New Structures for AlGaAs Lasers and Avalanche Photodetectors - CaltechTHESIS
CaltechTHESIS
A Caltech Library Service
About
Browse
Deposit an Item
Instructions for Students
New Structures for AlGaAs Lasers and Avalanche Photodetectors
Citation
Blauvelt, Henry A.
(1983)
New Structures for AlGaAs Lasers and Avalanche Photodetectors.
Dissertation (Ph.D.), California Institute of Technology.
doi:10.7907/DJMZ-S785.
Abstract
This thesis describes the fabrication and the properties of five new semiconductor laser diode structures. All of these devices were grown from the GaAs-AlGaAs ternary system using the liquid phase epitaxial technique. In addition, a new low noise avalanche photodetector is proposed.
The first example is a new technique for fabricating cleaved mirrors without cleaving through the substrate. This technique, called micro-cleavage, has potential applications for both opto-electronic integrated circuits and for the fabrication of short cavity length lasers. In this technique, cantilevers are formed by a sequence of etching steps. These cantilevers are subsequently cleaved using ultrasonic vibrations.
Three devices related to high power single mode lasers are described. The first of these is the large optical cavity buried heterostructure window laser. The output power of semiconductor lasers, particularly during pulsed operation is limited by catastrophic mirror damage which occurs at power densities above a pulse width dependent damage threshold. The damage occurs due to local heating up to the melting point of the active region in the vicinity of the cleaved mirror facets. However, catastrophic mirror damage can be avoided by isolating the active layer from the cleaved mirrors, as is done in these window lasers. The second device related to high power that is described is the Inverted Strip Buried Heterostructure laser. These lasers combine many of the best features of both the buried optical guide lasers and the strip buried heterostructure that have been previously developed elsewhere. The inverted strip buried heterostructure lasers have significantly better beam quality than buried optical guide lasers and can be operated in the fundamental spatial mode for larger emitting areas (and therefore greater output power). The third device related to high power lasers is a variation of a buried heterostructure laser in which the injected current is confined to a narrow section in the center of the active layer. The optical gain is therefore also confined to a narrow section in the center of the active layer. By doing so the fundamental mode is much better matched to the optical gain than the higher order spatial modes. The result is that fundamental mode operation is possible for buried heterostructure lasers with active layer widths up to 8 µm. When the current is injected uniformly into the active layer, fundamental mode operation is possible only for active layer widths less than 2 µm. In addition to the descriptions of these devices a theoretical chapter on high power single mode lasers is included.
The final laser structure that is described is a single liquid phase epitaxial growth laser structure in which the current is restricted to flow between two narrow stripes located above and below the active layer. This structure, which is fabricated using a meltback-growth technique allows the current injection to be restricted to a very narrow section of the active layer, which results in several interesting properties which are described and explained using a simple model.
The final subject of this thesis is a multilayer avalanche photodetector (APD) which has been proposed for low noise applications. The noise generated by an APD is dependent on the statistics of the carrier multiplication process, since positive feedback effects, which exist when both electrons and holes produce secondary pairs, can greatly amplify any current fluctuations. Significantly more noise is generated if the electron and hole ionization rates (α, β) are equal than if only one carrier produces secondary pairs. The multilayer structure described and analyzed in this chapter is expected to have impact ionization which is dominated by electrons and therefore would be of importance for low noise applications.
Item Type:
Thesis (Dissertation (Ph.D.))
Subject Keywords:
(Applied Physics) ; AlGaAs lasers; avalanche photodetectors
Degree Grantor:
California Institute of Technology
Division:
Engineering and Applied Science
Major Option:
Applied Physics
Thesis Availability:
Public (worldwide access)
Research Advisor(s):
Yariv, Amnon
Thesis Committee:
Yariv, Amnon (chair)
Bridges, William B.
McCaldin, James Oeland
McGill, Thomas C.
Goodstein, David L.
Defense Date:
26 October 1982
Funders:
Funding Agency
Grant Number
Caltech
UNSPECIFIED
NSF
UNSPECIFIED
Office of Naval Research (ONR)
UNSPECIFIED
Record Number:
CaltechETD:etd-09072006-080636
Persistent URL:
DOI:
10.7907/DJMZ-S785
Default Usage Policy:
No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:
3373
Collection:
CaltechTHESIS
Deposited By:
Imported from ETD-db
Deposited On:
25 Sep 2006
Last Modified:
23 Jul 2025 17:09
Thesis Files
Preview
PDF
- Final Version
See Usage Policy.
9MB
Repository Staff Only:
item control page
CaltechTHESIS is powered by
EPrints 3.3
which is developed by the
School of Electronics and Computer Science
at the University of Southampton.
More information and software credits
NEW STRUCTURES FOR AIGaAs LASERS
AND AVALANCHE PHOTODETECTORS
Thesis by
Henry A. Blauvelt
In Partial Fulfillment of the Requirements
for the Degree of
Doctor of Philosophy
California Institute of Technology
Pasadena, California
1983
(Submitted October 26, 1982)
-ii-
To my Paren ts
-iii-
-Acknowledgements-
I would like to take this opportunity to express my appreciation to my
adviser, Professor Amnon Yariv, for his encouragement and support throughout
my graduate studies at Caltech. I have profited greatly from the experience
gained under his supervision in the quantum electronics group.
I am especially indebted to Dr. Shlomo Margalit for many fruitful discussions
and for his guidance. His constant willingness to discuss new ideas was of
tremendous value. Special thanks are due also to Dr. Nadav Bar-Chaim for his
collaboration on some of the work described in this thesis and for teaching me
how to grow crystals by liquid phase epitaxy.
I would also like to express my thanks to the other members, past and
present, of the quantum electronics group with whom 1 have worked these past
years: Dr. Israel Ury, Dr. Kam Lau, Dr. Daniel Wilt, Dr. Pei-Chuang Chen, Dr.
Joseph Katz, Dr. Uziel Koren, Mr. Chris Harder, Mr. Liew-Chuang Chiu, Mr. Paul
Yu, and Mr. Stephen Smith. I am also indebted to Mr. Desmond Armstrong for his
assistance with the experimental apparatus. I also thank Mr. Lawrence Begay for
his help in constructing some of the experimental apparatus used in this study.
I am thankful for the financial support received from the California Institute
of Technology, the National Science Foundation, and the Office of Naval
Research.
Finally, my deepest appreciation goes to my parents for thier constant support and encouragement of my undertaking this work.
-ivAbstract
This thesis describes th~ fabrication and the properties of five new semiconductor laser diode structures. All of these devices were grown from the GaAsAlGaAs ternary system using the liquid phase epitaxial technique. In addition, a
new low noise avalanche photodetector is proposed.
The first example is a new technique for fabricating cleaved mirrors without
cleaving through the substrate. This technique, called micro-cleavage, has
potential applications for both opto-electronic integrated circuits and for the
fabrication of short cavity length lasers. In this technique, cantilevers are
formed by a sequence of etching steps. These cantilevers are subsequently
cleaved using ultrasonic vibrations.
Three devices related to high power single mode lasers are described. The
first of these is the large optical cavity buried heterostructure window laser.
The output power of semiconductor lasers, particularly during pulsed operation
is limited by catastrophic mirror damage which occurs at power densities above
a pulse width dependent damage threshold. The damage occurs due to local
heating up to the melting point of the active region in the vicinity of the cleaved
mirror facets. However, catastrophic mirror damage can be avoided by isolating
the active layer from the cleaved mirrors, as is done in these window lasers. The
second device related to high power that is described is the Inverted Strip
Buried Heterostructure laser. These lasers combine many of the best features of
both the buried optical guide lasers and the strip buried heterostructure that
have been previously developed elsewhere. The inverted strip buried hlterostructure lasers have significantly better beam quality than buried optical guide
lasers and can be operated in the fundamental spatial mode for larger emitting
-v-
areas (and therefore greater output power). The third device related to high
power lasers is a variation of a buried heterostructure laser in which the
injected current is confined to a narrow section in the center of the active layer.
The optical gain is therefore also confined to a narrow section in the center of
the active layer. By doing so the fundamental mode is much better matched to
the optical gain than the higher order spatial modes. The result is that fundamental mode operation is possible for buried heterostructure lasers with active
layer widths up to 8 µm. When the current is injected uniformly into the active
layer, fundamental mode operation is possible only for active layer widths less
than 2 µm. In addition to the descriptions of these deVices a theoretical chapter
on high power single mode lasers is included.
The :final laser structure that is described is a single liquid phase epitaxial
growth laser structure in which the current is restricted to ftow between two
narrow stripes located above and below the active layer. This structure. which is
fabricated using a meltback-growth technique allows the current injection to be
restricted to a very narrow section of the active layer, which results in several
interesting properties which are described and explained using a simple model.
The :final subject of this thesis is a multilayer avalanche photodetector (APD)
which has been proposed for low noise applications. The noise generated by an
APD is dependent on the statistics of the carrier multiplication process. since
positive feedback effects, which exist when both electrons and holes produce
secondary pairs, can greatly amplify any current tluctuations. Significantly
more noise is generated if the electron and hole ionization rates (a.,(J) are equal
than if only one carrier produces secondary pairs. The multilayer structure
described and analyzed in this chapter is expected to have impact ionization
which is dominated by electrons and therefore would be of importance for low
noise applications.
-vi-
CONTENTS
Chapter I Introduction
1.1
Semiconductor laser diodes
1.2
Outline of thesis
References- Chapter I
11
Chapter II Lasers with Micro-cleaved Jlirrors
13
2.1
Introduction
13
2.2
Fabrication of lasers with micro-cleaved mirrors
14
2.3
Optoelectronic integrated circuits
19
2.4
Short cavity length lasers
22
References- chapter II
26
Chapter IlI High Power Single Mode AlGa.A.s Lasers
28
3.1
Introduction
28
3.2
Mode control of semiconductor lasers
31
3.3
High power single mode buried heterostructure lasers
47
Appendix 3.1 Optical waveguide theory
52
Appendix 3.2 Effective index method
53
Appendix 3.3 Vector variational method
59
References- chapter III
61
Chapter IV Large Optical Cavity AlGaAs Buried Heterostructure
Window Lasers
64
4.1
Introduction
64
4.2
Fabrication or buried heterostructure window lasers
69
4.3
Properties or buried heterostructure window lasers
73
-viiReferences- chapter IV
79
Chapter V AlGaAs Inverted Strip Buried Heterostructure Lasers
80
5.1
Introduction
80
5.2
_t:'abrication of inverted strip buried h·?terostructure lasers
86
5.3
Properties of inverted strip buried heterostructure lasers
90
References- chapter V
96
Chapter VI AlGaA.s Buried Heterostructure Lasers with Narrow
97
Carri.er Injection
6.1
Introduction
97
6.2
Fabrication of narrow injection buried heterostructure lasers
100
6.3
Properties of narrow injection buri~d heterostructure lasers
102
References- chapter VI
107
Chapter VII Narrow Stripe AlGaAs Lasers Using Double Current Confinement
109
7.1
Introduction
109
7 2
Fabrication of narrow stripe lasers with double current confinement
111
7.3
Properties of narrow stripe lasers with double current confinement
115
References- chapter VU
127
Chapter VIII Sin&le Carrier Type Dominated Im.pact Ionization in
Multilayer Structures
References- chapter VIII
130
142
- 1-
Cb.apter I
INTRODUCTION
1.1 Semiconductor Laser Diodes
Since
the
first
observations
of
lasing
action
in
GaAs
in
1962 1 -4,
semiconductor laser diodes have been the focus of great attention because of
their tremendous commercial potential.
improvements
have
been
made
in
the
In the past twenty years great
performance and
reliability
of
semiconductor lasers. By reducing the threshold currents of the laser diodes, to
as low as 4 mA Ln buried heterostructure lasers 5 ; continuous operation at room
temperature has been made possible. Lasers which operate continuously and
stably in a single spatial mode and frequency at output powers up to 40 mW
have been demonstrated6 •7 . Laser reliability has been improved to the point
where some laser structures have expected lifetimes exceeding 10 6 hours.
Throughout this thesis, a familiarity with semiconductor lasers is assumed.
Comprehensive treatments of semiconductor laser diodes can be found in the
books by Casey and Panish 8 , Kressel and Butler9 , and Thompson 10 •
Laser diodes have numerous unique properties which make them attractive
for a number of important applications. Laser diodes are very small in size, they
are highly efficient, and they can be modulated at frequencies up to several GHz
by directly modulating the laser current. Semiconductor laser diodes also have
the potential for duplicating the low cost and high reliability characteristics of
semiconductor
electronic
devices.
Most
of
the
major
applications
of
semiconductor laser diodes involve the use of laser diodes in the storage and
transmission of information. Laser diodes are being used as the transmitters
for high speed communication through optical fibers. Laser diodes will soon
appear in the rapidly expanding consumer electronics market to be used to
-2optically read high density audio and video information stored on discs. Storage
and retrieval of information from high density disc recording has potential
applications in the computer industry. The characteristics of laser diodes also
make them potentially attractive for document generation applications such as
laser printers and copiers.
Fiber optic communication has been the subject of intense research recently.
Of particular interest are fiber optic communication systems which use glass
fibers as the transmission medium and semiconductor devices as the light
sources and detectors 11 . Fiber optic communication systems have a number of
advantages as compared to conventional microwave transmitting systems.
Optical fibers are light weight, have very low attenuations, are free from
interference, and have large transmission bandwidths. The semiconductor light
sources and detectors in optical communication systems are small, highly
efficient, capable of operating at GHz frequencies, and have the potential for low
cost and high reliability.
Semiconductor laser research for optical communications has concentrated
primarily on IIl-V materials. In particular, the ternary AlxGa 1 _xAs grown on GaAs
substrates and the quaternary InxGa 1-xAs1 P 1 -y lasers grown on InP substrates
have received the most attention. By adjusting the mole fractions of the atomic
constituents, the emitting wavelength of these semiconductor lasers can be
varied. AlGaAs lasers have been fabricated with wavelengths ranging from 0.68
µm to 0.88 µm. InGaAsP lasers have been fabricated with wavelengths ranging
from 1.0 µm to 1.7 µm. Of particular interest for optical communications is the
wavelength range of 1.3-1.6 µm. This is the wavelength region where silica optical
fibers have the lowest loss and least optical dispersion. Attenuations as low as
0.2 dB/Km have been reported at 1.55 µm 12 . By comparison the best optical
fibers have losses of 2 dB/Km at the wavelength of GaAs lasers of 0.88 µm.
-3Because of this, InGaAsP lasers are the preferred transmitting sources for long
distance optical communication systems. However, AlGaAs lasers have many
advantages which make them more attractive for short distance, high speed
communications applications where low fiber losses and dispersion are not
critical. AlGaAs laser technology has received the most attention and is more
advanced than that of lnGaAsP lasers.
AlGaAs lasers are significantly less
sensitive to temperature variations. which is a major problem with long
wavelength quaternary lasers. The technology for fabricating GaAs electronic
circuits is also more advanced than that of lnP making monolithic integration of
AlGaAs lasers with GaAs electronic devices more attractive than the integration
of InGaAsP lasers with InP electronic circuits.
The advantages of integrated optoelectronic circuits (IOEC) combining AlGaAs
lasers and GaAs electronic circuits have been recognized for some time 13 . These
advantages include size, cost, and reliability. In addition, the reduction of
parasitic
reactances,
which
would
otherwise
result
from
device
interconnections, can lead to significant improvements in the speed and noise
performances of the optoelectronic circuits.
The problem of designing and fabricating IOECs is mostly one of devising
means to make the technologies of GaAs electronic circuits and AlGaAs lasers
compatible with one another. AlGaAs lasers are usually fabricated on highly
conductive substrates. GaAs electronics, such as MESFETs (metal semiconductor
field effect transistors), however, are fabricated on n-type layers which are
formed on top of semi-insulating (SI) substrates. The use of the SI substrates
reduces parasitic capacitances and also facilitates isolation between devices
fabricated on a common substrate.
It is therefore desirable for IOECs to
fabricate lasers on SI substrates. One problem in fabricating lasers on SI
substrates is that such lasers must be mounted on heat sinks with the substrate
-4-
side down, because of the requirement that both the P and N contacts be made
to the top surf ace. Since the active layer is typically located about 1 micron
beneath the top surface, the heat generated in the active layer can be removed
more efficiently when the lasers are mounted substrate side up (active layer
near the heat sink). For IOECs, it is therefore important to have lasers which
have threshold currents as low as possible to minimize the heat dissipation of
the devices. A second difficulty is the limitation on the size of the IOEC imposed
by the laser cavity formed by the cleaved facets of the chip. Several methods of
fabricating laser mirrors without cleaving through the substrate have been
reported. Using techniques, such as that of microcleavage 14 , which v.·m be
described in detail later in this thesis, lasers can be fabricated on substrates
which are not restricted in size by the laser cavity length, which is typically 300
µm.
A second major area of application of semiconductor laser diodes is in the
storage and retrieval of information from optical discs. In this application
information stored on an optical disc is read out by a focused laser beam as
illustrated in fig 1. Laser diodes are particularly well suited to this application
because of their small size. Consumer electronics products in which audio
and/ or video information stored on a disc is read out by a laser diode have
already been demonstrated. The digital audio disc, in which audio information
is stored on the disc in a digital format, making possible nearly perfect sound
reproduction, is expected to appear in the marketplace by late 1982.
Fig. 1
INFORMATION
ENCODED IN
PITS ON THE
DISC
BEAM
SPLITTER
C>
c:>
FOCUSING
LENS
Schematic representati on of an optical disc informalion storage system
COLLIMATING
LENS
LASER
DIODE
DETECTOR
V1
-6-
The first generation or optical disc recording systems are read only systems
in which information can be encoded by using specialized and expensive
equipment to evaporate pits in a metal film on the disc. These optical discs are
analogous to conventional audio records in which prerecorded information can
be played. Unlike audio and videotape, first generation optical disc systems do
not have any provisions for recording information by the user. However, simple
systems for recording information on discs using laser diodes has already been
demonstrated 15 and recording capability can be expected in future generations
of optical disc systems. The addition of a recording capability will make optical
disc systems very attractive as an alternative to magnetic disc systems for
computer memory.
For the reading of optical discs the most important laser characteristics are
the laser beam quality and the emitting wavelength. High quality optical beams
are essential for error free reading of the stored information.
Since the
minimum spot size of a focused laser beam is proportional to the wavelength, it
is desirable to use lasers with as short a wavelength as possible. AlGaAs lasers
emitting in the .7-.78 µm range have received the most attention for this
application. The reading of information from optical discs requires relatively
little laser output power, about 1 mW, which can be easily obtained from
commercially available lasers.
Optical recording of information requires
substantially more power, about 30 mW. Although continuous power outputs of
as high as 40 mW have been reported from lasers with good beam quality8· 7 ,
reliable laser operation has not as yet been achieved for output powers greater
than about 20 mW. The design and fabrication of high beam quality, high power
lasers has been an area of intense research lately, and will be discussed in more
detail later in this thesis.
-7-
1.2 Thesis Outline
Six subjects related to AlGaAs lasers will be discussed in this thesis in
chapters II-VII. Chapter II describes a new technique for fabricating cleaved
mirrors without cleaving through the substrate. This technique, called microcleavage, has potential applications for both opto-electronic integrated circuits
and for the fabrication of short cavity length lasers. In this technique,
cantilevers are formed by a sequence of etching steps. These cantilevers are
subsequently cleaved using ultrasonic vibrations.
This makes possible the
fabrication of lasers with cleaved mirrors on a substrate which is not restricted
in size to the laser cavity length. The fabrication and performance of oxide
stripe lasers with micro-cleaved mirrors is discussed in detail in this chapter.
Chapter III is a theoretical chapter concerned with the spatial mode
properties of semiconductor lasers and the closely related subject of high power
single mode laser design. For most applications of laser diodes, it is highly
desirable that the laser operates stably in the fundamental spatial mode. In
chapter III, the factors which determine the spatial mode patterns of
semiconductor lasers are examined. Spatial mode control of lasers actually
involves two separate problems. First, the laser structure must be such that at
low power levels the laser operates in the fundamental spatial mode. In addition,
the laser structure must be such that as the power level of the laser is
increased, the spatial mode pattern does not change. Both aspects of spatial
mode control are discussed in chapter III. Calculations related to the mode
gains and mode reflectivities of buried heterostructure lasers are presented.
Predictions of the spatial mode properties of buried heterostructure lasers,
based on these calculations are presented and compared with experimentally
observed properties. The factors which limit the amount of single mode power
- Bavailable from laser diodes are also discussed in detail in chapter III.
Chapter
describes
the
fabrication
and
performance
of
buried
heterostucture window lasers. The output power of semiconductor lasers,
particularly during pulsed operation is limited by catastrophic mirror damage
which occurs at power densities above a pulse width dependent damage
threshold. The catastrophic damage thresholds are approximately 15 mW I µm 2
for continuous operation and 70 mW I µm 2 for 100 nsec pulses. The damage
occurs due to local heating up to the melting point of the active region in the
vicinity of the cleaved mirror facets. However, catastrophic mirror damage can
be avoided by isoiating the active iayer from the cleaved mirrors. Chapter IV
describes a variation on the buried heterostructure laser which has transparent
"window" sections near the mirrors and is immune from catastrophic mirror
damage at power levels up to three times the catastrophic damage limit of
conventional buried heterostructure lasers.
Chapter V describes the fabrication and performance of Inverted Strip Buried
Heterostructure (ISBH) lasers. ISBH lasers combine many of the best features of
both the buried optical guide lasers developed by Chinone et al. 16 and the strip
buried heterostructure (SBH) lasers developed by Tsang and Logan 17 . These
lasers have significantly better beam quality than buried optical guide lasers
and can be operated in the fundamental spatial mode for active layer widths up
to 4 µm. These spatial mode characteristics are similar to those of strip buried
heterostructure lasers. However, ISBH lasers have lower threshold currents and
better thermal characteristics than SBH lasers. ISBH laser fabrication is also
simpler and therefore of potentially higher yield than SBH laser fabrication.
Chapter VJ describes a variation of a buried heterostructure laser which
applies some of the concepts discussed in chapter III. In the narrow injection
-9-
buried heterostructure lasers the injected current is confined to a narrow
section in the center of the active layer. The optical gain is therefore also
confined to a narrow section in the center of the active layer. By doing so the
fundamental mode is much better matched to the optical gain than the higher
order spatial modes. The result is that fundamental mode operation is possible
for buried heterostructure lasers with active layer widths up to 8 µm. When the
current is injected uniformly into the active layer, fundamental mode operation
is possible only for active layer widths less than 2 µm. The fabrication and
performance of these lasers is discussed in detail in chapter V1.
Chapter V11 describes a single liquid phase epitaxial growth laser structure in
which the current is restricted to flow between two narrow stripes located above
and below the active layer. This structure, which is fabricated using a meltbackgrowth technique allows the current injection to be restricted to a very narrow
section of the active layer. The primary emphasis in this chapter will be simply
to explain the structure and the properties of lasers with very narrow injection.
Among the more significant characteristics are far field patterns characteristic
of leaky mode waveguides and operation in a very large number of longitudinal
modes. In addition, the properties of the lasers suggest that the laser structure
may result in useful applications and potential applications, such as low
threshold laser structures and arrays of optically coupled lasers are discussed
briefly
Chapter VIII describes a multilayer avalanche photodetector (APD) which has
been prop~sed for low noise applications. The noise generated by an APD is
dependent on the statistics of the carrier multiplication process, since positive
feedback effects, which exist when both electrons and holes produce secondary
pairs, can greatly amplify any current fluctuations. Significantly more noise is
generated if the electron and hole ionization rates (a,{:J) are equal than if only
- 10 one carrier produces secondary pairs~ It is therefore highly desirable to have a
detector in which the multiplication process is dominated by one carrier type.
Unfortunately, most III-V materials have cxRl (J. The multilayer structure
described and analyzed in this chapter is expected to have impact ionization
which is dominated by electrons and therefore would be of importance for low
noise applications.
- 11 References for Cb.apter 1
1. R.N. Hall, G.E. Fenner, J.D. Kingsley, T.J. Soltys, and R.0. Carlson "Coherent
Light Emission from GaAs Junctions", Phys. Rev. Lett. 9 , pp366-368 ( 1962).
2. M.I. Nathan, W.P. Dumke, G. Burns, F.N. Dill, and G.J. Lasher "Stimulated
Emission of Radiation from GaAs PN Junctions", Appl.Phys. Lett. 1 , pp 62-64
(1962).
3. M. Holonyak and S.F. Bevacqua "Coherent (Visible) Light Emission from
Ga(As 1-xPx) Junctions", Appl. Phys. Lett. 1 , pp 82-83 ( 1962).
4. T.M. Quist, R.H. Rediker, R.J. Keyes, W.E. Krag, B. Lax, A.L. McWhorter, and H.J.
Zeigler "Semiconductor Masers of GaAs" Appl. Phys. Lett. 1 , pp 91-92 (1962).
5. K. Saito, N. Shige, T Kajimura, T. Tsukada, M. Maeda, and R. Ito "Buried
Heterostructure Lasers as Light Sources in Fiber Optic Communications",
Technical Digest Integrated Optics and Optical Communications (IOOC Tokyo
1977) p 65.
6. D. Botez "CW High Power Single Mode Operation of Constricted Double
Heterojunction AlGaAs Lasers with a Large Optical Cavity", Appl. Phys. Lett.
36' pp190-192 (1980).
7. M. Nakamura, K. Aiki, N. Chinone, R. Ito, and J. Umeda, '1.ongitudinal-Mode
Behaviors of Mode Stabilized AlxGa 1 -xAs Injection Lasers", J. Appl. Phys. 49 ,
pp 4644-4648 (1978).
8. H.C. Casey, and M.B. Panish Heterostructure Lasers Academic Press, New
York 1978.
9. H. Kressel and J.K. Butler Semiconductor Lasers and Heterojunction LEDs
Academic Press, New York, 1977.
- 12 10. G.H.B. Thompson Physics of Semiconductor Laser Devices John Wiley and
Sons, Chichester, England 1980.
11. D. Botez and G.J. Herskowitz ''Components for Optical Communications
Systems: A Review" Proc. IEEE 68, pp 689-731 (1980).
12. T. Miya, Y. Terunuma, T. Hosaka, and T. Miyashita ''Ultimate Low-Loss Single
Mode Fibers at 1.55 µm" Electron. Lett. 15, pp 106-108 (1979).
13. A.
Yariv
"Active
Integrated
Optics"
Proc.
Esfahan
Symposium
on
Fundamental and Applied Laser Physics" M.S. Feld, A. Javan, and N.A. Kurnit
Eds. Wiley Interscience, New York, 1973 pp 897-919.
14. H. Blauvelt, N. Bar-Chaim, D. Fekete, S. Margalit, and A. Yariv ''.AlGaAs Lasers
with Micro-cleaved Mirrors Suitable for Monolithic Integration" Appl. Phys.
Lett. 40 , pp 289-290 ( 1982).
15. R. Bartolini. A. Bell. and F. Spong 'Diode Laser Optical Recording Using
Trilayer Structures" IEEE J. Quantum Electron. QE-17, pp 69-77 (1981).
16. N. Chinone, K. Saito, K. Aiki, and N. Shige "Highly Efficient (GaAl)As Buried
Heterostructure Lasers with Buried Optical Guide", Appl. Phys. Lett. 35 , pp
513-516 (1979).
17. W.T. Tsang, and R. Logan, "GaAs-AlxGa 1 _xAs Strip Buried Heterostructure
Lasers", IEEE J. Quantum Electron. QE-15, pp 451-469 (1979).
- 13 Chapter II
Lasers with Micro-cleaved Mirrors
2.1 Introduction
The conventional method of obtaining optical feedback in a semiconductor
laser is to cleave opposite facets of the substrate. Although this provides nearly
perfect mirror surf aces, this method is undesirable for many applications. For
the fabrication of optoelectronic integrated circuits (OEICs), it is necessary to
be able to fabricate lasers on relatively large substrates so that there is
sufficient area for fabricating other optoelectronic devices. If opposite cleaved
facets of the substrate are used as mirrors then the number of optoelectronic
devices that can be integrated onto the chip with the laser is severely restricted
by the laser cavity length, which is typically 300 µm. For other applications, it is
desirable to fabricate lasers with very short cavity lengths.
However, it is
difficult to fabricate lasers with cavity lengths less than 100 µm by cleaving
through the substrate.
In this chapter, a new fabrication process is described in which laser mirrors
are cleaved without cleaving through the substrate. These lasers, which will be
referred to as having micro-cleaved mirrors, have threshold currents and
differential quantum efficiencies which are essentially identical to those of
lasers with conventionally cleaved mirrors. However, this process makes possible
the fabrication of lasers on substrates which are not limited by the laser cavity
length. These lasers can therefore be fabricated on large area substrates which
makes possible the integration of many electronic devices onto the same chip
for applications such the integrated optical repeater 1 • Short cavity length lasers
can also easily be fabricated using this process.
- 14-
Many other techniques for fabricating lasers that do not rely upon opposite
cleaved facets of the substrate for mirrors have been previously reported. The
techniques which have received the most attention are mirrors formed by
etching2 and the use of distributed Bragg reftectors 3 • The difficulty with etched
mirrors is due to the fact that any irregularities in the mirror surface must be
much smaller than the wavelength of the laser light to obtain a high quality
mirror. Using chemical etching it is both difficult to obtain vertical surfaces
and almost impossible to reduce the irregularities to sizes small compared to
the wavelength of light, due to limitations in the photolithograph ic processes
involved. Reactive ion etching 4 is an alternative which gives vertical surfaces, but
the irregularities tend to be even larger than is the case for chemical etching.
Similar technical problems are involved in the fabrication of lasers with
distributed Bragg reflectors. Irregularities in the reflectors result in unwanted
scattering of the laser light. Other techniques for fabricating lasers that do not
rely upon opposite cleaved facets of the substrate for mirrors include lasers
with grown mirrors 5 , lasers with ion milled mirrors 6 , and lasers with curved
cavities having mirrors on the same cleaved facet or on a cleaved corner7 . For
all of these approaches, the resulting lasers have been found to have
significantly higher threshold
currents and significantly lower quantum
efficiencies than lasers with conventionally cleaved mirrors.
2.2 Fabrication of Lasers with Micro-cleaved Mirrors
The basic approach to obtaining micro-cleaved mirrors is to selectively etch
underneath the double heterostructure , leaVing a cantilever structure as is
shown in figure 1. This thin cantilever of GaAs-AlGaAs is quite fragile and will
easily break when subjected to mechanical stress. The idea of micro-cleavage is
to break this cantilever in such a way as to produce a high quality cleaved
- 15 -
Met al Contact
Si02 - - - -
GaAs---AlxGa 1_xAs -~~~~~
Active Layer
GaAs Buffer
AIYGa 1_YAs
GaAs Substrate
Fig. 1
Schematic diagram of a cantilever prior to micro-cleavage
Micro-cleaved Mirror
Fig. 2
Schematic diagram of a laser with a micro-cleaved mirror
- 16 mirror. It was found that ultrasonic vibrations were a particularly convenient
way in which to cleave the cantilevers. By using ultrasonic vibrations, all the
mirrors on a wafer can be cleaved simultaneously, making the process of microcleavage well suited to batch processing. Microcleavage enables the fabrication
of lasers with cleaved mirrors on a chip which is not restricted in size by the
laser cavity length, as is shown in fig. 2. The remaining area of the chip can then
contain other optoelectronic devices.
The method used to fabricate the lasers described in this chapter was to grow
the double heterostructure on top of a layer of .A1yGa 1-yAs of high aluminum
content and to subsequently selectively etch this layer. The etchant that was
used to etch the high aluminum content layer was concentrated HCl.
Concentrated HCl at room temperature will etch layers with aluminum
concentrations greater than 0.6 without significantly attacking the other AlGaAs
layers. For y=0.8 the etch rate was approximately 1.5µm/min. To obtain the
structure shown in fig. 1, 25 µm wide channels were first etched down to the
AlyGa 1 -yAs layer using a nonselective etch H2 S04 :H 2 0 2 :H 2 0 (1:8:8). The high
aluminum content layer was then selectively etched in HCl until the double
heterostructure was undercut by 20-25 µm. Next 1 :8:8 was used to form a series
of 20 µm cantilevers from the overhanging double heterostructure . This
sequence of etching steps is illustrated in fig. 3. Figure 4 shows a scanning
electron microscope (SEM) photograph of a cantilever prior to micro-cleavage.
The cantilevers were then cleaved using ultrasonic vibrations. Figure 5 is an
SEM photograph of a micro-cleaved mirror. Micro-cleavages were typically found
to have small terraces ( < 100 A), but these terraces did not significantly affect
the performance of the lasers.
Oxide stripe lasers with 7 µm wide stripes were fabricated with micro-cleaved
mirrors. For 150 µm laser cavity lengths, threshold currents for devices with
N+ GoAs
N Alo,6 Goci4As
(c)
Fig. 3
(b)
Processing steps in the fabrication of micro-cleaved mirrors
a) Layers grown for lasers with micro-cleaved mirrors
b) Structure arter selective etching of high aluminum content layer
c) Structure after etching of cantilevers
d) Structure after cantilevers have been micro-cleaved
(a)
~::--t==============::t:f.
P AioAGao.As
(d)
"""'
-....J
- 18
- 19 -
either two micro-cleaved mirrors or one micro-cleaved and one conventionally
cleaved mirror were 80-120 mA. The differential quantum efficiencies were
typically 15-20% per facet. Results obtained for lasers having micro-cleaved
mirrors were no different from the results obtained for lasers of identical
structure having conventionally cleaved mirrors.
After being subjected to ultrasonic vibrations, approximately 50% of the
cantilevers cleaved properly. The main reason for failure of some of the
cantilevers to cleave satisfactorily can be attributed to irregularities in the
undercut edge. When the edge of an undercut was parallel to the cleavage plane,
the cantilevers almost always cleaved properly.
There are undoubtedly many other approaches to obtaining micro-cleaved
mirrors. For instance the position of the cleave can probably be determined by
scribing the top surface of the cantilever rather than relying on the edge of the
undercut. It is also possible to form double heterostructure cantilevers without
growing an AlGaAs layer of high aluminum content. In this case the double
heterostructure is undercut by selectively etching the GaAs substrate with H2 0 2
(pH=7). Micro-cleaved mirrors have also been fabricated in this case although
the yield was not as good as the case in which a high aluminum content layer
was used.
2.3 Optoelectronic Integrated Circuits
One of the major areas of application of lasers with micro-cleaved mirrors is
for
optoelectronic
integrated
circuits
(OEICs),
because
micro-cleavage
eliminates the restriction of the chip size to the laser cavity length. The
fundamental requirements of OEICs are a material system and fabrication
technology capable of integrating light sources, detectors, and electronic devices
onto a single chip. The ternary system (AlGa)As best meets these requirements,
- 20 at the present time. This system has the following properties:
1. The ternary (AlGa)As is very nearly lattice matched to GaAs over the entire
alloy composition range.
2. The alloy has a direct bandgap over the range from GaAs (Eg=l.43eV) to
A1. 45 Ga. 55As (Eg=1.95eV). Properties 1 and 2 are essential requirements for
the fabrication of efficient light sources and matching detectors for the
wavelength range of 0.7 µ,m to 0.88 µ,m.
3. Semi-insulating substrates, which are doped with Cr are readily available.
This facilitates
isolation of electronic devices and reduces parasitic
capacitances in integrated electronic circuits improving the high frequency
characteristics.
Because of these properties well developed technologies have evolved for the
fabrication of electronic and optical devices in this material system. However,
they have evolved almost completely independent of one another and are
therefore in many instances not compatible with one another. Micro-cleavage
enables the elimination of one of the major areas of imcompatibility, namely the
restriction of the chip size to the laser cavity length for lasers with
conventionally cleaved mirrors. Elimination of this restriction enables OEICs
such as the hypothetical chip shown in figure 6. In this chip a detector has been
integrated with the laser to monitor the light output of the laser, a V-groove has
been etched to align a fiber to the laser, and space remains for electronic
circuits such as a multiplexor and laser driving circuitry.
Optical Fiber
Fig. 6
.r
Sem-I nsulat ing GaAs
___ .../ /
L. ___
Space For
/ Go As Electr onic/
Circui ts /
--- -/'
,-/
Hypoth etical optoele ctronic integra ted circuit ilncorp orating a
laser with a micro-c leaved mirror.
_..
Loser Diode
Optical Detector
- 22 -
2.4 Short Cavity Lasers
When laser mirrors are fabricated by the conventional method of cleaving
through the substrate it is difficult to fabricate lasers with cavity lengths less
than 100 µm. Chips of such small dimension are also very difficult to handle for
mounting on heat sinks. However, using micro-cleavage very short cavity lengths
can be easily obtained without encountering the difficulties associated with
small chips. Laser cavity lengths as short as a few microns should be possible
using micro-cleavage. Such short cavity lengths are of interest for the
fabrication of extremely low threshold lasers and for the fabrication of single
frequency lasers.
For some applications of semiconductor laser diodes it is desirable to have
the laser operate in a single longitudinal mode. Many index guided laser
structures such as the buried heterostructure8 and the transverse junction
stripe 9 lasers operate in a single longitudinal mode for intermediate power
levels. However, at both low and high power output levels and during high
frequency modulation most lasers tend to exhibit multi-longitudinal mode
operation. The number of longitudinal modes is usually small, even in extreme
operating conditions, with the total bandwidth generally being less than 10 A.
The spacing between longitudinal modes of a laser diode is given by
"A
o"A=------
ZL[-n_en___
dn_e_n]
"A
(6.1)
dA.
where nen is the effective index of the transverse mode and Lis the laser cavity
length. For GaAs active layers with A.=8800 A , reducing the cavity length to 50
µm will increase the longitudinal mode spacing to oA.~ 15A. With such large
mode spacings longitudinal mode operation might be expected even for extreme
operating conditions.
- 23 The second area of application for short cavity length lasers is low threshold
current lasers. The threshold current density of a GaAs laser diode 10 can be
expressed as:
Jth(A/cm 2 ) =4.5x10 3 dlr]+(20d/7]r) [ a 1+( ltL)ln( 1/R)]
(6.2)
where d is the active layer thickness in microns, 7J is the internal efficiency, r is
the confinement factor, a 1 is the sum of the internal losses, Lis the laser cavity
length, and R is the mirror reflectivity. The first term of (6.2) represents the
current density which is required to pump the active layer to transparency and
the second term is the additional current required to provide gain to
compensate for the internal and mirror losses. The threshold current density
versus laser cavity length is plotted in figure 7 for various mirror refl.ectivities.
Uncoated laser facets have refl.ectivities of approximately 32%, but refl.ectivities
as high as 81% have been reported for facets coated with a six layer Al 2 0 3 /Si
coating 11 . Recently, short cavity length buried heterostructure lasers with
micro-cleaved lasers have been fabricated by Levine et al 12 at the Bell
Laboratories following the technique just described, which was developed at Cal
Tech. For laser cavity lengths of 40 µm threshold current as low as 7 m.A were
achieved. As can seen from fig. 7, for a laser cavity length of 40 µm, significant
reductions in the threshold current can be expected by increasing the mirror
reflectivity. Thus, much lower threshold currents, perhaps less than 1 mA.
should be possible for micro-cleaved lasers. The lower limit to the threshold
currents that are achievable with short cavity lasers will ultimately be
determined by the mirror refl.ectivities that can be achieved. The threshold
current of a GaAs laser will decrease in proportion to the cavity length, if the
mirror reflectivity is adjusted so as to keep (
)In( ~ ) a constant. Short cavity
length lasers with mirrror reflectivities of 80% (ln( ~) = 0.22) can therefore be
- 24 -
d=0. 2µm
f=0.6 7
-E
-I
a·=
20cm
77=1.0
~ 3
<(
__..
.......
R = 0.3
R =0.5
(/)
Q)
.......
Q)
:J
--00
..c
(/)
Q)
..c
I-
50
100
150
200
250
Laser. Cavity Lengt h (µm)
1'..Je. 7
Calculate d threshold current vs. laser cavity length
300
- 25-
expected to have threshold currents that are approximately 5 times lower than
lasers with mirror reflectivities of 32% (ln( ~) = 1.14). Improvements in the
fabrication of high reflectivity laser diode mirrors could result in even further
reductions in the threshold currents.
In conclusion, lasers with micro-cleaved mirrors have been fabricated which
have threshold currents and quantum efficiencies which are comparable to
those of lasers with conventionally cleaved mirrors. The technique of microcleavage makes possible the fabrication of lasers on substrates which are not
restricted in size to the laser cavity length. Very short cavity length lasers which
have the potential for both very stable single longitudinal mode operation and
very low threshold currents, can also be fabricated using the technique of
micro-cleavage. Finally, the technique has the advantage of allowing the
simultaneous formation of all of the mirrors on a wafer, making this technique
desirable for the batch processing of lasers.
- 26 -
References for Chapter Il
1. M. Yust, N. Bar-Chaim, S.H. Izadpanah, S. Margalit, I. Ury, D. Wilt, and A. Yariv,
"A Monolithically Integrated Optical Repeater", Appl. Phys. Lett. 35 , pp 795797 (1979).
2. J.L. Merz and R.A. Logan '1ntegrated GaAs-AlxGa 1-xAs Injection Lasers and
Detectors with Etched Reflectors", Appl. Phys. Lett. 30, pp 530-533 (1977).
3. W. Ng, H.W. Yen, A. Katzir, I. Samid, and A. Yariv, "Room Temperature
Operation of GaAs Bragg-Mirror Lasers", Appl. Phys. Lett. 29 , pp 684-686
( 1976).
4. E.L. Hu and R.E. Howard "Reactive Ion Etching of GaAs and InP Using
CC12 F2 /ArAJ 2 ", Appl. Phys. Lett. 37, pp 1022-1024 (1980).
5. F.A. Blum, K.L. Lawley, and W.C. Holton, "Monolithic Ga 1 -xinxAs Mesa Lasers
with Grown Optical Facets", J. Appl. Phys. 46, pp 2605-2611 (1975).
6. Y. Suematsu, M. Yamada, and K. Hatashi, "A Multi-Hetero-AlGaAs Laser with
Integrated Twin Guide", Proc. IEEE 63, p 208 (1975).
7. I. Ury, S. Margalit, N. Barchaim, M. Yust, D. Wilt, and A. Yariv "Whispering
Gallery Lasers on Semi-Insulating GaAs Substrates", Appl. Phys. Lett. 36 , pp
629-631 (1980).
8. K. Saito, R. Ito, "Buried Heterostructure AlGaAs Lasers" IEEE J. Quantum
Electron. JQE-16 ,pp 205-215 (1980).
9. H. Kumabe, T. Tanaka, H. Namizaki, M. Ishii, and W. Susaki, "High
Temperature Single Mode CW Operation with a Junction Up TJS Laser",
Appl.Phys. Lett. 33 , pp 38-40.
10. H.C. Casey, and M.B. Panish Heterostructure Lasers, chapter 7, Academic
Press, New York 1976.
- 27-
11. M. Ettenberg, D Botez, D. Gilbert, J. Connolly, H.V. Kowger, The Effect of Facet
Reflectivity on the Spectrum of Single-Mode CW Constricted DoubleHeterojunction Diode Lasers", IEEE J. Quantum Electron. QE-17 , pp 22112214 (1981).
12. B. Levine, J.P. Van Der Ziel, R. Logan, C.G. Bethea, 'High Quantum Efficiency
Low Threshold Micro-cleaved AlxGa 1-xAs Lasers", Electron. Lett. 18 , pp 690-
691 (1982).
- 28 -
Chapterm
High Power Single Mode AlGaAs Lasers
3.1 Introduction
Designing and fabricating high power single mode AlGaAs lasers consists
primarily of identifying the factors that limit the output power of AlGaAs lasers,
and when possible, altering the design to avoid these limitations. Many factors
can influence the reliability of a laser in long term operation. However, for short
term operation, the output power of an AlGaAs laser is usually limited by either
catastrophic optical mirror damage (COMD) or heating.
The active layer of an AlGaAs laser is typically absorbing in the immediate
vicinity of the mirrors due to non-radiative surface recombination of carriers at
the mirrors. The interface between GaAs and air contains a large density of
surface states at the cleaved mirror facets. Absorption of laser light by the end
sections results in the heating of the active layer near the mirrors which makes
these sections even more absorbing. Above a critical optical power density,
thermal runaway results, causing the active layer to melt at the mirrors. The
catastrophic optical mirror damage occurs at a threshold intensity of
approximately 15 mW/µm 2 for CW operation and 70 mW/µm 2 for 100 nsec.
pulses 1 . However, it is possible to avoid COMD by fabricating "window" lasers in
which the active layers are terminated short of the mirror facets. These lasers
have window sections located at the mirrors which are transparent to the laser
light. A technique for fabricating window lasers, first demonstrated by Yonezu et
al. 2 consists of diffusing zinc into the active layer everywhere except near the
mirrors. The zinc diffusion reduces the bandgap of the active layer shifting the
laser output to a longer wavelength which is not absorbed in the active layer
near the mirrors, where no zinc diffusion takes place. Using this technique the
available pulsed output power from a stripe geometry laser has been increased
- 29 -
by a factor of 10. A second technique, which was developed at Caltech involves
eliminating, by selective etching, the active layer in the vicinity of the mirrors 3 .
In this case, only transparent AlGaAs layers extend to the laser mirrors. These
two types of window lasers are shown in fig. 1.
The selectively etched window laser structure, fabricated in the process of the
research reported here, consists of a 200 µm long center active section, and two
25 µm long passive window sections in which the active layer has been removed
by selective etching. The optical power is mostly contained in the Al. 22 Ga. 78 As
optical guide layer in both the active and window sections. This enables low loss
coupling of the laser light between the sections. This approach has been used to
fabricate buried heterostructure window lasers 3 which have threshold currents
and quantum efficiencies which are nearly identical to those of conventional
buried heterostructure lasers, but can be operated without degradation at
pulsed power outputs up to three times the catastrophic damage threshold of
otherwise identical lasers without windows. The power output was limited by
heating of the laser during the pulse. By improving the thermal properties of
these lasers, still higher power outputs can be expected. This laser will be
discussed in more detail in chapter 4.
The second important factor for high power operation is designing lasers to
minimize the heating of the laser under high power operation.
The power
dissipation of a laser can be expressed as follows
(3.1-1)
(3.1-2)
where R9 is the series resistance(primarily contact resistance), Vi is the junction
voltage, 1J .is the differential quantum efficiency, Ith is the threshold current, Eg is
- 30 -
WINDOW
REGION
ACTIVE REGION
" -~
...___,... Al_ Ga. 7 As
N-+ Al. 06 Ga.94 As
/:
ACTIVE LAYER
CLEAVED
MIRROR
p• Zn DIFFUSED STRIPE
fig 1 a) Zinc diffused window laser
PASSIVE
WINDOW
PASSIVE
ACTIVE SECTION OF LASER
WINDOW
...,.125µml
200µm --C-A_ _
r- u~
.___ r--
p x= 0.35
,___.
CLEAVED MIRROR
p x= 0.22
N-
x=0.31
P- x=0.3
N x=0.6
_..,..-x = 0.05
Zn DIFFUSION
CLEAVED MIRROR
['\._ OPTICAL
INGUIDE
x=0.3
N + Go As
P- x=0.3
Au Ge-Au
Fig. lb Selective ly etched window laser
- 31 -
the bandgap, and q is the electron charge. A more relevant property, especially
for high power operation, where the temperature rise of the active layer is
significant, is the power dissipation per unit area.
(3.1-3)
where Re is the specific contact resistance in 0 -cm2 and J is the current density
in A/cm 2 . To minimize power dissipation per unit area a laser should have a low
Jth, large TJ, and small Re.
The
threshold
current
densities
and
quantum
efficiencies of most laser structures are relatively insensitive to the stripe width.
Increasing stripe width is therefore an effective way of increasing the power
output of a laser without increasing the power dissipation per unit area.
Increasing the stripe width also increases the catastrophic optical damage
threshold. The difficulty is in obtaining stable fundamental mode operation for
wide stripes, stable meaning that the beam profile does not change as the power
level is increased.
3.2 Mode Control of Semiconductor lasers
The spatial mode properties of semiconductor lasers can be divided into two
categories:
I. Spatial Mode at Low Output Powers
II. Stability of the Spatial Mode as the laser Power is Increased
For example consider the buried optical guide (BOG) laser developed by Chinone
et al 4 which is shown in fig 2.
This laser has the following spatial mode
characteristics:
1. For stripe widths less than 2-3 µm the laser operates in the fundamental
lateral spatial mode and the output is stable as the power is increased.
2. As the stripe width is increased up to 5 µm the lasers continue to operate
Fig. 2
Al.30 Ga.1o As
Cross section of a buried optical guide laser
Ga As
Al.25 Ga.75 As OPTICAL
GUIDE
n+
Al.35 G~s 5 As
P Al. 30 Ga.70 As
Al.3o Ga.70 As
"'------~
A!o5Ga. 95 As
ACTIVE LAYER
VI
- 33 stably, but in successively higher order spatial modes.
3. Above 5 µm stripe widths the laser operates in higher order modes and the
beam profile is not stable as the power is increased.
The spatial mode at the onset of lasing is the mode which reaches the threshold
condition first, the threshold condition being that the round trip gain inside the
cavity equals the round trip loss.
for
e -i{i mz
z dependence of the laser field
the threshold condition can be written as
Power gain of mode m = 2Im~.Bm~ = aim + ( ~ )ln( Rl )
(3.2-1)
where
(3.2-2)
aim= internal losses of modem
Rm= reflectivity of modem
To obtain stable fundamental mode operation it is therefore necessary to
have only the fundamental mode reach the threshold condition. Because the
fundamental mode is best confined to the gain region, the mode gain of the
fundamental mode is generally higher than any other spatial mode, at least at
low output power levels. At high output power levels spatial hole burning 5 ·6 can
distort the gain profile in such a way that higher order modes can have mode
gains exceeding that of the fundamental mode. The simplest approach to mode
control is to obtain a preference for the fundamental spatial mode by designing
- 34 -
the laser structure so that the fundamental mode has a significantly larger gain
than the higher order modes. This is done by designing waveguides in which the
fundamental mode is significantly better matched to the gain than are any of
the other modes. The simplest way to do this is to have weak lateral waveguiding
so that the there is only one guided spatial mode or if the waveguide supports
more than one mode, so that the higher order modes have much lower
confinement factors than that of the fundamental mode. The confinement
factor of a semiconductor laser waveguide is defined to be the fraction of the
optical power that is propagating in the active layer where there is gain. For
instance the lateral waveguiding in a strip buried heterostructure laser is
weaker than that of a conventional buried heterostructure laser. This is why
the strip BH lasers developed by Tsang and Logan 7 can be operated in the
fundamental mode for active layer widths up to 5 µm (see discussion in chapter
5). Inverted strip BH lasers which have been developed recently at Caltech8 also
are weakly guiding and fundamental mode operation has been achieved in this
case for active layer widths up to 4 µm. The inverted strip BH laser structure
which will be discussed in more detail in chapter 5 is shown in fig. 3
A slightly more complicated approach is to cause the gain to possess a
nonuniform profile across the active layer. If there is only gain in the middle of
the active layer, the fundamental mode will be preferred over the higher order
modes since it has relatively more power in the center of the waveguide, and will
thus exercise a larger gain. This has been achieved in a buried heterostructure
laser to obtain fundamental mode operation for a laser with a 8 µm wide active
layer. The structure studied is shown in fig. 4 and will be referred to as a narrow
injection buried heterostructure (NIBH) laser. If the active layer of a NIBH laser
is much wider than the injecting stripe width, then a NIBH laser is expected to
behave like a gain guided stripe laser. If the active layer width is approximately
Active Layer
Fig. 3
Cross section of an inverted strip buried heterostructure laser
--------- --------- --------- --------- ---'---'
N GaAs
N Al. 50Ga. 5 ~s ----./JE:-W __.,
P A1_ 24Ga .76 As
---w
___-P'I,
P Al.37Ga •63 As
""'
AuGe-Au
p A~GOG~40As
N Aly Ga 1_ YAs
Si02
- Zn Diffusion
____:-:-..:t- Cr -Au
V1
VI
Fig. 4
A~37 G~63 As
LN
A! 2 G~ As
A~ 5 G~ 5 As
N A~ 5 G~ 5 As
Cross section of a narrow injection buried heterostructure laser
N+ GoAs
N GaAs
N- A! 37 Gq63 As
Narrow Injection Buried Heterostructure Laser
VI
°'
- 37 -
the same as the injecting stripe width then a NIBH laser should behave like a
conventional BH laser. The intermediate case where the active layer is a few
microns wider than the injecting stripe width is the most interesting case and
will be discussed in more detail in chapter 6.
Mode control can also be achieved by making the internal losses of the higher
order modes larger than that of the fundamental. One way to achieve this in
semiconductor lasers is to fabricate a leaky wavguide in the lateral direction.
Mode control in the high power constricted double heterostructure (CDH) lasers
developed by Botez 9 is achieved by incorporating such a leaky waveguide. These
lasers have been operated at higher single mode CW power levels than any other
laser structure yet developed. The structure has been operated at CW single
mode output powers of 40 mW. The output power of the high power CDH laser is
limited by heating. The greatest amount of CW single mode power reported from
a BH laser to date has been 25 mW4 . However, the BH lasers are much more
efficient than the CDH lasers and the power output has been limited not by
heating, but rather by catastrophic mirror damage. Thus incorporation of
passive window sections at the mirrors may enable greater single mode output
powers from BH lasers than can be obtained from CDH lasers.
The final factor which determines the spatial mode that lases is the mirror
reflectivity. Unfortunately, the mode reflectivities of the higher order TE-like
modes generally are higher than the mode reflectivity of the fundamental mode.
The problem of solving for the mode reflectivity and transmitted power
distribution at the boundary between a dielectric slab waveguide and a uniform
dielctric medium has been analyzed by several authors 10 - 14 . The basic approach
to solving the problem involves matching the tangential electric and magnetic
fields at the boundary. The geometry of the problem is shown below in fig. 5.
03
n2
n,
Fig. 5
REFLECTED <
GUIDED MODES :
n=I
4.NSMITTED
~IATION MODES ~
Transmitted and reflected fields at a cleaved mirror f acC't
.......>INCIDENT
: GUIDED MODE
REFLECTED
RADIATION MODES
- 39 For the problem of calculating the mode refiectivities of laser modes, the
solution method must be extended to two dimensional waveguides. Although this
makes the calculation more complicated numerically, the extension to two
dimensions does not introduce any fundamental difficulties. For simplicity, the
spatial modes of the two dimensional waveguides were solved approximately
using the effective index method 15 - 17 . If more precision is required the more
rigorous vector variational method ia-2o or the finite element method 22 ·23 can be
used. The effective index method and the vector variational technique will be
described in more detail at the end of this chapter.
The mode refiectivities of BH lasers have been calculated following the
method used by Gelin etal. 14 to analyze the refiectivities of dielectric slab
waveguides. This method uses an iterative approach to converge on the solution
for the mode refiectivities and transmitted power distribution. For TE-like
modes the dominant field components are Ey and Hx. The requirement that
these tangential field components be continuous across the interface can be
written as follows.
aiEi + l:bnEn
=jc(p)E(p)dp
.BiaiEi -2:.BnbnEn = j,B(p)c(p)E(p)dp
(3.2-3)
(3.2-4)
where
ai = amplitude of the incident mode
bi = amplitudes of the reflected modes
c(p) = amplitude of the transmitted radiation modes (plane waves in this calculation)
.Bn = propagation constant of mode n
p=transverse propagation constant of the radiation modes
,B(p) = (ko 2_p2)*
- 40 -
i\= laser wavelength in free space
and the relation between the tangential electric and magnetic fields has been used.
(3.2-5)
To calculate the mode refiectivities, zeroth order estimates of bn are first made.
Using these values of bn, equation (3.2-3) is used to calculate c(p). Next using
equation (3.2-4) and c(p), first order estimates of bn can be calculated which are
then used in equation (3.2-3) to obtain a refined estimate of c(p). This process is
continued until the desired accuracy is obtained. Typically, 5 iterations are
sufficient. Some of the results obtained for BH lasers are shown in fig 6-9.
In fig 6-9, the structure on which the calculations are based is shown in part
(a). Part (b) of the figures shows the calculated mirror reflectivities of the four
lowest order transverse modes. Part (c) shows the lateral confinement factor of
the modes which is defined as
(3.2-6)
Assuming the internal losses of all the modes are the same, the spatial mode in
which the laser oscillates is determined by the mode refiectivities and the mode
confinement factors. Neglecting internal losses, the threshold gain is given by:
(3.2-7)
where
3.0
4.0
J:a:::
6~ 55t-
Li:
t3
_J
a:
1.0
~45
3.0 (b) 2.0 M•O 3.0 (d) ACTIVE LAYER WIDTH (µ.m) Fig. 6 a) Cross section of a buried heterostruc ture laser (c) ACTIVE LAYER WIDTH (µ.m) _J 1.0 1-..JI a:::>-50 ~w Z0.4 o- ~0.6 ~60 ...... 2.0 ACTIVE LAYER WIDTH (µ.m) 4.0 4.0 55.-~~~~~~~~~~~~~~~ 1- __. 1.0 2.0 ___. (a) ~35 W40 x =0,4 j:: x=0.4 >':: r.n.!\c: x=0,4 45 ~0.8 1- a: X s0,4 >--' a:: ~>-70 1.5 2.0 2.5 3.0 3.5 (b) ACTIVE LAYER WIDTH (µ.m) 1.0 2.0 3.0 ACTIVE LAYER WIDTH (,u.m) (d) (c) ACTIVE LAYER WIDTH (µ.m) <( ...J 1.0 a:: >65 1-...J 2.0 ::t: (f)w ...J LL 80.4 ~0.6 1.0 4.0 32~L-~1.---1~.....1.~~:--~:---:6':---: 4.0 85r-~~~~~~~~~~~~~~- O<{ 4.0 -E W35.0 a:: LL ....I ;::: ~80 3.0 x •0.25 (a) x•0.25 r,n,dc: >- 1- 1- x •0.25 x•0.25 ~40.0 -1::> 3.0 4.0 5.0 6.0 o5~ 195 ...I 2.0 3.0 4.0 5.0 (b) 2.0 3.0 4.0 5.0 (d) ACTIVE LAYER WIDTH (µ.m) 1.0 M•O Fig. B a) Cross section of a buried optical guide laser (c) ACTIVE LAYER WIDTH (µ.m) 2: 185 a:: IJ.I ...I I- cnw ~~190 Z0.4 1.0 ACTIVE LAYER WIDTH (µ.m) 6.0 6.0 205..-~~~~~~~~~~~~~-, ~200 rE 0 32 a:: Li: 2.0 !J.134 x = 0.35 ::ca:: 1.0 ~ ~- (a) x•0.30 x• 0,25 2= hl x =0.05 >- 38 IJ.I ~0.6 IJ.I LL 0.8 1-- 151.01 x •0.35 x •0.35 40~~~~~~~~~~~~~~~~~~..., VJ 2101 ...J a: ...J 2.0 > 190 I...J 1- 2.0 4.0 5,0 2.0 3.0 4.0 5.0 (d) ACTIVE LAYER WIDTH (µ.m) 1.0 'M •O (b) ACTIVE LAYER WIDTH (µ.ml 1,0 M•3 Fig. 9 a) Cross section of a buried optical guide laser (c) ACTIVE LAYER WIDTH (µ.m) ~~195 80,4 Vlw ...J~200 o- ~0.6 Li: 1- 1.0 0 32 ~205 W34 LL. a: LL. 0.8 1- ~ :::::> x 0.30 h:l i= 36 --- -- (a) x =0.30 x = 0.25 , x =0.05 >- 38 ~ 1.0 I x =0.30 x •0.35 0~ 40r- 6,0 6.0 - 45 - (3.2-8) where - 46 - of uniform index of refraction. This then gives the Fourier transform of the The spatial modes of a semiconductor laser structure can be classified as either TE-like (dominant fields Ey and Hx) or TM-like (dominant - 47 - burying layer, and the stronger the lateral waveguiding. In addition, the of the four lowest order modes approach one another asymptotically as the stripe width is increased. This would tend to indicate that Buried heterostructure lasers have many characteristics which are desirable during short term operation is limited by either catastrophic mirror damage or heating of the device. Conventional BH lasers, - 48 - the buried optical guide lasers were developed by Chinone et al4 . This structure Fundamental mode operation in this can be reproducibly obtained for active layer widths up to approximately 3 µm. For an optical guide layer thickness of 1 µm and an active 49 N AJ30 Go.10 As t=.============i~- A~ 05 Go. 95 As ACTIVE LAYER N A~ 3 s Gq 65 As n+ P Al. 30 Go.70 As Ga As Fig. 10 Cross section of a buried heterostructure laser A~ 35 Gq 65 As A~ 05 Ga. 95 As ACTIVE LAYER A~ 25 Ga. 75 As OPTICAL A~ 30 Ga.70 As n+ GaAs Cross section of a buried optical guide laser Al.30 Ga. 70 As - 50 eliminate optical damage as the limiting factor for short term operation. More - 51 be preferable for high power applicatio ns to have the active layer on top of the However, window laser structures may result in an improvem ent in the reliability of BH lasers. In a conventio nal laser, even at - 52 - Appendix 3.1 Optical Waveguide Theory (A3.1-1) H=(Ht+kHz)ei(r.rt-.Bz) (A3.1-2) where k is a unit vector in the longitudinal direction and {3 is the propagation V xH = ic..ieE (A3.1-3) gives the following four equations: (A3.1-4) V xHt = ic..iµkEz (A3.1-5) ip>kxEt + kxV Ez = ic..iµHt (A3.1-6) ip>kx Ht + kx V Hz = -ic..iµEt (A3.1-7) substituting (A3.1-1) and (A3.1-2) into the equations: V ·µH = O (A3.1-8) gives the equations: - 53 - Using these expressions the wave equation for the transverse E fields can be (A3.1-10) a similar expression holds for the transverse H fields (A3.1-11) Optical waveguide theory consists of methods of obtaining solutions to these The effective index method is an extremely useful method for obtaining - 54- direction. In general, all six field components (Ex,y.zHx.y.z) of the modes of a two Semiconductor lasers generally oscillate in pseudo-TE modes because of the larger modal reflectivities of the pseudo-TE modes. With (A3.2-1) (A3.2-2) (A3.2-3) the wave equation can be written as where Ey here refers to a vector directed along the y-axis. this can be further (A3.2-5) V ·eEy = 0 (A3.2-6) but so the wave equation becomes - 55 - (A3.2-7) (A3.2-9) where F is a slowly varying function of y. Substituting this into (AS.2-8) gives (A3.2-10) The next step in the effective index method is to solve - 56 - (A3.2-14) 57 s.. CJ VJ ro >- >< ~N ---- --c ..... c:: ....,0 CJ VJ s... 58 n, n, n3 n3 n4 n4 inde x meth od is to solve for the mode s ,flll(x ), II eff e ff ~~ '-'' \' slab wave guide show n below is to a strip burie d hete rostr uctu re laser - 59Appendix 3.3 Vector Variational Method The vector variational method is an extremely powerful technique for obtaining accurate solutions to optical waveguide problems. It is, however, considerably more complex computationally, and the vector variational method is used p2 =NA> (A3.3-1) where and integration - 60- The surf ace integrals are carried out over the whole cross section of the then the variational expression can be written as (A3.3-5) this can be written in matrix form as follows: The condition that p2 is stationary is equivalent to the condition22 Thus the problem of finding solutions to the optical waveguide problem is - 61 - References for Chapter 3 Mechanism in (AlGa)As DH Lasers", Japan. J. Appl. Phys. supplement 19-1, pp ture Laser", IEEE J. Quantum Electronics QE-15, pp 775-781 (1979). Heterostructure Window Lasers" Appl. Phys. Lett. 40 ,pp 1029-1031. (1982). Heterostructure Lasers with Buried Optical Guide", Appl. Phys. Lett. 35 , pp Geometry Injection Lasers", IEEE J. Quantum Electron. QE-15 , pp 718-726 iconductor Lasers", IEEE J. Quantum Electron. QE-17, pp 453-468 (1981). Lasers", IEEE J. Quantum Electron. QE-15 ,pp 451-469, (1979). tructure Lasers", Appl. Phys. Lett. 41 ,1982 tures and Electrooptical Characteristics", IEEE J. Quantum Electron. QE-17 - 62 4466-4479, ( 1971). - 63 20. K. Morishita, and N. Kumagai. 'Unified Approach to the Derivation of Variational Expressio n for Electroma gnetic Fields", IEEE Trans. Microwave Theory - 64 - Chapter IV Many failure mechanisms of AlGaAs lasers can be attributed to the presence of This has been attributed to the greater stability of the native oxide of AlGaAs compared to the The catastrophic optical mirror damage occurs at a threshold - 65 intensity of approximately 15 mW/µm 2 for CW operation and 70 mW/µm 2 for 100 - 66 - Cr-Au CROSS SECTION OF THE ___ .,,I ;' ;' CROSS SECTION P- x= 0,3 Figure 1 a Schem atic didgram of a window Lor BH la~ ·,r structu re PASSIVE ACTIVE SECTION OF LASER 125; PASSIVE Cr-Au CLEAVED MIRROR P x= 0.22 N- x =0.3 P- a:i:0.3 Zn DIFFUSION N x= 0.6 CLEAVED MIRROR _....F--"'----------------....1.-~~ -P- 0.3 x= N+ GoAs Figu.re 1 b. Side view of a window LOC BH laser - 67 - sections from which the active layer and lower cladding layers have been removed PASSIVE WINDOW SECTION ACTIVE SECTION REFLECTED MODES'! TRANSMITTED REFLECTED' I A1_ 35Go_ 65 As ==:::>'• TRANSMITTED -----------+-.'--,---------- OPTICAL GUIDE Al_ 05Ga _95 As Figure 2. Coupling of power between the center and window sections In general this problem has to be solved by techniques such as the iterative - 68 - As is the case in determining mirror reflectivities the coupling problem is solved by (4.1) where En (i. 2 ) = Guided mode n in region 1,2 =amplitude of reflected mode i f3n(i. 2 ) =propagation constant of mode n in region 1,2 using the orthogonality relation for guided modes, the transmitted and reflected (4.3) (4.4) where the modes have been normalized to satisfy JE (1.2)E (1.2)ds = 6 (4.5) For the laser structure shown in fig. 1 with a 0.8 µm thick optical guide layer, a 0.1 the reflection and transmission coefficients are listed below. The modes which the - 69 - the fundamental mode to modes of odd order in the lateral direction. Mode % Total Power transmitted 00 95.2 transmitted 10 1.8 all other transmitted and reflected guided modes <.01 transmitted forward radiation modes 3.0 4.2 Fabrication of Buried Heterostructure Window Lasers Al 0 .6 Ga 0 .4 As layer in the unmasked sections is then removed by selective etching in unmasked sections. Fig. shows scanning electron microscope (SEM) photographs of mesas prior to the second LPE growth showing the selective removal Fig. 5 is a scanning electron micrograph of a cross section of a window section of the laser, after the second 70 - Op1ical Layer {a ) { b) Fabricatio n steps for a buried heterostr ucture lac;er - 71 - Fig. 4 - 72 - - 73 - Normally, LOC BH lasers are grown with N type optical guide layers to minimize Since the transparent Al0 .22 Ga 0 .78 As guide layer, to which most most of the optical - 74 -3: -a:: 240 210 WINDOW LOC BH ITH= IBmA a:: 0:: w 150 0.. a:: 3: 120 0.. f- 90 CATAS TROP HIC MIRROR 0.. f- 60 200 400 600 800 1000 I (mA ) Llght vs. curren t charac teristic s of a regular and window 75 The far fields typically had irregularities due to scattered light as is shown in fig 7. Irregularities in the far field - 76 - chapter have significantly improved beam quality and are compatible with this The window lasers also appear to be less susceptible to damage due to facet oxi.dation. _ -··- -45° F'igure 7. -30° __L_ ___ 15° ___J__ 1=90mA 30° IrH = 30 mA Far field pattern of a window LOC BH laser -15° WINDOW LASER I= 120mA 45° ---.] 78 - ACCELERATED FACET EROSION >- -uw WINDOW BH LASER LL w 0.75 :::> 1- :::> _J 0.50 10:: LL 0.25 1 0:: 10 15 20 25 30 BOILING TIME MIN. Degradation due to accelerated mirror oxidation - 79 References for Chapter 4 1. T. Kamejima and H. Yonezu, "Catastrophic Optical Damage Generation Mechanism in (AlGa)As DH Lasers", Japan. J. Appl. Phys. supplement 19-1 , pp - 80 - ChapterV 5.1 Introduct ion Large optical cavity buried heterostru cture (LOC BH) lasers have recently developed by Chinone et al. 1 which have many highly desirable characteri stics. These lasers have low threshold currents (8-10 mAlµm stripe By compariso n the layers of the second LPE - 81 - growth of the LOC BH lasers are grown starting from the GaAs substrate and can As was mentioned in chapter 3, the transverse spatial mode in which the laser .. N GaAs Active Layer p A~soG~40As N- AlyGa,_YAs - Zn Diffusion Si02 Cr-Au Fig. 1 Cross section of an ISBH laser L....---------------------------------~=..__-AuGe-Au N Al. 5Jia. 5 aAs P Al. 37Ga .63As '--------------' ---------~~1 00 0.50 <[ -6.0 -4.5 >I< -3.0 width L Active layer _j -1 6.0 Fig. 2 Mode profile in the lateral direction of the fundamental uo-------------~---L-----'------1..--~-----""----'---' 0:: 0.25 LL _J <[ n. _J ~0.75 f .00 I Optical guide layer t,,.;i c:i _J a::
_J Z0.4 ~0.6 Li: 1- Lt 0.8 1- ~ 1.01 2.0 3,0 M•3 x•0.5 x•0.2 5 5.0 Fig. 3 6.0 ::::::t x•0.0 5 x•0.3 5 1- ;::: 195 ~~200 :r: a:: _Jl!) 205 O<{ ~210 'e 2.0 4.0 s.o ACTIVE LAYER WIDTH (µ.m) 1.0 M•O M•3 2.0 3.0 4.0 5,0 ACTIVE LAYER WIDTH (µ.ml 1.0 6.0 6,0 215,------'"T"---.,.----- 0 31 a:: W32 LL _J I- 33 >- 34 Cros s secti on of an ISBH lase r ACTIVE LAYER WIDTH (µ.ml 1.0 x•0.3 5 x•0.3 5 ...... ~~~~~ 35r·~~~~~~~~~~~~ +:> 00 - 85 using the method described in chapter 3. Fig. 3c shows the confinement factors If these confinement factors are compared to the lateral confinement factors of a LOC BH laser with the same aluminum content in the (5.1) (5.2) where =Active layer width d = Active layer thickness - 86 wide an active layer (wide emitting area) as possible. In practice, fundamental Next a thin P A10 .6 Ga0 .4 As layer and an N AlyGa 1 -yAs layer are grown in the second LPE growth. For Fig. 4 'I - '1 P Al Go 1 As Al. 05Go •95 As N Al yGo -y As -- v Optical Guide Steps in the fabrication of ISBH lasers (c ) (a) N+ GoAs N Al. 50Go. 50 As P Al. 24Go. 76 As P Al. 37 Go. 63 As ( d) ( b) 00 - 88 - Fig. 5 Fig. 6 Optical ... ( c) GoAs Active Loyer P Al. 36Ga. 64 As N Al •36Go.64 As ( b) Steps in the fabricatio n of strip buried heterostru cture lasers (a) N+ GoAs N Al. 36Go. 64 As N Al. 10Ga.90 As P Al •36 Ga •64 As t.D 00 - 90 - purposes of comparison, the steps involved in the fabrication of strip buried ISBH lasers were found to have threshold currents and differential quantum thresholds were approximately 10.% higher. Differential quantum efficiencies - 91 stably in a single transverse and longitudinal mode. For lasers with Al. 3 Ga. 7As The temperature dependence of the pulsed threshold current for ISBH -3 0° -2 0° -10° oo W2 =4µ..m l 1h=40mA 30 ° -3 00 -2 0° -10° Fa r fi el d pa tte rn s of 20 ° 8II = 12° FWHM Fi g. 7 75 mA 90 mA CJ> BOmA 10° 20 ° 30 ° 8II =18° FWHM W2 =2 .5µ .m 11h=30mA t.D 93 5. 0- -- -- -- -- -- -- -- -- -- -, 4.0 C\J ....r-3.0 -<>- ....- y=0.3 IN BURYING LAYE R ..... LLJ a:: 2.0 ::> Cl :r: 0:: :r: ..... Cl ~ 1.0 a:: 20 40 60 80 100 120 -TEMPERATURE (°C) Temp eratur e depen dence of the pulse d thres hold curre nt - 94- lasers is shown in fig. 8. The improvem ent in the high temperatu re operating - 95 in our lab. Finally, the ISBH laser structure is compatible with a recently developed window laser technology. Fabrication of ISBH window lasers may - 96 - Reference s Heterostr ucture Lasers with Buried Optical Guide", Appl. Phys. Lett. 35 , pp 6. H. Blauvelt, S. Margalit, and A. Yariv, "Large Optical Cavity Buried Heterostr ucture Window Lasers" Appl. Phys. Lett. 40, pp 1029-1031 (1982). - 97 - Chapter VI As was mentioned in chapter 3, this provides a strong mode selecting mechanism, which makes possible fundamental mode operation for much wider A~37G~53As Layer A~ 5 G~ 5 As N A~ 5 G~ 5 As Fig. 1 Cros s s0cti on of an NIBH lase r N+ GaAs N GaAs N- A!37 G~ 63 As t::=================:=;~Active Narrow Injec tion Buried Heterostructure Laser 00 \.0 - 99 - to behave like conventi onal BH lasers. The most interesti ng case, however , is the quantum efficienc y (typicall y 40-60%) . This is due to poor optical confinem ent in the lateral direction . - 100 - For some applications of semiconductor lasers it would be desirable to combine 6.2 Fabrication of Narrow Injection Buried Heterostructure Lasers P Al. 5Go. 5 As N Al. 5Go.5 As N +GoAs (d) ( b) Fjg. 2 Sti~ps iP the FabrL:a tion of NIBll laserc; (c) (a) N GoAs P Al. 37 Go. 65 As ---.,.__ f-1 - 102 - selective etching. The selective H2 0 2 etch typically etches down twice as fast as it The results to be presented for the narrow injection BH lasers were for devices fig. 3 -20° oo 10° Lateral Far Field Pattern -10° 20° TJ = 16% I Facet FWHM = 7° Ith = 80mA Narrow Injection Buried Heteros true tu re Laser VI ...... - 104 more than one spatial mode. For sufficiently narrow gain profiles, it would be 7760 7750 Fi~:. 7760 7770'A 7750 1=120m A NISH LASER 4 Spectr 7770'A 7750 I= IOOmA SPECTRA OF AN 1110!. 7760 I= 140mA Cf1 !--' - 106 - narrow stripe gain guided lasers and the buried heterostructure lasers and these - 107 References for Chapter 6 Nash ''Mode Guidance Parallel to the Junction Plane of Double Heterostructure Lasers", J. Appl. Phys. 44, pp4696-4707, (1973). - 108 - 11. K.L. Yu. K.Y. Lau, U. Koren, T.R. Chen, and A. Yariv ''Mode Stabiliza tion Mechani sm - 109Chapter VD Narrow Stripe AIGaAs Lasers Using Double Current Confinement 7.1 Introduction One of the most widely studied classes of injection lasers are the gain guided great amount of work has been reported, both experimentally and theoretically2·3 •4 on the influence of the stripe width on the _ _ N Al_ fig. 1 Scher riatic diagr am or the narro w strip e DCC laser ture 35 Go. 65 As P + GoAs N Al. 2 Ga. 8 As P At. 35 Go. 65 As GaAs Active Layer ----------------------------------------------------- " " n +GaAs ---------- -- -- -- -- -/ ? .. _~~ --------------~ Twi n Ver tica l Stri pe Las er ...... - 111 DCC structure it was reported to be difficult to obtain injecting stripes with widths Current confinement in the narrow stripe DCC lasers is provided by an oxide solute concentration near a curved solid surface of radius R is given by the following - 112 - (7.1) LPE growth was carried out at an initial - 113 - b.O ·r-1 - 114 - Fig. 3 - 115 - temper ature of 800°C and at a cooling rate of 0.4°CA nin .. With these times betwee n 45 second s and 3 minute s. For growth times less than 45 second s the blockin g layer is too thin or is discont inuous and for growth times greater than 3 minute s, growth on top of the mesas resulte d. For growth times within this range the meltba ck and growth process was found to be very reprodu cible, provide d the etching of the mesa in the substra te prior to the growth was done reprodu cibly. One be very narrow . Injectin g stripes as narrow as 0.5 µm have been fabrica ted using Narrow stripe DCC lasers were found to have several interes ting propert ies. The results to be present ed were obtaine d for a laser with the feature s listed below. The width of the top GaAs contac t layer and the opening in the Si0 were 3 µm, the were 1 cm-3 . As is the case with other narrow stripe lasers the thresho ld current s of these lasers are for 250 per facet. For temper atures betwee n 0°C and 70°C the thresho ld curren t was found to be given by the relation Ith=Ith(0°C)exp(T/I'0 ). The charact eristic temper ature, T , for one of these lasers in the directio n paralle l to the junctio n. The lasers with thresho field Fip. 4 -20° -10° oo 10° 20° 30° Late ral far field patte rn of a narro w strin c DCC lase -30° I =150mA I=20 0mA I =250 mA I=30 0mA °' ....... ....... - 117 - patterns with the most pronounced antiguiding characteristics. The variation in the Far field patterns exhibiting leaky mode characteristics have been reported previously in narrow stripe lasers, however the (7.2) Fig. 5 n=nlr+ inli x= t/2 n=nlr+inli Gain guided dielectric slab waveguide x= -t/2 1-l co - 119 - layer and the cladding layers is very small and there is little difference between the E(x)=cos(hx) x (7.3) I xi> t12 for the even modes and x E(x)=sin(ht/Z)exp(-px) x> t/2 E(x)=-sin(ht/Z)exp(px) x< -t/2 (7.4) for the odd modes. and (7.6) pt=(ht)tan(ht/2) (7.7) pt= (ht) cot(ht/2) (7.8) for the even modes and for the odd modes. The lowest order mode of gain guided dielectric slab waveguides - 120 I.Or---------~----------, STRIPE WIDTH= 2µ.m >..,_ w 0.5 -> _J a:: 4.0 - 2.0 2.0 6.0 -Tr Cf) a... -3-rr -4-rr _____._______ ______ _____ -6.0 -4.0 - 2.0 2.0 4.0 ~-- 6.0 DISTANCE ALONG JUNCTION (µm) 121 - STRIPE WIDTH= 3µ.m I-CJ) .-z -> 1 _J a:: -9.0 -6.0 -3.0 3.0 6.0 9.0 DISTANCE ALONG JUNC TION (µm) .,,. Cf) a.. -2.,,. ·W -> f _J a:: -3Tr -4Tr -9.0 -6.0 -3.0 3.0 6.0 9.0 DISTANCE ALONG JUNCTION(µm) - 122 1.0 r----------~-----------, >- I- -en f- 0.5 -f- _J a::: -5.0 -2.5 2.5 5.0 7.5 DISTANCE ALONG JUNCTION(µ.m) -7T en a_ -37T -47T-- -------- -------- -------- ---' - 123 1.0--------------------STRIPE WIDTH= 3µ.m >- STRIPE WIDTH= 2µ.m ( /) STRIPE WIDTH= 5µ.m -tz t- ~ 0.5 _J a:: -6.0 -3.0 3.0 6.0 9.0 DISTANCE ALONG JUNCTION(µ.m) >- 2µ.m tz Cf) -w 0.5 ti _J a:: -20° -10° 10° 20° 30° FAR FIELD ANGLE - 124 - I= 2 Ith 8760 8780 8800 Wavelength ( $.) tlg. 10 Spectrum of a narrow stripe DCC laser - 125 was adjusted to give a mode gain of 40 cm- 1 , which is approximately the mode gain (7.9) exp(-p!xl) - 126 - !E!2(y)dy (7.10) The spontan eous emission into a mode with spontan eous emission factor K can be They have also been observed to operate in a large number of longitud inal modes. Finally, the techniqu e - 127 of curren t injectio n throug h a narrow stripe in the substra te appear s to be well suited to several applica tions such as laser arrays, and also to low thresho ld laser - 128 Referen ces for Chapte r 7 1. F.R. Nash "Mode Guidan ce Paralle l to the Junctio n Plane of Double Hetero structu re Lasers" , J. Appl. Phys. 44 , pp4696 -4707, ( 1973). or in Narrow Stripe Lasers" , IEEE J. Quant. Electro n. QE-15 ,pp 727-730 , (1979). Gain Guided Lasers at Thresho ld", IEEE J. Quant. Electro n. QE-18 ,pp 856-864 (1982). "Criteri a for Designi ng V-Groove Lasers" , IEEE J. Quant. Electro n. QE-17 ,pp Peterm ann. "Calcul ated Sponta neous Emissio n Factor for Double Hetero structu re Injectio n Lasers with Gain Induced Guiding", IEEE J. Quant. Diode Lasers" , Appl. Phys. Lett. 40 ,pp 305-307 (1982). P-N Junctio n", J. Appl. Phys. 49 ,pp 2629-26 38 (1978). - 129 Channels ", Appl.Phy s. Lett. 28 ,pp 234-237 ( 1976). - 130 - Chapter VIII The noise generated by an avalanche photodiode (APD) is dependent on the enhance the ratio of the ionization rates a/(i. This structure has been fabricated 131 0 GaAs (--450A) N+ P+ P+ N+ Fig. 1 a) Schematic diagram sho~ing the layer structure - 132 - p+ p- p+ n- n+ n- n- p+ n- n+ n- n- p+ n+ .._--------M ULTIPLIC ATION REGION - - ABSORPTION -t- ( f) 0 Z :::> ->LL 0::: ~tum w 0::: ( b) - 133 large to fully deplete the absorpti on and multipli cation regions. Typically , the layers Since GaAs-AlGaAs superlat tices can be fabricate d using MEE, we expect GaAs (Eg=1.43eV) and Al. Ga. As (Eg=2.0eV) to be - 134 - paths for electron s and holes respectiv ely, Ein,p are the electron and hole ionizatio n eV for holes and optical phonon mean free paths of 50 A for electron s and 40 A for holes. Fig. 3 shows a single unit cell of the structur e the position x=-12 satisfies: similarly So(L.3)-So(-O)=Eip-oEv (B.3) ~ (~)-So(D) =Eip-OEv (B.4) for devices consistin g of layers of GaAs and Al. 45 Ga. 55As OEcR:t 0.5 eV and oEvR:t 0.1 eV. w, E=E I x= -L 2 fjg. --1~ x=O GaAs Al.4 5Ga _55 As w2 E=E2 o-~--l~oe--- E=E2 x= D x=L3 ~I /\unit cell of the multilayer J\PD structure used in the AI .45Ga_ 55 As x=-L! E=E I x=L4 (.N U"1 - 136 In this calculation, all of the relevant material parameters, except for the x> -L:a P(x)= 0 (8.5) -x P(x) "'exp( Ln ) x<-11 It is assumed that the high field region is sufficiently wide so that the fraction of the - 137 - case 1: 4< W2 +D P(x)=O x< La -La P(x) "'exp( -(x-D) Lp (8.6) Ls< x< 4 case 2: ~< (W2 +D)< 4 P(x)=O x< La -Ls (8.7) La< x -Ls+(x-(W 2+D))(l- E2 ) W2+D< x< 4 Lp case 3: ~> W2 +D P(x)=O x -Ls+(x-L a)(l- E2 ) Lp (8.8) La< x< 14 Integratin g over all starting positions the electron and hole ionization rates are a "' [ Ln +Lno ( 1-exp( i::)) Jexp( -::: ) (8.9) - 138 - {3 "'(D+Lp)exp( -Ls (8.10) easel case2 case3 (8.11) 1pE2 (8.12) I-'Jlo = (8.13) The values of a/{3 calculated from the Shockley model are shown in fig. 4. The curve 10 0::: 6E=O ---------6E =0.2 V/cm 2.5 3. 5 V/cm) Ionization ratio for various values of the electric Electric Field in High Field Layers (10 3.0 4.0 1cf--~~~~~~--L--~~~~~·~_._~~~~~~__. 1o'k_ 6E = l.OxlO ~ AF=O~ ~-~---,---- - ~ Cj 10 \D iJ-1 ......, - 140 bias of 6-10 volts per cell ). A practica l detector would therefor e probably have However , several qualitati ve features of the detector design can be stated without precisely knowing the energy However , it is undesira ble for the fields to be so high that multipli cation in the AlGaAs layers becomes significa nt. To minimiz e seconda ry - 141 In conclusion a new III-V avalanche photodetector in which the multiplication - 142 - References 2. R. Chin, N. Holonyak, G.E. Stillman, J.Y. Tang, and K. Hess, "Impact Ionisation in Price, Monte Carlo Calculation of Electron Transport in Solids", in Semiconductors and Semimetals vol. 14 ch. 4. R.K. Willardson and A. Beer eds.
b) Mode refiectivities of the transverse modes of a buried heterostruc ture laser
c) Lateral confinement factor of the transverse modes of a buried heterostruc ture laser
d) Active layer threshold gain of the transverse modes of a buried heterostruc ture laser
<( 0.2
::x::
0 l.Ok2
••-n
::t:<(
Fig. 7 a) Cross section of a buried heteros tructur e laser
b) Mode reftecti vities of the transve rse modes of a buried heteros tructur e laser
c) Lateral confine ment factor of the transve rse modes of a buried heteros tructur
e laser
d) Active layer thresho ld gain of the transve rse modes of a buried heteros tructur
e laser
;:::
<(0.2
...J<.!>75
u 37.5
b) Mode reftectivi.ties of the transverse modes of a buried optical guide laser
c) Lateral confinement factor of the transverse modes of a buried optical guide laser
d) Active layer threshold gain of the transverse modes of a buried optical guide laser
1--
::c _J
36
1--
1--
3.0
4.0
5.0
6,0
i=
3.0
b) Mode reflectivities of the transverse modes of a buried optical guide laser
c) Lateral confinement factor of the transverse modes of a buried optical guide laser
d) Active layer threshold gain of the transverse modes of a buried optical guide laser
Ia:
...J
I-
.;:,.
.;:,.
W = Active layer width
d = Active layer thickness
gmth = active layer gain at which mode m reaches threshold
The transverse mode that oscillates is the mode which reaches threshold at the
lowest active layer gain. It is almost always desirable to have the laser oscillate
in the fundamental mode. For many applications, such as those which require
high output powers it is desirable to obtain fundamental mode operation for as
wide an active layer (wide emitting area) as possible. In practice, fundamental
mode operation is generally achievable in buried heterostructure lasers only for
active layer widths less than 1.5-2 microns. Buried optical guide lasers are
slightly better , but fundamental mode operation is still typically observed only
for active layer widths less than 2-3 microns. This spatial mode behavior is
consistent with the results shown in part (d) of fig. 6-9
There are several significant features of these calculated curves of mirror
reflectivities. Most significant for understanding the operation of semiconductor
lasers is that the higher order modes have greater mode reflectivities than the
fundamental mode. Although it is necessary to use an approach such as that
outlined above to calculate the mirror reflectivities accurately, these curves can
be understood qualitatively by the following simple model. In this model the
mirror reflectivities of the spatial modes are obtained by first decomposing the
incident mode into plane waves. Each plane wave component is then assumed to
be reflected at the laser mirror according to the Fresnel relations for the
reflection and transmission of light at an interface between semi-infinite slabs
reflected radiation. This in turn can be expanded in terms of the actual modes
of the waveguide with the coefficients of this expansion giving the mode
reflectivies.
fields Ex and Hy). The TE-like modes are analogous to plane waves with the
polarization perpendicular to the plane of incidence and the TM-like modes are
analogous to plane waves with the polarization in the plane of incidence. For
plane waves with the polarization perpendicular to the plane of incidence the
reflectivity increases as the angle of incidence increases. Based on this result
for plane waves, one would expect that the higher order TE-like modes of a
semiconductor laser would have larger refiectivities than the lower order modes.
T.Us is because if the modes are decomposed into plane waves, a greater
fraction of the power of the higher order modes are in the plane waves with
large angles of incidence. For the polarization in the plane of incidence the
reflectivity at first decreases as the angle of incidence increases, reaching 0 at
the Brewster's angle. The reflectivity of plane waves polarized in the plane of
incidence is less than or equal to that of the other polarization for all angles of
incidence. One would then also expect from this model that semiconductor
lasers would oscillate in the TE-like modes because these modes would be
expected to have greater reflectivities. This is inf act what is observed.
A second interesting feature of the curves shown in fig 6-9 is the fact that the
reflectivity advantage of the higher order modes is more pronounced when the
lateral waveguiding is stronger. This can be seen by comparing fig 8 and 9 which
are for structures which are otherwise identical except for the aluminum
content of the burying layer. The greater the aluminum content the greater is
the difference between the index of refraction of the active layer and the
reft.ectivities
fundamental mode operation in very wide stripe structures may be possible if
the gain is matched best to the fundamental mode, such as was done in the
narrow injection buried heterostructure laser (see chapter 6).
3.3 High Power Single Mode Buried Heterostructure Lasers
for high power single mode lasers. They have low threshold currents, are very
efficient, and operate in a single longitudinal mode even at high output power
levels and when being modulated. However, until now the power output of BH
lasers has been relatively low due to the small emitting area of most BH lasers.
In this section, the factors limiting the output power of BH lasers will be
examined and some projections will be made as to the ultimate power
capabilities of single mode BH lasers.
As mentioned in the beginning of this chapter, the power output of a
semiconductor laser
being small emitting area devices, are generally limited by catastrophic mirror
damage. For the structure shown in fig. 10, the active layer width typically is less
than 2 µm so that fundamental spatial mode operation can be achieved. The
vertical size of the mode is typically about 0.3 µm. This gives an emitting area of
only about 0.6 µm 2 . For the structure shown in fig. 10 with a 0.2 µm thick active
layer the peak intensity inside the active layer reaches the CW catastrophic
damage threshold intensity of 15 mW I µm 2 at a power output of only 5 mW. A
significant advancement in the power capabilities of BH lasers was made when
is shown in fig. 11. The most significant change is the incorporation of a four
layer large optical cavity waveguide which increased the vertical size of the
optical mode to approximately 1 µm.
structure
layer thickness of 0.08 µm, the peak intensity inside the active layer reaches the
CW catastrophic damage limit at a power output of approximately 25 mW.
To further improve the power handling capabilities of BH lasers, it is
desirable to further increase the emitting area and possibly to incorporate a
window structure. Strip buried heterostructure lasers, such as the inverted strip
buried heterostructure laser 8 shown in fig. 3 are capable of fundamental mode
operation for active layer widths up to 4 µm. Further refinements could
probably consistently enable fundamental mode operation for stripe widths up
to approximately 5 µm. In addition, the thickness of the optical guide layer can
probably be increased to approximately 1.5 µm while still maintaining
fundamental mode operation in the vertical direction. For the structure shown
in fig. 12 with a 0.08 µm thick active layer, the peak intensity in the active layer
reaches the catastrophic damage limit at a power output of approximately 50
mW. Further increases in the stripe width, while maintaining fundamental mode
operation may be possible using techniques such as narrow injection of current
into the active layer, but the effect of this on the spectral characteristics of the
lasers is unclear.
From the preceding analysis it appears that CW power outputs up to 50 mW
should be possible from BH lasers, without using a window laser structure. To
further increase the power output, transparent window sections at the mirrors
will probably be necessary. The incorporation of window sections will most likely
GUIDE
Fig. 1 ~
likely, the power output of window lasers will be limited by heating of the device.
The power level can then be expected to saturate above a certain current
density. In the high power CDH lasers 9 fabricated by RCA and the channeled
substrate planar 24 lasers of Hitachi, the power saturates at a current density of
approximately 2x10 4 A/cm 2 • For a laser stripe width of 5 µm and a laser cavity
length of 300 µm this corresponds to a current of 300 mA. A laser of these
dimensions can be expected to have a threshold current of approximately 50 mA
and a differential power output of 0.4-0.5 mW /mA per facet. If one mirror is
coated for high reflectivity the differential power output from the other mirror
can be increased by nearly a factor of two to O.B-1 mW /mA. Based on these
estimates, heating can be expected to limit the power output to
(1mW /mA)(300mA-50mA)=250 mW
Since the BH lasers are more efficient than the CDH lasers. operation to even
higher current densities should be possible for a BH laser with the same series
resistance and thermal resistance as the CDH lasers. For a series resistancestripe area product of 4x10-5 0 -cm 2 , which is the value for the RCA lasers, the
power from BH lasers can be expected to saturate at a current density of 22
kA/cm 2 rather than the 20 kA/cm 2 of the CDH laser. This increases the limit on
the power due to heating slightly to approximately 280 mW. This figure should
be viewed as an upper limit on the short term CW single mode power capabilities
of BH lasers. Since this is an order of magnitude greater power than has been
thus far achieved from BH lasers, it is difficult to predict the maximum power
level at which BH lasers can be operated at reliably. One problem with the
window BH lasers, such as shown in fig lb, is that the optical guide layer is above
the active layer. This increases the distance between the active layer and the
heat sink, which in turn increases the thermal resistance of the laser. It would
optical guide layer.
power levels below the catastroph ic damage threshold, defects are generated at
a fast rate in the active layer near the mirrors. Window structures enable the
passivatio n of the end of the active layer and may result in a decrease in the
rate of defect generation .
An understanding of the optical modes of the dielectric waveguides of various
laser structures is essential to design lasers for high power operation. In this
section some basic concepts related to optical waveguide theory are reviewed
and two numerical methods for calculating the modes of optical waveguides are
described.
The starting point for optical waveguide theory is Maxwell's equations. It is
convenient to resolve the fields into transverse and longitudinal components
E=(Et+kEz)ei(r.rt-.Bz)
constant. Substituting these expressions into Maxwell's equations
V x E = -ic..iµH,
V xEt_ = -ic..iµkHz
V ·eE = 0,
(A3.1-9)
obtained
µV x-V
xEt,-V -V
·eEt,-(c.h;µ-(J 2 )Et,=O
wave equations. Unfortunately, there are relatively few waveguides for which an
exact solution is known. Therefore, waveguide theory is comprised primarily of
numerical methods for obtaining approximate solutions. In the following section, the effective index method, and the vector variational method will be
described and the approximations included in these techniques will be analyzed.
Appendix 3.2 Effective Index Method
approximate solutions to optical waveguide problems. The effective index
method is useful primarily because it is very simple. However, because of the
many simplifying approximations that are made, the technique should not be
used when great accuracy is required. Nonetheless for most optical waveguide
problems associated with laser structures, the effective index method provides
sufficient accuracy and this method is the most widely used approximation
method for laser waveguide problems.
The effective index method is best applied to waveguides in which there is
strong optical guiding in one direction and weak optical guiding in the other
direction. This is generally true for laser structures, which typically have strong
guiding in the vertical. x, direction and relatively weak guiding in the lateral, y,
dimensional waveguide will be nonvanishing. However, for the case of strong
guiding in one direction and weak guiding in the other, the solutions can be
classified as pseudo-TE and pseudo-TM. Pseudo-TE modes have dominant field
components Ey.Hx, and Hz and pseudo-TM modes have dominant field components Hy.Ex, and Ez.
the assumption of only one dominant electric field component the wave equation can be significantly simplified. Using
The wave equation for Ey can be written as
Making the further assumption that µis constant and using the vector identity
V xV xA = V (V ·A)-V 2 A
(A3.2-4)
simplified using the relation
V ·Ey = -V
·eEy--Ey·V
Generally, laser structures are made up of layers which have uniform e. In this
case the right hand side of (A3.2-7) is zero and we arrive at the form of the wave
equation that is generally used to analyze laser waveguide problems
(A3.2-8)
In the effective index method, the solutions are taken to be of the form
Ey(x,y) = F(x,y)G(y)
G a2F + a2(FG) i k 2n2(x y)FG - R2FG =a
,. .
ay2
ax 2
(A3.2-11)
for F(x,y) and Px(Y) for all values of y, where F(x,y) and {3(y) are the solutions of
a dielectric slab waveguide (infinite in the f direction) with an index of refraction n(x,y). The values of f3x(Y) thus obtained are used to obtain an equation for
they variation.
(AS.2-12)
F and G thus satisfy
(A3.2-13)
Comparing this to (AS.2-11), we can see that the effective index method makes
the approximation:
which is an approximation which is good if F(x,y) is slowly varying in y. This, in
turn is true when n(x.y) is slowly varying in y. The entire procedure is best
explained by means of an example.
Fig 12 shows a strip buried heterostructure laser. To apply the effective index
method, the modes and effective indices of dielectric slab waveguides of infinite
width corresponding to the layers of sections I and II are first calculated. Then
the shape of the modes in the lateral direction are determined approximately by
solving for the modes of an infinite slab waveguide having indices of refraction
equal to the effective indices of sections I and II. Fig 13 illustrates this procedure.
In this case, n(x,y) is actually not slowly varying in y, but changes abruptly.
The use of the effective index is therefore not completely justified for this type
of waveguide. However for most purposes the effective index method will provide
solutions of sufficient accuracy. For laser waveguide problems, the uncertainty
associated with difficulties in precisely controlling the layers that are grown is
generally larger than the error made by using the effective index method. However, it is sometimes desirable to solve waveguide problems with more accuracy
than is possible with the effective index method. In such cases, methods such as
the vector variational method or the finite element method can be used. These
methods unfortunately require orders of magnitude more computer time and
are therefore not practical for calculations where the modes of many
waveguides need to be determined.
CJ
VJ
VJ
n2
a) The first step in the effec tive
10
wave guide s corre spon ding to
and the effec tive indic es, nen · • of diele ctric slab
s are mad e up of the layer s of
secti ons 1 and 11 of fig 12. Thes e slab wave guide
tion.
secti ons 1 and II, but are infin ite in the y direc
neff
b) Havi ng solve for ncn and neffn, the diele ctric
and the wave guide prop agati on
now solve d for the later al mod e varia tion G(y)
to be infin ite in the x direc tion
cons tant (3. In this case. the wave guide is take n
Fig. 13 Appl icatio n of the effec tive inde x meth od
only when a high degree of accuracy is required. For the research described in
this thesis, this method was only used to check the accuracy of the application
of the effective index method to laser waveguide problems. The method is based
on the variational expression for the propagation constant, which in a lossless
dielectric waveguide is given by17 :
(A3.3-2)
(A3.3-3)
waveguide and the line integrals are carried out over all the boundaries at which
the material constants µ,t change discontinuously. n is a unit vector normal to
the boundary and directed as shown in fig 14. If trial functions are chosen such
that (1/µ)V ·µHt and k(nxHt) are continuous across the boundaries between the
media then the variational expression is stationary for trial functions which
satisfy the wave equation and the boundary conditions that the tangential H
fields are continuous. This method does not yield solutions which satisfy the
boundary conditions related to the derivatives of the H fields at the boundaries.
It is most convenient to take trial functions to be of the form:
(A3.3-4)
(A3.3-6)
(A3.3-7)
reduced to an eigenvalue problem which can be easily solved using standard
numerical techniques.
1. T. Kamejima and H. Yonezu, "Catastrophic Optical Damage Generation
425-429 (1978).
2. H. Yonezu, M. Ueno, T. Kamejima, and I. Hayashi, "An AlGaAs Window Struc-
3. H. Blauvelt, S. Margalit, and A. Yariv, "Large Optical Cavity AlGaAs Buried
4. N. Chinone, K. Saito, R. Ito, K. Aiki, N. Shige, 'Highly Efficient (GaAl)As Buried
513-516 (1979).
5. R. Lang, '1.,ateral Transverse Mode Instability and Its Stabilization in Stripe
(1979).
6. S. Wang. C. Chen, A. Liao, and L. Figueroa, "Control of Mode Behavior in Sem-
7. W.T. Tsang and R. Logan, "GaAs-AlxGa 1 _xAs Strip Buried Heterostructure
8. H. Blauvelt, S. Margalit, and A. Yariv "AlGaAs Inverted Strip Buried Heteros-
9. D. Botez, "Constricted Double-Heterojunction AlGaAs Diode Lasers: Struc-
,pp 2290-2309, (1981).
10. F.K. Reinhart, I. Hayashi, and M.B. Panish. "Mode Reflectivity and Waveguide
Properties of Double-Heterostructure Injection Lasers" J. Appl. Phys. 42 ,pp
11. T. Ikegami, "Reflectivity of Mode at Facet and Oscillation Mode in Double
Heterostructure Injection Lasers", IEEE J. Quantum Electron. QE-8 ,pp 470476, (1972).
12. L. Lewin, "A Method for the Calculation of the Radiation Pattern and Mode
Conversion Properties of a Solid State Heterojunction Laser", IEEE Trans.
Microwave Theory and Tech., MTI'-23 ,pp 576-585, ( 1975).
13. T. Rozzi, and G. in't Veld 'Variational Treatment of the Diffraction at the
Facet of D.H. Lasers and of Dielectric Millimeter Wave Antennas", IEEE Trans.
Microwave Theory and Tech .. MTI'-28 ,pp 61-73, (1980).
14. P. Gelin, M. Petenzi, and J. Citerne, "Rigorous Analysis of the Scattering of
Surface Waves in an Abruptly Ended Slab Dielectric Waveguide", IEEE Trans.
Microwave Theory and Tech., M'IT-29 ,pp 107-114 (1981).
15. W. McLevige, T. Itoh, and R. Mittra, "New Waveguide Structures for Millimeter
Wave and Optical Integrated Circuits", IEEE Trans. Microwave Theory and
Tech., M'IT-23 ,p788, (1975).
16. P. Kirkby, and G.H.B. Thompson, "Channeled Substrate Buried Heterostructure GaAs-(GaAl)As Injection Lasers''. J. Appl. Phys. 47, p4578, (1976).
17. W. Streifer, and E. Kapon, "Application of the Equivalent Index Method to DH
Diode Lasers", Appl.Opt. 18 ,p 3724, (1979).
18. K. Kurokawa, "Electromagnetic Waves in Waveguides with Wall Impedance"
IEEE Trans. Microwave Theory and Tech., MTT-10 ,pp 314-320, {1962).
19. M. Matsuhara, and N. Kurnagai, "Theory of Coupled Open Transmission Lines
and Its Application".IEEE Trans. Microwave Theory and Tech., MTT-22 ,pp 378382, (1974).
and Tech., MTT-25 ,pp 34-40, (1977).
21. M. Ohtaka, M. Matsuhara , and N. Kumagai, "Analysis of the Guided Modes in
Slab Coupled Waveguide s Using a Variationa l Method".IEEE J. Quantum Electron. QE-12 ,pp 378-382, 1976.
22. C. Yeh, "Optical Waveguide Theory".IE EE Trans. Circuits and Systems, CAS-26
,pp 1011-1019 , (1979).
23. C. Yeh, K. Ha, S. Dong, and W. Brown, "Single Mode Optical Waveguide s"
Applied Optics, 18 ,pp 1490-1505 , (1979).
24. K. Aiki. M. Nakamura , T. Kuroda, J. Umeda, R. Ito, N. Chinone, and M. Maeda,
'Transvers e Mode Stabilized AlxGa 1 -xAs Injection Lasers with Channeled Substrate Planar Structure" , IEEE J. Quantum Electron. QE-14, pp 89-94 (1978).
25. see for example P. Morse and H. Feshbach, Methods of Theoretica l Physics ,
McGraw-Hill Book Co. New York, 1953.
Large Opticru Cavity Al.GaAs Buried Heterostructure Window Lasers
4.1 Introduction
the active layer at the laser mirrors. Local heating due to optical absorption in the
active layer near the mirrors can result in catastrophic mirror damage 1 at high
output powers. At somewhat lower output powers, facet erosion due to oxidation of
the active layer can occur 2 ·3 . Even at low output powers, defects are generated at a
fast rate in the active region in the vicinity of the mirrors 4 . Catastrophic mirror
damage can be avoided by making the laser structure transparent to the light
output in the vicinity of the mirrors. This has been accomplished previously by
selective Zn diffusion in stripe lasers and results in an order of magnitude increase
in the available pulsed optical power 5 . The power output of Zn diffused window lasers
is limited by catastrophic damage due to local heating in the bulk rather than at the
laser mirrors. There is also evidence that laser degradation due to facet oxidation is
greatly reduced in lasers with transparent AlGaAs mirrors 6 .
native oxide of GaAs.
Although many factors can influence the long term reliability of AlGaAs lasers, for
short term operation, the power output is generally limited either by catastrophic
mirror damage or heating. The active layer of an AlGaAs laser is typically absorbing
in the immediate vicinity of the mirrors due to surface recombination of carriers at
the mirrors. The cleaved GaAs facet contains a high density of surface states.
Absorption of laser light by the end sections results in the heating of the active layer
near the mirrors which makes these sections even more absorbing. Above a critical
optical power density, thermal runaway results, causing the active layer to melt at
the mirrors.
nsec. pulses 7 .
Buried heterostructure lasers have many characteristics which are desirable for
high power single mode lasers. They have low threshold currents, are very efficient,
and operate in a single longitudinal mode even at high output power levels and when
modulated. However, until now the power output of BH lasers has been relatively low
due to the small emitting area of most BH lasers. A conventional BH laser (without a
large optical cavity) reaches the catastrophic optical damage threshold at a power
output of only 5 mW. Large optical cavity BH lasers are better primarily because the
vertical size of the mode is substantially increased. Because fundamental mode
operation can be obtained for slightly wider stripe widths in LOC BH lasers the
lateral size of the mode is also increased. Still the CW power output is limited by
catastrophic mirror damage to approximately 25 mW. If the catastrophic damage
can be avoided then it may be possible to increase the CW power of a BH laser by as
much as a factor of 10 (see chapter 3).
In this chapter large optical cavity (LOC) buried heterostructure (BH) window
lasers are described in which only the transparent AlGaAs layers, and not the active
layer, extend to the laser mirrors. These lasers have threshold currents and
quantum efficiencies comparable to those of LOC BH lasers without transparent end
sections and have been operated, without degradation at up to three times the
power at which regular LOC BH lasers degrade by catastrophic mirror damage.
Window LOC BH lasers also appear to be more resistant to degradation due to mirror
oxidation.
The window LOC BH laser structure is illustrated in fig. 1. The laser actually
consists of two separate structures, a 200 µm central section which is a regular LOC
BH structure8 grown by liquid phase epitaxy (LPE) and 25µm transparent window
CENTER PORTION OF THE
LASER
OF LASER
AT THE MIRRORS
N+GoAs
x::Alum inum cor.ten t of the layers
WINDOW
WINDOW
P x= 0.35
x=0.05
OPTICAL GUIDE
Au Ge-Au
after the first growth and replaced by transparent AlGaAs layers that are added
during a second LPE growth. The optical mode is mostly confined to the Al 0 _22 Ga 0 78 As
guide layer in both the center and window sections, which results in low loss
coupling of the laser mode between the waveguides in the center and window
sections.
The problem of determining the coupling loss between the center and window
sections is similar to the problem of calculating mode reflectivities that was
discussed in chapter 2. In general there will be one forward traveling incident guided
mode, reflected guided and radiation modes, and transmitted radiation and guided
modes. The geometry is shown in fig 2.
RADIATION
INCIDENT
MODE
RADIATION MODES
GUIDED MODES 1
-1
,..
DES
UIDE MO
_ _ _ Al.22Go.78 A s
technique described in chapter two. However, for the case where the waveguides
which are coupled together are well matched, the radiation modes and the reflected
modes other than than the incident mode can be neglected to a first approximation.
matching the tangential E and H fields at the boundary between the two regions. For
TE-like modes the dominant field components are Ex and Hy. The requirement that
these tangential components be continuous across the boundary can be written as
follows:
KC
1 )(1+r·)
1 = I:t n En C)
(4.2)
tn =amplitudes of the transmitted modes
ri
amplitudes can be expressed as follows:
nm
µm thick active layer, a 2 µm wide active layer, and an incident fundamental mode
waveguide can support are labeled according to the order of the mode in the vertical
and horizontal directions, respectively. Because of symmetry there is no coupling of
The fabrication of the window LOC BH laser requires two LPE growths. The layers
grown in the first growth are N Al 0 .6 Ga 0 .4 As (1.5µm Te doped), undoped A1 0 .05 Ga 0 .95 As
(0.12µ,m), P Al 0 .22 Ga 0 . 78 As (0.7µm Ge doped), and P Al 0 .s 5 Gao.a 5 As (1.Dµm Ge doped).
Next, mesas are etched as is done for a regular BH laser. These first two fabrication
steps are illustrated in fig. 3. The mesas are then masked by photoresist, except for
50µm sections, which will eventually be the window sections of the laser. The
HF. This is followed by a brief nonselective etch to remove the active region as well in
the
of the bottom cladding layer in the window section. After this, a thin P-A1 0 .3 Ga 0 .7As
layer and an N- A1 0 .3 Ga 0 .7As layer are grown in the second LPE growth. These layers
grow on the sides of the unetched sections of the mesas, and both underneath and
on the sides of the selectively etched sections.
growth.
Guide
Ac1ive
N+ GaAs
Figure 3.
electron leakage. For a double heterostructure with a given difference in the
aluminum contents of the active and cladding layers, the leakage of electrons into
the P cladding layer will be greater than the leakage of holes into the N cladding
layer. If a relatively low aluminum content optical guide layer is incorporated into
the laser structure, it is therefore desirable that it be N type. When n type GaAs
substrates are used this means the active layer is on top of the guide layer. However,
it is difficult to fabricate these lasers with the active layer on top of the guide layer,
since this would require growing over the AlGaAs guide layer in the second LPE
growth. Having the active layer underneath the optical guide layer, enables the
layers of the second growth to be grown starting from the GaAs substrate. ISBH
lasers can still have N type optical guide layers if p type GaAs substrates are used.
4.3 Properties of Buried Heterostructure Window Lasers
power is confined, extends from mirror to mirror, there is very little coupling loss
between the center and window sections of the laser. Typically, our window LOC BH
lasers have threshold currents of 8-10 mA per µm stripe width of the lasers for
pulsed operation and 9-12 mA per µm for CW operation and differential quantum
efficiencies of 20-30% per facet. The best results obtained for 2 µm wide stripes were
pulsed and CW threshold currents of 15 mA and 18 mA, respectively, and a
differential quantum efficiency of 35% per facet. These results are nearly identical to
those of regular LOC BH lasers fabricated from the same wafer. The most significant
difference between the window and regular LOC BH lasers was the ability of the
window lasers to operate at high pulsed output powers without catastrophic mirror
damage. Fig. 6 shows the light vs. current characteristics of a LOC BH laser and a
window LOC BH laser for 2 µm wide devices driven by 75 nsec. pulses at a repetition
rate of 1 KHz. Window LOC BH lasers were operated up to 130 mW I µm stripe width,
LASE R
TJ = 32°/o I FACE T
a:: 180
DAMAGE FOR REGU LAR
LOC BH LASE RS
Figure 6.
LOC BH laser
which is three times the power at which regular LOC BH lasers fabricated from the
same wafer failed due to catastrophic mirror damage. The output power of the
window lasers was limited by heating and not by mirror damage.
To further increase the output power of window BH lasers it would be desirable to
both increase the laser stripe width and improve the thermal characteristics of the
laser. By going to a strip BH structure (see for instance chapter 5) the stripe width
can be roughly doubled while maintaining fundamental mode operation. Much more
important, however, for increasing the power output would be to decrease the series
resistance of the devices which results primarily from the contact resistance. Since
BH lasers have in the past been primarily low current devices, the contacts have
been relatively unimportant. The usual fabrication procedure results in the P
contact being made to the upper cladding layer which typically has an aluminum
content of 0.35. Much lower contact resistances are possible for contacts made to
GaAs layers. Therefore, significant reductions in the power dissipation of the lasers
can be expected if the fabrication process is modified to allow GaAs contacts. Such a
modification should be possible with only a slight increase in the fabrication
complexity. The main difficulty is that it is normally undesirable to have growth on
top of the mesa during the second growth and growth will occur on top of GaAs, but
not on Al. 35 Ga.a5As.
For stripe widths of less than 2 µm , operation of the window LOC BH lasers in the
fundamental transverse mode can be obtained.
typically are present in regular BH lasers as well due to scattering from the sidewalls
of the laser 8 . In window LOC BH lasers scattered light can also result from losses in
the coupling of the laser light between the center and window sections. However, it
could not be determined whether coupling losses were significantly affecting the
beam quality. The inverted strip BH lasers which will be described in the next
window laser technology.
The effect of mirror oxidation on the laser performance has also been examined
by oxidizing the mirrors in boiling water. At regular intervals during this accelerated
mirror oxidation test the pulsed laser characteristics were measured. Lasers
without windows were found to degrade approximately four times as fast as window
lasers, as is shown in fig. B. One possible explanation for this is that the oxide that is
grown on the AlGaAs window sections is more stable than the oxide grown on the
active layer of lasers without windows.
In conclusion window LOC BH lasers with transparent waveguides at the mirrors
have been fabricated. These lasers have threshold currents and quantum efficiencies
that are nearly identical to those of regular LOC BH lasers, but have been operated
at pulsed power outputs which are three times the level at which regular LOC BH
lasers degrade by catastrophic mirror damage. Even greater output powers may be
possible since the power is currently limited by heating of the window LOC BH lasers
rather than catastrophic damage.
FAR FIELD
---.]
DUE TO BOILING WATER
LL
LL
Figure 8.
425-429 (1978).
2. T. Yuasa, M. Ogawa, K. Endo, and H. Yonezu, "Degradation of (AlGa)As DH Lasers
Due to Facet Oxidation" Appl. Phys. Lett. 32, pp 119-121 (1978).
3. Y. Shima, N. Chinone, and R. Ito, "Effects of Facet Coatings on the Degradation
Characteristics of GaAs-Ga 1-xAlxAs DH Lasers" Appl. Phys. Lett. 31 , pp 625-627
(1977).
4. F. R. Nash, R.L. Hartman, N.M. Denkin, and R.W. Dixon "GaAs Laser Reliability and
Protective Facet Coatings" J. Appl. Phys. 50, pp 3123-3132 ( 1979).
5. H. Yonezu, M. Ueno, T. Kamejima, and I. Hayashi, "An AlGa.A.s Window Structure
Laser", IEEE J. Quantum Electronics QE-15, pp 775-781 (1979).
6. S. Takahashi, T. Kobayashi, H. Saito, and Y. Furakawa, "GaAs-AlGaAs DH Lasers
with Buried Facet", Japan J. Appl. Phys. 17, pp 865-870 (1978).
7. N. Chinone, K. Saito, R. Ito, K. Aiki, N. Shige, ''Highly Efficient (GaAl)As Buried
Heterostructure Lasers with Buried Optical Guide", Appl. Phys. Lett. 35, pp 513516 (1979).
8. K. Saito and R. Ito ''Buried Heterostructure AlGaAs Lasers" IEEE J. Quantum
Electron. QE-16 ,pp 205-215 (1980).
AlGaAs Inverted Strip Buried Heterostr ucture Lasers
been
width), high differentia l quantum efficiencie s (60-80%), and operate stably in a
single transverse and longitudin al mode. However, fundamen tal transverse mode
operation is achievable only for relatively narrow stripe widths. The maximum
allowable stripe width is typically between 2 and 3 microns depending on the
exact compositi on of the layers. As a result the power output of these lasers is
limited by catastroph ic mirror damage to approxim ately 20 mW CW and 80 mW
for 100 nsec pulses. Scattering of light by irregularit ies in the sidewalls of the
waveguide can also result in a significan t degradati on of beam quality. This is of
particular importanc e for lasers with stripe widths less than 2 microns 2 . A
second important type of buried heterostru cture laser is the strip buried
heterostru cture (SBH) laser developed by Tsang et al. 3 These lasers can be
operated in a stable fundamen tal mode for strip widths up to 5 microns and are
much less susceptibl e to degradati on of beam quality due to scattered light.
However, fabricatio n of this laser requires epitaxial growth on top of an
Al:"Ga 1 -xAs guide layer. Initiation of growth on air exposed AlGaAs layers of
aluminum content greater than 0.1 is very difficult due to the formation of an
oxide layer on the exposed AlGaAs surface. To obtain good growth on top of the
guide layer requires a precisely controlled meltback prior to the second liquid
phase epitaxial (LPE) growth.
be simply and reliably grown without a melt back. In this chapter the fabrication
of inverted strip buried heterostructure lasers (ISBH), which combine many of
the desirable features of the LOC BH and SBH lasers, is described. A detailed
comparison between the fabrication steps of the ISBH and SBH lasers is also
included.
The ISBH laser structure is shown schematically in fig. 1. The structure
resembles that of the LOC BH laser except the active and the lower cladding
layers are narrower in width than the Al. 24 Ga.76As optical guide layer. This
results in two significant changes in the modal characteristics of the laser. First,
the laser mode is isolated from the sidewalls of the optical guide layer, in which
most of the power propagates. This greatly reduces irregularities in the far field
patterns of the lasers due to scattered light. The lateral near field intensity of
an ISBH laser, calculated using the effective index method (see chapter 3) is
shown in fig. 2. As can be seen most of the power is confined within the width of
the optical guide. Second, the confinement of the mode in the lateral direction
by a weakly guiding strip loaded waveguide makes possible operation of these
lasers in the fundamental transverse mode for wider stripe widths than can be
achieved for LOC BH lasers. 'Fhe reason for this is that the higher order modes
are less well confined to the active layer and therefore have significantly lower
modal gains than the fundamental mode.
operates is determined primarily by the mode refiectivities and the mode
confinement factors. Fig. 3b shows the mode retlectivities of the 4 lowest order
lateral spatial modes of an ISBH laser. In this case the aluminum contents of the
burying layers were taken to be 0.35. The mode refiectivities were calculated
--
-1.5
0.0
1.5
3.0
4.5
DISTANCE ALONG JUNCTION (µ.m)
mode of an ISBH laser
width
4.0
a)
b)
c)
d)
1-
:r: _J
U)w
3.0
t:
Refl ectiv ity of the mode s of an ISBH lase r
Late ral confi neme nt facto rs of the mode s of an
ISBH lase r
Activ e laye r thres hold gain s of the mode s of an
ISBH lase r
calculated by the effective index method of the same 4 lowest order modes vs.
stripe width.
burying layers (see chapter 3 ), it can be seen that the ISBH laser has a
significantly stronger preference for the lowest order mode due to the difference
in confinement factors. This is because the strip BH laser structure provides
weaker lateral optical guiding than does the LOC BH laser structure. Combining
the results shown in fig. 3b and 3c the active layer gain required for each mode
to reached threshold is plotted in fig. 3d vs stripe width. For the calculation
shown in fig. 3d it is assumed that there are no internal losses, such as from
free carrier absorption or optical scattering. In this case, the threshold gain is
given by:
where
gmth = active layer gain at which mode m reaches threshold
The transverse mode that oscillates is the mode which reaches threshold at the
lowest active layer gain. It is almost always desirable to have the laser oscillate
in the fundamental mode. For many applications, such as those which require
high output powers it is desirable to obtain fundamental mode operation for as
mode operation is generally achievable in buried heterostructure lasers only for
active layer widths less than 1.5-2 microns. The results shown in fig. 3 predict
that the ISBH laser will operate in the fundamental mode for active layer widths
up to 6 microns, which was the maximum stripe width calculated. In practice,
fundamental mode operation was achieved only for active layer widths up to 4
microns. The discrepancy is most likely due to inaccuracies introduced by the
use of the effective index method, which probably resulted in an overstating of
the confinement factor advantage of the fundamental mode.
5.2 Fabrication of ISBH Lasers
The steps involved in the fabrication of ISBH lasers are illustrated in fig. 4.
The fabrication of ISBH lasers requires two LPE growths. The layers grown in the
first growth are an N Al. 5 Ga. 5 As bottom cladding layer ( 1.5 µm thick, Te doped),
an A1. 05 Ga. 95 As active layer (.10 µm thick, undoped), a P A1. 24 Ga. 76As optical guide
layer (0.8 µm thick, Ge doped), and a P A1.:37 Ga. 63As top cladding layer (1.0 µm
thick, Ge doped). After the first growth, mesas are formed by etching down to
the GaAs substrate using H2 S04 :H2 0 2 :H 2 0 1 :8:8. The width of the guide layer after
the mesa etch, W1 , is typically 5-7 microns. Next the bottom cladding layer is
selectively etched in an HF etch until the top of the mesa is undercut by 1-1.5
microns. The portion of the active layer which is exposed after the HF etch is
then selectively removed using H2 0 2 (pH=7.0). The width of the remaining
portion of the active layer, W2 , is typically 2.5-4 microns.
As in the case of the LOC BH lasers, these layers are grown starting from the
GaAs substrate. A meltback step is not required, nor are there any limitations
on the aluminum content of the optical guide layer. Fig. 5 is a scanning electron
microscope photograph of a cleaved cross section of an ISBH laser.
Active Loyer
'-I
Guide
heterostructure lasers are shown in fig. 6. The main disadvantage of the SBH
laser is that it requires an LPE growth initiated on top of the optical guide layer.
Initiation of growth on air exposed AlGaAs layers of aluminum content greater
than 0.1 is very difficult due to the formation of an oxide layer on the exposed
AlGaAs surface.
Normally, LOC BH lasers are grown with N type optical guide layers to
minimize electron leakage. For a double heterostructure with a given difference
in the aluminum contents of the active and cladding layers, the leakage of
electrons into the P cladding layer will be greater than the leakage of holes into
the N cladding layer4 . If a relatively low aluminum content optical guide layer is
incorporated into the laser structure, it is therefore desirable that it be N type.
The use of N type GaAs substrates causes the active layer to be on top of the
guide layer. However, it is difficult to fabricate these lasers with the active layer
on top of the guide layer, since this would require growing over the AlGaAs guide
layer in the second LPE growth. Having the active layer underneath the optical
guide layer, makes it possible to grow the layers in the second growth starting
from the GaAs substrate. ISBH lasers can still have N type optical guide layers if
p type GaAs substrates are used.
5.3 Properties of ISBH Lasers
efficiencies that were nearly identical to those of LOC BH lasers fabricated in our
lab. For 300 µm cavity lengths, pulsed threshold currents were typically 25-30
mA for 2.5 µm wide active layers and 35-40 mA for 4 µm wide active layers. CW
were generally 50-60.%, and exceeded 70.% in a few devices. ISBH lasers operated
burying layers ISBH lasers operated predominantly in the fundamental
transverse mode for active layer widths, W2 , up to 4 microns. By comparison,
LOC BH lasers fabricated in our lab operated predominantly in the fundamental
mode only for active layer widths less than 2 microns. Since fundamental mode
operation of LOC BH lasers for stripe widths greater than 2 microns has been
achieved elsewhere 2 , it is possible that predominantly fundamental mode
operation of ISBH lasers with stripe widths greater than 4 microns can also be
achieved by optimizing this structure. A second important characteristic of
ISBH lasers was that irregularities in the far field patterns of the lasers were
greatly reduced as compared to LOC BH lasers that we have fabricated. Fig. 7
shows lateral far field patterns for ISBH lasers with active widths of 2.5 and 4
microns. Although the LOC waveguide structure can support two modes in the
direction perpendicular to the junction, ISBH lasers typically operated in the
fundamental mode in the direction perpendicular to the junction. The full width
half maximum of the far field in this direction was typically 30-35°.
The effect of varying the aluminum content of the AlyGa 1 _yAs burying layer
has also been investigated. The modal characteristics were found to be much
less sensitive to the aluminum content of the burying layers than were
conventional LOC BH lasers, although increasing the aluminum content, y, of the
burying layer from 0.3 to 0.6 decreased the maximum stripe width for which
fundamental mode operation was obtained from 4 µm to 3 µm. However, the
lasers with y=0.6 had significantly improved thermal characteristics. Similar
results have recently been reported for conventional BH lasers with high
aluminum content burying layers 5 . Lasers with y=0.6 operated CW at output
powers up to the catastrophic optical damage limit even when mounted junction
up.
10°
ISBH la se rs
00
-+- y =0.6 IN BURYING LAYE R
0::
_J
CJ)
_J
<[
Fig. 8
of ISBH laser s
character istics is believed to be due to the reduction in the shunt currents
ft.owing on the sides of the mesa when high aluminum content burying layers are
used. The shunt currents are expected to decrease due to both the higher
bandgap of the burying layer and the larger resistivity that is typically obtained
for layers with high aluminum contents.
In chapter 4, the fabricatio n of large optical cavity buried heterostru cture
window lasers in which the active layers did not extend to the mirrors 6 was
described. These lasers were operated at powers up to 130 mW I µm stripe width
per facet. This was three times the power density at which otherwise identical
lasers in which the active layer extended to the mirrors failed by catastroph ic
mirror damage. Since the output power was limited by heating, even larger
power densities should be possible from this structure by improving the thermal
characteri stics of these lasers. In the LOC BH window lasers, fundamen tal mode
operation was achieved only for stripe widths up to 2 µm and the far fields were
frequently highly irregular due to scattered light. However, the window laser
technolog y is completel y compatibl e with the ISBH laser structure and it is
believed that the applicatio n of the window technolog y to the ISBH structure will
result in significan t improvem ents in output power and beam quality as
compared to the LOC BH window lasers.
In conclusion , inverted strip buried heterostru cture lasers have been
fabricated which have threshold currents and quantum efficiencie s that are
comparab le to those of conventio nal LOC BH lasers. However, predomin antly
fundamen tal transverse mode operation of ISBH lasers was obtained for stripe
widths up to 4 µm which was twice as wide as fundamen tal mode operation was
obtained in LOC BH lasers. The beam quality of ISBH lasers was found to be
significan tly improved as compared to LOC BH lasers which have been fabricated
result in significant improvemen ts in the output power and beam quality over
that obtained with LOC BH window lasers.
1. N. Chinone, K. Saito, R. Ito, K. Aiki, N. Shige, 'Highly Efficient (GaAl)As Buried
513-516 (1979).
2. H. Nakashim a and K. Aiki, 'Transvers e Mode Control and Reduction of
Threshold Current in (GaAl)As Buried Heterostr ucture Lasers with a Buried
Optical Guide Japan J. Appl. Phys. 19, pp L591-L594 (1980).
3. W.T. Tsang and R. Logan, "GaAs-AlxGa 1 -xAs Strip Buried Heterostr ucture
Lasers", IEEE J. Quantum Electron. QE-15 ,pp 451-469, (1979).
4. H.C. Casey ''Room Temperat ure Threshold Current Dependen ce of GaAs
AlxGa 1 _xAs Double Heterostr ucture Laser on X and Active Layer Thickness" , J.
Appl. Phys, 49, pp 3684-3692 (1978).
5. C. Henry, R. Logan, and F. Merritt, ''Single Mode Operation of Buried
Heterostr ucture Lasers by Loss Stabilizati on", IEEE J. Quantum Electron.
QE-17 pp 2196-2204 (1981).
AIGaAs Buried Heterostructure Lasers with Narrow Carrier Injection
6.1 Introduction
Unlike most laser structures, which only require a single liquid phase epitaxial
(LPE) growth, buried heterostructure (BH) lasers require two LPE growths. This is
generally considered to be an undesirable feature of the BH lasers because it makes
the task of achieving a high device yield substantially more difficult. However, the
use of two LPE growths affords a large degree of flexibility, which makes possible the
fabrication of many structures that are impossible with just a single LPE growth.
The window BH and the inverted strip BH lasers described in chapter 3 and 4 are
examples of the types of the devices that can be fabricated by exploiting the
flexibility inherent in two LPE growth devices. In this chapter, another example of
this flexibility will be described. In this case wide stripe (8 µm) BH lasers have been
fabricated with the gain region confined to a narrow region in the center of the
stripe.
stripe widths than in a conventional BH laser.
This laser structure, which will be ref erred to as a narrow injection buried
heterostructure (NIBH) laser is shown in fig 1. The most significant feature of this
laser structure is that the current is forced to ft.ow through a narrow injecting stripe
(N GaAs) beneath the bottom cladding layer. This narrow injection is achieved by
the incorporation of a reversed biased PN junction. In this structure, the active
layer width and the width of the gain region are independent of one another. If the
active layer width is chosen to be much larger than the gain width then NIBH lasers
would be expected to behave like gain guided stripe lasers. If the active layer width is
chosen to be the same width as the gain region, then NIBH lasers would be expected
A~2 G~a As
intermed iate situation where the NIBH laser is behaving like a hybrid of the gain
guided and BH lasers.
Before describin g the properti es of NIBH lasers, some of the more relevant
properti es of the gain guided lasers 1 - 4 and BH lasers 5 - 8 will be first reviewed . These
properti es are listed below:
BH lasers
1. Very low threshol d currents (8 mA/ µm stripe width) and high different ial
2. Single longitud inal mode
3. For narrow stripe widths(l ess than 2 µm), single and fundame ntal spatial mode
4. For wider stripe widths, higher order and multi-sp atial mode operatio n occurs,
spatial hole burning results in changes in the beam pattern as the power is
increase d.
Narrow Stripe Gain Guided lasers
1. Relativel y high threshol d currents and lower quantum efficienc y (30-40%)
2. Many longitud inal modes.
3. For stripe widths less than about 5 µm the beam pattern is stable as the power
is increase d. However , the far field pattern is characte rized by two peaks rather
than a fundame ntal gaussian -like beam profile. For wider stripe widths the far
field pattern is similar to a fundame ntal gaussian beam pattern at low power
levels. At higher output power levels the beam pattern is distorted due to spatial
hole burning.
some of the properties of these two types of lasers. For instance, for multimode
fiber-optic communications it is desirable to have a laser that operates stably in the
fundamental transverse spatial mode (properties of BH lasers), but oscillates
simultaneously in many longitudinal modes (a property of gain guided lasers). The
reason that many longitudinal modes are desirable is the attendant reduction of
modal noise 10 •11 . Another example of a potential advantage of the NIBH laser
structure is for a high power single mode laser structure. As was explained in
chapter 2, for high power operation it is desirable to increase the size of the
emitting area as much as possible while still maintaining fundamental mode
operation. The confinement of the gain to the center of the waveguide provides a
very strong mode selecting mechanism because the fundamental mode is best
confined to the center of the waveguide, thus exercising the maximum gain.
The fabrication of NIBH lasers is similar to that of conventional BH lasers,
requiring two LPE growths. The fabrication steps are illustrated in fig. 2. In the first
growth, the layers grown are an N A1. 2 Ga. 8 As layer (0.3 µm thick, Sn doped), an N
GaAs layer (2 µm thick, Sn doped), an N A1. 40 Ga.60 As bottom cladding layer ( 1 µm
thick, Sn doped), an Al_ 12 Ga. 88 As active layer (0.3 µm thick, undoped), and a P
Al. 40 Ga. 60 As top cladding layer (1 µm thick, Ge doped). Next mesas are etched down
to the N GaAs epitaxial layer. Typically, the mesa etched is 6-B µm wide at the active
layer. Next the GaAs layer is selectively etched in H2 0 2 (pH=7). This etching is
stopped when the width of the GaAs is approximately 3 µm. The H202 etch will etch
AlGaAs layers of aluminum content less than 0.1. We choose aluminum content of
the active layer to be 0.12, which is higher than normal. to minimize the etching of
the active layer during the GaAs selective etching. The purpose of the lower A1. 2 Ga.aAs
layer is to limit the amount of downward etching that occurs during the GaAs
N Al. 20Go.80As
Al. 12Gq88 As
N Al. 37 Ga. 63As
f-1
etches laterally. Next. two layers are grown in the second LPE growth. The first layer
is a P A1. 5 Ga. 5 As. It is a property of LPE growth that the growth proceeds in such a
manner as to flatten out any irregularities in the surf ace that were present prior to
the growth. As a result, the P layer rapidly fills the undercut that resulted from the
selective etching. This P layer provides a reversed biased PN junction, which restricts
the current to ft.ow through the narrow top of the N GaAs epitaxial layer. The second
layer is N type A1. 5 Ga. 5 As and this layer is grown until a fiat top surface is obtained.
6.3 Characteristics of Narrow Injection Buried Heterostructure Lasers
with an active layer width of 6-8 µ,m and a lower injecting stripe width of 2-3 µ,m. A
conventional BH laser with a stripe width this wide would operate in a high order
spatial mode and, because of spatial hole burning, the mode profile would change as
the power output is increased. The intention in the laser structure that was
investigated was to obtain stable fundamental mode operation by having a narrow
gain profile which is best matched to the fundamental spatial mode. Obtaining
fundamental mode operation in wide stripe BH lasers would enable the fabrication
of single mode lasers with significantly greater output powers. The far field pattern
of one of these lasers is shown in fig. 3. As can be seen, for low output powers, the
laser operates in the fundamental spatial mode. Fundamental spatial mode
operation is never observed in BH lasers of 8 µm stripe width when the active layer
is uniformly pumped. Achieving fundamental mode operation in wide stripe lasers is
of great importance for high power laser operation since the maximum power
available from a semiconductor laser is approximately proportional to the stripe
width (see chapter 2). It can also be seen in fig. 3 that at higher output powers
there is some distortion of the beam profile, which indicates the laser is operating in
expected that the laser would operate only in the fundamental mode at all power
levels 11 . If the gain profile is sufficiently narrow, as compared to the optical mode
that the optical intensity is nearly uniform over the gain region, then spatial hole
burning is impossible. The reason for the failure of these lasers to do so is probably
due to current spreading in the lower cladding layer. As a result of this current
spreading the width of the gain profile is sufficiently large so that spatial hole
burning effects become important. By increasing the sheet resistivity of this layer
and narrowing the injecting stripe width, improved mode control shoud be possible
in this structure.
As mentioned previously, NIBH lasers can be thought of as a hybrid between
narrow stripe gain guided lasers and BH lasers. It is therefore of interest to see
whether the characteristics of the laser described in the previous chapter
correspond more to those of narrow stripe lasers or to those of BH lasers. Three
important characteristics of a laser are the threshold currents, the differential
quantum efficiency, and the spectrum. For the laser described above the threshold
currents were typically 80-100 mA and the differential quantum efficiency was
typically 15% per facet. In both cases, these figures are more characteristic of gain
guided stripe lasers than BH lasers. A BH laser with an 8 µm wide active layer would
be expected to have a threshold current of 50-60 mA and a quantum efficiency of 2030% per facet. However, the spectrum of NIBH lasers is more characteristic of BH
lasers, which typically operate in a single longitudinal mode. The spectrum of a NIBH
laser is shown in fig. 4 . Although, the laser is not single longitudinal mode, there are
many fewer modes than the typical 5-10 modes found in narrow stripe gain guided
lasers (see chapter 7).
In conclusion, results for a buried heterostructure laser with narrow current
injection have been presented in this chapter. These lasers are a hybrid of the
P=9mW
P=4mW
P = 14mW
lasers show some of the characteristics of each of these classes of lasers. The NIBH
structure makes possible the fabrication of a continuum of laser structures in
between the two extremes of gain guided lasers and BH lasers. By combining
characteristics from the gain guided and BH lasers, it may be possible to fabricate
lasers which are better suited to certain applications than are either of the extreme
cases.
1. F.R.
2. P.M. Asbeck, D. Cammack, J. Daniele, and V. Klebanoff. "Lateral Mode Behavior in
Narrow Stripe Lasers", IEEE J. Quant. Electron. QE-15 ,pp 727-730, (1979).
3. R. Lang. '1.ateral Transverse Mode Instability and Its Stabilization in Stripe
Geometry Injection Lasers", IEEE J. Quant. Electron. QE-15 ,pp 718-726 (1979).
4. W. Streif er, R. Burnham, and D. Scifres. "An Analytic Study of (GaAl)/ As Gain
Guided Lasers at Threshold", IEEE J. Quant. Electron. QE-18 ,pp 856-864 (i 982).
5. T. Tsukada, "GaAs-Ga 1-xAlxAs Buried Heterostructure Injection Lasers", J. Appl.
Phys. 45 ,pp 4899-4906 ( 1974).
6. K. Saito, R. Ito, "Buried Heterostructure AlGaAs Lasers" IEEE J. Quantum
Electron. JQE-16 ,pp 205-215 (1980).
7. H. Nakashima and K. Aiki, 'Transverse Mode Control and Reduction of Threshold
Current in (GaAl)As Buried Heterostructure Lasers with a Buried Optical Guide"
Japan. J. Appl. Phys. 19 ,pp L591-L594 (1980).
8. N. Chinone, K. Saito, R. Ito, K. Aiki. N. Shige, 'Highly Efficient (GaAl)As Buried
Heterostructure Lasers with Buried Optical Guide", Appl. Phys. Lett. 35 , pp 513516 (1979).
9. R. Epworth, 'The Phenomena of Modal Noise in Analog and Digital Optical Fiber
Systems", Proc. 4th Eur. Conf. Optical Communications (Genoa Italy) , pp 492501 (1978).
10. R. Epworth, 'The Phenomena of Modal Noise in Fiber Systems", Tech. Dig. Topical
Meeting on Optical Fiber Communication (Washington DC) paper ThD1. 1979.
in Buried Wavegui de Lasers with Lateral Diffused Junction s", submitte d for
publicat ion.
stripe geometry lasers. In these laser structures, there is no built in waveguide to
confine the light in the lateral direction. The primary advantage of these laser
structures is the simplicity of their fabrication. However, there are several
disadvantages of gain guided stripe lasers, such as, higher threshold currents than
the more complicated buried heterostructure and transverse junction stripe laser
structures. Due to spatial hole burning, the transverse mode patterns of wide stripe,
gain guided lasers do not remain stable if the current is increased much above
threshold 1·2 . The output beams of gain guided lasers can also have a large amount of
astigmatism. Recently,
properties of gain guided lasers. The problem of spatial hole burning has been
found to be greatly reduced in narrow stripe lasers, such as the V-groove laser 5 .
Narrow stripe lasers having very large astigmatism factors have been reported 6 .
These lasers oscillate simultaneously in many longitudinal modes which has been
explained by the high spontaneous emission factors expected in lasers with large
astigmatism 7•8 . In this chapter, gain guided lasers are described in which the current
is confined to flow between two narrow stripes located above and below the active
layer. A simple theoretical model of gain guided lasers is also presented which
explains many of the characteristics mentioned above The laser structure described
in this chapter is a modification of the double current confinement (DCC)
configuration developed by Tsang and Logan9 and is shown in fig. 1. In the original
struc
-~ .~ . _
. __ --~-- -~~~llization
......
less than 10 microns. The structure which will be described is capable of restricting
carrier injection to an extremely narrow width of the active layer. Injecting stripe
widths of 2 microns were routinely obtained and injecting stripe widths as narrow as
0.5 microns have been obtained. The main emphasis in this chapter will be simply
describing the structure and the properties of these lasers. However, it is believed
that the narrow stripe DCC configuration has potential applications in the
fabrication of low threshold lasers structures and arrays of optically coupled lasers.
7.2 Fabrication of Narrow Stripe Lasers with Double Current Confinement
stripe above the active layer and by N-type blocking layers below the active layer.
The fabrication of the narrow stripe DCC structure requires only one liquid phase
epitaxial (LPE) growth and is based on the fact that LPE crystal growth proceeds in
such a manner as to flatten out any features such as mesas or channels that are
present in the substrate prior to the growth 10- 12 . In the case of GaAs substrates with
etched mesas, in the first stages of the LPE growth, the tops of the mesas can be
being melted back at the same time that AlGaAs layers are being grown on the sides
of the mesas. It has been shown from thermodynamic considerations 13 that for a
given temperature, the surface curvature of a solid effects the chemical potential of
the solid. If a liquid is in contact with a solid with a non-planar surface , to maintain
equilibrium, variations in the chemical potential of a solid have to be matched by
variations in the chemical potential of liquid. The chemical potential of the liquid is
in turn directly related to the solute concentration. The deviation in the equilibrium
relation 14 :
where 7 is the surface tension, Vm is the crystal molar volume, and R is taken to be
positive for a convex solid surface. Thus curved solid surfaces will be in equilibrium
at a temperature Te with a solute concentration Ce± L.\C, depending on whether the
surface is convex or concave. A melt with a solute concentration of Ce at a
temperature of Te is therefore undersaturated if above a convex solid surface and
supersaturated if above a concave solid surface. The undersaturated case results in
a melt back of the solid surf ace and the supersaturated case results in rapid
deposition on the solid surf ace.
Fig. 2a shows a scanning electron microscope (SEM) photograph of the mesas
that were etched in the substrate before the growth. The surface of such a mesa is
convex at the top and concave at the base. Thus the melt above the top surface is
effectively undersaturated while the melt near the base is supersaturated. Fig. 2b
shows the growth that results when the substrates were placed under a solution of
gallium, saturated with GaAs, for 30 seconds. As can be seen, in this 30 seconds, the
top of the mesa was melted back to approximately the position of the waist of the
mesa, while at the same time layers were being grown on the sides of the mesa. In
the narrow stripe DCC laser structure this effect is used to grow a blocking layer,
which restricts the current to !low through a narrow mesa etched in the substrate
prior to the growth. The layers grown for the narrow stripe DCC lasers are shown in
fig 3. If the blocking layer is grown for a sufficiently short time it will not grow at all
above the mesa. In practice, we found that when we used n type substrates that the
I-V characteristics of the devices sometimes indicated the presence of a thin
blocking layer above the mesa. However, if Zn doped substrates are used this
problem is avoided due to the rapid diffusion of Zn through any thin blocking layers
that may form above the mesas.
µ.,
growth
conditi ons the DCC configu ration could be obtaine d for blockin g layer growth
signific ant feature of this structu re is that the lower injectin g stripe can
this
techniq ue.
7.3 Proper ties of Narrow Stripe DCC Lasers
width
of the lower injectin g stripe was 2 µm, the lower and upper claddin g layers
µm thick and doped with Ge and Sn respect ively to approx imately 5x10 16
relative ly high. Thresh old current s varied widely, ranging from 70-150 mA
µm long devices . Differe ntial quantu m efficien cies were typicall y 15-20%
these
lasers was a relative ly high 170°K.
One of the most interes ting propert ies of these lasers are the far field pattern
which are charact eristic of a leaky guide. Fig. 4 shows the far field pattern of
ld
curren ts at the high end of the 70-150 mA range were found to have far
threshold currents is believed to be due variations in the width of the injecting
stripe with the high threshold lasers corresponding to the devices with the most
narrow injection. Both the far field patterns and the relatively high threshold
currents can be understood by the fact that the real index of refraction is
suppressed in the gain region by the high carrier density present there. The
resulting strong real index antiguiding causes both the leaky mode far field patterns
and the high threshold currents.
effect was much more pronounced in these narrow stripe DCC lasers. This is an
indication of the very narrow injection that can be achieved in these lasers.
The dependence of the far field pattern on the stripe width can be qualitatively
understood by examining the following simple model. In this model the actual gain
guided waveguide is represented by a dielectric slab waveguide which has gain in the
center layer and loss on the sides. This waveguide is shown in fig. 5. The actual gain
guided laser waveguide has continuously varying gain and real index of refraction
and can only be solved properly by an elaborate calculation in which Poisson's
equation and the wave equation are solved self consistently for the carrier
distribution and the optical modes 4 . Nonetheless, most of the important features of
the modes of a gain guided laser waveguide can be understood by considering the
gain guided dielectric slab waveguide model, which can be solved exactly.
The modes of the gain guided dielectric slab waveguide are the solutions to the
wave equation:
where n is the complex index of refraction and f3 is the complex propagation
constant. Typically, the differences between the indices of refraction of the core
n=nzr+in2i
1-l
TE and TM modes of this waveguide. The solutions of the wave equation are then
given by:
E(x)=cos(ht/Z)exp(-p Ixi)
E(x)=sin(hx)
where
(7.5)
{3 is determined by the requirement that ~~ be continuous. This is satisfied for
with stripe widths of 2,3,and 5 µm are shown in fig 6-9. In all cases the imaginary
part of the index of refraction in the cladding layers was taken to be -0.0005, which
corresponds to a loss of 38 cm- 1• The imaginary part of the index in the core layer
(j)
zw
..,_
-6.0
DISTANCE ALONG JUNCTION (µ.m)
-4.0
fig. 6 Near field intensity and phase for gain guided laser with 2 µm. wide stripes
1-
fig. 7 Near field intensi ty and phase for gain guided laser With 3 µm Wide stripes
STRIPE WIDTH= 5µm
-7.5
-7.5
-5.0
-2.5
2.5
5.0
7.5
DISTANCE ALONG JUNCTION(µ.m)
fig. 8 Near field intensity and phase for gain guided laser with 5 µm wide stripes
-9.0
tz
-30°
fig. 9 Near and far fields of modes of gain guided dielectric slab waveguides
P= 20mW
at threshold for a typical laser. The real index suppression in the center layer was
related to the gain by the expression
Because the indices of refraction of these waveguides are complex, p and h are also
complex. The phase fronts of the modes are therefore not planar in gain guided
lasers and the output beams are astigmatic. For instance the solutions in the lossy
cladding layers of these waveguides are of the form
where p is complex. The solutions are therefore plane waves, except that the
amplitude decays exponentially away from the core.
The most striking feature of these gain guided modes is the appearance of
pronounced side lobes in the far field patterns of the narrow stripe gain guided
lasers. This can be understood by the fact that for narrow stripe widths, a large
fraction of the power is propagating in the cladding layers. These decaying plane
wave solutions in the cladding layers give rise to the side lobes in the far field, which
are at an angle determined by the phase fronts of these decaying plane waves.
A second interesting property of narrow stripe DCC lasers was that they oscillate
simultaneously in a large number of longitudinal modes. The spectrum of a narrow
stripe DCC laser under pulsed operation is shown in fig. 10. This is consistent with
previous findings that the number of longitudinal modes of narrow stripe lasers
increases as the stripe width is decreased 6 . The reason narrow stripe lasers oscillate
in many longitudinal modes is that as the stripe width is increased the astigmatism
of the modes increases with the astigmatism factor being given by
K= =---- --
shown to be proporti onal to K7. Thus as the stripe width decrease s, the spontan eous
emission into the modes increase s. This in turn makes multilon gitudina l mode
operatio n more probable .
Although the primary purpose of this chapter is to report on the properti es of
DCC lasers with very narrow injection , it is believed that there are several
advantag es of the techniqu e of injecting current though a narrow stripe in the
substrat e that may result in useful applicat ion for this techniqu e. First, this
techniqu e allows the fabricati on of narrow stripe lasers with broad area contacts ,
thus reducing contact resistanc e. In addition to lasers with a narrow top contact
describe d above, lasers have been fabricat ed with top stripe widths of 15 µm. These
lasers do not have as pronoun ced antiguid ing characte ristics as the narrow stripe
DCC lasers. They also have lower threshol d currents than the DCC lasers, with some
devices having threshol d currents as low as 50 mA. By incorpor ating some real index
waveguid ing mechani sm, to provide good optical confinem ent, it may be possible to
fabricate lasers with still lower threshol ds using this techniqu e of restricte d
injection through the substrat e. In addition , we believe, that the leaky guide nature
of the narrow stripe DCC structur e combine d with the simplicit y of its fabricati on
makes the narrow stripe DCC structur e an excellen t choice for the fabricati on of
arrays of optically coupled lasers.
In conclusi on, double current confinem ent lasers which feature very narrow
carrier injection have been fabricate d. These lasers exhibit strong antiguid ing
resulting in far fields characte ristic of leaky waveguid es.
structu res if combin ed with a real index wavegu iding mechan ism.
2. P.M. Asbeck , D. Camma ck, J. Daniele , and V. Klebano ff. "Latera l Mode Behavi
3. R. Lang. "Latera l Transv erse Mode Instabi lity and Its Stabiliz ation in
Stripe
Geome try Injectio n Lasers" , IEEE J. Quant. Electro n. QE-15 ,pp 718-726 (1979).
4. W. Streif er, R. Burnha m, and D. Scifres. "An Analyti c Study of (GaAl)/ As
5. C. Wolk, H. Gottsm ann, P. Marsch all, K. Peterm an, W. Pfister, and H. Vollme
756759 (1981).
6. W. Streiffe r, D. Scifres , and R. Burnha m, "Sponta neous Emissio n Factor of Narrow
Stripe Gain Guided Diode Lasers" Electro nics Letters 17 ,pp 933-934 ( 1981).
7. K.
Electro n. QE-15 ,pp 566-570 (1979).
8. W. Streiffe r, D. Scifres, and R. Burnha m. "Longit udinal Mode Spectra of
9. W.T. Tsang and R Logan, "Latera l Curren t Confine ment in a GaAs Planar StripeGeome try and Channe led Substra te Buried DH Laser Using a Reverse d-Biase d
10. D. Botez, W. Tsang, and S. Wang "Growth Charac teristic s of GaAs-Ga
1 -xAlxAs
Structu res Fabrica ted by Liquid- Phase Epitaxy over Prefere ntially Etched
11. P.A. Kirkby and G.H.B. Thomps on ''Channe led Substra te Buried Heteros tructure
GaAs-(GaAI)As Injection Lasers" J. Appl. Phys. 47 ,pp 4578-458 9 (1976).
12. K. Funakos hi, A. Doi, K. Aiki. and R. Ito "Liquid Phase Epitaxia l Growth of
Ga 1-xAlxAs on Channel ed Substrat es" J. Cryst. Growth, 45 ,pp 252-257 (1978).
13. J.W. Cahn and D.W. Hoffman "A Vector Thermod ynamics for Anisotro pic Surf aces
II. Curved and Faceted Surfaces " Acta Metallur gica, 22 ,pp 1205-121 4 (1974).
14. D. Botez, "Constri cted Double- Heteroju nction AIGaAs Diode Lasers: Structur es
and Electroo ptical Characte ristics", IEEE J. Quantum Electron . QE-17 ,pp 22902309 (1981).
Single Carrier Type Dominated. Impact Ionization in Multilayer Structures
statistics of the carrier multiplication process, since positive feedback effects, which
exist when both electrons and holes produce secondary pairs, can greatly amplify
any current fluctuations. Significantly more noise is generated if the electron and
hole ionization rates (cx,{3) are equal than if only one carrier produces secondary
pairs. 1 It is therefore highly desirable to have a detector in which the multiplication
process is dominated by one carrier type. Unfortunately, most III-V materials have
a:.R:i f3 .Recently, Chin etal. 2 proposed a multilayer GaAs-AlGaAs structure designed to
using molecular beam epitaxy (MBE) by Capasso et.al. 3 . The expected enhancement
of O'./{J is due primarily to the fact that the discontinuity of the conduction band is
larger than the discontinuity of the valence band. Thus, electrons enter the GaAs
multiplication region with more kinetic energy than do holes and are therefore more
likely to produce a secondary pair. This multilayer APD structure is shown in fig. 1.
There is little impact ionization in the AlGaAs layers due to the higher ionization
threshold of AlGaAs. In this chapter a proposed modification of this structure is
described which is expected to significantly further increase O'./{J.
To illustrate the basic principles of this photodetector, first consider the
hypothetical structure shown in fig. 2. A unit cell of this multilayer structure
consists of five layers and two different materials. Material A has a low ionization
threshold energy and is the material in which the multiplication occurs in this
device. Material B has a much larger ionization threshold energy and negligible
multiplication occurs in the layers of material B. The applied voltage is sufficiently
Ga As
Ga As
of the superlattice APD grown by MBE3.
b) Energy band diagram of the superlattice APD. The
band edge discontinuities are AE c =0.48eV, AE v =0.08eV.
(a)
REGION
_J
U- 0:::
..J<{
w-
Fig. 2a) Schematic diagram of the multilayer APD structure
showing the doping of each layero Shaded regions are
material A; white regions are material B.
b) Electric field in each layer of the multilayer structure.
of the multipli cation region are fully depleted even at zero applied voltage. Since the
gradient of the electric field in a depleted layer is proporti onal to the doping, the
electric field changes abruptly in the thin, heavily doped layers. In the lightly doped
layers, the field is nearly constant . By doping the layers as shown in fig. 2 the
electric field on one side of material A (left side in fig. 1) can be made to be larger
than the electric field on the other side of A. If the electron s are injected into A
from the high field side and the holes are injected into A from the low field side, then
the fraction of the electron s that are injected with energies above the ionizatio n
threshol d can be significa ntly larger than the fraction of holes that are injected with
energies above the ionizatio n threshol d.
To obtain the hypothe tical avalanch e photode tector just describe d all that is
needed are two material s with sufficien tly different ionizatio n threshol ds. Since the
ionizatio n threshol ds of semicon ductors are generall y proporti onal to the bandgap ,
any two semicon ductors with sufficien tly different bandgap s could be used for the
material s A and B. For the specific case of the ternary material s AlxGa -xAs, the
bandgap increase s as x increase s.
45
55
suitable choices for material s A and B respectiv ely.
Since the electric fields vary significa ntly over very short distance s, the
common ly used Baraff theory4 cannot be readily applied to calculate the ionizatio n
rates of this device. However , to estimate the enhance ment of al{:l we have applied a
simple model of impact ionizatio n due to Shockley >. In this model impact ionizatio ns
occur only for the case of a carrier starting from zero energy and accelera ting,
without suffering any collision s, to an energy above the ionizatio n threshol d. In the
Shockley model, the electron and hole ionizatio n rates are proporti onal to exp(Dn(Ein)/ Ln) and exp(-Dp(Eip)/Lp). where Ln and Lp are the optical phonon mean free
threshol d energies , and Dn,p(Ei) are the distance s the electron s and holes must
travel without undergo ing a collision to accelera te to the threshol d energy for
ionizatio n. For this calculati on the thicknes ses of the GaAs, high field AlGaAs, and
the low field AlGaAs layers were chosen to be 400 A, 700 A, and 900 A respectiv ely.
For the ionizatio n threshol ds and optical phonon mean free paths we used the
values given in Chin et al. These are ionizatio n threshol ds of 2.0 eV for electron s and
1.5
for which the ionizatio n ratio cx./{3 was calculate d. For the calculati on the simplifyi ng
assumpt ion was made that the electric field changes abruptly . This is equivale nt to
replacin g the heavily doped layers of finite thicknes s by a sheet of charge of
infinites imal thicknes s.
In fig. 3 the position x=-1 1 satisfies:
(B.1)
(B.2)
The position s x=-1 1 .-1..:2 are therefor e the starting points for electron s, which result
in impact ionizatio ns at x=O,D in the Shockley model x= O,D respectiv ely, assumin g
they suffer no collision s. Similarly the position s x=1:3,L are the starting position s of
holes that result in impact ionizatio ns at x=O,D in the Shockley model.
Shockley model ionization rate calcul;1tjon.
ionization threshold energy, are assumed to be the same in Al. 45 Ga. 55As as in GaAs.
Although is certainly not the case, the enhancement of a.1(3 in this structure as
compared to that fabricated by Capasso et al. is significant for any reasonable set of
assumptions about the material parameters.
To calculate the ionization ratio a../(3, the probability for a carrier being
accelerated so that it results in an impact ionization is first determined as a
function of the starting position. Integrating over all starting positions then gives
the total ionization rates. For electrons
electrons contributing to the secondary multiplication that start outside of the high
field region is negligible. A similar assumption for the holes is not always reasonable
since the voltage drop across the low field region may be less than the ionization
energy for some operating conditions of interest (this is actually the most desirable
situation). The ionization probabilities must therefore be separated into three cases
depending on whether the positions x=L3 ,L4 are in the low or high field regions.
P{x) "'exp(
4)
x> 4
P(x) "'exp(
r:;->
P(x) "'exp(
Lp
Ei )
P(x) "'exp ( -{x-D) )
x> 4
P(x) "'exp (
Lp
Ei )
P(x) "'exp( -{x-D))
x> 4
obtained.
and
r;-)
where
1nE2
Lno = E -E2
E1-E2
and
for tiE=O, corresponds to a detector similar to that reported by Capasso etal. 3 . The
main purpose of this calculation is to show that significant enhancement of a/{1 is
expected for .6E7- 0 . The enhancement of a/{1 is most likely somewhat exagerated
for the upper left hand part of fig. 4 due to shortcomings of the Shockley model.
This calculation has been repeated for other commonly quoted values for the
ionization thresholds and optical phonon mean free paths, which are consistent with
experimentally measured ionization rates of GaAs The predicted enhancement of a/{3
was found to be relatively insensitive to which set of parameters we chose. The
enhancement of a/{1 for .6Et' 0 is most pronounced for lower electric field strengths.
However, even with higher fields, which means greater multiplication per unit cell,
the enhancement of a/fl is significant. To reduce the complexity of fabricating such
a detector it is desirable to have no more than 10-15 unit cells ( with an external
Fig. 4
ficlcls in the high and low field layers.
electric field strength s correspo nding to the right hand side of fig. 4 since higher
electric fields would mean that fewer unit cells would be required .
To optimize the design of the propose d detector , it would be necessar y to
calculate the electron and hole energy distribut ion function s fe(G),fh(G) at each
position as the carriers move through the layers of the detector . Such a calculati on
can be done using a Monte Carlo simulati on6 , however this has not been attempte d
for this structur e. Although the Shockley model is useful for estimati ng rx/{3, it can
not be used to accurate ly calculate fe(G) and fh(G).
distribut ion function s. First, the n + AlGaAs layers should be as thin as possible, so
that the hot electron s do not lose much energy in these layers. The high field AlGaAs
layers should have thicknes ses and electric fields such that a significa nt fraction of
the electron s are injected into the GaAs layers with enough energy to produce
seconda ry pairs.
ionizatio n by holes the low field layers have to be sufficien tly thick that holes can
lose ( by phonon collision ) the kinetic energy gained in the precedin g high field
layer. In addition , the differenc e between the electric fields in the high and low field
regions should be as large as is practical . A 50 A thick n+ layer with a doping of
2x10 18 cm-3 will result in a change, ~E. in the electric field of approxim ately
l.6x10 5 V/cm. It is also desirabl e to have the total number of donors in a unit cell
nearly equal to the total number of acceptor s, so that the electric field pattern
repeats itself in each unit cell. One final consider ation is that electron s that ft.ow
across an abrupt interface from GaAs to Al. 45 Ga. 55As have to overcom e an energy
barrier of nearly 0.5 eV. For this reason, it may be desirable to grade the interface s
between the GaAs and the n- Al. 45 Ga. 55As layers.
process is dominated by a single carrier type has been proposed. For a GaAsA1.45Ga.55As detector of this type a/{3 has been estimated based on a simple model.
The results of this calculation suggest that detectors with greatly enhanced ratios of
a/{3 should be practical.
1. R.J. Mcintyre "Multiplication Noise in Uniform Avalanche Diodes". IEEE Trans.
ED-13' 1966, pp164-167
Multilayered Heterojunction Structures", Electron. Lett., 16 , 1980, pp 467-469
3. F. Capasso, W.T. Tsang, A.L. Hutchinson, and G.F. Williams, Appl. Phys. Lett. 40 ,
Jan. 1 1982, pp 38-40
4. G.Baraff, Phys. Rev. 128, 1962 pp 25075. W. Shockley "Problems Reiated to P-N Junctions in Silicon" Solid State Electron.,
2' 1961 pp 35-67
6. P.
Academic Press, New York, 1979.