23-24, 57, 60 and zero-dimensional exciton
in microcavity, 69 confinement, 140-143
Creation operator, 32 Differential gain
Crystal quality in quantum dot lasers, 242
and carrier relaxation lifetime, 91 in quantum dots, 245
and laser performance, 268 Differential optical gain, 92
CW lasing, 249, 257, 267 Diirerential quantum efficiency, for
quantum dot lasers, 248, 257-258
Dirac state vectors, 13
D Direct modulation
of quantum dot lasers, 298- 300
Damping factor, 92 and wavelength chirp, 295-296
Decay Discrete energy levels
double-exponential, 215 and carrier relaxation, 21 1-212
of time-resolved photoluminescence interband transitions and
spectra, 217-218 photoluminescence, 186- 187
Defects in InAs islands, 125 and quantum dot device application, 294
Delta function, 241, 243 in quantum dots, 215-218
Density-matrix approach Discrete subbands in quantum dot lasers,
and gain of quantum dots, 244-245 259
to linear optical susceptibility derivation, Dislocation lines, 206
14- 19 Distributed feedback (DFB) lasers, 295
to nonlinear optical susceptibility Double exponential decay curves, 215. 219f
derivation, 19-21 calculated vs. measured, 22 1-225
to optical susceptibility derivation, 12- 14 Double heterostructure lasers, 326
Density of states Dressed excitons, 66
bulk materials, 6-7 Dry etching for narrow ridge waveguide
quantum dots, 3, 5f, 10 structure, 255
and quantum nanostructure dimension, 6 Dynamic characteristics of quantum dot
quantum wells, 8 lasers, 248-249
quantum wire. 9
Dephasing in quantum dots, 307f
Dephasing rate, 12, 104
E
calculation, 308 -3 10
of polarization, 21 Effective-mass approximation
and quantum dot geometry, 311-313 in semiconductor quantum wells. 4-7
for third-order linear susceptibility ( ~ ‘ ~ 9, Effective-mass equations
307 electron-hole system, 36
Diamagnetic energy, 43 excitons, 33, 38
Diamagnetic shift quantum wells, 5, 39-40
of exciton optical absorption, 51 quantum wells under magnetic field.
of exciton resonance in quantum wells, 43 42-43
of exciton spectra due to localization, quantum wells with in-plane confinement,
64-65 40-42
342 INDEX
Electric dipole approximation, 55 cavity effect, 67-73

Electric field, 14 in planar microcavity, 69
Electric wave function Exciton effect, semiconductor optical
bulk materials, 108 response, 12
quantum dots, 110- 1 1 1 Exciton effective-mass equations. see
quantum wells, 109-1 10 Effective mass equations
Electroluminescence Exciton emission spectra
for ALS quantum dots vs. SK dots, temporal evolution, 76
188- 189 Exciton lasing
for quantum dot lasers, 255-256 conditions for, 78
simulation of spectra, 225-227 in wide-gap semiconductor quantum
spectra, 2 15- 2 17 wells, 77-78
spectra and phonon bottleneck, 215-217 in wide-gap semiconductors, 31-32
temperature dependence for quantum dot Exciton optical absorption spectra,
lasers, 262 41-54
Electron-hole plasma, 47 effect of magnetic field compression,
Electronic coupling 51-53, 52f
of quantum dots, 250 as a function of temperature, 53-54,
of stacked dot layers, 272 54f
Electronic states Exciton-photon interactions, 43-47
in quantum dot microcavity, 280f control of, 66
in semiconductor quantum nano- quantum dots, 46-47
structures, 3- 1 1 quantum wells, 45-46
Emission efficiency semiconductor microcavities, 66-77
of ALS dots, 206 Exciton-polari ton
of columnar-shaped quantum dots, 274 dispersion relations, 66
Emission spectra and photon emission direction, 45
broadening and island height, 130 Exciton spontaneous emission
for closely stacked quantum dots, 271 cavity enhanced, 74
multiple-peaked for ALS quantum dots, effect of, 31
187-188 lifetime temperature dependence, 62-63
single island, 138f and localization, 63-66
width (FWHM) for ALS quantum dots mesoscopic enhancement of rate, 47
vs. temperature, 185- 186, 188f in mesoscopic quantum disks, 55,
Emission wavelength and laser 60-62
performance, 269 in microcavity, 68f, 75f
Energy-balance model for island formation, in quantum wells, 55-60
119-121 in quantum wells and photon emission
Energy conservation rule, 25 direction, 45
and carrier relaxation lifetime, 210 reversibility, 77
and two-phonon processes, 212 Excitonic enhancement of gain, 80
Energy-dispersive X-ray microanalysis Excitons
(EDX), 162 binding energy vs. quantum well width,
Energy separation and quantum dot sue, 49, 50f
191 coupling with cavity mode, 71-72
Envelope wave function, 6, 25, 34 decay in planar microcavity, 69
Excited states definition, 32
in self-assembledquantum dot lasers. 252 effect on optical absorption spectra,
temporal variation, 22-23 47 -54
Exciton decay gain generation for hound state, 98
INDEX 343
integrated intensity vs. quantum well Gaussian inhomogeneous broadening, 87-
width, 49, 50f 88
localization vs. spontaneous emission. Ground-state emission energy, and
63-66 quantum dot size, 191
optical properties, 30-32 Growth rate for 1.3 pm emitting quantum
photon emission direction, 45 dots, 333-335
and population inversion, 32
radius vs. quantum well width, 49, 50f
resonance in quantum wells, 31, 39, 41,43
H
state vectors, 32-36
temporal evolution of emission spectra, Harmonic-oscillator type quantum
73 confinement, 189
External modulation scheme, 296-298 Heisenberg equation, 79
Highly lattice-mismatched semiconductors,
118
F Hole, definition, 32
Fabry-Perot lasers, 265,295 Homogeneous broadening. 77.82
Fermi-Dirac distribution, 16-17 effect on lasing emission spectra, 95-98,
Fermi's golden rule, 57, 61, 310 328-331
Fiber probe, 193 origin, 330-331
Fiber to the curb (FTTC) network systems, for single quantum dots, 85, 192
320 for single SK dots, 185
Fiber to the home (FTTH) network
systems, 320
Four-wave mixing, 304, 304f I
Free excitons, 65, 70, 73-74
Frequency rollor InAs islands, 30
and carrier transport, 299 alignment, 131
Full width at half maximum (FWHM) close stacking, 133- 137
linewidth of quantum dots, 156 growth, 121-125
for self-assembled quantum dot lasers, InAs quantum dots, 249
25 1 covered with In,,,,Ga, 8 3 A ~3341
.
and uniformity of ALS dots, 164 Infrared photodetector, 319
InCaAs cap, 333-335
InGaAs quantum dots, 249
G Inhibited spontaneous emission, 74
Inhomogeneous broadening, 53-54,77, 85,
Gain properties, 47-54 96, 106
quantum dot lasers, 242, 244-245 for single SK dot, 186
quantum dots, 258 for SK-mode islands, 128-129
of two-level quantum dots, 289-291 for two-level quantum dots, 293
Gain saturation, 330 Injection current, relationship to carrier
quantum dot lasers, 261-262, 262f density, 246
semiconductor lasers, 30 Integrated intensity
two-level quantum dots, 293 band-gap dependence in quantum wells,
and wavelength jump on columnar-dot 49, 50f
lasers, 327 of exciton resonance, 51-53
Gain spectra, 245, 245f Intensity modulation, 302-303
Gaussian distributions, 42, 53, 87-88 Interband optical transition. 11- 12. 19, 22,
of ALS dot diameters, 161, 187f 186
344 INDEX
Interband optical transition (Contd.) rate equations, 27-30

and electroluminescence emission peaks, simulation with Auger process, 235-237
215 and steady state, 89
in semiconductor quantum nano- Lasers
structures. 11-30 with 1.3 pm wavelength, 332, 333-335
Interband transition matrix active region, 87f
in bulk materials, 111 crystal quality, 201
and carrier relaxation, 212 limiting factors to performance, 267-269
in quantum dots, 112 output power and Auger effect, 236-237
Interdiffusion, 23 1 Lateral confinement potentials, 2
Internal loss, 95 Lattice constant change and confinement
of quantum dot lasers, 257-258 control, 198-201
Intersubband carrier relaxation, 205, 21 1 Lattice mismatch, 196, 198
in time-resolved photoluminescence, 2 13- Lattice-mismatched epitaxy, 120
214 Linear optical susceptibility, 14-19
Intraband relaxation LO phonons, 210-21 1
and relaxation broadening, 245 energy and thermal excitation, 307
and width-of-gain spectrum, 243 Local area cable systems, 320
Island growth, 118 Localized exciton, 65-66
energy balance model, 119- 121 Long-haul cable systems, 320
multiple layers and perpendicular Long-wavelength infrared lasers, 28
alignment, 125-130 Long-wavelength quantum dot lasers,
on prepatterned substrates, 131 - 132 276-279
rates, 122 Longitudinal relaxation, 13, 16
Islands Lorentz distribution, 18, 49, 71, 87
height and emission spectrum Luttinger-Kohn parameter, 51
broadening, 130
in-plane alignment, 130-132
as quantum dots, 156
M
stacking vs. size, 131
Magnetic field, 51, 64
Magneto absorption, 51
K Magneto-optical spectroscopy, 200-201
Many-body effect, 196
k selection rule, 79
Matrix element, 44, 47
k - p band calculation method, 196
-
k p perturbation, 6, 30
Memory duration, persistent hole burning,
315
and transition matrix in bulk materials,
Mesa stripes, 132
111
Mesoscopic enhancement, 31, 47, 104
Kramers-Kronig relation, 290
Mesoscopic region, 77
exciton spontaneous emission in, 60-62
L Metal-organic chemical vapor deposition
(MOCVD), 3, 121, 242
LA phonons, 210-211 in ALS dot growth by InGaAs sequence,
Laser operations 166
closely stacked quantum dots, 272-273 for self-assembled quantum dot
long-wavelength quantum dot lasers, fabrication, 250
276-279 Metal-organic vapor phase epitaxy
quantum-dot vertical-cavity surface- (MOVPE), 156
emitting lasers, 279-282 and quantum dot wafer growth, 190
INDEX 345
Microcavity, 55, 66-77 Nonlinear coefficient gain, 21

in quantum-dot vertical-cavity surface- Nonlinear optical susceptibility, 19-21
emitting lasers, 279 Nonradiative centers, 206t
Microprobe photoluminescence, 192- 196 Nonradiative channel, 231-232
Microscope photoluminescence, 192 Nonradiative recombination, 23, 65
Misfit dislocation of superlattices, I56 ALS quantum dots, 201-207
Modal gain closely stacked SK-mode islands, 144
of closely stacked quantum dots, 272 columnar-shaped quantum dots, 148
current density relationship, 260 and laser performance, 242
Modulation rate for closely stacked quantum dots,
direct, 298-300 273
external, 296-298 Normalized broadening function, 88
intensity. 301-302
in multiple quantum well lasers, 296-298 0
of quantum dot lasers, 267
very high speed and quantum dots, Optical absorption spectra
295-298 caused by excitons, 47-54
Modulation bandwidth, 92-93 intensity experiment, 50-53
and phonon bottleneck, 301 from linear susceptibility, 19
in quantum dot lasers, 242, 249 resonance spectrum profile, 53-54
in SK-mode islands, 129 Optical confinement factor, 28, 89
Molecular beam epitaxy (MBE), 3, 119. 242 Optical data storage, 320
growth of InAs quantum dots, 155 Optical gain
InAs island growth, 121- 122 bulk semiconductor lasers, 29
self-assembled quantum dot fabrication, closely stacked quantum dots, 271
250 Coulomb effect on spectra, 77-83, Slf, 83f
Mott density of bi-excitons, 99, 103 effect of homogeneous broadening, 328-
Multi-mode oscillation, 265 33 1
Multiple gas injectors in MOVPE system, homogeneous broadening vs.
166 inhomogeneous broadening in
Multiple layer growth single-dots, 329
in islands, 125-130 from linear susceptibility, 19
in SK dots vs. ALS dots, 180 from nonlinear susceptibility, 21
in type A and type B quantum dots, quantum dots, 96
172- 176 relationship with current density, 247
Multiple peak emission spectra for ALS semiconductors, 27
quantum dots, 187, 1 8 8 Optical image processing, 320
Multiple quantum well (MQW) lasers, Optical interconnections, 320
296-298 Optical mode distribution, 26
Multiplication factors (MPFs), 314 Optical nonlinearity, 2
applications, 19
N Optical response categories for
semiconductors, 12
Near-field optical microscopy system Optical susceptibility, 12-14
(NSOM), 193, 193f linear, 14-19
ALS spectra dependence on excitation nonlinear, 19-21
power, 194-196 Optical switches, 2
Nondegenerate four wave mixing, 21 Optically-coupled quantum dots, 250
Nonexponential decay curves of time- Optoelectronics, 320
resolved photoluminescence, 217 Oscillation, 76-77
346 INDEX
Oscillator strength. 98. 106 reduction of temperature induced shift,

effect on quantum dot lasers, 104 336-337
Output power single island layer, 138f
and carrier relaxation lifetime in SK dots vs. excitation power, 214f
quantum dot lasers, 91 time-resolved, 217-225
columnar-shaped quantum dot lasers. TMIDMEA grown ALS dots, 160, 161f
274,325-328 wavelength tunability in ALS dots, 164
vs. current in columnar-dot lasers, 326f wavelength vs. ALS dot growth
quantum dot lasers, 90f temperature, 168, 169f
semiconductor lasers. 29 wavelength vs. quantum dot size, 164
Photon lifetime, 22, 74-75
Photons
P in quantum dot microcavity, 280f
reabsorption in microcavity, 71
Pauli blocking, 188- 189 wave vector, 72
Pauli's exclusion principle, 88 Plasma dispersion in quantum dots, 290
Periodic boundary condition, 40 Polarization, 14, 16, 80
Perpendicular stacking, 119 bandwidth and dephasing rate, 21
Persistent hole burning memory, 3 14-319, and interband optical transition, 14
316f, 317f of quantum-dot vertical-cavity surface-
calculated vs. measured results, 318f emitting lasers, 281
Phonon bottleneck, 84, 189, 209-238, 298, Population inversion, 27
300,301 in bi-excitons, 99, 101
description, 17, 209 in bound-state excitons, 78
and internal quantum efficiency of in excitons, 32
quantum dot lasers, 258 Pulse jet epitaxy (PJE), 157
and laser performance, 269 and ALS dot growth by InAsGaAs
Phonon scattering, 17, 49, 54, 57 sequence, 160
Photoluminescence and laser devices. 164
after annealing, 229-231
ALS quantum dots, 277f
ALS quantum dots vs. SK dots, 185- 186
Q
closely stacked islands, 137-138 Quantized energies, controlling, 196-200
closely stacked quantum dots, 271f Quantum-box laser, 241
columnar-shaped quantum dots, 148,273 Quantum-confined Stark effect, 2, 31, 95
columnar-shaped quantum dots vs. SK Quantum confinement, 118, 156
dots, 2741 controllability, 196-201
and discrete state interband transitions, effect on exciton radius, 49
186- 187 harmonic oscillator type, 189
excitation in ALS quantum dots vs. SK potential evaluation by diamagnetic
dots, 187 shifts, 200-201
InAs quantum dots with 1.3 pm emission, Quantum disks, 32, 55
335f mesoscopic and spontaneous photon
island growth, 128-129, 130f emission, 77
linewidth of quantum type A vs. type B spontaneous photon emission in
dots, 171 microcavity. 66-77
and nonradiative recombination channel Quantum-dot infrared photodetector, 294
evaluation, 201-207 Quantum dot injection layers, 243
quantum dots, 10, 1If Quantum dot lasers, 83-107
quantum wells, 63, 65 characteristics, 255-267
INDEX 347
Coulomb effect, 84 photoluminescence spectra, 10, 1 1 f

direct modulation, 298-300 properties, 288-294
efficency and carrier relaxation lifetime, selective growth on patterned substrate,
91 242
emerging technologies, 269-282 self-assembly process, 242
emission efficiency, 257 single dot optical properties, 85
on GaAs substrates, 95 size control by close stacking, 139
hierarchy, 94, 94f third-order nonlinear susceptibility, 291-
on I n P substrates, 95 292
lasing characteristics, 256-262 as a two-level system, 288-293
mirror loss dependence, 257-258 uniformity and laser performance, 268
output vs. current, 25% and very high speed light modulation,
properties, 243-249 295-298
schematic, 246f wave function, 10
self-assembly process, 84 Quantum nano-structures
spectral linewidth, 242, 249 density of states, 4, 4f
speed of carrier relaxation, 84 optical response, 1 1-30
temperature dependence of Quantum well infrared photodetector
characteristics, 262-267 (QWIP), 319
wavelength chirp, 300, 301f Quantum-well lasers, 2
Quantum dot layers, optical coupling, 250 wavelength chirp, 300, 301f
Quantum dot optical devices, 320-321 Quantum well thickness
Quantum dot size comparison t o SK mode wetting layer,
and excited states: 252 171, 176
and laser performance, 268 Quantum wells, 2, 7-8
uniformity, 251 energy calculation by effective-mass
Quantum dots approximation, 4-7
approaches to fabrication, 242 spontaneous emission. 55-66
atomic layer epitaxial growth, 152- 160 Quantum wire, 140- 143
bleaching, 316 density of states, 9
and Coulomb interactions, 85 fabrication, 1 1 8
density and laser performance, 268 wave function, 9
density in SK dots, 250 Quantum-wire lasers, 300. 301f
density of states, 10
direct modulation, 295-296
R
electroluminescence properties, 185- 189
emitting at 1.3 pm wavelength, 333-335 Rabi frequency, 72
fabrication, 118 Radiative emission efficiency of ALS dots,
on InGaAs substrates, 331-333 201-207
intensity modulation, 302-303 Radiative lifetime, 247
interaction and homogeneous Random carrier capturing, 21 5--216
broadening, 33 1 Random initial occupation (RIO), 217-21 8
light emission from discrete energy states, and decay curves of time-resolved
185-196 photoluminescence, 225
as nonlinear medium, 303-314 Rate equations
optical characterization, 183-207 carrier-photon system, 87
optical nonlinearity, 303-306 laser operations, 27-30
optimization of structure, 250 photons, 292
photoluminescence excitation, 185- I89 Reading signal, 315, 317f
photoluminescence properties, 185- 189 Reciprocal space map, 198
INDEX
Recombination and dephasing, 307 self-organization on disk structures, 242

Recombination lifetimes spontaneous formation in SK mode, 242
for ALS quantum dots, 203, 203f structure of ALS type, 163
Red chirp, 297 Self-generated electro-optic effect devices
in quantum dots, 290 (SEEDS), 2, 31
Reduced effective mass, 40 Self-limiting growth for ALS quantum dots,
Reflection high-energy electron diffraction 158-159, 167
(RHEED) Semiconductor Bloch equation, 78-83
InAs closely stacked islands, 133, 134f Semiconductor lasers
InAs island growth, 121, 122-123. 122f edge-emitting type, 27
stacked island growth, 144, 145f effect of nonlinearity, 19
Refractive index dispersion principle of operation, 27
and external modulation, 297 Semiconductor quantum nano-structures
and quantum dot device application, 294 electronic states, 3-1 1
quantum well, 290f fabrication, 118
Relative motion, 35-36, 41, 43 history of development, 1-3
Relaxation oscillation frequency, 92 interband optical transition, 11-30
quantum dot lasers, 248 optical susceptibility, 12-21
Resonant excitation, 21 1-212 spontaneous emission rate, 21-26
Resonant modes in microcavity, 280 Semiconductor superlattice, 2
Retarded carrier relaxation. see Phonon Separate confinement heterostructure
bottleneck (SCH) layer, 86
Ridge-waveguide structure, 253-254, 253f in columnar-shaped quantum dot lasers,
Rolloff, 92, 94, 299 326
Shottkey contact structure, 316
Single quantum dot
S homogeneous broadening of optical gain,
85, 328-331
Saha equation, 100 optical properties, 85
Schrodinger equation, 56 SK-mode islands, 128- 129, 133
time-dependent, 13, 22 Sommerfeld factor, 48, 82
Second quantization, 32,43 Source materials
Selection rule, for wave vectors, 43 for ALS quantum dots, 158
Self-assembled growth, 119 amount and ALS dot growth, 71
Self-assembled quantum dot lasers role in ALS dot growth, 179
electroluminescence properties, 255 -256 role in columnar SK dots, 179
fabrication, 249-255 Spectral hole-burning effect, 305
modulation characteristics, 267 Spectral hole shape, 31 8-319
optical properties, 251 -253 Spin, 3 10
temperature dependence, 262-267 in bi-excitons, 98
Self-assembled quantum dots, 2 in carrier number rate equations, 226
alternative supply in atomic layer epitaxy and optical susceptibility, 17
(ALE), 242 in quantum dots, 102
fabrication, 253-255 Spontaneous emission
fabrication by SK mode, 250 rate for electron-hole transition, 21-26
grown by metal organic chemical vapor wavelength variation, 336-337
deposition (MOCVD), 155-181 Spontaneous emission of bi-excitons, 98-
InAsGaAs sequence growth. 161 106
molecular beam epitaxial growth, 117- Spontaneous emission of photons, 21-26
151 cavity-enhanced, 74
INDEX
and exciton localization, 63-66 quantum dot lasing characteristics, 262-

inhibited, 74 267
lifetime variation in quantum disks vs. recombination lifetimes, 229-235, 233f
quantum wells, 55-62 relaxation lifetimes, 229-235. 233f
in mesoscopic disks, 55-66 Ternary substrates, 332
in quantum dot microcavity, 66-77 Thermal broadening, 53-54
in quantum wells, 55-66 Thermal excitation and dephasing. 307
reversible, 77 Thermal treatment and phonon bottleneck,
temperature dependence, 62-63 229-235
Stacked islands, 148-150 Thermal velocity of intersubband carrier,
Stark effect, 31 205
Stark shift, 302-303 Third-order linear susceptibility (I[.’’),20,
State vectors, 32-36 305
Stimulated emission, 12, 27 amplitude, 3 I3
Strain, 196 asymmetry, 305, 313-314
Strain energy, 196, 198 bandwidth, 313
Strain fields and island perpendicular calculation, 306-31 I
alignment, 127-128 multiple quantum wells, 306f
Strain relaxation, 198 and quantum dot device application,
Strained-layer epitaxial growth, 121 294
Strained quantum-well laser two-level quantum dots, 29 1-292
design, 30 Third-order nonlinear coefficient, 20--21, 30,
modulation, 95 89
Stranski-Krastanov (SK) mode, 3-4, 87, Three-dimensional quantum confinement.
155 241
energy-balance model for island Threshold current
formation, 119-121 columnar-shaped quantum dot lasers.
InAs island growth, 121-125 273-274, 325-328
island growth, 120- 121 quantum dot lasers, 242, 246-248
multiple layer growth, 125- 130 quantum-dot vertical-cavity surface-
and quantum dot growth, 118-119 emitting lasers, 279-280
for self-assembled quantum dot semiconductor lasers, 30
fabrication, 250 Threshold current density, 256
Strong coupling, 66-67, 71, 74 Threshold current in quantum dot lasers
Superlattices, 156 and carrier relaxation lifetime, 9 1
Supply cycle number and quantum temperature dependence, 265-267, 265f
confinement control, 196- 198, 199f Threshold gain, 29, 89
Surface density, 294 quantum dot lasers, 247
Susceptibility. see Optical susceptibility Time-dependent Schrodinger equation, 13,
22,210
T Time division multiplexing (TDM), 320
Time-resolved photoluminescence, 217-225
Temperature dependence carrier capture/relaxation process. 21 3-
ALS emission spectra width, 185- 186, 214
188f T M G a (trimethylgallium), 158
ALS quantum dot growth, 159, 160f, 171 TMIDMEA, 157, 158
ALS self-limiting growth, 167 TMIn (trimethylindium), 158
InAs island growth, 124 Transition matrix, 4f
photoluminescence wavelength for ALS Transmission electron microscopy (TEM)
dots, 168, 169f closely stacked InAs islands, 270f
350 INDEX
Transmission electron microscopy (Conrtf.) Wave function

closely stacked islands structure, 135- bulk materials, 6
136, 136f exciton-photon system, 56
columnar-shaped quantum dots, 146. exciton state, 33
147f, 274f normalized, of quantum wells, 7
InAs island growth, 121, 124-125, 126f quantum dots, 10
InAs slow growth rate quantum dots, quantum wells, 7
335f quantum wire, 9
quantum dots, 4f Wave vector, 22, 35, 66
quantum wells, 4f optical mode, 13, 14
triple-stacked quantum dots, 251f and selection rule for center-of-mass
type A quantum dots, 171, 172f motion, 45
Transverse relaxation, 13, 16 and selection rule for quantum well
Tunneling, 212 in-plane component, 55
Two-band model Waveguide structure, 248
for Bloch equation derivation, 79 Wavelength
for linear optical transition, 14 behaviors in quantum dot lasers, 259f
for nonlinear optical transition, 19 jump in columnar-dot lasers, 327
Type A quantum dots, 168-171 in long-wavelength quantum dot lasers,
growth process vs. SK islands, 176 279
multiple layer growth, 172- 176 Wavelength chirp, 93, 95, 291
Type B quantum dots, 168-171 in quantum dot lasers, 249
multiple layer growth, 172- 176 in quantum nano structures, 300, 301f
Wavelength conversion. 12, 304
U efficiency, 305f
Wavelength division multiplexing (WDM),
Undersea cable systems, 320 304, 320
Weak coupling, 67, 71, 74
V Wetting layer, 4. 250
absence in ALS dots, 163, 278
Vector potential, 13
in columnar-shaped quantum dots, 179
Vent-and-run configuration, 167
comparison to quantum well thickness,
Vertical-cavity surface-emitting lasers
171. 176
(VCSELs), 27, 156, 279-282
SK islands vs. ALS dots, 161, 171
with 1.3 pm wavelength, 332
and strain energy release, 155
schematic, 281f
Wide-gap lasers, 31 -32
Vertically-stacked dots
and dot density, 250
Z
W Zeeman term, 43
Zero-dimensional confinement, 140- 143
Wafer mapping, 190-192
Contents of Volumes in This Series
Volume 1 Physics of 111-V Compounds

C. Hilsuni, Some Key Features of 111-V Compounds
F . Bassani, Methods of Band Calculations Applicable to 111-V Compounds
E . 0. Kune, The k - p Method
V; L. Bunch-Brueuich, Effect of Heavy Doping on the Semiconductor Band Structure
D. Long, Energy Band Structures of Mixed Crystals of 111-V Compounds
L. M . Roth and P. N. Argyres, Magnetic Quantum Effects
S. M. Puri und T. H. Gebulle, Thermomagnetic Effects in the Quantum Region
W. M. Becker, Band Characteristics near Principal Minima from Magnetoresistance
E. H . Putley, Freeze-Out Effects, Hot Electron Ellects, and Submillimeter Photoconductivity
in InSb
H . Weiss, Magnetoresistance
B. Ancker-Johnson, Plasma in Semiconductors and Semimetals

M. G. Hollund, Thermal Conductivity
S. I. Novkovu, Thermal Expansion
U. Piesbergen, Heat Capacity and Debye Temperatures
G. Giesecke, Lattice Constants
J, R. Dmhhle, Elastic Properties
A. Li. Mac Rue und G. W Gobeli, Low Energy Electron Diffraction Studies
R. Lee Mieher, Nuclear Magnetic Resonance
B. Goldstein, Electron Paramagnetic Resonance
T. S. Moss, Photoconduction in I l l - V Compounds
E. Antoncik und J. Trruc, Quantum Efficiency of the Internal Photoelectric Effect in InSb
G. W. Gobeli and I. G. Allen, Photoelectric Threshold and Work Function
P. S. Pershan, Nonlinear Optics in 111-V Compounds
M . Gershenzon, Radiative Recombination in the 111-V Compounds
F. Stern, Stimulated Emission in Semiconductors
351
352 CONTENTS IN THISSERIES
OF VOLUMES
Volume 3 Optical of Properties 111-V Compounds

M. Huss, Lattice Reflection
W. G. Spitzer, Multiphonon Lattice Absorption
D. L. Stierwalt and R. F. Potter, Emittance Studies
H. R. Philipp and H. Ehrenveich, Ultraviolet Optical Properties
M. Cardonu, Optical Absorption above the Fundamental Edge
E. J. Johnson, Absorption near the Fundamental Edge
J. 0. Dimmock, Introduction to the Theory of Exciton States in Semiconductors
B. Lux and J. G. Mavroides, Interband Magnetooptical Effects
H. Y. Fan, Effects of Free Carries on Optical Properties
E. D. Palik and G. B. Wright, Free-Carrier Magnetooptical Effects
R. H. Buhe, Photoelectronic Analysis
B. 0. Seruphin and H. E. Bennett, Optical Constants

N. A. Goryunova. A . S. Borschevskii, and D. N. Tretiakov, Hardness
N. N. Sirota, Heats of Formation and Temperatures and Heats of Fusion of Compounds A"'BV
D. L. Kendall, Diffusion
A. G. Chynoweth, Charge Multiplication Phenomena
R. W. Keyes, The Effects of Hydrostatic Pressure on the Properties of 111-V Semiconductors
L. W. Aukerman, Radiation Effects
N. A. Goryunova, F. P. Kesamanly, and D. N . Nasledov, Phenomena in Solid Solutions
R. T. Bate, Electrical Properties of Nonuniform Crystals
Volume 5 Infrared Detectors

H. Levinstein, Characterization of Infrared Detectors
P. W Kruse, Indium Antimonide Photoconductive and Photoelectromagnetic Detectors
M. B. Prince, Narrowband Self-Filtering Detectors
I. Melngalis and T. C. Harman, Single-Crystal Lead-Tin Chalcogenides
D. Long and J. L. Schmidt, Mercury-Cadmium Telluride and Closely Related Alloys
E. H. Putley, The Pyroelectric Detector
N. B. Stevens, Radiation Thermopiles
R. J. Keyes and T. M. Quist, Low Level Coherent and Incoherent Detection in the Infrared
M. C. Teich, Coherent Detection in the Infrared
F R. Arums, E. W. Surd, B. J. Peylon. undF. P. Pace, Infrared Heterodyne Detection with
Gigahertz IF Response
H. S. Sommers, Jr., Macrowave-Based Photoconductive Detector
R. Sehr und R. Zuleeg, Imaging and Display
Volume 6 Injection Phenomena

M. A. Lampert and R. B. Schilling, Current Injection in Solids: The Regional
Approximation Method
R. Williams, Injection by Internal Photoemission
A . M. Barnett. Current Filament Formation
CONTENTS
OF VOLUMES
IN THISSERIES 353
R. Baron and J. W Mayer, Double lnjection in Semiconductors

W. Ruppel, The Photoconductor-Metal Contact
Volume 7 Application and Devices

Part A
J. A . Copeland and S. Knight, Applications Utilizing Bulk Negative Resistance
F. A. Paduvuni, The Voltage-Current Characteristics of Metal-Semiconductor Contacts
P. L. Hower, W. W. Huoper, B. R. Cairns, R. D. Fairman, and D. A. Tremere, The GaAs
Field-Effect Transistor
M. H. White, MOS Transistors
G. R. Antell, Gallium Arsenide Transistors
T. L. Tansley, Heterojunction Properties
Part B
T. Misawu,lMPATT Diodes
H. C. Okean, Tunnel Diodes
R. B. Campbell and Hung-Chi Chang, Silicon Junction Carbide Devices
R. E. Enstrom, H. Kressel, and L. Krassnur, High-Temperature Power Rectifiers of GaAs, -.P,
Volume 8 Transport and Optical Phenomena

R. 1. Stirn, Band Structure and Galvanomagnetic Effects in 111-V Compounds with Indirect
Band Gaps
R. W. Ure, Jr., Thermoelectric Effects in IIILV Compounds
H. Piller, Faraday Rotation
H. Barry Bebb and E. W. Williams,Photoluminescence I: Theory
E. W. Williams and H. Barry Bebb, Photoluminescence 11: Gallium Arsenide
Volume 9 Modulation Techniques

B. 0. Seruphin, Electroreflectance
R. L. Aggarwal, Modulated Interband Magnetooptics
D. F. Blossey and Paul Handler, Electroabsorption
B. But=, Thermal and Wavelength Modulation Spectroscopy
I. Balslev, Piezopptical Effects
D. E. Aspnes and N . Bortka, Electric-Field Effects on the Dielectric Function of Semiconductors
and Insulators
Volume 10 Transport Phenomena

R. L. Rhode, Low-Field Electron Transport
J. D. Wiley, Mobility of Holes in 111-V Compounds
C. M. Wove and G. E. Stillmun, Apparent Mobility Enhancement in Inhomogeneous Crystals
R. L. Petersen, The Magnetophonon Effect
354 CONTENTS I N THISSERIES
OF VOLUMES
Volume 11 Solar Cells

H . J. Hovel, Introduction; Carrier Collection, Spectral Response, and Photocurrent; Solar
Cell Electrical Characteristics; Efficiency;Thickness; Other Solar Cell Devices; Radiation
Effects; Temperature and Intensity; Solar Cell Technology
Volume 12 Infrared Detectors (11)

W. L. Eisemun, J. D. Merriam, and R. E Porrer, Operational Characteristics of Infrared
Photodetectors
P. R. Eraft, Impurity Germanium and Silicon Infrared Detectors
E. H . Putley, InSb Submillimeter Photoconductive Detectors
G. E. Stillman, C. M. Wolfe, and J. 0.Dimmock. Far-Infrared Photoconductivity in High
Purity GaAs
G. E. Stillman and C. M. Wolfe, Avalanche Photodiodes
P. L. Richards, The Josephson Junction as a Detector of Microwave and Far-Infrared
Radiation
E. H. Putley, The Pyroelectric Detector -An Update
Volume 13 Cadmium Telluride

K. Zanio, Materials Preparations; Physics; Defects; Applications
Volume 14 Lasers, Junctions, Transport

N. Holonyak, Jr. and M. H . Lee, Photopumped 111-V Semiconductor Lasers
H . Kressel and J . K. Butler, Heterojunction Laser Diodes
A Van der Ziel, Space-Charge-Limited Solid-state Diodes
P. J. Price, Monte Carlo Calculation of Electron Transport in Solids
Volume 15 Contacts, Junctions, Emitters

E. L. Sharma, Ohmic Contacts to 111-V Compounds Semiconductors
A. Nussbaum, The Theory of Semiconducting Junctions
J. S. Escher, NEA Semiconductor Photoemitters
Volume 16 Defects, (HgCd)Se, (HgCd)Te

H. Kressel, The Effect of Crystal Defects on Optoelectronic Devices
C. R. Whitsett, J. G. Broerman. and C. J. Summers, Crystal Growth and Properties of
Hg, _xCd,Se alloys
M. H . Weiler, Magnetooptical Properties of Hgl xCd,Te Alloys
P. W. Kruse and J. G. Ready, Nonlinear Optical Effects in Hg, -xCd,Te
Volume 17 CW Processing of Silicon and Other Semiconductors

J. F. Gibbons, Beam Processing of Silicon
A. Lietoila, R. E. Gold, J. F. Gibbons, and L. A. Chrisrel, Temperature Distributions
and Solid Phase Reaction Rates Produced by Scanning CW Beams
CONTENTS IN THISSERIES
OF VOLUMES 355
A. Leitoilu und J. F. Gibbons, Applications of CW Beam Processing to Ion Implanted

Crystalline Silicon
N . M . Johnson, Electronic Defects in CW Transient Thermal Processed Silicon
K. F. Lee, T. J. Stultz, und J. F. Gibbons, Beam Recrystallized Polycrystalline Silicon:
Properties, Applications, and Techniques
T. Shibutu. A. Wukitu, T. W. S i p o n . und J. F Gibbons, Metal-Silicon Reactions and
Silicide
Y. I. Nissim and J. F Gibbons, CW Beam Processing of Gallium Arsenide
Volume 18 Mercury Cadmium Telluride

P. W. Kruse, The Emergence of (Hg, _.,Cd,)Te as a Modern Infrared Sensitive Material
H. E. Hirsch, S. C. Liung, and A. G. White, Preparation of High-Purity Cadmium. Mercury,
and Tellurium
W. F H. Micklethwuite, The Crystal Growth of Cadmium Mercury Telluride
P. E. Petersen, Auger Recombination in Mercury Cadmium Telluride
R. M. Broudy und V J. Muzurczyck, (HgCd)Te Photoconductive Detectors
M. B. Reine, A . K. Soud, and T. J. TredwI/, Photovoltaic Infrared Detectors
M. A. Kinch, Metal-Insulator-Semiconductor Infrared Detectors
Volume 19 Deep Levels, GaAs, Alloys, Photochemistry

G. F. Neurnurk und K. Kosui, Deep Levels in Wide Band-Gap 111-V Semiconductors
D. C. Look, The Electrical and Photoelectronic Properties of Semi-Insulating GaAs
R. F. Brebrirk, Ching-Huu Su, and Pok-Kni Liuo, Associated Solution Model for Ga-In-Sb
and Hg-Cd-Te
Y. Yu. Gurevich und Y. V. Pleskon, Photoelectrochemistry of Semiconductors
Volume 20 Semi-Insulating GaAs

R. N . Thomas, H. M. Hobgood3 G. W. Eldridge, D. L. Burreft. T. T. Braggins. L. B. Tu, and
S. K. Wung, High-Purity LEC Growth and Direct Implantation of GaAs for Monolithic
Microwave Circuits
C. A. Stolte, Ion Implantation and Materials for GaAs Integrated Circuits
C. G. Kirkputrick. R. T. Chen. D. E. Holrnc)s, P. M. Asbeck, K. R. Elliott, R. D. Fuirrnan. und
J. R. Oliver, LEC GaAs for Integrated Circuit Applications
J. S. Blukemore and S. Ruhimi, Models for Mid-Gap Centers in Gallium Arsenide
Volume 21 Hydrogenated Amorphous Silicon

Part A
J. I. Punkove, Introduction
M. Hirose, Glow Discharge; Chemical Vapor Deposition
Y. Uchida, di Glow Discharge
T. D. Moustukus, Sputtering
I. Yumudu, Ionized-Cluster Beam Deposition
B. A. Scott, Homogeneous Chemical Vapor Deposition
356 CONTENTS
OF VOLUMESIN THISSERIES
F. J. Kampas, Chemical Reactions in Plasma Deposition

P. A. Longeway, Plasma Kinetics
H. A. Weakliem, Diagnostics of Silane Glow Discharges Using Probes and Mass Spectroscopy
L. Gluttman, Relation between the Atomic and the Electronic Structures
A. Chenevas-Paule, Experiment Determination of Structure
S. Minomura, Pressure Effects on the Local Atomic Structure
D. Adler, Defects and Density of Localized States
Part B
J. I. Pankove, Introduction
G. D. Cody, The Optical Absorption Edge of a-Si: H
N. M. Amer and W. 5.Jackson, Optical Properties of Defect States in a-Sk H
P. J. Zanzucchi, The Vibrational Spectra of a-Si: H
Y. Hamakawa, Electroreflectance and Electroabsorption
J. S. Lannin, Raman Scattering of Amorphous Si, Ge, and Their Alloys
R. A. Street, Luminescence in a-Si: H
R. S. Crandall, Photoconductivity
J. Taur, Time-Resolved Spectroscopy of Electronic Relaxation Processes
P. E. Vanier, IR-Induced Quenching and Enhancement of Photoconductivity and Photo
luminescence
H. Schade, Irradiation-Induced Metastable Effects
L. Ley, Photoelectron Emission Studies
Part C
J. D. Cohen, Density o f States from Junction Measurements in Hydrogenated Amorphous
Silicon
P. C. Taylor, Magnetic Resonance Measurements in a-Si: H
K. Morigaki, Optically Detected Magnetic Resonance
J. Dresner, Carrier Mobility in a-Si: H
T. Tiedje, Information about band-Tail States from Time-of-Flight Experiments
A. R. Moore, Diffusion Length in Undoped a-Si: H
W. Beyer a i d J . Uverhqf, Doping Effects in a-Si: H
H. Fritzche, Electronic Properties of Surfaces in a-Si: H
C. R. Wronski, The Staebler-Wronski Effect
R. J. Nernanich, Schottky Barriers on a-Si: H
B. Abeles and T. Tiedj,, Amorphous Semiconductor Superlattices
Part D
D. E. Carlson, Solar Cells
G. A. Swartz, Closed-Form Solution of ILV Characteristic for a a-Si: H Solar Cells
I. Shimizu, Electrophotography
S. Ishioka, Image Pickup Tubes
OF VOLUMES 351
P. G. LeComher and W. E. Speur, The Development of the a-Si: H Field-Effect Transistor and
Its Possible Applications
D. G. Ast, a-Si: H FET-Addressed LCD Panel
S. Kaneko, Solid-state Image Sensor
M . Mutsuinuru, Charge-Coupled Devices
M. A . Bosch, Optical Recording
A . D'Amico and G. Fortunato, Ambient Sensors
H. Kukimoto, Amorphous Light-Emitting Devices
R. J. Phelun, Jr., Fast Detectors and Modulators
J. I. Punkove, Hybrid Structures
P. G. LeComher, A . E. Owen, W. E. Speor, J. Hajro, and W. K. Choi, Electronic Switching in
Amorphous Silicon Junction Devices
Volume 22 Lightwave Communications Technology

Part A
K. Nukfljima, The Liquid-Phase Epitaxial Growth of InGaAsP
W. T. Tsang, Molecular Beam Epitaxy for 111-V Compound Semiconductors
G. B. Stringfellow, Organometallic Vapor-Phase Epitaxial Growth of 111- V Semiconductors
G. Beuchet, Halide and Chloride Transport Vapor-Phase Deposition of InGaAsP and GaAs
M. Razeghi, Low-Pressure Metallo-Organic Chemical Vapor Deposition of
Ga,In, -,ASP, - y Alloys
P. M . Petrqg Defects in 111-V Compound Semiconductors
Part B
J. P. van der Z i d , Mode Locking of Semiconductor Lasers
K. Y. Luu und A . Yariv, High-Frequency Current Modulation of Semiconductor Injection
Lasers
C. H. Henry, Special Properties of Semiconductor Lasers
Y. Suemutsu, K. Kishino, S. Arai, and F. Koyctrm, Dynamic Single-Mode
Semiconductor Lasers with a Distributed Reflector
W. T. Tsnng, The Cleaved-Coupled-Cavity (C') Laser
Part C
R. J. Nelson und M. K. Duttu, Review of InGaAsP InP Laser Structures and Comparison of
Their Performance
N . Chinone and M. Nukumuru, Mode-Stabilized Semiconductor Lasers for 0.7-0.8- and
1.1-1.6-pm Regions
Y. Horikoshi, Semiconductor Lasers with Wavelengths Exceeding 2 pm
B. A . Dean and M. Dixon, The Functional Reliability of Semiconductor Lasers as Optical
Transmitters
R. H. Suul, T. P . Lee, and C. A. Burus, Light-Emitting Device Design
C. L. Zipfel, Light-Emitting Diode-Reliability
T. P. Lee and T. Li, LED-Based Multimode Lightwave Systems
K Ogaivu, Semiconductor Noise-Mode Partition Noise
358 CONTENTS
Part D
F. Cupusso, The Physics of Avalanche Photodiodes
T. P. Peursull and M. A. Polluck, Compound Semiconductor Photodiodes
T. Kunedu, Silicon and Germanium Avalanche Photodiodes
S. R. Forresl, Sensitivity of Avalanche Photodetector Receivers for High-Bit-Rate Long-
Wavelength Optical Communication Systems
J. C. Cumpbell, Phototransistors for Lightwave Communications
Part E
S. Wung, Principles and Characteristics of Integrable Active and Passive Optical Devices
S. Murgufit und A. Yuriv, Integrated Electronic and Photonic Devices
T. Mukui, Y. Yumamoto, and T. Kimuru, Optical Amplification by Semiconductor
Lasers
Volume 23 Pulsed Laser Processing of Semiconductors

R. F. Wood, C. W. White. and R. T. Young, Laser Processing of Semiconductors: An Overview
C. W. White, Segregation, Solute Trapping. and Supersaturated Alloys
G. E. Jellison. Jr., Optical and Electrical Properties of Pulsed Laser-Annealed Silicon
R. F. Wood and G. E. Jeifison, Jr., Melting Model of Pulsed Laser Processing
R. F. Wood and I? W. Young, Jr., Nonequilibrium Solidification Following Pulsed Laser
Melting
D. H. Lowndes and G. E. Jellison, Jr., Time-Resolved Measurement During Pulsed Laser
Irradiation of Silicon
D.M. Zebner, Surface Studies of Pulsed Laser Irradiated Semiconductors
D.H. Lowndes, Pulsed Beam Processing of Gallium Arsenide
R. B. James, Pulsed CO, Laser Annealing of Semiconductors
R. T. Young and R. I? Wood, Applications of Pulsed Laser Processing
Volume 24 Applications of Multiquantum Wells, Selective Doping,

and Superlattices
C. Weisbuch, Fundamental Properties of 111-V Semiconductor Two-Dimensional Quantized
Structures: The Basis for Optical and Electronic Device Applications
H. Morkoc and H. Unlu, Factors Affecting the Performance of (Al, Ga)As/GaAs and
(Al,Ga)As/InGaAs Modulation-Doped Field-Effect Transistors: Microwave and Digital
Applications
N . T. Linh, Two-Dimensional Electron Gas FETs: Microwave Applications
M. Ahe rt af., Ultra-High-speed HEMT integrated Circuits
D. S. Chemlu, D.A. B. Miller, and P. W. Smith, Nonlinear Optical Properties of Multiple
Quantum Well Structures for Optical Signal Processing
I.: Cupusso, Graded-Gap and Superlattice Devices by Band-Gap Engineering
W. T. Tsung, Quantum Confinement Heterostructure Semiconductor Lasers
G. C. Oshourn et ul., Principles and Applications of Semiconductor Strained-Layer
Superlattices
ok VOLUMESI N THISSERIES
CONTENTS 359
Volume 25 Diluted Magnetic Semiconductors

W. Giriar and J. K. Furdynu, Crystal Structure, Composition, and Materials Preparation of
Diluted Magnetic Semiconductors
W. M. Becker, Band Structure and Optical Properties of Wide-Gap Af'xMn,B,, Alloys at
Zero Magnetic Field
S. Oseroff and P. H. Keesom, Magnetic Properties: Macroscopic Studies
T. Giebultowicz and T. M. Holden, Neutron Scattering Studies of the Magnetic Structure and
Dynamics of Diluted Magnetic Semiconductors
J . Kossur, Band Structure and Quantum Transport Phenomena in Narrow-Gap Diluted
Magnetic Semiconductors
C. Riquaux, Magnetooptical Properties of Large-Gap Diluted Magnetic Semiconductors
J. A . Guj, Magnetooptical Properties of Large-Gap Diluted Magnetic Semiconductors
J. Myridski, Shallow Acceptors in Diluted Magnetic Semiconductors: Splitting, Boil-off. Giant
Negative Magnetoresistance
A. K, Rcrmudas and R. Rudriquez, Raman Scattering in Diluted Magnetic Semiconductors
P . A. Wo@ Theory of Bound Magnetic Polarons in Semimagnetic Semiconductors
Volume 26 111-V Compound Semiconductors and Semiconductor

Properties of Superionic Materials
Z. Yuanxi, 111-V Compounds
H. V. Winston, A . T. Hunter. H. Kiniuru, uritl R. E. Lee, InAs-Alloyed GaAs Substrates for
Direct Implantation
P. K . B l ~ ~ i r c i c h a and
r y ~ S. Dhar, Deep Levels in I l l - V Compound Semiconductors Grown by
MBE
Y. Ya. Gurcvich and A . K. fvcinov-Sliir.s. Semiconductor Properties of Supersonic Materials
Volume 27 High Conducting Quasi-One-Dimensional Organic Crystals

E. M . Conwell, Introduction to Highly Conducting Quasi-One-Dimensional Organic Crystals
I. A . Howard, A Reference Guide to the Conducting Quasi-One-Dimensional Organic
Molecular Crystals
J. P. Pouquer, Structural Instabilities
E. M . Conwell, Transport Properties
(7. S. Jucohsen, Optical Properties
J. C. Scort, Magnetic Properties
L. Zuppiroli, Irradiation Effects: Perfect Crystals and Real Crystals
Volume 28 Measurement of High-speed Signals in Solid State Devices

J . Frey and D.Ioannou, Materials and Devices for High-speed and Optoelectronic Applications
H . Schumarher and E. Strid, Electronic Wafer Probing Techniques
D.H. Auston, Picosecond Photoconductivity: High-speed Measurements of Devices and
Materials
J. A. Vulrlinunis,Electro-Optic Measurement Techniques for Picosecond Materials, Devices,
and Integrated Circuits.
J. M. WiesenfeldandR. K. Jain, Direct Optical Probing of Integrated Circuits and High-speed
Devices
G. Plows, Electron-Beam Probing
A. M. Weiner und R. B. Mrtrcus, Photoemissive Probing
360 OF VOLUMES
Volume 29 Very High Speed Integrated Circuits: Gallium Arsenide LSI

M. Kuzuhara and T. Nazaki, Active Layer Formation by Ion Implantation
H. Hasirnoto, Focused Ion Beam Implantation Technology
T. Nozaki and A. Higashisaka, Device Fabrication Process Technology
M. Zno and T. Takadu, GaAs LSI Circuit Design
M. Hirayama, M. Ohmori, and K. Yamasaki, GaAs LSI Fabrication and Performance
Volume 30 Very High Speed Integrated Circuits: Heterostructure

H. Watanabe, T. Mimtani, and A. Usui, Fundamentals of Epitaxial Growth and Atomic Layer
Epitaxy
S. Hiyamizu, Characteristics of Two-Dimensional Electron Gas in IIILV Compound
Heterostructures Grown by MBE
T. Nakanisi, Metalorganic Vapor Phase Epitaxy for High-Quality Active Layers
T. Nimura, High Electron Mobility Transistor and LSI Applications
T. Sugeta and T. Zshibashi, Hetero-Bipolar Transistor and LSI Application
H. Matsuedu, T. Tanaka, and M . Nakumuru. Optoelectronic Integrated Circuits
Volume 3 1 Indium Phosphide: Crystal Growth and Characterization

J. P. Farges, Growth of Discoloration-free InP
M. J. McColluni and G. E. Stillman, High Purity InP Grown by Hydride Vapor Phase Epitaxy
T. Znada and T. Fukuda, Direct Synthesis and Growth of Indium Phosphide by the Liquid
Phosphorous Encapsulated Czochralski Method
0. O h . K. Katagiri, K. Shinohara, S. Karsura, Y. Takahashi, K Kuinosho, K. Kohiro. and R.
Hirano, InP Crystal Growth, Substrate Preparation and Evaluation
X Tada, M. Tatsumi, M. Morioka, T. A r k i , und T. Kawase, InP Substrates: Production and
Quality Control
M Razeghi, LP-MOCVD Growth, Characterization, and Application of InP Material
T. A. Kennedy and P. J. Lin-Chung, Stoichiometric Defects in InP
Volme 32 Strained-Layer Superlattices: Physics

T. P. Pearsall, Strained-Layer Superlattices
F. H. Pollack, Effects of Homogeneous Strain on the Electronic and Vibrational Levels in
Semiconductors
J. Y. Marzin, J. M. Gerirrd, P. Voisin, and J. A. Erum, Optical Studies of Strained 111-V
Heterola yers
R. People and S. A. Jackson, Structurally Induced States from Strain and Confinement
M. Jaros, Microscopic Phenomena in Ordered Superlattices
Volume 33 Strained-Layer Superlattices:

Materials Science and Technology
R. Hull and J. C. Bean, Principles and Concepts of Strained-Layer Epitaxy
W. J. Schafi P. J. Tusker, M. C. Foisj, and L. F. Eustman, Device Applications of
Strained-Layer Epitaxy
OF VOLUMES 361
S. T. Picraux, B. L. Doyle, and J. Y. Tsao, Structure and Characterization of Strained-Layer

Superlattices
E. Kasper and F. Schafer, Group IV Compounds
D. L. Martin, Molecular Beam Epitaxy of IV-VI Compounds Heterojunction
R. L. Gunshor, L. A. Kolodziejski, A. I/. Nurmikko,and N. Otsuka, Molecular Beam Epitaxy of
11-VI Semiconductor Microstructures
Volume 34 Hydrogen in Semiconductors

J. I. Pankove and N. M. Johnson, Introduction to Hydrogen in Semiconductors
C. H. S a g e r , Hydrogenation Methods
J. I. Pankove, Hydrogenation of Defects in Crystalline Silicon
J. W . Corbett, P. Deak, U.!L Desnicu, ond S. J . Pearton, Hydrogen Passivation of Damage
Centers in Semiconductors
S. J. Pearton, Neutralization of Deep Levels in Silicon
J. I. Pankove, Neutralization of Shallow Acceptors in Silicon
N. M . Johnson, Neutralization of Donor Dopants and Formation of Hydrogen-Induced
Defects in n-Type Silicon
M. Stavola and S. J. Pearron, Vibrational Spectroscopy of Hydrogen-Related Defects in Silicon
A. D. Marwick, Hydrogen in Semiconductors: Ion Beam Techniques
C. Herring and N . M. Johnson, Hydrogen Migration and Solubility in Silicon
E. E. Huller, Hydrogen-Related Phenomena in Crystalline Germanium
J. Kakalios, Hydrogen Diffusion in Amorphous Silicon
J. Chevalier. B. Clerjaud, and B. Pajot, Neutralization of Defects and Dopants in 111-V
Semiconductors
G. G. DeLeo and W. B. Fowler, Computational Studies of Hydrogen-Containing Complexes in
Semiconductors
R. F. Kiefl and T. L. Estle, Muonium in Semiconductors
C. G. Van de W a l k , Theory of Isolated Interstitial Hydrogen and Muonium in Crystalline
Semiconductors
Volume 35 Nanostructured Systems

M . Reed, Introduction
H. van Houten, C. W J. Beenakker, and B. J. van Wees, Quantum Point Contacts
G. Timp, When Does a Wire Become an Electron Waveguide?
M. Biittiker, The Quantum Hall Effects in Open Conductors
W. Hansen, J. P. Korthaus, rind U. Merkr, Electrons in Laterally Periodic Nanostructures
Volume 36 The Spectroscopy of Semiconductors

D. Heitnan, Spectroscopy of Semiconductors at Low Temperatures and High Magnetic Fields
A. V. Nurmikko, Transient Spectroscopy by Ultrashort Laser Pulse Techniques
A. K. Ramdas and S. Rodriguez, Piezospectroscopy of Semiconductors
0. J. GIemhocki and B. V. Shanabrook, Photoreflectance Spectroscopy of Microstructures
D. G. Seiler, C. L. Lirtler, and M. H. Wiler, One- and Two-Photon Magneto-Optical
Spectroscopy of InSb and Hg, -xCd,Te
362 OF VOLUMESIN THISSERIES
CONTENTS
Volume 37 The Mechanical Properties of Semiconductors

A.-B. Chen, A. Sher and W. T. Yost, Elastic Constants and Related Properties of
Semiconductor Compounds and Their Alloys
D. R Clarke, Fracture of Silicon and Other Semiconductors
H. Siethof, The Plasticity of Elemental and Compound Semiconductors
S. Guruswumy, K. T. Faber and J. P. Hirfh, Mechanical Behavior of Compound
Semiconductors
S. Mahajun, Deformation Behavior of Compound Semiconductors
J. P. Hirth, Injection of Dislocations into Strained Multilayer Structures
D.Kendull, C. B. Fleddermunn. and K. J. Malloy, Critical Technologies for the Micromachining
of Silicon
1. M a m b a and K. Mokuya, Processing and Semiconductor Thermoelastic Behavior
Volume 38 Imperfections in IIIlV Materials

U. Scherz and M. Schefler, Density-Functional Theory of sp-Bonded Defects in III/V
Semiconductors
M . Kaminska and E. R. Weber, El2 Defect in GaAs
D. C. Look, Defects Relevant for Compensation in Semi-Insulating GaAs
R C. Newwan, Local Vibrational Mode Spectroscopy of Defects in IIIjV Compounds
A. M. Hennel, Transition Metals in III/V Compounds
K. J. Malloy and K. Khachaturyan, DX and Related Defects in Semiconductors
V. Swaminuthan and A. S. Jordun, Dislocations in IIIjV Compounds
K. W. Nauka, Deep Level Defects in the Epitaxial IIIjV Materials
Volume 39 Minority Carriers in 111-V Semiconductors:

Physics and Applications
N. K. Dutfa, Radiative Transitions in GaAs and Other 111-V Compounds
R. K , Ahrenkiel, Minority-Carrier Lifetime in IIILV Semiconductors
T. Furuta, High Field Minority Electron Transport in p-GaAs
M . S. Lundstrom, Minority-Carrier Transport in 111-V Semiconductors
R A. A b r m , Effects of Heavy Doping and High Excitation on the Band Structure of GaAs
D. Yevick and W. Bardyszewski. An Introduction to Non-Equilibrium Many-Body Analyses of
Optical Processes in 111-V Semiconductors
Volume 40 Epitaxial Microstructures

E. F. Schuberr, Delta-Doping of Semiconductors: Electronic, Optical, and Structural Properties
of Materials and Devices
A . Gossard, M. Sundarum, and P. Hopkins, Wide Graded Potential Wells
P. Petroz Direct Growth of Nanometer-Size Quantum Wire Superlattices
E. Kupon, Lateral Patterning of Quantum Well Heterostructures by Growth of Nonplanar
Substrates
H. Temkin. D. Gershoni, and M . Punish. Optical Properties of Ga, _,In,As/lnP Quantum
Wells
CONTENTS 363
Volume 4 1 High Speed Heterostructure Devices

F. Capasso, F. Beltram, S. Sen. A. Puhlevi, and A. Y. Cho, Quantum Electron Devices: Physics
and Applications
P. Solomon, D. J. Frank, S. L. Wright, trnd F. Cmora, GaAs-Gate Semiconductor-.Insulator-
Semiconductor FET
M. H. Hasherni and I/. K. Mishra, Unipolar InP-Based Transistors
R. Kiehl, Complementary Heterostructure FET Integrated Circuits
T. Ishibashi, GaAs-Based and InP-Based Heterostructure Bipolar Transistors
H. C. Liu and T. C. L. G. Sollner, High-Frequency-Tunneling Devices
H. Ohnishi, T. More, M. Takatsu, K . Imuniuru. and N. Yokoyumu, Resonant-Tunneling
Hot-Electron Transistors and Circuits
Volume 42 Oxygen in Silicon

F. Shimura, Introduction to Oxygen in Silicon
W. Lin, The Incorporation of Oxygen into Silicon Crystals
T. J. Schujier and D. K. Schroder, Characterization Techniques for Oxygen in Silicon
W. M. Bullis, Oxygen Concentration Measurement
S. M. Hu, Intrinsic Point Defects in Silicon
8. Pajor, Some Atomic Configurations of Oxygen
J. Michel and L. C. Kimerling, Electical Properties o f Oxygen in Silicon
R. C. Newman and R. Jones, Diffusion of Oxygen in Silicon
T. Y. Tan and W J. Tuylor, Mechanisms of Oxygen Precipitation: Some Quantitative Aspects
M. Schrerns, Simulation of Oxygen Precipitation
K. Simino und I. Yonenriga, Oxygen Effect on Mechanical Properties
W. Bergholz, Grown-in and Process-Induced Effects
F. Shimura, Intrinsic/Internal Gettering
H. Tszrya, Oxygen Effect on Electronic Device Performance
Volume 43 Semiconductors for Room Temperature

Nuclear Detector Applications
R. B. James and T. E. Schlesinger, Introduction and Overview
L. S. Darken and C. E. Cox, High-Purity Germanium Detectors
A. Burger, D. Nason, L. Van den Berg. rind M. Schieher, Growth o f Mercuric Iodide
X J. Boo, T. E. Schlesinger. and R. B. Jtrnies, Electrical Properties of Mercuric Iodide
X. J. Brio, R. B. James. and T. E. SchIesitIger, Optical Properties of Red Mercuric Iodide
M. Hage-Ali and P. Si$erf, Growth Methods of CdTe Nuclear Detector Materials
M. Huge-Ali and P S f e r r , Characterization o f CdTe Nuclear Detector Materials
M. Huge-Ali and P. Sifferf, CdTe Nuclear Detectors and Applications
R. B. Jumes, T. E. Schlesinger. J. Lund, rind M . Scliieber, Cd, -xZn,Te Spectrometers for
Gamma and X-Ray Applications
D. S. McGregor. J. E. Kammeruad Gallium Arsenide Radiation Detectors and Spectrometers
J. C. Lund. F. Olschner. and A. Burger, Lead Iodide
M . R. Squillunte, and K . S. Shah. Other Materials: Status and Prospects
V. M. Gerrish, Characterization and Quantification of Detector Performance
J. S. iwunciyk and B. E. Part, Electronics for X-ray and Gamma Ray Spectrometers
M. Schieher, R. B. James, and T. E. Schle.~ingt~r. Summary and Remaining Issues for Room
Temperature Radiation Spectrometers
364 CONTENTS IN THISSERIES
OF VOLUMES
Voiume 44 11-IV BluefGreen Light Emitters:

Device Physics and Epitaxial Growth
J. Hun and R. L. Gunshor, MBE Growth and Electrical Properties of Wide Bandgap
ZnSe-based 11-VI Semiconductors
S. Fujita w d S. Fujitu, Growth and Characterization of ZnSe-based IILVI Semiconductors by
MOVPE
E. Ho and L. A . Kolodziejski, Gaseous Source UHV Epitaxy Technologies for Wide Bandgap
11-VI Semiconductors
C. G. Van de Walle, Doping of Wide-Band-Gap 11-V1 Compounds-Theory
R Cingoluni, Optical Properties of Excitons in ZnSe-Based Quantum Well Heterostructures
A. Ishibashi and A . V. Nurmikko, 11-VI Diode Lasers: A Current View of Device Performance
and Issues
S. Guha and J. Perruzello, Defects and Degradation in Wide-Gap IILVI-based Structures and
Light Emitting Devices
Volume 45 Effect of Disorder and Defects in Ion-Implanted Semiconductors:

Electrical and Physiochemical Characterization
H. Ryssel, Ion Implantation into Semiconductors: Historical Perspectives
You-Nian Wang and Teng-Cai Ma, Electronic Stopping Power for Energetic Ions in Solids
S. T. Nakagawa, Solid Effect on the Electronic Stopping of Crystalline Target and Application
to Range Estimation
G. Miiller, S. Kalbirzer and G. N. Greczves. Ion Beams in Amorphous Semiconductor Research
J. Eoussey-Said, Sheet and Spreading Resistance Analysis of Ion Implanted and Annealed
Semiconductors
M . L. Polignano and G. Queirolo, Studies of the Stripping Hall EKect in Ion-Implanted Silicon
J. Stoemenos. Transmission Electron Microscopy Analyses
R. Nipoti and M. Servidori, Rutherford Backscattering Studies of Ion Implanted
Semiconductors
P. Zaumseil, X-ray Diffraction Techniques
Volume 46 Effect of Disorder and Defects in Ion-Implanted Semiconductors:

Optical and Photothermal Characterization
M. Fried, T. Lohner and J. Gyulai, Ellipsometric Analysis
A. Seas and C. Christofides, Transmission and Reflection Spectroscopy on Ion Implanted
Semiconductors
A. Othonos and C. Christofides, Photoluminescence and Raman Scattering of Ion Implanted
Semiconductors. Influence of Annealing
C. Christojides, Photomodulated Thermoreflectance Investigation of Implanted Wafers.
Annealing Kinetics of Defects
U. Zamrnit, Photothermal Deflection Spectroscopy Characterization of Ion-Implanted and
Annealed Silicon Films
A. Mendelis, A. Budiman and M. Vargos. Photothermal Deep-Level Transient Spectroscopy of
Impurities and Defects in Semiconductors
R. Kalish and S. Charbonneau, Ion lmplantation into Quantum-Well Structures
A. M. Myasnikov and N. N. Gerasimenko, Ion Implantation and Thermal Annealing of 111-V
Compound Semiconducting Systems: Some Problems of 111-V Narrow Gap
Semiconductors
OF VOLUMES 365
Volume 47 Uncooled Infrared Imaging Arrays and Systems

R. G. Buser ond M. P. Tompsett, Historical Overview
P . W Kruse, Principles of Uncooled Infrared Focal Plane Arrays
R . A. Wood, Monolithic Silicon Microbolometer Arrays
C. M . Hanson, Hybrid Pyroelectric-Ferroelectric Bolometer Arrays
D. L. Polla and J . R. Choi, Monolithic Pyroelectric Bolometer Arrays
N . Teronishi, Thermoelectric Uncooled lnfrared Focal Plane Arrays
M . F . Tompsett, Pyroelectric Vidicon
7: W Kenny, Tunneling Infrared Sensors
J . R . Vig, R. L. Filler and Y. Kim, Application of Quartz Microresonators to Uncooled Infrared
Imaging Arrays
P . W Kruse, Application of Uncooled Monolithic Thermoelectric Linear Arrays to Imaging
Radiometers
Volume 48 High Brightness Light Emitting Diodes

G. B. Stringfellow, Materials Issues in High-Brightness Light-Emitting Diodes
M . G. Craford, Overview of Device issues in High-Brightness Light-Emitting Diodes
F . M . Steranka, AlGaAs Red Light Emitting Diodes
C. H . Chen, S. A. Stockman, M . J . Peanasky, and C . P . Kuo, OMVPE Growth of AlGaInP for
High Efficiency Visible Light-Emitting Diodes
F. A . Kish and R. M . Fletcher, AlGaInP Light-Emitting Diodes
M . W Hodapp, Applications for High Brightness Light-Emitting Diodes
I . Akasaki and H . Amano, Organometallic Vapor Epitaxy of GaN for High Brightness Blue
Light Emitting Diodes
S. Nakamura, Group 111-V Nitride Based Ultraviolet-Blue-Green-Yellow Light-Emitting
Diodes and Laser Diodes
Volume 49 Light Emission in Silicon: from Physics to Devices

D. J. Lockwood, Light Emission in Silicon
G. Abstreiter, Band Gaps and Light Emission in Si/SiGe Atomic Layer Structures
T. G. Brown and D. G. Hall, Radiative lsoelectronic Impurities in Silicon and
Silicon-Germanium Alloys and Superlattices
J. Michel, L. V. C. Assali, M.T. Morse, and L. C. Kimerling. Erbium in Silicon
Y Kunemitsu, Silicon and Germanium Nanoparticles
P. M. Fauchet, Porous Silicon: Photoluminescence and Electroluminescent Devices
C. Delerue, G. Allan, and M.Lannoo, Theory of Radiative and Nonradiative Processes in
Silicon Nanocrystallites
L. Buus, Silicon Polymers and Nanocrystals
Volume 50 Gallium Nitride (GaN)

J. I. Pankove and T. D. Moustakas. Introduction
S. P. DenBaars and S. Keller, Metaloryanic Chemical Vapor Deposition (MOCVD) of Group
111 Nitrides
W. A . Bryden and T. J. Kistenmacher, Growth of Group IIILA Nitrides by Reactive Sputtering
N. Newman, Thermochemistry of IIILN Semiconductors
S. J. Pearton and R. J. Shul, Etching of 111 Nitrides
366 OF VOLUMESIN THISSERIES
CONTENTS
S. M. Bedair, Indium-based Nitride Compounds

A. Tramperf, 0.Brandt, and K. H. Ploog, Crystal Structure of Group Ill Nitrides
H. Morkoc, F. Hamdani, and A. Salvador, Electronic and Optical Properties of 111-V Nitride
based Quantum Wells and Superlattices
K. Doverspike and J. I. Pankove, Doping in the 111-Nitrides
T. Suski and P. Perlin, High Pressure Studies of Defects and Impurities in Gallium Nitride
B. Monemar, Optical Properties of GaN
W. R. L. Lambrecht, Band Structure of the Group 111 Nitrides
N. E. Christensen and P. Perlin. Phonons and Phase Transitions in GaN
S. Nakamura, Applications of LEDs and LDs
I. Akasaki and H. Ammo, Lasers
J. A. Cooper, Jr.. Nonvolatile Random Access Memories in Wide Bandgap Semiconductors
Volume 5 1A Identification of Defects in Semiconductors

G. D. Watkins, EPR and ENDOR Studies of Defects in Semiconductors
J.-M. Spaeth, Magneto-Optical and Electrical Detection of Paramagnetic Resonance in
Semiconductors
T. A. Kennedy and E. R. Glaser, Magnetic Resonance of Epitaxial Layers Detected by
Photoluminescence
K . H. Chow, B. Hitfi, and R. F. Kiej!, pSR on Muonium in Semiconductors and Its Relation to
Hydrogen
K. Saarinen, P. Hautojiirvi, and C. Corbel. Positron Annihilation Spectroscopy of Defects in
Semiconductors
R Jones and P. R. Briddon. The Ah Initio Cluster Method and the Dynamics of Defects in
Semiconductors
Volume 5 1B Identification of Defects in Semiconductors

G. Davies, Optical Measurements of Point Defects
P. M. Mooney, Defect Identification Using Capacitance Spectroscopy
M. Stavolu, Vibrational Spectroscopy of Light Element Impurities in Semiconductors
P. Schwander, W. D. Rau. C. Kisielowski. M . Gribelyuk, and A. Ourmazd, Defect Processes in
Semiconductors Studied at the Atomic Level by Transmission Electron Microscopy
N. D. Jager and E. R. Weber, Scanning Tunneling Microscopy of Defects in
Semiconductors
Volume 52 Sic Materials and Devices

K. Jarrendalil and R. F Davis, Materials Properties and Characterization of SIC
V. A. Dmitriev and M. G. Spencer, Sic Fabrication Technology: Growth and Doping
V. Saxena and A. J. Sfeckl, Building Blocks for SIC Devices: Ohmic Contacts, Schottky
Contacts, and p-n Junctions
M. S. Shur, Sic Transistors
C. D. Brandt. R. C. Clarke, R. R. Siergiej, J. B. Casady, A. W. Morse, S.Sriram, and A. K.
Agarwal, S i c for Applications in High-Power Electronics
R. J. Trew. Sic Microwave Devices
OF VOLUMES 367
J. Edmond, H. Kong, G. Negley. M. Leonard, K. Doverspike, W. Weeks, A. Suvorov, D. Waltz,

and C. Carter. Jr., Sic-Based UV Photodiodes and Light-Emitting Diodes
H . Morkag, Beyond Silicon Carbide! Ill-V Nitride-Based Heterostructures and Devices
Volume 53 Cumulative Subject and Author Index

Including Tables of Contents for Volume 1-50
Volume 54 High Pressure in Semiconductor Physics I

W. Paul, High Pressure in Semiconductor Physics: A Historical Overview
N . E. Christensen, Electronic Structure Calculations for Semiconductors under Pressure
R. J. NeimeJ and M . I. McMahon, Structural Transitions in the Group IV, 111-V and 11-VI
Semiconductors Under Pressure
A. R. Goni and K. Syassen, Optical Properties of Semiconductors Under Pressure
P. Trautman, M. Baj, and J. M . Buranowski, Hydrostatic Pressure and Uniaxial Stress in
Investigations of the EL2 Defect in GaAs
M. Li nnd P. Y Yu, High-pressure Study of DX Centers Using Capacitance Techniques
T. Suski, Spatial Correlations of Impurity Charges in Doped Semiconductors
N. Kuroda, Pressure Effects on the Electronic Properties of Diluted Magnetic Semiconductors
Volume 55 High Pressure in Semiconductor Physics I1

D.K. Maude and J. C. Portal. Parallel Transport in Low-Dimensional Semiconductor
Structures
P. C. Klipstein, Tunneling Under Pressure: High-pressure Studies of Vertical Transport in
Semiconductor Heterostructures
E. Anastussakis and M. Cardona, Phonons, Strains, and Pressure in Semiconductors
F. H. Pollak, Effects of External Uniaxial Stress on the Optical Properties of Semiconductors
and Semiconductor Microstructures
A. R. Adams, M. Silver. and J. Allum. Semiconductor Optoelectronic Devices
S. Porowski and I. Grzegory, The Application of High Nitrogen Pressure in the Physics and
Technology of 111-N Compounds
M. Yousuf; Diamond Anvil Cells in High Pressure Studies of Semiconductors
Volume 56 Germanium Silicon: Physics and Materials

J. C. Beun, Growth Techniques and Procedures
D. E. Savage, F. Liu, V. Zielasek, und M . G. Lagally, Fundamental Crystal Growth
Mechanisms
R. HUN, Misfit Strain Accommodation in SiGe Heterostructures
M. J. Shaw and M. Juros, Fundamental Physics of Strained Layer GeSi: Quo Vadis?
F. Cerdeira, Optical Properties
S. A . Ringel and P. N. Grillot, Electronic Properties and Deep Levels in Germanium-Silicon
J. C. Campbell, Optoelectronics in Silicon and Germanium Silicon
K. Eberl, K. Brunner. and 0. G. Schmidi. Si,_,C, and Si,-,-,Ge,C, Alloy Layers
368 CONTENTS
Volume 57 Gallium Nitride (GaN) I1

R. J. Molnar, Hydride Vapor Phase Epitaxial Growth of 111-V Nitrides
T. D. Moustakas, Growth of 111-V Nitrides by Molecular Beam Epitaxy
Z. Liliental- Weber, Defects in Bulk GaN and Homoepitaxial Layers
C. G. Van de Walk and N. M. Johnson. Hydrogen in 111-V Nitrides
W. Gotz and N. M. Johnson, Characterization of Dopants and Deep Level Defects in Gallium
Nitride
B. Gil, Stress Effects on Optical Properties
C. Kisielowski, Strain in GaN Thin Films and Heterostructures
J. A. Miragliottu and D. K. Wickenden. Nonlinear Optical Properties of Gallium Nitride
B. K Meyer. Magnetic Resonance Investigations on Group 111-Nitrides
M. S. Shur and M. AsifKhun, GaN and AlGaN Ultraviolet Detectors
C. H. Qiu, J. I. Pankove. and C. Rousington. 111-VNitride-Based X-ray Detectors
Volume 58 Nonlinear Optics in Semiconductors I

A. Kost, Resonant Optical Nonlinearities in Semiconductors
E. Garmire, Optical Nonlinearities in Semiconductors Enhanced by Carrier Transport
D. S. Chemlu. Ultrafast Transient Nonlinear Optical Processes in Semiconductors
M. Sheik-Bahae and E. W. Van Stryland Optical Nonlinearities in the Transparency Region of
Bulk Semiconductors
J. E. Millerd, M. Ziari, and A. Partoit, Photorefractivity in Semiconductors
Volume 59 Nonlinear Optics in Semiconductors I1

J. B. Kkurgin, Second Order Nonlinearities and Optical Rectification
K. L. Hull, E. R. Thoen, and E. P. Ippen. Nonlinearities in Active Media
E. Hanamura, Optical Responses of Quantum Wires/Dots and Microcavities
U. Keller, Semiconductor Nonlinearities for Solid-state Laser Modelocking and Q-Switching
A. Miller, Transient Grating Studies of Carrier Diffusion and Mobility in Semiconductors
I S B N O-L2-75216 3-0

(Mitsuru Sugawara, Robert K. Willardson, Eicke R.

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Mitsuru Sugawara, Robert K. Willardson, Eicke R.

Uploaded by

Copyright:

Available Formats

Self-Assembled InGaAslGaAs

Edited by R. K. Witlardson Eicke R. Weber

International Standard Book Number: 0-12-752169-0

PRINTED I N THE UNITED STATES OF AMERICA

Chapter 1 Theoretical Bases of the Optical Properties of Semiconductor

Chapter 2 Molecular Beam Epitaxial Growth of

Chapter 3 Metalorganic Vapor Phase Epitaxial Growth

Chapter 4 Optical Characterization of Quantum Dots

Chapter 5 The Photon Bottleneck Effect in Quantum Dots

Chapter 6 Self-Assembled Quantum Dot Lasers

Chapter 7 Applications of Quantum Dot to Optical Devices

2. Direct Modulation of Quantum-Dot Lasers . . . . . . . . . . . . . . . 298

Chapter 8 The Latest News

This volume is devoted to the crystal growth, optical properties, and

Experimental results and numerical calculations are included to improve

a theoretical problem; however, the recent advent of self-formed quantum

Numbers in parenthesis indicate the pages on which the authors’

MITSURUEGAWA,(1 55) Optical Semiconductor Devices Laboratory, Fujitsu

Theoretical Bases of the Optical Properties of

Research on synthesized semiconductor quantum nano-structures was

structures consisting of alternating, differing layers, whose thickness was less

(MBE) and metal-organic vapor phase epitaxy (MOCVD), which created

11. Electronic States of Semiconductor Quantum Nano-Structures

Recent progress in nano-scale growth techniques such as MBE and

FIG. 1.1. Cross-sectional transmission electron microscope (TEM) photograph of (a)

(SK) mode. In particular, they emit at a much longer wavelength of 1.3 pm

N,: Areal density

band-edge electron states of semiconductors can be described by the

Here, m* is the effective mass; h is the Planck's constant divided by 2.n;

where s2 is the unit-cell volume. The 111-V and 11-VI semiconductor

Setting V(r) = 0, Eq. (1.1) gives the energy of

and the envelope function of

Here, R, = 4n is the integration in the angular part. The factor of two

with the corresponding normalized wave function of

where R, = 2x is the integration in the angular part and O(x) is the

Quan rum Wive

Here, L is the length of the quantum wire, and k = ( k x ,nyn/Ly,n,n/L,).

Assuming an infinitely high potential for all directions, the confinement

E = E(k) = Enx + EnY+ En== __

where n,, ny,n, = 1,2,3.. . and k = (n,n/L,, nyz/Ly,n,z/L,). The energy

Letting N , be the volume density of quantum dots, the density of states is

1 1.2 1.4 1.6

111. Interband Optical Transition

The interband transition of an electron between the conduction and

Let us derive linear and nonlinear optical susceptibilities using the

By taking )I as the electron wave function of the material system and

where H , is the electron unperturbed Hamiltonian, and H r is the Hamil-

The interaction Hamiltonian with the electromagnetic field is given as

where e is the electron charge, m, is the electron mass, and p is the

w. It is written in a classical manner as

The electric field is related to the vector potential as

Using the linear susceptibility, x ( l ) , and the third-order susceptibility, x(~),

P(r, t ) = Eofl)E(r, t ) + E ~ X ( ~ t)E(r,

where E~ is the permittivity of a vacuum and subscripts of q and u are

Thus, the interband optical transition occurs perpendicularly in the energy

The diagonal terms, pi:) and p!:,', are zero.

is used. Qd (d = 0,1,2,3) is the inverse of the crystal volume; Q3 = V-' for

where p i is the quasi-Fermi level; k , is the Boltzmann's constant; and T is

Thus, we get the first-order susceptibility as

Degeneracy due to spin is taken into account. The contribution of other k

is the normalized broadening function and nr is the background refractive

where E , is the band gap;

x @(hw - E , - E$) - Eid))[.f,(w)