Growth, Characterization, and Simulation

Growth, Characterization, and Simulation of Novel Semiconductor Tunnel Structures - CaltechTHESIS
CaltechTHESIS
A Caltech Library Service
About
Browse
Deposit an Item
Instructions for Students
Growth, Characterization, and Simulation of Novel Semiconductor Tunnel Structures
Citation
Chow, David Hsingkuo
(1989)
Growth, Characterization, and Simulation of Novel Semiconductor Tunnel Structures.
Dissertation (Ph.D.), California Institute of Technology.
doi:10.7907/96gc-kc14.
Abstract
This thesis presents investigations of novel semiconductor heterostructure devices based on quantum mechanical tunneling. Due to their small characteristic dimensions, these devices have extremely fast charge transport properties. Thus, it is expected that tunnel structure devices will be well-suited to high frequency and optoelectronic applications. The work presented here can be divided into three sections. In the first section, a theoretical model for simulating current-voltage behavior in single barrier heterostructures is developed. The simulations are then used to design a novel single barrier negative differential resistance (NDR) device. The second section consists of detailed experimental characterizations of single barrier Hg
1-x
Cd
Te heterostructures, including the first demonstration of the novel single barrier NDR mechanism. Growth of III-V semiconductor heterostructures by molecular beam epitaxy (MBE) is the subject of the third section. Several aspects of tunneling are explored through characterization of these III-V structures.
In chapter 2, a theoretical model is developed to simulate tunneling currents in single barrier heterostructures. The model includes band bending effects and a two band treatment of electron attenuation coefficients in the barrier. It is proposed that certain material systems have the appropriate band alignments to realize a novel single barrier negative differential resistance mechanism. A thorough theoretical analysis of these single barrier NDR structures is presented.
The first experimental demonstration of the single barrier NDR mechanism is reported in chapter 3. The HgCdTe/CdTe material system was selected for the demonstration. In this material system, low temperatures (<20 K) are needed to observe the NDR effect. However, it has been demonstrated recently that room temperature NDR can be obtained from InAs/GaAlSb single barrier structures. High temperature (190-300 K) current-voltage curves from the single barrier Hg
1-x
Cd
Te heterostructures have also been investigated, leading to a direct electrical measurement of the controversial HgTe/CdTe valence band offset.
In chapter 4, results are presented from several studies of III-V heterostructures grown by MBE. A measurement of the GaAs/AlAs valence band offset by xray photoemission spectroscopy yields a value of 0.46 ± 0.07 eV, independent of growth sequence. Optical measurements of electron tunneling times in GaAs/AlAs double barrier heterostructures are performed by growing structures with very thin cap layers. Tunneling times as short as ≈ 12 ps are measured. Triple barrier GaAs/AlAs tunnel structures are found to display strong NDR, indicating that the tunneling process is coherent (as opposed to sequential) in nature. Finally, a technique for depositing high quality InAs buffer layers on GaAs substrates is developed.
Item Type:
Thesis (Dissertation (Ph.D.))
Subject Keywords:
Applied Physics
Degree Grantor:
California Institute of Technology
Division:
Engineering and Applied Science
Major Option:
Applied Physics
Thesis Availability:
Public (worldwide access)
Research Advisor(s):
McGill, Thomas C. (advisor)
Bellan, Paul Murray (co-advisor)
Thesis Committee:
McGill, Thomas C. (chair)
Bellan, Paul Murray
McCaldin, James Oeland
Atwater, Harry Albert
Cross, Michael Clifford
Nicolet, Marc-Aurele
Defense Date:
10 May 1989
Funders:
Funding Agency
Grant Number
IBM
UNSPECIFIED
TRW Automotive
UNSPECIFIED
Caltech
UNSPECIFIED
Record Number:
CaltechETD:etd-11212003-115412
Persistent URL:
DOI:
10.7907/96gc-kc14
Related URLs:
URL
URL Type
Description
DOI
Article adapted for Chapter 2.
DOI
Article adapted for Chapter 2.
DOI
Article adapted for Chapter 2.
DOI
Article adapted for Chapter 3.
DOI
Article adapted for Chapter 3.
DOI
Article adapted for Chapter 3.
DOI
Article adapted for Chapter 4.
DOI
Article adapted for Chapter 4.
Default Usage Policy:
No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:
4620
Collection:
CaltechTHESIS
Deposited By:
Imported from ETD-db
Deposited On:
21 Nov 2003
Last Modified:
24 May 2024 18:33
Thesis Files
PDF
- Final Version
See Usage Policy.
6MB
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
GROWTH, CHARACTERIZATION, AND
SIMULATION OF NOVEL
SEMICONDUCTOR TUNNEL STRUCTURES

Thesis by
David H. Chow

In Partial Fulfillment of the Requirements
for the Degree of
Doctor of Philosophy

California Institute of Technology
Pa.sadena., California

1989

(Submitted May 10, 1989)

Acknowledgements
I am grateful to my advisor, Dr. T.C. McGill, for providing me with the benefits
of his experience and insight. He has consistently offered me sound guidance about
scientific and personal matters during my stay at Caltech.

It is a pleasure to acknowledge Dr. J.O. McCaldin, who on many occasions has
shared his knowledge of materials and physical chemistry with me. I have always
found our interactions to be enjoyable.
Many of the current and former members of Dr. McGill's group have contributed to my personal and professional development. I am indebted to Dr. Alice
Bonnefoi for "showing me the ropes" with unyielding patience when I first joined
the group. Dr. Matthew .Johnson taught me a great deal during our collaborative efforts in designing the MBE machine and cleanroom. I also enjoyed many
nonprofessional interactions with Matthew. I would like to acknowledge Dr. Ted
Woodward for patiently showing me how to use most of the equipment in the electrical lab, and for his significant role in our early III-V MBE work. Dr. Richard
Miles has always been a pleasure to interact with professionally and personally.
I would particularly like to thank Richard and his wife, Mayumi, for welcoming
me to dinner at their home on numerous occasions. I have enjoyed many discus-

sions with Mike Jackson during the three years that we have shared an office. I
am particularly grateful to Mike for proofreading Chapter 1 of this thesis, and
for his major role in the tunneling time measurements presented in Chapter 4.
Jan Soderstrom has always been a pleasure to work with, and has taught me a
great deal about 111-V MBE. I would also like to acknowledge Jan's very large
contribution to the InAs on Ga.As work presented in Chapter 4. I have had many
interesting discussions with Ed Yu, who performed all of the x-ray photoemission
measurements for the GaAs/ AlAs band offset project. I have enjoyed working with
Doug Collins, who had a major role in the coherent tunneling project discussed in

iii

Chapter 4. I am grateful to David Ting for patiently explaining theoretical results
to me on many occasions. Brian Cole has been highly enjoyable to work with, and
has provided valuable technical assistance in repairing numerous problems with the
MBE system and cleanroom. I would like to acknowledge Dr. Bob Hauenstein for
numerous interesting discussions, and for performing Hall measurements for us on
several occasions. Pete Zampardi contributed crucial photoluminescence measurements during our early III-V work. I would like to thank Mark Phillips for bailing
me out of several computer problems recently. It is also a pleasure to acknowledge
many useful discussions with Yasantha Rajakarunanayake, Ed Croke, Yixin Liu,
Dr. George Wu, Dr. Wesley Boudvillc, Dr. Steve Hetzler, Dr. A. Zur, Jean-Ma.re

Langlois, Rex Burrington, Todd Rossi, Joel Seely, Dr. T.E. Schlesinger, and Dr.
R. T. Collins.

I would like to thank Marcia Hudson and Carol McCollum for always providing
excellent administrative assistance and friendly conversation during my stay at
Caltech. I would also like to acknowledge the secretarial support of Vere Snell,
whose cheerful manner and insight are greatly missed.
I have had many valuable discussions with Dr. O.J. Marsh of Hughes Research
Laboratories. Peter Ericson, Mike Longerbone, and Sue Swiggart of Perkin Elmer
have provided much appreciated assistance in matters pertaining to MBE growth.

I.K. Sou and J.P. Faurie of the University of illinois and F.A. Shirland and O.K.
Wu of Hughes Research Laboratories provided HgCdTe samples which were crucial
for obtaining the results presented in Chapter 3 of this thesis. C.W. Nieh provided
all of the TEM data contained in this thesis.

I would like to acknowledge the financial support of International Business
Machines corporation, TRW, and the California Institute of Technology.
I would like to thank my parents, who have always vigorously encouraged me
to pursue my education and career. Without their support, I am quite certain that

this thesis would never have been written. Finally, I would like to acknowledge my
wife, Sharon, whose love gives my life purpose and meaning.

Abstract
This thesis presents investigations of novel semiconductor heterostructure devices based on quantum mechanical tunneling. Due to their small characteristic
dimensions, these devices have extremely fast charge transport properties. Thus,
it is expected that tunnel structure devices will be well-suited to high frequency
and optoelectronic applications. The work presented here can be divided into three
sections. In the :first section, a theoretical model for simulating current-voltage
behavior in single barrier heterostructures is developed. The simulations are then
used to design a novel single barrier negative differential resistance (NDR) device.
The second section consists of detailed experimental characterizations of single barrier Hg 1 _..,Cd.., Te heterostructures, including the first demonstration of the novel

single barrier NDR mechanism. Growth of III-V semiconductor heterostructures
by molecular beam epitaxy (MBE) is the subject of the third section. Several aspects of tunneling are explored through characterization of these III-V structures.

In chapter 2, a theoretical model is developed to simulate tunneling currents
in single barrier heterostructures. The model includes band bending effects and
a two band treatment of electron attenuation coefficients in the barrier. It is
proposed that certain material systems have the appropriate band alignments to
realize a novel single harrier negative differential resistance mechanism. A thorough
theoretical analysis of these single barrier NDR structures is presented.
The first experimental demonstration of the single barrier NDR mechanism is
reported in chapter 3. The HgCdTe/CdTe material system was selected for the
demonstration. In this material system, low temperatures ( <20 K) are needed

to observe the NDR effect. However, it has been demonstrated recently that
room temperature NDR can be obtained from InAs/GaAlSb single barrier structures. High temperature (190-300 K) current-voltage curves from the single barrier Hg1-zCdz Te heterostructures have also been investigated, leading to a direct

vi
electrical measurement of the controversial HgTe/CdTe valence band offset.

In chapter 4, results are presented from several studies of III-V heterostructures
grown by MBE. A measurement of the GaAs/ AlAs valence band offset by xray photoemission spectroscopy yields a value of 0.46 ± 0.07 eV, independent of
growth sequence. Optical measurements of electron tunneling times in GaAs/ AlAs

double barrier heterostructures are performed by growing structures with very thin
cap layers. Tunneling times as short as ~ 12 ps are measured. Triple barrier
Ga.As/ AlAs tunnel structures are found to display strong NDR, indicating that
the tunneling process is coherent (as opposed to sequential) in nature. Finally,
a technique for depositing high quality InAs buffer layers on GaAs substrates is
developed.

vii

Parts of this thesis have been or will be published under the following titles:

Chapter 2:
Current Transport Mechanisms in GaAs/ AlAs Tunnel Structures
Grown by Metalorganic Chemical Vapor Deposition,
A.R. Bonnefoi, D.H. Chow, and T.C. McGill, J. Vac. Sci. Teclmol. B . 4,

988 (1986).
Tunneling in MOCVD Grown GaAs-AIAs-GaAs Heterostructures,
A.R. Bonnefoi, D.H. Chow, and T.C. McGill, Bull. Am. Phys. Soc. 31, 395

(1986).
Negative Differential Resistances from Hg1 _eCda: Te-Cd Te Single
Quantum Barrier Heterostructures,

D.R. Chow and T.C. McGill, Appl. Phys. Lett. 48, 1485 (1986).
Negative Differential Resistances from Hg 1 _a:Cda:Te-CdTe Single
Barrier Heterostructures,

D.H. Chow and T.C. McGill, Bull. Am. Phys. Soc. 31, 395 (1986).
Energy Band Diagrams and Current-Voltage Characteristics of
Single Barrier Tunnel Structures,

A.R. Bonnefoi, D.H. Chow, and T.C. McGill, J. Appl. Phys. 62, 3836
(1987).

viii

Chapter 3:
Electrical Determination of the Valence Band Discontinuity in
HgTe-CdTe Heterojunctions,
D.H. Chow, J.O. McCaldin, A.R. Bonnefoi, T.C. McGill, I.K. Sou, and J.P.
Faurie, Appl. Phys. Lett. 51, 2230 (1987).
Observation of Negative Differential Resistance from a Single Barrier Heterostructure,
D.H. Chow, T.C. McGill, I.K. Sou, J.P. Faurie, and C.W. Nieh, Appl. Phys.

Lett. 52, 54 (1988).
Electrical Studies of Single Barrier Hg1 -a,Cda: Te Heterostructures,
D.H. Chow, J.O. McCaldin, A.R. Bonnefoi, T.C. McGill, I.K. Sou, J.P. Faurie, F.A. Shirland, and O.K. Wu, J. Vac. Sci. Tech. A 6, 2614 (1988).

Chapter 4:
Commutativity of the GaAs/ AIAs (100) band offset,
E.T. Yu, D.H. Chow, and T.C. McGill, Phys. Rev. B 38, 12764 (1988).
Electron Tunneling Time Measured by Photoluminescence Excitation Correlation Spectroscopy,
M.K. Jackson, M.B. Johnson, D.H. Chow, T.C. McGill and C.W. Nieh, Appl.

Phys. Lett. 54, 552 (1989).

Evidence for Coherent Tunneling in Resonant Triple Barrier Structures,

D.A. Collins, D.H. Chow, D.Z. Ting, E.T. Yu, J.R. Soderstrom, and T.C.
McGill, submitted to Phys. Rev. Lett.
MBE Growth of InAs and GaSb Epitaxial Layers on GaAs Substrates,

J.R. Soderstrom, D.H. Chow, and T.C. McGill, to appear in Proc. Mat. Res.
Soc., Spring 1989.

Contents

Acknowledgements

Abstract

List of Publications

Vll

List of Figures

List of Tables

xviii

1 Introduction
1.1

1.2

1.3

Introduction to Thesis

1.1.1

Overview . . .

1.1.2

Summary of Results

1.1.3

Outline of Chapter

Motivation . . . . . . . . .

1.2.1

High Speed Devices .

1.2.2

Heterostructures - Novel Electrical Properties

1.2.3

Optoelectronics Applications ..

Theoretical Simulations .

1.3.1

Band Offsets

1.3.2

Band Bending Calculations

. .

1.4

1.5

1.6

1.3.3

Current-Voltage Simulations . .

1.3.4

Tunneling Times

Molecular Beam Epitaxy

1.4.1

General Approach .

1.4.2

III-V growth .

1.4.3

II-VI MBE . .

1.4.4

Group IV growth

Single Barrier Negative Differential Resistance

1.5.1

Overview . . . . . . . . . . . . . . . .

1.5.2

Origin of NDR in Single Barrier Structures .

1.5.3

Summary of Experimental Results .

Outline of Thesis

References . . . . . . . .

2 Single Barrier Electron Tunneling: Theory and NDR Devices

2.1

Introduction and Outline . .

2.2

Theoretical 1-V Simulations

2.2.1

Band Bending . . .

2.2.2

Tunneling Currents

2.3

Single Barrier GaAs/ AlAs Structures

2.4

Hg1_ 111 Cd 111 Te Negative Differential Resistance Devices

2.4.1

Growth Parameters . . . . . . . . . . . .

2.4.2

Effects of Valence Band Offsets on NDR

Other Material Combinations for Single Barrier NDR

2.5

2.5.1

InAs-Ga1 _ 111 Al111 Sb Single Barrier Heterostructures

2.5.2

Pb 1 _ 111 Sn111 Te Single Barrier Heterostructures

References . . . . . . . . . . . . . . . . . . . . . . . . . . .

xii

3 Electrical Studies of Hg 1 _ 111 Cd 111 Te Single Barrier Heterostructures 7 4
Introduction . . . . . . • . . . . . .

3.1.1

Motivation and Background

3.1.2

Summary of Results

3.1.3

Outline of Chapter

3.2

Samples . . . . . . . . . .

3.3

Demonstration of NDR at Low Temperatures

3.1

3.4

3.3.1

Sample TS6 . . . .

3.3.2

Device Fabrication

3.3.3

Low Temperature Electrical Contacts .

3.3.4

1-V Results at Low Temperatures . . .

Electrical Determination of the HgTe/CdTe Valence Band Offset .

3.4.1

Theoretical Simulations of Hole Currents

3.4.2

Device Fabrication . . . .

3.4.3

1-V Results and Analysis .

3.5

Other Samples

. . . . .

. 105

3.6

Summary of Conclusions

107

References . . . . . . . . . . .

. 108

4 Molecular Beam Epitaxy of III-V Heterostructures
4.1

4.2

110

Introduction . . . .

. 110

... .

4.1.1

Background

4.1.2

Summary of Results

112

4.1.3

Outline of Chapter .

114

Standard Ala:Ga1 _a:As Heterostructures

110

115

4.2.1

Substrate Preparation . . . . . .

. 115

4.2.2

Al111 Ga1 _::11As Growth Parameters .

. .. 116

4.2.3

Double Barrier Heterostructures . .

.. 118

xiii

4.3

4.4

4.5

4.6

4. 7

4.2.4

Quantum Well Photoluminescence .

121

4.2.5

Modulation doped GaAs layers

123

4.2.6

Characterization of Thick GaAs Films

124

Determination of the GaAs/ AlAs Valence Band Offset

125

. .

4.3.1

Motivation and Background . . . .

125

4.3.2

X-ray photoemission spectroscopy ..

125

4.3.3

Samples

126

4.3.4

Results .

128

Tunneling Times in GaAs/ AlAs Double Barrier Heterostructures .. 129
4.4.1

Motivation and Background

129

4.4.2

Measurement Technique

130

4.4.3

Samples

. 131

4.4.4

Results .

134

4.4.5

Ongoing Experiments .

135

Electron Coherence in Resonant Tunneling Structmes .

136

4.5.1

Motivation and Background

136

4.5.2

Samples

137

4.5.3

Results .

. 139

Growth of lnAs on Ga.As substrates

141

4.6.1

Motivation and Background

. 141

4.6.2

Growth and In Situ Analysis .

. 142

4.6.3

Characterization

Summary of Conclusions .

144
146

References . . .

149

A MBE System

152

A.1 Introduction and Outline .

152

A.2 Transfer Mechanism

154

...

xiv
A.3 III-V Chamber Modifications

A.3.1 Sb cracker . . . . . . . .
A.3.2

Source Loading Information

155
155
156

A.3.3 Liquid Nitrogen Plumbing

157

A.3.4 Water Cooling . . . . . .

157

A.3.5 Instrument Coating . . .

158

A.4 Prep/ Analysis Module Modifications

158

A.5 Metallization Chamber Modifications

159

B Cleanroom

161

B.1 Purpose

..............

B.2 Design of Cleanroom Facility.

162

...

. 162

Specifications

. 165

B.2.1 General
B.2.2

. 161

B.2.3 Utilities

....

166

List of Figures
1.1

Layer diagram of a. typical IIEMT . . . . . . . . . . . . . .

1.2

Conduction band edge in a double barrier heterostructure.

1.3

Band diagrams for a. Ga.As/ Al:i:Ga1 _:i:As single barrier heterostructure . . . . . . . . . . . . . . . . .

1.4

Schematic diagram of MBE system

1 .5

Band diagram for Hg 1 _..,Cd.., Te NDR structure

1.6

K vs. E in the energy gap of CdTe . . . . . . .

1.7

I-V curve from Hg1 _ 11 Cdm Te single barrier heterostructure

2.1

TTansmission coefficient for a square Al..,Ga 1 _.,As barrier .

2.2

Band diagram for a single barrier GaAs/ AlAs heterostructure

2.3

Measured and Calculated J-V curves for a single barrier Ga.As/ AlAs
heterostructure . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.4

Band diagram for a typical Hg 1 _mCdm Te single barrier structure

2.5

J-V curves for three electrode doping densities . . . . . . . . . .

2.6

Current density vs. barrier thickness for a Hg1 _ 111 Cd111 Te single barrier
heterostructure . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2. 7 Linear and Log current density vs. voltage plots for several choices

2.8

of A.E,,. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Band Diagram for a single barrier InAs/GaAlSb tunnel structure

xvi
2.9

J-V curve for a single barrier InAs/GaAlSb tunnel structure

3.1

TEM photograph of the GaAs/CdTe interface

3.2

Side view of a fabricated device (schematic) .

3.3

Log current vs. log mesa diameter for a typical preparation of sample
TS6 . . . . . . . . . . . . . . . . . . . .

3.4

Reverse bias I-V cmves a.t 4.2 a.nd 15 K

3.5

Forward bias 1-V curve at 4.2 K . . . . .

3.6

High-resolution TEM picture of the single barrier interfaces

3.7

Forward bias 1-V curve from a 67 µm diameter device

3.8

Reverse bias I-V curve from a 67 µm diameter device .

3.9

Experimental and theoretical J-V curves at 300 K from sample TS6 100

3.10 Experimental and theoretical J-V curves a.t 300 K from sample
ML26A' . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
3.11 Loge(J/T 2 ) vs. 1/kT

104

3.12 Log current vs. log mesa diameter for a preparation of sample TS4 . 106
4.1

I-V curves for a typical double barrier structure at 300 K and 77 K 120

4.2

Photoluminescence spectrum of a single GaAs quantum well

122

4.3

Four types of samples grown for XPS experiment . . . . . . .

127

4.4

Schematic diagram of growth sequence for tunneling time samples

132

4.5

High-resolution TEM print of a sample with 34 A barriers .

133

4.6

Schematic diagram of triple barrier heterostructure samples .

138

4.7 1-V curve from a triple barrier heterostructure .

140

4.8

Schematic layer diagram of InAs layer on Ga.As

. 143

4.9

RHEED oscillations for growth of InAs . . . . .

145

A.1 Schematic diagram of MBE system ..

153

xvii

B.1 Drawing of the cleanroom . . . . . . . . . . . . . . . . . . . . . . . 163

xviii

List of Tables
3.1

Table of growth parameters for Hsi-a:Cda: Te single barrier heterostruc-

ture samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Chapter 1
Introduction
1.1

Introduction to Thesis

1.1.1

Overview

This thesis is concerned with the design and realization of novel semiconductor electronic devices based on quantum mechanical tunneling. It is possible to
break the process of creating these devices into three distinct subprocesses: ( i)
theoretical simulations of device behavior, ( ii) growth of the ultrathin layered
heterostructures needed for tunneling, and ( iii) the fabrication and characteriza-

tion of the devices. This thesis includes work which would fall into each of these
three categories. The theoretical models developed here are intended to facilitate
the design of tunnel structures; emphasis is placed on simplicity and qualitative
accuracy. A few semiconductor growth techniques have been proven capable of
producing heterostructures which display reproducible tunneling behavior. In this

thesis, the technique of molecular beam epitaxy (MBE) is employed to produce
almost all of the tunnel structures studied. This choice is made because of the :flexibility and straightforward process offered by MBE. Characterization of the tunnel
structures is directed towards observing novel electronic properties and measuring

heterostructure material parameters which influence tunneling behavior strongly.

1.1.2

Summary of Results

One of the major results of this thesis is the proposal of a novel single barrier
negative differential resistance (NDR) mechanism. A theoretical model is developed to simulate the current-voltage (I-V) behavior of single barrier heterostructures. The model includes band bending effects and a two band model for the
electron imaginary wa.vevector ( attenuation coefficient) in the barrier. The simulations are first applied to GaAs/ AlAs single barrier structures. It is found that
simple elastic tunneling through the AlAs r-point does not adequately explain the
observed experimental I-V curves. Next, the theoretical model is used to analyze
a Hg 1 _ 111 Cd111 Te single barrier heterostructure. It is found that strong NDR behavior due to the novel mechanism can be expected from Hg 1 _a:Cda: Te single barrier

structures with appropriately chosen parameters. Other material systems are also
suggested as candidates for single barrier NDR.
The first demonstration of the novel single barrier NDR mechanism is also
discussed in this thesis. I-V curves from a Hgi-zCd111 Te single barrier structure
are taken at low temperatures ( <20 K) to realize this result. The observation of
NDR suggests that the HgTe/CdTe valence band offset is very small ( <100 meV)

at low temperatures. High temperature I-V curves in these heterostructures are
also investigated and analyzed. It is found that the valence band offset is large
(>300 meV) at 300 K. The high temperature 1-V data (190-300 K) are consistent

with a temperature dependent valence band offset.
This thesis also reports results from a number of studies of III-V heterostructures grown by molecular beam epitaxy. Several initial Ala:Ga.1 _:uAs "calibration"

structures are discussed, including double barrier heterostructures, single quantum
wells, and high electron mobility transistors. An x-ray photoemission spectroscopy

measurement on MBE grown heterojunctions yields a value of 0.46 ± 0.07 eV for
the GaAs/ AlAs valence band offset, independent of growth sequence. Samples
grown for an optical measurement of electron tunneling times in double barrier
heterostructures are discussed. The measurement yields tunneling times as short
as 12 ps for electrons escapi11g from a. Ga.As quantum well sandwiched by two 16 A

AlAs barriers. The tunneling times are found to depend exponentially on barrier
thickness, in good agreement with theory. Triple barrier GaAs/ AlAs heterostructures are found to yield strong resonant tunneling effects, indicating a coherent
nature to the electron tunneling process. Finally, a technique for depositing high
quality thick InAs layers on GaAs substrates is developed.

1. 1.3

Outline of Chapter

The purpose of chapter 1 is to provide some introduction and background
for the thesis, and to give an overview of the following chapters. Section 1.2
attempts to describe some of the motivations for developing heterostructure electronic devices whose behavior is governed by tunneling. Section 1.3 discusses the
successes and limitations of a few theoretical models which are frequently applied
to tunnel structures. Included in the discussion are band bending calculations,
current-voltage simulations, band offset models, and tunneling time calculations.
Section 1.4 describes the growth process of molecular beam epitaxy. Particular

emphasis is placed on III-V semiconductor growth, although brief discussions of
II-VI and group IV techniques are included. In section 1.5, a novel single tunneling
barrier device is introduced. This device displays current-voltage characteristics
which yield a negative differential resistance region due to elastic electron tunneling. The remainder of the thesis is summarized in section 1.6.

1.2

Motivation

1.2.1

High Speed Devices

Over the past two decades, increases in the frequencies at which semiconductor
electronic devices can operate have triggered large scale technological changes in
the world. Computer operations which previously required days can now be performed in fractions of a second. Microprocessors are commonly found in virtually
every electronic system produced, including communications, medical, manufacturing, and consumer electronics equipment. It is a strong possibility that the
next generation of high performance computers will rely upon GaAs devices, which
are faster than their silicon counterparts because electrons in Ga.As move approximately six times faster than in silicon. It seems likely that significantly faster
devices will find applications. The degree to which older devices are replaced by
newer, faster ones will probably be determined by cost/performance comparisons.
In general, the smaller an electronic device can be made, the faster it will
be. Quantum mechanical tunneling effects through semiconducting films become
important when layer thicknesses are reduced to 10-200 A, depending upon the
specific choices of materials. By comparison, it is difficult to obtain field effect
transistors (FETs) with submicron channel lengths. One might therefore expect
tunneling devices to be much faster than the current generation of semiconductor

electronic devices. The prospect of extremely fast electrical devices is one of the
major motivations for developing tunnel structures. The question of exactly what
operating frequency can be attained for tunnel structures is a matter of debate.
Many conflicting theoretical predictions for tunneling times in various structures
have been published over the past twenty years.[1,2,3,4,5,6, 7] Conversely, experimental measurements of these times have only recently been attempted because of
the lack of adequately fast electronics. In fact, all of the measurements which have

been made utilize optical probing to resolve tunneling response times as short as
a few picoseconds.[8,9,10,11] Structures grown by MBE for one of these measurements are discussed in Section 4.4.

1.2.2

Heterostructures - Novel Electrical Properties

In addition to having characteristic dimensions small enough for quantum mechanical tunneling effects, all of the devices studied in this thesis are heterostruc-

tures, consisting of thin epitaxial layers of diffP.rP.nt sP.m1cond11cting materials. The
ability to fabricate heterostructures by techniques such as MBE creates additional
degrees of freedom in designing devices with novel properties. There are many
examples of heterostructure devices which yield device characteristics not obtainable from bulk semiconductors. One application of heterostructures which does
not involve tunneling is the high electron mobility transistor (HEMT). The layer
sequence for a typical HEMT is illustrated in Fig. 1.1. The heterostructure consists of a. heavily doped n-type AlmGa.1_ 111 As layer grown on top of a.n undoped

AlmGa1 _ 111 As spacer layer and a thick film of undoped GaAs. In the structure, the
dopants in then-type Ala,Ga1 _ 111 As layer become ionized, yielding a high concentration of electrons in the GaAs :film.• These electrons carry current in the device by
moving laterally through the GaAs film, with high mobilities at low temperatures
resulting from the lack of ionized impurities in the GaAs. This modulation doping

yields a heterostructure :field effect device which is much faster than conventional
FETs.
Another good example of the additional degree of freedom provided by heterostructures is the double barrier tunnel structure. This heterostructure requires
at lea.st two semiconducting materials with different energy gaps. The basic struc•These electrons actually become localized near the interface with the Al.Ga1-eAs spacer as
a two-dimensional electron gas.

50 A

GaAs, n=1x10 18 cm- 3

600 A

Al Ga As, n:;::1x10 18 cm- 3
.3

200 A

Al Ga As, undoped
.3

1.0 µm

GaAs, undoped

Semi - Insulating GaAs
Substrate

Figure 1.1: Schematic layer diagram of a typical GaAs - Ali:Ga1 _ 111 As high electron
mobility transistor. Free electrons are placed in the undoped GaAs layer by growing a. thin, heavily doped Alii,Ga1 __ eAs layer, which becomes fully depleted. High
mobilities result from the spatial separation of the ionized donor atoms from the
free electrons.

GaAs

AIAs GaAs AIAs

GaAs

Figure 1.2: Schematic diagram for the conduction band edge in a GaAs/ AlAs double barrier heterostructure as a function of distance in the direction perpendicular
to the layers. The dashed line indicates the energy of the confined state in the
quantum well. The shaded rectangles represent free electron gases in each of the
n-type electrodes.

ture begins with a thick, doped layer of the smaller band gap material, forming
an electrode. This is followed by two ultrathin, undoped layers uf the larger band
gap material which sandwich an ultrathin, undoped layer of the smaller band gap
material. These three layers form a quantum well for electrons (and/or holes).
Finally, a thick doped cap layer of the smaller band gap material is grown to form
a second electrode. A schematic diagram for the conduction band edge in a typical

Ga.As/ AlAs double barrier heterostructure is given in Fig. 1.2. In this example,

GaAs has the smaller energy gap, and is used for the quantum well and electrodes.
The AlAs is used for the thin quantum barriers. Because the barriers are thin
(~ 50 A), it is possible for electrons in the n-type GaAs cladding layers to tunnel
across the structure. When a bias is applied between the electrodes, a tunneling
current can be measured. This current is strongly enhanced when the applied

voltage becomes sufficiently large that the tunneling electrons have energies equal
to the confined state energy in the quantum well. When the voltage becomes large
enough that the conduction band edge in the negatively biased electrode is at a
higher energy than the confined state in the quantum well, the enhanced tunneling vanishes, lea.ding to a. pronounced negative differential resistance. It has been

proposed that these resonant tunneling structures could be used in high frequency
oscillators, amplifiers, and mixers.

1.2.3

Optoelectronics Applications

In general, silicon is not a good material for optical and optoelectronic applications because its energy gap is not direct. Therefore, it is often necessary to use
other semiconducting materials, such as GaAs. Many of these materials are well
suited to epitaxial growth techniques like MBE. It is often difficult to make conventional electrical devices from these semiconductors. For example, metal-oxidesemiconductor (MOS) devices cannot be easily fabricated on semiconductors other
than silicon because of the lack of a stable oxide. Some materials, such as InAs,
are not suitable for conventional FETs because their energy gaps are too small
to support the electric field induced by a metal gate. These difficulties increase
the likelihood that tunnel structures could become important electrical devices for
integration with optical and optoelectronic elements. Furthermore, optical devices

are presently being explored as high speed alternatives to electrical devices in certain situations. In many of these applications, it may become necessary to provide
high speed electronic links between the fast optical devices to realize their full
potential. Tunnel structures seem to be a likely alternative for this purpose.

1.3

Theoretical Simulations

Published theoretical treatments of the various aspects of tunneling in semiconductors range from highly sophisticated models to calculations which could be
performed on the back of an envelope, with no direct correlation between usefulness and sophistication. This section attempts to review some of the theoretical
treatments of the relevant issues for the tunnel structures studied in the thesis.

Where possible, comparisons are made to published experimental observations.

1.3.1

Band Offsets

Whenever a semiconductor is grown epitaxially on top of another semiconducting material, a discontinuity exists in the valence (and conduction) band edge at
the interface. This discontinuity is often referred to as the band offset. The value
of the valence band offset* is crucial for many heterostructure devices because it
determines the potential throughout a given structure. In cases of structures with
quantum-sized dimensions, the band offsets play large roles in determining confinement energies and barrier heights. Without reasonably accurate knowledge of the
band offset it is virtually impossible to predict even qualitative device behavior.
Several theoretical treatments of band offsets in semiconductors have been published. One of the earliest models proposed was the "common anion rule", which
states that the valence band offsets in polar semiconductors are determined solely
by the anions, i.e., the column V and VI elements in III-V and II-VI semiconductors, respectively.(12] Some predictions of this rule are that the valence band
offset between InAs and Ga.As should be zero, and that the CdTe/ZnSe valence
band offset should be the same as for ZnTe/CdSe (assuming transitivity of band
*The conduction band offset can always be determined if the valence band offset a.nd the two
energy gaps a.re known.

offsets). The physical reasoning behind this model is that the states near the valence band maximum are derived mainly from the p-like atomic states of the anion.
Therefore, the electronegativity of the anion largely determines the position of the
valence band maximum. It should be pointed out that McCaldin et al.[12) do not
claim that this rule will work for Al or Hg containing compounds.
More recently, it has been proposed that semiconductor-semiconductor interfaces should be treated analogously to metal-metal interfaces, with dipoles forming
at the interfaces to shift the valence band offset away from the common anion value
in certain cases.[13] The argument for the existence of dipoles is based upon the
midgap states which form in semiconductors when they are terminated at an interface. For a particular value of the valence band offset, ordinary bulk states
from one semiconductor are able to tunnel into these midgap states in the other
semiconductor, leading to interface dipoles. It is argued that materials will tend
towards a zero dipole band lineup. Another recent theoretical paper proposes that
since semiconductor energy gaps change with temperature due to the electronphonon interaction, the band offsets between different semiconductors should have
some temperature dependence also.[14] It is proposed that in some cases, this
temperature dependence can be quite large.
To date, it has been difficult to verify the results of most of the band offset theories. This difficulty can be attributed to a lack of experimental data
for most of the semiconductor heterojunctions and conflicts between different
published experimental measurements. Even for the heavily studied Ga.As/ AlAs
band offset, a large range of experimental values of the valence band offset has
been reported.[15,16,17,18]

Furthermore, the two most heavily studied cases,

Ga.As/ AlAs and HgTe/CdTe, do not test the common anion rule as stated by McCaldin et al.[12] The more recent theory of Tersoff[13) is in reasonable agreement
with most published results for GaAs/ AlAs and InAs/GaSb.[19]

However, fur-

ther experimental data is needed to test the predictive value of this theory. In
chapter 3 of the thesis, experimental res.ults from electrical studies of single barrier
HgCdTe heterostructures are given, and used to show evidence for a temperature
dependent band offset in HgTe/CdTe. The theoretical model of Malloy et al.(14)
is in reasonable agreement with these results.

1.3.2

Band Bending Calculations

Whenever a bias is applied between the two outer cladding layers of a semiconductor heterostructure, the conduction and valence band edges in the heterostructure must bend to accomodate the voltage. A calculated diagram of the
conduction band edge vs. distance in the direction perpendicular to the layers of
a Ga.As/ Al 111 Ga1 -mAs single barrier tunnel structure is given in Fig. 1.3(a).

The

method used to calculate the diagram is described briefly in section 2.2, and in
complete detail by Bonnefoi.[20]

It is assumed in Fig. 1.3( a) that a constant quasi-Fermi level in each electrode
can be defined, with the applied voltage equal to the difference between the two
quasi-Fermi levels. The meaning of this assumption is that the chemical potential
in each electrode is constant, i.e., ohmic voltage drops are ignored. The band diagram is then calculated by solving Poisson's equation self-consistently throughout
the heterostructure. At each semiconductor-semiconductor interface, two bound-

ary conditions are satisfied: ( i) a step discontinuity equal to the conduction band
offset between the two materials is placed in the conduction band edge at the interface, and ( ii) the electrostatic displacement is continuous, i.e., the first derivative
of the conduction band edge with position changes by the ratio of the dielectric
constants of the two semiconductors as the interface is crossed. It can be seen from

Fig. 1.3( a) that part of the applied voltage drops across the cladding layers as well
as the barrier region.

500---------.---~--.........- - ~

GaAs-Al _ Ga _ As-GaAs
04
06
(n)
(i)
(n)

>Cl)

Cl)

250

-0
C:
C,

300 mV

0 r------

:;;

:::,

-g -250
()

(a)
-200

>Cl)

250

Cl)

Cl
"O

_!
E' _________

Cl)

lJ
.0

200

300 mV

.2
.....

::J

"O
C:

()

-250

----------(b)

-200

200

.)
Distance (A

Figure 1.3: Conduction band edge of a GaAs - AI:11 Gai-zAs single barrier heterostructure with a 100 A thick barrier layer under an applied bias of 300 m V.
The electrodes haven-type doping densities of 5 x 1017 cm- 3 • (a) is calculated by
the method of Bonnefoi et al.(23], while (b) is calculated by assuming that all of
the applied voltage drops across the barrier. E} (Ej) represents the quasi-Fermi
level in the left (right) electrode.

The treatment described above is classical in that it does not include two dimensional electron gases which result. from quantum mechanical confinement in
regions such as the accumulation layer in the left electrode of Fig. 1.3( a). However, the model does provide a more realistic picture of the conduction band edge
in a heterostructure than the diagram obtained by assuming that the cladding lay-

ers are metal-like conductors. Figure 1.3(b) depicts the diagram obtained when
the second boundary condition given above is replaced by the requirement that the
electrodes maintain zero electric field. In this diagram, all of the applied voltage
appears across the barrier region. It has been shown that predicted current-voltage

(I-V) behavior in tunnel structures is considerably different both qualitatively and
quantitatively when voltage drops in the cladding layers are ignored.[21,22,23,24]
Furthermore, experimental I-V curves from GaAs-AlAs tunnel structures are in
better agreement with the predictions of models which incorporate band bending than with those which assume that the barrier sustains all of the applied
voltage.[23,25]

1.3.3

Current-Voltage Simulations

Although it is hoped that tunnel structures will operate as very high frequency
electrical devices, DC current-voltage (I-V) measurements are often used to help
cha.1:a.cterize the structures. This is done largely beca.use low frequency measure-

ments can be performed with straightforward electronic techniques, and device
response does not change, in general, until very high frequencies are reached. An
example of a common DC characterization is the negative differential resistance

(NDR) region in double barrier heterostructures. These structures are usually
characterized by the peak current densities and peak-to-valley current ratios in
their I-V curves. "State oi tb.e art" results for GaAs-AlAs double barriers are
peak current densities of 104 A/cm2, and peak-to-valley current ratios of 20:1 at

14
77 K.

Theoretical simulations of I-V behavior in tunnel structures range from nearly
closed form equations to large scale computer calculations.[26,27,28,22)

In almost

all cases, it is possible to obtain correct qualitative behavior, but nearly impossible
to generate quantitatively consistent I-V curves. For example, all of the models
correctly predict that the double barrier should show NDR, and some of them can
predict the resonant voltages in close agreement with experiment. However, all of
the models predict peak-to-valley current ratios which are orders of magnitude
larger than those observed experimentally.[22,24] Two possible explanations for
these failures exist. The first is that the transmission coefficient for electrons to
tunnel across a barrier is strongly exponential with the thickness and height of
the barrier, a.nd the effective mass and energy of the tunneling carriers. Small
errors in the choices of these parameters can produce differences of several orders
of magnitude in the tunneling current. A second possibility is that transport
mechanisms other than elastic tunneling may contribute strongly to the current,
especially under conditions such as off-resonant biases in double barriers.
A fairly simple calculation of I-V curves for single barrier tunnel structures is

developed in section 2.2.

1.3.4

Tunneling Times

In section 1.2.1, it was mentioned that the high frequency behavior of tunnel
structures is a topic of interest. A number of different theoretical approaches to
calculating the time required for an electron to tunnel through a potential barrier
have been published.[1,2,3,4,5,6,7] Many of the approaches produce conflicting
results, although most of them agree that the theoretical limit is shorter than 50
picoseconds and longer than 1 femtosecond. One fairly straightforward approach
is to use the time dependent Schrodinger equation, starting with a.n electron wave

packet incident on the potential barrier(s ).[1,5,3] The time required for the packet
to traverse the structure is labeled the tunneling time. Another method explicitly
calculates the RC time constant for a double barrier heterostructure by examining
the theoretical small signal response of the current density to an applied voltage
which modifies the potential profile of the double barrier.[6] A third approach is

to calculate the tunneling time for monoenergetic electrons incident on a timevarying potential barrier.[2,7)

Reconciling the differences between these models

is well beyond the scope of this thesis.
For double barrier heterostructures, it is reported by Harada et al.[3] that the
energy width of the resonance in the electron transmission coefficient, r, can be

related to the tunneling time, T, through the uncertainty principle, i.e.,

(1.1)
This approach is particularly attractive because the electron transmission coefficient can be determined analytically by solving the time independent Schrodinger
equation. The values of T determined by this method were found to agree with
those produced by the time dependent Schrodinger equation approach.[3]

Fur-

thermore, recent photoluminescence experiments appear to yield tunneling times
which are in reasonable agreement with the values predicted by this method.[9,10]
These experiments measure the rate of decay of charge in the quantum well of a
double barrier heterostructure. However, it is claimed by Guo et al.[5] that the
time required for charge to build up in the well can be much longer than that

needed for decay. It is likely that further experimental results will be needed to
establish a valid theoretical model of tunneling times.

1.4

Molecular Beam Epitaxy

Molecular beam epitaxy (MBE) is a process for growing crystalline solids in
which the constituent atoms or molecules are deposited on a heated substrate under
ultrahigh vacuum (UHV) conditions. Beams of the constituent atoms or molecules
are produced by evaporating ultrapure elemental or compound source materials.
High quality epitaxial* films are achievable because of precisely controlled substrate
temperatures and molecular beam fluxes. Typically, MBE can produce films with
monolayer abruptness. Other keys to good crystalline quality include substrate
preparation (both chemical etching outside of vacuum, and oxide removal within
the vacuum system) and surface structure during film deposition.
Two advantages of MBE for semiconductor research are the straightforwardness

and :flexibility of the growth process. Other techniques, such as chemical vapor
deposition, require specific chemical reactions to occur at the substrate surface, in
which many molecules other than those which are constituents of the epitaxial film
are present. These chemical processes are often more difficult to understand and
control than UHV deposition. Furthermore, the mean free paths of molecules in
UHV are longer than the distances from the evaporation sources to the substrate.
This means that complications involving gaseous or liquid flow patterns do not

arise in MBE. The :flexibility of MBE derives from the fact that it is possible, at
least in principle, to evaporate any pure semiconductor source material under UHV
conditions. It has been proven possible to grow most (but not all) of the known
III-V, II-VI, and group IV semiconductors by choosing substrate temperatures and
beam fluxes appropriately. Furthermore, two in situ growth analysis techniques

are available in most MBE systems. Reflection high energy electron diffraction
(RHEED) provides a method for analyzing the surface structure of the epitaxial
• Epitazial means along the 11ame azi11. In the case of epitaxial :films, it is taken to mean that

the film has a crystallographic structure which is related to that of the substrate.

17
film during MBE growth, further enhancing the flexibility of MBE for research
applications. Residual gas analysis {RGA) can provide information about possible
sources of impurities in the UHV environment. A final consideration is the relative
safety of MBE as compared to chemical vapor deposition, which often involves
extremely toxic materials at high volume levels.

This factor is often strongly

weighted in the university research environment.

1.4.1

General Approach

Fig. 1.4 is a schematic diagram of the MBE system used to produce the heterostructures described in chapter 4. The system includes three separate growth
chambers dedicated to III-V, II-VI, and group IV semiconductors, respectively, and
an ESCA/ Auger system, all connected by ultrahigh vacuum transfer tubes. The
internal transfer mechanism is designed to allow the transfer of samples throughout the entire system without removing the samples from the UHV environment.
The system also features a separate chamber for in situ metalization, a substrate

heating and cleaning stage, and three sample introduction loadlocks. Gate valves
are included at each of the points of connection between the systems and transfer
tubes, allowing the chambers to be run independently, if desired. Brief descriptions
of some of the special features of each of the chambers are given in Appendix A.
Whenever a semiconductor surface (or just about any other material surface)
is exposed to atmosphere, radical changes in the surface composition occur due
to the reaction of surface molecules with the molecules which make up the atmosphere, such as oxygen. Even at partial pressures as low as 10- 6 Torr, the surface
molecules undergo collisions with the gaseous molecules once per second. In general, a surface which has been exposed to non-UHV conditions for any period of
time has large concentrations of impurity atoms and poor crystalline properties.
This degraded semiconductor surface is generally not suitable for further epitaxial

"...

:,:
(.)

....w
..J
,(
(J

...w

S::JINOIH::l3H 38~ 1,\-11

(J)

SI MBE ELECTRONICS

□D

,(
(J

- \----------- 5

:Iii

r-----------__,.;;

..J
:;)

:li

....
::>

1,.----~------. g

:z:
;:::
,(

::;

....

1 ~ - - - - ' - - - - - _ _ / t.i
:li

Figure 1.4: Schematic diagram of MBE system. The system has separate III-V,
II-VI, and group IV growth chambers, and an ESCA/ Auger system, connected by
ultrahigh vacuum transfer tubes.

19
growth, and cannot be used to give an accurate picture of the surface conditions
during growth. The intent of the MBE system depicted in Fig. 1.4 is to maintain
maximum flexibility in the choices of materials for heterostructure devices (e.g., II-

VI films on top of III-V films) and to allow each of the growth chambers to access
the surface analysis capabilities of the ESCA/ Auger chamber without removing

the samples from vacuum. It would not be nearly as appropriate to place all of
these growth and analysis capabilities into a single chamber (i.e., no gate valves
between the processes) because even small background levels of impurities from
different classes of semiconductors can produce large changes in the optical and
electrical properties of a material. For example, an Arsenic concentration of 1 part
per million in Silicon produces a background n-type doping level of 5 x 1016 cm-3,
whi.ch is too large for many devices.

1.4.2

III- V growth

All of the heterostructures discussed in chapter 4 were produced in the HIV growth chamber portion of the MBE system depicted in Fig. 1.4. Detailed
descriptions of the growth parameters used for each of these structures are given in
chapter 4. The purpose of this section is to introduce some of the guiding principles
behind III-V growth, and to review the specific materials characteristics of several
of the III-V compounds. Of these compounds, Ga.As and AlAs (and Ala:Ga1 _a:As)
are by far the most extensively grown by MBE, with InAs a distant third.[29]
Sufficient work has been done on GaSb, AlSb, and InSb to prove that MBE is
capable of growing high quality epitaxial films of these materials. Phosphides
have proven difficult to produce by MBE because of severe pumping difficulties
associated with high phosphorous vapor pressures.[30}
An important characteristic of all of the III-V compounds is that they preferentially desorb group V atoms from their surfaces at their growth temperatures.

20
Therefore, it is necessary to provide an overpressure of the group V species at all
ti~es in order to maintain a stable crystalline structure at the surface of a III-V
semiconductor. In the case of GaAs, which is typically grown at 600°C, a steady
As flux is usually maintained on the material throughout the heating, oxide desorption, growth, and cooling processes. Growth is usually initiated by exposing

the substrate to a Ga flux, with the growth rate determined by the arrival rate of
Ga atoms at the surface. The arrival rate of As atoms must always be larger than
that of the Ga atoms in .order to maintain a stable surface. However, the ratio
of the group V flux to the group III flux is usually not chosen to be significantly
higher than what is needed to maintain the surface. For GaAs, an As/Ga ratio of
approximately 6 is considered to be optimal.[29] Stoichiometry is easily preserved
during growth because the excess group V atoms are not incorporated into the
film. The congruent sublimation temperature is an important material constant
for the III-V compounds. Above this temperature, the group V atoms are preferentially desorbed from the surface even under a steady group V flux. Optimal
growth temperatures are usually near to, but not above, the congruent sublimation
temperature for a material.

GaAs/ Ala:Ga1 -a:As Heterostructures

The combination of GaAs and Ala:Ga1 _aiAs has been by far the most popular
choice for MBE grown heterostructures, in which abrupt changes in energy gaps
and refractive indices are dcsircd.[31]

The popularity of Ga.As-Ala,Ga.1 _a:As is

largely due to the nearly perfect lattice match between GaAs and AlAs, which
facilitates the growth of high quality epitaxial layers by removing considerations
of strain. The usual procedure for growing these structures is to start with an
etched GaAs substrate, and to remove its oxide by heating in an arsenic flux
at approximately 600°0. Next, a thick epitaxial GaAs layer is often grown to

21
provide a very smooth buffer layer for the heterostructure. The GaAs growth
rate is usually chosen to be approximately 1 µm/hr (~ 1 monolayer/sec), at a
substrate temperature of 600°0. For reasonably small values of :c, the Al source
oven temperature is varied to produce the desired Ala:Ga1-a:As composition. It
has been shown that higher quality Al,.,Ga1 _,.,As can be obtained by increasing the
substrate temperature to as much as 700°C.[32] However, it is generally agreed
that Ala:Ga 1 _:i:As and AlAs bulk films cannot be grown with as smooth a surface
as GaAs. Furthermore, for :c > 0.8, the Al:i:Ga1 -mAs surface is not stable upon
removal from vacuum due to its reactivity with atmosphere. N-type doping of
these materials is usually accomplished by coevaporating a small amount of silicon
during growth. Although Si would be an acceptor if it were to occupy As sites, it
appears to prefer the group III sites, making it a donor.[33} Beryllium is the most
commonly used p-type dopant for Ga.As and Al:i:Ga1 -:i:As.
In chapter 4 of this thesis, the growth procedures used for several varieties of
Ga.As-Al:r:Ga.1 _:i:As heterostructures are discussed in detail. The following is a brief
summary listing of each type of heterostructure studied, the purpose for the study,
and a brief statement of the results obtained.
1) Double Barrier Heterostructures. These structures were prepared as two-

terminal electrical devices. The 1-V curves were used as a benchmark for
interface smoothness and material quality. The best results for GaAs-AlAs
structures are peak-to-valley current ratios of 2.5:1 and 10:1 at 300 K and
77 K, respectively. Peak current densities of 104 A/cm 2 have been obtained.

2) Single Quantum Wells. Photoluminescence spectra were taken from these
samples, with the linewidth of the peak from the confined state in the quantum. well being used as a measure of the sharpness of the heterostructure
interfaces. Full widths at half maximum of 3.8 meV have been observed for

22
50 A GaAs quantum wells, corresponding to fluctuations in well thickness of
one monolayer or less.

3) High Electron Mobility Transistors. The electron mobility in these modulation doped structures can be used as a measure of the background impurity quality and interface smoothness. Both the inverted and noninverted
geometries (GaAs on Al:i:Ga1 -mAs and vice-versa) have been studied. The
heterostructures yielded electron mobilities which were far less than the best
published results, but which nevertheless displayed enhancement due to the
spatial separation between the electron gas and the ionized donors.

4) GaAs-AlAs Superlattice. One of these structures was grown for a calibration of a Raman scattering experiment. Superlattice phonon modes were
observed.

5) GaAs/ AlAs Heterojunctions. Several of these structures were grown for a
measurement of the GaAs-AlAs valence band offset by x-ray photoemission
spectroscopy. The band offset was found to be commutative, with a value of
0.46 ± 0.07 eV.[15]

6) Double Barriers for Optical Tunneling Rate Experiments. These
structures were designed with very thin {300 A) Ga.As cap layers so that
the double barrier could be probed optically. In particular, the decay of the
photoluminescence from the confined state in the quantum well was used as
a measure of the time required for electrons to tunnel out of the well. This
time is of extreme interest because it may govern the maximum frequencies
at which double barrier devices can be operated. An exponential dependence
of decay times with barrier thickness was observed.[10)
7) Triple Barrier Heterostructures. These structures were grown in order to

23
explore the issue of coherent vs. sequential tunneling. Two terminal I-V
curves were observed to vary stroµgly with middle barrier thickness, yielding
an estimate for the coherence time, i.e., the amount of time that an electron
retains its phase information in the quantum well.[34)

InAs
InAs is a potentially useful material for semiconductor heterostructures because
it has a high electron mobility (10 4 cm 2 /V-s at 300 K) and a small energy gap (350
meV at 300 K). In combination with InAs, larger band gap materials can provide
very high potential barriers for tunneling electrons or holes. This can greatly reduce the thermionic effects which tend to compete with tunneling currents. For
example, the largest peak-to-valley current ratio ever reported for a double barrier heterostructure was observed in a structure which had In0 •53 Gao. 41As for the
electrode and well material, and AlAs for the barrier material.(35] InAs may also
have optical and optoelectronic applications due to its infrared energy gap. Growth

considerations for InAs are remarkably similar to those for GaAs, with the major
difference being the lower substrate temperature (:::::::: 520°0) needed for InAs.
The major stumbling block to widespread use of InAs in combination with
Ga.As and AlAs in heterostructures is the severe lattice mismatch (7%) between
the materials. Extremely large strain energies result when more than one or two
monolayers of InAs are grown with an in plane lattice constant equal to that of
Ga.As. For thicker InAs films, the strain energy is usually relieved by the appearance of misfit dislocations. Unfortunately, these dislocations are generally
detrimental to device behavior.[36] One approach for avoiding the formation of
dislocations in heterostructures containing InAs is to use large band gap materials, such as GaSb, AlSb, ZnTe, or CdSe which lattice match reasonably well with
InAs. None of these materials is nearly as well understood as Ga.As and AlAs for

MBE growth. Furthermore, high quality substrates are scarce and/or expensive
for all of these materials (including InAs ). Therefore, methods for growing buffer
layers on GaAs which terminate with dislocation-free InAs are highly desirable.
In section 4.6, a scheme is discussed, in which superlattice interfaces are used to
getter dislocations in an InAs buffer layer on GaAs. The resulting bulk InAs layers
are characterized by RHEED, Hall measurements, and x-ray diffraction.
Sb containing compounds
The ability to grow antimonides (InSb, GaSb, AISb) in addition to arsenides
by III-V MBE should allow for increased flexibility in tailoring band edges and
lattice constants in heterostructures. Unfortunately, little systematic work has
been published regarding the influence of growth conditions on epitaxial layer
properties for these materials. It is generally agreed that the material quality is
more strongly dependent on the Sb to group III flux ratio in these semiconductors
than for the arsenides.[37,38] This is largely due to the stronger tendency for excess
Sb to become incorporated into the epilayers at the growth temperatures used. It
is interesting to note that while the antimonide growth temperatures are somewhat
lower than those for the arsenides, Sb evaporation is usually done near 600°C, as
compared to near 300°C for As. Several applications have been proposed and/or
realized for heterostructures containing Sb compounds, including high electron
mobility devices, single and double barrier negative differential resistance devices,
and infra.red detectors.(39,40,41,42]

1.4.3

II-VI MBE

Most of the known II-VI semiconductors ha.ve large direct energy gaps. Hence,
many of the proposed applications for these materials are optical, with emphasis
on the visible portion of the spectrum. Very little work has been reported on

the MBE growth of II-VI semiconductors, with most of the available literature
coming from Japan. Although little of this thesis is devoted to the growth of
II-VI materials, a few brief comments are made here for contrast to the growth
considerations previously discussed for III-V semiconductors.

It is well known that most of the II-VI materials sublime congruently up to
fairly high vapor pressures. In fact, it is common practice for MBE growth of
these semiconductors to be accomplished by evaporating bulk material of the desired compound, e.g., evaporating Zn Te at 600°C in a Knudsen cell and growing it
on a ZnTe substrate at 300°C. Elemental sources can also be used, with stoichiometry usually preserved in excess overpressures of either constituent.(43] Substrate
temperatures for these materials are usually considerably lower than for III-V
semiconductors. In general, an MBE system which is intended to be used for the
growth of II-VIs can have the same design as one for III-Vs. Notable exceptions
to this rule are systems designed to grow Hg-containing compounds. Because of
the extremely high vapor pressure of mercury, special design considerations must

be ma.de to handle and remove it from the system routinely. In spite of these difficulties, a considerable amount of effort has been directed towards the MBE growth
of Hg1 -:a:Cd:e Te. This interest stems mostly from the unique infrared tunability of
the Hg 1 _:a:Cd:a:Te energy gap.[43]

1.4.4

Group IV growth

Although silicon is by far the most extensively studied and best understood
of all of the semiconductors, MBE growth of Si is a relatively new technology.
The lag in the development of Si MBE is largely due to the high temperatures
required to evaporate significant quantities of silicon. Standard Knudsen cells are
not capable of achieving these temperatures without significant chemical reactions
occuring between the crucible material and the silicon charge. This problem has

26
been solved by bombarding the silicon charge with a beam of high energy electrons
which evaporate it locally within the charge. Unfortunately, this technique requires
the additional complication of installing electron guns inside the UHV chamber.
The potential applications for Si MBE stem from the low growth temperatures
which are possible, and the prospect of heterostruct11Tes which combine Si and
Ge.t[44,45]

1.5

Single Barrier Negative Differential Resistance

1.5.1

Overview

Perhaps the most significant result reported in this thesis is the proposal and
realization of a single barrier tunnel structure which displays a novel negative differential resistance (NDR). The active portion of this device consists of a thin
epitaxial layer of one semiconductor which forms a quantum barrier between two
thick cladding layers of another semiconductor. Electrons tunnel through the quantum barrier under an applied bias, producing a current. The 1-V curves from
this class of structures can have NDR regions when the tunneling electrons have
energies close to the valence band edge of the barrier material. Although the device is demonstrated here for a Hg1 -a:Cdai Te heterostructure, a few other material
combinations are candidates for the single barrier NDR phenomenon. In fact,
Hg1 _ 111 Cd 111 Te is probably not the best choice for practical applications because its
thermionic currents are high at room temperature, and it is a fragile material,
ma.king it difficult to achieve reproducible devices. At the time of this writing, the
tsmcon and germanium are the only two conventional group IV semiconductors. Diamond
has a prohibitively large energy gap, making it an insulator, and tin displays metallic behavior.

27
single barrier NDR concept has been demonstrated in a second heterostructure
combination: InAs-Ga1-zAla:Sb.[40]
The motivation for developing single barrier NDR structures is to produce electrical devices which can operate at extremely high frequencies. In these applications, the NDR feature is useful because it can be exploited in designing oscillators,

amplifiers, and mixers.[46] Furthermore, single barrier heterostructures may have
better high-frequency response and are easier to fabricate than double barrier devices, which produce NDR due to resonant electron tunneling. The possibility of
enhanced speed is due to the absence of a quantum well, which must be charged
and discharged as electrons pass through it. To date, no experimental measure-

ments of the time required for an electron to tunnel through a single barrier have
been made.

1.5.2

Origin of NDR in Single Barrier Structures

This section attempts to give a basic physical argument for expecting NDR
from single barrier heterostructures under certain conditions. A more detailed
mathematical simulation of single barrier tunneling is provided in chapter 2. The
specific case of ND R from single barrier structures is treated in section 2.4.
Figure 1.5 is a calculated energy band diagram for the single barrier heterostructure studied experimentally in chapter 3 under an applied bias of 50 m V. The

device consists of a thin CdTe layer sandwiched between two Hg0 _78 Cd0 _22 Te electrodes, doped n-type. These materials were selected for the single barrier heterostructure because their band alignments satisfy the requirement for observing
NDR. Assuming a small valence band offset between HgTe and CdTe,* the tunneling electrons which originate in the Hg0 , 78 Cd0 _22 Te electrodes lie much closer in
energy to the valence band edge in CdTe than to the conduction band edge. This
•The actual value of the valence band offset is still a matter of debate

HgCdTe - CdTe - HgCdTe
(n)

,.-....

>Q.) 1500

(i)

(n)

...._,,,,
(/)

Q)
CJ)

-0
Q)

-0

1000

I--

:;:;
(.)

:::J

-0

500 ...

-0
Q.)
(.)

(])

0 --------------

- - ----------f -

-200

200

400

Distance (A)
Figure 1.5: Valence and conduction band edges in the Hg 1 _aiCdai Te single barrier
heterostructure under an applied bia.s of 50 m V. A 170 A CdTe barrier layer is

sandwiched between two n-type Hg0 ,78 Cd 0•22 Te electrodes with carrier densities of
4 X 1016 cm- 3 •

29
situation is uncommon among the well-studied combinations of semiconductors.
In addition to these band alignments, there are two conditions which must be satisfied in order to observe NDR from a single barrier heterostructure. The first of
these is that the total current must be dominated by elastic electron tunneling.
The second condition is that a reasonably large fraction of the total applied voltage must be dropped across the barrier, instead of across the cladding layers. For
example, the band diagram in Fig. 1.5 depicts a situation in which roughly 40%
of the total bias appears across the CdTe layer. The remainder of the voltage is
lost in creating depletion and accumulation regions in the electrodes. Simulations
indicate that the quantum barrier must drop at least 25% (roughly) of the total
voltage in order to observe NDR.

In the WKB approximation, an electron has a transmission probability of tunneling through a single quantum barrier, T, given by:

T oc exp

[-2 J

K dz] ,

(1.2)

where K is the imaginary part of the electron wavevector in the forbidden region,
and z is the distance in the direction perpendicular to the layers, with z = 0 and
z = w defined to be the positions of the interfaces between the barrier and the
electrodes. In this situation, K behaves as the attenuation constant for an electron
tunneling through the single barrier. The numerical value of K is determined by the
specific barrier material and the energy of the tunneling electron. In a direct band
gap semiconductor, such as CdTe, the electron wavevector is purely imaginary for
energies in the forbidden gap, and purely real for energies in the conduction and
valence bands. Since the electron wavevector must be continuous with energy, both
the real and imaginary parts must go to zero at the conductio11 and valence band
edges. Figure 1.6 contains the results of a two band model, k • p theory calculation
which gives K as a function of energy, E, in the energy gap of CdTe.[47)

In this

plot, K is seen to go to zero at the conduction and valence band edges, with a

____________________ E CdT e

,,,--...,

Q)
.......__,

0 ____________________ E CdTe

Figure 1.6: Calculation of the imaginary part of the electron wavevector, K, in

the energy gap of Cd Te. E 11 (Ee) denotes the valence ( conduction) band edge in
CdTe. Two band k · p theory was used to calculate the curve.[47]

31
maximum near the middle of the energy gap. Hence, the transmission probability
for electrons incident upon a CdTe barrier is largest for electrons with energies near
the band edges, and smallest for electrons with energies near midgap. Since the
tunneling electrons in the heterostructure depicted in Fig. 1.5 have energies which
are closer to the valence band edge energy than to the conduction band edge in the

CdTe barrier, the low energy part of the E vs. K curve in Fig. 1.6 is the portion of
interest for this device. As the voltage applied to the heterostructure is increased,
the tunneling electrons move to higher energies with respect to the CdTe barrier.
Hence, they are confronted with increasing values of K, leading to decreasing
transmission probabilities in Eqn. 1.2. As the transmission probability decreases,

the tunneling current also decreases, yielding negative differential resistances. The
NDR is a direct result of the uncommon band alignment in this heterostructure.

1.5.3

Summary of Experimental Results

To date, single barrier NDR has been observed in semiconductor heterostruc•
tures fabricated from two distict material combinations. The first demonstration of
single barrier NDR was made for the Hg1 _ 111 Cd111 Te heterostructure described above.
Details of growth, device preparation, and measurements from this structure are
explored in chapter 3. An I-V curve, taken at 4.2 K from the Hg 1 _ 111 Cda,Te single
barrier heterostructure, is given in Fig. 1.7.

The curve displays a peak-to-valley

current ratio of 2:1, with a peak current density of 0.51 mA/cm2 • Both of these
figures of merit are considerably lower than typical results from double barrier
tunnel structures in GaAs-AlAs. Furthermore, it is likely that room temperature
operation will be highly desirable in most of the practical applications of tunnel
structures. The second realization of single barrier NDR was obtained from an

InAs-Ga1 _ 111 Al 111 Sb structure.[40] Room temperature NDR was observed in this
structure, with enhanced behavior at 100 K. Although the current density from

I-V Curve

T = 4.2 K
Reverse Bias

,,.--....,
<(
'-"'

.........

(I)
:)

0 1.-.------===;;;..i._-....L-_..____--.J._ __.__
40
80
120
160

_J_____J

Voltage (mV)
Figure 1. 7: Experimental I-V curve, obtained at 4.2 K from a Hg1 _ 111 Cd111 Te single
barrier heterostructure.

33
this structure was reasonbly good(~ 300A/cm2 ), the peak-to-valley current ratio
was a very low 1.1:1. Further development of growth and processing techniques
for these structures could improve behavior significantly.

1.6

Outline of Thesis

In chapter 2, a. theoretical model is developed for simulating the electrical
behavior of single barrier tunnel structures. Included in the model are calculations
of band bending, imaginary wavevectors, transmission coefficients, and I-V curves.
Simulations of single barrier GaAs-AlAs structures are presented and compared
to experimental results. The model is then used to quantify the predictions of
NDR from single barrier heterostructures fabricated from certain combinations of

semiconductors.
Chapter 3 contains an experimental study of the current-voltage behavior of
Hg 1 _ 111 Cd111 Te single barrier heterostructures. Growth and processing considerations are discussed, in addition to measurement techniques. Low temperature I-V
curves are shown to display NDR dne to the novel single barrier mechanism discussed previously. At higher temperatures, the thermionic hole currents are used
to determine the valence band offset between HgTe and CdTe. The measured
currents are found to be consistent with a temperature dependent band offset.
Several experimental studies which involved samples produced by the III-V
MBE chamber are explored in chapter 4. Emphasis is placed upon the specific
growth requirements and designs for each study. Several standard heterostructures are employed as diagnostic tools for the quality of materials and interfaces
produced by the MBE chamber. Four major projects are then undertaken: ( i) A
measurement of the GaAs/ AlAs valence band offset by x-ray photoemission spectroscopy, ( ii) An optical measurement of the tunneling escape rate of electrons in

the well of a double barrier heterostructure, ( iii) A test of coherent vs. sequential
tunneling in a triple barrier heterostructure, and ( iv) Development of a method
(involving superlattice buffer layers), for reducing dislocations in bulk InAs films
grown on GaAs.
Appendix A contains information regarding special features in the designs of

several of the chambers of the MBE system depicted in Fig. 1.4. Appendix B
describes the cleanroom in which the MBE system and most of the processing
facilities used in this thesis are housed. Special design considerations are discussed,
with emphasis upon adaptations made to tailor the cleanroom to the university
research environment.

References
[lJ K.K. Thornber, T .C. McGill, and C.A. Mead, J. Appl. Phys. 38, 2384 {1967).

[2] M. Biittiker and R. Landauer, Phys. Rev. Lett. 49, 1739 (1982).

[3] N. Harada and S. Kuroda, Jpn. J. Appl. Phys. Pt.2 25, 1871 (1986).
[4] S. Collins, D. Lowe, and J.R. Barker, J. Phys. C 20, 6213 (1987).
[5] H. Guo, K. Diff, G. Neofotistos, and J.D. Gunton, Appl. Phys. Lett. 53, 131

(1988).

[6] S. Luryi, Appl. Phys. Lett. 47, 490 (1985).
[7] S.C. Kan and A. Yariv, J. Appl. Phys. 64, 3098 (1988).

[8] J.F. Whitaker, G.A. Mourou, T.C.L.G. Sollner, and W.D. Goodhue, Appl.
Phys. Lett. 53, 385 {1988).

[9] M. Tsuchiya, T. Matsusue, and H. Sakaki, Phys. Rev. Lett. 59, 2356 (1987).
[10] M.K. Jackson, M.B. Johnson, D.H. Chow, T.C. McGill, and C.W. Nieh, Appl.

Phys. Lett. 54, 552 (1989).

[11] T.B. Norris, X.J. Song, W.J. Schaff, L.F. Eastman, G. Wicks, and G.A.
Mourou, Appl. Phys. Lett. 54, 60 (1989).
(12] J.O. McCaldin, T.C. McGill, and C.A. Mead, Phys. Rev. Lett. 36, 56 (1976).

[13] J. Tersoff, Phys. Rev. B 80, 4874 (1984).
(14] K.J. Malloy and J.A. Van Vechte:ri, Appl. Phys. Lett. 54, 937 (1989).
[15] E.T. Yu, D.H. Chow, and T.C. McGill, Phys. Rev. B 38, 12764 (1988).
[16] J. Batey and S.L. Wright, J. Appl. Phys. 59, 200 (1986).
[17] R. Dingle, W. Wiegmann, and C.H. Henry, Phys. Rev. Lett 33, 827 (1974).
[18] R.C. Miller, D.A. Kleinman, and A.C. Gossard, Phys. Rev. B 29, 7085 (1984).
(19] G.J. Gualtieri, G.P. Schwartz, R.G. Nuzzo, R.J. Malik, and J.F. Walker, J.

Appl. Phys. 61, 5337 (1987).
[20] A. R. Bonnefoi, Ph.D. Thesis, California Institute of Technology, 1986.
[21] H. Ohnishi, T. Ina.ta., S. Muto, N. Yokoyama., and A. Shiba.tomi, Appl. Phys.

Lett. 49, 690 (1986).
[22] M. Cahay, M. McLennan, S. Datta, and M.S. Lundstrom, Appl. Phys.·Lett.
50, 612 (1987).

(23] A.R. Bonnefoi, D.H. Chow, and T.C. McGill, J. Appl. Phys. 62, 3836 (1987).
(24] K.F. Brennan, J. Appl. Phys. 62, 2392 (1987).
[25] A.R. Bonnefoi, T.C. McGill, R.D. Burnham, and G.B. Anderson, Appl. Phys.

Lett. 50, 344 (1987).
[26] R. Tsu and L. Esaki, Appl. Phys. Lett. 22, 562 (1973).

[27] M.O. Vassell, J. Lee, and H.F. Lockwood, J. Appl. Phys. 54, 5206 (1983).
[28] D.D. Coon and H.C. Liu, Appl. Phys. Lett. 49, 94 (1986).

37
[29] J.D. Grange, in The Technology and Physics of Molecular Beam Epitazy,
edited by E.H.C. Parker (Plenum, New York, 1985), Chapter 3.
(30) C.R. Stanley, R.F.C. Farrow, and P.W. Sullivan, in The Technology and

Physics of Molecular Beam Epitazy, edited by E.H.C. Parker (Plenum, New
York, 1985), Chapter 9.
[31] A.Y. Cho, in The Technology and Physics of Molecular Beam Epita;r;y, edited
by E.H.C. Parker (Plenum, New York, 1985), Chapter 1.

[32] H. Morko~, in The Technology and Physics of Molecular Beam Epitazy, edited
by E.H.C. Parker (Plenum, New York, 1985), Chapter 7.
[33] C.E.C. Wood, in The Technology and Physics of Molecular Beam Epitazy,
edited by E.H.C. Parker (Plenum, New York, 1985), Chapter 4.
[34] D.A. Collins, D.H. Chow, D.Z. Ting, E.T. Yu, J.R. Soderstrom, and T.C.
McGill, unpublished.
[35] T. Inata, S. Muto, Y. Nakata, S. Sasa, T. Fujii, and S. Hiyamizu, Jpn. J.

Appl. Phys. 26, 11332 (1987).
[36] R.H. Miles, Ph.D. Thesis, California Institute of Technology, 1988.
[37] M. Lee, D.J. Nicholas, K.E. Singer, and B. Hamilton, J. Appl. Phys. 59, 2895
(1986).
[38] Chin-An Chang, H. Takaoka, L.L. Chang, and L. Esaki, Appl. Phys. Lett. 40,
983 (1982).

(39] K. Ploog, in The Technology and Physics of Molecular Beam Epitazy, edited
by E.H.C. Parker (Plenum, New York, 1985), Chapter 18.

[40] H. Munekata, T.P. Smith, and L.L. Chang, J. Vac. Sci. Technol. B, to be
published.
(41) L.F. Luo, R. Beresford, and W.I. Wang, Appl. Phys. Lett. 58, 2320 (1988).

[42] D.L. Smith and C. Mailhiot, J. Appl. Phys. 62, 2545 (1987).

(43] T. Yao, in The Technology and Physics of Molecular Beam Epitazy, edited by
E.H.C. Parker (Plenum, New York, 1985), Chapter 10.

[44) R.J. Hauenstein, Ph.D. Thesis, California Institute of Technology, 1986.
[45] Y. Shiraki, in The Technology and Physics of Molecular Beam Epitazy, edited
by E.H.C. Parker (Plenum, New York, 1985), Chapter 11.

[46] S.Y. Liao, Microwave Devices and Circuits (Prentice-Hall, Englewood Cliffs,
NJ, 1985), pp. 294-376.

[47] E.O. Kane, in Physics of III-V Compunds (Academic, New York, 1966), Vol.1,
Chap.3, pp. 75-100.

Chapter 2
Single Barrier Electron
Tunneling: Theory and NDR
Devices
2.1

Introduction and Outline

In this thesis, single barrier heterostructures are defined to be epitaxial layered
structures in which a single, thin layer of one semiconducting material forms a
potential barrier to electrons or holes traveling between two bulk layers of another
semiconductor. Although structures containing multiple quantum barriers and/or
wells offer increased flexibility and variety in device behavior, ample motivation
exists for studying single barrier heterostructures. For most of the possible combinations of semiconducting materials, fundamental heterojunction parameters, such
as the valence band offset, are still unknown. These parameters are often critical
in designing heterostructures for a particular purpose. A single barrier structure
generally offers the most straightforward measurement of these parameters because
its behavior is the most directly determined by them. Furthermore, single barrier

tunnel structures may be more suitable than multiple barrier devices for high speed
applications, because quantum well charging and discharging times are eliminated.
This chapter develops a theoretical model for predicting the tunneling current

through a single barrier heterostructure under an applied bias. Due to the relative
simplicity of the potential in a single barrier structure ( as compared to multiple
barrier structures), it is possible for straightforward simulations of current-voltage
(I-V) behavior to yield reasonable qualitative and quantitative accuracy. The
theoretical model developed here is divided into two sections. In the first section,
the band diagram for the single barrier structure is calculated as a function of
applied bias. The band diagram is then used in the second section to calculate
the tunneling current. Considerations of electron transmission probabilities and
complex handstructure are included in the second section.
In section 2.3 the 1-V simulation is applied to single barrier Ga.As/ AlAs heterostructures. At high voltages, experimental tunneling currents are found to be
larger than those predicted for elastic tunneling across the AlAs r-point. A com-

bination of tunneling through the AlAs X-point with r-point tunneling is found
to be consistent with the experimental 1-V data.
As discussed in section 1.5, the proposal and realization of a single barrier negative differential resistance (NDR) device is one of the major results of this thesis.
The Hg1-mCdm Te version of this device is analyzed theoretically in section 2.4.
Theoretical predictions of peak-to-valley current ratios and peak current densities
are given. Two other material combinations which are candidates for single barrier
NDR, InAs/GaAlSb and PbSnTe, are also analyzed in section 2.5.

2.2

Theoretical I-V Simulations

2.2.1

Band Bending

As discussed previously in section 1.3.2, the conduction and valence band edges
in a semiconductor heterostructure must bend to accomodate an applied bias. This
section gives a brief summary of the method used in this thesis to calculate band
diagrams for single barrier heterostructures. A more detailed explanation is given
by Bonnefoi.[l]

It is assumed here that a constant quasi-Fermi level in each

electrode can be defined, with the applied voltage equal to the difference between
the two quasi-Fermi levels. Ohmic voltage drops, i.e., changes in the conduction
band edge due to electrical current passing through the electrode material, are
ignored in this assumption. For most of the structures studied here, ohmic voltage
drops are negligible because the total device resistance is much greater than the
ohmic resistance of the cladding layers.
The ha.sic formula. for calculating band diagrams of heterostructures is to solve

Poisson's equation in one dimension for each of the layers:

d2 Ec(z) = ep(z)
dz 2
€of,.

(2.1)

where Ec(z) is the conduction band edge as a function of z, the distance in the
direction perpendicular to the layers, e is the electron charge, p(z) is the charge

density as a function of z, 1:0 is the permitivity of free space, and 1:,. is the dielectric
constant of the material. In this equation, Ee( z) is used instead of the more familiar
electrostatic potential, (z), for reasons of convenience. The two can be related
by:

E,,(z) = -e¢,(z).

(2.2)

The charge density, p( z ), can be determined by summing the densities of ionized

dopants and free charge carriers;

p(z) = -e (n(z) - p(z) + NA(z) - Nv(z)),

(2.3)

where n(z) and p(z) are the free electron and hole densities, respectively, NA(z)
is the density of ionized acceptors and ND(z) is the density of ionized donors.
For the structures of interest here, the densities of acceptor and donor atoms are
determined from growth parameters. It is usually reasonable to assume that the
donors and acceptors are fully ionized. The free carrier densities are related to the
relative positions of the conduction and valence band edges with respect to the
Fermi level. In the model used here, the T

O K expressions for n( z) and p( z)

are used as an approximation to the actual carrier densities;

{2.4)
and

p(z)

E1 > Eu

3 '11"2

[2ml
v(E,,(z) - E1) ]a/2 E1 ~ Eu '

(2.5)

where E1 is the Fermi level, m; (mh.) is the electron (hole) effective mass, and Eu(z)
is the valence band edge. Upon substituting the carrier densities in Eqns. 2.4 and

2.5 into Eqn. 2.1, a second order, nonlinear differential equation is obtained for

Ec(z). This equation can be solved numerically.
Once the solutions to Eqn. 2.1 have been obtained for each of the layers of a
heterostructure, they can be joined by using two boundary conditions. The first is
that the conduction band edge at an interface (z = z0 ) between two semiconductors, must take a discontinuous step equal to the conduction band offset between
those two semiconductors:

(2.6)

43
where Ef is the conduction band edge in material A, E! is the conduction band
edge in material B, and AEfB is the conduction band offset between materials
A and B. The second boundary condition is that the electric displacement, D(z),
must be continuous across an interface, i.e.,

dEf(zo) DA( )
dz
= zo

DB( zo ) = EB dE:1(zo)
dz .

(2.7)

In addition to employing the T = 0 K expressions for the free carrier densities,
this model neglects two dimensional subbands 1 which should form in the potential
wells created by accumulation layers. Both of these assumptions simplify the
mathematics greatly, but can produce erroneous results under certain conditions.

In particular, the T = 0 K approximation is most valid for degenerate doping
concentrations and low temperatures. Two dimensional subbands become more
prominent at large applied biases, i.e., where the conduction band edge is the
most steeply sloped. Although these approximations restrict the validity of the
model, it is clear that the band diagrams generated are more physically correct
than an assumption of no voltage drop in the electrodes.

2.2.2

Tunneling Currents

Once the hand diagram for a single barrier heterostructure has been determined, it is possible to calculate the tunneling current through the structure.

Several methods for performing this calculation have been published.(2,3,4,5,6,7l
In this thesis, the method of Tsu et al.[3] is chosen because of its straightforward
formalism. This method utilizes the time independent Schrodinger equation to
compute transmission coefficients and current densities through the barrier. Modifications to this approach have been incorporated to include band bending results
and non-square barriers.
The tunneling current calculation is based upon an assumption of nearly free

44
electron gases in the left and right electrodes. The electron wavefunctions, ii!, are
written:
,Tr -

exp(ik. r)

'l' -

-v'V

{2.8)

where r is the distance vector, k is the electron wavevector, and V is the crystal
volume (used for normalization). The electrons follow the dispersion relation

(2.9)

where Ee is the conduction band edge, E is the total energy, and m; is the electron
effective mass.
Due to the layered nature of the heterostructure, the potential can be treated
as varying in only one dimension, i.e.,

(2.10)
where z is the distance in the direction perpendicular to the layers of the heterostructure. The three dimensional Schrodinger equation can then be separated
into perpendicular and parallel parts (relative to the layers of the heterostructure):

(2.11)
Two mathematical simplifications result from this separation. The first is that the
total energy is the sum of perpendicular and parallel energies;

{2.12)
where E-t is defined to be the perpendicular energy. The second result is that W

is the product of a function which depends only on the parallel distances y and~,
with a function which depends only on z;

{2.13)

It should be noted that, since Ec(z) is independent of :v and y, it is assumed that k~
and ky are constant for an electron tunneling elastically across the heterostructure.
The current density in the z direction for an occupied state is given by :

Jz = ~('11* 8'11 _ '118'11* ).
2m;i
8z
8z

(2.14)

For an occupied state tunneling from the left electrode to an available state in the
right electrode, we assume that

1/J1.. Jt(z) =

)v [exp(ik.,z) + r exp(-ik.,z)],

(2.15)

and
"Pf'ight(z)=

(texp(ikzz)].

(2.16)

Application of Eqn. 2.14 to Eqn. 2.15 yields
JleJt = e1ikz {1 - r2)
m*V

(2.17)

and
(2.18)
The requirement that the current density be constant throughout the structure in
steady state then gives:
1- r 2 = t 2

(2.19)

as expected. The quantity t 2 (r 2 ) is called the transmission (reflection) coefficient,
and is assumed to be a function of the perpendicular energy of the tunneling
electrons only. Furthermore, it is assumed that t 2 is the same for particles moving

left to right as for particles moving right to left.
To calculate the elastic tunneling current from the left electrode to the right
electrode, it is necessary to multiply Eqn. 2.18 by the probability that the states
in the left are occupied and that the states in the right are empty;
e1ikz 2 !1 ( E )[ 1
J,_" = -Vt
m*

(2.20)

where / 1( E) (fr( E)) is the Fermi distribution function in the left (right) electrode.
Letting E11 and Etr represent the quasi-Fermi levels in the left and right electrodes,
respectively, yields:
J,(E) = 1 + exp[(E E11)/kT]'

fr(E) = 1 + exp[(E - E1r)/kT]

(2.21)
(2.22)

Similarly, the tunneling current from the righi dectrode to the left electrode is
given by:
(2.23)
The net tunneling current at a particular energy, Jnei(E), is then given by:
(2.24)
Equation 2.24 can be multiplied by the density of states in k-space, and integrated over all available states to obtain an expression for the total tunneling
current;
(2.25)
Utilizing the identity for the electrodes,
(2.26)

Eqn. 2.25 becomes:

Finally, performing the integration over all k2 and ky, and rewriting the integral
over kz in terms of the perpendicular energy yields:
(2.28)
The lower integration limit, Ec(zo), is the conduction band edge at the interface
between the barrier and the negatively biased electrode.

47
Equation 2.28 is easily integrated numerically once the transmission coefficient
and band diagram are known. The formula is sufficiently general that it can be
adapted to include a transmission coefficient for an arbitrary barrier potential.
Calculated band diagrams are used to determine the conduction band edge in the
barrier and the parameters E1i, E,.,., and Ee(zo), It should be noted that the
inclusion of band bending effects is slightly inconsistent since non-flat electrode
bands do not yield plane wave solutions. However, this inconsistency is ignored
here for the sake of simplicity.

Transmission Coefficients
In general, it is possible to determine the transmission coefficient for tunneling through a single barrier heterostructure by solving the Schrodinger equation
(Eqn. 2.11) in each of the layers of the structure. At each interface (z = a), the
solution for layer i, 1Pi(z ), can then be matched to the solution for layer i + 1,

'1/Js+i(z), by the boundary conditions:
(2.29)

__!_ d'f/,;,+1)

__!_ d-r/Ji)
m*e dz

m*e

(2.30 )

The effective mass enters Eqn. 2.30 because of the condition of constant probability
current across the interface.
For a single "square" potential barrier between two identical flat electrodes, t 2
can be written in closed form;
(2.31)

mn is the electron mass in the electrodes (bar-

where w is the barrier width, m~ (

rier), ki (k.,.) is kz in the left (right) electrode, and Kb is the imaginary part of
the electron wavevector in the barrier. In this formula, the exponential factor,

exp{-2Kbw), tends to dominate the prefactor for typical problems of interest.[8]
For an arbitrary potential barrier, the Wentzel-Kramers-Brillouin (WKB) approximation is used in place of the closed form exponential factor in Eqn. 2.31:

(2.32)
where the integration limits, z = 0 and z = w, are the left and right interfaces
between the barrier and the electrodes.

Imaginary Part of the Electron Wavevector
As discussed in section 1.5.2, the imaginary part of the electron wavevector, K,
is determined by the energy of the tunneling state, and the properties of the barrier
material. It was argued previously that K should go to zero at the conduction
and valence band edges of a direct energy gap semiconductor, with a maximum
near midgap. This section gives an abbreviated derivation of a two band model
k · p theory formula for K as a function of energy, E, in the direct bandgap of

an arbitrary semiconductor. A more general derivation is given by Kane.[9] This
formula differs substantially from the commonly used "one band" formula:

K-_ (2m:( Ee: - E))

112

(2.33)

Fig. 2.1 depicts the transmission coefficient for tunneling a.cross a square

Alo.3Gao.7As barrier as calculated with the one and two band formulas. The error
made by the one band model becomes increasingly more pronounced as the energy
is lowered towards the valence band edge.
The k · p method gives the bandstructure of a semiconductor by constructing
bands from well-known parameters ( energy gaps and effective masses). Initially,

the electronic states are written as Bloch functions:

(2.34)

4 --------,-----,-------T""-----,

Transmission Coefficient
one band

two band

-4

-+-'
Ci)

-8

///

-12

·"

,,,/"

.~/

,,,,,,

-16

,,./

,,/

Ev
-20

Energy (eV)
Figure 2.1: Log of the transmission coefficient for electrons tunneling across a 100 A
Alo.3Gao.1As barrier. The zero of energy is taken to be the valence band edge in
Al0•3 Gao. 7 As. E,, (Ee) represents the valence (conduction) band edge energy. The
dashed and solid lines represent the curves generated by calculating K with the
one and two band formulas, respectively. The prefactor in Eqn. 2.31 is taken to
be unity, since it is a £unction of the electrode material parameters.

50
where the index n refers to the band being described (e.g., conduction or valence), k
is the electron wavevector (real), r is the distance vector, and the functions Un have
the periodicity of the crystal potential, V( r). iii n(k, r) satisfies the Schrodinger
equation,

HW.(k,r) = (:~ + V(r)] w.(k,r) - E.(k)W.(k,r),

(2.35)

where H is the Hamiltonian, P is the momentum operator, m is the free electron mass, and En(k) are the energy eigenvalues. Substituting the Bloch function
expression for 'Pn(k, r) into the Schrodinger equation yields:

[ 2m

1ik. p

t,,2k2

+ --+
-2m + V(r) Un(k,r) = En(k)un(k,n).

(2.36)

For direct semiconductors, the valence band maximum and conduction band
minimum both occur at the center of the Brillouin zone. Hence, the energy eigenvalues between the two bands a.re known to be separated by the energy gap, E 9 ,

at k = 0. Defining the valence band edge to be the zero of energy, Eqn. 2.36 can
be evaluated at k = 0:

[:~ + V(r)] u,(O,r) = Ou.(O,r),

(2.37)

[:~ + V(ri)u<(o,r)

(2.38)

and

E9 u,(O,r),

where u11(0, r) (uc(0, r)) is the valence (conduction) band eigenstate at k = 0. Since
we are considering only two bands in the semiconductor,* the states at k = 0 are

a complete basis in which eigenstates for arbitrary k can be expanded:

(2.39)
(2.40)
•More bands can be included in the calculation through a straightforward extension of the

basic principles.[9]

51
where the complex coefficients, Cmn, satisfy the condition
(2.41)
Rewriting Eqn. 2.36 with the expanded expression for the conduction band state
gives;
(2.42)
where it has been assumed that Ec(k) = Ec(k) (this is true for direct gap semiconductors near k = 0).
At this point, several identities are needed. The eigenstates at k

O are

orthogonal and chosen to be normalized,

j d ru;(o,r)u (0,r) = Jd ru:(o,r)uc(O,r) O,
j d r u:(o, r)u (0, r) = j d r u:(o, r)uc(O, r) =

(2.43)
(2.44)

Next, axes are chosen such that

k•P = kP.

(2.45)

Since Pis a Hermitian operator,
(2.46)
Finally, since the k = 0 eigenstates are band extrema,[9]
(2.47)
All of the integrals above are taken over the unit cell for the crystal.

Utilizing the above identities, Eqn. 2.42 can be multiplied by u:(o, r) and integrated over the unit cell, yielding:
(2.48)

Performing the same operation with u;(o, r) gives:
(2.49)
The requirement for a solution to Eqns. 2.48 and 2.49 is that
li.2k2 2m

E (k)

E g + t,,22mk2 - E (k)

=0.

(2.50)

Solving the determinant in Eqn. 2.50 for Ec(k) yields

- Eo t,,2 k2 Eu [
4,-,,2 k2p2] t/2
Ee( k) - 2 + 2m ± 2 1 + E;m 2

(2.51)

Only the root with the positive sign in Eqn. 2.51 gives Ec(k = 0) = E0 as required
by the choice of energy eigenvalues. In order to calculate the bandstructure (i.e.,

find Ec(k)), it is necessary to determine a value for p. For small values of k,
(2.52)
and Eqn. 2.51 can be simplified:
(2.53)

The electron effective mass is defined by:

(2.54)
It follows that,
(2.55)

2P E (m 1).

m•

(2.56)

Since the electron effective mass is well known for most direct gap semiconductors,
p is easily determined.

It is assumed that p is independent of k in this model.

Eqn. 2.51 is then a closed form dispersion relation for the conduction band energy
as a function of k.
From Eqn. 2.51, values of Ee which are between O and Eg are obtained when k
is imaginary. Making the substitution,
{2.57)

k =iK

and requiring real values of K yields an expression for the imaginary part of the
electron wavevector in the energy gap:

The formula in Eqn. 2.58 has the predicted qualitative behavior, with K = 0 at

Ee.= (0, E9 ), and maximized near midgap.

2.3

Single Barrier GaAs/ AlAs Structures

Although AlAs is often used in combination with GaAs in tunnel structures,

there remains some doubt as to how it should be treated as a barrier material.
The ambiguity arises from the fact that AlAs is an indirect semiconductor, with its
conduction band minimum towards the Brillouin zone edge in the [100)-direction.*
In fact, the indirect (X-point) energy gap is only 2.17 eV at 300 K, as compared
to 3.02 eV for the direct (f-point) gap.[10]

The potential harrier height seen

by electrons tunneling from GaAs into AlAs is therefore considerably lower at
the X-point than at the !'-point. Fig. 2.2 is a band diagram for a single barrier
GaAs/ AlAs/GaAs tunnel structure which depicts this situation. In the figure, a
valence band offset of 0.55 eV is assumed.[11]
•The AlAs valence band maximum is at the zone center, with characteristics similar to most
direct gap semiconductors.

~ 1000

(I)
CJ)

-0
(I)

-0

500 -X
- - ------- ---- --- --- ----

ho~

........... __ .,.

...

X----------- -- --

-1-'

,__

.. --X........

_________________ ~---- ------------------

o i-:r;;_______

GaAs

AIAs

GaAs

-200

-100

100

200

Distance (A)
Figure 2.2: Calculated r- and X-point band edges in a single barrier GaAs/ AlAs
heterostructure at zero applied bias. The electrodes are doped n-type, with a
carrier density of 4 x 1018 cm- 3 • The barrier is doped 3 x 10 18 cm- 3 , p-type.

_Tunneling through the structure depicted in Fig. 2.2 can be treated as the parallel combination of tunneling through the AlAs r- point with tunneling through
the AlAs X-point. Since the r-point is a. local minimum for the AlAs conduction

band which lies at k

0, the two band k • p method presented in section 2.2.2

is applicable to tunneling through the direct gap. Hence, the imaginary part
of the electron wavefunction for f-point tunneling is given by Eqn. 2.58, with
m! = 0.15m.[10] For tunneling through the X-point, the one band formula is
more appropriate because the conditions needed for the two band method are not

satisfied. Furthermore, the tunneling electrons are much closer in energy to the
X-point, making the one band formula a reasonable approximation. The imaginary part of the electron wavefunction for X-point tunneling can be obtained from
Eqn. 2.33, with m: = 0.78m, the longitudinal mass.[10]
Despite the low barrier height of the AlAs X-point, there are strong physical
factors which enhance tunneling through the f-point. The large effective mass at
the X-point leads to large values of K whenever the energies of the tunneling elec-

trons are not very close to the conduction band edge. In fact, the value of K at the
X-point becomes larger than the r-point value for energies more than:::::: 100 meV
below the indirect conduction band minimum. Furthermore, the tunneling states
in the Ga.As electrodes lie near the Ga.As r-point. Hence, their wavefunctions
are considerably more like the r-sta.tes in AlAs than the X-sta.tes. The result is

an enhancement of approximately 104 in the tunneling current through the AlAs
r--point relative to the X-point. For these reasons, it has often been predicted
theoretically that tunneling in Ga.As/ AlAs structures should be governed by the
AlAs f-point.[12,13)
Figure 2.3 contains an experimental current density vs. voltage curve for the
heterostructure represented in Fig. 2.2.

Also presented is a theoretical curve,

generated by the method outlined in section 2.2. The calculation only includes

J-V Curves

,,.--.....

'-......
.....__,,

4.2 K

T -

-----

exp

J el

>........,

·( j)

Cl)

........,
Q)
I....
I....

::::,

,,,. /

,, /

.,,.,. .

,, /

,,.. . /

100

200

Voltage (mV)
Figure 2.3: Experimental and theoretical current-voltage characteristics for the
single barrier structure depicted in Fig. 2.2. The solid line corresponds to the
experimental current density, Jelllp• The dashed line is the calculated r-point
elastic current density, J!;.

57
elastic tunneling through the AlAs r-point. Near zero bias, the agreement between
the theoretical and expedmental current densities is surprisingly good. However, it
is likely that this agreement is at least somewhat fortuitous, given the uncertainties

in experimental and theoretical parameters. The most notable characteristic of
Fig. 2.3 is the qualitative and quantitative disagreement between the theoretical
and experimental curves at larger voltages. The divergence of the two curves
indicates that current transport mechanisms, other than simple elastic tunneling
through the f-point, contribute to the total current. It has been suggested that
these alternative transport mechanisms could be elastic and/ or inelastic tunneling
via the AlAs X- point.[1]

2.4

Hg 1_:vCdz Te Negative Differential Resistance
Devices

In this section, the current-voltage simulation developed earlier in the chapter is used to analyze Hgi-a:Cda: Te single barrier heterostructures theoretically.
Emphasis is placed upon optimizing the predicted negative differential resistance
behavior by maximizing current densities and peak-to-valley ratios. The effects

of varying growth parameters, such as electrode composition, doping density, and
barrier thickness, are studied. It is shown that careful choices of these parameters
must be made in order to obtain NDR. Finally, changes in device behavior due to
variations in the value of the HgTe-CdTe valence band offset, AE,,, are explored.
The tunneling current is found to depend strongly on AE", both quantitatively
and qualitatively. In fact, NDR disappears completely when the HgTe valence
band edge is more than 100 meV above that of CdTe. A calculated band diagram
for a typical Hg1-a:Cda: Te single barrier heterostructure is displayed in Fig. 2.4. It
is assumed in the figure that AE,, = 0 eV.

Hg0. 7 Cd 0. 3Te-CdTe-Hg 0. 7 Cd 0. 3Te

r----_

,,-....

>(l)

....._.,,

1000

(l)

'"'O
(l)

'"'O

..0

-+-'

:::::s

'"'O

E'f

0 .... -------------

Er -

----------

-250

250

500

Distance (A)
Figure 2.4: Cakulated band diagram for a Hg 1 _..,Cd.., Te single barrier heterostruc-

ture under an applied bias of 150 m V. Both the conduction and valence band edges
a.re shown, along with the quasi-Fermi levels. The electrodes are Hg0 .rCd0 , 3 Te,
with n = 1 x 1017cm-3 • The barrier is a. 200 A thick CdTe layer. The valence band
offset is taken to be zero.

It should be noted that the calculations in this section are performed at 4.2 K.
This choice is made because of the requirement for NDR that electron tunneling dominate the current through the heterostructure. At higher temperatures,
thermionic mechanisms need to be considered. A treatment of thermionic hole
currents in these structures is developed in chapter 3.

2.4.1

Growth Parameters

Electrode Composition

The alloy Hg1_:i:Cda: Te is a highly unusual semiconductor. For :z: < 0.16, the energy gap is zero, with the conduction and heavy hole bands degenerate at k = 0.[8]
This situation comes about because HgTe has an inverted band order, with the

"conventional" "' conduction band having negative curvature in k-space and lying
below the "conventional" light hole band, which has positive k-space curvature.
The "conventional" light hole band then becomes the conduction band, degenerate with the heavy hole band as in most semiconductors. For values of :z: > 0.16,
Hg1-0Cd0 Te is a more traditional semiconductor with an energy gap which varies
nearly linearly with :z: (there are quadratic corrections, but they are relatively
small) over the range 0-1.6 eV:[8]

Ea(T = 4.2K) ~ (-0.3eV) + (1.9eV):z: 0.16 < ai < 1.0.

(2.59)

Examination of Fig. 2.4 reveals that, as the electrode band gap is widened, the
energies of the tunneling electrons increase with respect to the valence band edge
in the CdTe barrier. Hence, larger electrode energy gaps weaken the NDR effect,
because the transmission coefficient changes most sharply for energies near the
valence hand edge of the barrier (see, for example, Fig. 2.1). It follows that small
• "Conventional" is taken to mean the band assignment in more standard semiconductors,

such as GaAs and CdTe.

values of :z: are desirable for the Hg 1 _a:Cda: Te cladding layers. However, a nonzero
energy gap in the electrodes is needed because filled states in the valence band of
one electrode can tunnel across the barrier into empty states in the conduction band
of the opposite electrode if the two bands overlap. In fact, valence to conduction
band tunneling is allowed whenever the voltage applied to the structure becomes
larger than the electrode band gap. Therefore, it is necessary to choose :z: such
that the Hg 1 _a:Cd 111 Te energy gap is somewhat larger than the operating voltages
for the strnctnre. 'ryptcally, the optimal electrode energy gap is in the range 100
- 200 me V, corresponding to 0.21 < :v < 0.26. This range is quite narrow in terms
of growth considerations.

Barrier Composition
Due to the high vapor pressnre of Hg at room temperature, MBE systems which
grow Hg1 -a:Cd11 Te require special design features for the handling and removal of
excess Hg. Despite these features, all CdTe layers grown in a. Hg 1 _ 111 Cda: Te MBE
system incorporate some Hg. Estimates of the minimum attainable background
concentration vary from 5 to 15 percent.[5,20] The consequence of background
Hg incorporation for the single barrier heterostructure is an alloying of the barrier
layer. For uniform alloying, the energy gap of the barrier is in the range 1.32 eV

< E';tlTe < 1.5 eV, as compared to 1.6 eV for CdTe. In the single barrier NDR
structure, most of the energy gap difference between the cladding layers and barrier
is absorbed in the conduction band offset, with the valence band offset affected
to a lesser degree by changes in barrier composition. Since device behavior is
largely determined by the valence band offset, little qualitative change is seen in
theoretical I-V curves generated for slightly alloyed barriers a.s compared to pure
CdTe barriers. The overall current density is found to decrease somewhat due to
a decrease in the effective mass used to calculate K in Eqn. 2.58.

Electrode doping
The concentration of free carriers in the Hg 1 -z Cda: Te electrodes plays an important role in determining the 1-V behavior of the single barrier heterostructure.
Two effects are involved. First, the amount of voltage dropped in the cladding layers as determined by band bending is strongly affected by the charge concentration
in the electrodes. Since the NDR effect relies upon changing electron transmission
coefficients by placing a voltage across the barrier, minimizing the voltage "lost" in
bending the cladding layer bands is critical. The voltage lost in the cladding layers
becomes larger as the electrode doping is decreased due to less effective screening of
the electric field. Second, the peak of the I-V curve occurs when additional states
are no longer added to the tunneling integral as the voltage is increased. This
point is reached when the conduction band edge in the negatively biased electrode
is raised to an energy equal to the quasi-Fermi level in the opposite electrode. For
larger free carrier concentrations in the electrodes, the quasi-Fermi levels become
larger, and the voltage required to reach the peak condition is increased. Low
operating voltages in these structures are desirable because NDR disappears when
the voltage surpasses the energy gap in the electrodes, as discussed previously.
Thus, the two effects of electrode doping density work in opposite directions; low

densities are better for low operating voltages, while high densities are better for
dropping applied voltage in the barrier. Fig. 2.5 displays J-V curves for a single
barrier structure, calculated by choosing three different electrode doping concentrations. All three curves were generated with the model developed in section 2.2,
choosing a CdTe barrier thickness of 170 A, zero valence band offset, and :z: = 0.22

in the Hg 1 _zCd 111 Te electrodes. The best NDR performance is obtained for the
middle value of electrode doping. As predicted, the peak of the curve calculated
for higher carrier concentration is shifted to higher voltage. The weak NDR in the
case of low electrode doping is due to large voltage drops in the cladding layers.

0.4

J-V Curves

,.,--....

0.3

(.)

"E
<(

....__,;

>.......,
(/)

0.2

(l)

.......,

~,--~ -----

---~·- ----~----.

-·---~

(l)

!,...

:::l
(.)

0.1

n -

1 X

10 17
10 16
1016

Voltage (mV)
Figure 2.5: Current density vs. voltage for a Hg1 _ 21 Cdai Te single barrier heterostruc-

ture with three different electrode doping densities. The electrode ma.terial is
Hgo,1sCdo.22Te. The barrier is a 170 A thick CdTe layer. The valence band offset
is taken to be zero.

Barrier Thickness
Figure 2.6 is a plot of the current density vs. barrier thickness for a single barrier
structure with Hg0•78 Cd0•22 Te cladding layers and a CdTe barrier under an applied
bias of 50 m V. The magnitude of the tunneling current in the Hg1-:i: Cd:i: Te single
barrier heterostructure is strongly dependent upon the thickness of the barrier
layer. As is the case in many tunnel structures, this dependence arises from the
exponential factor in the transmission coefficient (see Eqn. 2.32). In order to
observe NDR, electron tunneling must dominate the current through the single
barrier structure. Thin barriers are desirable because they enhance the tunneling
current as compared to other currents, and yield higher operating current densities.
However, two problems arise from barriers which are too thin. First, the strength
of the NDR effect is directly related to the magnitude of the exponential factor:
(2.60)
As the barrier becomes thinner, the effect of increasing K values as the energies
of the tunneling electrons increase is reduced. As a result, smaller peak-to-valley
current ratios are observed, or NDR is lost completely. Second, the amount of
voltage dropped in the barrier decreases as the barrier thickness decreases be-

cause the increased electric fields require more band bending in the electrodes.
As discussed previously, large cladding layer voltage drops can weaken or eliminate NDR. Clearly, barrier thicknesses can be too small or too large for the single
barrier heterostructure to yield NDR. However, optimal barrier thicknesses are
difficult to quantify heca11se competing current transport mechanisms are not well
understood.

J vs. w

,--,
.,,.--.....

(.)

<(
......__,,,

-+-'
(/)

(I)

-2

+->

(I)
LL-

::J

i........i

-4

-6..___....________.____.....__ _,
120

160

200

CdTe Barrier Thickness (A)
Figure 2.6: Log of the current density through a Hg0 ;18 Cdo.22Te/CdTe heterostruc.

ture vs. the CdTe layer thickness. The electrode doping density is 4 x 1016 cm,- 3 •
The applied bias is 50 m V.

2.4.2

Effects of Valence Band Offsets on NDR

In contrast to most of the growth parameters discussed in the previous section,
the effect of the valence band offset, A.Ev, on NDR from the Hg1-a:Cda: Te single
barrier heterostructure is monotonic.t Smaller values of A.Ev yield significantly
better NDR than larger values. Figs. 2.7 (a) and (b) contain linear and log current
density vs. voltage plots for several choices of A.Ev in a single barrier Hg1_.,Cd., Te
heterostructure. The linear plots demonstrate that peak-to-valley current ratios
and total current densities are strongly reduced as the valence band offset is increased. In fact, NDR disappears altogether for A.Ev > 100 meV. The log plots
quantify the drop in current density at higher A.Ev's. As the valence band offset
is increased, the conduction band edges in the Hg 1 -a:Cd 21 Te electrodes are moved
up in energy with respect to the barrier. The drop in current density is due to the
increasing values of K seen by tunneling electrons at energies further away from

the valence band edge in the barrier. The reduction in peak-to-valley current
ratios is a result of the gentler slope of K with energy at higher energies in the
barrier.

2.5

Other Material Combinations for Single Barrier NDR

A few material combinations other than Hg1_.,,Cd.,,Te-CdTe are possible candidates for single barrier NDR. In this section, two of these heterostructures are
analyzed conceptually and theoretically. A second demonstration of single barrier NDR has recently been made in one of these material combinations, InAs-

flt is relatively certain that AE.,, is not extremely negative, i.e., the HgTe valence band edge

is not well below the CdTe valence band edge.

1-V Curve

AEV = 0 meV
AE:v=65 meV

,,_,

:::,

(a)

L:::::=======::::.====---_J

100

200

A£,,
0 m•V

e, ....v

.._,

:::,

c::n

-4

-a 0

256 me\/

4$0 mrN

(b)
100

200

Voltage (mV)

Figure 2.7: Linear (a) and log (b) plots of the current density vs. voltage behavior calculated for a. single barrier Hg 1 _ 11 Cd:i, Tc heterostructure. Curves for several

values of l::i.E11 are plotted. The electrodes have :ll = 0.22, and an electron concentration of 4 x 1016 cm- 3 • The barrier is a 170 A thick CdTe layer.

2.5.1

InAs-Ga1_ 2 Ala:Sb Single Barrier Heterostructures

InAs-GaSb heterostructures have been of interest for some time.[18)

However,

considerably less experimental literature exists for this system than GaAs/ AlAs,
largely because growth techniques for the antimonides have lagged behind the those
for the arsenides. The most unique feature of the InAs-GaSb heterojunction is a
completely staggered band alignment. Recent x-ray photoemission measurements
have placed the valence band offset between the two materials (EfaSb - EJnA 11 )
at 0.51 eV.[19] Since the InAs energy gap is only 0.35 eV at 300 K, the InAs
conduction band minimum lies below the GaSb valence band maximum by 0.16
eV. Thus, InAs-GaSb tunnel structures have no barriers, since neither material
can have free carriers which lie at energies in the forbidden gap of the other.
The semiconductor AlSb has a large indirect energy gap, with a "normal" type I
band alignment against GaSb. X-ray photoemission experiments have yielded a

valence band offset (E;t 08" - Ef15b) of 0.40 eV.[17]

Assuming a linear variation

of the Ga1-aiAla:Sh valence band edge with :z:, it is possible to deduce a relation for
the InAs-Gai-a:Ala:Sb valence band offset:

EJnA1t ~ E~aAlSb - {0.51eV) + (0.40eV):z:.

(2.61)

For single barrier NDR structures, device behavior is optimized when the conduction band edge in the electrodes is nearly aligned with the valence band edge in

the barrier. For an InAs-Ga1-a:Ala:Sb-InAs heterostructure, this alignment can be
obtained by choosing:
:z: ~

0.51eV - 0.35e V
= 0.4.
0.40e

(2.62)

A calculated band diagram for a single barrier structure with z = 0.45 in the
Ga1-aiAlaiSb layer is given in Fig. 2.8. In the diagram, the InAs conduction band
edge lies 20 me V above the Ga.1-a:Ala:Sb valence band edge.
This material combination has several advantages over Hg1_aiCda: Te for single

lnAs

1000

Al
0.45

0.55

lnAs

I-

,,--.....

>(1)

....._,,,
(/)

(1)
CJ')

-0
Cl)

-0

-200

200

400

Distance (A)
Figure

2_8:

Calculated

valence

and

conduction

band

edges

for

InAs-Ga,o~4sAlo.ssSb-InAs heterostructure under an applied bias of 50 m V. The
electrode doping density is taken to be 5 x 1016 cm- 3 •

69
barrier NDR structures. (i) The ability to tune the position of the valence band
in the barrier by varying the Al mole fraction provides a degree of freedom which
does not exist for Hg 1 _zCgiven above are correct, it is possible to position the tunneling electrons very close
in energy to the valence band maximum in the barrier. (iii) III-V semiconductors are better for device fabrication and are more stable than Hg1-zCdm Te. ( iv)
Thermionic hole currents in this system are limited by the InAs energy gap, which is
significantly larger than that of llg0 •78 Cd0 •22 Te. Thus, room temperature operation
may be possible . A calculated J-V curve for a single barrier InAs/Gao.ssAlo.45Sb
structure with a 100 A thick barrier layer is displayed in Fig. 2.9.

2.5.2

Pb 1_ 111 Sn 111 Te Single Barrier Heterostructures

The ternary semiconductor Pb 1_2 Sn2 Te has an extremely narrow energy gap,

E0 , over the entire composition range:[21]
E0 (T = 12K) ~

0.19 + 0.54z a: < 0.35

(2.63)

z > 0.35.

It has been proposed by Heremans et al.[22] that single barrier NDR could be observed in a heterostructure fabricated from different compositions of this material.
This argument is based upon the pinning of the Fermi level in these materials by
the addition of indium doping during growth. It is assumed that the bands in
the electrodes and barrier will strongly screen electric fields, resulting in the bulk
(pinned) alignments of the bands. Thus, band offsets are ignored. In fact, it is
proposed that the barrier should have z = 0.3, with the electrodes having z = 0.2,
i.e., the electrode energy gap is larger than that of the barrier. This choice is made
because the Fermi level in bulk, indium doped Pb 1_2 Sn2 Te rises strongly with increasing Pb content, so that it rests in the conduction band for :z: = 0.2, and near
the valence band edge for :z: = 0.3. Alignment of the Fermi level across the entire

1400
J-V Curve
1200
,--.....

1000

"<(
....__,,,.
>+-'

800

(/)

(1)

600

+-'

(1)
\.,_

!,_

:::)

400
200
0 .___

__._____.__...____

____._____,

100

300

200

Voltage (mV)
Figure 2.9: Calculated current density vs. voltage curve for the single barrier
structure depicted in Fig. 2.8.

71
heterostructure results in the required single barrier NDR band alignment, with
the conduction band in the Pb 0•8 Sn 0•2 Te electrodes lying slightly higher in energy
tha.n the valence band in the Pb0 •7 Sn 0 .aTe barrier.

Several conceptual and practical problems make it unlikely that the Pb 1 _ 0 Sn.ll) Te
single barrier device will show NDR. First, a physically correct view of the band
diagram (accounting for band offsets) will shift the band alignment away from the
desired configuration. Furthermore, including band bending effects weakens predicted NDR behavior significantly. Second, the energy gap in the barrier is only
30 meV. Hence, extremely thick barriers are needed to produce any attenuation of
the tunneling electrons, and voltages larger than 30 m V push the tunneling electrons to energies higher than midgap of the barrier. Third, the extremely narrow
energy gaps lead to large competing thermionic currents. Finally, the addition of
indium doping in the barrier is likely to introduce scattering centers, substa.ntially
reducing the probability of elastic tunneling across the thick barrier layer.

References
[1] A.R. Bonnefoi, Ph. D. Thesis, California Institute of Technology, 1987.

[2] J. Bardeen, Phys. Rev. Lett. 6, 57 (1961).
[3] R. Tsu and L. Esaki, Appl. Phys. Lett. 22, 562 (1973).
[4] M.O. Vassel, J. Lee, and H.F. Lockwood, J. Appl. Phys. 54, 5206 (1983).

[5] B. Ricco and M.Y. Azbel, Phys. Rev. B 29, 1970 (1984).
[6] K. Araki, J. Appl. Phys. 62, 1059 (1987).

[7] D.D. Coon and H.C. Liu, Appl. Phys. Lett. 49, 94 (1986).
[8] E.O. Kane, in Tunneling in Solids (Academic Press, New York, 1969), Chapter
1.

[9] E.O. Kane, in Physics of 111-V Compounds (Academic Press, New York,
1966), Vol. 1, Chapter 3, pp. 75-100.

[10] H.C. Casey, Jr., and M.B. Panish, Heterostructure Lasers (Academic Press,
New York, 1978).
[11] J. Batey and S.L. Wright, J. Appl. Phys. 59, 200 (1986).

[12] G.C. Osbourn, J. Yac. Sci. Technol. 17, 1104 (1980).

73
[13] C. Ma.ilhoit and T.C. McGill, J. Vac. Sci. Technol. B 1, 637 {1983).
[14] R. Dornhaus and G. Nimtz, in Narrow-Gap Semiconductors (Springer-Verlag,
Berlin, 1983), Part 2.

[15] J.P. Faurie, private communication.
[16] O.K. Wu, private communication.

[17] H.Munekata, T.P. Smith, and L.L. Chang, to appear in J. Va.c. Sci. and Technol. B, March/ April,1989.

[18] L.Eso.ki, in The Technology and Physics of Molecular Beam Epita:r.y, edited
by E.H.C. Parker (Plenum, New York, 1985), Chapter 6.

[19] G.J. Gualtieri, G.P. Schwartz, R.G. Nuzzo, R.J. Malik, and J.F. Walker, J.
Appl. Phys. 61, 5337 (1987).

[20] G.J. Gualtieri, G.P. Schwartz, R.G. Nuzzo, and W.A. Sunder, Appl. Phys.

Lett. 49, 1037 ( 1986).
[21] G. Nimtz and B. Schlicht, in Narrow-Gap Semiconductors (Springer-Verlag,
Berlin, 1983), Part 1.
[22] J. Heremans, D.L. Partin, P.D. Dresselhaus, and B. Lax, Appl. Phys. Lett.
48, 644 (1986).

Chapter 3
Electrical Studies of Hg 1 _xCdxTe
Single Barrier Heterostructures
3.1

Introduction

3.1.1

Motivation and Background

This chapter presents results from an experimental study of the current-voltage

(I-V) behavior of single barrier Hg1 _ 0 Cd2 Te heterostructures. The original intent
in undertaking this study was to demonstrate the novel single barrier negative
differential resistance (NDR) mechanism described in section 1.5. The motivation
for developing single barrier NDR structures is to obtain electrical devices which
can operate at extremely high frequencies. In these applications, the NDR feature
can be exploited in designing oscillators, amplifiers, and mixers. Furthermore, single barrier heterostructures may have better high-frequency response than double
barrier (resonant tunneling) devices, in which a quantum well must be charged and
discharged during high speed switching. Further motivation for studying the particular choice of Hg 1 _ 0 Cd 111 Te single barrier devices is derived from current interest
in heterostructures containing Hg 1 _11:Cd 2 Te. The ternary compound Hg 1 -a,Cd2 Te

75
has been a far-infrared detector material of choice because of its tunable narrow
energy gap.[1) More recently, a great deal of attention has been focused upon the
high quality layered growth of HgTe, CdTe, and Hg 1 _..,Cd.., Te, largely due to the potential applications of the HgTe-CdTe superlattice as an infrared material.[2,3,4,5)
Another device of interest is the double barrier HgTe/CdTe heterostructure, from
which room temperature NDR has been demonstrated.[6,7]

Electrical studies

of single barrier heterostructures can provide fairly direct measurements of material and heterojunction properties, such as barrier penetration distances and band
offsets. Detailed knowledge of these properties can facilitate the design of more
elaborate heterostructure devices.
Published theoretical and experimental values of the HgTe/CdTe valence band
offset (AE,,) range from Oto 500 meV.[9,l0,11,12,13,14,15,16,15,18]

A few expla-

nations can be suggested for the disagreement between the various experimental
measurements. Firstly, many of the experimental techniques measure the band
offset in a.n indirect fashion. For example, most of the measurements on superlattice samples rely on superlattice bandstructure calculations to determine the band
offset. Secondly, many of the experiments are performed under different conditions
(e.g., different temperatures). Finally, growth techniques for Hg1 _ 111 Cd111 Te are not
yet reproducible enough to be certain that samples grown in different places and/or
times are of comparable quality. Many of the current transport mechanisms in single barrier heterostructures are strongly dependent on the valence band offset.
In the case of the Hgi-mCda: Te structures studied here, thermionic hole currents
have a direct exponential dependence on (AE,,/kT). At high temperatures this
thermionic mechanism can dominate transport, yielding a fairly direct measurement of AE,,. When conditions are such that elastic electron tunneling dominates
the current, the theoretical simulations of section 2.4 show that NDR can be observed only if AE,, is less than 100 meV. Hence, the successful demonstration of

NDR places an upper limit on (.~E11 ).

3.1.2

Summary of Results

NDR has been observed from a single barrier Hg1_:Cda: Te heterostructure at
temperatures accessible with liquid helium, but not at higher temperatures. Due to
the extremely tedious nature of the low temperature bonding technique used here,
only a few devices were studied, with roughly half displaying NDR. The best device
yielded a peak-to-valley current ratio of 2:1, with a peak current density of 0.51

mA/cm- 2 • The strongly temperature dependent behavioris probably attributable
to competing current transport mechanisms which freeze out at extremely low
temperatures. It is possible that a temperature dependent valence band offset also
plays a role in restricting NDR to low temperature I-V curves. The observation
of NDR places an upper limit of 100 meV on t:,.Ev at 4.2 K. The demonstration of

NDR in this material system verifies the predicted single barrier NDR mechanism.
It is possible that other material combinations could yield room temperature NDR
and better device performance.
The theoretical analysis of Hg1-a:Cda: Te structures given in section 2.4 indicates
that stringent selection of growth parameters is a requirement for achieving single
barrier NDR in this material system. Thus, the successful demonstration of NDR
is indicative of a relatively controllable and reproducible growth technique. Furthermore, reasonable material quality is indicated by these results, because crystal
defects and impurities tend to provide transport paths which can compete with or
dominate elastic tunneling.
At higher temperatures (>80 K), a more detailed investigation of I-V behavior
was ma.de. Two HgTe/CdTe single barrier samples were studied in addition to the

Hg1-111Cd~Te sample which displayed NDR at low temperatures. Measured currents were found to vary linearly with device area in all three samples, indicating

the absence of surface leakage currents. Over the temperature range 190-300 K,
the observed current is attributed to the sum of two transport mechanisms: ( i)
thermionic emission of holes across the barrier layer, and ( ii) holes tunneling across

a "triangular-shaped" barrier. This interpretation of the current is supported by
good agreement between observed current-voltage curves and theoretical simulations which include only these two mechanisms. The dependence of the measured
current with temperature was then used to determine AE11 • Results from the three
samples at 300 K yielded values of AE., between 290±50 and 390±75 meV. In all

three samples, data taken over the range 190-300 K are consistent with a valence
band offset which decreases at lower temperatures.
Several other Hg1 -aiCda: Te single barrier samples were grown for this study.
Most of them displayed reproducible 1-V behavior, with measured currents varying linearly with device area. However, the CdTe barriers in these samples were

considerably thinner than in the three samples discussed above. The effect of
thin barriers is to enhance elastic and inelastic tunneling currents while leaving
thermionic currents unchanged. The measured currents in these samples could not
be attributed to the thermionic hole mechanism at high temperatures. Therefore,
the data from these samples were not used to determine AE.,. Current-voltage
experiments at liquid helium temperatures were not attempted in these samples.

3.1.3

Outline of Chapter

A description of the samples grown for the Hg1 _ 111 Cda: Te single barrier study
is given in section 3.2. Section 3.3 describes the experiment used to demonstrate
NDR at liquid helium temperatures. I-V data from a few devices are presented,
including two curves taken at different temperatures from a single device. Higher
temperature current-voltage behavior from three single barrier structures is investigated in section 3.4. A straightforward theoretical model of thermionic and

tunneling hole currents is developed and used to analyze the 1-V data. The HgTeCd Te valence band offset is then determined over the temperature range 190-300 K.
Section 3.5 gives brief descriptions of the electrical behavior nhRerved from ot.her

samples grown for this study. Finally, conclusions are summarized in section 3.6.

3.2

Samples

Table 3.1 contains a listing of the single barrier Hg 1 _a,Cd:e Te heterostructures
grown for this study by molecular beam epitaxy (MBE). The majority of the
samples (TSl through TS16) were grown at the University of Illinois at Chicago

(UIC) by I.K. Sou and J.P. Faurie. Three additional samples (ML26 A, A', B)
were provided by F.A. Shirland and O.K. Wu at Hughes Research Laboratories
(HRL).in Malibu, California.
Semi-insulating Ga.As was used as a substrate material for the samples grown
at UIC. Growth commenced with a thick CdTe buffer layer(~ 3 µm), which provided a high quality, lattice-matched template for growth of the first Hg 1 _:eCd:it Te
electrode.* High resolution transmission electron microscope (TEM) pictures of
the GaAs/CdTe interface indicate that an amorphous CdTe layer forms initially
on the Ga.As {100) surface. After the first few hundred angstroms of growth, highly
crystalline (111)-oriented CdTe is nucleated and maintained for the remainder of
the layer. A typical high resolution TEM print of this interface is shown in Fig. 3.1.
Following the CdTe buffer, a thick Hg1 -a,Cdz Te layer was grown to form the bottom electrode of the heterostructure. This layer was usually doped n-type with
indium, although defect doping was attempted in a few samples (TSl, TS3, TS5).
A thin CdTe layer was then deposited to form the single quantum barrier. Due
to the constant Hg overpressure in MBE systems which grow Hg1 _a,Cdz Te, the
•The lattice constant of CdTe (6.481 A) is nearly identical to that of HgTe (6.462 A).[8]

Growth Parameters for Hg1_111 CdzTe Single Barrier Samples

Sample
No.

Bottom Electrode

CdTe

Top Electrode

Electrode

Thick-

Barrier

Thick-

:ll

ness

Width

ness

(µm)

Doping
(101scm-3)

(A)

(µm)

Doping
(1016cm-3)

TSl

0.20

3.0

p - 0.45

100

0.5

p - 0.45

TS2

0.23

3.0

2.25

100

0.5

2.25

TS3

0.20

2.6

0.08

0.43

0.08

TS4

0.22

3.0

3.57

100

0.5

3.57

TS5

0.22

2.6

p - 0.01

0.46

p - 0.01

TS6

0.22

3.0

3.58

150

0.5

3.58

TS7

2.59

80.
50.

80
80

0.44

TS8

0.33
0.32

TS9

0.295

3.3

TSlO

U.315

2.84

40.

0.51
0.59

80.
50.

0.46

TSU

0.33

3.26

60.

0.53

TS12

0.21

2.82

2.0

0.46

TS13

0.235

3.15

6.0

0.52

100.
4.0
8.0

TS14

0.27

2.45

20.

0.41

50.

TS15

0.30

2.66

20.

0.44

100.

TS16
ML26A

0.31
0.0

2.73
1.0

20.

80
180

0.46
0.5

100.

ML26A'
ML26B

0.0

1.0
1.0

180
180

0.5

60.
60.

0.0

3.17

60.
60.
60.

0.5

40.

60.

Table 3.1: Table of growth information for Hg1 -zCd 111 Te single barrier samples.
Samples TS1-TS16 were grown at the University of Illinois at Chicago on GaAs
(100) substrates. Samples ML26 A, A', B were grown at Hughes Research Laboratories on CdTe (111), CdZnTe (111), and CdTe (100), respectively. All doping
densities are n-type, unless otherwise noted with a 'p'. Carrier concentrations for
the samples grown at UIC (HRL) were measured at 30 K (300 K). All CdTe layers
are nominally undoped.

I II

I:: I I

}fl I'

·~ I {

· !rn·111111fn

~,.

\., ......

Figure 3.1: High resolution transmission electron microscope photograph of the
GaAs(lOO)/CdTe(lll) interface.

81
barrier layer is alloyed to some extent. The actual composition is estimated to be
Hg0 •05 Cd 0 •95 Te.[19] Finally, an n-type Hg 1 _ 0 Cd111 Te top electrode was grown, with
a layer thickness of ~ 0.5 µm.

The three samples from HRL were grown on three different substrates:
CdTe(l00), CdTe(lll), and Cd0 •5 Zn 0 •5 Te (111). A thick CdTe buffer layer (~
3 µm) was grown on each of these substrates prior to the growth of the single
barrier heterostructure. In contrast to the samples grown at UIC, the intent in
designing these samples was solely to determine l!:..E., by measuring thermionic hole
currents across the single barrier (i.e., NDR was not sought from these samples).
This shift in focus resulted in a choice of pure HgTe a.s the electrode material
instead of a narrow gap Hg1 _zCd 2 Te alloy. The advantage in using HgTe is to
simplify the analysis of the measured thermionic hole current. In addition, the
CdTe barriers in the samples grown at HRL were made thicker to suppress elastic
and inelastic tunneling mechanisms. Hg incorporation in the barriers of these
samples is estimated to reduce their composition to Hg0 ,15Cd0 .a 5 Te.[20]

3.3

Demonstration of NDR at Low Tempera-tures

3.3.1

Sample TS6

Upon including band bending effects in the current-voltage simulations developed in section 2.2, sample TS6 was found to be the only sample from which
strong NDR could be expected. Thus, major efforts were concentrated on TS6 in
this study. Several TEM runs were performed on this sample in addition to over
10 preparations for electrical measurements. Sample TS6 was the only sample
studied at liquid helium temperatures.

82
The major difference between TS6 and the other samples grown at UIC is the
thickness of its barrier layer. In table 3.1, this width is listed a.s 150 A, which is
the thickness estimated from growth rate calibrations. All of the ot.hel' sa.mpleR al'e

estimated to have barrier layers of 100 A or less. As discussed in section 2.4, thicker
barriers allow less of the total applied bias to be lost in bending the cladding layers.
Furthermore, peak-to-valley current ratios rise dramatically with thicker barriers
because the exponential factor in the electron transmission coefficient becomes
more heavily weighted. The disadvantages of thicker barriers are the reduction of
the peak current density and the increased role of competing transport mechanisms.
These competing currents are often more strongly temperature dependent than
elastic tunneling. Hence, low temperature measurements are often used to observe
tunneling effects while freezing out other mechanisms. Extremely low temperatures
were needed to observe NDR in sample TS6 ( < 20 K).

3.3.2

Device Fabrication

To study current-voltage behavior in a semiconductor heterostructure, it is almost always necessary to electrically isolate small areas of the sample. Isolation
is usually accomplished by forming mesa.a in the sample. The mesa. heights are
selected to be greater than the film thickness from the sample surface to the active
portion of the structure. This approach has several benefits. ( i) Sample nonuniformities, such as large crystal defects, can cause an electrical device to become short
circuited. If sufficiently small mesa areas can be achieved, these nonuniformities
can be excluded from a large number of the mesas. ( ii) Operating currents are
reduced by smaller device areas, lessening the detrimental effects of parasitic series
resistance sources ( such as contacting wires and measurement circuitry). ( iii) Fabricating a large number of small devices on a single sample, instead of a few very
large devices, allows a more extensive study of electrical behavior. ( iv) For samples

83
in which electrical contact cannot be made to the substrate, the bottom electrode
can only be contacted by removing the active region of the heterostructure over a
portion of the sample area.

The result of the device fabrication process is depicted schematically in Fig. 3.2.
The figure shows the side view of a single mesa along with contacts to the top and
bottom electrodes.

The following list is the procedure used to fabricate two

terminal devices in sample TS6.
1. A small piece(~ 3 mm x 4 mm) of the sample is selected by cleaving with a

Circon microscribe. In the remainder of the procedure list, this small piece
is referred to as the 'sample'.

2. The sample is cleaned by successive immersion in acetone, methanol, and
deionized water for 1 minute each. The water is then blown off with dry
nitrogen.
3. Positive photoresist (Shipley AZ 5214) is spun onto the sample. Photolitho-

graphic procedures are then used to expose and develop the photoresist film.
The result is an array of circular photoresist dots, with varying diameters
ranging from 35 to 70 µ,m.

4. Mesas are fabricated in the sample by wet etching with Br 2 :HBr:H 20 in a
0.005:1:3 ratio by volume. The etch rate is approximately 0.4 µm/min. This
chemical etch permits the use of conventional positive photoresist in pho-

tolithographic procedures, as opposed to the more commonly used Br 2 :methanol which attacks most positive resists. The etch leaves a surface which is
somewhat less smooth than the unetched surface. The photoresist circles
are then removed with acetone. Etching depths are checked with a stylus
profilometer.

Au contact

Hgo."/s Cd 0.22 Te

Au contact

CdTe

..
.."

Hgo.1s Cd 0.22Te

CdTe Buffer

GaAs Substrate

Figure 3.2: Schematic diagram depicting the result of the device fabrication process
for sample TS6. The diagram shows the side view of a single mesa along with Au
contacts to the top and bottom Hgi-mCdai Te cladding layers.

85
5. A second coating of photoresist is spun onto the sample, exposed, and devel-

oped. The second mask is inverted with respect to the first mask, and has
slightly smaller circular features. Holes result in the photoresist film; they

are aligned with the tops of the mesas.
6. A gold film is evaporated over the entire sample. The sample is then immersed in acetone and exposed to an ultrasonic cleaner. As a result, the gold

film is lifted off of the sample everywhere except for on top of the mesas,
forming a contact to the top electrode. This contact was found to be ohmic
by probing circular gold dots on an unetched sample.
7. A contact to the bottom electrode is made by evaporating gold on the sample

near one end only. The series resistance which arises due to electrons traveling
laterally through the bottom Hg1 _ 0 Cdg: Te layer was found to be negligible by
comparing devices which were different distances away from the gold 'back'
contact.
Room temperature electrical tests of fabricated devices in sample TS6 have
been made by probing the mesas with a 25 µ,m diameter gold wire. Fig. 3.3 is a
log-log plot of the current at 300 K vs. mesa diameter for a typical preparation

of sample TS6. The data were obtained at an applied bias of 25 m V. Although
the spread of measured currents is fairly large, the slope of the best fit line to the
data in Fig. 3.3 is nearly 2, indicating that the current varies linearly with device
area. This condition verifies that surface leakage currents and/or isolated defects
do not dominate current transport in the fabricated devices. Roughly 25% of the
devices probed were "short-circuited", with markedly higher currents and nearly
linear I-V curves. These "short-circuited'' devices were not included in the plot
of Fig. 3.3.

-8

Sarnple TS6
-6.2

.....,

( I)

-6.4

I()

""'C

""'C(I)
''::I

-6.6

.....

-6.8

Slope=1.96

-7
1.5

1.6

1.7
1.8
log [Device Diameter {microns)]

1.9

Figure 3.3: Log-log plot of the current vs. mesa diameter for a typical preparation

of sample TS6. The current measurements were made at 300 K under an applied
bias of 25 mV.

3.3.3

Low Temperature Electrical Contacts

Current-voltage measurements at low temperatures ( <80 K) were performed
by placing the prepared samples into a liquid helium immersion dewar. Since probe
wires could not be used inside the dewar, it was necessary to develop a permanent
bonding scheme for the fabricated devices. Conventional wire bonding machines
were not suitable for this purpose because the Hg1_.,Cd., Te devices were damaged
by the ultrasonic bonding pulse. The following is a description of the permanent
bonding technique used.
1. The technique creates bonds which are approximately 100 µm diameter cir-

cles. Since the fabricated devices are smaller than the bonds, it is necessary
to insulate the etched surface of the sample. Photoresist is spun on to the
sample, exposed, and developed by standard photolithographic procedures.
The second mask of the two mask process described in section 3.3.2 is used
during exposure. Thus, an insulating photoresist film is placed on the sample,
with holes aligned such that only the tops of the mesas are exposed.
2. The sample is mounted on an 8-pin header with silver print.
3. A conductive epoxy, Acme E-Solder No. 3021, is mixed. This epoxy can be
cured at room temperature.
4. A thin gold wire (25 µm diameter) is mounted on a manipulator. The end
of the wire is then dipped in the conductive epoxy.
5. With the assistance of a microscope, the wire is guided to a device by the
manipulator. Good electrical contact is verified by measuring the I-V characteristic on a curve tracer during this process.
6. The epoxy is left to cure overnight.

7. Epoxy is applied to a pin on the header. The wire is then carefully clipped
near the manipulator mount, and guided with tweezers to the header pin.
The epoxy on the pin is then left to cure overnight.
8. The excess wire beyond the header pin is clipped.
9. Contacts to the etched surface are made in the same fashion, with the end
of the wire epoxied to the gold pad at the end of the sample.
Room temperature I-V behavior from bonded devices was fouI1d to be identical
to that obtained prior to bonding. The technique described above is both tedious
and time consuming, and results in bonds which are not particularly resilient at
low temperatures. In fact, cracking of the photoresist film and/or the epoxy bond
often occurs upon cooling the sample to liquid helium temperatures. Consequently,
only a few devices have been studied. It should be noted that low temperature
stations which allow in situ probing of samples do exist. The use of such a station
should enhance the reproducibility of this experiment considerably.

3.3.4

I-V Results at Low Temperatures

As discussed previously, the low temperature I-V behavior of sample TS6 was
investigated by immersing bonded devices in liquid helium. Sample headers were
mounted on a probe with electrical feedthroughs and a temperature sensor. The
current-voltage curves shown in this subsection were obtained with a HP 4145A
Semiconductor Parameter Analyzer. At low biases, it was possible to limit back•
ground current fluctuations to less than 10 pA by shielding all of the measurement
cables. Data were stored in digital form on a HP9816 computer, and transferred
to a DEC Microvax computer for plotting via an IEEE 488 bus.
Figure 3.4 contains I-V curves, taken from a 37 µm diameter device at two
temperatures, 4.2 and 15 K. The curves were obtained under reverse biased con·

I-V Curves
4.2 K

15 K

,,,.--.....
<:(

....__..,

,/"'\

,4.-J

(I)

::)

'\
'\

(.)

80
Voltage

120

160

(mV)

Figure 3.4: Reverse bias I-V curves from a 37 µ,m diameter device. The solid
(dashed) line is the curve obtained at 4.2 K (15 K).

90
ditions (negative voltage on the top electrode). The curve taken at 4.2 K displays
negative differential resistance, with a peak current density of O.51 mA / cm- 2 at
109 m V. The peak-to-valley current ratio is slightly greater than 2:1, in reasonable agreement with the simulation results of section 2.4. However, the simulations
also indicated that NDR should be displayed over the voltage range 50-100 m V
(roughly), in contrast to the 109-139 mV range observed in Fig. 3.4. It is possible
that a large contact resistance develops in the epoxy bonds at low temperature,
leading to a shift of the tunneling characteristics to higher biases.
As shown in Fig. 3.4, the effect of lowering the temperature from 15 K to 4.2 K
is to increase the peak-to-valley current ratio from 1.6:1 to 2.3:1. This increase
is mainly due to an enhancement of the peak current, indicating that tunneling is
enhanced at low temperatures. In contrast, double barrier tunnel structures often
display increased peak-to-valley ratios at low temperatures due to the freeze out of
nonresonant transport mechanisms. This freeze out results in a drop in the valley
current. Increases in the tunneling current with decreasing temperature are difficult to understand because tunneling is a fairly temperature independent process.
Two possibile explanations are consistent with the observed data. ( i) The tunneling
electrons are nearer to the conduction band edge in the Hg 1 _a:Cda: Te electrodes at
lower temperatures due to sharpening of the Fermi distribution. This effect would
result in smaller imaginary wavevectors ( and therefore, larger transmission coefficients). ( ii) The valence band offset decreases as temperature decreases, leading
to smaller imaginary wavevectors. Section 3.4 presents evidence for a temperature
dependent valence band offset in this material system. The observation of NDR
implies that the low temperature valence band offset is less than 100 meV, as
discussed in section 2.4.
Figure 3.5 displays the forward bias 1-V curve at 4.2 K from the 37 µ,m diameter device discussed above. The curve shows two distinct NDR regions, with peak

I-V Curve
T = 4.2 K

400

Forward Bias

,,,--....

Q_
....__,,
.+-J

(])
!r....
!r....
:::)

200

0 i-=;....__1..---_.____'---_.____.____...___,
40
80
120

Voltage (mV)
Figure 3.5: Forward bias I-V curve from same device as in Fig. 3.4. The curve
was obtained at 4.2 K.

92
current densities of 0.01 mA/cm2 at 57 mV and 0.039 mA/cm 2 at 109 mV. This
bimodal characteristic is not predicted by our straightforward electron tunneling
model. It is possible that nonuniformity in the portion of the sample covered by
this device is responsible for the observed behavior. TEM studies indicated that
the CdTe barrier thickness varies laterally from 170 to 250 A in sample TS6. Another possible explanation is the presence of filled interface states, lying within the
energy gap of the Hg 1 -mCd111 Te electrodes. Such states could contribute electrons
to the tunneling current, yielding discrete peaks as they become aligned with the
conduction band edge in the opposite electrode.
The asymmetry between the forward and reverse bias I-V curves may be caused
by an asymmetery between the interfaces on either side of the CdTe barrier.
Fig. 3.6 is a high-resolution TEM photograph of the active region of the sample.

A twin boundary is seen at the interface between the top Hg1 -mCdm Te

electrode and the CdTe barrier. In contrast, the interface between the barrier and
the bottom Hg 1 _mCda: Te layer shows no evidence of twinning. The two types of
interfaces are similar to the type A and B orientations which have been observed
for NiSi 2 on Si.[21] In both cases, the [111] growth direction gives rise to the two
possible orientations. The NiSi 2 :Si barrier height was shown to vary by greater
than 100 mV, depending upon which type of interface was grown.[21,22]
Other device1i te1itt:d at low temperatures gave a variety of results. A set of

devices was fabricated on a second piece of sample TS6 to test the reproducibility
of the observed NDR. A forward biased I-V curve taken from a 67 µm diameter
device on the second piece is shown in Fig. 3.7. The curve displays NDR over the
voltage range 48-65 m V, with a peak-to-valley current ratio of 1.4:1. However,
the peak current density is almost two orders of magnitude lower than that dis-

played in Fig. 3.4. This device also displayed inflections ( d2 I/ dV 2 changed from
positive to negative) at a higher positive bias, and in reverse bias. The reverse bias

iii
i I·
I I

Figure 3.6: High resolution TEM picture of the active region of the Hg1 _.Cd 11 Te
single barrier heterostructure. The picture reveals a twin boundary at the interface
between the top Hg1 -a:Cda:Te layer and the CdTe barrier.

400

I-V Curve
T = 4.2 K
Forward Bias

300

0..
..........,,,
-+-'

200

Q)
I...
I...

::J

100

0 i..c:;._...___...___...___..._________________,

Voltage (mV)
Figure 3.7: Forward bias I-V curve from a 67 µm diameter device on sample TS6.
The curve was obtained at 4.2 K.

characteristic is displayed in Fig. 3.8. Two other devices also displayed inflections
at 4.2 K. However, three tested devices did not display any inflections or NDR.
The variations in the I-V behavior of different devices may have been caused by

nonuniformity in the sample or by the poor resiliency of the epoxy bonds at low
temperatures.

3.4

Electrical Determination of the HgTe/CdTe
Valence Band Offset

This section presents a study of the high temperature ( > 160 K) I-V behavior of
three Hg1 _11:Cd 111 Te single barrier samples: TS6, ML26A, and ML26A'. The growth
parameters for these samples were described in section 3.2. At high temperatures,
the measured current ia interpreted to be the sum of thermionic and tunneling

h.ole curtents. Both. of these transport mechanisms depend exponentially upon the
HgTe-CdTe valence band offset, A.Eu, Hence, analysis of the I-V data can yield a
determination of A.Eu.

3.4.1

Theoretical Simulations of Hole Currents

Energy band diagrams of Hg1 _ 111 Cd111 Te single barrier heterostructures with thick
CdTe barriers (see for example, Fig. 1.5) suggest that the dominant source of current at high temperatures is the thermionic emission of holes from the Hg 1 _..,Cd.., Te
cladding layers across the CdTe valence band barrier. It is important to note that
the n-type doping of the electrodes does not prohibit this transport mechanism
because the electrode energy gaps are small (in this case, ~ 200 meV). In fact, the
thermionic barrier for electrons is much larger than that for holes, due to the large
conduction band offset in these heterostructures.
A simple theoretical treatment, similar to the Bethe model for Schottky

120

1-V Curve
T = 4.2 K
Reverse Bias

80
.....--....

<(
.............,

.......

(I)

::J

120

160

Voltage (mV)
Figure 3.8: Reverse bias I-V curve at 4.2 K from the same device as in Fig. 3.7.

barriers,[23] can be developed to calculate thermionic and tunneling hole current
densities a.cross a single barrier as a function of applied voltage. The portion due
to thermionic emission, Jthe:rm, can be written:

Jth ..... tn

ceV) [1 - exp (-eV)]
kT

= A *T 2 exp (-+
kT

(3.1)

where A* is the modified Richardson constant,¢ is the potential barrier height, Tis
the temperature, e is the hole charge, and c is the fraction of the total applied voltage which drops across the positively biased electrode. For these heterostructures,

A* is 120(mi:) in A/cm 2 K 2 , where m;: is the unitless hole mass. The contributions

from the light and heavy hole bands are summed to give the total current from
this mechanism in the analysis of sample TS6. For samples ML26A and ML26A',
only the heavy holes contribute to the current significantly, because the light hole
band is split off in HgTe. It is important to note that the factor c is a function of
the voltage applied across the heterostructure, and must therefore be derived from
the energy band diagram for each individual bias condition. In this study, c has
been calculated by the method outlined in section 2.2. The value of c is generally
in the range 0.1 - 0.4 for the heterostructures studied here, as compared to the
case of a Schottky barrier, where c = 1.
For sample TS6, the potential barrier height, ¢, in Eqn. 3.1 can be written:
(3.2)
where E1 is the Fermi energy relative to the conduction band minimum in the
.electrodes, E;= 0•22 is the energy gap in the electrodes, and the quantity (E:= 0 •22 -

E;= 0 •95 ) is the valence band offset at the Hgo.78 Cd0 •22 Te-Hg0 •05 Cd 0 •91;Te interface.
The barrier height is reduced in samples ML26A and ML26A' due to the zero ba.nd

gap HgTe electrodes:

(3.3)

where the quantity (E:= 0 - E:= 0·85 ) is the valence band offset at the HgTeHg0.15Cd0.85Te interface. It should be noted that the reduction in barrier height due
to the HgTe cladding layers is partially compensated for by the larger composition
difference between the barrier and the electrodes.
For applied voltages of~ 50 mV and higher, hole tunneling across the triangular shaped CdTe barrier makes a contribution to the total current through the
heterostructure. This transport mechanism can be treated theoretically in a manner which is analogous to the model for the thermionic hole current. The resulting
expression for the hole tunneling current density, Jht,m, differs from that for Jtherm
by an integral term which replaces the contribution to offset. For sample TS6, we obtain,
Jhtun

0 22
Ea:= • +

= A*T2 exp ( -E1

ceV) [1 - exp (-eV)]
JA
(-u
kT
t u exp
- ) du.

(3.4)
In this expression,

· (3.5)
where

v, is the group velocity of the holes in the growth direction,

'ILA

= [2(E;=0·22 -

E:=0·95)/kT]1l2, and t 2 is the transmission coefficient for holes tunneling through
the CdTe barrier. Similarly, for samples ML26A and ML26A',

i .... = A'T'

exp cEk; V) [

1 - exp ( ~";)

lr exp (;}u,
t'u

(3.6)

where UB = [2(E:= 0 - E;= 0 •86 )/kT] 1 l 2 • In this study, t 2 is calculated by the
method described in section 2.2. The two band k • p theory formula was used to
find imaginary light hole wavevectors in the CdTe barrier, while imaginary heavy
hole wavevectors were determined from the one-band formula.

3.4.2

Device Fabrication

Device preparation for sample TS6 was performed according to the procedure
described in section 3.3.2. Samples ML26A and ML26A' were prepared by a slightly
different procedure. Chemical etching was accomplished by the use of Br2:ethylene
glycol instead of Br 2:HBr:H2 O. The change in procedure was adopted because
Br 2:HBr:H 2O was found to etch these samples nonuniformly, with material removed
in large Hakes. However, the Br2:ethlylene glycol recipe left a surface which was
comparable in quality to the unetched surface.

Circular mesa diameters were

reduced to 15-40 µm for preparations of ML26A and ML26A'. Measured currents
were found to vary linearly with device area in all three samples, indicating the
absence of surface leakage currents.

3.4.3

1-V Results and Analysis

In this subsection, experimental current-voltage measurements are presented,
and analyzed via the theoretical model developed in section 3.4.1. High temperature (>80 K) current-voltage curves were measured by probing the fabricated
devices with a thin (25 µm diameter) gold wire. Temperatures below 300 K were
reached with an MMR Joule-Thompson cooling station.
Figure 3.9 contains an experimental current density-voltage (J-V) curve, taken
from sample TS6 at 300 K. Also plotted is the J-V curve generated by the theoretical model discussed above for a barrier height r/, = 514 meV. This value of¢,
was chosen by requiring the theoretical and experimental current densities to be
equal at 50 mV, and was the only adjustable parameter used. Selecting a different
value of the applied bias results in changes in depicted in Fig. 3.9.
The electrode carrier concentrations given in table 3.1 for sample TS6 were
determined at low temperature {30 K). Hall measurements have been performed

100

400

J-V Curve
T -

300 K

Sample TS6

--.....

300
Experiment
Theory

...........

200

........

Q)
I...

::J

100

0 ____._____._________.._________

100

200

Voltage (mV)
Figure 3.9: Experimental J-V curve taken from sample TS6 at 300 K. Also plotted
is a theoretical curve calculated for a HgTe-CdTe valence band offset of 390 meV.
A.Ev is the only adjustable parameter used to generate the theoretical curve.

101
at 300 K, yielding a free electron concentration of 1.5 x 1011 cm- 3 in the cladding
layers of sample TS6. Assuming that the electrons form a nearly free Fermi gas,
E1 can be estimated to be 44 ± 10 meV above the conduction band edge in the

Hg0 _18 Cd0 •22 Te electrodes. The value of E;=0 •22 is taken here to be 185 ± 20 meV
at 300 K.[8] In addition, the uncertainty of the cladding layer compositions is
estimated to produce an uncertainty of ~ 15 me V in the value of the electrode
energy gap. Eqn. 3.2 then gives,
( E:=0 •22

E:=0 •95 ) = 285 ± 55 meV.

(3.7)

A linear extrapolation of this expression to a pure HgTe-CdTe heterojunction
yields AEv = 390 ± 75 meV.
As discussed previously, samples ML26A and ML26A' are expected to yield
higher current densities than sample TS6. Furthermore, the higher electron den-

sities in the pure HgTe electrodes result in less of the applied voltage being
dropped there, i.e., , the factor c in Eqn. 3.1 is smaller for samples ML26A and
ML26A'. Thus, the current density varies more slowly with voltage in these samples. Fig. 3.10 contains an experimental J-V curve taken from sample ML26A' at
300 K. Also plotted is the theoretical curve generated for a barrier height tp = 332
meV, which was selected in the same manner as the tp used in Fig. 3.9.

As ex-

pected, Fig. 3.10 displays larger current densities than Fig. 3.9, and shows a weaker
voltage dependence. E1 is estimated to be 75 ± 30 meV for the HgTe cladding
layers at 300 K. Eqn. 3.3 then gives,

(3.8)
Linear extrapolation of this expression to a pure HgTe-CdTe heterojunction yields

AE,, = 300±,'50 meV. A similar analysis of 300 K data from sample ML26A obtains
AE,, = 290 ± 50 meV.

102

200

J-V Curve
T -

300 K
Sample ML26A'

,,,--.....

Experiment
------ Theory

<(
....__,,,,

4-1

(/)

100

(l)

-+-'

(l)
II-

::i

,,. /

0 _____...._____,_______.._______.

100

200

Voltage ( mV)
Figure 3.10: Experimental J-V curve taken from sample ML26A' at 300 K. Also
plotted is a theoretical curve calculated for a HgTe-CdTe valence band offset of
300 meV.

103

Figure 3.11 contains the measured current density from each of the three samples as a function of temperature over the range 190-300 K. The data were taken
for an applied bias of 50 mV and are diApla.yed -in the Ata.nda-rd log(J/T 2 ) vs. 1/kT

format. As discussed previously, the current density in sample TS6 is considerably less than in the other two samples at all temperatures. In addition, samples
ML26A and ML26A 1 yield currents with nearly a T 2 temperature dependence,
while sample TS6 varies more strongly with temperature. These results can be
shown to be consistent with a valence band discontinuity which decreases nearly
linearly as the temperature decreases.
In samples ML26A and ML26A', E1 behaves roughly as (const xT) due to the
nearly intrinsic HgTe cladding layers. Thus, if

(3.9)
then Eqn. 3.3 yields,
ex T.

(3.10)

The current density in Eqn. 3.1 would then have a T 2 dependence , in agreement
with the data shown in Fig. 3.11.
On the other hand, Eqn. 3.2 has an extra term, E;=0 · 22 , which has a temperature independent part:(8]
E;=0 •22 me V ~ 100 + 0.284 x T.

(3.11)

In addition, the Fermi level in the cladding layers of sample TS6 is not due to
intrinsic carriers, and will therefore have a temperature independent pa.rt over the
range of interest here."' This part can be estimated to be 25 me V from the carrier
densities given previously. Therefore, we suggest that the barrier height in sample
•sample TS6 was indium doped; no intentional impurity doping was used in the growth of

samples ML26A and ML26A'.

104

log ( J /T 2 ) vs. ( 1/kT)

-8

,-.

-t::_
2,

-12

Q.)

• Sample TS6

O'I

• Sample ML26A
• Sample ML26A'

-16

Figure 3.11: Measured current density from samples TS6, ML26A, and ML26A'
as a function of temperature, plotted in a loge(J/T 2 ) vs. 1/kT format. The data
from all three samples are consistent with a valence band offset which decreases
linearly as the temperature decreases.

105
TS6. can be written:

(3.12)

This behavior would give a current density in Eqn. 3.1 which depends on temperature as [T 2 x exp(-125/kT)]. The line in Fig. 3.11 has a slope of-120 meV, in
reasonably good agreement with this hypothesis.
It should be noted that if an unknown transport mechanism is contributing to
the observed currents, the above analysis will lead to false determinations of the
band offsets. However, the fact that the experimental J-V behavior is very close
to that predicted by the theoretical model used here supports the assertion that
the observed current is due solely to thermionic and tunneling hole currents.

3.5

Other Samples

As discussed previously, the remainder of the samples listed in table 3.1 were
unsuitable for single barrier NDR effects because of their thin barrier layers. None
of these samples has been studied at liquid helium temperatures. Furthermore,
these samples were not used for the band offset determination of section 3.4 because
the observed currents could not be attributed to the thermionic hole mechanism.
Nevertheless, the majority of the samples display nonlinear I-V curves, indicating
that the CdTe barrier plays an active role in limiting current. It is possible that
elastic and/or inelastic tunneling dominate the high temperature transport across
thin CdTe layers.
Figure 3.12 is a. log-log plot of current at 300 K vs. mesa diameter for a preparation of sample TS4. Although the spread of measured currents is large for each
device diameter, the slope of the best fit line to the data in Fig. 3.12 is uea.rly 2,
indicating the absence of surface leakage currents. Most of the samples listed in
table 3.1 yielded currents which scaled linearly with device area. However, four

106

Sample TS4

-4.6

....,
,.... -4.8

(I)

=e
II)

...,

-5

...,

t..

(.)
.......

'·,·,

.. .. ..

'-------- Slope=1.8

-5.2

-5.4

1.5

1.6

1.7

1.8

1.9

log [Device Diameter (microns)]
Figure 3.12: Log-log plot of current vs. mesa diameter for a preparation of sample
TS4. The current measurements were made at 300 K under an applied bias of

25 mV.

107
of the samples, TSl, TS3, TS5, and ML26B, did not satisfy this criterion. In the
cases of the UIC samples, doping difficulties (TSl and TS5 were p-type, while TS3

had virtually no carriers) are the most likely cause of the unusual I-V behavior.
Sample ML26B may suffer from structural defects due to the (100) orientation of
the CdTe substrate.

3.6

Summary of Conclusions

We have reported the first experimental observation of NDR due to electron
tunneling in a single barrier heterostructure. The sample used to demonstrate NDR
consisted of a thin CdTe layer sandwiched between two Hg0 .rsCd0 . 22 Te electrodes.
The largest peak-to-valley current ratio attained was slightly greater than 2:1. In
the Hg1 _a:Cd 11 Te material system, NDR can only be achieved at low temperatures

(T < 20 K) due to the dominance of thermionic hole currents at high temperatures.
The observation of NDR in this system implies that the low temperature valence
band discontinuity at the HgTe-CdTe interface is less than 100 meV.
High temperature current-voltage behavior from three Hg 1 _ 11 Cd11 Te heterostructures has been investigated. The measured currents have been interpreted to
be the sum of thermionic and tunneling hole currents. This analysis yielded values
of the HgTe-CdTe valence band offset between 290 ± 50 me V and 390 ± 75 me V at
300 K. In all three samples, data taken over the range 190-300 K were consistent

with a valence band offset which decreases at lower temperatures.

108

References
[1) J. Ameurlaine, J. Coester, and H. Hofheimer, Opt. Spectra 27, 1973.

[2] J .N. Schulman and T .C. McGill, Appl. Phys. Lett. 34, 663 {1979).

[3] C.E. Jones, T.N. Casselman, J.P. Faurie, S. Perkowitz, and J.N. Schulman,
Appl. Phys. Lett. 47, 140 (1985).

[4] S.R. Hetzler, J.P. Baukus, A.T. Hunter, J.P. Faurie, P.P. Chow, and T.C.
McGill, Appl. Phys. Lett. 47, 260 (1985).

[5] J.P. Faurie, IEEE J. Quantum Electron. QE-22, 1656 (1986).

(6) J.N. Schulman and C.L. Anderson, Appl. Phys. Lett. 48, 1684 (1986).
[7] M.A. Reed, R.J. Koestner, and M.W. Goodwin, Appl. Phys. Lett. 40, 1293
(1986).

[8] R. Dornhaus and G. Nimtz, in Narrow-Gap Semiconductors (Springer-Verlag,
Berlin, 1983), Part 2.
[9] J.O. McCaldin, T.C. McGill, and C.A. Mead, Phys. Rev. Lett. 36, 56 (1976).

[10] W.A. Harrison, Phys. Rev. B 24, 5835 (1981).

[11] J. Tersoff, Phys. Rev. B 30, 4874 (1984).
[12] W.A. Harrison and J. Tersoff, J. Vac. Sci. Technol. B 5, 1068 (1986).

109
[13] S.H. Wei a.nil A. Znnge-r, .J. Vac. Sci. Technol. A 6, 2597 (1988).

[14] K.J. Malloy and J.A. Van Vechten, Appl. Phys. Lett. 54, 937 (1989).
[15] Y. Guldner, G. Bastard, J.P. Vieren, M. Voos, J. P. Faurie, and S. Million,

Phys. Rev. Lett. 51, 907 (1983).
[16] D.J. Olego, J.P. Fa.urie, and P.M. Racca.h, Phys . .Rev. Lett. 55, 328 (1985).

[17] S.P. Kowalczyk, J.T. Cheung, E.A. Kraut, and R.W. Grant, Phys. Rev. Lett.
56, 1605 (1985).
[18] J.P. Faurie, C. Hsu, and T.M. Due, J. Vac. Sci. Tecbnol. A 5, 3074 (1987).

[19] J.P. Faurie, private communication.
[20] O.K. Wu, private communication.
[21] R.T. Tung, Phys. Rev. Lett. 52, 461 {1984).
[22] R.J. Hauenstein, T.E. Schlesinger, T.C. McGill, B.D. Hunt, and L.J. Sch~walter, Appl. Phys. Lett. 47, 853 (1985).
[23] S.M. Sze, Physics of Semiconductor Devices (Wiley, New York, 1981), pp.
255-259.

110

Chapter 4
Molecular Beam Epitaxy of
111-V Heterostructures

4.1

Introduction

4. 1. 1

Background

The purpose of the molecular beam epitaxy (MBE) system described in section 1.4 is to supply semiconductor heterostructures for many different research
projects. Thus, a large menu of available semiconducting materials is desirable
because it enhances the :flexibility of the system for growing novel heterostructure devices. Unfortunately, a combination of many different source materials
in a single MBE chamber often results in a loss of semiconductor purity due to
cross-contamination of the various materials. Hence, the pursuit of novel heterostructures as a general goal results in a trade off of material purity for variety
of samples. Structures which require extremely high purity and reproducibility are
not well-suited to such a research program, and are generally better produced in a
more development oriented situation. Due to the nature of semiconductors, even
systems which are designed for high flexibility can be contaminated to the extent

111

that desired structures cannot be grown. Clearly, a balance between purity and
flexibility is needed.
As was discussed in section 1.4, the MBE system used here has chambers for
each of the three major classes of semiconductors. This greatly reduces the probability of contamination due to elements from different columns of the periodic
table, which is considerably worse than contamination within a class. For example, a one part per thousand concentration of indium in MBE grown GaAs is
considerably less serious than the same incorporation of tellurium. This is due to
the fact that tellurium is a dopant in Ga.As, and would result in very degenerate ntype material, in contrast to the small indium alloying effect which has little effect
upon the behavior of many Ga.As devices. The separation of the MBE system into
independent growth chambers for different classes of semiconductors prevents the
more serious cross-contamination possibilities, while maintaining a large degree of
:flexibility in designing novel heterostructures.
The III-V growth chamber portion of the MBE system has been used to produce all of the samples discussed in this section. Most of the work presented
here involves the heavily studied Ga.As/ AlAs system. This material combination
continues to be of interest for high speed and optoelectronics applications, and
for studies of fundamental quantum phenomena. Although much of the current
research being performed with the III-V chamber involves other materials, it is

likely that Al111 Ga1 _ 111 As heterostructures will continue to be of interest for quite
some time. Thus, it is highly desirable to retain the capability to grow high
quality GaAs, AlAs, and Al 21 Ga1 _ 111 As. Periodic growths of a few standard, wellcharacterized Al111 Ga1 _ 21 As heterostructures are performed in the III-V chamber
routinely to check for contamination problems.

112

4.1.2

Summary of Results

A few standard Al:eGa1 _:eAs structures have been characterized to provide reference points for the status of the system at any given time. GaAs/ AlAs double
barrier heterostructures have been found to yield I-V curves which display reproducible negative differential resistance (NDR) behavior. Peak-to-valley current
ratios of 2.5:1 at 300 Kand 10:1 at 77 Kare routinely obtained under good system
conditions. Photoluminescence spectra from single quantum well heterostructures
have been shown to yield sharp exciton peaks at the expected confinement energy.
The full width at half maximum of these peaks corresponds to :fluctuations of
less than one monolayer in well thickness. Modulation doped GaAs layers (or high
electron mobility transistors) have been characterized by Hall effect measurements.
A pronounced enhancement in the electron mobility at 77 K is observed due to
the spatial separation of free carriers a.nd ionized impurities. Bulk (lightly doped

n-type) Ga.As layers have also been characterized by photoluminescence and Hall
effect measurements.
A measurement of the GaAs/ AlAs valence band offset by x-ray photoemission spectroscopy has been made. This experiment was performed by growing
GaAs/ AlAs heterojunctions and bulk samples in the III-V chamber and transfering them through the UHV transfer tube to the ESCA chamber. The GaAs
valence band maximum was found to lie 0.46 ± 0.07 eV above that of AlAs, independent of growth sequence (the offset is commutative). The capability to study
the MBE grown samples without exposing them to atmosphere was ideal for this
measurement.
A set of double barrier GaAs/ AlAs heterostructures has been grown for optical
measurements of electron tunneling rates. These structures were designed to have

varying AlAs barrier thicknesses, constant Ga.As well thicknesses, and a very thin
top GaAs cladding layer ( ~ 300 A). This sample geometry permitted time resolved

113

measurements of the quantum well photoluminescence, which was found to decay
as electrons tunnel out of the well. Measured tunneling times ranged from 12 ps
for 16 A barriers to 800 ps for 34 A barriers. An exponential dependence of decay
time .with barrier thickness was observed.
Triple barrier GaAs/ AlAs heterostructures have been grown and prepared as
two terminal electrical devices as a test of electron coherence in resonant tunneling structures. NDR has been observed in the I-V curves of these devices, with
multiple resonances indicating a coherent nature of the tunneling process. Thin
middle AlAs barriers have been found to yield the best NDR behavior in terms
of peak-to-valley current ratios and number of resonances. As the middle AlAs
barrier thickness is increased, the NDR behavior is degraded due to a loss of coherence. Unusually large peak-to-valley current ratios (as large as 19:1) have been
observed at 77 K from these structures.
A method for growing high quality (relaxed) InAs on GaAs substrates has
been developed and tested thoroughly.

The method relies on a short period

In0 .1Ga.o.3 As/GaAs superlattice at the InAs/GaAs interface. It is hypothesized
that the superlattice supresses island formation during the initial, heavily dislocated growth, allowing a high quality bulk InAs layer to be deposited once the film
has reached the unstrained InAs lattice constant. A set of 2 µm thick InAs layers
has been grown in this fashion, and tested by in situ RHEED analysis, Hall effect
measurements, optical surface morphology, and x-ray diffraction. Electron mobilities comparable to bulk InAs values have been obtained. RHEED oscillations,
comparable in quality to the best reported for InAs growth, have been observed.

Recent Results
Current research activities on the III-V growth chamber involve a larger variety of materials. Unfortunately, most of these results are too recent for detailed

114

reports to be included in this thesis. A quick summary of some of the current activities is given here. InAs/ AlSb double barrier heterostructures have been grown and

demonstrated to show resonant tunneling behavior which rivals that of GaAs/ AlAs
double barriers. Furthermore, InAs/ AlSb structures are potentially much faster
than their GaAs/ AlAs counterparts due to the high InAs mobility and the high
conduction band offsets in this system. The InAs/GaAlSb single barrier NDR
structure discussed in section 2.5 has been grown and demonstrated to have considerably better performance than that reported by Munekata et al.[1] InAs/GaSb
superlattices have been grown and demonstrated to yield infrared photoluminescence. These structures have been proposed to be well suited to applications as
far-infrared detectors with the addition of In to the Ga.Sb layers, and/or the use
of thicker superlattice layers. All of these newer structures have been grown on
thick InAs buffer layers, deposited on Ga.As substrates by the previously described
method.

4. 1.3

Outline of Chapter

Section 4.2 presents growth parameters and characterization data for a number
of standard Ga.As/ Al11:Ga 1 _:11As heterostructures. Included in the discussion are
double barrier structures, single quantum wells, high electron mobility transistors,
and bulk film properties. Brief descriptions of substrate cleaning procedures and
general growth information are also included in this section. The samples grown for
the measurement of the Ga.As/ AlAs valence band offset by x-ray photoemission
are discussed in section 4.3. Section 4.4 contains results from an optical study
of tunneling times in Ga.As/ AlAs double barrier heterostructures. Emphasis is
placed on design and MBE growth of samples for the study. Section 4.5 presents
an investigation of electron coherence in triple barrier GaAs/ AlAs heterostructures.
The development of a method for growing high quality InAs bulk layers on GaAs

115

substrates is discussed in section 4.6. Finally, the conclusions of this chapter are
summarized in section 4. 7.

4.2

Standard AlxGa 1_:vAs Heterostructures

As discussed previously, several Ala:Ga1 -mAs heterostructures have been characterized and used as standards for the status of the III-V chamber. This section
contains descriptions of the standard structures and typical figures of merit obtained under good system conditions. Brief discussions of substrate preparation
procedures and typical GaAs/ Al111 Ga1 -a,As growth conditions are also included.

4.2.1

Substrate Preparation

All of the samples discussed in this chapter have been grown on GaAs sub-

strates, which are cheaper and generally of higher quality than most commercially
available III-V substrates. These substrates can be obtained in either conductive
(heavily doped n-type) or insulating form. Etch pit densities are typically on the
order of 103 cm- 2 to 104 cm- 2 for (100)-oriented substrates.
Generally, GaAs substrates are etched and polished mirror smooth by the companies that sell them (such as Sumitomo). The procedure for preparing these
substrates for MBE growth is fairly well known,[2] although minor variations are
found from one laboratory to the next. The procedure followed in our laboratory is
given here. Initially, the substrates are immersed successively in warmed solvents
(trichloroethane, acetone, and isopropyl alchohol) to remove organic contaminants
from the surface. Next, the substrates are rinsed in deionized water, blown dry, and
etched in 5:1:1 H 2 S0 4 :H 2 0:H 2 0 2 for 2 minutes. This etch removes approximately
10 µ.m of material from the surface, and serves to eliminate polishing damage and
contaminants near the surface. A protective oxide is left on the GaAs surface.

116

Finil..lly, the substrates are rinsed in deionized water and blown dry with filtered
nitrogen gas.
Once etched, GaAs substrates are carefully cleaved into suitably sized pieces
and indium bonded to molybdenum blocks. These blocks are then loaded into a
small "intro hatch" which can be pumped down from atmosphere to the 10-s Torr
quickly. Each of the three intro hatches on the MBE system can accomodate six
blocks, and can be heated to approximately 100°C to remove water vapor prior to
entry into the UHV chamber.
The protective oxide on a GaAs substrate is removed once it has entered the IIIV chamber by heating it to approximately 600°C in an As flux. The desorption of
the oxide can be monitored by reflection high energy electro~ diffraction (RHEED).
At the time of oxide desorption, the RHEED pattern is observed to change from
a hazy uniform background with a few diffraction spots to a clear set of streaky
spots on a dark background. It has been observed that the power output from the
substrate heater power supply required to reach the oxide desorption temperature

is nearly constant for a given block once it has been well coated.

4.2.2

Al0 Ga 1_ 0 As Growth Parameters

Optimum growth parameters for GaAs, Al11 Ga1 _ 111 As and AlAs have been thoroughly studied.[2] The three major parameters are substrate temperature, As/Ga
(or As/ Al) flux ratio, and growth rate.
For all of the Ga.As/ Al 111 Ga1 _ 111 As heterostructures discussed in this chapter, the
substrate temperature was chosen to be~ 600°C. This temperature is considered
to be optimal for GaAs growth. However, it has been reported that better quality
Al 111 Ga1 _ 111 As and AlAs layers are usually obtained at higher temperatures (up to
~ 700°C).[2,3]

Five methods have been used to monitor substrate temperatures

in this study: ( i) optical pyrometry, ( ii) thermocouple readings on the back of

117

the molybdenum block, ( iii) block color, ( iv) output power of the substrate heater
power supply, and ( v) observation of oxide desorption by RHEED. Of these methods, the optical pyrometer has proven to be the most reliable. Output power is

reproducible for a given block, but is more cumbersome to use, since it doesn't
give a direct temperature readout. Oxide desorption provides a good calibration
point for the other methods, and block color is useful as a rough check at high
temperatures. Thermocouple readings generally provide only relative information
about substrate temperatures.

Flux ratios are difficult to measure quantitatively. However, it has been reported that the optimum As flux for GaAs growth is slightly greater than that
required to maintain the As-stabilized surface. This surface is characterized by
a 2 x 4 reconstruction, which can be observed by RHEED. We have calibrated
residual gas analyzer (RGA) scans to the transition from the As-stabilized surface
to the Ga-stabilized surface. The desired flux ratio is then obtained by adjusting
the As evaporator temperature until the RGA peak heights are in the appropriate range. A flux monitor ("nude" ion gauge) is also used to check the flux ratio.
Since the congruent sublimation temperature for Al 111 Ga1 _ 111 As is high (compared to
GaAs), a smaller As flux than that used for GaAs is usually sufficient to maintain
the As-stabilized surface for AlmGa 1 _ 111 As.
As discussed previously, III-V growth rates are usually determined solely by
the group III flux. We have measured growth rates for bulk MBE grown films by
shadowing a small portion of a substrate with a tantalum wire, and measuring the
depth of the resulting trench with a stylus profilometer. The rates obtained this
way have been found to be reproducible to within 2% for films grown within a few
da.ys (i.e., within a few growths) of each other. RIIEED intensity oscillations have

also been.used to calibrate thin film growth rates. The thin film rates are usually
found to be approximately 10% higher than the bulk rates. This disparity may be

118

ca.used by temperature transients in the ovens initiated by opening the shutters,
or by enhancement of the growth rate due to the incident electron beam. We have
usually grown GaAs at a rate of 1 pm/hour, or approximately 1 monolayer/second.

AlAs has been grown at rates ranging from 0.1 µm/hour to 1 µm/hour. The
AlmGai-a:As growth rate has been found to be identical to the sum of the GaAs
and AlAs rates for given Ga and Al oven temperatures, respectively.

4.2.3

Double Barrier Heterostructures

As discussed in chapter 1, extensive work has been reported on resonant tunneling in double barrier AI:,,Ga1 -a:As heterostructures. Since much of our research is
directed towards tunneling in semiconductors, these double barrier structures are
a logical choice for standard growths. Two terminal electrical devices are defined
in these structures, and tested for negative differential resistance (NDR). Peak

to-valley cunent ratios and peak current densities are used as measures of growth
quality.
We have employed two standard double barrier geometries. Each of these
heterostructures begins with a 0.5 µm, heavily doped n-type GaAs layer, grown on
a conductive Ga.As substrate. Next, a 500 A lightly doped (n ~ 2x 10 16 cm- 3 ) GaAs

spacer layer is grown, followed by a 25 A undoped GaAs spacer layer. These lightly
doped and undoped layers have been found to greatly improve NDR behavior in

double barrier structures.[4] One of the standard heterostructures has an active
region consisting of a 60 A GaAs quantum well sandwiched between two 60 A
At.,Ga1-mAs barriers, with ~ = 0.45. The other standard structure has a 45 A
GaAs well between two 25 A AlAs barriers. Each structure is capped with a
25 A undoped GaAs spacer layer, followed by a 500 A lightly doped Ga.As layer
(n ~ 2 x 1016 cm- 3 ), and a 2500 A heavily doped n-type GaAs top elect-rode.
Typical I-V curves for a standard double barrier structure with Al0 •45 Gao. 55 As

119

barriers are shown in Fig. 4.1. The curves were measured at 300 Kand 77 K from
a 147 µm diameter device. NDR is displayed in both bias directions, with nearly
symmetric current peak positions. The reverse bias peak-to-valley current ratio is
5:1 (1.3:1) at 77 K (300 K). The forward bias peak-to-valley ratio is 2:1 {1.3:1) at
77 K (300 K). Peak current densities are on the order of 200 A/cm2 in both bias
directions. We consider these results to be fairly typical for this heterostructure
geometry under good system conditions (i.e., we have observed both better and
worse peak-to-valley current ratios and current densities). A noticeable degradation in NDR behavior from these standard double barrier structures has been
observed when material quality is relatively poor. For example, it is well known
that the first few growths following a venting and bakeout of an MBE chamber
tend to incorporate high background impurity levels. This effect is thought to be
caused by contaminants which condense oTI the source material during venting.
Standard double barrier structures (with Alo,45Ga0 •55 As barriers) grown shortly
after a venting and bakeout of the III-V chamber usually show no NDR at room
temperature, and small peak-to-valley ratios at 77 K. The sensitivity of the I-V
curves from this double barrier heterostructure to system conditions makes for a
useful standard.
Improved resonant tunneling characteristics have been observed from the other
standard double barrier geometry described above. The improvement is largely due
to the thin AlAs barriers, which allow larger tunneling currents, while presenting
a high potential barrier to thermionic currents.[5) These structures typically yield
room temperature NDR with peak-to-valley current ratios greater than 2:1. Peak
current densities are in the range 104 A/cm 2 • Low temperature I-V measurements
are somewhat more difficult for these structures because the high current densities
require the use of small diameter devices (5 to 20 µm). Permanent wire bonded
contacts a.re too large to be made to these devices without shorting to the etched

120

I-V Curves
;1

/, !'

300 K

77 K

.,../

/,,..

(1,)
I...

I...
:)

-40

-0.8

-0.4

0.4

0.8

Voltage (V)
Figure 4.1: I-V curves from a double barrier tunnel structure with AlmGa1 -a:As
barriers. The curves were taken from a 14 7 µm diameter device. The solid (dashed)
curve was taken at 300 K (77 K).

121
surface. Thus, immersion of the devices in liquid nitrogen is difficult. Nevertheless, these double barrier heterostructures make excellent standards, because their
room temperature NDR. behavior is reproducibly good under reasonable system
conditions.

4.2.4

Quantum Well Photoluminescence

A thin layer of GaAs sandwiched between two Al:i,Ga1 -:iiAs barriers forms a
quantum well for both electrons and holes. The energy difference between the
electron and hole ground states in the quantum well, E 0 , is larger than the energy
gap of GaAs:
( 4.1)

where E;a:A, is the energy gap of GaAs, and f~ ( 1:i) is the confinement energy
of the electron (hole) ground state in the quantum well. In a photoluminescence
experiment, electrons and holes are generated by an incident photon flux (usually
from a laser), and allowed to recombine, yielding photons at characteristic energies.
Thus, the photoluminescence spectrum from a single quantum well should show a

peak near E 0 •
The quantities f~ and f~ can each be obtained through a straightforward numerical solution for the ground state of a square quantum well, using the band offsets,
effective masses, and well thickness as input parameters.[6] The thicknesses of the
barrier layers do not influence the confined state energies significantly.

Figure 4.2 is a typical photoluminescence spectrum taken from a sample with
a 58 A GaAs quantum well.

The peak of the spectrum is centered at 1.624 eV

(7633 A), 105 meV above the GaAs energy gap. The value of E 0 calculated for

this structure is 1.648 eV. The width of the photoluminescence peak in Fig. 4.2 can
be used as a standard for interface abruptness. A change in the GaAs quantum
well thickness of 1 monola.yer would result in an energy shift of 9 me V for the

122

Quantum Well Photoluminescence
T = 5 K

,,-...
Cl)

:!::
C:

:::::s

.ci
I,._

...__,
>.

..l,J

·oo

(I.)
..l,J

7550

7600

7650

7700

7750

Wavelength (Angstroms)
Figure 4.2: Photoluminescence spectrum of a. single 58 A Ga.As quantum well at

5 K. The excitation source was set at 5145 A with an intensity of 1 mW /cm 2 • The
spectrum shows a narrow peak at 7633 A (1.624 eV).

123

structure discussed here. Thus, the width of the photoluminescence peak can be
an indication of the layer thickness fluctuations in the lateral area covered by
the incident laser spot. The full width at half maximum of the peak in Fig. 4.2

is 4.7 meV (22 A), consistent with fluctuations of one monolayer or less in the
thickness of the quantum well.

4.2.5

Modulation doped GaAs layers

As discussed briefly in chapter 1, a free electron distribution can be placed
in an undoped GaAs layer by growing a heavily doped n-type Al 2 Ga1-:i:As layer
within a distance of 200 A. These modulation doped GaAs layers have high electron
mobilities for lateral transport, due to the spatial separation of electrons from

ionized impurities. High electron mobility transistors (HEMTs) are based upon
this concept.
Two different HEMT geometries exist.[3] 'Normal' structures begin with undoped GaAs on insulating GaAs substrates, followed by an undoped Al:i:Ga1 _ 111 As
spacer and the heavily doped Ala1Ga1_ 2 As layer. Electron mobilities greater than

106 cm 2 /V-s have been reported for these structures.[7] 'Inverted' HEMTs have

the undoped GaAs grown on top of the Al2 Ga 1 -:i:As layers. These structures generally yield lower mobilities than the 'normal' versions because Al 31 Ga 1 _:i:As tends to
outdiffuse impurities to the surface (including dopants). These impurities provide
scattering centers for the free electrons, reducing mobilities. Recently, superlat-

tice buffer schemes have been shown to greatly reduce this problem by providing
impurity-gettering interfaces in the large band gap (Al containing) material.[8]
We have grown both HEMT geometries as a means for checking material and
interface quality. A home built Hall effect system has been constructed and used to
measure mobilities and carrier concentrations in the HEMTs. The van der Pauw
method has been used to make the Hall measurements (four corner contacts). A

124
magnetic field of 4000 Gauss was provided by two permanent magnets on opposite
sides of the sample holder.* For both structures, we have obtained mobilities which
are cousidera.bly lower than the best reported values ( approximately one order of

magnitude lower), even under good system conditions. The low mobilities are
probably due to nonoptimized structure parameters and relatively high background
impurity levels. It should be noted that the MBE systems which produce the
highest HEMT mobilities are generally dedicated to these structures.

4.2.6

Characterization of Thick GaAs Films

We have characterized thick Ga.As films(~ 1 µm) by Hall effect measurements
and photoluminescence. These characterizations can be used as reference points
for the status of the III-V chamber at any given time.
Thick Ga.As layers for Hall effect measuremeuts have usually been lightly doped
n-type to avoid total depletion of the material by the surface potential. A carrier
concentration of~ 1 x 1016 cm- 3 is obtained reproducibly for a GaAs growth rate
of 1 µm/hour and a silicon oven temperature of 950°C. The measured mobility for
these lightly doped films is typically 4000 (15,000) cm2 /V-s at 300 (77) K, which
is comparable with the best reported mobility for MBE grown films.[9]
Photoluminescence spectra from bulk Ga.As films typically show two major
peaks. The higher energy peak is near the band gap of Ga.As, and is attributable
to free exciton luminescence. The lower energy peak has been identified as an
exciton bound to the carbon acceptor level.[9]
• An electromagnet has :recently been installed to :replace the permanent magnetis.

125

4.3

Determination of the GaAs/ AlAs Valence
Band Offset

This section presents results from a study of the GaAs/ AlAs valence band offset
by x-ray photoemission spectroscopy. The independence of the band offset with
respect to growth sequence (commutativity) is verified.

4.3.1

Motivation and Background

Band offsets are extremely important in the design of most semiconductor heterostructure devices, because of their impact on potential profiles in the structures.
In many devices, predicted behavior is drastically changed by even small changes
in the band offsets. Furthermore, band offsets are physically interesting because
of their fundamental nature.
Although the GaAs/ AlAs material system is the most extensively studied of
all of the heterojunctions, published experimental results for the GaAs/ AlAs valence hand offset, A.E,,, vary significantly. Even recent experimental papers have
reported a range of values from 0.36 eV to 0.55 eV for A.E11 .(10,ll,12,13,14] Furthermore, the commmutativity of the band offset has been an unresolved experimental issue.[10,11]

4.3.2

X-ray photoemission spectroscopy

X-ray photoemission spectroscopy (XPS) has become a well-established technique for measuring valence band offsets over the last few years. Several heterojunction material systems have been studied, including GaAs/ AlAs,[10,11]
HgTe/CdTe,[15] InAs/GaSb,[16] and GaSb/ AlSb.[17] The applicability of XPS
to band offset measurements is derived from the surface sensitivity of the tech-

nique. A typical escape depth for photoemitted electrons is 25 A. It is usually

126
straightforward to maintain a constant potential profile within this distance of a
heterojunction interface (i.e., band bending effects can be made negligible over this
distance).

The XPS band offset measurement is usually performed on a heterojunction
interface which is near the surface of a sample so that the photoemitted electrons
originate from regions near the interface. Our MBE system is ideal for these measurements because heterojunction samples can be transferred to the ESCA chamber for XPS analysis without removing them from ultrahigh vacuum conditions.
Thus, contaminants are not introduced on the surfaces of the samples, eliminating
the possibility of potential changes due to chemical bonding of the surface atoms.
We have found that Ga.As and AlAs samples can be left in ultrahigh vacuum conditions for 48 hours before surface contaminants can be detected by Auger electron
spectroscopy.
Three different samples are required for an XPS measurement of the valence
band offset between two semiconductors, A and B. The photoemission spectrum of
a bulk sample of semiconductor A is used to obtain the energy separation between
its valence band edge and a convenient core level, E:, - EfOf'e. A similar sample
of sem1cnnductor B is used to obtain the energy separation between its valence
band edge and a core level, E;l - E~Of'e' Finally, a sample with a heterojunction
interface between semiconductors A and B near the surface is scanned to obtain

Ei,,.e - E!.e• These three energy separations then give the valence band offset,
E:,-E:.

4.3.3

Samples

Four types of samples were grown in the III-V chamber for the XPS measurement of the Ga.As/ AlAs valence band offset. Fig. 4.3 depicts these four sample
geometries schematically.

All of the samples were grown on conductive Ga.As

127

GaAs

AIAs

0.2µm

GaAs

1.5µm

1µm

Substrate

GaAs Bulk

AlAs Bulk

AIAs

GaAs

25A

GaAs

25A

AIAs

200A

GaAs

0.5µm

0.1µm

Substrate

AlAs/GaAs

GaAs/AlAs

Figure 4.3: Schematic diagrams of the growth sequences used to grow the four

different types of samples required for the XPS band offset measurement. The
upper left (right) diagram depicts the bulk GaAs (AlAs) sample. The two types
of heterojunction samples, AlAs/GaAs and GaAs/ AlAs, are depicted in the lower
left and lower right corners, respectively.

128
(100) substrates at 600°0. The substrates were cleaned prior to growth by the
procedure described in section 4.2.1. (As )/(group III) flux ratios and growth rates
were selected by the method discussed in section 4.2.2. All of the samples were

lightly doped n-type with Si (n ~ 1 x 1016 cm- 3 ) to avoid sample charging effects.
Heavy doping was avoided because short surface depletion lengths can result in a
significant amount of band bending near the surface.
Thick GaAs layers ( >1 µm) were grown for measurements of the energy separation between the valence band edge and the Ga3d core level in Ga.As. Thick

AlAs layers (>2000 A) were used to obtain the energy spacing between the valence
band edge and the Al2p core level in AlAs. To check the commutativity of the
band offset, two types of heterojunction samples were grown: thin AlAs {25 A) on
thick Ga.As (1000 A) and thin Ga.As (25 A) on thick AlAs (100, 200, or 500 A).
These heterojunctions were used to measure the energy separation between the

Ga3d level in Ga.As and the Al2p level in AlAs.

4.3.4

Results

XPS measurements on GaAs and AlAs bulk films yielded
{4.2)
and

Eif/; - E11A" = 72.71 ± 0.04 eV.

(4.3)

For the AlAs on Ga.As heterojunction, the core level separation was found to be:

Ejff; - E3:fJ = 54.43 ± 0.02 eV.

( 4.4)

It follows that the valence band offset for AlAs on GaAs is 0.45 ± 0.07 eV. The
GaAs on AlAs heterojunction yielded a core level separation:

Eift; - Eg:fJ = 54.45 ± 0.02 eV,

{4.5)

129
leading to a value of 0.47 ± 0.07 for the valence band offset. These results indicate
that the band offset is commutative within experimental uncertainties, as expected
for ideal interfaces. The measured GaAs/ AlAs valence band offset is 0.46 ± 0.07
eV. This value of the band offset is in agreement with most of the recent published
results for the GaAs/ AlAs (100) interface.

4.4

Tunneling Times in GaAs/ AlAs Double Barrier Heterostructures

This section contains results from an investigation of electron tunneling times in
double barrier heterostructures by photoluminescence excitation correlation spectroscopy.

4.4.1

Motivation and Background

One of the major reasons for the current interest in double barrier tunnel structures is their potential for high speed applications. Due to the small charateristic
dimensions of these structures, it is possible for charge carriers to traverse the
active regions of the devices in very short times. As was discussed in section 1.3,
tunneling times have been the subject of much debate for several years. Many different theoretical approaches have been applied to this problem, with conflicting
results.[18,19,20,21,22,23,24] However, most of the theoretical predictions agree
that the theoretical limit is shorter than 50 ps and longer than 1 fs.
Experimental measurements of tunneling times are difficult because the time

scales are shorter than those which can be accessed by conventional electronic
means. Hence, optical excitation and sampling techniques are usually needed. In
fa.ct, most of the published experimental measurements of double barrier response
times have been either optical or optoelectronic.[25,26,27,28)

130

4.4.2

Measurement Technique

The photoluminescence (PL) intensity, Ip1, from the quantum well of a double
barrier heterostructure is given by:
]pl oc n X p,

(4.6)

where n and p are the number of electrons and holes, respectively, in the quantum
well. If electrons and holes are placed in the quantum well by a short excitation

pulse ( at an energy above the quantum well peak energy), then the PL peak

intensity will decay with time as the carriers escape from the well. Hence, "timeresolved" photoluminescence can be used to measure tunneling rates out of the
quantum well when tunneling is the dominant escape mechanism for the carriers.
The most obvious technique for performing time-resolved PL is to generate
carriers with short la.ser pulses and detect the luminescence with a very fast detec-

tor. This method has been recently demonstrated to yield a reasonable tunneling
time measurement.[26] However, signal-to-noise problems and detector (streak
camera) response limitations prevented measurement of tunneling times shorter
than 60 ps.
We have used photoluminescence erdtation correlation spectroscopy (PECS)

to measure tunneling times in GaAs/ AlAs double barrier heterostructures with
varying barrier thicknesses. The PECS technique has been described in detail
elsewhere.(29] A very brief description is given here. A colliding pulse modelocked ring dye laser is used to generate a train of very short pulses (200 fs full
width at half maximum). This beam is then split into two pulse trains. Next,
one set of pulses is forced to traverse an extra distance, delaying it with respect
to the other pulse train by a short time, "Y (-500 ps :5 "Y :5 500 ps ). The two
beams are then chopped at different frequencies, Ji and h, and focused down to a
spot on the sample surface. Finally, the photoluminescence signal from the sample

131
is synchronously detected through a. lock-in amplifier set to the sum of the two
chopping frequencies, Ji + f2. It can he shown that the the photoluminescence
signal at the sum frequency decays with increasing 'Y exponentially, with a time
constant equal to the characteristic carrier escape time.

4.4.3

Samples

The growth sequence used to produce samples for the tunneling time vs. barrier
thickness investigation is depicted schematically in Fig. 4.4.

All of the samples

were grown on Ga.As (100) substrates at 600°C. The substrates were cleaned prior
tc-,growth by the procedure described in section 4.2.1. (As)/(group III) flux ratios
and growth rates were selected by the method discussed in section 4.2.2. None of
the samples were intentionally doped.
Growth commenced with a 0.5 µm GaAs layer, followed by a superlattice buffer
layer consisting of five periods of Al 0 ,3 5 Gao.esAs (50 A) and GaAs (500 A).• A
0. 7 µm Ga.As layer was grown next to eliminate any optical effects from the super-

lattice. The GaAs/ AlAs double barrier structure was then grown symmetrically,
with a GaAs well thickness of 58 A. Seven different samples were studied, with
AlAs barrier thicknesses of 16,22,28,34,34,48, and 62 A. Finally, a 300 A GaAs cap
layer was grown. This thickness was sufficient to prevent quantum confinement
effects in the cap layer, while allowing optical probing of the quantum well.
Barrier and well thicknesses were determined from bulk growth rates. However,
high-resolution transmission electron microscopy (TEM) was used to measure the
layer thicknesses in the 16 A barrier sample and one of the 34 A barrier samples.
Fig. 4.5 is a high-resolution TEM print of the double barrier region in the 34 A
barrier sample.

The TEM print confirms the 34 A AIAs barrier thickness, with

•The intention in growing this buffer layer was to improve material quality. However, we have
:f'ound no difference in experimental results with the supe:rln.ttice deleted.

132

Ga
0,35

GaAs

300 A

AIAs

GaAs

58 A

AIAs

GaAs

0.7 µm

As/GaAs superlattice
0.65

GaAs

50 A/500 A
5 periods

0.5 µm

Substrate

Figure 4.4: Schematic diagram depicting the growth sequence used to produce
double barrier heterostructures for the optical tunneling time measurement. The
AlAs barrier thickness, LB, was varied, with all other growth parameters held
constant.

133

Figure 4.5: High-resolution TEM photograph of the double barrier region of a
sample with 34 A AlAs barriers and a 58 A GaAs quantum well. The layer thicknesses can be obtained by counting monolayers, to an uncertainty of one monolayer
1>er interface.

134

an uncertainty of two monolayers.[30] Possible lateral fluctuations in layer thicknesses have been investigated by checking the linewidths displayed in one beam
photoluminescence. The resulting quantum well peaks are narrow (~ 6 meV),
consistent with fluctuations of 1 monolayer or less in layer thickness.
A heterostructure with symmetric barriers consisting of three 8.5 A AlAs layers,

separated by two 8.5 A GaAs layers has also been grown for the PECS experiment.
The structure was grown with a 49 A GaAs quantum well, and a 300 A GaAs cap
layer. This double barrier geometry has been shown to yield the highest reported
peak-to-valley ratio for GaAs/ Ala:Ga1 _ 111 As resonant tunneling structures.[4]

4.4.4

Results

Each of the samples discussed in the previous section displayed a clear photoluminescence peak attributable to the lowest energy confined state in the quantum

well. For the thinnest AlAs barriers, the photoluminescence intensity was relatively weak due to the rapid escape of tunneling electrons out of the well. For each
of the samples with varying AlAs barrier thicknesses, scans were made of the sum
frequency photoluminescence intensity, I.um, as a function of delay time, --y. These
scans showed a simple decaying exponential dependence of J,,.,,m. with increasing
--y. The characteristic decay times of Iaum were identified as the times required for

electrons to tunnel out of the quantum wells.
For the sample with the thinnest AlAs barriers (16 A), the tunneling time was
measured to be ~ 12 ps, the shortest such time ever reported. The tunneling
time for the samples with 34 A barriers was measured to be :::::: 800 ps. A simple
exponential dependence of tunneling time on barrier thickness was observed for
samples with barrier thicknesses between 16 and 34 A. This result is consistent
with a straightforward method of calculating tunneling times for double barrier
structures. In this approach, the electron tunneling time, r, is related to the

135
width of the resonance in the transmission coefficient, r, through the uncertainty
principle:
T=

r•Ii

(4.7)

This method is particularly attractive because the transmission coefficient for double barriers is well known and easily calculated.
The sample with barriers consisting of three 8.5 A AlAs layers separated by two
8.5 A Ga.As layers was found to display an electron tunneling time of 350 ± 60 ps
out of the 49 A quantum well.

4.4.5

Ongoing Experiments

Further PECS experiments on Ga.As/ AlAs double barrier heterostructures are
currently being pursued. Doped structures have been grown and prepared as two
terminal electrical devices with thin (60 A) Au/Ge contacts on the tops of the
mesas. These devices can be biased into the NDR region, and probed optically to
measure tunneling times under conditions in which significant tunneling current is
present.
A set of undoped structures with progressively narrower Ga.As quantum wells
has been grown. As the wells become thinner, it is expected that the quantum well

conduction band states will rise in energy until they are comparable to the AlAs
X-point energy. At this point, significant mixing between the quantum well state
and the AlAs X-point state is expected. The onset of this mixing should cause a
sharp transition in the observed photoluminescence decay behavior.

136

4.5

Electron Coherence in Resonant Tunneling
Structures

This section presents a study of electron tunneling in triple barrier heterostructures. The results provide evidence for a coherent model ( as opposed to a sequential
model) of electron tunneling through thin AlAs barriers.

4.5.1

Motivation and Background

Double barrier tunnel structures are the subject of a considerable amount of
current research. The first proposal of negative differential resistance (NDR) in
these structures was based on a model of tunneling in which the electron wavefunctions are coherent across then entire structure.[31) In this context, coherent
is taken tu mean that the electrons retain their phase information throughout the
tunneling process. In more intuitively meaningful terms, coherent electrons are
not scattered as they tunnel through the double barrier structure. In the coherent picture of double barrier tunneling, the electron transmission coefficient has a
resonance, similar to the Fabry-Perot effect for optical waves.
Recently, Luryif23) has proposed that a sequential tunneling model can ade-

quately explain the observation of NDR in double barriers. The basis of Luryi's
argument is that electron tunneling from a 3-dimensional set of states to a 2dimensional set of states will always show NDR if the total energy and parallel
wavevector of each tunneling electron is conserved. Under these conditions, NDR
occurs when the conduction band edge in the negatively biased electrode is raised
above the 2-dimensional subband minimum in the quantum well by an applied
voltage. The basic difference between the two tunneling models is that the electrodes and quantum well are treated as independent sets of states in the sequential
model. In the coherent model, the electron states are extended throughout the dif-

137
ferent regions of the heterostructure. The two models are thought to be virtually
indistinguishable in double barrier I-V curves.
We have grown triple barrier heterostructures in an attempt to resolve the coherent vs. sequential tunneling issue. Electrons tunneling through these structures
must pass between two quantum wells. Hence, sequential tunneling conditions become more stringent, because the electrons must hop from one set of 2-dimensional
states to another. In fact, if the different regions of the triple barrier heterostructure are treated independently, tunneling currents can only become significant
when a. subband minimum in the first quantum well aligns precisely with a subband minimum in the second well. Thus, we expect NDR to be difficult to achieve
if electron tunneling between the two wells is sequential in nature. In contrast, the
coherent tunneling model allows the electron wavefunctions to penetrate from one
quantum well to the next, so that NDR behavior should be similar to the double
barrier case (provided that the middle barrier is sufficiently thin).

4.5.2

Samples

Figure 4.6 is a schematic diagram of the growth sequence used for the triple
barrier heterostructure samples.

All of the samples were grown on conductive

Ga.As (100) substrates at 600°0. Growth commenced with a 0.5 µm heavily doped
n-type GaAs layer. Next, a 500 A lightly doped (n = 2 x 1016 cm- 3 ) Ga.As spacer
layer was grown, followed by a 25 A undoped Ga.As spacer. The symmetric triple
barrier portion of the structure was then grown; it consisted of 30 A AlAs outer
barrier layers, 54 A Ga.As quantum wells, and an AlAs middle barrier with a
variable thickness. Undoped (25A) and lightly doped (500 A) Ga.As spacer layers
wel'e grown on top of the triple barrier region. Finally, a 2500 A heavily doped

n-type Ga.As cap layer was deposited. Four samples were studied, with middle
barrier thicknesses of O, 3, 6, and 12 monolayers.

138

Triple Barrier Heterostructure
GaAs, n=Sx 10 18 cm- 3

0.25 µm

GaAs, n=2x 10 16 cm- 3

500 A

GaAs, undoped

25 A

AIAs,
GaAs,

AIAs,
GaAs,
AIAs,
GaAs,

,,
,,
,,

30 A
54 A

54 A

30 A
25 A

GaAs, n=2x 10 16cm- 3

500 A

GaAs, n=Sx 10 18 cm- 3

0.5 µm

Substrate

Figure 4.6: Schematic diagram depicting the growth sequence used to produce

GaAs/ AlAs triple barrier tunnel structures. The middle barrier thickness, W, was
varied, with all other growth parameters held constant.

139
The samples were prepared as two terminal electrical devices by standard photolithographic and wet etching techniques. Wire bonds were attached to the devices
for liquid nitrogen temperature experiments.

4.5.3

Results

All four of the samples studied displayed I-V curves with NDR, and the three
thinnest middle barrier (0 13, and 6 monolayers) samples had multiple resonances.
These results suggest that the electron wavefunctions do penetrate across the mid-

dle barrier region, and are inconsistent with the sequential picture of tunneling as
proposed by Luryi. Furthermore, the resonances are extremely strong. Fig. 4. 7 is
an I-V curve taken from the sample with a 3 monolayer barrier at 77 K. The curve
displays three distinct NDR regions, with a peak-to-valley current ratio of 19.3:1
for the second peak. This is comparable to the largest peak-to-valley ratio ever
reported for a GaAs/ AlAs structure,[4] and is higher than any of those reported
for conventional GaAs/ AlAs double barrier heterostructures. These results clearly
indicate that the electron wavefunction is coherent across the middle barrier layer.
The samples with 3 and 6 monolayer middle barriers both show stronger NDR
behavior than the sample with no middle barrier. However, the sample with a
12 monolayer middle barrier shows only one degraded NDR region in its 1-V curve.
This poor performance may indicate a loss of coherence across the thicker middle
barrier. A possible explanation for this result is the longer tunneling time expected
for tunneling through the thicker barrier. Longer time scales for tunneling may
make scattering more likely, reducing the probability of coherent transport.

140

225.-----------~---------,
(C)

3 M.L. Middle Barrier (77K)

--<(

+-'

Q)
,_
,_

::J

()

0 1--..-:c:::__...i..-_---1-__,_--..L..-----L--..___--l--_........__--,A
2.25

Voltage (Volts)

Figure 4.7: 1-V curve from a GaAs/AlAs triple barrier tunnel structure, taken at
77 K. The AlAs middle barrier thickness is 3 monolayers. The curve displays three
distinct NDR regions, with a peak-to-valley current ratio of 19.3:1 for the second
NDR region.

141

4.6

Growth of InAs on GaAs substrates

This section presents the development of a method for growing high quality
thick InAs films on GaAs substrates.

4.6.1

Motivation and Background

Combinations of the nearly lattice-matched semiconductors InAs, GaSb, and
AlSb are promising for a number of interesting heterostructures. Some of these
structures have already been realized, as discussed previously in section 4.1.2.
However, lattice-matched substrates for these materials are a problem. Even the
highest quality InAs substrates have etch pit densities (dislocations) that are more
than ten times higher than those of standard Ga.As substrates. Reasonably high

quality GaSb substrates can be obtained, but cleaning procedures for these substrates have not been well studied. Furthermore, both InAs and GaSb are considerably more expensive than GaAs (by a factor of approximately 6). AlSb is not
stable to atmospheric exposure. Hence, a technique for depositing a high quality
buffer layer of InAs, GaSb, or AlSb on GaAs is highly desirable.

The difficulty with growing a thick buffer layer on a poorly lattice matched
substrate is that relaxation of the buffer material to its natural lattice constant
occurs through dislocation formation. These dislocations tend to thread through
the entire buffer layer, with adverse effects on optical and electrical properties.
Recently, it has been reported that a short period strain ]aye:rerl snperlattice can

reduce the number of dislocations in extremely thick InAs films (6 µm) grown on
GaAs substrates.[32] The dislocation densities were inferred from measurements
of electron mobilities. However, MBE growth rates are usually limited to about
1 µm/hour, so that extremely thick buffers are impractical. We have attempted
to obtain electron mobilities comparable to those reported by Kalem et al.[32]

142

for reasonable InAs buffer layer thicknesses (2 µm) on GaAs. Growth parameters
such as substrate temperature, As/In flux, and superlattice buffer layers have been
varied, with in situ monitoring by RHEED, optical pyrometry, and residual gas

analysis (RGA).

4.6.2

Growth and In Situ Analysis

Figure 4.8 is a schematic layer diagram of the growth sequence used to deposit
InAs layers on GaAs substrates. All of the samples were grown on insulating Ga.As
(100) substrates. The substrates were cleaned prior to growth by the procedure
described in section 4.2.1. Initially, a 0.5 µm undoped GaAs layer was deposited at
a substrate temperature of 600°C. The growth was then interrupted {As flux only)
while the substrate temperature was lowered to that desired for the InAs layer.
Two different superlattice buffers at the GaAs/InAs interface were studied. Some

of the samples had a five period In0.1Gao.3 As/GaAs (2 monolayer/2 monolayer)
superlattice only. Other samples had an additional five period InAs/Ino.1Ga0 •3 As
(100 A/6 A) superlattice grown immediately after the first superlattice. Finally, a
2 µ,m undoped lnAs layer was grown.
Several different substrate temperatures were used for the InAs layers, spanning the range 490-550° C as measured by the optical pyrometer. The RHEED
pattern from the InAs surface was observed to indicate a transition from a 2 X 4

reconstructed surface to a 4 x 2 reconstructed surface at a substrate temperature of
530-535°C. This transition is believed to be due to a change from an As-stabilized
surface at low temperatures to an In-stabilized surface at high temperatures. The
As2/In flux ratio was estimated to be between 10:1 and 20:1 for all of the samples
by measuring RGA peak heights.

143

lnAs

lnAs/ln Ga As superlattice (opt.)
0.7

0.3

In Ga As/GaAs superlattice
0.7

0.3

GaAs

2.0 µm

100 A/6 A
5 periods

6 A/6 A
5 periods

0.5 µm

GaAs Substrate

Figure 4.8: Schematic diagram depicting the growth sequence used to deposit thick

InAs layers on GaAs. The second (InAs/In 0 .1Gao. 3 As) superlattice was used for a
few of the samples.

144

InAs RHEED oscillations
RHEED intensity oscillations have been observed for InAs layers grown under
As-stabilized conditions. These oscillations in diirraction intensity are believed to

be caused by periodic surface roughness changes which occur during deposition
of each monolayer of material.[33] Thus, the oscillations can be used to calibrate
growth rates for thin layers. Furthermore, strong oscillations are usually indicative
of good layer-by-layer growth. Fig. 4.9 shows the RHEED intensity oscillations
observed during growth of an InAs layer.

The RHEED intensity was measured

by placing one end of a thick fiber optic cable on the phosphor RHEED screen,
and directing the other end of the cable to a photodiode. A chart recorder was
used to record the signal from the photodiode as a function of time. The data in
Fig. 4.9 were obtained by interrupting the InAs growth for one minute after 0.5 µm
of material had been deposited. The specular spot in the 2-fold RHEED pattern
([110]-azimuth) was chosen as the point seen by the fiber optic cable. Over twenty
oscillations can be seen in Fig. 4.9, indicating a growth rate of 3.17 A/s. To our
knowledge, these are the most intense RHEED oscillations ever reported for InAs
growth.[34]

4.6.3

Characterization

All of the InAs samples grown under As-stabilized conditions had good surface
morphology, though not quite as good as is usually obtained for high quality GaAs
growth. Samples grown under In-stabilized conditions had rough hazy surfaces.
Hall effect measurements were made at 77 and 300 K on all of the InAs layers.
Van der Pauw (four corner) contacts were made by evaporating Au/Ge through
a foil mask. The homemade Hall apparatus was calibrated against results from a
Hall effect experiment at Hughes Research Labs. All of the samples were found
to be n-type, with background carrier concentrations in the 1016 cm- 3 range.

145

( /)

."t:

..0
,._
co

."t:

(/)

Q.)
......

Time (seconds)
Figure 4.9; RHEED intensity oscillations observed during the growth of InAs. The
oscillations were obtained by interrupting the growth (As flux only) after 0.5 µm
of InAs had been deposited.

146
The highest room temperature electron mobility observed was 18,900 cm 2 /V-s,
comparable to the results obtained by Kalem et al. on much thicker (6 µm) layers.
The best 300 K mobilities were obtained from samples grown under As-stabilized

conditions with only one short period superlattice interface layer. We believe that
the lowest number of dislocations in the InAs film is acheived under these growth
conditions. In contrast, samples grown under Ga-stabilized conditions yielded
the best 77 K mobilities (35,500 cm 2 /V-s), independent of whether one or two
superlattice interface layers were used. The 77 K results can be explained by
measured background carrier concentrations, which were found to be lower for
samples grown under Ga-stabilized conditions by a factor of four. The difference
in background doping is probably attributable to lower impurity incorporation
during growth under Ga-stabilized conditions.
X-ray diffraction was performed on an InAs sample grown under As-stabilized
conditions with two superlattice interface layers. The lattice constant of the film
was found to be 6.08

0.04 A. This result is close to the bulk InAs lattice con-

stant of 6.06 A, and indicates that the lnAs film is almost completely relaxed.
However, the x-ray diffraction peak was found to be extremely broad for a 2 µm
film, indicating a variation in lattice constant. The broadening may be caused by
the portion of the InAs layer near the GaAs interface, which is probably heavily
dislocated.

4. 7

Summary of Conclusions

A number of standard GaAs/ Ala:Ga1 -a:As growths have been characterized,
including double barrier tunnel structures, single quantum wells, HEMT's, and
bulk GaAs. These structures have been used as reference points for the status
of the III-V chamber at various times. Under good system conditions, most of

147
the standard structures yield results comparable with the best results reported for
MBE growth.
A measurement of the GaAs/ AlAs valence band offset has been made by x-ray
photoemission spectroscopy (XPS). Samples for this experiment were produced in
the III-V growth chamber, and transferred under ultra.high vacuum to the surface

analysis chamber. This arrangement is ideal for the surface sensitive XPS technique
because the sample surface is not contaminated by exposure to atmosphere. The
measured Ga.As/ AlAs valence band offset is commutative, with a value of 0.46 ±
0.07 eV.
Double barrier heterostructures have been produced in the III-V chamber £or

a measurement of electron tunneling times by photoluminescence excitation correlation spectroscopy. The structures were grown with thin (300 A) GaAs cap
layers to permit optical probing of the quantum well region. Decay times as short
as 12 ps have been measured for electrons escaping from a 58 A GaAs quantum
well surrounded by two 16 A AlAs barrier layers. The electron tunneling times are
found to depend exponentially upon barrier thickness, in good agreement with a
calculation based on resonance widths in the electron transmission coefficient.
A study of coherent vs. sequential tunneling in triple barrier Ga.As/ AlAs heterostructures has been performed. For thin AlAs middle barriers ( <12 monolayers), these structures yield I-V curves with multiple strong NDR regions. Particularly large peak-to-valley current ratios are observed (19.3:1 at 77 K), indicating
that the electron wa.vefunctions a.re coherent a.cross the middle barrier. A loss of

coherence for thick AlAs middle barriers is suggested.
A method for growing high quality 2 µm InAs layers on Ga.As substrates has
been developed.

Good results are obtained by inserting a short period strain

layered superlattice at the InAs/GaAs interface. A careful optimization of the
InAs growth parameters has also been performed. Intense RHEED oscillations

148
have been observed for growth under As-stabilized conditions. The resulting InAs
layers have electron mobilities comparable to the best ever reported. This method
for depositing InAs buffer layers on GaAs has since been used in growths of novel

heterostructures from combinations of the nearly lattice-matched materials InAs,
GaSb, and AlSb.

149

References
[1] H. Munekata, T.P. Smith, and L.L. Chang, J. Vac. Sci. Technol. B, to be
published.

[2] J.D. Grange, in The Technology and Physics of Molecular Beam Epitazy
(Plenum Press, New York, 1985), Chapter 3.

[3] H. Morko~, in The Technology and Physics of Molecular Beam Epitazy
(Plenum Press, New York, 1985), Chapter 7.
[4] C.I. Huang, M.J. Paulus, C.A. Bozada, S.C. Dudley, K.R. Evans, C.E. Stutz,

R.L. Jones, and M.E. Cheney, Appl. Phys. Lett. 51, 121 (1987).
[5] J.R. Soderstrom and T.G. Andersson, Superlattices and Microstructures 5,
109 (1989).

[6] A.R. Bonnefoi, Ph.D. Thesis, California Institute of Technology, 1986.

(7] J.H. English, A.C. Gossard, H.L. Stormer, and K.W. Baldwin, Appl. Phys.
Lett. 50, 1825 (1987).
[8] 'I'. Sajoto, M. Santos, J.J. Heremans, M. Shayegan, M. Heiblum, M.V. Weckwerth, and U. Meirav, Appl. Phys. Lett. 54, 840 (1989).
[9] M. Ilegems, in The Technology and Physics of Molecular Beam Epitazy
(Plenum Press, New York, 1985), Chapter 5.

150
[10] A.D. Katnani and R.S. Bauer, Phys. Rev. B 33, 1106 (1986).
[11] J.R. Waldrop, R.W. Grant, and E.A. Kraut, J. Vac. Sci. Technol. B 5, 1209
(1987).

[12] G. Duggan, J. Vac. Sci. Technol. B 3, 1224 (1985).
[13] B.A. Wilson, P. Dawson, C:W. Tu, and R.C. Miller, J. Vac. Sci. Technol. B

4, 1037 (1986).
[14] W.I. Wang and F. Stern, J. Vac. Sci. Technol. B 3, 1280 (1985).

[15] S.P. Kowalczyk, J.T. Cheung, E.A. Kraut, and R.W. Grant, Phys. Rev. Lett.

56, 1605 (1985).
[16] G.J. Gualtieri, G.P. Schwartz, R.G. Nuzzo, R.J. Malik, and J.F. Walker, J.

Appl. Phys. 61, 5337 (1987).
[17] G.J. Gualtieri, G.P. Schwartz, R.G. Nuzzo, and W.A. Sunder, Appl. Phys.

Lett. 49, 1037 (1986).
[18] K.K. Thornber, T.C. McGill, and C.A. Mead, J. Appl. Phys. 38, 2384 (1967).
[19) M. Biittiker and R. Landauer, Phys. Rev. Lett. 49, 1739 (1982).

[20] N. Harada and S. Kuroda, Jpn. J. Appl. Phys. Pt.2 25, 1871 (1986).
[21] S. Collins, D. Lowe, and J.R. Barker, J. Phys. C 20, 6213 (1987).

[22] H. Guo, K. Iliff, G. Neofotistos, and J.D. Gunton, Appl. Phys. Lett. 53, 131
(1988).
[23) S. Luryi, Appl. Phys. Lett. 47, 490 (1985).

[24] S.C. Kan and A. Yariv, J. Appl. PbyD, 64, 3098 (1988).

151
(25] J. F. Whitaker, G.A. Mourou, T.C.L.G. Sollner, and W.D. Goodhue, Appl.
Phys. Lett. 53, 385 {1988).
[26] M. Tsuchiya., T. Ma.tsusuc, and H. Sa.ka.ki, Phys. Rev. Lett. 50, 2356 (1087).

(27] M.K. Jackson, M.B. Johnson, D.R. Chow, T.C. McGill, and C.W. Nieh, Appl.

Phys. Lett. 54, 552 (1989).
[28] T.B. Norris, X.J. Song, W.J. Schaff, L.F. Eastman, G. Wicks, and G.A.
Mourou, Appl. Phys. Lett. 54, 60 (1989).
[29] M.B. Johnson, T.C. McGill, and A.T. Hunter, J. Appl. Phys. 63, 2077 (1988).
(30] C.W. Nieh, private communication.
(31] R. Tsu and L. Esaki, Appl. Phys. Lett. 22, 562 (1972).
(32] S. Kalem, J.I. Chyi, H. Morko-;;, R. Bean, and K. Zanio, Appl. Phys. Lett. 53,
1647 (1988).
(33] R. Ludeke, R.M. King, and E.H.C. Parker, in The Technology and Physics of
Molecular Beam Epitazy (Plenum Press, New York, 1985), Chapter 16.
(34] B.F. Lewis, R. Fernandez, A. Madhukar, and F.J. Grunthaner, J. Vac. Sci.

Tech. B 4, 560 (1986).

1.'i2

Appendix A
MBE System
A.1

Introduction and Outline

Figure A.1 is a schematic diagram of the McGill group MBE system. The figure is identical to Fig. 1.4, but is shown again here for convenience. As discussed
previously, the MBE system was designed to have a high degree of flexibility; in
terms of both the types of semiconductor heterostructures which can be grown, and
the different experiments which can be performed. Although ea.ch of the chambers

on the system are adaptations of standard Perkin Elmer equipment, integration
of the different chambers required a number of special design considerations. The
purpose of this appendix is to document these considerations, in addition to modifications of the standard Perkin Elmer chambers which have been made. Features
of the chambers which are considered to be standard are not discussed here, as

they are well-documented in the system manuals.
Section A.2 discusses the transfer of samples throughout the MBE system.
Modifications made to the III-V chamber are documented in section A.3. Some
practical considerations for III-V source loading are also given. Section A.4 lists
the modifications made to the standard Perkin Elmer prep/analysis module. Fi-

153
:c

....
..J
<(

':,
.::l

....uJ

:l!

iii

::E

..J

S:lltlOIH0313 39Vi IA·II

□D

SI MB E ELECTRONICS

<1.

.,;;::c::.i.....·t---------.. 5

::E

"'OJ::I!
r--")-==-------_;lii

..J

::E

,,,uJ
::I!

"'u

I=0

uJ
uJ
Ill

Figure A.1: Schematic diagram of the McGill group MBE system. The system
has separate III-V, II-VI, and group IV growth chambers, and an ESCA/Auger

system, connected by ultrahigh vacuum transfer tubes. The system also includes
a. prep/analysis module and a metallization chamber.

154
nally, modifications made to the standard Perkin Elmer metallization chamber are
documented in section A.5.

A.2

Transfer Mechanism

The MBE system is designed to accomodate wafers up to 3" in diameter. Substrates are mounted on molybdenum blocks by either In bonding or specially designed clips, depending upon which variety of block is used. Each of the molybdenum blocks has a groove ( ~ 0.5 cm wide) around its side. This feature allows
the block to be held by a transfer fork which has prongs designed to slide into the
groove.•
Each of the three transfer tubes depicted in Fig. A.1 ha.a a. transfer fork which

runs its length. The fork in the middle transfer tube (the one attached to the
ESCA and group IV chambers) has been specially designed to permit blocks to be
transferred out both ends of the tube. Four additional transfer forks are mounted
perpendicular to the two long transfer tubes, opposite the metallization, III-V,
group IV, and ESCA chambers. These forks a.re used to move blocks from the

transfer tubes into the chambers.
Parking stages have been installed at each place in the system where it is
necessary to release blocks from one transfer fork and pick them up with another
(six places, including the stage in the prep/analysis module). Blocks can be set on
and taken off of the stages with the transfer forks. In addition, each of the transfer
tubes has an elevator which can hold a cassette for storage of up to six blocks.
These eleva.tors are positioned directly beneath the three intro hatches, and a.re

the mechanism for transferring the blocks into and out of ultrahigh vacuum.
At least two gate valves must be opened in order to transfer a block from
*The blocks for the group IV growth chamber are designed differently, but are compatible
with the system transfer mechanism.

155
one chamber to another. In general, it is a good practice to open these valves
sequentially, so that only adjacent chambers are exposed to each other at any
given time. Moving a block from one end of the system to the other has been
found to take approximately 15 minutes.

A.3

111-V Chamber Modifications

The III-V chamber is basically a standard Perkin Elmer 430 system, with no
major modifications. However, some procedures not discussed in the manufacturer's manual have been developed from our experience with the system, and
should be documented. In addition, an Sb cracker made by another vendor has
been installed on our source flange.

A.3.1

Sb cracker

We have purchased and installed an EPI PE-75 cracking effusion cell, along
with a matching EPI-PSC power supply. Both the bulk evaporator and cracker
sections of the cell are water cooled, allowing the Sb evaporation temperature
{ R'j

600°C) to be accessed. The crucible for the cell consists of a cylindrical tantalum

can with a long thin tube ( cracking zone) extending out its top. The crucible can
be dismounted from the system for source loading without removing the body of

the cell from the system. A simple calculation indicates that dimers (Sb 2 ) are
produced by the cell whenever the cracker section is held above 700°C. Typical
operating conditions during growth are 625°C at the bulk evaporator section, and
850° C along the cracking zone.
Unfortunately, the EPI power supply requires 120 VAC power, and the Perkin
Elmer electronics rack is designed for mostly 220 VAC (single phase) power. It
has been determined that running the Sb cracker from the 120 VAC section of the

156

pressure interlocked power can cause the system circuit breaker to trip, resulting
in an abrupt power shutdown. Therefore, we connect the power supply to a wall
outlet in the cleanroom during growth.• When the system is idling, we connect the

EPI power supply to the pressure interlocked power so that it will be safeguarded
in the event of vacuum loss.

A.3.2

Source Loading Information

Although the lll-V source flange has eight ports, several of the source ovens
have loading restrictions which limit the possible arrangements of sources. These
restrictions are listed here.
I. The Sb cracker is better placed such that it tilts back slightly, since the Sb is

sometimes melted during operation and can flow out under gravity. However,
the cell will not fit in the bottom two ports of the source flange due to its
length.
2. The 60 cc In cell should not be placed in a port which has a shutter pivot
mechanism below it, because the In can flow down the shutter into the mechanism. This can result in a freezing of the shutter. The two bottom ports
on the source flange are ideal for this cell.
3. The 60 cc Ga cell should be placed such that it tilts back. The four bottom
ports of the source :flange can accomodate this cell.
4. The 60 cc Al cell should not be loaded with a large quantity of Al ( <12 g).
This will prevent cracking of the crucible provided that slow ramp downs of
the oven temperature are used.
•The wall outlet is supplied through e. 20 Amp brea.ker. The EPI powe:t supply is rated for
20 Amps maximum.

157

5. The As cracker should be placed on one of the top two ports on the source
flange. The cell can be loaded by .dismounting the bulk evaporator from the
system.

A.3.3

Liquid Nitrogen Plumbing

We have found that the III-V system can be run in a closed cycle liquid nitrogen
mode in spite of its small source flange tubing (the II-VI and group IV growth
chambers have larger diameter tubing). Our plumbing arrangement is a 'T' at the

end of the triax feed line from the liquid nitrogen phase separator, with one branch
feeding the source flange directly, and the other feeding the system cryopanels
through the Perkin Elmer manifold. The return lines from the source flange and
manifold are then merged in another 'T' with the semiflex return line to the phase
separator. We generally cool the system down to liquid nitrogen temperature with
open loop operation, then switch to closed loop once the system is cold. It is
occasionally necessary to switch to open loop operation when most of the source
ovens are hot (>900°C).

A.3.4

Water Cooling

We run each of the parts of the system which require water cooling in series.
This arrangement assures that the flow rate through each of the parts is the same.
The EPI cracking cell presents the largest resistance to flow as it has the smallest
diameter tubing. Therefore, it is best to run the supply line to the EPI cell first
to prevent building up large back pressures on the other parts of the system. We

do not run water into the bellows for the substrate heater. Perkin Elmer has
determined that water cooling of the substrate heater is unnecessary, and can lead
to direct leakage of water into the chamber if (and when) the bellows wear out.

158

A.3.5

Instrument Coating

We have found that the analysis instruments (ion gauge, RHEED gun, RGA,
flux monitor) in the III-V chamber tend to become coated due to the high As and
Sb pressures during growth. This coating eventually results in short-circuiting of

the electrical elements in these instruments. Removing the shorted elements for
cleaning requires breaking vacuum and a. bakeout, resulting in at least one week

of down time. It has been found that satisfactory functioning can be restored by
passing electrical current through the short (particularly a fairly resistive short),
reevaporating the coating material. We generally use a curve tracer for this operation, set to the minimum amount of output power necessary to remove the short.
High voltages ( up to 1.6 kV) can be used, provided that the elements are wellinsulated (other than the short). Of course, safety precautions should be exercised
when applying high voltages to prevent bodily injury and/or equipment damage.

A.4

Prep/ Analysis Module Modifications

The prep/analysis module depicted in Fig. A.1 is a modified version of a standard Perkin Elmer chamber. Standard design features include a heater stage which
can reach 1200°C, a water jacketed chamber, an ion pump, and two 8" gate valves.
The changes and additions made to the standard design are listed here.

1. An 8'' :flange has been added at the top of the chamber. The :flange is

positioned appropriately for a. Princeton Research Instruments reverse view
LEED (low energy electron diffraction) analyzer.
2. A heat shield for the LEED analyzer has been installed on a 1.33" flange.
The shield is connected to the flange through a rotary feedthrough which
allows it to be moved into and out of position.

159
3. A sputter gun port (nonstandard, 4.75" flange) has been designed for a dif-

ferentially pumped Ion Tech gun: The port is approximately 10" from the
sample holder, at an angle of 45°.
4. The docking viewport on the system is mounted on a 4.5" flange, at an angle

of 30° to the substrate.
5. An optical access port has been added at the angle of reflection to the docking
viewport (for reflection of light from the sample surface). The port has a

2. 75" flange, and could be used for a number of optical applications, such as
pyrometry, laser annealing, ellipsometry, and interferometry.
6. An ion gauge has been mounted on a. port with a. 2. 75" flange.

7. A 6" flange has been added in case a cryopump is desired.

8. A 4.5" flange for a UTI 100 residual gas analyzer has been added to the
chamber.
Drafted layouts for this chamber are stored in the file cabinet in Watson building,
room 270.

A.5

Metallization Chamber Modifications

The metallization chamber depicted in Fig. A.1 is a modified version of a stan-

dard Perkin Elmer chamber. The changes and additions made to the standard
design are listed here.
1. A 'T' has been placed between the standard CTI CT-8 cryopump and the

chamber so that an ion pump can be added at a later date.
2. A nonstandard 4.75" flange has been added for an Ion Tech sputter gun.

160

3. A 6" flange has been added at an angle of 135° from the sample introduction
port. This flange is intended to be used for attaching additional modules to
the MBE system. Alternatively, the docking viewport (90° from the intro-

duction port, 135° from the additional 6" port) could be used as a point of
connection to additional modules. The window for viewing sample docking
would then be moved to the additional 6" flange.
4. A 4.5" flange for a UTI 100 residual gas analyzer has been added at the end
of a long, shuttered tube.
5. Two 10" e-gun ports have been placed on opposite sides of the chamber.

These ports should be sufficiently large to house multiple crucible or single
crucible guns from Airco Temescal, Leybold-Herraeus, or another manufacturer.
6. Nine 2. 75" con:flats have been added at various points on the chamber for
other sources, cryopanels, crystal thickness monitors, and shutters.
Drafted layouts for this chamber arc stored in the file cabinet in Watson building,

room 270.

161

Appendix B
Cleanroorn
B.1

Purpose

The principal purpose of the McGill group cleanroom is to provide a clean,
controlled environment for molecular beam epitaxy and photolithography. Both
of these experimental facilities encompass a number of pieces of equipment and

procedures, which require varying degrees of cleanliness. Attempting to house all
of the activities under the most stringent cleanliness conditions would be very
expensive and difficult to maintain. Instead, we have separated the cleanroom into
sections, with different cleanliness requirements in each section.
Many pieces of equipment housed in the cleanroom have special utility requirements (such as liquid nitrogen). Another purpose of the cleanroom is to organize
the supply of these utilities such tha.t connections to equipment are convenient.

Furthermore, the cleanroom serves to partially isolate utility supplies from those
routed to other laboratories, reducing their vulnerability to external fluctuations.
Finally, an important purpose of the cleanroom is to enhance safety (both
for people and equipment). Some of the MBE and photolithography procedures
involve hazardous materials, which are more easily isolated in a controlled environ-

162

ment. Large electrical power loads can also be safeguarded better in the controlled
cleanroom environment.

B.2

Design of Cleanroom Facility

This section contains general information about the design of cleanroom, including specifications for cleanliness, environment, and materials. Utilities built
into the facility are also discussed. Complete as-built drawings, manuals, and
specifications for the cleanroom facility can be found in Watson building, room
270.

B.2.1

General

Figure B.1 is a diagram of the cleanroom, drawn to scale.

The space for the

facility was obtained by combining Labs 251 and 253 of the Thomas J. Watson
Sr. La.bora.tory of Applied Physics. All interior walls in and between the two
rooms were demolished. Building plumbing and ducting were rerouted around the
cleanroom.
The cleanroom facility has been partitioned into six separate rooms. The following is a listing of each room and its function.
1. The MBE room houses the Perkin Elmer MBE system.

2. The prep room contains two hoods for MBE substrate preparation.
3. The lithography room houses all of the equipment for photolithography
(mask aligner, photoresist spinner, ovens, microscope, etc.).
4. The Vestibule provides a space for people to put on cleanroom gowns, caps,
and booties prior to entering the rest of the facility.

163

,....

".,,,-~,c~l

1 _:::; H~r-'

v,'.:1;)

,')

J517f!/().:J

CD,_,,
ji;?/-';JS

,,
i::

00)1

).] _JJ.'JA1o_\'

""?I

\"()
I)

J111"'1Xlf7

'<'.

IQ,JIW/1'-"'

I)'<

':,.

".
,:,

.::i

.--,

'[§

~I
\,~1 ~

.'c

Figure B.1: Diagram of the cleanroom facility, drawn to scale. The facility has six
separate rooms, with an overall specification of Class 10,000. The facility houses
equipment fur molecular beam epitaxy an

164

5. Utility room 1 is a. nonclean room which houses utilities and dirty equipment needed by the MBE system.
6. Utility room 2 is a nonclean room which houses utilities and dirty equipment needed by the photolithography and substrate preparation equipment.

Connection of Rooms
Access between the various rooms is provided as follows:
1. Sliding glass doors connect the vestibule to the lithography and MBE

rooms. The door to the lithography lab is tinted yellow. Another sliding
glass door connects the prep room to the MBE room.

2. Windows are located on the wall between utility room 1 and the MBE
room, and on the wall between utility room 2 and the lithography room.

3. Removable walls have been placed between utility room 1 and the MBE
room, and between utility room 2 and the lithography room. These walls
can be dismounted to move large pieces of equipment into and out of the
cleanroom facility.
4. Plumbing passthroughs are used as points of entry for utility lines from
the utility rooms to the MBE and lithography labs.
5. A chemical passthrough cabinet is located on the wall between the vestibule and the prep room. The cabinet is large enough to hold several one
gallon chemical bottles. Transparent cabinet doors are mounted on both the
vestibule and prep room sides, so that bottles can be transferred into the
prep room without being carried through the MBE room.
6. A sample passthrough cabinet (identical in design to the chemical passthrough cabinet) is located on the wall between the lithography room and

165

Lab 255. This cabinet facilitates the removal of samples from the cleanroom.
The transparent cabinet doors are tinted yellow.

7. Doors to the corridor from the vestibule and the two utility rooms have been
made.

Air Handling
Air handling for the cleanroom facility is separated into two sections. One air
handler is dedicated to the MBE room. Air is supplied through the HEPA filter
units in the ceiling, pushing air out of the room through grates at the bottom of
return air walls and the wall to utility room 1. The return air plenum consists
of the return air walls, utility room 1, and the space bet.ween the MBE room
ceiling and the Watson roof. Air is drawn from the plenum for refiltering and
recirculation into the MBE room. A purge fan has been installed to exhaust air
from the room H desired. Make up air is drawn from the outside. The second
air handler services the lithography room, prep room, and vestibule. Once again,
air is supplied through HEPA filter units in the ceiling. Considerably more air is
exhausted from these rooms due to four exhausted hoods. Hence, less of the air
is recirculated. The return air plenum consists of return air walls, utility room 2,
and the space between the Watson roof and the lithography room, prep room, and
vestibule ceilings. All of the air handling equipment is located on the Watson roof,
above the cleanroom facility.

B.2.2

Specifications

The following is an abbreviated list of specifications for the cleanroom facility.
1. The overall cleanroom classification is Class 10,000, which means that there

are fewer than 10,000 particles of size 0.5 µm or larger per cubic foot. Particle

166
counter measurements consistently yield smaller numbers of particles than
this specification.
2. Several specific areas are designed to be cleaner than the rest of the facility.

The work spaces under the laminar flow hoods and benches should be Class
100 areas due to complete HEPA filter coverage. Three 4' x 4' vinyl curtained
a.rea.s have been set up above ca.ch of the three intro hatches on the MBE

system. These areas enclose two 2' x 4' HEPA :filters each, making them
roughly Class 100. Finally, the prep room is probably Class 1000 due to the
large number of HEPA filters per unit area in the room.
3. The temperature in the facility is thermostatically controlled. The air handling system should be capable of maintaining a temperature of 72 ± 5°F.
4. The humidity in the facility should be controllable to 40% ± 5% RH.
5. The existing walls in the cleanroom have been coated with an epoxy paint

which generates very few particles.
6. Modular cleanroom wall panels have been used for all new cleanroom walls.
7. The celling of the clean room is a. Comp-Aire 2" system. Con-web clean

room ceiling panels have been used.
8. The flooring material in the cleanroom facility is mipolan. This material is
resistant to chemical spills, but can he cracked by liquid nitrogen.

B.2.3

Utilities

The following is a list of all of the utilities available in the cleanroom and how
each of thein can be accessed.

167
1. Liquid nitrogen is needed for the three growth chambers on the MBE

system and for sorption pumps. A Vacuum Barriers phase separator has
been installed on the roof of Watson above utility room 1 for this purpose.
Three triax. feed lines {inner-liquid, middle-liquid and gas, outer-vacuum)
have been dropped from the phase separator and plumbed to each of the three
growth chambers. The lines have been routed via a plumbing passthrough
and cable trays in the MBE room. Semifl.ex return lines (inner-liquid and
gas, outer-vacuum) run back to the phase separator from the chambers. The

phase separator is filled by 160 liter liquid nitrogen dewars, housed in utility
room 1.

2. Electrical panels for the facility have been placed in the utility rooms. Three
225 A panels, located in utility room 1, service the MBE room. One 225 A
panel, located in utility room 2, services the lithography and prep rooms.

All of the large (permanent) equipment in the cleanroom facility has been
hard-wired to the electrical panels.
3. Unfiltered deionized water is available in the utility rooms. It has also

been plumbed to the water filtration systems installed in the prep and lithog~
raphy rooms.
4. Filtered, sterilized deionized water is generated by the water filtration

systems in the prep and lithography rooms. Each of these systems consists of
a Millipore filtration unit and an Aqua:fine ultraviolet sterilyzer. The water
is then delivered to each of the four wet hoods in the facility.
5. Closed loop (Watson building) water for equipment cooling can be
accessed in both of the two utility rooms.

168
6. A NESLAB filtered, closed loop, cooling water recirculator has been

installed in utility room 1. The NESLAB unit supplies cooling water for the
MBE system via a cooling water manifold located in the MBE room. Heat is
exchanged from the NESLAB to the Watson equipment cooling water lines.
7. Hot and cold tap water are available in both 0£ the utility rooms, as well
as the fume hood in the lithography room.

8. Natural Gas is available in both 0£ the utility rooms.
9. High pressure air {85 PSI) is available in both of the two utility rooms,
as well as two of the hoods in the lithography room and both of the hoods

in the prep room.
10. Unfiltered nitrogen gas is available in both of the utility rooms.
11. Filtered, dried nitrogen gas has been plumbed to all of the hoods in
the facility. A separate, regulated supply has been connected to a manifold
in the cable tray above the MBE system via a plumbing passthrough. The
manifold has six output connectors (Swagelok) for hooking up the different
MBE chambers.
12. High pressure nitrogen (for MBE pneumatics) is supplied by a gas cyliner

in utility room 1. A high pressure hose carries the nitrogen to a cross in the
cable tray above the MBE system. The three remaining branches of the cross
are connected to each of the three growth chambers via high pressure hoses.
A stainless steel manifold has been placed in the cable tray for this utility,
but has never been used.
13. Helium gas is supplied by a gas cylinder in utility room 1. A stainless steel
manifold has been placed in the cable tray for this utility, but has never been
used.

169
14. A vacuum pump has been placed in the Watson mechanical pod nearest
to the cleanroom. The pump can be switched on in either the lithography
room or the prep room. Vacuum lines have been plumbed to each of the
hoods in the facility.
15. Drains: all of the hood drains have been tied together and plumbed to an
open drain in utility room 2. Because of this connection scheme, all solvents
(except alcohols) must be stored in waste bottles, and all acids must be
neutralized before they are :flushed down the hood drains. An open drain
has also been left in utility room 1 for use in the event of :flooding.
16. RS2S2 lines (for computer terminals) have been placed at convenient points
throughout the cleanroom.