Optoelectronic structure fabrication by organometallic vapor-phase epitaxy and selective epitaxy - CaltechTHESIS
CaltechTHESIS
A Caltech Library Service
About
Browse
Deposit an Item
Instructions for Students
Optoelectronic structure fabrication by organometallic vapor-phase epitaxy and selective epitaxy
Citation
Tsai, Charles Su-Chang
(1996)
Optoelectronic structure fabrication by organometallic vapor-phase epitaxy and selective epitaxy.
Dissertation (Ph.D.), California Institute of Technology.
doi:10.7907/vrda-c377.
Abstract
NOTE: Text or symbols not renderable in plain ASCII are indicated by [...]. Abstract is included in .pdf document.
The internal configuration and external supports of OMVPE reactors are examined. The quality of epitaxial layers deposited by an OMVPE reactor is strongly influenced by its internal configuration. The quality of the external supports determines the safety, the environmental impact, and the operating efficiency of the OMVPE reactor.
Optoelectronic structures are fabricated by selective epitaxy. The morphology and growth behavior of GaAs, AlGaAs, and InGaAs using selective epitaxy are presented. Highly selective growth can be achieved through the use of organometallic compounds which contain halogens. The selective growth of nanometer-scale GaAs wire and dot structures is demonstrated. Spectrally-resolved cathodoluminescence images as well as pectra from single dots and wires, passivated by an additional AIGaAs layer, are presented. A blue shifting of the GaAs luminescence peak is observed as the size scale of the wires and dots decreases. Formation of highly-uniform and densely-packed arrays of GaAs dots by selective epitaxy is described. The smallest GaAs dots formed are 15-20 nm in base diameter and 8-10 nm in height with slow-growth crystallographic planes limiting growths of individual dots. Completely selective GaAs growth within dielectric-mask openings at these small size-scales is also demonstrated. The technique of facet-modulation selective epitaxy and its application to quantum-well wire doublet fabrication are described. The smallest wire fabricated has a crescent cross-section less than 140 [...] thick and less than 1400 [...] wide.
The development of OMVPE epitaxial layers for a visible-wavelength vertical-cavity surface-emitting laser (VCSEL) is presented. The defect density of the mirror layers was reduced to a negligible level by optimizing gas switching. Electroluminescence spectrum of an InGaP heterostructure p-n diode is presented. The defect density of the active region was also reduced to a negligible level by optimizing the gas-switching sequences.
Item Type:
Thesis (Dissertation (Ph.D.))
Degree Grantor:
California Institute of Technology
Division:
Engineering and Applied Science
Major Option:
Applied Physics
Thesis Availability:
Public (worldwide access)
Research Advisor(s):
Vahala, Kerry J. (advisor)
Scherer, Axel (co-advisor)
Thesis Committee:
Unknown, Unknown
Defense Date:
16 May 1996
Record Number:
CaltechETD:etd-12222007-114128
Persistent URL:
DOI:
10.7907/vrda-c377
Default Usage Policy:
No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:
5125
Collection:
CaltechTHESIS
Deposited By:
Imported from ETD-db
Deposited On:
23 Jan 2008
Last Modified:
16 Apr 2021 22:28
Thesis Files
Preview
PDF (Tsai_csc_1996.pdf)
- Final Version
See Usage Policy.
17MB
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
Optoelectronic Structure Fabrication by
Organometallic Vapor-Phase Epitaxy
and Selective Epitaxy
Thesis by
Charles Su-Chang Tsai
In Partial Fulfillment of the Requirements
for the Degree of
Doctor of Philosophy
California Institute of Technology
Pasadena, California
1996
(Submitted May 16, 1996)
ii
Charles Su-Chang Tsai
Acknowledgements
I would like to express my gratitude to Professor Kerry J. Vahala, my thesis advisor,
for his guidance and support. He provided many ideas of fundamental importance to the
work presented here, including that of using selective epitaxy in fabricating
nanostructures. His advice, both in regard to this thesis and other matters, was
indispensable.
I thank Professor Axel Scherer for his encouragement and technical advice, and his
graduate students, especially Oskar Painter, for processing the samples from the visible
VCSEL project. I also thank Robert Lee for the many many hours he spent at JPL
assisting me during the OMVPE growths.
I thank John Lebens and Professor Thomas Kuech for their guidance and assistance
during the initial work in selective epitaxy. John taught me the critical skills of sample
processing, e-beam lithography, and scanning electron microscopy. Professor Kuech
introduced me to the OMVPE growth technique.
I am grateful to everyone who had helped me in the course of my graduate work. I
thank Mike Hoenk, Mike Newkirk, Channing Ahn, Gang He, Peter Sercel, Paul Maker,
Rich Muller, Akbar Nouhi, James Singletery, and Siamak Forouhar for their assistance
during various stages of this work. And most of all, Iam grateful to my family for their
support.
I am also grateful to the AT&T Foundation and the National Science Foundation for
providing fellowship support for my graduate studies.
iv
Abstract
The internal configuration and external supports of OMVPE reactors are examined.
The quality of epitaxial layers deposited by an OMVPE reactor is strongly influenced by
its internal configuration. The quality of the external supports determines the safety, the
environmental impact, and the operating efficiency of the OMVPE reactor.
Optoelectronic structures are fabricated by selective epitaxy. The morphology and
growth behavior of GaAs, AlGaAs, and InGaAs using selective epitaxy are presented.
Highly selective growth can be achieved through the use of organometallic compounds
which contain halogens. The selective growth of nanometer-scale GaAs wire and dot
structures is demonstrated. Spectrally-resolved cathodoluminescence images as well as
spectra from single dots and wires, passivated by an additional AlGaAs layer, are
presented. A blue shifting of the GaAs luminescence peak is observed as the size scale of
the wires and dots decreases. Formation of highly-uniform and densely-packed arrays of
GaAs dots by selective epitaxy is described. The smallest GaAs dots formed are 15-20
nm in base diameter and 8-10 nm in height with slow-growth crystallographic planes
limiting growths of individual dots. Completely selective GaAs growth within dielectric-
mask openings at these small size-scales is also demonstrated. The technique of facet-
modulation selective epitaxy and its application to quantum-well wire doublet fabrication
are described. The smallest wire fabricated has a crescent cross-section less than 140 A
thick and less than 1400 A wide.
The development of OMVPE epitaxial layers for a visible-wavelength vertical-cavity
surface-emitting laser (VCSEL) is presented. The defect density of the mirror layers was
reduced to a negligible level by optimizing gas switching. Electroluminescence spectrum
of an InGaP heterostructure p-n diode is presented. The defect density of the active
region was also reduced to a negligible level by optimizing the gas-switching sequences.
Table of Contents
1. Introduction
1.1. Introduction to OMVPE .......... 02-0 ee eee
1.2. Outline of the Thesis... 2... 5.2.2... eee ee eee eee
2. Internal Configuration of OMVPE Reactors
2.1. Basic Modules within OMVPE Reactors... 2... eee eee eee eee
2.2. Gas and OM Sources .. 2... 2-2 ee eee ee te eee ee
2.3. Gas Switching Manifold ......... 22.2.6 2c ee eee eee eee
2.4. Deposition Chamber... 2.6... - eee eee ee ees
2.5. Pressure Control System .... 22... ee ee ee eee eee
3. External Supports for OMVPE Reactors
3.1. Basic Supports for OMVPE Reactors ..........-.---25+2 0555s
3.2. Safety Monitoring and Control] Station ...........-.+--+.----5:
3.3. Gas Bunker for Hazardous Gas Cylinders ...........-.2.--20--
3.4. Isolated Air Enclosure... 2... 2... ee eee
3.5. Primary Exhaust Treatment ...........-.-- 20 eee eee eee ee
4. Selective Epitaxy of GaAs, AlGaAs, and InGaAs
4.1. Introduction to Selective Epitaxy............. cee cece eee
4.2. Substrate Preparation and Growth Conditions ................---
4.3. Selective Epitaxy on Patterned Substrates ........-.-.----------
4.3.1. GaAs Selective Epitaxy .. 2.0... 22. e eee ee ee eee eee
4.3.2. Al,Ga,_,As Selective Epitaxy .....-..--.--.-2----200---
4.3.3. In,Ga,_,As Selective Epitaxy ......--.---- 2-2 eee eee eee
4.4. Growth Uniformity during Selective Epitaxy .................---
4.4.1. Global Uniformity ........ 0.0.2... 0 2 eee eee eee eee
I-1
I-]
II-1
4.4.2. Local Uniformity .......... 5.0.0.0... 0. e ee eee eee
4.5. Applications and Selective Heterostrucure Formation .............
4.6. Discussion and Conclusion ...........-..02 202.2. e ee eee
5. Fabrication of Nanometer-Scale Wire and Dot Structures by Selective
Epitaxy
5.1. Introduction to Quantum Wires and Dots ..................0-.
5.2. Sample Preparation and Growth .............0.00000000000.
5.3. Characterization of Wire and Dot Growths .................-.-.
5.4. Conclusion 2.2... ee eee eee
6. Formation of Highly-Uniform and Densely-Packed Arrays of GaAs Dots
by Selective Epitaxy
6.1. Introduction .. 2.2.2... ee ee ee ene
6.2. Dot Formation by Selective Epitaxy...............-.--------
6.3. Characterization and Analysis by Atomic-Force Microscopy ........
6.4. Conclusion .. 2... eee eee
7. Facet-Modulation Selective Epitaxy - a Technique for Quantum-Well
Wire Doublet Fabrication
7.1. Introduction to Facet-Modulation Selective Epitaxy ..............
7.2. Substrate Preparation and Growth Conditions ..................
7.3. Characterization of Growth Samples ................---202--
7.4, Conclusion ............2.2.0202000- oe ceeetceeteesees
8. Visible-Wavelength Vertical-Cavity Surface-Emitting Laser by OMVPE
8.1. Introduction to Visible-Wavelength VCSEL ...................
8.2. Growth and Characterization of A/4-Stack Mirror................
8.3. Growth and Characterization of MQW Active Region.............
8.4. Growth of Complete VCSEL Structure... 2... ..........0..... VUI-14
vil
List of Figures
2.1. Schematic diagram of basic modules in OMVPE reactors ............. I-2
2.2, Schematic illustration of gas and OM sources ...........-2-...---. II-6
2.3. Schematic illustration of two gas switching manifolds ............... I-11
2.4. Schematic illustration of two deposition chambers ...............--. I-14
2.5. Schematic diagram of pressure control system ............-...000- I-17
3.1. Schematic diagram of external supports for OMVPE reactors .......... I-2
3.2. Schematic illustration of gas bunker for hazardous gas cylinders, coaxial tubing,
and isolated air enclosure... 2... 2... 2 eee ee ee ee eee I-7
3.3. Schematic illustration of two primary exhaust treatment systems ........ I-11
4.1. Schematic illustration of OMVPE selective epitaxy ................-. IV-4
4.2. GaAs growth process in conventional OMVPE ..................4. IV-5
4.3. GaAs growth process in selective epitaxy using DEGaCl ............. IV-7
4.4. Basic fabrication procedures for selective epitaxy ..........-......-. IV-9
4.5. Cross-sectional images of [Ol 1] GaAs stripes ..........02.-2....06- IvV-11
4.6. Cross-sectional images of [011] GaAs stripes... 2. .....0.2.-..000-. IV-12
4.7. Cross-sectional images of [010] GaAs stripes ...........2.-...004. IV-13
4.8. Contour plot of growth-rate uniformity of GaAs ..... eve ceeeceeuen IV-17
4.9. Profile of patterned regions grown using different chemistries .......... IV-19
4.10. Schematic illustration of local growth-rate variation for [011] stripes.... IV-21
4.11. Schematic illustration of selective multilayer [010] stripes........... IV-23
4.12. Cross-sectional images of a multilayer [010] stripe................ TV-25
4.13. Schematic diagram of a narrow-stripe laser by selective epitaxy ........ IV-26
5.1. Schematic drawings of quantum structures... ........2........0-. V-3
5.2. Density of states functions for quantum structures .........0.......4. V-4
5.3.
5.4.
5.5.
5.6.
5.7.
6.1.
6.2.
6.3.
6.4.
6.5.
7.1.
7.2.
7.3.
7A.
7.5.
7.6.
8.1.
8.2.
8.3.
8.4.
8.5.
8.6.
8.7.
8.8.
Vill
SEM images of selectively grown GaAs wires and dots .............. V-7
Cathodoluminescence (CL) spectra of single wires and single dots ....... V-9
CL image of an array of wires .. 2... 2 ee eee V-11
CL image of an array of dots... 2... eee eee V-12
Cross-sectional images of thick AlGaAs overgrowths..............-. V-13
Schematic diagram of an atomic-force microscope (AFM) ............ VI-4
AFM images of GaAs dots after growth... 2... 0.022. ee eee ee VI-6
3-D AFM images of GaAs dots after growth. ................00-. VI-7
3-D AFM images of GaAs dots after nitride removal ................ VI-8
Plot of GaAs dot height versus base diameter after nitride removal ....... VI-10
Precursor to facet-modulation selective epitaxy .................20.- VIi-3
Schematic illustration of facet-modulation selective epitaxy ........... VII-5
Cross-sectional images of a wire-doublet structure. ................ VIL-7
CL spectra taken near a wire-doublet structure .................04. VII-8
CL images of a wire-doublet structure at different wavelengths ......... VIU-10
CL images of different wire-doublet structures ..............2..... VIil-11
Schematic illustration of VCSEL structures ..............0....00.0. VItI-3
Schematic illustration of a passive fabry-perot cavity ................ VII-5
Reflectance plot of a passive fabry-perot cavity... .. - Lee ee ee eee VII-6
Optical and TEM micrographs of A/4-stack mirrors .............---- VIll-8
Optical micrographs of InGaP/InAIP p-n diodes . See eee eee VIU-10
Electroluminescence spectrum of a InGaP/InAIP p-n diode ........... VIH-12
Optical and TEM micrographs of MQW active regions............. VI-13
Optical micrographs of full VCSEL structures by OMVPE ........... Vul-15
I-1
Chapter 1
Introduction
1.1 Introduction to OMVPE
Organometallic vapor-phase epitaxy (OMVPE), often referred to as metal-organic
chemical-vapor deposition (MOCVD) or other permutations of these same letters
(MOVPE or OMCVD), began with the work of Manasevit [1-3] in the late 1960s. In
OMVPE, organometallic (OM) compounds such as trimethylgallium (TMGa) or
trimethylindium (TMIn), and hydride gases such as arsine (AsH3) or phosphine (PH3),
are utilized in an OMVPE reactor for the epitaxial growth of semiconductor materials
such as GaAs or InP [4]. In the OMVPE reactor, the OM compounds and the hydride
gases flow over a semiconductor substrate that is heated and supported by a hot graphite
susceptor. The OM compounds and the hydride gases decompose thermally and deposit
epitaxial material on the substrate. Normally an additional large flow of a carrier gas
such as hydrogen conveys the OM compounds and the hydride gases onto the substrate.
Compared to other epitaxial growth techniques such as molecular-beam epitaxy
(MBE) or liquid-phase epitaxy (LPE), the chemistry of the OMVPE process is more
complex. In MBE, simple elemental sources such as gallium and arsenic are evaporated
at a controlled rate onto a heated substrate under ultra-high-vacuum (UHV) conditions for
the epitaxial growth of semiconductor materials such as GaAs. In LPE, epitaxial
materials such as GaAs are grown on a substrate as the substrate contacts molten
solutions of semiconductor compounds dissolved in metal, such as GaAs dissolved in Ga.
Although both MBE and LPE offer simpler chemistry, OMVPE is more versatile. For
{-2
example, the growth of compounds containing both Al and In is nearly impossible in
LPE, and the growth of alloys containing both As and P is particularly difficult in MBE,
but the growth of these materials is routine in OMVPE.
As will be shown in this thesis, the more complex chemistry of the OMVPE process
offers many useful variations in the epitaxial-growth process. Selective epitaxy, or the
laterally controlled growth of epitaxial material within dielectric-mask openings on a
substrate, was achieved here with complete selectivity by substituting the conventional
OM compounds with halogenated OM compounds. In selective epitaxy, the epitaxial
growth of material is controlled in three dimensions instead of in one dimension as in
conventional OMVPE. This three-dimensional control of epitaxy has resulted in the first
demonstration of selectively grown nanometer-scale structures.
1.2 Outline of the Thesis
This thesis can be divided into three parts. Chapters 2 and 3 describe the physical
configuration of the OMVPE reactors utilized for this work. Chapters 4 to 7 describe the
experimental efforts and results in the fabrication of optoelectronic structures using
selective epitaxy. Chapter 8 describes the experimental efforts to fabricate a visible-
wavelength vertical-cavity surface-emitting laser (VCSEL) by OMVPE.
In chapter 2, the internal configuration of OMVPE reactors is examined in detail. The
quality of epitaxial layers deposited by an OMVPE reactor is strongly influenced by its
internal configuration. Design variations are described in relation to their potential
impacts on the quality of epitaxial layers produced. Comparisons are made between a
"Reactor A" and a "Reactor B" to illustrate the similarities and differences between an
earlier-generation reactor and a state-of-the-art reactor.
In chapter 3, the external supports for OMVPE reactors are examined in detail. The
external supports for an OMVPE reactor do not directly influence the quality of epitaxial
layers deposited by the OMVPE reactor. Instead, the quality of the external supports
1-3
determines the safety, the environmental impact, and the operating efficiency of the
OMVPE reactor. Specific examples will be used to illustrate possible variations in the
external supports and their consequences on the safety, the environmental impact, and the
operating efficiency of OMVPE reactors.
In chapter 4, experimental results on the morphology and growth behavior of GaAs,
Al,Ga,.,As and In,Ga,_,As using selective epitaxy are presented. Highly selective
growth can be achieved through the use of organometallic (OM) sources which contain
halogens, such as diethylgallium chloride (C,Hs5),2GaCl (DEGaC]), diethylaluminum
chloride (C)Hs5),AICI (DEAICI), and dimethylindium chloride (CH3)InCl (DMInCl).
The single-layer materials and selectively grown heterostructures produced by this
technique have been characterized. The interface between the selectively grown material
and the underlying substrate was investigated and the conditions for achieving high-
quality defect-free interfaces were determined.
In chapter 5, the selective growth of nanometer-scale GaAs wire and dot structures
using OMVPE is demonstrated. Spectrally-resolved cathodoluminescence images as well
as spectra from single dots and wires, passivated by an additional Al,Ga,_,As layer, are
presented. Growth behavior of GaAs wires with thick Al,Ga,_,As overgrowths is also
presented for potentia] device applications. A blue shifting of the GaAs luminescence
peak is observed as the size scale of the wires and dots decreases.
In chapter 6, formation of highly-uniform and densely-packed arrays of GaAs dots by
selective epitaxy is described. The arrays of GaAs dots are imaged using atomic-force
microscopy (AFM). Accounting for the AFM tip radius of curvature, the smallest GaAs
dots formed are 15-20 nm in base diameter and 8-10 nm in height with slow-growth
crystallographic planes limiting growths of individual dots. Completely selective GaAs
growth within dielectric-mask openings at these small size-scales is also demonstrated.
The uniformity of the dots within each array ranged from 6% for the larger dots to 16%
for the smallest dots.
1-4
In chapter 7, the technique of facet-modulation selective epitaxy, a variation of
selective epitaxy, and its application to quantum-well wire doublet fabrication are
described. Successful fabrication of wire doublets in the Al,Ga,_,As material system is
achieved. The smallest wire fabricated has a crescent cross-section less than 140 A thick
and less than 1400 A wide. Backscattered-electron images, transmission electron
micrographs, cathodoluminescence spectra, and spectrally-resolved cathodoluminescence
images of the wire doublets are presented.
In chapter 8, the development of OMVPE epitaxial layers for a visible-wavelength
vertical-cavity surface-emitting laser (VCSEL) is presented. Passive fabry-perot cavities
with Alp 5Gap 5As/AlAs A/4-stack mirrors were grown as test structures for mirror
calibration. The defect density of the mirror layers was reduced to a negligible level by
optimizing gas-switching sequences during OMVPE growth. Transmission electron
micrographs and reflectance measurements of one calibration cavity are presented.
Ing, 5Gag 5P/Ing 5sAlg 5P heterostructure p-n diodes, processed by using the actual
fabrication procedures for VCSEL, were prepared as an intermediate test structure for the
Ing 5Gag 5P/Ing 5Alg Gap 3P/Ing 5Alo 5P multiple-quantum-well (MQW) active region.
Electroluminescence spectrum of the p-n diode is presented. Test structures containing
only the MQW active region were grown. The defect density of the MQW active region
was also reduced to a negligible level by optimizing the gas-switching sequences. Full
structures containing both the mirror and the MQW active region were grown. Optical
micrographs of these full structures are also presented.
1-5
Bibliography
[1] H. M. Manasevit, Appl. Phys. Lett. 12, 156 (1968).
[2] H. M. Manasevit and W. I. Simpson, J. Electrochem. Soc. 116, 1725 (1969).
(3] H. M. Manasevit, J. Crystal Growth 13/14, 306 (1972).
[4] G. B. Stringfellow, Organometallic Vapor-Phase Epitaxy, Academic Press (1989).
Il-1
Chapter 2
Internal Configuration of OMVPE Reactors
The quality of epitaxial layers deposited by an OMVPE reactor is strongly influenced
by the internal configuration of the OMVPE reactor, where the quality of epitaxial layers
is gauged by the purity of the deposited materials, the uniformity of material composition
and thicknesses, the abruptness of transitions between different epitaxial layers, and the
density of crystalline defects.
In this chapter, the internal configuration of OMVPE reactors is examined in detail.
Design variations are described in relation to their potential impacts on the quality of
epitaxial layers produced. Comparisons are made between a “Reactor A" and a “Reactor
B" to illustrate the similarities and differences between an earlier-generation reactor and a
state-of-the-art reactor. Specifically, "Reactor A" is the General-Air OMVPE reactor, and
"Reactor B" is the AIXTRON 200/4 OMVPE reactor [1]. Both reactors are currently
located at the Microdevices Laboratory (MDL), NASA Jet Propulsion Laboratory (JPL),
Pasadena. Both reactors have been utilized extensively during the course of this thesis
project. "Reactor B" was selected, procured, designed, manufactured, inspected, and
brought on-line as a significant portion of this thesis project.
2.1 Basic Modules within OMVPE Reactors
Most OMVPE reactors, although complex in appearance, can be partitioned into
several basic modules as illustrated in figure 2.1. These basic modules include: gas
sources and carrier gases module, organometallic (OM) sources module, gas switching
Computer
Control
External
Interface
TI-2
Input Gases
Gas Sources &| Carrier | Organometallic
Carrier Gases Gas, (OM) Sources in
(including gas
purification)
temperature-
controlled baths
Gas Switching Manifold
j, v
Reactor Cs
Safety
Logic ad
Manual
Control
Vent v Run
Deposition Chamber
Pressure Control System
Exhaust Gases
Figure 2.1. Schematic diagram of basic modules within an OMVPE reactor.
I-3
manifold, deposition chamber, pressure control system, computer control module, manual
control module, and reactor-safety-logic module.
Input gases, for example hydrogen (H>), nitrogen (N2), arsine (AsH3), phosphine
(PH3), and silane (SiH4) required for the epitaxial growth of doped GaAs or InP
semiconductor materials, are directed into the gas sources and carrier gases module. The
input gases are purified and filtered to remove solid particles from the gas delivery lines
and impurities such as water vapor (H>O) or oxygen (O2), which are impurities
detrimental to the epitaxial growth of any material containing aluminum (AJ) or silicon
(Si).
Typically H, is purified by gas diffusion through a heated palladium (Pd) membrane,
because H, is the only gas that can diffuse through such a membrane. Purification by Pd
membrane results in exceedingly pure H, that is suitable for use as the main carrier gas in
the OMVPE epitaxial growth process. Purification by forcing gases through chemically-
activated-resin matrix, also commonly known as Nanochem®, is another popular
technique for purifying N> and hydrides (AsH3, PH3, SiH) where the diffusion
technique using a Pd membrane cannot be applied.
After purification, the gases required by the OMVPE epitaxial growth process are
precisely measured and controlled by various arrangements of gas valves, mass-flow
controllers, and/or pressure controllers, and subsequently directed into the gas switching
manifold. In addition, carrier gases such as H» and N> are directed into the OM sources
module.
In the OM sources module, carrier gases such as H> or N> are used as vapor carriers
for OM materials. For each OM source, a carrier gas is forced through the OM source
bubbler maintained at a constant temperature by immersion in a temperature-controlled
recirculating-coolant bath. The carrier gas conveys a quantity of the OM material
proportional to the vapor pressure of the OM material. The amount of OM material
required by the OMVPE epitaxial growth process is precisely measured and controlled by
1-4
various arrangements of gas valves, mass-flow controllers, and/or pressure controllers,
and subsequently directed into the gas switching manifold.
The source gases, supplied by the gas sources and carrier gases module, and the OM
materials, conveyed by carrier gases from the OM sources module, are mixed together in
the gas switching manifold. Those source gases and/or OM materials required during a
particular step of the OMVPE epitaxial growth process are directed into one or more run
lines, and those source gases and/or OM materials not required are directed into the vent
line. The run lines are routed to the deposition chamber, and the vent line bypasses the
deposition chamber and is routed directly to the pressure control system.
In the deposition chamber, a substrate is placed on a susceptor. The susceptor and the
substrate are heated to a temperature up to 800°C. For most OMVPE epitaxial growth
processes, the gases flowing in the run lines are mostly Hp gas that is mixed with trace
amounts of OM materials (<< 0.1% of the total gas flow) and small amounts of hydride
gases such as AsH3 and/or PH3 (< 10% of the total gas flow). When heated by the
susceptor and the substrate, the OM materials and hydride gases decompose and deposit
epitaxial layers on the substrate. After flowing over the hot susceptor and substrate, the
depleted gases are directed into the pressure control system.
The pressure control system maintains the deposition chamber at a constant pressure,
usually below atmospheric pressure. The operation of the pressure control system in an
OMVPE reactor is made more complicated by the solid decomposition products and
particles generated in the deposition chamber and by the acutely toxic gases such as AsH3
and PH3 present in the gas stream. Without a properly designed pressure control system,
the solid components in the gas stream would quickly clog and damage the vacuum pump
and gas valves in the system and result in an unusable OMVPE reactor.
After proceeding through the pressure control system, the gaseous waste from the
vent line and the deposition chamber is conducted through stainless-steel tubing to the
exterior of the OMVPE reactor for further treatment.
II-5
An OMVPE reactor is incomplete without the means to control its operations. For
most reactors, computer-automated operation and manual operation are possible.
Computer-automated operation allows an operator to program the sequences of a
complex OMVPE epitaxial growth process before the growth. The actual growth process
would be controlled entirely by the computer without any human intervention. Manual
operation requires an operator to actuate valves, adjust mass-flow controller setpoints,
etc., by physically moving toggle switches and turning knobs. Manual operation is
possible only for the growth of the simplest epitaxial structures.
Reactor-safety-logic modules are included in certain OMVPE reactors, especially in
newer reactors such as “Reactor B." This module oversees all of the operations of the
OMVPE reactor to prevent actions or procedures which might result in damage to the
reactor, injury to the human operator, release of toxic gases, or any other combinations
thereof. In addition, this module verifies the correct operation of key components within
the reactor. In the event of a serious malfunction internal to the reactor, such as a pump
failure, or an emergency situation external to the reactor, such as a toxic gas release, the
reactor would automatically revert to a safe condition, thereby preventing further
problems from occurring.
2.2 Gas and OM Sources
The gas sources module and the OM sources module in an OMVPE reactor
determines the overall performance of the OMVPE reactor. Figure 2.2 illustrates the
different types of gas sources and OM sources, ranging from the simplest sources
containing only one mass-flow controller to more elaborate sources containing multiple
mass-flow controllers and one pressure controller. In general, the performance of the
more elaborate sources is superior to the simplest sources in all respects except for the
higher complexity, lower margin of stability, and higher initial cost.
Gas Source Type A
II-6
OM Source Type C
CG —+o-»o- MFC1 CG +=
SG sero h GSM MGT. |4Vow ‘ i
cr |e
Gas Source Type B T
CG GSM
oeol, MFC2 qT
SG— 0+ GSM OM
Ve MFC1
OM Source Type D
Gas Source Type C CG
CG MFC2
oy MFC4 =
Va | “MEG Ly C NTP
P ent | C
OM Source Type A
ce-{H — GSM r GSM
MFC1 Wow ‘ er
| OM Source Type E
OM —> |-GSM
OM Source Type B Vent
CG
MFC2
— . GSM T
MFC1 OM
x * OM
be
OM
Figure 2.2. Schematic illustration of different types of gas sources and OM sources.
II-7
In gas source type A, the simplest gas source, a gas valve (Vg) selects between the
carrier gas (CG) and the source gas (SG) and directs the gas flow through a mass-flow
controller MFC1 into the gas switching manifold (GSM). The flow through MFC1 is
called the source flow. The actual flow rate F (in sccm, standard cubic centimeter per
minute) of the source gas delivered to the gas switching manifold is regulated by MFC1,
and is equal to:
_ Furci XsG
~ 100 ’
and the partial pressure P of the source gas in the deposition chamber is equal to:
P= Por
where Pp is the absolute pressure measured in the deposition chamber, Fyypcy is the
flow-rate setpoint (in sccm) of MFC1, Xgq is the actual concentration (in percent) of
source gas supplied by the gas cylinder, and Fy is the total flow rate (in sccm) of gases
entering the deposition chamber.
Although simple and low cost in construction, this type of gas source suffers from
significant drawbacks. For flow rates Fyypc, less than about 50 sccm, it takes a
significant amount of time for changes in flow rates Fyypc, to reach the gas switching
manifold and to stabilize, which results in prolonged delays and compositional drifts
during an epitaxial growth process. Excessively long interruptions between epitaxial
layers can significantly increase impurity incorporation and defect density at the
interfaces between epitaxial layers. This type of gas source is used exclusively in
"Reactor A." As a result, it is essentially impossible to grow complex multilayer
structures in this particular reactor.
In gas source type B, a gas valve selects between carrier gas and source gas and
directs the gas flow into MFC1. In addition, a separate flow of carrier gas is directed into
another mass-flow controller MFC2. The flows from MFC1 and MFC2 are combined as
I-8
close to MFC] as possible. The combined flow is directed into the gas switching
manifold. The flow through MFC2 is called the pusher flow. The actual flow rate F of
the source gas delivered to the gas switching manifold is regulated by MFC1, and is
identical to the formulas presented for gas source type A.
The addition of MFC2 significantly improves the operation of this gas source. As the
name suggests, the pusher flow increases the velocity of gas flow between MFC1 and the
gas switching manifold. Changes and stabilization of flow rates that required rainutes
using gas source type A would take seconds to accomplish using gas source type B. This
type of gas source significantly improves the OMVPE growth process, and is used
exclusively in "Reactor B" for all non-dopant gas sources.
In gas source type C, a gas valve selects between carrier gas and source gas and
directs the gas flow into MFC1. In addition, a separate flow of carrier gas is directed into
another mass-flow controller MFC3. The flows from MFC1 and MFC3 are combined as
close to MFC1 as possible. A portion of the combined flow is directed through a mass-
flow controller MFC4 into the gas switching manifold. Remainder of the combined flow
is directed through a pressure controller (PC) into the vent line of the pressure control
system. The flow through MFC3 is called the dilution flow, and the flow through MFC4
the injection flow. The actual flow rate F (in sccm) of the source gas delivered to the gas
switching manifold is equal to:
_ _Fecs Futrci Xsc
~ 100 'vrci + Frcs)’
where Fyyrc3 is the flow-rate setpoint (in sccm) of MFC3 and Fyyrcy is the flow-rate
setpoint (in sccm) of MFC4.
The larger number of interacting active elements in this type of gas source contributes
to the lower margin of stability, where flow-rate oscillations may occur under certain
conditions. The flow rate F is more susceptible to variations and drifts, because F is
dependent on the calibrations of multiple mass-flow controllers where cumulative errors
II-9
may occur. In addition, a portion of the source gas is always discarded as waste gas
during the operation of this gas source. This type of gas source is primarily used for
controlling dopant gases where the required range of flow rates F is large in order to
accommodate the order-of-magnitude changes in doping concentrations during an
OMVPE epitaxial growth process. It is uncommon to use this type of gas source for main
components of the epitaxial layers such as AsH3 for As or PH; for P in GaAs or InP
materials, where the precise control of flow is more important than the available dynamic
range.
The different types of OM sources are also illustrated in figure 2.2. In OM source
type A, the simplest OM source, a flow of carrier gas is directed through the mass-flow
controller MFC1 into a gas valve assembly (Voyy). The gas valve assembly either directs
the flow through the OM bubbler or allows the flow to bypass the OM bubbler. The flow
from the gas valve assembly is directed into the gas switching manifold. The OM
bubbler is maintained at a constant temperature T by immersion in a temperature-
controlled recirculating-coolant bath. The flow through MFC1 is also called the source
flow. The actual flow rate F (in sccm) of the OM material as a vapor delivered to the gas
switching manifold is equal to:
_ F'eci PomiT]
~ Pp-PomiT] ’
where Poy (T] is the vapor pressure of the OM material at temperature T.
This type of OM source suffers from the same drawbacks as the gas source type A. In
addition, F is strongly dependent on the chamber pressure Pp. Any small fluctuations in
Pp is mirrored in F, which causes compositional fluctuations in epitaxial layers. In order
to keep F constant, Fy4pc, adjusted for one chamber pressure has to be changed when
another chamber pressure is used. This type of OM source is also used exclusively in
"Reactor A." As a result, it is essentially impossible to grow complex multilayer
structures in this particular reactor.
II-10
In OM source type B, the addition of MFC2 results in the same improvements over
OM source type A as gas source type B over gas source type A. The formula specified
for OM source type A applies to OM source type B. Unfortunately, this OM source still
suffers from the same sensitivity to chamber pressure as OM source type A.
In OM source types C and D, the addition of pressure controllers (PC) significantly
improves the operation of these OM sources over OM source types A and B respectively.
The actual flow rate F (in sccm) of the OM material as a vapor delivered to the gas
switching manifold is equal to:
_ ¥'ueci Pomi)
~ Ppc- PomIT] ’
where Ppc is the pressure setpoint of the pressure controller.
The additional pressure controller isolates the OM bubbler from the chamber
pressure. The OM bubbler pressure can be set independently from the chamber pressure.
F is independent of Pp. OM source type D significantly improves the OMVPE growth
process, and is used exclusively in "Reactor B" for all non-dopant OM sources.
For OM source type E, the previous discussions on gas source type C also apply. The
actual flow rate F (in sccm) of the OM material as a vapor delivered to the gas switching
manifold is equal to:
Fyrc4 Furci PomiT]
(Furci + Furc3) (Pec - PomiT) °
f=
2.3 Gas Switching Manifold
In an OMVPE reactor, a properly designed gas switching manifold is critical to the
production of defect-free abrupt transitions between epitaxial layers. Two gas switching
manifolds are illustrated in figure 2.3.
The source gases, supplied by the gas sources module, and the OM materials,
conveyed by carrier gases from the OM sources module, are combined in the gas
I-11
Gas Switching Manifold (Single Run Line)
CG Oo O Oo O re rey O O Run
pa ST TN TT TST TTT Vent
J Eve} | Mon! | LMac} | L¥an! | Lan] Eve | Lae | vi
TMIn TMAl TMGa DEZn HeSe SiH4 AsH3 PH3
Gas Switching Manifold (Multiple Run Lines)
CGi o res o o o o o o Run1
PGt [TS PASTS PPS TS im PAS TA
va [oa | [||
TMIn TMIn TMA] TMAI TMGa TMGa DEZn Cp2Mg
CG2 o o o ° o © Run2
PG2 fis, [Ts [Ts fis, fs, i #— Vent
Ef
AsH3 AsH3 PH3 PH3 HoeSe SiH4
CG3 ° ° ° ° Run3
Pas Aa TT TA TNS
Va Vu Vu | Vu
OM1 OM2 HCI DEGaCl
Figure 2.3. Schematic illustration of two different gas switching manifolds.
Y-12
switching manifold. Those source gases and/or OM materials required during a particular
step of the OMVPE epitaxial growth process are directed by the manifold gas valves
(Vy) into the run lines, and those source gases and/or OM materials not required are
directed by the manifold gas valves into the vent line. The run lines are routed to the
deposition chamber. The vent line bypasses the deposition chamber and is routed directly
to the pressure control system.
Separate high flows of carrier gas (CG, CG1, CG2, and CG3) are maintained in the
run lines (Run, Run1l, Run2, and Run3). These flows, also called the main flows, provide
a steady source of carrier gas into the deposition chamber when none of the manifold gas
valves are directed into the run lines. These main flows also serve to quickly propel the
source gases and/or OM materials into the deposition chamber, thereby making the exact
placement of the manifold gas valve within the manifold inconsequential. For some
OMVPE reactors, separate flows of purge gas (PG, PG1, PG2, and PG3) are also
maintained in the vent lines (Vent) to quickly flush the unneeded source gases and/or OM
materials into the pressure control system.
Properly designed manifold gas valves are critical for an OMVPE reactor. Any
unpurged space (dead space) within a manifold gas valve must be minimized or
eliminated. The unpurged spaces within the valve may act as reservoirs for source gases.
Source gases from these reservoirs would leach out and cause compositional shifts or
blurred interfaces between epitaxial layers in the OMVPE epitaxial growth process.
Another critical issue for the gas switching manifold is the equalization of pressure
between the run lines and the vent lines. If large pressure imbalance exists between these
lines, the switching of the manifold gas valves would produce glitches in the gas flow
that disturb the OMVPE epitaxial growth process with effects ranging from minor
compositional shifts to major increases in the density of crystalline defects.
Two different gas switching manifolds are illustrated in figure 2.3, one with single
run line, and the other with multiple run lines. The manifold with single run line is
II-13
implemented in “Reactor A." This manifold is restricted to source gases and OM
materials which are chemically compatible. If there are parasitic reactions between the
source gases and/or the OM materials, the manifold and the run line could be coated with
the by-products from the parasitic reactions and the epitaxial layers could be
contaminated. In addition, the number of sources that can be used at any one time is
limited by the physical size of the single-run-line manifold.
The manifold with multiple run lines is implemented in "Reactor B." Compatible
OM materials and source gases are grouped together for each run line, thereby
eliminating any possible parasitic reactions. For example, conventional OM materials
such as trimethylindium (TMIn), trimethylaluminum (TMA\), trimethylgallium (TMGa),
diethylzinc (DEZn), and cyclopentadienylmagnesium (CpyMg) are grouped together for
Runl. Hydride gases such as AsH3, PH3, hydrogen selenide (H7Se), and SiH, are
grouped together for Run2. Others such as diethylgallium chloride (DEGaCl), hydrogen
chloride (HCI), and experimental OM materials (OM1 and OM2) are grouped together for
Run3. Many more sources can be used with this multiple-run-line manifold, including
duplicate source gases and OM materials. Duplicate sources are necessary for the growth
of complex multilayer structures because very little time, if any, is available between
epitaxial layers to stabilize large changes in flow rates.
2.4 Deposition Chamber
The deposition chamber is where the deposition of epitaxial layers takes place. As
shown in figure 2.4, a substrate, wafer or sample, is placed on a graphite susceptor. A
quartz tube contains the susceptor and the substrate. In addition, the susceptor disk
supporting the substrate in "Reactor B" also rotates during an epitaxial growth in order to
achieve excellent uniformity in the composition and thicknesses of the epitaxial layers.
As shown in figure 2.4, the susceptor disk rotates parallel to the main susceptor body
about the rotation axis [2]. The susceptor disk is held in sustentation and caused to rotate
I-14
Deposition Chamber (Reactor A)
y Cooling A
i Water il
Run ——> —?> Pressure Control System
Graphite URS
Susceptor Lo) * Quartz
IR Lamp & Tube
Reflector
Deposition Chamber (Reactor B)
h Cooling Plate
Gap_vJ = ~<_— Cooling
Runt x Water
Run3 > >
Run2 —— —— Water : <_ quartz
Graphite DQ SS EERO
Susceptor
Le) O oO Oo
Pressure Control System
IR Lamps & Reflectors
Gas-Flow
Susceptor Rotating Susceptor Disk by
Rotation
(Reactor B)
| Rotation Axis
—\———
_-
~ _
~~ —
Main Susceptor Body 4 4 5
Gas Flow Inlets
Figure 2.4. Schematic illustration of two different deposition chambers. Gas-flow
susceptor rotation utilized in the deposition chamber of "Reactor B" is also shown.
If-15
in the indicated direction by the action of gas flow through several inlets into several
helical grooves on the main susceptor body. The rotary movement is obtained by a force
of viscosity of the gas. No particles are generated during susceptor rotation, because no
frictional contact occurs between the susceptor disk and the susceptor body.
The susceptor and the substrate are heated by one or more infrared lamps with built-in
water-cooled parabolic reflectors. The actual temperature of the susceptor is measured by
a thermocouple placed into the susceptor. Feedback corrections are made to the power
supplies for the infrared lamps to maintain the susceptor at a constant temperature.
In "Reactor A," the gases from one run line (Run) flow into the deposition chamber
through a single inlet. In "Reactor B," the gases from three run lines (Run1, Run2, and
Run3) flow into the deposition chamber through two separate inlets. The OM materials
(Run1 and Run3) enter through the top inlet and the source gases (Run2) enter through
the bottom inlet. The flows from these two inlets combine directly above the leading
edge of the susceptor. This flow arrangement prevents the premature decomposition of
the OM materials, and at the same time, enhances the decomposition of the source gases
(hydride gases), because the OM materials decompose much more readily than the source
gases when heated.
When heated by the susceptor and the substrate, the source gases and the OM
materials in the gas flow decompose and deposit epitaxial layers on the substrate. After
flowing over the hot susceptor and substrate, the depleted gas flow is directed into the
pressure control system. |
In "Reactor A,” the ceiling of the quartz tube above the susceptor is cooled directly by
water flowing through a quartz pocket incorporated into the quartz tube. In "Reactor B,”
the ceiling of the quartz tube above the susceptor is cooled indirectly by water flowing
through a metallic cooling plate that is positioned at a small distance h away from the
quartz tube. In addition, the deposition chamber of "Reactor B" actually consists of two
concentric quartz tubes, an inner tube for containing the gas flow as shown in figure 2.4
I-16
and an outer tube to withstand the pressure from outside atmosphere (not shown). The
space between the outer tube and the inner tube is purged by an adjustable mixture of H
and Np.
Cooling the ceiling of the quartz tube prevents excessive decomposition and depletion
of the source gases and OM materials on the quartz tube. However, the cooling method
as implemented in "Reactor A" results in excessive condensation of OM materials on the
quartz tube. The temperature of the ceiling is too low. The indirect cooling method
utilized in "Reactor B" allows adjustments in the temperature of the ceiling by changing h
and by changing the H>/N> mixture in the gap (the thermal conductivity of H2 and N> are
different) [3]. With the appropriate h and H>/N> mixture, the optimum ceiling
temperature can be achieved, where the decomposition, depletion, and condensation of
the source gases and OM materials on the inner quartz tube are minimized and the
deposition of epitaxial materials on the substrate is maximized.
2.5 Pressure Control System
Stable performance of the pressure control system is critical to the reliable operation
of an OMVPE reactor. During an epitaxial growth process, the pressure control system
maintains the deposition chamber at a constant pressure, usually below atmospheric
pressure. The solid decomposition products and particles generated in the deposition
chamber and the acutely toxic gases such as AsH3 and PH;3 present in the gas stream
complicate the operation of the pressure control system. Without a properly designed
pressure control system, the solid components in the gas stream would quickly clog and
damage the throttle valve and the vacuum pump in the system which results in an useless
OMVPE reactor. Figure 2.5 illustrates the major components within the pressure control
system for an OMVPE reactor. This illustration presents a simplified version of a
pressure control system implemented in an actual OMVPE reactor.
I-17
isneuxy
'10}0831 Td AINO Ue JOJ uIa\sAs [oINUOD sinssoid oY} UTYITM s}UsUOdUIOD JofeU OY} Jo WIBIZEIP MRUTOYIS *¢°Z BIN3T{
eBing | sayjouop | Jeonpsuel |
wInjey aN eINSSald BINSSOId
anen | i lo 1 SOAIEA SOAIEA
yooug |} 4010} Uonejos| UONeIOS|
duind Jaquieug
; winnden uoilsodeq
aed ean joyg/dedL ojomed
Ie) oMOuUL plowuey\
<1 B<] $—<-— Buiyoyms
OA[BA YOOUD BAleA YO8UD JUSA SB)
Wa}SAS jO1]U0D a1NSSald
I-18
During normal operation, the depleted gases from the deposition chamber and the
waste gases from the gas switching manifold are combined and directed through a
particle trap/filter to remove solid particles in the gas flow. Deleterious contamination of
the pressure control system is prevented by this placement of the particle trap/filter. A
motorized throttle valve regulates the gas flow into the vacuum pump. The gaseous
waste from the vacuum pump is conducted through stainless-steel tubing to the exterior
of the OMVPE reactor for further treatment.
The absolute pressure in the deposition chamber is measured by a pressure transducer.
Feedback corrections are made to the motorized throttle valve through a pressure
controller module to maintain a constant chamber pressure. The placement of the
pressure transducer is less than ideal as presented in figure 2.5. The solid particles
generated in the deposition chamber can clog the tubing connected to the pressure
transducer. A more favorable arrangement would be to place the pressure transducer
somewhere upstream of the deposition chamber at a position where the flow of gas into
the deposition chamber is not disrupted.
For safety reasons, several check valves are added to the pressure control system to
prevent over-pressure or reverse-flow conditions in the OMVPE reactor. A check valve
is placed in parallel with the particle trap/filter to allow for a path of gas flow when the
particle trap/filter is clogged or is removed for maintenance or cleaning operations. A
second check valve is placed in parallel with the vacuum pump to allow for a path of gas
flow when the vacuum pump is shutdown for any reason. Another check valve is placed
at the output of the vacuum pump to prevent reverse flow of gases into the vacuum pump.
This reverse flow could be caused by the reverse rotation of the pump motor.
Two pairs of isolation valves are added in the path of the particle trap/filter. These
valves facilitate the removal of the particle trap/filter without exposing the inside of the
reactor or the particle trap/filter to outside air. The inner surfaces of the reactor and, to a
much greater extent, the filter element inside the particle trap/filter are coated with
I-19
arsenic and phosphorous residues. If either is exposed to outside air, spontaneous
combustion would instantly occur and produce toxic gases such as AsH3 and PH3 and
various oxides of arsenic and phosphorous.
For pumping toxic hydride gases such as AsH3 and PH3, the vacuum pump is filled
with a special pump fluid, Fomblin® (PFPE: perfluoropolyethers), that is chemically
inert when exposed to these gases and their decomposition products such as phosphoric
acid. In addition, a continuous flow of purge gas N> into the vacuum pump is maintained
to flush out the toxic and corrosive gases entering the vacuum pump. An oil trap with an
oil return line connected to the vacuum pump is also highly recommended, because
Fomblin® is extremely expensive.
Based on the basic design presented here, pressure control systems can be developed
to meet the specific needs of individual OMVPE reactors.
I-20
Bibliography
[1] Product brochures for AIX 200/44, AIX 200, AIXTRON GmbH, Aachen, Germany,
1992.
[2] P. M. Frijlink, Device Comprising A Flat Susceptor Rotating Parallel To A Reference
Surface About A Shaft Perpendicular To This Surface, United States Patent No.
4,860,687, August 29, 1989.
[3] P. M. Frijlink, Epitaxial Reactor Having A Wall Which Is Protected From Deposits,
United States Patent No. 5,027,746, July 2, 1991.
T-1
Chapter 3
External Supports for OMVPE Reactors
The quality of epitaxial layers deposited by an OMVPE reactor is not directly
influenced by the external supports for the reactor. Instead, the quality of the external
supports determines the safety, the environmental impact, and the operating efficiency of
the OMVPE reactor.
In this chapter, the external supports for OMVPE reactors are examined in detail.
Specific examples will be used to illustrate possible variations in the external supports
and their consequences on the safety, the environmental impact, and the operating
efficiency of OMVPE reactors. The specific examples illustrated here are derived from
the external supports implemented for "Reactor A" and "Reactor B" (see chapter 2)
located at the Microdevices Laboratory (MDL), NASA Jet Propulsion Laboratory (JPL),
Pasadena [1]. During the course of this thesis project, extensive interactions with these
external supports for the two OMVPE reactors were mandatory and essential.
Experiences accrued from these interactions are included here.
3.1 Basic Supports for OMVPE Reactors
OMVPE reactors require comprehensive external supports to meet stringent safety
and environmental regulations. One specific arrangement of the external supports for two
OMVPE reactors is illustrated in figure 3.1. These basic supports include: building safety
monitoring and control station, gas bunker for hazardous gas cylinders, isolated air
enclosure, primary exhaust treatment, and secondary exhaust treatment. The dotted lines
T]I-2
Building Safety
Monitoring and [4@---- .
Control Station
Aidt
* .
ee ee. oe
ee ee LL oe
Gas Bunker ]| | Primary
for Hazardous }--> OMVPE |» Exhaust
Gas Cylinders Treatment
ee eee 7
Y (- Isolated Air |
v Enclosure }
secondary :
EXHAUST Mebernecenececececeeeecececececeeeeeeee?
Treatment
Outside Air
Figure 3.1. Schematic diagram of one specific arrangement of the external supports for
two OMVPE reactors.
I-3
with arrows indicate paths for electronic control and status lines and hazardous gas
sensing lines. The solid lines with arrows indicate paths for gas flow, including process
gas flow and exhaust gas flow from the ventilated enclosures. Enclosing the OMVPE
reactors and the primary exhaust treatment systems, the box drawn with dashed lines is
the isolated air enclosure.
Hazardous gas cylinders, containing gases such as hydrogen (H>), arsine (AsH3),
phosphine (PH3), and silane (SiH4) required for the epitaxial growth of doped GzAs or
InP semiconductor materials, are stored in the gas bunker. The gas bunker includes
various sensors to monitor for toxic gas leaks and other dangerous conditions. The gases
supplied by the gas cylinders are delivered to the OMVPE reactors for epitaxial growths.
In addition, the air extracted from the ventilated enclosures containing the hazardous gas
cylinders is directed into the secondary exhaust treatment system as a safety precaution.
The hazardous gases delivered from the gas bunker are utilized in the OMVPE
reactors for epitaxial growths. The depleted waste gases are sent into the primary exhaust
treatment system. The air extracted from the ventilated enclosures containing the
OMVPE reactors is directed into the secondary exhaust treatment system as a safety
precaution.
The depleted waste gases from the OMVPE reactors are treated by the primary
exhaust treatment systems. Residual toxic hydrides and organometallic materials are
removed from the gas stream. Any remaining gases are directed into the secondary
exhaust treatment system for further treatment. |
The OMVPE reactors and the primary exhaust treatment system are located in an
isolated air enclosure (room) to shield the rest of the building from potential hazards of
the toxic gases.
The secondary exhaust treatment system removes harmful fumes that remain in the
gas stream. The gas stream passes through a wet chemical scrubber. The chemical fumes
entering the scrubber react with an intensive spray of chemical solutions (pH 9.0-10.0)
IiI-4
supplied through a system of nozzles and a recirculation pump. The liquid waste
produced is periodically drained as hazardous waste. After passing through the wet
chemical scrubber, the gas stream is directed to the outside of the building through a
rooftop stack.
The correct operational status of the gas bunker, the isolated air enclosure, the
OMVPE reactors, the primary exhaust treatment systems, and the secondary exhaust
treatment system are monitored constantly by the building safety monitoring and control
station (control room). In addition, sensors are placed in each location to detect
dangerous conditions such as toxic gas leaks or fire. The safety equipment in the control
station monitors these sensors constantly. If potentially dangerous conditions are
detected, the gas cylinders in the gas bunker are shutdown automatically by the safety
control equipment, and other automatic responses would also take place to prevent further
problems.
3.2 Safety Monitoring and Control Station
The safety monitoring and control station is an extremely important external support
for OMVPE reactors. Extremely hazardous and toxic gases are utilized within OMVPE
reactors. High temperatures and large electrical currents are also present in OMVPE
reactors. The safety monitoring and control station provides constant monitoring for
toxic/pyrophoric gases, combustible gases, oxygen deficiency, seismic activity, manual
emergency push button activation, fluid leaks, and fire/smoke. Additionally, the control
station also verifies the proper operational status of the exhaust ventilation and treatment
systems, fire-suppression systems, gas cabinets for hazardous gases at the gas bunker,
cooling water system, and door alarms for gas bunker access and emergency-only exits in
the building.
The toxic/pyrophoric gases are detected by using Chemcassettes®, which are
chemically impregnated dry reaction substrates (paper) used to detect and measure
TiI-5
specific toxic gases. The gases are drawn through the dry reaction substrate by suction
from a vacuum pump. If any toxic/pyrophoric gases are present, the dry reaction
substrate changes color from normally white to brown. The color change is detected and
measured by a built-in optical system and displayed in terms of ppb (parts per billion) or
ppm (parts per million) at the readout. For rooms where the operators occupy, the level
for audio alarm is set at 0.5 TLV and the level for building evacuation is set at 1 TLV.
For inside ventilated enclosures of equipment, the level for audio alarm is set at 1 TLV
and the level for building evacuation is set at 10 TLV. TLV is the threshold limit value
determined for each particular substance at a concentration where workers may be
exposed to day after day without any adverse effects. In the event of a building
evacuation, caused either by toxic-gas detection or by manual activation of emergency
push button, the gas bunker is shutdown, the fire department is notified, and the OMVPE
reactor room is switched from positive room pressure to negative room pressure as
compared to the rest of the building.
The presence of combustible gases or oxygen deficiency are detected by using sensors
containing catalytically-aided platinum-wire bridge. Exposing the wire bridge to specific
gases changes the electrical potential across the wire bridge according to gas
concentration. The changes in electrical potential are detected and measured by an
electrical amplifier and displayed in terms of concentration in percent at the readout. For
combustible gases, the level for audio alarm is set at 0.2 LEL and the level for building
evacuation is set at 0.5 LEL. LEL is the lower explosive limit, the concentration of a
substance in air that is capable of causing an explosion or flash. For oxygen deficiency,
the level for audio alarm is set at 19.5% Oy and the level for building evacuation is set at
16.5% Op.
Seismic activity is detected by an accelerometer mounted in direct contact with the
foundation of the building. Ground accelerations that exceed a set limit will cause a
complete shutdown of the gas bunker.
III-6
Fluid leaks are monitored by electrical electrodes placed within various dry recesses
in the floor. Fluid leaks significant enough to flow into the dry recesses will cause a flow
of electrical current between the electrodes and cause an alarm condition.
Fire, or prelude to fire including smoke, is detected by an incipient fire detection
system. This system detects the pre-combustion submicron particles that are generated
before an actual fire. When the concentration of pre-combustion particles reaches a first
level, an audio alarm is initiated. When the concentration of pre-combustion particles
reaches a higher second level, building evacuation commences.
This particular arrangement of safety monitoring allows centralized control and
maintenance of the building safety equipments. There are drawbacks, however. The
centralized location of the toxic-gas monitoring equipment results in long lag times
before an actual toxic gas leak is detected, because there are long tubings between the
monitored locations and the monitoring equipment. In addition, the toxic gas monitoring
equipment is time-shared between different locations which causes further delays.
An improved arrangement would be to locate additional toxic gas monitoring
equipment at the individual locations. The local monitoring equipment would be much
smaller and less expensive. They would be electrically connected to the central control
station, resulting in much faster responses while providing local audio alarm capabilities
for immediate warning. The early warning may allow operators to avert complete
building evacuation by reacting promptly to the initial local alarm condition.
3.3 Gas Bunker for Hazardous Gas Cylinders
Hazardous gas cylinders, containing gases such as hydrogen (Hj), arsine (AsH3),
phosphine (PH3), and silane (SiH4), are stored in a gas bunker away from the main
building. A schematic diagram of the gas bunker is shown in figure 3.2. The entire
structure of the gas bunker is constructed to be earthquake-resistant and is constructed
from steel-reenforced concrete. The two entrances into the gas bunker are protected
Ill-7
Gas Bunker Coaxial Tubing for
ASSSUIIEIT®IY Hazardous Gases
Exhaust Output N
H2 Gas in
Sensor _| A —_ Coaxial
<= Gas Tubing
Toxic Control >
Gas
Sensor Panel_I | Panel
Cylinder
Status
=| Double
> Support
a Straps
Le
Hazardous
Gas Flow
Noe Purge Gas
Monitored for
Hazardous Gas
Oa
Gas Cylinder Cabinet
KEEN
Earthquake-Resistant Construction
LLLLLELLLLLLLLLLLLLLLL ELLE ELLE ELLE
LLLLLLLELLLLELLLLELLELLE ELLE ELLE
Isolated Air Enclosure (Floor Plan)
Temperature- Direct Exhaust
Controlled Air to ae Air
— (~
Reactor A Reactor B
Exhaust Exhaust Air
Treatment Treatment | {Interlock
Figure 3.2. Top left: schematic diagram of the gas bunker for hazardous gas cylinders.
Top right: schematic illustration of the coaxial tubing for delivery of hazardous gases.
Bottom: schematic illustration of the isolated air enclosure for two OMVPE reactors.
IiI-8
from vehicle collisions by thick barriers constructed from steel-reenforced concrete. The
two entrances are secured by two security doors. The doors are constructed of closely-
spaced steel bars to allow air ventilation in the gas bunker to prevent entrapment of toxic
gases. The two doors at the two entrances are locked at all times except during service
operations. The two doors are monitored at the control station for any unauthorized
accesses.
Inside the gas bunker, there are many gas cylinder cabinets, one of which is shown in
figure 3.2. The actual gas cylinder is fastened to the gas cylinder cabinet using two
support straps. The two straps are necessary to prevent the toppling of the gas cylinder
during an earthquake. If the gas cylinder is secured using only one strap, it is possible for
the gas cylinder to slide on the ground and slip out of the single strap during an
earthquake. The gas cylinder cabinets are anchored to the floor and to the walls of the
gas bunker. The gas cylinder cabinets are ventilated through the building exhaust system.
The air extracted from the ventilated gas cylinder cabinets is directed into the secondary
exhaust treatment system as a safety precaution.
The gas cylinder is connected to a gas cylinder control panel. This computer-
controlled panel allows the automated purging, testing, and evacuating of the tubing
connections when gas cylinders are replaced or service operations are performed. The
panel is also controlled from the control station, allowing for the remote shutdown of the
gas cylinders. Anomalies in the gas delivery system such as excess flow or excess
delivery pressure are detected by the panel and indicated in the control station. The
specific gas cylinder is also shutdown in such events.
The hazardous gas from the gas cylinder is delivered to the OMVPE reactors through
coaxial tubings, as shown in figure 3.2. The inner tubing carries the hazardous gas flow.
A flow of N» purge gas is maintained between the inner and the outer tubing. At the gas
bunker, the Nz purge gas flow is sampled for the presence of toxic gases. If toxic gases
are detected, the gas bunker is shutdown.
IL-9
Various sensors are placed inside and outside the gas cylinder cabinets to monitor for
toxic gas leaks and other dangerous conditions. Toxic gas monitoring points are placed
inside the gas cabinets to detect any toxic gas leaks. Hy» sensors are additionally placed
inside the gas cylinder cabinets for gas cylinders containing hydrogen.
Building a separate gas bunker away from the main building isolates the large depot
of hazardous gases, thereby reducing the risk of exposing the occupants of the main
building to any major toxic gas releases. However, the separate gas bunker also makes it
more difficult to adjust or check parameters for each gas cylinder, such as adjusting the
delivery pressure of the source gas or checking the amount of source gas available from
the gas cylinder. Currently, these adjustments or checks can only be made in the gas
bunker. In addition, the users of these source gases do not know if the gas cylinder is
shutdown or not, since there are no local readouts indicating gas bunker status.
For a gas bunker to function more efficiently and effectively, it is necessary to
electronically monitor all of these additional parameters and distribute this electronic
signal to local readouts located at the equipments utilizing these source gases. It would
be even better if these additional parameters can be adjusted remotely without having to
enter the gas bunker.
3.4 Isolated Air Enclosure
As shown in figure 3.2, the OMVPE reactors and the primary exhaust treatment
systems are completely enclosed in an isolated air enclosure. Two sets of double doors
that form an air interlock are the only entrance to the isolated air enclosure. The doors
open outward for ease of escape during emergencies.
As shown in the diagram, temperature-controlled air is pumped into the isolated air
enclosure through one path. Some of the air exits through the ventilated enclosures of the
OMVPE reactors and the primary exhaust treatment systems into the secondary exhaust
treatment system. The remaining air exits directly to the outside of the building through
TiI-10
another path. In order to prevent potentially contaminated air from escaping through the
doors and other openings, the isolated air enclosure is maintained at a negative pressure
as compared to the rest of the building when a toxic gas leak is detected by the toxic-gas
monitoring equipment. Normally, the isolated air enclosure is maintained at a slightly
positive pressure as compared to the rest of the building to allow for efficient operation of
the exhaust ventilation system.
The incoming air must be maintained at a constant temperature for the proper
operation of the OMVPE reactors. Since air is not recycled in the isolated air enclosure,
the ambient temperature measured at the OMVPE reactors is substantially the same as the
temperature of the incoming air. Fluctuations in the temperature of the incoming air
would disrupt the temperature-controlled baths for the OM sources and shift the operating
points of the mass-flow and pressure controllers. Unwanted shifts in the material
composition of epitaxial layers would be the result. For the proper operation of the
OMVPE reactors, it is therefore extremely important to maintain a constant temperature
in the incoming air.
3.5 Primary Exhaust Treatment
The depleted waste gases from the OMVPE reactors are treated by the primary
exhaust treatment systems. Residual hydride gases and organometallic materials are
removed from the gas stream. Any remaining gases are directed into the secondary
exhaust treatment system for further treatment. |
There are two primary exhaust treatment systems illustrated in figure 3.3. One
system is the CDO unit, an acronym for the combustion-decomposition-oxidation unit.
In this system, the depleted waste gases from the OMVPE reactors are mixed with a gas
containing oxygen and then passed through a chamber with heating coils heated to
800°C. The waste gases from the OMVPE reactors decompose and oxidize in the
presence of high temperature and oxygen. The treated exhaust gas is directed into the
I-11
Primary Exhaust Treatments:
A) C.D.O.: Combustion-Decomposition-Oxidation Unit
Flame
Arrestor
exhaust_, {/1— p> Treated
Input Ur —— Exhaust
| | Heating Coils |
O2 Source A at 800°C
B) AIXTOX Scrubbing System
Exhaust Treated
Input Exhaust
Stage 1 Stage 2
Solution Solution
Figure 3.3. Schematic illustration of two different types of primary exhaust treatment
systems.
WH-12
secondary exhaust treatment system. A flame arrestor is installed on the inlet side of the
CDO unit to prevent combustion from traveling backwards into the exhaust tubing from
the OMVPE reactors.
The CDO unit suffers from extremely serious drawbacks. The unit is not enclosed in
a self-contained ventilated enclosure. If any gas leaks occur in the unit, the gas would
permeate throughout the whole room. Since the CDO chamber is at a high temperature,
there is significant risk for explosions if flow surges occurred in the oxygen source or in
the waste gases which may contain large amounts of hydrogen. The combustion by-
products, composed of oxides of arsenic and phosphorous, are still highly toxic and
corrosive. There is no effective way to dispose of these combustion by-products except
to leave them where they remain in the exhaust ducts. When the exhaust ducts are
clogged, then the ducts have to be replaced.
There were so many problems and drawbacks to the CDO units that the CDO units
installed initially for the OMVPE reactors were replaced by the AIXTOX® scrubbing
system [2,3], also illustrated in figure 3.3.
In an AIXTOX® scrubbing system, there are typically two stages, each containing
different liquid chemical solutions. Within each stage, a pump for corrosive liquids pulls
the chemical solution from a large reservoir (90 liter capacity) and force the solution
through nozzles to form a fine spray of droplets that trickle down through a column
packed with plastic o-rings. The plastic o-rings enhance the total available surface for
chemical reactions. |
In the AIXTOX® scrubbing system, the depleted waste gases from the OMVPE
reactors flow through stage 1 and react with an acidic solution of bromate or iodate salts.
Hydride gases such as AsH3 or PH3 are oxidized and converted into water-soluble acids.
These water-soluble acids remain dissolved in the chemical solution of stage 1. In this
chemical reaction, free molecules of bromine or iodine are also formed. The bromine or
iodine molecules are removed in stage 2 by a basic solution that reacts with the free
I-13
bromine or iodine and forms water-soluble bromate or iodate salts. The treated exhaust
gas, containing only hydrogen, is diluted to concentration levels much below the lower
explosive limit by the air flowing through the ventilated enclosure containing the
scrubbing system. The treated exhaust is directed into the secondary exhaust treatment
system as a safety precaution.
When the chemical solution in stage 1 is nearly depleted, trace amounts of hydride
gases appear in the treated exhaust and is detected by toxic-gas monitoring equipment to
indicate the depletion of the chemical solutions. The chemical solutions would be
drained and replaced with new solutions. The depleted chemical solutions, containing
water-soluble forms of arsenic and phosphorous, are easily drained and disposed of.
The AIXTOX® scrubbing systems perform extremely well as the primary exhaust
treatment systems for the two OMVPE reactors. The only maintenance required, besides
the routine drain and refill of the chemical solutions, is the periodic checks for leaks and
the replacements of rubber seals to ensure liquid-tight and air-tight systems.
I-14
Bibliography
[1] C. Matus, Microdevices Laboratory Safety Manual: Operations, Policies, and
Procedures Plan, NASA Jet Propulsion Laboratory, September 1992.
[2] W. Fabian, H. Roehle, and P. Wolfram, Method Of Removing Noxious Substances
From Gases, United States Patent No. 4,612,174, September 16, 1986.
[3] Product brochures for AIXTOX Scrubbing System, ATXTRON GmbH, Aachen,
Germany, 1990.
IV-1
Chapter 4
Selective Epitaxy of GaAs, AlGaAs, and InGaAs
Many device structures benefit from the ability to selectively deposit epitaxial
materials. Through the use of a masking material, such as silicon nitride (Si3N4) or
silicon dioxide (SiO), on the substrate surface, patterns generated through standard
lithographic procedures can be used to define regions for selective deposition.
Highly selective growth can be achieved through the use of organometallic (OM)
sources which contain halogens, such as diethylgallium chloride (C)Hs)yGaCl (DEGaC]),
diethylaluminum chloride (CzHs5)2AIC] (DEAICI), and dimethylindium chloride
(CH3)2InCl (DMInCl). These compounds decompose, most probably, to the volatile
mono-halogenated species, for example gallium chloride GaCl, and also generate
hydrogen chloride HC] in the gas phase as a reaction by-product.
Experimental results on the morphology and growth behavior of GaAs, Al,Ga,_,As
and In,Ga,_,As using this technique are presented. The single-layer materials and
selectively grown heterostructures produced by this technique have been characterized.
The interface between the selectively grown material and the underlying substrate was
investigated and the conditions for achieving high-quality defect-free interfaces were
determined.
4.1 Introduction to Selective Epitaxy
The drive to smaller device dimensions has led to the development of epitaxial
growth techniques which controllably deposit material on a nanometer scale. Thin layer
IV-2
structures can be produced in which the local band structure is modified by changes in the
material composition along the growth direction. As device dimensions become smaller,
lateral control over the composition and band structure is also needed. Such laterally
defined and controlled growth of semiconductor materials require new growth techniques
and growth chemistries. Selective epitaxy has been developed, in many forms, to address
some of these new technological challenges. Laterally controlled growth has been
investigated in all the major methods of epitaxial growth [1-18]. The term selective
epitaxy has been used to refer to several forms of laterally controlled growth. There are
three current research areas which use the term selective epitaxy:
(1) The first area is conformal deposition on a partially masked substrate [19]. In this
case, the deposited material grows as an epitaxial single-crystal only within the mask
openings to the substrate. Polycrystalline deposition occurs on the masking material and
the edge morphology of the epitaxial regions typically exhibit enhanced growth rates.
(2) Growth on non-masked, patterned substrates, also referred to as selective epitaxy, is a
second area of active research. The substrate patterning is accomplished prior to growth
through chemical or dry etching and standard lithographic procedures. Growth on such
patterned substrates can exhibit lateral variations in the composition and growth rate [20].
These lateral variations result from selective adsorption and surface diffusion of growth
precursors. Certain crystallographic surfaces will preferentially incorporate a particular
growth precursor, locally altering the material characteristics. This technique has been
widely investigated for the formation of advanced optoelectronic structures within a
single growth procedure.
(3) The last form of selective epitaxy refers to the local preferential growth of epitaxial
materials without extraneous or polycrystalline growth on a masking material. External
energy sources, such as a focussed laser, have been used to locally heat the substrate
surface to promote the thermally-driven decomposition reactions [21]. This form of
selective epitaxy allows the ‘writing’ of epitaxial structures on the substrate surface. The
IV-3
lateral definition of the selective epitaxy regions is limited by the wavelength of the laser
and the thermal diffusion characteristics of the substrate material. This last definition of
selective epitaxy can be extended to truly selective growth with masked and patterned
substrates. Selective epitaxy with no deposition on the masked areas is particularly
applicable to device applications where the polycrystalline deposits and variable edge
morphologies are potentially destructive to device performance.
This chapter will present results on this third form of selective epitaxy. As illustrated
in figure 4.1, epitaxial materials such as GaAs or AlGaAs are deposited within openings
of a masking material, such as silicon nitride, on a substrate. For device fabrication, one
distinct advantage of selective epitaxy is the efficient injection of current carriers into the
active region of the device. In addition to defining the active region of the device, the
dielectric layer also performs well as a current focusing element that funnels the injected
carriers through the active region of the device. When selective epitaxy is applied to
semiconductor lasers, the threshold current for lasing could be reduced significantly.
Selective epitaxy of GaAs, Al,Ga,_,As, and In,Ga,_,As were demonstrated using
alternative halogen-based organometallic precursors such as DEGaC] and DEAIC] [22].
The use of halogen-based organometallic compounds can result in the epitaxial growth of
compound semiconductors within openings of a dielectric-coated substrate [13,22]. No
deposition will occur on the masking material under appropriate conditions unlike the use
of conventional OMVPE precursors, such as trimethylgallium (T'MGa), trimethyl-
aluminum (TMA\), and trimethylindium (TMIn). |
GaAs growth process using conventional OMVPE precursors is illustrated in figure
4.2 [23]. Selective epitaxy using this process results in polycrystalline deposits over the
masked area in addition to deposition occurring within the openings of the masking
material. The poor selectivity is the result of the surface reaction, where no mechanism
exists for the Ga atom to return to the gas phase once nucleation has taken place over the
masked area.
IvV-4
OMVPE Selective Epitaxy
/ Ke AlGaAs SN
Mask
Nitride Nitride 2 7 Gans Ninide
Nitride Opening
i GaAs Substrate
LSA LP SP EPP PP PES SE
—s
v7, GaAs Substrate
Efficient Carrier Injection
Insulator
MIE
= t\* —
GaAs >
Insulator
Figure 4.1. Top: schematic illustration of epitaxial deposition within mask openings by
OMVPE selective epitaxy. Bottom: efficient electrical-carrier injection is a distinct
advantage of device fabrication by selective epitaxy.
IV-5
Gallium Arsenide Growth Process
in Conventional OMVPE
(CH,),Ga = trimethylgallium
ASH, = arsine
Complete Reaction
(CH,),Ga + ASH, => GAAS + 3 CH,
Gas-Phase Reaction
(CH.),Ga => CH,Ga + 2 CH,
ASH, => ASH + 2H
Surface Reaction
CH,Ga + AsH => GaAs + CH,
Surface reaction is trreversible
Ga is not volatile
Figure 4.2. Proposed chemical-reaction steps in GaAs growth process using conventional
OMVPE growth precursors trimethylgallium and arsine.
IV-6
GaAs growth process using alternative halogen-based precursors is illustrated in
figure 4.3 [13]. In contrast with the conventional process, this process results in complete
selectivity under appropriate conditions. The surface reaction is reversible. If GaAs
nucleated over the masked area, the surface reaction would reverse under the appropriate
growth condition and convert the nucleated GaAs, in the presence of HCl, back into gas
phase as As and GaCl.
The growth process using alternative halogen-based precursors has the advantage of
being compatible with conventional photolithographic and electron-beam lithographic
practices allowing for the selective epitaxy of sub-micron structures. The properties of
these materials grown on patterned masked substrates will be presented. The results
indicate that these precursors provide a chemical route to achieve selective epitaxy within
conventional OMVPE reactors under growth conditions compatible with the deposition
of high-purity and high-quality epitaxial structures.
4.2 Substrate Preparation and Growth Conditions
Selective epitaxy was studied using the chlorine-based organometallic compounds
DEGaCl and in situ formed DMInCl. DEGaCl was supplied in a standard organo-
metallic-source-bubbler arrangement. The low vapor pressure of this source (0.13 Torr at
40°C for DEGaC]l) may require the use of elevated source temperatures and heated lines
in certain OMVPE reactors. For the initial study, the DEGaCl source was held at ~40°C
and used heated source lines. For subsequent studies, the DEGaCl source was held at
~17°C, slightly below room temperature. DMInCl has a prohibitively low vapor pressure
(~0.03 Torr at 40°C), preventing its direct use in a conventional bubbler. In,Ga,_,As-
based growths are carried out using DEGaCl and TMIn. In this case, DMInCl is probably
generated in situ in the OMVPE reactor through the reaction of DEGaCl and TMIn.
IV-7
Gallium Arsenide Growth Process
in Selective Epitaxy
(C,H;)>GaCl = diethylgallium chloride
ASH, = arsine
Complete Reaction
(C,H;)2>GaCl + AsH, => GaAs + HCl + 2 CoH,
Gas-Phase Reaction
ASH, => ASH + 2 H
Surface Reaction
GaCl + ASH <=> GaAs + HCI
Surface reaction is reversible
As, GaCl are volatile
Figure 4.3. Proposed chemical-reaction steps in selective GaAs growth process using
halogenated-organometallic precursor diethylgallium chloride and arsine.
IV-8
The OMVPE reactors utilized for these experiments are conventional horizontal
reactors as described in chapter 2. Deposition chamber pressure was maintained between
100 and 200 mBar. Growth temperatures within the range 600-800°C were investigated
with most of the growths undertaken at 650 or 700°C. Substrates of semi-insulating and
n-type GaAs were used in the studies. The studies of the selectively grown GaAs used
substrates consisting of 2 im layer of Alp 3Gag 7As capped by 10-20 nm layer of GaAs to
prevent oxidation of the Alp 3Gag 7As layer. The luminescence from the GaAs substrate
is prevented from contributing to the luminescence signal by the Alp 3Gag 7As buffer
layer. The patterned substrates were masked with SizN4 or SiO», deposited by plasma-
enhanced chemical-vapor deposition (PECVD). Most studies used Siz3N4 masked
substrates. The openings in the mask were formed by conventional photolithographic
techniques. Micron-sized openings were defined through reactive-ion etching (RIE)
using CF, plasma. In larger structures, the masking material was etched using buffered
HF solutions. The patterned substrates were cleaned in organic solvents to remove the
resist materials. An intermediate cleaning step of RIE using O> plasma was also applied
to ensure the complete removal of resist materials. Prior to loading into the OMVPE
reactor, the samples were dipped in concentrated HCI for 20-60 seconds in order to
remove the surface oxides, and then rinsed in deionized water. This step was found to be
crucial in achieving high-quality interfaces between the substrate and the selectively
grown material. The samples were annealed at the growth temperature under a H>/AsH3
ambient for 5-10 minutes prior to the beginning of the regrowth step. The basic steps in
the fabrication procedure are illustrated in figure 4.4.
The selectively grown materials were characterized by scanning electron microscopy
(SEM) and optical microscopy in order to evaluate surface and edge morphology. The
luminescence properties of some selectively grown heterostructures were obtained in a
modified SEM with room and low temperature (77°K) cathodoluminescence (CL)
IV-9
‘Axeyida satogyas Joj saInposoid uoreorqey oIseq sy) Jo UOTeNSN]I oNeUIOYIS “pp aIN3Iy
Axeyida aaljoa]9s FdAINO (4 BHuluesjo pue Hulyoye Jy (3 1siso1 uO AydesBou? (gq
KA sienisang see NS SUNIL A Roeasang S20
SSsrensans syey eyensqns syey \Siaisang syey
sVEDIV
SSS Ss
SPUUN | ‘OPUUN :
Cae ee
SEE OPIN
: SPLAIN B
yMmoib o19a/9IP CADAd (9 ~—s (IeUONdo) yyMo1H AqAWO (GQ ~—s BJBASQns Syed Bulyiels (Vv
KA aensang see NS SORE AS
Wa syey Ssreusens sve Ao SEAN
WAR EEREEET
IV-10
collection system [24]. Cathodoluminescence was used to determine the alloy
composition of the ternary growth of AlGaAs.
4.3 Selective Epitaxy on Patterned Substrates
4.3.1 GaAs Selective Epitaxy
The selective epitaxy of GaAs using DEGaCl has been previously reported [13]. The
DEGaCl-based GaAs growth is extremely selective with respect to all of the
conventionally-deposited dielectric materials SizN4, SiO2, and SiON,. The range of
growth temperatures over which complete selectivity is obtained extends over the entire
investigated range of 600-800°C.
The morphology of the GaAs selective epitaxy is dominated by the appearance of the
slow-growth crystallographic planes noted in the inorganic-based growth of GaAs [25].
These crystallographic planes are predominantly the (110) and mixed (111) types of
planes. The appearance of a particular plane is dependent on the growth temperature, the
growth pressure, the growth rate, and the orientation of the mask opening with respect to
the substrate crystallographic orientation.
Cross-sectional views of selectively grown GaAs stripes of various orientations are
shown in figures 4.5, 4.6, and 4.7. The stripes shown in figures 4.5, 4.6, and 4.7 are
oriented along the [01 1], [011], and [010] directions respectively, where the vertical
direction is [100]. Each figure contains backscattered-electron images of various stripes
from the same sample. Stripes of different orientations show different facetting
characteristics. The thicknesses of the GaAs stripes are reduced as the widths of the
initial mask openings are increased. No epitaxial material was deposited over the
dielectric masking material outside of the openings.
The selectively grown regions are free of extended defects as characterized by
transmission electron microscopy (TEM) [22]. The initial growth interface cannot be
resolved using TEM, unlike TMGa-based GaAs layers where the initial growth interface
IV-11
[0 11] cross-sections
Figure 4.5. Cross-sectional backscattered-electron images of selectively grown GaAs
stripes oriented along the [Ol 1] crystallographic direction.
IV-12
[O11] cross-sections
Figure 4.6. Cross-sectional backscattered-electron images of selectively grown GaAs
stripes oriented along the [011] crystallographic direction.
IV-13
-sections
oa)
o2)
al
1)
a)
me
ee
ly grown GaAs
ive.
f select
images o
-sectional backscattered-electron i
Figure 4.7. Cross
irection.
stripes oriented along the [010] crystallographic d
IV-14
can be delineated clearly. This high-quality interface can be produced without
measurable chemical or structural defects.
The complete selectivity seen in the DEGaCl-based growth of GaAs has been
attributed to the in situ formation of GaCl as a result of the decomposition of DEGaC1.
The growth mechanism near the substrate surface should parallel that of the inorganic-
based GaAs growth [25]. In that case, the selectivity arises from the low local
supersaturation and the lack of adsorption of GaCl on the dielectric masking material.
4.3.2 Al,Ga,_,As Selective Epitaxy
The selective epitaxy of Al,Ga,_,As was also undertaken using growth conditions
similar to those employed in the studies of GaAs selective epitaxy, except for the use of
TMAI. The addition of DEGaCI during the growth of Al,Gay_,As results in the selective
deposition of Al,Ga,_,As over a range of alloy compositions and growth temperatures.
The selectivity, however, also depends on the density and pattern of openings in the
masking material. Al,Ga,_,As selectivity was not extensively studied below 700°C
growth temperature.
The data indicate that layers with low Al content grown at high temperatures exhibit
good selectivity. The overall chlorine to metal (Al and Ga) ratio in the gas phase is
increased under these growth conditions. Selective GaAs growth will always occur under
these growth conditions and will generate HCl. This HCl, in turn, prevents the formation
of AlAs deposits on the masked regions, thereby improving selectivity. Higher growth
temperatures decrease the overall supersaturation of the growth ambient and reduce the
overall driving force for deposition and thus improving selectivity.
The morphology of the selectively grown Al,Ga,_,As regions is similar to that of the
selectively grown GaAs. The features are once again dominated by slow-growth
crystallographic planes. There is a tendency to have several planes appear along the
IV-15
growth edge with a shorter spatial extent to each facet face. This tendency of AlGaAs
growth is exploited effectively in the growth technique to be described in chapter 7.
4.3.3 In,Ga,_,As Selective Epitaxy
The selective epitaxy of In,Ga,_,As is similar to that of Al,Ga,_,As, except for the
substitution of TMA] with TMIn. There is a limited range of growth temperature and In
composition over which complete selectivity is achieved. Low In concentrations and
higher growth temperatures result in complete selectivity [22]. The higher In content
layers grown at 600 and 700°C do not show complete selectivity. The morphology of the
selectively grown In,Ga,_,As regions is also similar to that of the selectively grown
GaAs. The features are once again dominated by slow-growth crystallographic planes.
The reduced selectivity seen in the In,Ga,_,As materials is attributed to the low
chlorine to metal (In and Ga) ratio employed here. The lower chlorine to metal ratio in
the gas phase during In,Ga,_,As growth, compared to the binary growth of GaAs, leads
to higher supersaturation. Since TMIn and DEGaC] are used in this study, the chlorine to
metal ratio is always less than 1 whereas DEGaC]I growth of GaAs has a ratio of 1. The
addition of HCI to the gas phase should reduce the overall growth supersaturation and
lead to greater selectivity for In,Ga,_,As selective epitaxy [18].
4.4 Growth Uniformity during Selective Epitaxy
4.4.1 Global Uniformity |
The uniformity of material deposition is an important parameter for any epitaxial-
growth technique. Uniformity during selective epitaxy can be viewed from both a global
(i.e. whole wafer) and local perspective. The utility of the deposition process for device
applications is affected by uniformity on both length scales. The global uniformity
during both selective epitaxy and conventional growth is affected by the therma]-fluid
environment within the deposition chamber of the OMVPE reactor. As such, the
IV-16
uniformity of deposition will be influenced by the total flow rate, the thermal boundary
conditions at the deposition-chamber walls, the chamber pressure, and the chamber
geometry. An additional complicating factor associated with the use of DEGaCl is the
introduction of chlorine which may act as an etchant under certain conditions.
The global uniformity was investigated for the growth of GaAs under two different
conditions, as shown in figure 4.8. The growth temperature was 700°C in both cases.
TMGa and DEGaC] were used in separate growth runs on non-patterned substrates. The
uniformity on masked and patterned substrates using DEGaCl was also determined. The
main variable changed between these runs was the total flow rate into the deposition
chamber. The uniformity of the TMGa-based growth at 7.5 slm (standard liters per
minute) total flow is dominated by a monotonic change in the growth rate across the
wafer. The growth rate uniformity down the wafer in the flow direction is quite good.
This uniformity pattern is attributed to a single convective roll in the gas flow, in the
deposition chamber, induced by the thermal expansion of the gas and the slight
asymmetry in the position of the susceptor within the quartz tube. Since the susceptor is
slightly closer to one of the chamber walls, the gas expansion will be asymmetrical with
respect to the tube axis and a roll develops in the gas flow.
The global uniformity of the DEGaCl-based growth on a non-patterned substrate
indicates a monotonic decrease in the growth rate down the flow axis of the quartz tube.
The lower growth rate on the downstream side of the wafer is probably due to the
decrease in the GaAs growth resulting from the generation of HCI from the GaAs growth
upstream. Growth on the patterned substrate does not seem to be as strongly influenced
by the growth-generated HC] as on the non-patterned substrate. The overall amount of
GaAs deposited on the patterned wafer is less than on the non-patterned wafer due to the
selective nature of the growth and the small amount of exposed substrate (< 5%). The
global uniformity is otherwise very similar to that of the TMGa-based growths.
IV-17
DOWNSTREAM
7.5 splm
(8 Torr
+12% +20% +9%
L 13.5 splm
78 Torr
+9% t7% +2%
UPSTREAM
Figure 4.8. Contour plot of growth-rate uniformity of GaAs for: (left) DEGaCl-based
growth on patterned substrates, (middle) DEGaCl-based growth on non-patterned
substrates, and (right) TMGa-based growth on non-patterned substrates. The arrows
indicate the direction of increasing layer thickness. The average thickness deviations
from the mean thickness are also shown.
IV-18
Increasing the total flow rate into the deposition chamber improved the global
uniformity in all cases as shown in figure 4.8. The convective roll present in the growth
with 7.5 slm total flow rate has been eliminated due to the higher gas velocity in the
deposition chamber. The uniformity in the TMGa-based growth is greatly improved.
The uniformity is mainly affected by the GaAs deposition on the warm chamber walls
which causes some gas-phase depletion. The DEGaCl-based growths on both patterned
and non-patterned substrates are similar. The increase in the growtl. rate across the wafer
is quite similar to the roll-generated non-uniformity seen in the TMGa-based growth.
The main effect of the higher gas velocity is to greatly improve the uniformity of the
growth rate across the wafer.
4.4.2 Local Uniformity
The local growth rate of the DEGaCl-based GaAs growth is affected by several
primary factors. The local ratio of the mask opening to masked regions appears to be the
primary determinant of the local growth rate. The specific growth chemistry also
influences the local growth rate and growth profile, as shown in figure 4.9. These growth
profiles are obtained by a mechanical profilometer.
The InAs was grown using TMIn at a growth temperature of 700°C. Under these
conditions, the InAs growth was partially selective. The pattern used in this study was an
array of lines surrounded by wide borders. The TMIn-based growth exhibits an overall
"U shape” to the growth profile. Selectivity in the InAs case can be attributed to the
surface diffusion of the deposited In across the mask surface to a mask opening. The "U"
profile results from the accretion of material diffusing from the surrounding masking
material onto the exposed GaAs with limited diffusion over the exposed growing surface.
The surface diffusion length inferred from this profile is about 100 um at 700°C growth
temperature. The surface diffusion length on the masking material may be even higher.
"SJUSMIINS BIW Jo}OWOTYOId oY} JO sJovstE ore soptyod [Qe AC oY) Jo o8po oy} uo syead dreys oy ‘suot8e1 posodxa
SYED oY) UO JosINooId YIMOIS oY) JO UONdIOSpe DANDIIAS DY) WOY SINS YMOIZ paseq-[OeH Aq 9y) uo uses sopyosd yey
YL ‘Sdutuedo yseu oy} 0} UOIseI poysew Surpunouns sy} Woy satoeds yIMoIT uy Jo UOIsNFIp oy) Woy sy{nsor op_yo3d
SVU] OY], ‘SOLNSTWAYO WIMOIZ JOJOFJIP OM] BuISN UMOIS SUOIZaAI pouJoed Om) Jo sjold o[yoId-ssouyomI oy], “6p OINLY
(w7/) JONV.LSIC
O0v
008 009 00d O
yl
p09 %| SHe9)
SYDO =
—_ -
ut ®(£H9) : a
SyUT
ot NUN O +N O
(YW) LHOISH
IV-20
The lower profile in figure 4.9 is from a GaAs growth using DEGaCl. The profile is
distinctly different from the InAs case. The wide borders possess sharp peaks at the
edges of the growth. This "peak" results from the facet formation at the edges of the
selectively grown regions. The sharpness of this peak is exaggerated by the difference on
the vertical and horizontal distance scales used in this figure and the finite size of the
profilometer tip. Closer examination of the peak profile indicates that the actual slope is
about 1°-3°. The growth rate within the stripe openings does not exhibit the "U-shaped"
profile as in the InAs case. Growth in these small stripe openings exhibit similar edge
peaks to the wider border areas which yield an apparent, not actual, change in growth rate
between the border area and the stripes. The DEGaCl selective epitaxy is quite uniform
on a local level. The lack of a "U-shaped" profile indicates that the growth mechanism
responsible for the selectivity is not the same as in the InAs or conventional OMVPE
growth. The uniform growth rate indicates that the mechanism of the selectivity is the
preferential adsorption of the growth precursor within the mask openings.
Another important factor affecting the local growth rate is the orientation of the mask
opening in relation to the crystallographic directions of the substrate. This factor
becomes more important as the growth thickness approaches the width of the initial mask
opening. The selective growth of stripes oriented along the [011] direction will be used
here as an example. As the selective growth progresses, the trapezoidal cross-sections of
the [011] stripes evolve into triangles as shown earlier in figure 4.6. The two non-parallel
sides of the trapezoidal cross-section are the (11 1)-type crystallographic planes shown
schematically in figure 4.10. Since these (111)-type planes are slow-growth planes, most
of the growth precursor, in this case GaCl, adsorbed on the (111)-type planes will either
diffuse onto the masking material and desorb or diffuse onto the (100) plane and
contribute to the growth on the (100) plane. The growth-rate equation in figure 4.10
describes this process mathematically, where dH is the enhanced local growth
IV-21
Local Growth Rate of [011] Stripes
—~\
NC
Vv H wae
~< al
Ww
dH = dh ( L(1o0) + Laity )
L (100)
dH = Local Growth Rate
dh = Nominal Growth Rate
Thickness of Growth (um)
1.5 2. 2.5 3. 3.5 4.
W = Width of Mask Opening (um)
Figure 4.10. Top: schematic illustration of a growing [011] stripe with widening (111)
facets and a narrowing (100) facet. The widths of the individual facets affect the local
growth rate dH. Bottom: a plot of growth thickness versus mask opening width for
stripes oriented along the [011] crystallographic direction.
IV-22
rate on the (100) plane, and dh is the nominal growth rate without considering this effect.
In figure 4.10, a graph is also plotted relating the thickness of the growth H to the width
of the initial mask opening W for a given growth time. The straight portion of the graph
is the geometric limit reached by the triangular cross-section. The curved portion of the
graph shows the growth-rate enhancement as the trapezoidal cross-section approaches the
triangular limit. Also plotted are data points taken from actual growths for comparison.
This enhancement of local growth rate becomes more important as the growth thickness
approaches the width of the initial mask opening.
4.5 Applications and Selective Heterostructure Formation
The selective epitaxy of GaAs, Al,Ga,_,As, and In,Ga,_,As makes possible the
selective formation of heterostructures. This application of selective epitaxy was
explored in the growth of multilayer heterostructures in stripe mask openings. The
complete structures of the selectively grown stripes, oriented along the [010] direction,
are shown schematically in figure 4.11. The heterostructures are each composed of a
base GaAs layer and two thinner GaAs layers placed in between three AlGaAs layers.
The entire heterostructure is grown in a single run on a GaAs substrate precoated with a 2
tim layer of Alp 3Gag 7As, a 20 nm layer of GaAs to prevent the oxidation of the
Alo 3Gag.7As layer, and a 20 nm layer of Si3zN, as the dielectric mask. The additional
layers in the substrate were necessary in order to suppress the luminescence from the
underlying GaAs substrate. Stripe openings were made in the dielectric mask layer using
conventional photolithographic procedures. Two different stripes are shown
schematically in figure 4.11. The narrow stripe has a triangular cross-section. The wide
stripe has a trapezoidal cross-section with the narrow flat side on the top. The wide stripe
would eventually develop a triangular cross-section also, if additional material is
deposited.
IV-23
Schematic Cross-Sections
of [010] Stripes
AlGaAs
GaAs
ou
AlGaAs
GaAs ) AlGaAs
Narrow Stripe
Figure 4.11. Schematic illustration of selectively grown multilayer stripes containing
GaAs/AlGaAs heterostructure. These stripes are oriented along the [010]
crystallographic direction.
IV-24
The back-scattered electron (BSE) and cathodoluminescence (CL) images of the
actual wide stripe are shown in figure 4.12. In the BSE images, the brighter portion of
the image corresponds to GaAs and the darker portion corresponds to AlGaAs. Except
for the slight bowing on the top surface, the heterostructure shown in the BSE images
corresponds to the wide stripe shown in the schematic diagram.
One distinct characteristic of selective epitaxy emerges from the CL images. For the
selective epitaxy of binary compounds such as GaAs, compositional shift is not possible.
For the selective epitaxy of ternary or quaternary compounds such as AlGaAs on
patterned masked substrates, compositional shifts occur as the selective growth
progresses. For example, the peak luminescence wavelength of the bottom AlGaAs layer
(not the one in the substrate) is at ~745 nm, as indicated by the strong luminescence
shown in the CL image at 745 nm. The luminescence peak of the middle AlGaAs layer is
at ~708 nm, indicating that a shift in the material composition has occurred even though
all of the AlGaAs layers were grown using identical OMVPE parameters. The magnitude
of material-composition shifts for different layers within a heterostructure is determined
by a complex set of parameters including the crystallographic orientation and the area
filling fraction of the mask opening, the growth rate, the chamber pressure, and other
variables in the OMVPE growth process. Careful calibration of individual] layer
thicknesses and compositions is necessary for the successful growth of heterostructures
by selective epitaxy.
One potential device application of the selectively grown heterostructures is a narrow
stripe laser as illustrated in figure 4.13. The width of the active region is reduced to ~200
nm. The p-doped top region and the n-doped bottom region are entirely isolated from
each other by a layer of silicon nitride. This laser would have low threshold current, low
electrical circuit parasitics, and high electrical modulation speed, because there are no
parasitic p-n junctions and no leakage current. Additional advantages of this
IV-25
Figure 4.12. Backscattered-electron (top 2) and spectrally-resolved cathodoluminescence
(bottom 3) cross-sectional images of a wide stripe oriented along the [010]
crystallographic direction.
IV-26
‘Axeqide aanoapas ACAWO Aq pareouqey Jose] oporp JoJONpuooruies odiNs-MoLEU B JO WeIZEIP NVUIBYIS “Ep WNT
uoibey sAljoy MOEN
wu 00g —» | «—
jn 4
Bulppejo_,
NIH ~
LOOT gee Zo
joe]UOD
SIWUO
Axeyidg ent}o9}9S SdAINO Aq sase adisjs moweN a[buis
IV-27
laser include the high optical-mode confinement possible, the flexible control of
waveguide modes, and the relatively simple and straightforward fabrication procedures.
4.6 Discussion and Conclusion
The OMVPE growth chemistry has been known to determine many of the important
parameters which impact the utility of the grown materials. Most studies of the growth
chemistry have centered on the purity of the growth product and the uniformity of the
deposition. This present study utilized halogen-based organometallic compounds such as
DEGaCl to achieve extremely selective growth. These organometallic compounds are
believed to decompose into the volatile mono-chloride species which are used in the
inorganic-based growth of compound semiconductors. The in situ generation of these
growth reactants mimics the inorganic-based growth chemistry near the growth surface,
allowing for complete selectivity on patterned and masked substrates.
The fabrication of selectively grown heterostructures has been demonstrated using
these halogenated precursors in combination with the conventional precursors under
standard OMVPE growth conditions. The processing required for the selective regrowth
step does not seem to adversely impact the pre-processed structure, at least in the
Al,Ga,_,As/GaAs material systems. The use of lithographically-defined masked regions
to achieve selective epitaxy offers many advantages over the alternative approaches
mentioned in the introduction. Control of the growth morphology and facet formation
can be used to advantage in many applications. Since the growth is truly selective using
DEGaCl, the absence of polycrystalline material eliminates the additional processing
steps for the subsequent removal of this extraneous material. Lastly, selective epitaxy
can be achieved in a variety of material systems which allow the growth of many
combinations of heterostructures. The extension to the phosphorus-based material
systems should be straightforward.
IV-28
This novel growth technique requires further development in the understanding of the
factors which influence the local and global uniformity of the deposition. The range of
physical dimensions, microns to centimeters, and the many crystallographic directions,
each with different growth properties, encountered in the determination of the local and
global growth rates make this problem numerically difficult. In addition, shifts in the
material composition occur for layers within a heterostructure containing ternary or
quaternary materials. For the successful application of selective epitaxy, careful
calibration of the layer composition is mandatory for each specific heterostructure and
mask pattern. A better understanding of the growth chemistry should lead to the design
of more suitable precursors for the growth of alternative materials and expand the
application for selective epitaxy.
IV-29
Bibliography
[1] C. P. Lee, I. Samid, A. Gover, and A. Yariv, Appl. Phys. Lett. 29, 365 (1976).
[2] S. B. Kim and Y. S. Kwon, J. Appl. Phys. 61, 5478 (1987).
[3] U. Konig, U. Langmann, K. Heime, L. J. Balk, and E. Kubalek, J. Crystal Growth 36,
165 (1976).
[4] H. Heinecke, A. Brauers, F. Grafahrend, C. Plass, N. Piitz, K. Werner, M. Weyers, H.
Liith, and P. Balk, J. Crystal Growth 77, 303 (1986).
[5] K. Kamon, M. Shimazu, K. Kimura, M. Mihara, and M. Ishii, J. Crystal Growth 77,
297 (1986).
[6] R. Azoulay, N. Bouadma, J. C. Bouley, and L. Dugrand, J. Crystal Growth 55, 229
(1981).
[7] C. Ghosh and R. L. Layman, Appl. Phys. Lett. 45, 1229 (1984).
[8] K. Nakai and M. Ozeki, J. Crystal Growth 68, 200 (1984).
(9] Y. Takahashi, S. Sakai, and M. Umeno, J. Crystal Growth 68, 206 (1984).
[10] J. P. Duchemin, M. Bonnet, F. Koelsch, and D. Huyghe, J. Crystal Growth 45, 181
(1978).
[11] K. Kamon, S. Takagishi, and H. Mori, J. Crystal Growth 73, 73 (1985).
[12] Y. Nakayama, S. Ohkawa, and H. Ishikawa, Fujitsu Sci. and Tech. J. 13, 53 (1977).
[13] T. F. Kuech, M. A. Tischler, and R. Potemski, Appl. Phys. Lett. 54, 910 (1989).
[14] F. W. Tausch Jr. and A. G. Lapierre, J. Electrochem. Soc. 112, 706 (1965).
[15] D. W. Shaw, J. Electrochem. Soc. 113, 904 (1966).
[16] G. H. Olsen and V. S. Ban, Appl. Phys. Lett. 28, 734 (1976).
[17] A. Okamoto and K. Ohata, Appl. Phys. Lett. 51, 1512 (1987).
IV-30
[18] E. Colas, C. Caneau, M. Frei, E. M. Clausen Jr., W. E. Quinn, and M. S. Kim, Appl.
Phys. Lett. 59, 2019 (1991).
[19] S. H. Jones and K. M. Lau, J. Electrochem. Soc. 134, 3149 (1987).
[20] E. Kapon, S. Simhony, R. Bhat, and D. M. Hwang, Appl. Phys. Lett. 55, 2715
(1989).
[21] N. H. Karam, H. Liu, 1 Yoshida, B.-L. Jiang, and S. M. Bedair, J. Crystal Growth
93, 254 (1988).
[22] T. F. Kuech, M. S. Goorsky, M. A. Tischler, A. Palevski, P. Solomon, R. Potemski,
C.S. Tsai, J. A. Lebens, and K. J. Vahala, J. Crystal Growth 107, 116 (1991).
[23] G. B. Stringfellow, Organometallic Vapor-Phase Epitaxy, Academic Press (1989).
[24] M. E. Hoenk and K. J. Vahala, Rev. Sci. Instrum. 60, 226 (1989).
[25] D. W. Shaw, J. Crystal Growth 31, 130 (1975).
V-1
Chapter 5
Fabrication of Nanometer-Scale Wire and Dot
Structures by Selective Epitaxy
The selective growth of nanometer-scale GaAs wire and dot structures using
organometallic vapor-phase epitaxy is demonstrated. Spectrally-resolved cathodo-
luminescence images as well as spectra from single dots and wires, passivated by an
additional Al,Ga,_,As layer, are presented. Growth behavior of GaAs wires with thick
Al,Ga,.,As overgrowths is also presented for potential device applications. A blue
shifting of the GaAs luminescence peak is observed as the size scale of the wires and dots
decreases.
5.1 Introduction to Quantum Wires and Dots
Conventional crystal-growth techniques, such as molecular-beam epitaxy (MBE) and
organometallic vapor-phase epitaxy (OMVPE), have achieved atomic-scale
compositional control in one dimension along the growth direction. This capability has
resulted in a wealth of new physical phenomena and device applications, producing
structures exhibiting quantum confinement in one dimension (quantum wells). There has
been significant interest in developing techniques enabling the equivalent control in the
lateral directions perpendicular to the growth direction. Such a capability would greatly
enhance the variety of possible structures as well as provide advancements in applications
such as monolithic integration of photonic devices. A major goal of this effort is to
V-2
fabricate structures exhibiting quantum confinement in two or three dimensions, namely
quantum wires or quantum dots.
Schematic drawings of a quantum well, a quantum wire, and a quantum dot are
presented in figure 5.1 (top). Bulk material is also illustrated for comparison. A quantum
well has quantum confinement in one dimension (L,). A quantum wire has quantum
confinement in two dimensions (Ly and L;). A quantum dot has quantum confinement in
three dimensions (L,, Ly, and L,). For quantum structures using the GaAs/AlGaAs
material systems, L,, Ly and L, should be on the order of 10 nm for observable optical
effects at room temperature. Also shown in figure 5.1 (bottom) are three-dimensional
arrays of quantum wells, quantum wires, and quantum dots. Potential barriers containing
material of a higher band-gap confine and isolate the individual quantum structures
within each array.
An important application of quantum wires or dots would be their incorporation into
the active region of a laser diode, which has been predicted to significantly improve
present performance [1,2]. The improvements result from the narrower optical gain
spectrum provided by these quantum structures. The optical gain spectrum corresponds
roughly to the density of states function as shown in figure 5.2 for bulk material, quantum
well, quantum wire, and quantum dot. The narrowing of the density of states function
near E,,, with increasing quantum confinement, and consequently the narrowing of the
gain spectrum, is apparent. The narrowing of the gain spectrum would result in reduced
threshold current and linewidths for laser diodes containing quantum wires or dots.
A variety of approaches is being studied to achieve lateral confinement for quantum
wires and dots. Lateral confinement has been produced in an already existing quantum
well by defining a confinement pattern using techniques such as electron-beam or ion-
beam lithography and then transferring this pattern into the quantum well by etching
[3-6], impurity-induced disordering using ion-beam implantation [7], or selective
V-3
Quantum Wells, Quantum Wires, and Quantum Dots
Quantum
Wire
Ly
Ly
Y x Quantum
Dot
7 9) c)
ID-QW 2D-QW 3D-QW
“Potential
Barrier
(a) (b) (c)
Figure 5.1. Top: schematic drawings of a quantum well (a), a quantum wire (b), and a
quantum dot (c). Bulk material is also illustrated for comparison. Bottom: schematic
drawings of three-dimensional arrays of quantum wells (a), quantum wires (b), and
quantum dots (c).
Density of States Functions
f BULK { Q-WELL
p(E) PoE)
E sop E E sop E
A f
Q-WIRE Q-DOT
p, &E) PofE)
E gap E E gap E
Figure 5.2. The density of states function for bulk material (BULK), quantum well (Q-
WELL), quantum wire (Q-WIRE), and quantum dot (Q-DOT).
V-5
diffusion [8]. However, such methods suffer from free-surface effects, creation of a
damage field during implantation, or lack of interface control due to the random nature of
the disordering mechanism. Methods have emerged which allow the formation of
nanostructures during the epitaxial growth process itself. Growth on patterned substrates
has shown interesting compositional modulation effects [9] and has demonstrated
stimulated emission from a single quantum wire [10]. Wires grown by migration-
enhanced epitaxy on a tilted substrate have exhibited optical anisotropies associated with
confinement effects [11]. These latter methods have not yet been used to produce
quantum dot structures however.
In this chapter, the application of selective epitaxy to the fabrication of arrays of
nanometer-scale wires and dots is presented [12]. Selective epitaxy, described earlier in
chapter 4, refers to the laterally- and spatially-controlled growth of epitaxial material
within openings of a masking material. In the work presented here, the structures grown
are determined by patterning a masked substrate using electron-beam lithography. This
allows the selective growth of GaAs wires and dots which can be followed by a single-
crystal growth of Al,Ga,.,As to embed the structures in situ. Using cathodo-
luminescence (CL) scanning electron microscopy, spectra of single dots and wires as well
as spectrally-resolved CL images are presented.
5.2 Sample Preparation and Growth
The samples prepared for selective epitaxy comprised a 2 um undoped
Alp 3gGag 62AS layer followed by a 20 nm GaAs cap layer grown by OMVPE on semi-
insulating (100) GaAs substrates. For the masking material, 20 nm of silicon nitride was
formed by plasma-enhanced chemical-vapor deposition (PECVD) on the samples.
Arrays of 100-um-long stripe openings with widths ranging from 90 nm to 300 nm were
patterned into the silicon-nitride layer using electron-beam lithography and CF, plasma
etching. The stripe openings were oriented in the [011], [011], and [010] crystallo-
V-6
graphic directions. 100 im x 100 lum square openings were also written into each sample
as control.
The selective growth of GaAs was performed in a low-pressure horizontal OMVPE
reactor. Details of the selective growth process have been described in chapter 4. A
range of samples was grown with the growth temperatures varying from 600°C to 750°C.
Approximately 100 nm of GaAs was grown on the samples as measured from the cross-
section of the square area on the samples. The actual growth thicknesses of the dots and
lines were different depending on the crystallographic orientation of the initial mask
openings. The GaAs was found to grow selectively within the openings of the mask and
the selectivity did not depend on growth temperature. We observed growth of GaAs in
all openings, including the smallest.
5.3 Characterization of Wire and Dot Growths
Figure 5.3a shows a micrograph of an array of dots grown by this technique. The
growth displays a highly facetted profile with excellent size uniformity. The exact
orientations of the crystallographic bounding planes are difficult to measure for such
small structures. In general, the slow-growth crystallographic planes determine the facet
profile of each dot. For the inorganic-based growth of GaAs, these have been observed to
be the {110} and {111} families of planes [13,14]. Selectively grown GaAs wires also
exhibited facetted profiles with bounding planes depending on the crystallographic
orientation of the initial stripe opening. Figure 5.3b shows wires grown oriented in the
[010] direction. For the three wire orientations used in this study, a uniform facet profile
was observed along each wire.
To obtain strong luminescence from such small structures, it is important that no free
surface exists that would generate nonradiative-recombination centers. Also, both types
of carriers (electrons and holes) should be confined within each wire or dot by
surrounding the entire structure with a higher band-gap material. This was accomplished
V-7
i d dots.
res an
SW
ly grown GaA
ive
of selecti
graph
cro
Figure 5.3. Scanning electron m
ic
Crystallographi
les.
(b) show facetted growth profi
in
res i
(a) and the wi
sin
Both the dot
in the figures
icated
S are as ind
orientation
V-8
in a second set of samples in which the GaAs selective growth (identical to the previous
set) was followed by a growth of Al,Ga,_,As (nominally x = 0.3) with an approximate
growth thickness of 250 nm. These samples contained patterns with similar electron-
beam-exposure conditions as the previous set.
The luminescence properties of such passivated structures were investigated using
cathodoluminescence (CL). Details of the CL collection system are given elsewhere [15].
CL spectra at 77°K of single wires and dots were measured as a function of lateral width
and diameter, respectively. These spectra are shown in figures 5.4a and 5.4b. For
comparison, the spectrum of a selectively grown square area on the same sample is also
included in the figures. The luminescence intensity of these square areas was comparable
to conventionally grown GaAs layers. The exact size of the GaAs wire or dot is difficult
to ascertain for the passivated samples since the lateral growth of the Al,Gay_,As
passivation layer is dependent upon the initial size of the GaAs structures, the wire
orientation, the pattern density, and the growth conditions. The sizes were estimated by
using the first set of samples which have similarly exposed patterns on which only GaAs
was grown. For each size range, numerous spectra of dots and wires were taken, showing
only a small distribution in peak position about the luminescence peaks seen in the
figures. As the dimensions of the structures decrease, the luminescence peaks show an
increasing blue shift.
One possible cause of the blue shift is the increasing quantum confinement of the
carriers in the GaAs structures. The shifts, however, are larger than those predicted for
quantum confinement at the estimated structure sizes. There are other possible effects
that might contribute to lateral confinement. Electrostatic band-bending arising from the
free surface on top of the passivating Al,Ga,_,As layer will alter the confinement
potential. Strain created by the surrounding nitride layer could modulate the band gap in
the wires and dots. Both effects could potentially increase the quantum confinement of
the structures.
V-9
Line Spectra
Intensity (arbitrary units)
6] . T . T —
770 790 810 830 850
(a) Wavelength (nm)
Dot Spectra
Intensity (arbitrary units)
0 T “ T * T r
770 790 810 - 830 850
(b) Wavelength (nm)
Figure 5.4. Cathodoluminescence spectra of wires and dots for three different wire widths
and dot diameters, respectively. Plot (a) shows the wire spectra. For reference, spectra
(i) is from a square-area control region. The estimated widths of the wires are (ii) 160
nm, (iii) 140 nm, and (iv) 100 nm. Plot (b) shows the dot spectra. (i) is the control
region. The estimated dot diameters are (ii) 140 nm, (iii) 110 nm, and (iv) 80 nm.
V-10
Spectrally-resolved CL images with the wavelength set at the emission peak of the
arrays of wires and dots were also taken. Typical images of wires and dots are shown in
figures 5.5 and 5.6, respectively. The images shown consisted of 140-nm-wide wires
with 0.83 um center-to-center spacing and 110-nm-diameter dots with 0.66 [1m center-to-
center spacing. The wires and dots exhibit uniform emission intensity. This uniformity
decreased somewhat for the smallest wires and dots.
In a third set of samples, the GaAs selective growth was followed by a much thicker
growth of Al,Ga,_,As (nominally x = 0.2) with an approximate growth thickness of |
jim. These samples contained patterns with similar electron-beam-exposure conditions as
the previous sets. The growth behavior of the GaAs wires with thick Al,Ga,_,As
overgrowths can be seen from the cross-sectional backscattered-electron images in figure
5.7. The thick Al,Ga,_,As overgrowths from the GaAs wires coalesce and form a
continuous corrugated top surface. Depending on the crystallographic orientation of the
initial stripe openings, the corrugated top surface planarizes to form a smooth plane and
allows further growth or processing steps to form useful device structures. Overgrowths
on wires oriented along the [010] direction seem to join and planarize most readily and
perfectly, leaving no void within the final structure.
5.4 Conclusion
The application of selective epitaxy to the fabrication of nanometer-scale wires and
dots was demonstrated. Wires and dots with diameters as small as 70 nm were prepared.
Spectrally-resolved CL showed that the luminescence from single wires and dots was
blue-shifted with respect to the bulk GaAs. Cross-sectional images of GaAs wires with
thick Al,Ga;_,As overgrowths were presented. The Al,Ga,_,As overgrowths eventually
planarize, allowing further growth or processing steps to form useful device structures.
The success of this growth process shows promise for many potential device applications
including quantum-wire and quantum-dot lasers.
V-1l
Se
Bee ce
BIS
ae
BGs
ie
Le
Pan co
ae
Se
ee
es:
Seas
pic RN as ihe Serotec eee
Tne
ves
age of an array of w
scence im.
mine
-resolved cathodolu
Figure 5.5. Spectrally
The wavelength is set to 814 nm.
A(a)
) of figure 5
(iii
corresponding to spectra
V-12
809 nm
is set to
f an array of dots
Se Doe i < : ae
os
ee zs
- :
1 rt
inescence image o
4(b). The wavelength
gure 5
) of
at
ved
mt
resolved cathodolum
fi
gure 5.6. Spectrally
corresponding to spectra
V-13
SS
Ue
s of GaAs wires overgrown with
image
electron i
ional backscattered
The w
-sect
Figure 5.7. Cross
thick Al,Ga,
[O11]
and [011]
g the [010]
dalon
iente
ires are ori
As layers
Xx
irections.
crystallographic d
V-14
Bibliography
[1] Y. Arakawa, K. Vahala, and A. Yariv, Appl. Phys. Lett. 45, 950 (1984).
[2] M. Asada, Y. Miyamoto, and Y. Suematsu, J. Quantum Electron. QE-22, 1915
(1986).
[3] K. Kash, A. Scherer, J. M. Worlock, H. G. Craighead, and M. C. Tamargo, Appl.
Phys. Lett. 49, 1043 (1986).
[4] H. Temkin, G. J. Dolan, M. B. Panish, and S. N. G. Chu, Appl. Phys. Lett. 50, 413
(1987).
[5] D. Gershoni, H. Temkin, G. J. Dolan, J. Dunsmuir, S$. N. G. Chu, and M. B. Panish,
Appl. Phys. Lett. 53, 995 (1988).
[6] B. I. Miller, A. Shahar, U. Koren, and P. J. Corvini, Appl. Phys. Lett. 54, 188 (1989).
[7] J. Cibert, P. M. Petroff, G. J. Dolan, S. J. Pearton, A. C. Gossard, and J. H. English,
Appl. Phys. Lett. 49, 1275 (1986).
[8] H. A. Zarem, P. C. Sercel, M. E. Hoenk, J. A. Lebens, and K. J. Vahala, Appl. Phys.
Lett. 54, 2692 (1989).
[9] M. E. Hoenk, C. W. Nieh, H. Chen, and K. J. Vahala, Appl. Phys. Lett. 55, 53 (1989).
[10] E. Kapon, D. M. Hwang, and R. Bhat, Phys. Rev. Lett. 63, 430 (1989).
[11] M. Tsuchiya, J. M. Gaines, R. H. Yan, R. J. Simes, P. O: Holtz, L. A. Coldren, and
P. M. Petroff, Phys. Rev. Lett. 62, 466 (1989).
[12] J. A. Lebens, C. S. Tsai, K. J. Vahala, and T. F. Kuech, Appl. Phys. Lett. 56, 2642
(1990).
{13] T. F. Kuech, M. A. Tischler, and R. Potemski, Appl. Phys. Lett. 54, 910 (1989).
[14] D. W. Shaw, J. Cryst. Growth 31, 130 (1975).
[15] M. E. Hoenk and K. J. Vahala, Rev. Sci. Instrum. 60, 226 (1989).
VI-1
Chapter 6
Formation of Highly-Uniform and Densely-Packed
Arrays of GaAs Dots by Selective Epitaxy
Formation of highly-uniform and densely-packed arrays of GaAs dots by selective
epitaxy using diethylgallium chloride and arsine is described. The arrays of GaAs dots
are imaged using atomic-force microscopy (AFM). Accounting for the AFM tip radius of
curvature, the smallest GaAs dots formed are 15-20 nm in base diameter and 8-10 nm in
height with slow-growth crystallographic planes limiting growths of individual dots.
Completely selective GaAs growth within dielectric-mask openings at these small size-
scales is also demonstrated. The uniformity of the dots within each array ranged from 6%
for the larger dots to 16% for the smallest dots (normalized standard deviations of the
areas of individual dots within each array).
6.1 Introduction
Semiconductor structures that exhibit quantum-confinement effects in three
dimensions (quantum dots) have attracted considerable attention for their potential in
improving optoelectronic devices as discussed earlier in chapter 5. For these optical
effects to be observable at room temperature, the necessary dimensions are on the order
of 10-20 nm. Many approaches to form these nanostructures (quantum dots) have been
studied, including selective disordering [1] or physical patterning [2,3] of epitaxially-
grown quantum-well materials, in situ strain-induced formation of epitaxial islands [4,5],
and many others. Selective epitaxy, which refers to the laterally-controlled growth of
VI-2
epitaxial material within openings of a masking material, has also produced very
promising results to date. In particular, highly-organized dot structures exhibiting
excellent uniformity and good luminescence efficiency have been formed by selective
epitaxy [6-8].
In this chapter, the successful formation of highly-uniform and densely-packed arrays
of GaAs dots by selective epitaxy using diethylgallium chloride and arsine [9,10] is
described. Imaged by atomic-force microscopy (AFM), the GaAs dots are as small as 15-
20 nm in base diameter and 8-10 nm in height. Completely selective GaAs growth within
dielectric-mask openings is also confirmed at these small size-scales using AFM. This is
the first demonstration of selective epitaxial growth in dielectric-mask openings at these
small size-scales [11].
6.2 Dot Formation by Selective Epitaxy
The substrates prepared for this study contained a 2 um Si-doped Alp 3Gap 7As layer
and a 10 nm undoped GaAs cap layer, both deposited by low-pressure organometallic
vapor-phase epitaxy (OMVPE), on Si-doped (100) GaAs substrates. A dielectric
masking layer of silicon nitride was deposited by plasma-enhanced chemical-vapor
deposition on the substrates and then annealed under an arsine ambient at typical
OMVPE growth conditions. The thickness of the annealed silicon-nitride layer was
approximately 15 nm before lithographic patterning. 40 um x 40 um arrays of dot
openings with center-to-center spacings of 100 nm were patterned into the silicon-nitride
layer by high-resolution electron-beam lithography and reactive-ion etching in CF4
plasma. The electron-beam dosages used for the lithography were varied to produce
variations in the dot-opening sizes among different arrays.
The growths of the arrays of GaAs dots were performed in an AIXTRON 200/4 low-
pressure horizontal-flow OMVPE reactor ("Reactor B" as described in chapter 2) with
susceptor-disk rotation. Growth precursors for the organometallic chloride (III) and the
VI-3
hydride (V) were diethylgallium chloride (DEGaCl) and arsine (AsH3) respectively. The
DEGaCl-bubbler temperature was maintained at 17°C. The growth temperature was
700°C. The reactor-chamber pressure during growth was 200 mbar. Vapor-phase V/III
ratio was maintained at >100. The patterned growth samples were annealed under an
AsH3/H> ambient at the growth temperature for 10 minutes prior to the actual growth.
No dopant was intentionally introduced.
6.3 Characterization and Analysis by Atomic-Force Microscopy
Atomic-force microscopy (AFM) is based on sensing the microscopic forces between
the tip of an extremely sharp stylus and a surface of interest. The interatomic forces
induce the displacement of the stylus, and in its original implementation, a tunneling
junction was used to detect the motion of a diamond stylus attached to an electrically-
conductive cantilever beam [12]. Subsequently, optical interferometry was used to detect
the cantilever deflection. By maintaining a constant force (constant cantilever deflection)
between the stylus tip and a surface and by scanning the stylus over the surface, an image
of the surface profile can be generated.
The current AFM [13,14], utilized for this work, uses a position-sensitive photo-
detector to detect the cantilever deflection, as depicted in figure 6.1. A laser beam is
directed onto the back surface of the cantilever. The optical reflection of the laser beam
off the cantilever changes direction as the cantilever is deflected by surface features of a
sample. The optical reflection is detected by the position-sensitive photodetector,
comprising two juxtaposed semicircular photodiodes. Cantilever deflection is measured
as the difference between the output signals from these two photodiodes. For generating
an image of the surface profile of a sample, a piezoelectric XYZ translator moves the
sample and maintains a constant cantilever deflection, to provide a constant contact force
between the stylus tip and the sample.
VI-4
tht
XYZ Ih
TRANSLATOR
Figure 6.1. Schematic diagram of an atomic-force microscope using optical detection of
cantilever deflection.
VI-5S
The arrays of GaAs dots were imaged by the AFM as described. The operating
parameters for the AFM were optimized to image these small dot structures. The contact
force chosen for the etch-sharpened-silicon AFM stylus tip was an empirical compromise
between fast tip-wear and tip lift-off from the granular surface containing aciculate
features. For the AFM tip to closely track these surface features, the electrical gain for tip
feedback was maximized and slow scan rates and small scan areas were applied. Each
AFM image took approximately 20 minutes to complete. The acquired AFM image data
were plane-fitted to remove the effects of thermal drifts and small tilts from the samples
and/or the AFM. No other image-processing procedures were applied to enhance the
image data.
Plane views and three-dimensional views of the arrays of GaAs dots were generated
using the AFM. Figure 6.2 contains three plane-view AFM images of the GaAs dots
from different arrays. The apparent sizes of the dots are: (a) 22 nm, (b) 35 nm, and (c) 57
nm. Figure 6.3 contains two three-dimensional-view AFM images of the same arrays of
GaAs dots from figure 6.2 (figure 6.3a corresponds to figure 6.2b and figure 6.3b to
figure 6.2c). These AFM images of the arrays of GaAs dots illustrate that the dots are
highly uniform and densely packed within each array. By analyzing the AFM image
data, the areas of individual dots within each array were determined. The normalized
standard deviations of the areas of individual dots were 16% for the array illustrated in
Figure 6.2a, 10% for the array in Figure 6.2b, 8% for the array in Figure 6.2c, and 6% for
the arrays with larger dots. In addition, these images confirm that the GaAs deposition is
entirely selective under these growth conditions. No deposition occurred outside of the
dielectric-mask openings.
After the initial AFM imaging, the samples were etched in a low-bias-voltage CF,4/O
plasma to remove the silicon-nitride layer. Figure 6.4 contains two three-dimensional-
view AFM images of the arrays of GaAs dots after the removal of silicon nitride. The
array in figure 6.4a corresponds to the array in figure 6.2a, and figure 6.4b to figure 6.2b.
VI-6
200
-100
0 100 +~=}=200 nm
Figure 6.2. Atomic-force micrographs in plane view of arrays of (a) small, (b) medium,
and (c) large GaAs dots after growth. The actual scan area is 500 nm x 500 nm. The
height (Z) is represented by gray scale. The dot center-to-center spacing is 100 nm.
VI-7
Figure 6.3. Atomic-force micrographs in three-dimensional view of arrays of (a) medium
and (b) large GaAs dots after growth. The actual scan area is 500 nm x 500 nm and the
height range is 40 nm.
VI-8
Figure 6.4. Atomic-force micrographs in three-dimensional view of arrays of (a) small
and (b) medium GaAs dots after removal of silicon nitride. The actual scan area is 500
nm xX 500 nm and the height range is 40 nm.
VI-9
The apparent sizes of the dots are (a) 44 nm and (b) 55 nm. These images appear to be
noisier than the previous images. Possible reasons for the noise include chemical
contamination (a result of taking the samples off the AFM mount), uneven plasma
etching, and/or surface roughening from the plasma etching.
The height and the base diameter of the GaAs dots, with the silicon-nitride layer
removed, were measured to produce the plot in figure 6.5. A best-fit line is also drawn in
this figure. The x-intercept of this line is not at zero base diameter, but instead at ~29 nm,
indicating that the AFM tip is affecting the determination of the actual dot size. The inset
on figure 6.5 illustrates how an idealized AFM tip with a radius of curvature R could
affect the imaging of a small dot structure. The x-intercept of the best-fit line at ~29 nm
corresponds reasonably well with the tip specifications from the manufacturer (tip radius
of curvature of 10-30 nm) and tip-wear during scanning.
The slope of the best-fit line is ~0.48, indicating that the growth of the dots is self-
limited, most likely by 45° planes or (110)-type crystallographic planes (0.476 converts
to 43.6° facet angle from the horizontal plane), which are the slow-growth crystallo-
graphic planes previously observed in this growth process [9,10]. Corrected data,
accounting for the AFM tip radius of curvature, indicate that the dots are as small as 15-
20 nm in base diameter and 8-10 nm in height with the growth of individual dots self-
limited by slow-growth crystallographic planes.
6.4 Conclusion
The formation of highly-uniform and densely-packed arrays of GaAs dots by
selective epitaxy using diethylgallium chloride and arsine has been achieved. Atomic-
force micrographs of the arrays of GaAs dots have been presented. Accounting for the
limitations imposed by the AFM tip radius of curvature, the smallest GaAs dots
fabricated are 15-20 nm in base diameter and 8-10 nm in height with slow-growth
crystallographic planes limiting individual dot growth. Completely selective GaAs
VI-10
E 20+
~~ R tan(22.5°)
@ 15/4 4
ot) H
r=
Q 107 Scan Result J i j J j j I Dot Diameter (nm) Figure 6.5. Plot of GaAs dot height versus base diameter, measured after silicon-nitride microscope tip, with a radius of curvature R, on the imaging scan of a dot structure with 45° facets and base width D. VI-11 deposition within dielectric-mask openings is also confirmed at these small size-scales. within each array). VI-12 Appl. Phys. Lett. 49, 1275 (1986). [2] K. Kash, A. Scherer, J. M. Worlock, H. G. Craighead, and M. C. Tamargo, Appl. [3] H. Temkin, G. J. Dolan, M. B. Panish, and S. N. G. Chu, Appl. Phys. Lett. 50, 413 [4] D. Leonard, M. Krishnamurthy, C. M. Reaves, S. P. Denbaars, and P. M. Petroff, [5] J. Oshinowo, M. Nishioka, S. Ishida, and Y. Arakawa, Appl. Phys. Lett. 65, 1421 [6] J. A. Lebens, C. S. Tsai, K. J. Vahala, and T. F. Kuech, Appl. Phys. Lett. 56, 2642 [7] Y. D. Galeuchet, H. Rothuizen, and P. Roentgen, Appl. Phys. Lett. 58, 2423 (1991). [8] Y. Nagamune, S. Tsukamoto, M. Nishioka, and Y. Arakawa, J. Crystal Growth 126, [9] T. F. Kuech, M. S. Goorsky, M. A. Tischler, A. Palevski, P. Solomon, R. Potemski, [10] T. F. Kuech, M. A. Tischler, and R. Potemski, Appl. Phys. Lett. 54, 910 (1989). {11} C. S. Tsai, R. B. Lee, and K. J. Vahala, Mat. Res. Soc. Symp. Proc. 358, [12] G. Binnig, C. F. Quate, and C. Gerber, Phys. Rev. Lett. 56, 930 (1986). [14] A. L. Weisenhorn, P. K. Hansma, T. R. Albrecht, and C. F. Quate, Appl. Phys. Lett. VIil-1 Chapter 7 Facet-Modulation Selective Epitaxy - a Technique for The technique of facet-modulation selective epitaxy and its application to quantum- resolved cathodoluminescence images of the wire doublets are presented. 7.1 Introduction to Facet-Modulation Selective Epitaxy Semiconductor structures exhibiting quantum confinement in two or three dimensions organometallic vapor-phase epitaxy (OMVPE) growths on etched substrates. Wire and VIUl-2 dot structures have been selectively grown on substrates covered with patterned dielectric In this chapter, a new technique, facet-modulation selective epitaxy [17], is described and its application to quantum-well wire doublet fabrication in the Al,Ga,_,As material system is presented. The origina: concept that initiated this work is illustrated in figure Two precursor chemistries are used here, a combination of diethylgallium chloride (DEGaC]) and arsine (AsH3) for GaAs growth and a combination of trimethylaluminum morphology of GaAs selective growth using DEGaC] and AsH; is dominated by the appearance of the {111} and {110} families of slow-growth crystallographic planes, with the appearance of a particular plane dependent on the growth temperature and the mask opening orientation [11,13,18]. The morphology of Al,Gay_,As selective growth using TMAI, TMGa, and AsHs is similar, but the bounding planes include higher-index- VI-3 Fabrication of Quantum Wires by YS ‘Substrate Single Quantum Ha sei Ke Quantum Cla Figure 7.1. Schematic illustration of the original concept that preceeded facet- modulation selective epitaxy. VIl-4 number planes (one or more indices greater than one) in addition to the {111} and {110} well wire doublets in a single growth as illustrated in figure 7.2. Using DEGaCl and AsHsz, a GaAs buffer layer bounded by the low-index-number facets is grown on a produce nonradiative-recombination centers. The width of the starting Al,Ga,_,As facets and the amount of GaAs deposited determine the size of the wires and the spacing scopy, and cathodoluminescence scanning electron microscopy is presented. 7.2 Substrate Preparation and Growth Conditions The substrates used in this study contained a 2 um Si-doped Alp 4Gag ¢As layer, an undoped 100 nm Alp Gag gAs layer, and a 10 nm GaAs cap layer all deposited by OMVPE on Si-doped (100) GaAs substrates. The dielectric mask was 175 A of silicon along the [011] direction. VII-S Al Ga As rit) 109 porn GaAs Si _ Substrate Figure 7.2. Schematic illustration of quantum-well wire doublet fabrication by facet- modulation selective epitaxy. VIl-6 Sample growth was performed in an atmospheric-pressure OMVPE reactor with layers were DEGaC]I and AsH3. Precursors for the barriers were TMA] and AsH3. AlAs barriers were chosen here to test the extreme case. Growth interruptions were placed between layers while AsH3 flow was maintained. No dopant was intentionally introduced. 7.3 Characterization of Growth Samples Figures 7.3a and 7.3b show a complete structure grown in a stripe opening 1.2 um The wire doublet is shown clearly in figures 7.3c and 7.3d. The transmission electron The luminescence properties of these wire-doublet structures were investigated by distinct peaks appear in the CL spectra, one from the wire doublet at 690 nm and another VIL-7 (b) [100]4 ® é GaAs ISiN 1000 A ) AlAs [100] [211] “" Wire Doublet ~ AlAs Figure 7.3. (a) [011] cross-sectional backscattered-electron image of a complete structure grown by facet-modulation selective epitaxy. (b) Schematic illustration of the complete structure. (c) [01 1 ] cross-sectional transmission electron micrograph of the wire doublet. (d) Schematic illustration of the wire doublet. VIH-8 7 771 | nn Wavelength (nm) Figure 7.4. Cathodoluminescence spectra of the region near the wire doublet. The sample is at 14°K. VII-9 from the side-wall quantum wells at 625 nm. The red shifting of the wire-doublet peak Figures 7.5a-c are spectrally-resolved CL images of the wire-doublet structure in configurations of wire doublets. 7.4 Conclusion fabrication of quantum-well wire doublets have been described. Successful fabrication VIL-10 Figure 7.5, Spectrally-resolved cathodoluminescence images of a wire-doublet structure in cross-section. The sample is at 12°K, and the wavelengths are set at (a) 625 nm, (b) 693 nm, and (c) 820 nm. (d) Scanning electron micrograph of the same structure in cross-section. VI-11 Figure 7.6. Spectrally-resolved cathodoluminescence images of different wire-doublet structures on the same sample. The sample is at 77°K, and the wavelengths are set at (a)- (c) 690 nm and (d) 680 nm. VI-12 fabricated has a crescent cross-section less than 140 A thick and less than 1400 A wide. In,Ga,_,As, by using analogous precursor chemistries [13]. VII-13 [1] Y. Arakawa, K. Vahala, and A. Yariv, Appl. Phys. Lett. 45, 950 (1984). [2] M. Asada, Y. Miyamoto, and Y. Suematsu, IEEE J. Quantum Electron. QE-22, 1915 [3] P. C. Sercel and K. J. Vahala, Appl. Phys. Lett. 57, 545 (1990) [4] K. Kash, A. Scherer, J. M. Worlock, H. G. Craighead, and M. C. Tamargo, Appl. [5] H. Temkin, G. J. Dolan, M. B. Panish, and S. N. G. Chu, Appl. Phys. Lett. 50, 413 [6] B. I. Miller, A. Shahar, U. Koren, and P. J. Corvini, Appl. Phys. Lett. 54, 188 (1989). [7] J. Cibert, P. M. Petroff, G. J. Dolan, S. J. Pearton, A. C. Gossard, and J. H. English, [8] M. Tsuchiya, J. M. Gaines, R. H. Yan, R. J. Simes, P. O. Holtz, L. A. Coldren, and P. [9] E. Kapon, D. M. Hwang, and R. Bhat, Phys. Rev. Lett. 63, 430 (1989). [10] M. Walther, E. Kapon, C. Caneau, D. M. Hwang, and L. M. Schiavone, Appl. Phys. [11] J. A. Lebens, C. S. Tsai, K. J. Vahala, and T. F. Kuech, Appl. Phys. Lett. 56, 2642 [12] T. Fukui, S. Ando, and Y. K. Fukai, Appl. Phys. Lett. 57, 1209 (1990). [13] T. F. Kuech, M. S. Goorsky, M. A. Tischler, A. Palevski, P. Solomon, R. Potemski, [14] Y. D. Galeuchet, H. Rothuizen, and P. Roentgen, Appl. Phys. Lett. 58, 2423 (1991) [15] M. Nishioka, S. Tsukamoto, Y. Nagamune, T. Tanaka, and Y. Arakawa, J. Crystal vVil-14 [16] S. Ando and T. Fukui, J. Crystal Growth 98, 646 (1989) [17] C. S. Tsai, J. A. Lebens, C. C. Ahn, A. Nouhi, and K. J. Vahala, Appl. Phys. Lett. [18] T. F. Kuech, M. A. Tischler, and R. Potemski, Appl. Phys. Lett. 54, 910 (1989) [19] M. E. Hoenk and K. J. Vahala, Rev. Sci. Instrum. 60, 226 (1989) VIll-1 Chapter 8 Visible-Wavelength Vertical-Cavity The development of OMVPE epitaxial layers for a visible-wavelength vertical-cavity Ing 5Gap 5P/Ing 5Alg 5P heterostructure p-n diodes, processed by using the actual fabrication procedures for VCSEL, were prepared as an intermediate test structure for the micrographs of these full structures are also presented. 8.1 Introduction to Visible- Wavelength VCSEL low threshold currents, and ease of integrating multiple VCSEL into closely spaced one- VI-2 and two-dimensional arrays or integrating with other devices such as transistors. In Recently, photopumped [1,2] and electrically-injected [3,4] lasing of visible- of the demonstrated VCSEL devices comprised a p-type top mirror with alternating Alp 5Gag sAs/AlAs A/4 layers, an active region of Ing sGag 5P/Ing 5Al,Gap 5_,P/ Alg.5Gag 5As/AIAs A/4 layers, and the necessary electrical contact metallization and isolation. Proton implantation is illustrated as one possible method of isolating the An alternative VCSEL structure is shown in figure 8.1b which eliminates the high series resistance in the top contact. A SiO/SiN dielectric top mirror is utilized in place of the Alo sGap 5As/AIAs p-type top mirror. Annular metal contact is made directly to the active region of Ing 5Gag 5P/Ing sAlg »>Gag 3P/Ing sAlp 5P MQW. The device junction is entirely encircled by dielectric material. This alternative VCSEL structure would have alternative VCSEL structure. VIl-3 (a) Laser ae Figure 8.1. Schematic illustration of a conventional VCSEL structure (a) and an etched VCSEL structure with dielectric top mirror (b). VIli-4 8.2 Growth and Characterization of (/4-Stack Mirror The VCSEL structures pose extremely challenging requirements for any epitaxial VCSEL. several passive fabry-perot cavities, each comprising one Alo 5Gag 5As A/2 layer placed The reflectance of the passive fabry-perot cavities were measured in an optical setup at the measured wavelengths), and inaccurate layer thicknesses used in the VIL-5 GaAs cap AASAASAAASASSS ECC SS CC SSC Ci ANASAREASASAAAAAAAAAAAAAS ARSE CRASSA SASAASAAAA A AAAAAAASASAAAANARAARASAEAS EASA ACA AANA SATAS AAA A TC CCC ane A AAAKASAASASASAAAAA SEAS ASE SAAC CLES SC SLE SAS NN Buffer GaAs Figure 8.2. Schematic illustration of the passive fabry-perot cavity used as a test structure for mirror calibration. VII-6 Reflectance of passive fabry-perot cavity 1 ee Reflectance (relative) 600 625 650 675 700 725 750 775 800 Figure 8.3. Plot of the measured and calculated reflectance of the passive fabry-perot cavity illustrated in figure 8.2. VIU-7 simulation model. Further refinement of the calculated reflectance is possible by using a The surface defect-density of the A/4-stack mirror was reduced to a negligible level process. Cross-sectional transmission electron micrographs (TEM) of the A/4-stack mirror are shown in figures 8.4c at 52,000X and 8.4d at 122,000X. These micrographs show that the layers in the A/4-stack mirror are uniform and well-defined with extremely abrupt appear oxidized as a result of exposure to atmospheric oxygen. 8.3 Growth and Characterization of MQW Active Region Ing 5Al,Gag 5.,P materials [8] lattice-matched to GaAs are utilized in the active material composition and the lattice constant usually result from drifts in the vapor VIIL-8 Figure 8.4. Optical micrographs at 680X of the initial (a) and the final (b) test structures electron micrographs of the mirror at 52,000X for (c) and 122,000X for (d). VII-9 pressure of the trimethylindium (TMIn) source, because TMIn is utilized in the form of a For electrically-injected VCSEL devices, p-type and n-type dopants are also Ing 5Gag 5P/Ing 5Alo sP heterostructure p-n diodes, processed by using the actual VCSEL active region. The heterostructure p-n diodes comprised an undoped 100 nm- thick Ing 5Gag 5P layer placed in between a p-type | sm-thick Ing 5Alg 5P layer and a n- type 200 nm-thick Ing 5Gap 5P layer was grown on the very top of the structure to reduce electrical contact resistance. Optical micrographs of the completed diodes are shown in figures 8.5a at 380X and the bottom mirrors and the substitution of the MQW active region with the Ing 5Gag 5P/Ing 5sAlg 5P heterostructure. VIU-10 Figure 8.5. Optical micrographs of Ing 5Gag 5P/Ing 5Alp 5P heterostructure p-n diodes at 380X for (a) and 95X for (b). Vill-11 Strong electroluminescence was obtained from the heterostructure p-n diodes. The is Clearly visible in figure 8.5. After the successful demonstration of electroluminescence from the Ing 5Gag 5P/ four Ing sAlg Gag 3P barriers can be seen clearly in the center of the micrograph. VIll-12 Output spectrum of InGaP/InAIP p-n diode 0.6 + 0.4 + Intensity (arbitrary units) 0.2 + Wavelength (nm) Figure 8.6. Electroluminescence spectrum of the Ing 5Gag 5P/Ing 5Alo 5sP heterostructure p-n diode. The luminescence peak is at 685 nm. VIO-13 Figure 8.7. Optical micrographs at 680X of the various test structures (a)-(c) for the transmission electron micrograph (d) of the MQW active region at 122,000X. VIll-14 this micrograph are artifacts from TEM sample preparation. The sample imaged in this The defect density of the MQW active region was also reduced to a negligible level rearranging the gas-switching sequences during the growth of the p-type layers. 8.4 Growth of Complete VCSEL Structure The chronology of the OMVPE development of the visible-wavelength VCSEL defects arose solely from the MQW active region. The third full VCSEL structure VITL-15 Figure 8.8. Optical micrographs at 680X of four full VCSEL structures containing the VIU-16 grown is shown in figure 8.8c. The defect density of the MQW active region had been VCSEL. VITI-17 1830 (1992). [2] R. P. Schneider, Jr., R. P. Bryan, J. A. Lott, E. D. Jones, and G. R. Olbright, J. [3] R. P. Schneider, Jr. and J. A. Lott, Appl. Phys. Lett. 63, 917 (1993). [4] K. Tai, K.-F. Huang, C.-C. Wu, and J. D. Wynn, Appl. Phys. Lett. 63, 2732 (1993). [6] D. E. Aspnes, S. M. Kelso, R. A. Logan, and R. Bhat, J. Appl. Phys. 60, 754 (1986). [8] H. Tanaka, Y. Kawamura, and H. Asahi, J. Appl. Phys. 59, 985 (1986). [9] R. W. Glew, J. Crystal Growth 68, 44 (1984). [10] E. Veuhoff, T. F. Kuech, and B. S. Meyerson, J. Electrochem. Soc. 132, 1958
% 10 20 30 40 50 60 70 80
removal. A best-fit line of the measured data is also shown with x-intercept of 29.4 nm
and slope of 0.476. Inset: idealized schematic illustrating the effect of an atomic-force
The size uniformity of the dots within each array ranged from 6% for the larger dots to
16% for the smallest dots (normalized standard deviations of the areas of individual dots
Bibliography
[1] J. Cibert, P. M. Petroff, G. J. Dolan, S. J. Pearton, A. C. Gossard, and J. H. English,
Phys. Lett. 49, 1043 (1986).
(1986).
Appl. Phys. Lett. 63, 3203 (1993).
(1994).
(1990).
707 (1993).
C.S. Tsai, J. A. Lebens, and K. J. Vahala, J. Crystal Growth 107, 116 (1991).
Microcrystalline and Nanocrystalline Semiconductors, 969 (1995).
[13] G. Meyer and N. M. Amer, Appl. Phys. Lett. 53, 1045 (1988).
54, 2651 (1989).
Quantum-Well Wire Doublet Fabrication
well wire doublet fabrication are described. Successful fabrication of wire doublets in the
Al,Ga,_,As material system is achieved. The smallest wire fabricated has a crescent
cross-section less than 140 A thick and less than 1400 A wide. Backscattered-electron
images, transmission electron micrographs, cathodoluminescence spectra, and spectrally-
have attracted considerable attention for their potential in improving optoelectronic
devices [1,2] and in revealing new phenomena in solid state physics, such as polarization
anisotropy in quantum wires [3]. Many approaches to fabricate these quantum-
confinement structures have been studied. Already grown quantum well material has
been physically patterned by a combination of lithography and etching [4-6] or has been
selectively disordered by ion implantation [7] to achieve lateral confinement. In situ
formation of nanostructures during epitaxial growths has also been studied. Migration-
enhanced epitaxy on tilted substrates has exhibited quantum-confinement effects [8], and
stimulated emission from quantum wires [9,10] has been demonstrated by performing
masks [11-15]. The formation of crystal facets in single precursor chemistry (trimethyl)
selective growth has also been studied as a potential method for quantum-well wire
fabrication [16].
7.1, One or more alternating layers of GaAs and Al,Ga,_,As are deposited selectively
within stripe openings of a dielectric mask on a substrate. The facets thus formed would
create angular bends in the GaAs layers and form regions of quantum confinement
located at the apexes of the selectively grown structures. Switching of the growth
chemistries was not a part of the original concept. Due to limitations imposed by the
OMVPE reactor utilized here ("Reactor A" as described in chapter 2), trimethyl-based
precursor chemistry had to be used for Al,Ga,_,As growths. In order to maintain some
selectivity, GaAs layers were grown using diethylgallium chloride (DEGaCl). Asa
result, the original concept evolved into facet-modulation selective epitaxy.
Facet-modulation selective epitaxy, as it is defined here, is the application of different
precursor chemistries to layers within a single growth to alter sequentially the appearance
of facets on a growing structure and to thereby form heterostructures of novel geometry.
(TMAI), trimethylgallium (TMGa) and arsine (AsH3) for Al,Ga;_,As growth. The
Facet-Defined Selective Epitaxy
Y, 55 YY
families of slow-growth planes.
One application of facet-modulation selective epitaxy is the fabrication of quantum-
substrate covered with a stripe-patterned mask. Then an Al,Gaj_,As layer is grown using
TMAI, TMGa, and AsH3. Using DEGaC] and AsH3 again, a GaAs wire doublet is
formed as the higher-index-number Al,Ga,_,As facets are partially filled in by the GaAs
growth attempting to form low-index-number GaAs facets. The wire doublet is buried in
situ by another layer of Al,Ga;_,As growth to eliminate any free surface that would
between wires in a wire doublet. In the work presented here, successful fabrication of
wire doublets by facet modulation selective epitaxy is achieved. Analysis of these wire
doublet structures by backscattered-electron microscopy, transmission electron micro-
nitride deposited by plasma-enhanced chemical-vapor deposition. Arrays of 5-mm-long
stripe openings, with stripe widths varying from 1 Um to 8 um and center-to-center
spacing between stripes of 250 tm, were patterned into the silicon-nitride mask by
photolithography and reactive-ion etching in CF, plasma. The stripes were oriented
x I-x
Buffer
x Y [011] Cross Section x y
graphite susceptor temperature set to 730°C. Precursors for the buffer and quantum-well
wide. The structure exhibits a facetted profile bounded mainly by the {111} family of
planes. Measured along the [100] crystal direction from the substrate to the crystal apex,
the growth consists of 1.3 um of GaAs, 0.6 um of AlAs, first quantum well, 0.4 um of
AlAs, second quantum well, 0.2 um of AlAs, third quantum well, and finally 0.13 um of
AlAs. Measured along the [111] crystal direction on the (111) facet, the first, second, and
third quantum wells are approximately 10, 20, and 5 nm thick respectively.
micrograph in figure 7.3c reveals the higher-index-number {211}-type AlAs facets near
the crystal apex. The micrograph also shows the growth of the third quantum well
partially filling in the two {211}-type facets, thereby forming a wire doublet as predicted
earlier. The two wires in the wire doublet are almost identical in shape and size. Both
wires are approximately 140 A thick at the center and 1400 A wide, although the
effective width of these wires is probably smaller than 1400 A, because the wires have a
tapered cross-section.
applying low-temperature cathodoluminescence (CL) scanning electron microscopy [19].
Figure 7.4 contains the CL spectra of the sample region near the wire doublet. Two
[lil] ,
IAs
Buffer
Substrate xy
fs p2aay
As
On Quantum Well
g ust a
an *
~ X 1 On Wire Doublet
al
jm
1.04
Se)
or)
= 0.54
= 0.
0.04 ’ { + t } + } + —} t _
600 620 640 660 680 700 720
with respect to the side-wall quantum-well peak is as expected from the thickness
modulation seen in figure 7.3c. However, it is clear from the CL spectra and the
thicknesses measured from figure 7.3c that the wire doublet and the side-wall quantum
wells are most likely Alp 7Gag gAs, not GaAs, because a 140 A GaAs quantum well at
14°K luminesces at about 800 nm, not at 690 nm as shown in the CL spectra, and a 50 A
well luminesces at about 710 nm, not at 625 nm. Here the wire doublet is approximated
as a quantum well because the widths of the wires are much larger than their thicknesses.
The presence of aluminum is due to interlayer mixing, presumably caused by incomplete
purging of TMAI before the DEGaC] growth had begun. The wire-doublet luminescence
peak at 690 nm also appears broadened. Possible reasons for this broadening may
include the fluctuations in the widths of the {211}-type AlAs facets, the surface
roughness on the {211}-type AlAs facets, or some aluminum segregation resulting from
mixing the two precursor chemistries. |
cross-section. Figure 7.5a is imaged at 625 nm where the side-wall quantum wells and
the substrate Alp 4Gag ¢As luminesce. Figure 7.5b is imaged at 693 nm, the
luminescence peak of the wire doublet. Figure 7.5c is imaged at 820 nm, the
luminescence wavelength of the buffer and substrate GaAs. The scanning electron
micrograph in figure 7.5d shows the cross-section imaged in figures 7.5a-c. Figures
7.6a-d are spectrally-resolved CL images of structures seeded by stripe openings of
different widths on the same sample. These images demonstrate the various
The technique of facet-modulation selective epitaxy and its application to the
of wire doublets in the Al,Ga,_,As material system has been achieved. The smallest wire
By reducing the width of the starting Al,Ga,_,As facets and the amount of GaAs
deposited, further reduction in wire size is possible. By doping the layers above and
below the wire doublet p-type and n-type respectively, and by embedding the wire
doublet in an Al,Ga;_,As optical waveguide, an injection laser may be fabricated. Other
applications could also be explored, including the fabrication of quantum dots and the
extensions of this technique to other material systems besides Al,Ga,_,As, such as
Bibliography
(1986).
Phys. Lett. 49, 1043 (1986).
(1987).
Appl. Phys. Lett. 49, 1275 (1986).
M. Petroff, Phys. Rev. Lett. 62, 466 (1989).
Lett. 62, 2170 (1993).
(1990)
C. S. Tsai, J. A. Lebens, and K. J. Vahala, J. Crystal Growth 107, 116 (1991)
Growth 124, 502 (1992).
60, 240 (1992).
Surface-Emitting Laser by OMVPE
surface-emitting laser (VCSEL) is presented. Passive fabry-perot cavities with
Alp 5Gap 5AS/AIAs A/4-stack mirrors were grown as test structures for mirror
calibration. The defect density of the mirror layers was reduced to a negligible level by
optimizing gas-switching sequences during OMVPE growth. Transmission electron
micrographs and reflectance measurements of one calibration cavity are presented.
Ing 5Gap 5P/Ing 5Alp 9Gap 3P/Ing 5Alp 5P multiple-quantum-well (MQW) active region.
Electroluminescence spectrum of the p-n diode is presented. Test structures containing
only the MQW active region were grown. The defect density of the MQW active region
was also reduced to a negligible level by optimizing the gas-switching sequences. Full
structures containing both the mirror and the MQW active region were grown. Optical
Visible-wavelength vertical-cavity surface-emitting lasers (VCSEL) have attracted
considerable interest because of their surface-normal operation, potential for extremely
addition, the visible-wavelength operation of the VCSEL would allow two-dimensional
arrays for plastic-fiber-based optical communications, for high-brightness optical
displays or projectors, and for replacing conventional light-emitting diodes (LED) and
He-Ne lasers.
wavelength VCSEL has been demonstrated. As seen in figure 8.1a, the overall structure
Ing 5sAlp.5P multiple-quantum-well (MQW), a n-type bottom mirror with alternating
device by making the p-type mirror layers non-conductive surrounding the device. This
structure suffers from high series resistance in the top contact because of the long
circuitous conduction path through the thick p-type mirror (up to 3 um thick).
low series resistance and high modulation speed. In addition, the thickness variations in
the OMVPE epitaxial layers can be compensated by adjusting the thicknesses of the
SiO/SiN top mirror, since the device structure is formed in two separate steps. This
chapter presents the development of OMVPE epitaxial layers toward achieving this
cz LLL EE ZEEE Annular Contact
= < AlGaAs/AIAs
= 3s a Zz p-type mirror
Proton
—— [{_— SS _siImpiantation
inGaP/InAlGaP
‘2.2. £44 £££ 2 £2 2A&bL 2A 2 LA LLL LAE LL Lab aL aE .
MQW Active
f— a Region
L L \ AN
ZL i \ \
Z ji \ \
L L \ \
wi 7 N x AlGaAs/AIAs
wa 4. I" wa n-type mirror
L L \ \
VA vi \ \
L j \ \
VA i \ \
Z i \ \
(b) Laser aan
SiO/SIN
dielectric mirror
key, Annular Contact
4 InGaP/InAlGaP
t LL ot LE LA ed ee .
i MQW encircled
— —_—~ _ by dielectric
ZL. fi \ AN
VA L \ XN
L L \ \
/ L \ AN
ws vi \ — AlGaAs/AIAs
ws vi \ Se n-type mirror
Z. Vi \ \
Z. l \ N
Z [ A N
L. [ \ N
Z. / \ N
growth technique. The high optical efficiency required in the small (~ 1 um thick) laser
cavity demands essentially perfect material with precise control of layer thickness and
composition. The challenges are compounded by the many layers (up to over 100) and
thick epitaxial growths (up to 8 um) which magnify greatly any tiny imperfections in the
epitaxial growth process. Epitaxial problems which are considered minor perturbations
in conventional device structures become major obstacles to the successful operation of a
In order to establish the proper OMVPE growth parameters for the 1/4-stack mirror,
in between two Alo 5Gag 5As/AIAs A/4-stack mirrors, were grown using conventional
OMVPE precursors in "Reactor B" as described in chapters 2 and 4. In this case, the
target A is in the range of 670-680 nm. The complete test structure is illustrated in figure
8.2. The top mirror was capped by a thin layer of GaAs (10 nm) to prevent oxidation of
the mirror.
containing a broad-band white-light source and an optical spectrometer with an attached
photomultiplier tube. In figure 8.3, the measured reflectance of one cavity is plotted
with a solid line. The measured cavity-resonance peak is centered at 674 nm, well
within the target wavelength range. The calculated reflectance of the same cavity is also
plotted. The calculated reflectance was obtained by computer simulation using
published material parameters [5,6] in a conventional characteristic-matrix model [7].
The measured and the calculated reflectance correspond qualitatively. Some of the
factors preventing the exact correspondence between the two include measurement errors
in the optical setup, the use of a simulation model without absorption (GaAs is absorbing
SAAS AT CC Cc ERS
SANSA AAAI p eriods of
AANARARAANSAAAAAASAASAARERAAAAARAAASARAAN NARS ASA RAAT
ANNKAAAARAASANAAAANSAAANSASARANIAAS SA AAANASAN AS ARAL ASSN AAAS SSASAISAS AlGaAs/AlIAs
SSRN ASAAAAANARASAASAAARASASASAAAS ARAL SASASRASA AAAS SAA SONIA quarter-wave
AANAAARRASAAANAAARARASAARAALASAES AAS AAA | ayers
AAXAAEAANSSAAAAAANANAAAAAARASASAAARAAS AS AAR SAAAASARAS AAAS
XXNANARANAARAAAANANSUAANASSARARNANA SSAA AAAS AAAS SSSA SASAS ARIAS
AAR AAALAC ST RCS
AKAAARASAAAASA RE SASSARAA ASA AARA SAS AA EAE SS CC ST CCC NCS
AAAAAAASSSARAAALASEAAR SSSA AAAS TC CTE A AE NS
AAS CC SSCS CCS Ac
AAASARLSARSAAR AAAS ACCS SSAC SAAC ST .
Za
MA
! ¢
KA
KA
AY a
au A
KA
; Z
4 AV
au au
A A AY
A 4
WATT TCE ESS Ct
AAAS AAA AAR AS AC SR Nt
AAMAS AAS SAARC SELECTS CASAC Rt
ARASAAAAASSAS AAA AAAS AACA SCC Se
AAAS AAC CCC CSS TN
WAAC SECS RCC GG ie
Substrate
Measured
OS Calculated
0.6 +
Wavelength (nm)
more comprehensive model with independently and accurately determined layer
thicknesses.
by optimizing gas-switching sequences during OMVPE growth. Several test structures,
each containing only 10 periods of Alo sGag 5As/AIAs A/4 layers, were grown to
optimize the gas-switching sequences. Optical micrographs at 680X of the initial and
the final test structures are seen in figures 8.4a and 8.4b respectively. For figure 8.4b,
the single defect, seen at the center, had to be located specifically, since the defect
density was so low. The number of surface defects was reduced dramatically by
minimizing flow-rate changes during layer-switching transitions in the OMVPE growth
transitions between adjacent layers. The outer edges of the AIJAs layers (lighter color)
regions of visible-wavelength VCSEL structures. Conventional OMVPE precursors are
employed here in the growth of these materials. A precise match in the lattice constants
is required for acceptable growths of these materials. The differences in lattice constants
between the Ing 5Al,Gag 5_,P materials and the GaAs substrate are determined by x-ray
diffractometry. For this study, the relative mismatch measured on materials grown
immediately after calibration is typically less than 2x 10-4 (Aa/a). Variations in the
for growth optimization of the mirror. Also shown are cross-sectional transmission
solid. The total surface area of the solid will change gradually as the TMIn source is
consumed in the OMVPE growth process.
necessary. Zn from diethylzinc [9] and Si from silane [10] are used as p-type and n-type
dopants respectively. The actual doping level is measured by an electrochemical C-V
profiler. The test sample is mounted next to a fluid cell containing the appropriate
electrolyte. The electrolyte forms a Schottky junction with the sample surface. The
Schottky junction is reverse-biased for C-V measurements to determine the doping
levels. The same electrolyte is also used as an etchant to remove thin layers from the
sample surface after each measurement. A depth profile of the doping levels within the
sample is generated by this technique. In this study, the nominal doping levels used in
the layers ranged from 10!7 to 1018/cm3, with higher doping levels used for the
metallization contact layers.
fabrication procedures for VCSEL, were prepared as an intermediate test structure for the
type 1 m-thick Ing sAlp sP layer, all grown on a GaAs substrate. A heavily-doped p-
8.5b at 95X. The actual p-n junctions are the circular mesas seen in the micrographs.
The speckled regions are the electrical interconnections and contact pads. The overall
structure is similar to the one illustrated in figure 8.1b with the elimination of the top and
electroluminescence spectrum of a heterostructure diode is shown in figure 8.6. The
luminescence peak is at 685 nm. The heterostructure diode was electrically excited with
500 t1s-duration constant-voltage pulses with 30 kHz repetition rate. The peak voltage is
set at 6.2 V. The electroluminescence can be observed directly by the eye as an intense
deep-red glow coming from a very small point. Continuous electrical excitation was not
possible due to device heating. The electrical contact metallization would start to bubble
and eventually self-destruct when continuous high voltage (> 5 V) was applied. The
high excitation voltage required and the device heating were attributed to two factors.
First, the p-type Ing sAlp 5P layer was doped too lightly (ow 10!17/cm3) which resulted
in high series resistance and caused considerable device heating. Second, the
metallization for the electrical contact was highly resistive due to rough surface
morphology that had resulted from incorrect etching conditions during sample
processing. For a few of the devices tested, electrical arcing was actually observable in
the metallization as random blue flashes. The speckled morphology of the metallization
Ing sAlo sP heterostructure p-n diode, test structures containing the multiple-quantum-
well (MQW) active region were grown. The MQW active région comprised three
undoped 80A-thick Ing,.5sGap 5P layers separated by four undoped 60A-thick
Ing 5Alp.»Gap 3P layers. These layers are placed in between a p-type 130 nm-thick
Ing 5Alo.5P layer and a n-type 130 nm-thick Ing 5Alp 5P layer, all grown on a n-type
GaAs substrate. A p-type 430 nm-thick Ing 5Alg »Gag 3P layer followed by a heavily-
doped p-type 15 nm-thick Ing ;Gag 5P layer were grown on the very top of the structure
to complete the MQW active region. Cross-sectional TEM micrograph at 122,000X of
the MQW active region is shown in figure 8.7d. Here the three Ing 5Gag 5P wells and the
growth optimization of the MQW active region. Also shown is a cross-sectional
The p-type and the n-type Ing sAlo 5P layers can also be seen. The vertical striations in
TEM micrograph corresponds to the sample in figure 8.7b.
by optimizing the gas-switching sequences. This can be seen in the picture sequence
from figures 8.7a to 8.7c. Seen in figure 8.7a, the first test structure grown containing
only the MQW active region is densely packed with surface defects. The defects are so
dense that their individual sizes are limited by the close spacings between defects.
Through improved gas-switching and the elimination of flow-rate changes through the
OM source bubblers, the defect density was reduced to the extent where the individual
defects were allowed to develop into their full sizes, as seen in figure 8.7b, and a few flat
areas between the defects became noticeable. For figure 8.7c, the single defect, seen at
the center, had to be located specifically, since the defect density was so low. Between
figures 8.7b and 8.7c, the number of surface defects was reduced dramatically by
structure is summarized in the optical micrographs (680X) of figure 8.8. Complete
VCSEL structures including the 4/4-stack mirror and the MQW active region were
grown during various phases of the development. The very first full VCSEL structure
grown is shown in figure 8.8a. This structure contained the first mirror and the first
MQW active region before any optimization of the OMVPE growth parameters. The
surface morphology was extremely rough. The large and dense defects arose from both
the mirror and the MQW active region. The second full VCSEL structure grown is
shown in figure 8.8b. The )/4-stack mirror had been optimized at this point. The
\/4-stack mirror and the MQW active region grown by OMVPE.
reduced considerably. Flat clear areas between defects are clearly visible. Containing
the optimized mirror and the optimized MQW active region, the fourth and final full
VCSEL structure grown thus far is shown in figure 8.8d. The single defect, seen at the
center, had to be located specifically, since the defect density was so low. With this final
VCSEL structure, it should be possible to produce a working visible-wavelength
Bibliography
[1] R. P. Schneider, Jr., R. P. Bryan, J. A. Lott, and G. R. Olbright, Appl. Phys. Lett. 60,
Crystal Growth 124, 763 (1992).
[5] S. Adachi, J. Appl. Phys. 58, R1 (1985).
[7] A. Thelen, Design of Optical Interference Coatings, McGraw-Hill (1989).
(1985).