close

Вход

Забыли?

вход по аккаунту

?

Solid source molecular beam epitaxy of indium phosphide-based composite-channel high electron mobility transistor structures for microwave and millimeter-wave power applications

код для вставкиСкачать
INFORMATION TO USERS
This manuscript has been reproduced from the microfilm master. UMI films
the text directly from the original or copy submitted. Thus, some thesis and
dissertation copies are in typewriter face, while others may be from any type of
computer printer.
The quality of this reproduction is dependent upon the quality of the
copy submitted. Broken or indistinct print, colored or poor quality illustrations
and photographs, print bleedthrough, substandard margins, and improper
alignment can adversely affect reproduction.
In the unlikely event that the author did not send UMI a complete manuscript
and there are missing pages, these will be noted.
Also, if unauthorized
copyright material had to be removed, a note will indicate the deletion.
Oversize materials (e.g., maps, drawings, charts) are reproduced by
sectioning the original, beginning at the upper left-hand comer and continuing
from left to right in equal sections with small overlaps.
Photographs included in the original manuscript have been reproduced
xerographically in this copy.
Higher quality 6" x 9" black and white
photographic prints are available for any photographs or illustrations appearing
in this copy for an additional charge. Contact UMI directly to order.
Bell & Howell Information and Learning
300 North Zeeb Road, Ann Arbor, Ml 48106-1346 USA
800-521-0600
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Solid Source Molecular Beam Epitaxy of InP-Based CompositeChannel High Electron Mobility Transistor Structures for
Microwave and Millimeter-Wave Power Applications
A Thesis
Presented to
The Academic Faculty
By
Tong-Ho Kim
In Partial Fulfillment
of the Requirements for the Degree of
Doctor o f Philosophy in School of Electrical and Computer Engineering
Georgia Institute of Technology
June 2000
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
UMI Number: 9978400
UMI
UMI Microform9978400
Copyright 2000 by Bell & Howell Information and Learning Company.
All rights reserved. This microform edition is protected against
unauthorized copying under Title 17, United States Code.
Bell & Howell Information and Learning Company
300 North Zeeb Road
P.O. Box 1346
Ann Arbor, Ml 48106-1346
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Solid Source Molecular Beam Epitaxy of InP-Based CompositeChannel High Electron Mobility Transistor Structures for
Microwave and Millimeter-Wave Power Applications
Approved:
fyvd!
April Brown, Chairman
Nan Jokerst
ar
t i t t y
Phillip First
Date Approved
O/iof
2 .0 0
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
o
To my parents, Kye-Hwa Kim and Ki-Soon Ko
and memory o f grandfather, Won-Moo Kim.
i
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ACKNOWLEDGEMENTS
I would like to thank the many people who provided encouragement and support
throughout my thesis work. I am extremely grateful for the guidance, encouragement, and
support that I received from Dr. April. S. Brown. She has always nourished me with a
kindness and many fresh ideas to overcome million tons o f problems in front o f me. I
would also like to thank Dr. Nan Jokerst, Dr. Joy Laskar, Dr. Gary May, and Dr. Phillip
First for their services and comments as exam committee members during my doctoral
program.
I would like to extend my deepest appreciation to Dr. Robert Metzger who
introduced a MBE technology and guided right ways to do experiments with his
tremendous experience and experimental knowledge. I would also like to specially thank
Dr. Alan Doolittle for insightful discussions and technical assistance during my thesis
work. I would like to thank Dr. Stuart Stock for allowing me to use his equipments and
insightful discussions.
Special thanks go to my group members, Dr. Kyeong Lee, Dr. Georgiana Dagnall,
Changhyun Y i, J.J. Shen, Sangbeom Kang, Gon Namkoong, Terence Brown, and
Gregory Triplett for their encouragement, cooperation, friendship, technical help, and
insightful discussions related with my research.
ii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
I indebted to my parents, Kye-Hwa Kim and Ki-Soon Ko who have always
encouraged and supported with a great love in my life. Last, I gratefully acknowledge my
brothers, Tong-Heub and Tong-Ki, my wife, Hyeon-Ja Choi, and two sons, Sung-Hoon
and Sang-Hyun for their support, endurance, and love.
in
Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
CONTENTS
DEDICATION........................................................................................................ i
ACKNOWLEDGEMENTS...................................................................................ii
CONTENTS........................................................................................................... iv
LIST OF TABLES............................................................................................... vii
LIST OF FIGURES.............................................................................................. ix
SUMMARY.......................................................................................................... xv
CHAPTER I. INTRODUCTION...........................................................................1
CHAPTER II. BACKGROUND.......................................................................... 7
2.1 Solid Source Molecular Beam Epitaxy.............................................
7
2.2 Conventional InP-Based HEMTs............................................................ 9
2.2.1 Basic Device Structure......................................................................9
2.2.2 Two-Dimensional Electron Gas..................................................... 11
2.2.3 Calculation o f the Equilibrium 2DEG Density............................ 12
2.2.4 Electron M obility............................................................................ 26
2.3 Pseudomorphic InP-Based HEMTs.......................................................28
2.3.1 Critical Thickness............................................................................ 28
2.3.2 Energy Bandgap.............................................................................. 36
2.3.3 Photoluminescence Transition Energy..........................................48
2.4 InP-Based Power HEMTs......................................................................51
iv
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2.4.1 High Frequency Power Transistor Design.....................................51
2.4.2 InAlAs/InGaAs Power HEMTs...................................................... 55
2.4.2.1 Schottky Layer Designs......................................................... 56
2.4.2.2 Channel Layer Designs...........................................................58
2.4.2.3 Reduction o f Electric Field at Drain-End............................. 62
2.4.2.4 Parasitic Gate Leakage............................................................63
2.4.2.5 Buffer Layer Designs..............................................................64
2.4.2.6 Hole Barrier Approaches........................................................ 65
2.4.2.7 Doping Schemes...................................................................... 66
2.4.3 Power Performance.......................................................................... 67
CHAPTER III. GROWTH OF CONVENTIONAL InP-BASED H EM Ts....70
3.1 Calibration Procedures for InP-Based HEMT Growths......................71
3.2 Standard InAlAs/InGaAs MD Structures..............................................73
3.2.1 Optimization o f 2DEG Conductivity..............................................75
3.2.2 Impact o f Channel Depth on 2DEG Conductivity........................ 80
3.3 Doped-Channel InAlAs/InGaAs MD Structures.................................83
3.4 Double-Heterojunction InAlAs/InGaAs MD Structures....................87
CHAPTER IV. GROWTH AND DEVICE PERFORMANCE OF InGaAs/InP
COMPOSITE-CHANNEL HEM Ts.................................................................... 92
4.1 Preliminary Experiments........................................................................ 92
4.2 InGaAs/InP Composite-Channel MD Structures.................................94
4.2.1 Optimization o f 2DEG Conductivity............................................ 96
v
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4.2.2 Quantum Size Effects o f InGaAs Channel..................................105
4.2.3 Growth of InGaP Hole Barrier...................................................113
4.3 Device Performance.............................................................................118
CHAPTER V. GROWTH OF InAsxP,.x/InP COMPOSITE-CHANNEL
HEMT STRUCTURE.........................................................................................128
5.1 InAsxPj.x/InP Multi-Quantum Wells....................................................129
5.2 InAsxPi.x/InAlAs Multi-Quantum W ells............................................ 150
5.3 InAsxPi.x/InP Composite-Channel MD Structures.............................161
5.4 Strain Compensation o f InAsxP|.x/InP Composite-Channel..............177
CHAPTER VI. CONCLUSIONS...................................................................... 183
REFERENCE...................................................................................................... 187
PUBLICATIONS AND PRESENTATIONS.................................................. 201
VITA.................................................................................................................... 203
vi
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
LIST OF TABLES
Table 1. Properties of ni-V binary materials at 300 K[123]..............................................42
Table 2. Comparison of power performance of InP-based FETs. In this table, Lg and Wg
are the gate length and width, respectively. Pout represents the high frequency output
power................................................................................................................................ 69
Table 3. Comparison of Hall data o f conventional InAlAs/InGaAs MD structure with 250
nm InAlAs buffer and InAlAs/InGaAs MD structure grown with half growth rate and
InGaAs/InAlAs superlattice buffer................................................................................... 77
Table 4. Hall data for conventional InAlAs/InGaAs M D structures with different channel
depths and a planar-doping (PD, No = 5.1 x 1012 cm'2) scheme........................................ 82
Table 5. Hall data of InP single channel MD structures....................................................93
Table 6. Hall data of multi-channel InGaAs/InP MD structure.....................
98
Table 7. Influence of InP spacer thickness for planar-doping InP subchannel on the 2DEG
conductivity o f InGaAs/PD (tyo = 1.6 x 1012 cm'2) InP composite-channel MD structure.
An uniform-doping of ND= 8 x 1018 cm'3 for top donor layer was utilized...................... 99
Table 8. The relative peak position of the 4.2 K PL spectra for InAlAs/ InAsxP|.x MQWs
as compared to InP/ InAsxPi.x MQW grown with the same growth conditions and structure
without interruptions.......................................................................................................159
Table 9. Hall data of InAlAs/InAsxP|.x/InP composite-channel MD structure grown at 500
O,
C with various As compositions in the InAsxP|.x channel. Hall data of lattice-matched
InAlAs/InGaAs/InP composite-channel MD structure grown with the same epilayer design
(Figure 68 (a)) except the channel material was compared............................................. 165
Table 10. Hall data for the InALAs/InAso.6 Po.4/InP composite-channel MD structure grown
with various interface interruption times.........................................................................169
Table 11. Hall data for the InAlAs/InAso.6 Po.4 /InP composite-channel MD structure with
different spacer thickness of 2 to 8 nm...........................................................................172
Vll
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 12. Calculated compensation index of Ini.xGaxAs (10 nm)/InAso.6 Po.4 (12 nm)
epilayers grown on InP.................................................................................................... 179
Table 13. 300 K Hall data of the strain-compensated Ini.xGaxAs/InAso6Po4 /InP
composite-channel, pseudomorphic Ino.4 Gao.6As single channel, and pseudomorphic
InAso.6 Po.4/InP composite-channel HEMT structures.......................................................180
vui
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
LIST OF FIGURES
Figure 1. Schematic diagram of standard InAlAs/InGaAs H EM T.....................................11
Figure 2. Energy band diagram o f an ungated basic HEMT structure with two subbands in
a triangular potential well..................................................................................................15
Figure 3. Two-dimensional electron gas density at the channel and minimum donor layer
thickness in the InAlAs/InGaAs HEMT as a function of doping density with the spacer
thickness dj as a parameter at 300 K ..................................................................................22
Figure 4. Two-dimensional electron gas density and minimum donor layer thickness in the
InAlAs/InGaAs HEMT uniformly-doped with Nd = 8 x 1018 cm'3 as a function of the
spacer thickness dj.............................................................................................................23
Figure 5. Subbands energy (Eo, E|, and E2 ) and Fermi energy Ef (calculated from Eq. (3))
vs. 2DEG
density. Ef (dotted line) self-consistantly calculated in Ref. 33 is
compared........................................................................................................................... 24
Figure 6. Spatial width of subbands vs. 2DEG density calculated from Eqs. (7) and
(8)......................................................................................................................................25
Figure 7. C alculated electron m obility in Ino.53Gao.47As lattice-m atched to InP as a function
of temperature [34].......................................................................................................... 27
Figure 8. Critical thickness vs. As mole fraction in InAsxPi.*/InP (100)...........................32
Figure 9. Critical thickness vs. In mole fraction in InGaAs/InP (100)...............................33
Figure 10. Critical thickness vs. In mole fraction in InAlAs/InP (100).............................34
Figure 11. Critical thickness vs. In mole fraction in InGaP/InP (100)...............................35
Figure 12. The modification of energy band structure in a strained epilayer.................... 41
Figure 13. Strain comparison o f some InP-related alloys.................................................. 43
Figure 14. 300 K energy bandgap and strain of InAsP grown on InP (100) vs. As
composition....................................................................................................................... 44
ix
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure IS. 300 K energy bandgap and strain of InGaAs grown on InP (100) vs. Ga
composition...................................................................................................................... 45
Figurel6. 300 K energy bandgap and strain of InAlAs grown on InP (100) vs. AI
composition...................................................................................................................... 46
Figure 17. 300 K energy bandgap and strain of InGaP grown on InP (100) vs. Ga
composition...................................................................................................................... 47
Figure 18. Relation between Vt , Vp, Vs and
E °a , A a , A a for a heterojunction A-B
system................................................................................................................................50
Figure 19. Schematic representation for a FET operating in class A ................................. 54
Figure 20.
Schematic diagram of standard lattice-matched InAlAs/InGaAs
MD
structure.............................................................................................................................75
Figure 21. Hall data for the standard InAlAs/InGaAs MD structures versus As4 BEP flux:
(a) 300 K and (b) 77 K Hall data...................................................................................... 78
Figure 22. 2DEG properties of standard MD structure (Figure 20): (a) 300 K and (b) 77 K
Hall data............................................................................................................................79
Figure 23. Schematic diagram o f the uniformly-doped donor and PD InGaAs single
channel..............................................................................................................................85
Figure 24. Hall data of planar-doped InGaAs channel M D structure (see Figure 23)....... 86
Figure 25. Double Planar-doped (PD1 and PD2) InGaAs single channel M D structure.
During the planar-doping, the substrate temperature was reduced from 500 to 420 °C to
avoid the segregation of Si...............................................................................................90
Figure 26. 300 K Hall data o f DH InAlAs/InGaAs M D structure grown with various
planar-doping density, in which Hall data of SH MD structure (open circle and square) was
also compared.................................................................................................................. 91
Figure 27. Schematic diagram o f a basic InGaAs/InP composite-channel MD structure.. .96
Figure 28. Multi-channel InGaAs/InP M D structures........................................................ 97
x
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure 29. Schematic diagram of planar-doped InGaAs channel/undoped InP MD structure
with various doping density (N 2 D = 7.54 - 4.8 x 1012 cm'2) ............................................ 102
Figure 30. Comparison of 300 K. 2DEG conductivity for InGaAs/InP composite-channel
MD structures with various doping schemes...................................................................103
Figure 31. Comparison of the 300 K Hall mobility and sheet resistance versus 2DEG
density for various InGaAs/InP composite-channel M D structures..................................104
Figure 32. Uniform-doped donor and thickness varied InGaAs/15 nm InP spacer/PD InP
MD structures grown to investigate the channel-size quantization effects..................... 105
Figure 33. Hall data versus InGaAs channel thickness for InGaAs/InP composite-channel
MD structures: (a) 300 K and (b) 77 K Hall data............................................................ 106
Figure 34. Comparison of the impact of InGaAs channel thickness on 2DEG property of
InGaAs composite-channel MD structures with reference data [ 106]..............................110
Figure 35. Impact of InGaAs channel size on maximum transconductance and drain current
density in the InGaAs/InP composite-channel MD structure........................................... I l l
Figure 36. Influence of InGaAs channel thickness on 2DEG density and mobility product
(a) and on normalized intrinsic (gm)ma* (b)......................................................................112
Figure 37. InGaAs/InP composite-channel MD structures with an Ini.yGayP (2.5 nm) hole
barrier..............................................................................................................................113
Figure 38. Hall data for the InGaAs/InP composite-channel MD structure with Ini.yGayP
hole barrier (0 < y < 0.5): (a) 300 K and (b) 77 K Hall data............................................ 116
Figure 39. Comparison of 300 K Hall data of the InGaAs/InP composite-channel MD
structures with two different layer positions of the InGaP hole barrier.......................... 117
Figure 40. Comparison of 300 K Hall data of InGaAs/InP composite-channel HEMTs and
InP/InGaAs/InP HEMTs grown by various growth methods.......................................... 120
Figure 41. Maximum transconductance versus maximum drain current density of 0.15 pmgate InGaAs/InP composite-channel HEMTs with various doping schemes................. 121
Figure 42. 0.15 pm x 200 (4 x 50) pm InGaAs/InP composite-channel HEMTs
(TRW )............................................................................................................................. 122
XI
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure 43. Current-voltage (a) and transconductance-gate voltage characteristics (b) of
InGaAs/InP composite-channel HEMT(0.15 pm x 40 pm )............................................ 123
Figure 44. Energy band diagram and electron distribution of InGaAs/InP compositechannel HEMT (Figure 27) simulated by ATLAS (Device simulation software,
SILVACO)......................................................................................................................124
Figure 45. Comparison of ft, fmx, and gain of the conventional InAlAs/InGaAs HEMT and
InGaAs/InP composite-channel HEMT (Figure 42) fabricated with the same process
(TRW’s InP-based HEMT process line).......................................................................... 125
Figure 46. Burnout breakdown voltages of the conventional InP-based HEMT and the
InGaAs. InP composite-channel HEMT measured at various drain current levels. The
composite-channel structure improves the breakdown voltage by about 1.5 V ................126
Figure 47. Output power and gain as functions of power measured from two-stage
composite-channel HEMTs MMIC amplifiers under on-wafer pulse power conditions. A
saturated output power of 25 dBm (316 mW) achieved at 94 GHz................................ 127
Figure 48. Schematic structures of the InP/InAsP MQWs and InAlAs/InAsP MQWs
131
Figure 49. Energy band gap of InAsxPi.x on an (100) InP substrate versus As composition
x at 4.2 K. The compressive strain in InAsxPi.x increases the band gap for the heavy-hole
(HH), light-hole (LH), and split-off valence band, respectively...................................... 136
Figure 50. X-ray rocking curve (a) experimental and (b) theoretical of a three period
InP(20.8 nm)/InAso.47Po.53(5.4 nm) M Q W .......................................................................137
Figure 51. X-ray DCRCs simulation o f InP (5 nm)/InAsxP|.x (25 nm) (0.2 < x < 0.8)
MQWs............................................................................................................................. 138
Figure 52. X-ray DCRCs measured
in two InP/InAsxP|.x MQWs with
different
compositions (x = 0.47 and 0.49).....................................................................................139
Figure 53. X-ray DCRCs simulated
in two InP/InASxPi.x MQWs with
different
compositions (x = 0.43,0.44, and 0.45)..........................................................................140
Figure 54. X-ray DCRCs simulated in InP (20.8 nm) /InAsxPi.x (5.6 nm) MQW with
different well thicknesses (Lw = 5.5,5.6, and 5.7 nm).................................................... 141
xu
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure 55. X-ray DCRCs simulated in InP/InAso.4 sPo.55 (5.6 nm) MQW with different
barrier thicknesses (L b = 20.7,20.8, and 20.9 nm)..............................................................142
Figure 56. X-ray double crystal rocking curves (DCRCs) o f a three-period InP/InAsxP|.
x/InP MQW (Figure 48) grown with various arsenic fluxes............................................. 143
Figure 57. The ratio of arsenic flux and V element flux (arsenic flux + phosphorous flux)
versus As composition determined by X-ray diffraction measurement and simulation in the
abrupt InP/InAs*Pi.x MQWs...........................................................................................144
Figure 58. 300 K PL emission o f InP/InAsxP|.x (x = 0.47 and 0.49) MQWs (Figure
48(a))...............................................................................................................................145
Figure 59. 77 K PL emission of InP/InAsxPi.x (0.3 < x <0.54) MQWs(Figure 48(b))....146
Figure 60. Temperature dependence of PL emission measured in InP/InAsP MQW , A317
sample..............................................................................................................................147
Figure 61. The relationship between the calculated transition energies and As composition
x in InP (24 nm) / InAsxP|.x (5.9 nm) MQWs. A heavy-hole valence band offset of 25%
was assumed in this calculation. The circles represent the experimental data...................148
Figure 62. Comparison of x-ray DCRCs of InP/InAso.6Po.4 MQWs grown on an InP (100)
substrate with an anion flux interruption at InP/InAsP interfaces....................................149
Figure 63. The schematic diagram o f the growth sequence at the interface of InAlAs/InAsP
MQWs. t3 -t2 and ts-U are an interruption time under P flux at the top interface of InAsP
and under As flux at the bottom interface, respectively................................................... 151
Figure 64. A comparison of the 4.2 K PL spectra for InAlAs/ InAsxP|.x MQWs and InP/
InAsxP,.x M Q W ...............................................................................................................156
Figure 65. The relationship between the calculated transition energies and As composition
x for Ino.s2 Alo.48 As (24 nm)/InAsxP].x (5.9 nm) MQWs. A heavy-hole valence band offset
o f 25% was assumed in this calculation. The various marks represent the experimental PL
data at 4.2 K ....................................................................................................................157
xui
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure 66. A comparison of the FWHM (full width at half-maximum) of PL spectra for
InAlAs/ InAsxP|.x MQWs and for InP/ InAsxP].x MQW with the same As composition x in
InAsP...............................................................................................................................158
Figure 67. A comparison of the measured energy differences in the two peaks (numbered 3
and 4) and the calculated HH (heavy-hole)-LH (light-hole) and E2-E1 (Eland E2 are
electron bound-states) energy difference as a function of As composition x for InAlAs/
InAsxP|.x MQWs at 4.2 K................................................................................................ 160
Figure 68. The schematic diagram of the InAlAs/InAsxPi.x/InP composite-channel MD
structure: (a) an InAsxP|.x channel MD structure with As composition variation and (b) a
strained InAlAs/InAso.6Po.4 composite-channel MD structure with the InAso.6 Po.4 channel
grown at 420 °C. The thickness tj, t2 , and t3 were changed from 4 to 8 nm, from 2 to 8 nm,
and from 12 to 4 nm, respectively................................................................................... 163
Figure 69. The 300 K sheet resistance of InAlAs/InAsxPi.x/InP composite-channel MD
structure (circle) shown in Figure 68 (a) versus As composition in the InAsxP|.x channel.
The sheet resistance of lattice-matched InAlAs/lnGaAs/InP composite-channel MD
structure, which was grown in the same batch and with the same epilayer design except the
channel material, was compared..................................................................................... 167
Figure 70. The influence of InAso.6 Po.4 channel thickness on 300 K 2DEG mobility and
density o f the InAlAs/InAso.aPoVInP composite-channel MD structure......................... 171
Figure 71. Comparison of 300 K Hall data of InAsP channel HEMTs (The closed and open
marks are a mobility and 2DEG density, respectively.)................................................... 175
Figure 72. The influence of doping thickness product (No x t|) on 300 K. 2DEG mobility
and density o f the InAlAs/InAso.6 Po.4 /InP composite-channel M D structure................... 176
Figure 73. The schematic diagram of InAso.6 Po.4/InP composite-channel MD structure
strain-compensated with 10 nm Ini.xGaxAs (x = 0.55,0.6,0.7,0.75) grown by
SSMBE............................................................................................................................178
xiv
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
SUMMARY
The growth conditions and material design factors related to optimizing the twodimensional electron gas (2DEG) conductivity of InP-based composite-channel high
electron mobility transistor (HEM T) structures were investigated through a comparative
study of the 2DEG properties of various InP-based HEMT structures, such as
conventional InAlAs/InGaAs HEMT structures and lattice-matched InAlAs/InGaAs/InP
and pseudomorphic InAlAs/InAsP/InP composite-channel HEMT structures grown with
solid-source molecular beam epitaxy (SSMBE). Several approaches, such as channel
doping, the inclusion of InGaP hole barriers, and Al-rich InAlAs Schottky barriers were
combined with the InGaAs/InP composite-channel HEMT structure to enhance
breakdown voltage. The device performance fabricated with the optimized InGaAs/InP
composite-channel HEMT structure showed excellent DC characteristics and state-of-theart RF power performance at W-band. Furthermore, a study of the 2DEG properties of
pseudomorphic InAsxPi.x/InP composite-channel HEMT structures was carried out.
Structure variations included the effects of As composition (0 < x < 0.74), channel
thickness, InAlAs spacer thickness, arsenic and phosphorus flux switching scheme at the
interfaces,
doping
scheme,
and strain compensation (Ini.xGaxAs, x
>
0.53/
InAso.6 Po.4 /InP). The 2DEG conductivity results showed the great potential of the
InAlAs/InAsxP].x/InP material system for microwave and millimeter wave power
applications.
xv
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER I
INTRODUCTION
InP-based high electron mobility transistors (HEMTs) have demonstrated highfrequency characteristics superior to those of any other transistor, including the highest
fmax (600 GHz, as reported in [1]), the highest ft (340GHz, as reported in [2]), and the
lowest noise figure and the highest efficiency at millimeter wave frequency [3].
The unequaled performance of the InP-based HEMT arises directly from the
intrinsic properties of the InAlAs/InGaAs material system, where the high indium content
(typically 53-80%) InGaAs channel possesses high electron mobility and velocity, and
the large conduction band discontinuity at the InGaAs/InAlAs heterojunction permits
high two-dimensional electron gas (2DEG) densities to be obtained. These properties
enable high current and transconductance. It is the transconductance of the InP HEMT
that is most directly responsible for its increased operating frequency and excellent gain.
Transconductance values as high as 1500-1700 mS/mm have been reported [4] and
typical 0.1 pm InP low noise HEMTs exhibit gm of 800-1000 mS/mm, compared with
600 mS/mm for comparable GaAs pseudomorphic HEMTs (PHEMT) [5].
1
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Although the superior performance of InP-based high electron mobility transistors
(HEMTs) at microwave and millimeter-wave frequencies has been widely recognized,
the performance o f InP-based HEMTs for power applications has been limited by
relatively low power densities. This results primarily from the low breakdown voltage
and Schottky barrier height of InP-based HEMTs. The off-state and on-statc breakdown
voltage in InP-based power HEMTs have been shown to be dominated by impact
ionization in the channel layer [6 ], and are typically low due to the low breakdown field
of narrow bandgap Ino.s3Gao.47 As channel. It is crucial to overcome these problems in
order to enhance the power capability of InP-based power HEMTs. Many approaches to
improving operation under high electric fields in the InP-based power HEMTs have
focused on modification of the channel material and Schottky barrier material.
InP is an attractive channel material for high-speed and high-power devices due
to its large T-L valley separation and the low impact ionization coefficient in comparison
to InGaAs. However, the current gain cutoff frequencies of InP-channel HEMTs are
inferior to InGaAs HEMTs due to the low electron mobility of InP and small conduction
band discontinuity between InP and InAlAs. To compensate the disadvantage of the InPchannel HEMT, Enoki proposed a composite channel that combines a thin layer of
lattice-matched (LM ) InGaAs and InP as the channel material [7]. With this structure, it
is possible to exploit the advantageous physical properties of both materials, i.e., the high
electron mobility o f LM InGaAs at low electric fields, and the high breakdown and
velocity of InP at high electric fields [8, 9]. To investigate the achievable 2DEG
2
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
characteristics in this structure, it is necessary to systematically study InGaAs/InP HEMT
structures grown with various epitaxial layer designs and doping strategies.
The InAsxPi.x alloy is also a good candidate for the channel material of an InPbased pseudomorphic power HEMT, due to the high T-L valley separation [10, 11], and
higher conduction band discontinuity (0.75AEg in InAsP/InP heterostructure [12] as
compared to 0.65AEg of LM InAlAs/InGaAs heterostructure [13]) and critical field for
breakdown (as compared to LM InGaAs on InP). In the growth of this mixed-anion
material system, the strain present in the alloy can modify growth dynamics, in particular,
the incorporation coefficients of anions. In addition, anion exchange can occur during
interface formation and can thereby roughen the interface. This interface roughness can
negatively impact the conductivity of 2DEG formed in the InAlAs/InAs*Pi.x/InP
composite-channel modulation-doped (M D ) structures.
In general, the hole current generated via impact ionization in the channel reduces
the gate-drain breakdown voltage. To enhance the gate-drain breakdown voltage, a hole
barrier layer can be inserted between the Schottky layer and channel. Also, a larger
bandgap Schottky layer, such as Al-rich InAlAs, can be grown to enhance the Schottky
barrier height, resulting in improvement in breakdown voltage [14], In this thesis, an
InGaP hole barrier and Al-rich InAlAs Schottky layer will be combined to further
improve the breakdown voltage in InP-based power HEMTs.
In the growth o f MD structures for high frequency power applications, several
parameters, such as high 2DEG mobility, high 2DEG density, high Schottky barrier
3
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
height, low channel impact ionization rate (inversely proportional to the bandgap of
material), and high 2DEG velocity at high electric fields, have to be considered to
optimize power capability. In particular, a higher 2DEG density is required to enhance
the channel current density for power applications and can be achieved through the
appropriate material design and doping strategy to maximize a 2DEG density in the
channel of MD structures. In this thesis, the impact of various doping strategies, such as
single heterojunction doping, double heterojunction doping, and channel doping, on the
2DEG characteristics o f lattice matched InP-based MD structures will be investigated.
The electrical properties of M D structures strongly depend on growth methods
and conditions, material structure and quality, and doping strategy.
In general,
phosphides, such as InP, InAsP, InGaAsP, InAlP, and InGaP, have been introduced in
conventional structures of InP-based HEMTs to enhance their power properties and have
been mainly grown by gas-source molecular beam epitaxy (GSMBE) and metalorganic
chemical vapor deposition (MOCVD). As an alternative technology for the growth of
phosphorus-containing epitaxial structures, solid source molecular beam epitaxy
(SSMBE), an emerging growth technology, has been successfully utilized. However, the
use of this economic and relatively safe growth process has been primarily dedicated to
the development of optical devices.
The purpose o f this research is to develop lattice-matched InGaAs/InP and
pseudomorphic InAs*P|.x/InP composite-channel HEMT structures and devices grown by
SSMBE for microwave and millimeter-wave power applications. In this research, the
4
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
material design and growth conditions will be addressed in order to optimize the 2DEG
properties of these InP-based composite-channel (arsenic and phosphide mixed
heterostructure) modulation-doped structures grown by SSMBE. The growth conditions
and material design factors related to optimizing the 2DEG properties will be investigated
through a comparative study of the 2DEG properties of various InP-based MD structures,
such as the conventional InAlAs/InGaAs M D structures and InAlAs/InGaAs/InP and
InAlAs/InAsP/InP composite-channel M D structures grown with the same SSMBE
system.
This thesis consists of six chapters. In chapter n, background for the development
of InP-based power HEMTs will be discussed. In chapter m, the growth of conventional
InAlAs/InGaAs MD structures will be optimized to obtain optimum 2DEG conductivity
without the addition o f phosphorus-containing alloys. Utilizing the SSMBE growth
conditions that optimize the conventional MD structure, the 2DEG characteristics of
InGaAs/InP MD structures will be investigated in chapter IV . In addition, several
approaches, such as channel doping, InGaP hole barriers, and Al-riched InAlAs Schottky
layers, will be combined with this InGaAs/InP MD structure to enhance breakdown
voltage. The device performance fabricated with the optimized InGaAs/InP compositechannel HEMT structure will be described.
Next, in chapter V , a study of the 2DEG
properties of pseudomorphic InAlAs/InAsxP|.x/InP composite-channel M D structures as a
function of As composition (0 < x < 0.74), channel thickness, InAlAs spacer thickness,
arsenic and phosphorus flux switching scheme at the interfaces, doping strategy, and
5
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
strain compensation (Ini.xGaxAs, x > 0.53/ InAso^PoVInP) will be described. Finally the
conclusions will be presented in chapter VI.
6
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER II
BACKGROUND
2.1
Solid Source Molecular Beam Epitaxy
Phosphorus-based materials are of great
importance for many
advanced
optoelectronic and electronic devices. The most common techniques used for growing
phosphorus-containing epitaxial structures are MOCVD, GSMBE, metal organic (MO)
MBE, and chemical beam epitaxy (CBE). All these growth methods use highly toxic
hydrides or tertriarybutylphosphine (TBP) and tertiarybutylarsine (TBA) as group V
precursors. The conventional effusion cell MBE technique has been considered unsuited
to grow phosphorus-based materials due to problems associated with the high vapor
pressure of phosphorus required for growth, different allotropic forms of phosphorus (red
and white phosphorus) with significantly different vapor pressures, and depletion of the
phosphorus charge from high consumption.
These problems have been overcome by application of specific cell design, i.e., the
valved cracking cell. The initial phosphorus valved cell design [IS] evolved directly from
that of the arsenic valved cell [16]. While the arsenic valved cell (two-zone furnace) was
implemented successfully, the phosphorus cell was highly problematic. Although it made
7
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
growth with P2 possible, it did not resolve the problem of random allotrope
transformation internal to the cell. Consequently, the beam flux output from the cell was
very erratic. The inability to limit allotrope changes is due to a lack of thermal control
within the area between the red phosphorus oven and thermal cracking zone, resulting in
large flux transients when switching the valve positions. Control over allotrope
transformation and subsequent stabilization of the beam flux was found by utilizing a
valved cell with three temperature zones [17]. The cracker midsection, between the
cracker and the evaporator, is used to condense P4 vapor from red phosphorus, thus
transforming the material to white phosphorus.
Our equipment consists of a Riber 2300 MBE system with standard pumps, i.e., ion
pump, liquid nitrogen shrouds, and titanium sublimation. The growth chamber is
equipped with an EPI RB500VP arsenic valved cracking cell and a Riber KPC 40
phosphorus vaved cracking cell.
Baillargeon et al. [18] assessed the performance of the Riber KPC 40 valved
phosphorus cracking cell, which incorporates a temperature controllable reservoir for the
condensing of white phosphorus. In their study, the performance is investigated by
evaluation of the beam equivalent pressure (BEP) or flux, switching transient, stability,
and reproducibility. The molecule species compromising the beam is determined by the
temperature of the cracking zone. A cracking temperature of greater than 800 °C
produces a P2 beam flux, while below this temperature the flux remains P4. The flux is
8
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
controlled through valve positioner. The maximum BEP is determined by the temperature
of the white phosphorus reservoir and the conductance of the cell.
SSMBE has been successfully utilized for the development of optical devices such as
MQW lasers [19] and Bragg reflectors [20]. The SSMBE growth of arsenide/phosphide
heterojunctions for electronic device applications has lagged considerably. Hoke and
coworkers [111] reported on the electrical properties o f the InAlAs/InAsojPo7/InP
composite-channel HEMT structure grown by SSMBE. In their work, the 2DEG density
and mobility were 3.39 x 1012 cm' 2 and 4300 cm2/Vs at 300 K, and 3.31 x 1012 cm' 2 and
8520 cm2/Vs at 77 K. Recently, these InAlAs/InAsP/InP HEMT structures grown by
SSMBE have been studied in moe detail by Kim et al. [129] with an emphasis on various
layer designs and the achievable channel conductivity. Cowles et al. [130] demonstrated
excellent device performance with InAlAs/InGaAs/InP double heterojunction bipolar
transistors (HBTs) structures grown by SSMBE.
2.2 Conventional InP-Based HEMTs
2.2.1 Basic Device Structure
A cross section of a standard InP-based HEMT is shown in Figure 1. The basic
structure is comprised of a semi-insulating InP substrate followed by an InAlAs buffer
9
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
layer, an InGaAs channel layer, an InAlAs spacer layer, an InAlAs donor layer, dd, an
InAlAs Schottky barrier layer, and an InGaAs cap layer. The conductive path in the
InGaAs channel layer is generated by the transfer of free electrons out o f the doped
InAlAs layer into the InGaAs channel, where they form a quasi-two-dimensional electron
gas (2DEG) localized within an approximately 15-nm-thick layer at the heterointerface of
the InAlAs/InGaAs layer. Here the conduction band forms a quasi-triangular potential
well. The carriers in the InGaAs are localized in the potential well not by inversion, but
by the electric field that results from the parent donors in the InAlAs and the mobile
electrons in the InGaAs channel. Ideally, this very thin conductive layer in the quantum
well is the only path under the gate electrode where mobile carriers flow. The InAlAs
layers are entirely depleted of mobile carriers so that the gate modulates only the 2DEG.
The modulation-doping concept is described in section 2.2.
In the HEMT, a very thin (about 15 nm) conductive path provides a high carrier
sheet density (for electrons >
1 0 12
cm'2) with, simultaneously, excellent transport
properties (mobility, velocity), even in close proximity to the gate (15 to 30 nm). This
feature translates into clear advantages for high-frequency operation.
10
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
GATE
SOURCE
InGaAs:Si
InAlAs
DRAIN
5
/C a p
Schottky Barrier
InAlAs:Si
InAlAs
Donor
InGaAs
InAlAs
Channel
Spacer
Buffer
InP Substrate
Figure 1. Schematic diagram o f a standard InAlAs/InGaAs HEMT.
2.2.2 Two-Dimensional Electron Gas
If two semiconductors with different bandgaps and electron affinities are brought
in contact (grown on each other), a heterojunction is formed. The properties of the
heterojunction are governed by the energy band lineup at the interface [21]. The
heterojunction exhibits discontinuities in the conduction band (AEC) and valence band
(AEV). The standard heterojunction system used for InP-based HEMTs is InAlAs/InGaAs.
The wide-bandgap semiconductor (InAlAs) has a smaller electron affinity, so that the
conduction band of the narrow-bandgap semiconductor (InGaAs) lies lower in energy.
The valence band of the narrow-bandgap material lies higher in energy than in the wide-
11
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
bandgap material, in this heterojunction (type I heterojunction). The sum of AECand AE%
is equal to the bandgap difference Eg (InAlAs) - Eg (InGaAs). For conventional HEMT
application the wide-bandgap material is selectively n-doped, while the narrow-bandgap
material is undoped. This doping principle is called modulation doping [22]. Due to the
Fermi-leve! alignment, electrons are transferred from the InAlAs into the InGaAs. This
charge transfer leads to a dipole layer formed by the positively charged depletion layer in
the InAlAs and the electron accumulation (negative charge) in the InGaAs. Thus the
conduction band of the InGaAs layer very close to the heterojunction forms a quasitriangular quantum well with quantized density of states perpendicular to the
heterojunction. The transferred electrons are localized in this very thin 10 to 12 nm
quantum well [23] and have two degrees of freedom for motion parallel to the
heterojunction. Thus, they are considered a quasi-two-dimensional electron gas.
2.2.3 Calculation of Equilibrium 2DEG Density
In order to understand a HEMT structure and maximize channel conductivity for
this modulat ion-doped structure, a theoretical calculation is required. Thus, we calculate
an achievable equilibrium 2DEG density in a standard InAlAs/InGaAs HEMT structure
(see Figure 1) in this section.
To describe the electron transfer across the heterojunction exactly, Schrodinger's
and Poisson’s equations have to be solved self-consistently using numerical techniques.
However, analytical models well suited for determining the 2DEG concentration have
12
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
been reported for modulation-doped AlGaAs/GaAs HEMTs [24, 25] and InAlAs/InGaAs
HEMTs [26, 27], Neglecting interface states and assuming full depletion of the n-doped
region and a triangular electron well with two populated subbands, the 2DEG
concentration, ns, can be calculated by equating two expressions self-consistently for ns
deduced from both Poisson’s equation (depleted wide-bandgap material Ndd<i) and Schrti
dinger’s equation (maximum density of states in the quantum well).
The 2DEG density, ns, of the two-dimensional electron gas formed at the
modulation-doped heterostructure (see Figure 2) can be calculated based on the triangular
potential well approximation and Fermi-Dirac statistics. Taking the Fermi level to be
constant throughout the heterostructure in thermal equilibrium, the amount of charge
transferred across the heterojunction interface is obtained by equating the charge depleted
from the wide bandgap layer to the charge accumulated in the potential well. Since for
the depletion layer near the heterojunction the Fermi level is far below the conduction
band, all of the donors are assumed to be ionized and the free-electron density is assumed
to be negligible. In this case, a consequence of the depletion approximation is that the
charge depleted in the wide bandgap layer is given by [25]
qn, = J 2 c.N4( A E , - E n - E , ) * q 1N ) i f -qN ,d ,
13
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(1)
where csand N<j are the permittivity and doping level of the wide bandgap semiconductor,
AEC is the conduction-band discontinuity at the heterojunction, Er is the separation
between the conduction-band edge and the Fermi level in the wide bandgap layer away
from the interface, and Er, is the Fermi level at the interface relative to the conductionband edge o f the small bandgap layer, dj is the thickness of an undoped wide bandgap
layer, the so-called spacer layer, introduced between the doped wide bandgap layer and
the undoped small bandgap layer to reduce the coulombic interaction between the ionized
impurities and the 2DEG in the well.
14
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Doped ^
+
+
+
+
|
+
+
+
+
+
+
+
Undoped
u
o
u
C8
n
+ c /3
'
1
!
i
—
AE,
Hn . + .+ .- f.+ .+ .-+..4t.+ .+ .+ .+ + .+ .
InAlAs *
♦ InGaAs
Figure 2. Energy band diagram of an ungated basic HEMT structure with two subbands
in a triangular potential well.
15
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
From Schrodinger’s equation the maximum charge accumulated in the well is given by
[24]:
nf = DkTIn[ [1 + exp( Efi
kT
E° )].v[l + exp(E-
kT
E' )]}, (2)
where only the first two allowed subbands are included. D = m k T l K t r is the twodimensional density of states per unit energy and per unit area.
In Ref. 24. the subband energy, E„ which is experimentally proportional to a
population of electron in the quantum well, is given by Yim271, with the adjustable
parameter, y,. which may be modified by fitting with the experimentally obtained
subband populations. An empirical parameter, f, with a value between 0 and 1, is
introduced to replace ns by fns [28, 29], where f = 0.5 represents the average field in the
accumulated layer and f =
1
represents the field at the interface.
Since Eq. (2) is a quadratic equation with respect to exp(Er,/kT), the equation for
Ef, can be easily solved in terms of Eo, E|, an ns. The solution is:
Ef l = k T ln (i[-e x p (^ )-e x p (^ :)] +
(3)
J<
[exp(— ) -t- exp(— )]2 + 4exp( ^ - 5 )[eXp ( - ^ - ) -1 ]})
2V
v kT
v kT
v kT
DkT
16
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
In practice, the wide bandgap layer in an actual HEMT is heavily doped to obtain a
higher 2DEG density in the channel. Thus, the Fermi-level separation, Eq, in Eq. ( 1 )
must be evaluated using Fermi-Dirac statistics instead of Boltzmann statistics. The
difference between the Fermi level and the bottom of the conduction band away from the
heterojunction, E q , is given by [30]:
En, -0-Ard/4W(i-Ar,/4)2+4gX
e x p (^ ) =
F kT
where Nd = Nd /
(4)
2g
and g = ge\p(Ed / kT) . Here, Nd is the total donor density, g is the
degeneracy factor of the donor level, Ed is the donor activation energy, k is the
Boltzmann constant, and T is the lattice temperature.
Solving for Nd, Eq. (1) yields
Nd = n] /{2-^|-[A£
- E p - E fi] - 2 n tdt },
q2
(5)
where the term in brackets in the denominator represents the parabolic band bending in
the wide bandgap layer near the interface. Therefore, calculations of Nd as a function of
n, and di are much easier than those of n, as a function of Nd and dj.
17
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Fermi-Dirac statistics were used to calculate the conduction band bending in the
InAlAs donor layer, in which the donor level, Ed of Si is assumed to be 10 meV [31]. The
subband energy levels are analytically described in terms of the 2DEG charge density, ns,
i.e.,
(6)
E, = f ( — ?Y3( - x — 0 + - ) Y 3n?
2m2
3 e2
4
where mj
is the effective mass of electrons in InGaAs channel, e2 is the InGaAs
dielectric constant, and h (= h/27t) is the Planck constant. In our calculation, m'2 = 0.042
m0, f = 0.5, and e2 - 13.9 e0 were used.
Schubert et al. [32] calculated the approximate spatial width of the triangular
potential well with the electron de Broglie wavelength
A(E) = h / ( 2 E m ' / 2.
(7)
where the electron de Broglie wavelength is matched to the potential well according to
the condition
(l/2)(i+l)A.(E0 = Zi(Ei), with i = 0,1, •••
18
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(8 )
This condition means that the width of the ith band is (i+ l)/2 times the electron
wavelength associated with i band.
In designing HEMTs, a high shallow donor concentration is usually employed
( N d2 1018 cm'3), and the semiconductor is therefore degenerate. This places the electron
Fermi-level above the InAlAs conduction band minimum. An effective minimum
thickness (dm) o f the uniformly doped n-InALAs layer is required to supply the maximum
2DEG carrier concentration (ns) satisfying the simple relationship, dm = n t I N d. The
Debye length, L d , represents a transition between the end of the depletion and the
beginning of the neutral region near the conduction band minimum in the n-InAlAs layer.
e(kT / q)
(9)
For design purposes of HEMTs, a highly doped n-InAlAs layer with a minimum
thickness of (L d + dm) can be used for the selected doping concentration that would result
in the desired 2DEG concentration (ns) as depicted in Figures 3 and 4.
From Eqs. (1) and (2), the most important dependences of n» on the layer structure
parameters can be summarized as follows:
( 10)
19
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Eq. (10) reveals that nscan be maximized by increasing the doping concentration, Nd, and
the conduction band discontinuity, AEC. However, the square root in Eq. (10) also makes
it clear that ns can increase monotonically, but shows some saturation for large of Nd and
AEC.
The 2DEG density, n„ in the basic InAlAs/InGaAs HEMT was numerically
determined using the Newton method from Eq. (1) as a function of doping density, Nd,
with the spacer thickness dj as a parameter. The Fermi energy En (see Figure S)
calculated from Eq. (3) is not significantly different from the exact Ef self-consistantly
determined by Ando et al. [33].
Figure 3 shows the theoretically achievable 2DEG density and required minimum
donor layer thickness for a standard InAlAs/InGaAs HEMT structure with various spacer
thicknesses and doping concentrations. As can be seen in this Figure, the 2DEG density is
proportional to the square root of the doping concentration, Nd, and approaches a
different limit for various spacer thicknesses. For example, although a modulation-doped
structure with spacer thickness of 100 A is highly doped at greater than 1 x 1019 cm'3, the
ultimate 2DEG density that can be achieved is less than 2 x 1012 cm'2. For a structure
with a doping density o f 8 x
1 0 18
cm' 3 and a spacer thickness of 20 A, a 2DEG density of
3.1 x 1012 cm' 2 is calculated. As previously stated, the minimum donor layer thickness
was defined as the sum o f the Debye length (Ld) and effective minimum thickness (dm)
of the uniformly doped n-InAlAs layer required to supply the maximum 2DEG earner
concentration (n,), d m = r t , / N d. Therefore, the minimum donor layer thickness is
20
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
inversely proportional to the doping concentration and spacer thickness and proportional
to the 2DEG density.
Figure 4 shows the theoretically achievable 2DEG density and required minimum
donor layer thickness for a standard InAlAs/InGaAs HEMT structure with various spacer
thicknesses and a doping concentration of
8
x 1018 cm'3. As shown in this figure, a
minimum donor layer thickness of 56 A is calculated.
The subband energies and Fermi level energy in the quantum well vary according
to the population of carriers in the well, i.e., they are exponentially proportional to the
2DEG density confined in the channel. As shown in Figure 5, the subband energies and
Fermi level energy in the triangular well of a standard HEMT structure have the same
trend, as expected. In addition, Figure 5 shows that a 2DEG density of 3 x 10* 2 cm' 2 can
be achieved with two populated subbands.
Figure
6
shows the spatial width of subbands vs. the 2DEG density calculated
with Eqs. (7) and ( 8 ). Despite the use of simple relationship (Eq. (7)), these results
present good insight into understanding the trends of the spatial width of the subbands
related to the change of the 2DEG density in the triangular channel of an InAlAs/InGaAs
HEMT. As can be seen, a Zi (the spatial width of the second subband) o f about 120
Acorresponds to a 2DEG density of 3 x 1012 cm*2.
21
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Lattice-matched InAIAa/lnGaAa triangular wall HEMT
100 80 60 40
20 d 3 0 [angstroms]
100
500
300 K
o
T—
300
X
£(A
C
■0o)
O)
a
o
a
-- -
200
Exact Ep
(Y. Ando et al.)
100
,dt = 0 [angstroms]
Minimum donor thickness [angstroms]
400
100 80
0
10
20
30
40
50
2DEG density, n. ( x 10n cm’2)
S
Figure 3. Two-dimensional electron gas density at the channel and minimum donor layer
thickness in the lattice-matched InAlAs/InGaAs HEMT as a function o f doping density
with the spacer thickness dj as a parameter at 300 K.
22
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Lattice-matched InAlAs/InGaAs triangular well HEMT
45
uniformly-doped with NQ= 8 x 101* cm"3.
(A
O)
o
T-
C
JO
3.5
in
in
X
£ID
C
0)
T3
o
111
Q
0
)
c
*
£u-
o
c
o
■o
2.5
E
a
E
CM
c
S
0
20
40
60
80
100
Spacer thickness, d [angstroms]
Figure 4. Two-dimensional electron gas density and minimum donor layer thickness in
the InAlAs/InGaAs HEM T uniformly-doped with Nd - 8 x 1018 cm'3 as a function of the
spacer thickness dj.
23
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
600
Lattice-matched InAIAa/lnGaAa MD structure
500
300 K
400
I
300
>.
S’
0
)
c
UJ
200
E0
100
exact E (Y. Ando et al.)
-100
2DEG density, ng ( x 1011 cm'2)
Figure 5. Subband energies (Eo, E i, and E 2) and Fermi energy Ef (calculated from Eq. (3))
vs. 2DEG density. Ef (dotted line) s e lf- consistently calculated in Ref. 33 is compared.
24
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
600
Spatial width of subbands [angstroms]
300 K
S00
Electron da Broglie wavelength * h/ (2 E m’)1/J
400
Z (E ) - (1/2) (i ♦ 1) (h/ (2 E m )”*) with i « 0,1, 2
300
200
100
0
2DE6 density, n# ( x 1011 cm*2)
Figure 6. Spatial width of subbands vs. 2DEG density calculated from Eqs. (7) and (8).
25
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2.2.4 Electron Mobility
Low field mobility can be interpreted as the efficiency o f electron acceleration
toward peak velocity. Even for HEMTs with submicron gate length there exists a
mobility-limited region under the gate that has a direct impact on the average speed of the
electrons. For modulation-doped structures, ionized impurity scattering is greatly
reduced, due to the spatial separation of electrons in the 2DEG from their parent donors,
and due to the screening effect caused by the high 2DEG density.
Intensive investigations have been made to determine the mobility limit in InPbased heterostructures, primarily for InAlAs/InGaAs [34, 35]. Figure 7 shows the
mobility calculated for a modulation-doped structure over the temperature range T = 4 K
to 300 K. Polar optical phonon scattering is the limiting mechanism from about 300 K
down to 130 K. Below 130 K, alloy disorder scattering becomes dominant. At room
temperature interaction with polar optical phonons limits the mobility to about 11,000
cm2/Vs. Alloy disorder scattering limits the mobility to values in the range 4 - 7 x 104
cm2/Vs. Acoustic phonon scattering is relatively unimportant. Piezoelectric coupling
contributes the least and is negligible in InAlAs/InGaAs modulation-doped structures
[34, 35]. In actual growth experiments, the low field mobility depends on extrinsic
material properties, i.e., heterointerface quality, doping control, background carrier
concentrations, and alloy structural perfection. These parameters, in turn, depend on
growth conditions, such as growth rate, growth temperature, and V /m ratio.
26
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
•C
Ns
E
u
£
Utlloy
4
30
10
T#mp«r»tur» (IQ
100
300
Figure 7 . Calculated electron mobility in lno.s3Gao.47As lattice-matched to InP as a
function of temperature [34].
27
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2.3
Pseudomorohic InP-Based HEMTs
One important goal in layer structure design is to maximize the channel conductivity.
The conductivity is proportional to the product n, x p. As shown in Eq. 1, the 2DEG
concentration, ns, is proportional to the square of the conduction band continuity. AEe
can be modified by adjusting the compositions o f the materials forming the
heterojunction. In addition, an In-rich InGaAs has higher peak velocity and mobility than
those of lattice-matched InGaAs and thus, has been used to improve HEMT performance.
Modification o f alloy compositions may create a lattice constant mismatch with respect to
the InP substrate. This lattice mismatch leads to strain incorporation in these layers,
which has to be taken into account for HEMT layer structure design.
2.3.1 Critical Thickness
The concept of using strained layers to improve the performance of HEMTs has been
applied successfully to generate a new type of HEMT, the pseudomorphic HEMT.
However, there exists a limitation in the exploitation o f strained layers in HEMTs:
namely, the existence of a critical thickness up to which the lattice strain can be
accommodated without deteriorating the carrier transport properties. Above this critical
thickness, it is energetically favorable for the strain to be relieved by forming
28
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
dislocations. Generally, dislocations may generate recombination centers and act as
scattering centers and thus degrade carrier mobility.
To determine the critical layer thickness, he, with respect to the lattice mismatch,
two theoretical models have been developed [36, 37]. One model considers mechanical
equilibrium conditions [36], and the other, energy balance [37]. The lowest values for/if
correspond to the mechanical equilibrium model. Most values of hc used for HEMT
layer structures are close to this value. However, growth conditions have been found to
lead to variations in hc even for nominally the same layer structures. It is important to
note that the substrate determines the lattice constant, and the layers grown on top
conform to this lattice constant. In the case of InAlAs and InGaAs on InP, this leads to a
tetragonal deformation of their normal cubic crystalline structures. A material with a
lattice constant smaller than that of the substrate undergoes biaxial tension, while for the
opposite case, it undergoes biaxial compression. Up to the critical thickness, hc, the
strain can be accommodated elastically. Such layers are perfect crystals and exhibit
excellent transport properties. They are termed pseudomorphic.
The critical thickness calculated based on both models are shown in Figures 8 -11 for
several materials relevant to HEMT applications: InAsP, InGaAs, InAlAs, and InGaP. To
improve the device performance, compositions may potentially be changed. From Figures
8 -11, it becomes obvious that he imposes stringent limitations on material compositions
that can be exploited to improve HEMT performance.
29
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The critical thickness, hc, based on the mechanical equilibrium model (Matthews
and Blakeslee, M - B model) is given by
( 11)
where v is Poisson’s ratio, / is the lattice mismatch, and b is the magnitude of the
Burgers vector in the alloy; v , / , and b are functions o f the indium mole fraction x. For
the <100> axes
where c*j are the elastic stiffness constants of the strained epitaxial layer. The lattice
mismatch between the constituent layers is given by / = (aB - a A) / a A where aA is the
lattice parameter of the bulk substrate, and the Burgers vector in the alloy, b , is defined
as aB/ y j l . In our calculations, Cjj and b of the strained alloy grown on an InP(lOO)
substrate were interpolated from those of binary materials (see Table 1).
The critical thickness, he, based on the energy balance model (People and Bean, P B model) is given by
30
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[1 + v(x)]32nf(x)
{ b)
31
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
10*
InAsPI InP (100)
7
Critical thickness
P - B model
M - B model
0.1
0.2
0.3
0.4
0.5
0.6
0.7
As mole fraction
Figure 8. Critical thickness vs. As mole fraction in InAs*Pi.x/InP (100).
32
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
0.8
Critical thickness (angstroms)
10
P - B model
M - B model
0.2
0.3
0.4
0.5
0.6
0.7
0.8
In mole fraction
Figure 9. Critical thickness vs. In mole fraction in InGaAs/InP (100).
33
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
10
P - B model
Critical thickness (angstroms)
InAIAs/lnP (100)
M - B model
0.4
0.45
0.5
0.55
0.6
0.65
0.7
In mole fraction
Figure 10. Critical thickness vs. In mole fraction in InAIAs/lnP (100).
34
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
10'
Critical thickness (angstroms)
InGaP/lnP (100)
B model
M - B model
101
0.4
0.5
0.6
0.7
0.8
0.9
In mole fraction
Figure 11. Critical thickness vs. In mole fraction in InGaP/lnP (100).
35
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2.3.2 Energy Bandgap
In strained epitaxial layers, the compressive force or tensile force generated by
lattice-mismatch results in a modification of the electronic properties of the
semiconductor, such as the band structure, as shown in Figure 12. Compressive force
causes the in-plane atoms to move closer together. Thus, both the conduction and valence
band extrema move away from one another, and the bandgap increases. In the case of
tensile force, the in-plane atoms move apart, and the opposite situation occurs, i.e., the
bandgap decreases. Analysis shows that the valence band can undergo significant
changes. This is due to the presence of heavy-hole (large effective mass) and light-hole
bands in most compound semiconductors, which are degenerate at the zone center, as
shown in the center diagram of Figure 12. Under uniaxial tensile strain (exx > 0), the
light-hole band moves up faster than the heavy-hole band. In the uinaxial compressive
strain case (exx < 0), the light-hole band moves down away from the heavy-hole band. In
both cases, the degeneracy in the zone center is removed. The conduction band, and
therefore the electron properties do not change appreciably, except that the conduction
band moves up with compressive strain and moves down with tensile strain. Such strain
effects on the energy levels can be mathematically described following the treatment
given below.
When growing on an InP buffer, a strained layer such as InAsxPi.x sustains a biaxial
in-plane compression and a corresponding extension along the <100> growth direction.
The in-plane lattice constant o f such strained layers, ol is simply given by
36
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
fli = (a\d\ + a2d2)/(d\ + d2)
(14a)
where 1,2 denote the lattice constants and thicknesses of the strained alloy and InP layers,
respectively. In general, the InP buffer layer is much thicker than the strained layers (d2
» di), so we obtain a± = a2 = ai„p. If we assume the axis z is parallel to the growth
direction, <001>, the strain tensor e,j has only diagonal components, i.e.,
(14b)
(14c)
eu = 0
(14d)
where cj1
, 21, cf^1 are the elastic constants of the strained alloy and InP, respectively.
Equations (14a)-(14d) define the alloy strain tensor completely.
A 6 x 6 strain Hamiltonian is an accurate descriptionof the valence band splittings
under high stress, as given by Poliak
[38]. This strainHamiltonian
can be easily
diagonalized and yields three eigenvalues at k = 0 (each doubly generate). We obtain
X = <?(x) (heavy hole),
(15a)
I = - j [ A ( * ) + <*(*)] i j p w M 2
(+ for light hole and - for split-off),
(15b)
37
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
= - { A £ , = -b(es - e„) = 6[(c,, + 2cn ) / c,,]eu ,
(15c)
(15d)
I = A+ £ “(*)+ A£w,
where A is the shift of the valence-band edge due to the shear strain, A is the straindependent valence-band edge with the conduction-band edge as origin, AEsis the shear
term in the strain Hamiltonian, b is the valence-band deformation potential associated
with <100> distortions, and A£w is given by
= a' ( ex,+ e >y + e z ) = 2fl"[(cn - Cn ) l c u ]e„ ,
(15e)
where a is the hydrostatic deformation potential in the valence-band. Similarly, the shift
in the conduction-band edge due to the hydrostatic deformation is
AEe =2a'[(cu - c l2) / c u ]eu .
(150
Here, a = a' + a is the total hydrostatic contribution, which can be obtained from the
pressure dependence of the bandgap, dE° / d P :
a = a +a
(c„+2c,
(15g)
dP
38
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
E “ (x) is the composition-dependent bandgap of unstrained InAsxP|.x bulk material.
The parameters of the binary materials are summarized in Table 1. A linear
interpolation is used to obtain the parameters of alloy.
The strain-dependent energy gap is calculated as
Eg = E e- A = E° ' +AEc +& E H - A + E ° g(x).
(16)
We chose the bottom of the conduction band as the energy origin, so E° = 0. AEc
and AE „ are the band-edge shifts arising from the hydrostatic strain for the conduction
and valence bands, respectively. Only the sum AEe + AEH is observable experimentally.
Sometimes, two thirds of such a change is assigned to the conduction band and one third
to the valence band [39] (a = 2 a /3 , a = a / 3). The choice of a and a
will not
influence the final results in Eq. (16).
(17)
Et = 2a[(c,, ~ c n ) / c u ]eu - A + E ° (x)
Figure 13 shows a comparison of the strain as a function of composition in some
InP-related alloys: InAsxPi.x, Ini.xGaxAs, Ini.xAlxAs, and Ini.xGaxP /InP(lOO). The strain-
39
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
dependent bandgap, Eg, of bulk InAsJV,,, Ini.xGaxAs, Ini.xAlxAs, and Ini.xGaxP on an
InP(lOO) substrate is shown in Figure 14, 15, 16, and 17, respectively, as a function of x
at T = 300 K.
40
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
TENSION
UNSTRAINED
COMPRESSION
'LH
HH
HH
LH
Figure 12. The modification of energy band structure in a strained epitaxial film.
41
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 1. Properties of m -V binary materials at 300 K. [123].
Parameters
GaAs
Lattice constant, a (A)
Direct band gap, Eg (eV)
Effective mass (x m0)
Electron, me
Light hole, mm
Heavy hole, mnh
Elastic stiffness constant
Cii (x 1011 dyn/cm2)
C 12 (x 1011 dyn/cm2)
Deformation potential
Hydrostatic, a (eV)
Shear, b (eV)
Dielectric constant, es
Spin-orbit splitting, Ao (eV)
GaP
Materials
InAs
6.0584
InP
AlAs
5.6622
5.653
3
1.42
5.451
2
2.74
0.36
5.868
7
1.35
0.067
0.074
0.62
0.17
0.14
0.79
0.0231
0.012
0.34
0.076
0.026
0.472
0.149
0.09
0.76
11.88
5.38
14.12
6.253
8.329
4.526
10.11
5.61
12.5
5.34
-9.77
-1.7
13.1
0.34
-9.6
-1.65
11.1
0.127
-6.9
-1.7
14.6
0.38
-7.4
-1.7
12.56
0.11
-8.11
-1.5
10.06
0.28
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3.02
8
Strain comparison
6
4
In Ga As
M
Strain [%]
2
x
In AlAs
0
InP (100)
-2
InAs P
-4
-6
-8
0
0.2
0.6
0.4
0.8
1
x
Figure 13. Strain comparison of some InP-related alloys.
43
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1.6
300 K
1.4
1.2
I
a.
Q
O)
Strain [%]
SPLIT-OFF
an>
>.
?
c
0.8
UJ
0.6
LH
STRAIN
HH
0.4
UNSTRAINED
•3
0.2
0
0.2
0.4
0.6
0.8
1
As composition in InAsP alloy
Figure 14. 300 K energy bandgap and strain of InAsP grown on InP (100) vs.
composition.
44
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
300 K
In1«EGaXAa/lnP
1.4
SPLIT-OFF
1.2
UNSTRAINEI
I
1
a
n
HH
STRAIN
0.8
Strain [%]
ra
O)
■CD
•O
>.
S'
C
Ui
LH
0.6
0.4
Lattice-matched to InP
(x = 0.468)
0.2
-8
0
0.2
0.4
0.6
0.8
1
Ga composition in InGaAs alloy
Figure IS. 300 K energy bandgap and strain of InGaAs grown on InP (100) vs.
composition.
45
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
300 K
In Al As/lnP
2.5
I
o>
TC3
(0
A
- 0
SPLIT-OFF
>*
Strain [%]
STRAIN -
a
(Q
fc
HI
LH
Lattice-matched to InP
(x = 0.478)
0.5
UNSTRAINED
0
0.2
0.6
0.4
0.8
1
Al composition in InAIAs alloy
Figure 16. 300 K energy bandgap and strain of InAIAs grown on InP (100) vs.
composition.
46
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3.0
300 K
Ini - i Gai P/lnP
UNSTRAINED -
STRAIN
2.5
I
ID
-O
HH
Strain [%]
a.
n
o>
XI
c
SPLIT-OFF
2.0
>»
?
c
UJ
•8
0
0.2
0.4
0.6
0.8
1
Ga composition in InGaP alloy
Figure 17. 300 K energy bandgap and strain of InGaP grown on InP (100) vs.
composition.
47
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2.3.3 Photoluminescence Transition Energy
After obtaining the bandgap for heavy and light holes, the subband energies are
calculated in strained multi-quantum wells in the regime of the envelope function method
[40-42]. Only the T point is considered with k± = 0 . Kane’s three-band model is used.
The dispersion relation of the conduction subbands is determined by the following
equations [43]:
cos[q(LA + L B)] = cos(KALA)cos{KBLB) - + ( £ + \/£)$'\n(KAL A)sin(KBLB), (18a)
4 = K bM a ( E ) / K aM b( E).
(18b)
M . = 2 P - -------------------- £ —
■ ----- ------------— .
(£ + £ ; , - \ A E )(E+E'; a + a a ) - h a e r
(18c)
M ®B = 2P:
- v ^ e ;b + t A g - f A £ , fl
' „
:------------j-r ,
/ c
r "
a f f l \/ r
1/
C "
a
\
1 / a c H \ (£ - 1V,/ +x £"s
- 41 A
£f )(£ -_V s+
E'"B
+1 As)
- 4(A
£ f)‘
(£ - A£* )[(£ + E; a -
e
4
A£,4)(£ + £;, + A, ) -
4
(A £ "): ]
(18d)
(18e)
= r r Ki P 2[ E+ E: A+ } A A- i AE?}
(E-V,-AEB
H ) [ { E - V s +E';B - ± A E ? ) ( E - V s + E tuB+ A B) - ± { A E ? ) 2]
= t r K l P 2[ E - V s+E;B+ ± A B-i;AE!)
(180
In Eqs. (18a)-(18f). the unstrained conduction-band edge is chosen as the energy origin.
The relations between V,, V F, V Sand E f°A, E°gB, A A, A a are shown in Figure 18.
48
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The heavy-hole subband is treated separately and described with a standard
single-band Kronig-Penney model since it is decoupled from other states. The final
formulae are similar to the unstrained ones, but the strain-dependent band-edge shift must
be included.
In Eq. (18), the conduction-band discontinuity V,(= AEC) is used as an adjustable
parameter. The temperature dependence of the lattice constants is neglected and a linear
interpolation is used to calculate the lattice constant for alloy. The strain tensor, etj, is
obtained from Eqs. (14a)-(14c). The shear and hydrostatic parts of the energy shift are
given in Eqs. (15c) and (15e), respectively, and the parameters in Table 1 are used in the
calculation. Since the InP buffer layer and barrier layers are thick, the InP barrier is
assumed to be strain-free.
After obtaining the subband energies for the conduction and valence band strained
quantum well, the photoluminescence (PL) transition energy in a quantum well is
determined by
C-mH(L) = & f +
0 ^ )
where Enc and E ^ d are, respectively, the confinement energies o f the nth conduction
and mth valence heavy (light) hole subbands, E, is the bandgap of strained well material
(InAsP), and E ]° is the two-dimensional exciton binding energy.
49
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
B
■
*
r
<
\
E°
^ gB
E°
gA
\__
...
A „
.............. 1
I
p
A Aj
11
1
V.
\ !..................
« ...
i
Figure 18. Relations between V, , Vp, Vs and E 0
^ , £°a , A^, Aa for an A-B
heterojunction system.
50
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2.4 InP-Based Power HEMTs
2.4.1 High Frequency Power Transistor Design
In small-signal transistor applications such as low-noise amplification, the device
properties are of interest only near the bias point. This is completely different for the
power transistor: large-signal operation means that the gate voltage may sweep over the
full gate voltage range between maximum gate forward bias and pinch-off (or even
beyond) [44].
High gain is realized by short gate lengths, resulting in high transconductance and
reduced input capacitance, and by using thin, high carrier-density conduction channels
within which carrier flow is confined and efficiently modulated by the gate. Figure 19
schematically shows the case of class A power operation with an idealized, purely
resistive load. In class A, the large-signal bias point is such that Id is about half the full
channel current obtained under maximum forward gate bias, If, and that V d, is also about
half the maximum value that can be applied, Vbd - Vice*, where
is the knee voltage,
and Vbd is the drain-to-source breakdown voltage when the gate is biased to pinch-off.
The maximum output power for class A operation is given by
Pmax = ~ (Vbd —Vione) If
o
(20)
51
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
According to this relation, the usable drain voltage and drain current swings have to be
maximized for maximum output power. This means that a good power device should
achieve high values of VM - V W and If simultaneously.
Another important property of a power transistor, the class A power-added
efficiency, PAE, is defined by
PAE =
(Pou, - P,„)/ Pdc = Pou,(l - G’1)/ Pdc = | (1 -
Vds)(l - G'')
(21)
where Pout and Pj„ represent the high-frequency output and input power, G = P 0Ut/Pin is the
power gain, Pdc is the power of the dc bias, and Vds is the drain voltage bias. Efficiency is
an important figure of merit because high PAE means small power losses, and therefore
small self-heating effects and long battery operation. The theoretical limit in class A
operation of PAE is 1/2, and is reached for infinite gain, G. The gain at a particular signal
frequency is bias-dependent. This imposes an additional boundary condition for the
proper choice of the large-signal bias point.
From a small-signal model, the unity current gain cut-off frequency, ft, and
maximum oscillation frequency (or power gain cut-off frequency),
fmx. are
approximately given by
(22)
52
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
where Cg is the gate capacitance, v, is the average saturation velocity of device, Lg is the
gate length, go is the output conductance, and Rm is the input resistance of FET. Thus, the
speed of power transistors is limited by ft and fiw.
The basic requirements for a good power transistor for microwave and millimeter
wave applications are, therefore, the following: low knee voltage, high gate-to-drain and
drain-to-source breakdown voltage, high current density, and high gain. Depending on
the frequency of operation, certain device parameters become more critical.
Multiple interacting factors are involved in optimizing the performance of power
transistor. The optimization goal for a power transistor is to achieve an output power
level with a particular gain and power-added efficiency. The interaction between different
factors can be very complex and optimization of some of these factors are sometimes in
conflict, resulting in more difficulty in the optimization of devices for power applications
as compared for low noise applications.
53
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Bias for Class A
min
V,
Figure 19. Schematic representation for a FET operating in class A.
54
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2.4.2 InAlAs/InGaAs Power HEMTs
The InAlAs/lnGaAs on InP HEMT is a nearly ideal device for the realization of the
low knee voltage and high gain required for power applications. This is due to the high
mobility, high sheet charge density, and high velocity of the channel electrons [45].
These factors result in a higher microwave and millimeter wave gain and lower knee
voltage than other InP-based FETs and comparable GaAs-based three terminal devices.
However, InAIAs/InGaAs HEMTs have one fundamental weakness when compared to
GaAs-based HEMTs. This is their low breakdown voltage.
There are two important breakdown voltages that determine the power performance
of InP-based HEMTs: the off-state breakdown voltage and the on-state breakdown
voltage. The three terminal off-state breakdown voltage is the drain-to-source voltage of
the transistor with the device pinched-off. The off-state breakdown voltage of
InAIAs/InGaAs on InP HEMTs is believed to be dependent on a combination of
thermionic-field emission and tunneling of electrons from the gate into channel, and
impact ionization in the channel [46]. The on-state breakdown voltage in these HEMTs
has been shown to be dominated by impact ionization in the channel layer [47], and is
typically low due to the low breakdown field of the narrow bandgap Ino.53Gao.47As
channel.
Effort to improve the power performance of InP-based HEMTs can be divided into
seven categories: Schottky layer designs, channel layer designs, reduction of the electric
field between the gate and drain, reduction of the parasitic gate leakage, buffer layer
55
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
designs, hole barrier approaches, and doping schemes. They will be described in
following sections.
2.4.2.1
Schottky Layer Designs
Four approaches have been taken in modifying the Schottky barrier to improve the
gate-to-drain breakdown voltage of InP-based HEMTs as compared to the conventional
lattice-matched InAIAs/InGaAs HEMTs.
In the first approach the A1 mole fraction in the InAIAs Schottky layer can be
increased from 48 9c (lattice-matched case) to as high as 70 % [48]. The increase in the
Al mole fraction of the InAIAs layer increases the bandgap of the Schottky layer,
resulting in improvement of the turn-on voltage as well as the gate-to-drain breakdown
voltage. The Schottky barrier height can be improved from 0.6 V (Al= 48 9c) to 0.9 V
(Al= 63 9c) without deteriorated two dimensional electron gas properties. By using an
Ino.3 Alo.7 As or Ino.75 Alo.2sP barrier layer, the breakdown voltage can be doubled from 6 V
to 12 V for an optimized HEMT with a 0.25 pm gate, with a current density of 800
mA/mm and a current gain cut-off frequency of 86 GHz [49].
In a second approach, alternative material systems, such as InAlP [50], InAIAsP
[51], and InGaP [52], have been used instead of InAIAs for the Schottky layer to improve
the gate-to-drain breakdown voltage. The advantage obtained from the use of an
alternative material is that they have lower aluminum content, and can be grown with less
strain on an InP substrate. However, results are inconsistent for the use of these less
56
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
mature materials. The Schottky barrier heights determined by I-V measurements were
0.42, 0.65, and 0.63 eV for Ini.xGaxP with x=0, 0.25, and 0.5, respectively [53]. Thus,
Ini.xGaxP with a thickness of lOnm x d x l5 nm and a Ga mole fraction of 15 % x x x 25
% has not dramatically improved the gate-to-drain breakdown voltage of InP-based
HEMTs. However, a pseudomorphic InP/InxGai.xAs/InP (x a 53 %)/InP backside-doped
split-channel HEMT with a strained undoped Ino.75Gao.25P Schottky barrier was reported
[54], A 0.25 pm gate length HEMT demonstrated a transconductance of 510 mS/mm and
a current cut-off frequency of f, = 102 GHz coupled with an off-state drain-to-source
breakdown voltage of 10 V. At 18 GHz a power gain of 25.6 dB was measured,
corresponding to an fmax > 200 GHz. However, these novel material systems are not
mature and require more study to optimize growth conditions and achieve more
consistent results.
In the third approach, a pn-junction is formed instead of the Schottky gate to
modulate the electron density in the channel. In the so-called junction-HEMT (JHEMT)
approach, a heavily doped p-type InAIAs layer is grown on the top of a standard HEMT
structure [55]. In this structure, the breakdown voltage is increased to 22 V with different
source-drain spacings. In addition to improving the breakdown voltage of the HEMT, this
approach has the advantage of achieving very uniform threshold voltages across the
wafer. This technology is, however, immature and requires further optimization of the
layer and device structure to achieve good device operation at millimeter wave
frequencies.
57
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
In the final approach, different metal systems can be considered to improve
Schottky barrier heighs. Recently, platinum (Pt) instead of Ti has shown excellent
promise as a Schottky contact to Ino.5 2 Alo 4 8 As, with barrier height of 0.82 - 1.09 eV [56],
leading to the another choice to improve the breakdown voltage.
2.4.2.2 Channel Layer Designs
In the standard HEMT layer design, a lattice-matched InGaAs channel with a
thickness of 20 nm to 40 nm is used [57]. This channel design suffers from low on-state
breakdown voltage due to the low bandgap of the channel. One approach to increasing
the bandgap of the channel is to increase the Ga percentage in the channel, such as using
an Inu-iGaoftAs channel. This Ga-rich InGaAs channel has a higher energy bandgap
resulting in the reduction of impact ionization rate. However, adding Ga to the channel
lowers electron mobility. Thus, there is a trade-off between mobility and energy bandgap
in determining a suitable mole fraction of Ga.
Using the wider bandgap InP as the channel material also improves the drain-tosource breakdown [58]. InP has a very high electron saturation velocity, but a lower
mobility than lattice-matched InGaAs, resulting in a higher sheet resistance and a higher
source resistance compared to In0 .53 Gao4 7 As channel HEMTs. With a highly doped InP
channel, a typical maximum drain-current density of 430 mA/mm and a gate-to-drain
breakdown voltage of 12 V have been achieved in InAlAs/InP HEMT. It is difficult,
58
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
however, to achieve a low contact resistance to InP without doping the InAIAs Schottky
layer. This results in lower gain at millimeter wave frequencies and lower PAE for InP
channel HEMTs.
In the composite channel HEMT approach, a combination of a thin layer of
lattice-matched InGaAs and InP is used as the channel [59]. This structure has the
advantages of both materials (high mobility of InGaAs at low fields, and high breakdown
and saturation velocity of InP at high fields). By further reducing the thickness of the
Ino.53Gao.47As channel it is possible to efficiently increase the bandgap of channel due to
energy quantization in the quantum well [60]. Composite channel HEMTs with
In0 s?Gao 47 AS channel thicknesses as small as 3 nm have been reported for high
performance 0.15 fim InP-based power HEMTs with on-state breakdown voltages of
more than 8 V [61 ]. As the In0 s.iGao47A s channel thickness is decreased, both the gate-todrain breakdown voltage and the drain-to-source breakdown voltage are increased. By
decreasing the channel from 30 nm to 3 nm the breakdown voltage can be increased from
approximately 6 V to over 10 V [62]. However, in order to optimize power performance,
we need a trade-off between breakdown voltage and channel thickness because a channel
with reduced thickness results in lower drain saturation current density.
InAsxPi.K also can be used as a single channel material or a composite channel
material with InP for microwave and millimeter wave power applications. The InAs0 .6 Po4
alloy has a T-L valley energy separation and a peak drift velocity estimated to be larger
than 0.6 eV and 3xl0 7 cm/s, respectively [63]. This material has a greater lattice constant
59
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
than InP. resulting in the formation of compressive strain in the InAsP. Thus one can
achieve an enhancement effect in the energy bandgap from the compressively strained
InAsP. Hong et al. demonstrated InAso6 Po4/lnP quantum-well pseudomorphic HEMTs
(PHEMTs) grown by metalorganic chemical vapor deposition (MOCVD) [64]. A drain
saturation current density of approximately 700 mA/mm at Vgs = 0 V was obtained in this
device. The electron saturation velocity estimated from this saturation current was
approximately 2.0x 107cm/s. The ft was maximum at 56 GHz at 2 V, whereas the fmax was
maximum at 61 GHz at 4 V. However, this InAsP/InP PHEMT has a higher sheet
resistance than the lattice-matched InGaAs HEMTs due to a relatively low sheet charge
density and low mobility.
InAIAs/InAsP can be a better heterojunction system for power applications
because of a higher confinement of carriers in the channel resulting from a higher
conduction band discontinuity compared to InP/InAsP material system. In addition, this
system exploits the higher Schottky barrier height of InAIAs compared to InP [65]. In
this thesis, we report a lower sheet resistance than that of a InP/InAs0 .6 Po4 HEMT in
InAIAs/InAso6Po4 PHEMTs grown by SSMBE.
The wider bandgap (0.95 eV) Ino.73 Gao.27 Aso.6 P0 4 is an InP lattice-matched
channel material and has been investigated for achieving high breakdown voltage without
sacrificing the low field mobility [66]. This quaternary alloy has a threshold field similar
to that of Ino.53Gao.4 7 As (-3 kV/cm) and a low field mobility of around 6500 cm2/Vs if it
is a high quality. An extrinsic transconductance of 210 mS/mm was obtained for a gate
60
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
length of 1.0 fim. The HEMTs showed improved drain-to-source (> 5 V) and gate-todrain (> 15 V) breakdown characteristics, and a very small output conductance of less
than 2 mS/mm, resulting in a high dc voltage gain of greater than 100. The maximum
available gain was 14.5 dB at 20 GHz. A maximum oscillation frequency fmax of 105
GHz was achieved at the drain bias of 5 V.
Hong et al. have investigated the transport properties and device characteristics of
the Ino4Alo6As/InP PHEMT [67]. A 1.5 jim gate length HEMT showed an extrinsic
transconductance and drain current density as high as 160 mS/mm and 300 mA/mm,
respectively. The HEMT also showed a very small output conductance of less than 2
mS/mm. and a high gate-to-drain breakdown voltage of larger than 15 V.
Besides the three channel design concepts discussed above, different approaches
have been carried out to improve the breakdown voltage as follows. Bahl et al.
demonstrated an increased breakdown voltage of about 13 V with an Ino4|Al059 As/ n+Inu65Ga<U!iAs HEMT [68]. The f, and fmax were 14.9 GHz and 101 GHz, respectively.
These results stem from 1) a thin subchannel used to introduce quantization to increase
the effective channel bandgap, 2) a strained Ino4 iAl059 As to increase Schottky barrier
height, and 3) the elimination of parasitic side-wall gate-leakage.
Another approach is to vary the indium composition in the channel to shift the
maximum of the electron wave function deeper inside the channel, and alleviate the
electric field at the spacer/channel interface. The modification of the quantum-well
channel by grading the composition considerably changes the channel breakdown voltage
61
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
and output conductance characteristics [50]. InP-based HEMTs with graded InxGa|.*As
channel (x = 0.7 to 0.6) and Ino gAlo.iAs Schottky barrier showed an improved breakdown
voltage of 11 V and output conductance of 40 mS/mm compared with HEMT with an
uniform composition (x = 0.7) in the channel (breakdown voltage = 4 V and output
conductance g0 = 80 mS/mm).
2.4.2.3 Reduction of Electric Field at Drain-End
Reduction of the electric field at the drain-end of the device can be achieved by a
laterally extended gate-recess etch process [69] and used to obtain low leakage, low
output conductance, and high breakdown voltage.
Boos et al.. reported a lattice-matched InAlAs/InGaAs HEMT with a double­
recessed gate structure and asymmetrical gate spacing (closer to source) to reduce the
peak electric field at the drain-end. They achieved an improvement in breakdown voltage
[70] with a gate-to-drain breakdown voltage of 16 V and a drain-to-source breakdown
voltage near pinch-off of 16 V.
Another approach to reducing the field is to insert an undoped InGaAs cap under
the conventional InGaAs cap. This has led to an improvement in the gate-to-drain
breakdown voltage for power applications at W-band [71]. In this device, a gate-to-drain
breakdown voltage of 7 V and high drain current density of 880 mA/mm were achieved.
62
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
At 94 GHz, the output power density was 300 mW/mm with 21 % power-added
efficiency.
Ohmic source and drain contacts to HEMTs may be classified as either alloyed or
nonalloyed. Alloyed channel contacts consist of interdiffusing the contact metal and the
semiconductor, favorably modifying the energy band structure. Selective rcgrown ohmic
contacts using MOCVD has been investigated as a means to improving breakdown
voltage due to the alleviation of electric fields at the drain-end of gate [54]. In 0.25 pm
gate length InGaAs/InP composite channel HEMTs, low contact resistance (0.35 Q-mm),
high transconductance (600 mS/mm). high full channel current density (650 mA/mm),
and high peak cut-off frequency (f, = 70 GHz, fmax= 170 GHz) have been obtained. The
gate-to-drain breakdown voltage was improved up to 18 V.
In addition. Shealy et al., demonstrated a high gate-to-drain breakdown voltage
of 31 V, and an off-state breakdown voltage of 28 V for 1 pm gate length InAIAs/InGaAs
junction HEMTs with regrown ohmic contacts by MOCVD [72].
2.4.2.4 Parasitic Gate Leakage
Conventional mesa isolation in InAIAs/InGaAs HEMTs results in the gate
contacting the exposed channel at the mesa side-wall, forming a parasitic gate leakage
path. This may result in excessive gate leakage current and reduced breakdown voltage.
To eliminate the parasitic gate leakage path various approaches such as use of air-bridges
63
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[71], ion implantation [73], and recessing the channel edge into the mesa side-wall using
a selective etchant for InGaAs over InAIAs [74] have been utilized.
2.4.2.5 Buffer Layer Designs
Various approaches for the optimization of buffer layer, including a low
temperature growth and superlattice buffer, have been developed to reduce output
conductance, resulting in the improvement of breakdown voltage.
Pao et al.. proposed a model (referred to as premature current saturation) for the
physical origin of the high output conductance in the lattice-matched InAIAs/InGaAs
HEMTs [75]. The origin of the high output conductance was explained as result of the
real space transfer of hot electrons in the channel into the buffer layer at high electric
field. This deconfinement process is self-limiting due to the band bending at the buffer
interface resulting from the presence of deconfined electrons in the buffer. lno.52 Alo.4 8 As
buffer layers grown at low temperature have an extremely low electron mobility and
hence low saturation velocity (Vuiaias) resulting in the elimination of the current path in
the buffer layer.
Low temperature (420 °C) growth of the InAIAs buffer layer increases the buffer
resistivity [76]. This results in a large improvement in the breakdown voltage (by a factor
two), and also a decrease in output conductance. Consequently this causes a significant
increase of the maximum frequency of oscillation. The InAIAs buffer grown at relatively
64
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
low temperature (420 °C) exhibited a greater improvement in breakdown voltage and
output conductance than that of various superlattice buffers.
2.4.2.6 Hole Barrier Approaches
In order to obtain high breakdown voltage and low gate leakage the use of an
Ino sGaosP hole barrier between the gate and channel has been evaluated [77], This spacer
successfully suppresses hole-tunneling to the gate, resulting in an improvement of
breakdown voltage and gate leakage current. Moreover, the large bandgap of about 1.9
eV of Ino.5Gao.5 P (AEV = 0.37 eV in the In0 sGao sP/Ino^Gao4 7 As interface) results in
improved gate leakage, output conductance and drain-to-source breakdown voltage. At
the drain bias of 10 V, a voltage gain Av> 100 was attained. A 0.7 pm gate length HEMT
exhibited a transconductance of 360 mS/mm, a maximum drain current density of 550
mA/mm. and a maximum oscillation frequency fmax of 150 GHz.
Another approach is to use an Al-rich InAIAs spacer [78], AlAs spacer [79] and a
mixed InAIAs/InP spacer [80] to improve the breakdown voltage of the devices in taking
advantage of the large valence band offset. Inside the Ino.s2 Alo.4 8 As spacer a 3 nm
strained In0 .1AI0 .9 As layer is placed acting as a hole barrier [78]. While the energy
bandgap of the strained Ino 1AI0 9 AS is Eg= 2.4 eV, a large lattice-mismatch of 3 % has to
be taken into account. In addition, a valence band discontinuity as high as AEV= 0.5 eV
65
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
can be expected. When biased at peak transconductance, a gate leakage current of Ig <
100 nA/mm at the drain bias of 2.5 V was obtained.
2.4.2.7 Doping Schemes
For power applications, the doping level of the InAIAs donor layer under gate
must be low enough to obtain a low gate leakage current and a good device breakdown
voltage. This, however, reduces the total sheet charge density available in the device
channel. In order to achieve a higher carrier density, a doped-channel or doubleheterojunction structure can be used [44]. Delta-doping (or planar doping) is more
favorable than an uniform doping in the donor layer for power applications. The deltadoping layer is a monolayer of Si ( - 0.5 nm) with a doping level of approximately 5x10
l2/cm: located right above the spacer or two third above and one third of total doping
concentrations under the channel in a double doped structure.
In the doped-channel scheme, carriers are transferred to the channel from the
donor layer and the device channel is also doped to obtain a high sheet charge density.
The doped-channel scheme reduces the electron mobility in the device channel. The
doped-channel reduces the peak velocity of carriers due to a higher effective mass of
electron populated in the high subband energy, resulting in the extension of threshold
electric field [81]. In the double-heterojunction power HEMT, carriers are introduced into
the InGaAs channel by doping the InAIAs on both sides of the InGaAs. This results in a
high channel current density and hence a high power-handling capability. However, this
66
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
structure requires precise control of the doping level in the bottom InAIAs layer to
achieve good pinch-off characteristics by avoiding parallel conduction.
Somerville et al., have developed a tunneling-limited model for off-state
breakdown of power HEMTs [82], which is based on the assumption that off-state
breakdown is determined by tunneling and/or thermionic field emission. Their model
demonstrates that the most significant variables in determining the off-state breakdown
voltage of power HEMTs are the sheet carrier density in the extrinsic gate-drain region,
and the gate Schottky barrier height. Therefore, there exists a trade-off between the
breakdown voltage and the 2DEG density.
The factors required to improve the power performance of InP-based HEMTs are:
large Schottky barrier height, large energy bandgap of channel material, high sheet
charge density, high mobility of 2DEG, and low output conductance.
2.4.3 Power Performance
Optimization of the material layer structures has led to the development of InPbased power HEMTs with state-of-the-art power performance at microwave and
millimeter wave frequencies as compared to GaAs-based power HEMTs. Output power
densities of more than 1 W/mm with very high power-added efficiencies (see Table 2)
have been reported at 4 GHz and 12 GHz [83, 84]. Using a composite channel layer
design and by increasing the Schottky barrier height, a gate-to-drain breakdown voltages
67
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
of more than 20 V have been reported for 0.25 pm gate length high-performance InPbased HEMTs [83]. At 60 GHz output powers as high as - 500 mw [3] and power-added
efficiencies of as high as ~ 50% have been reported [3]. Table 2 shows a summary of the
power performance of discrete InP-based HEMTs.
InP-based HEMTs are ideal for millimeter wave power applications where high
gain and power-added efficiency are desirable. Also, these HEMTs are ideal for
applications requiring low voltage (battery operated), and high-efficiency operation [85].
For low voltage operation of power transistors, low knee voltage, high drain current
density, as well as high power gain are essential for achieving high power-added
efficiencies. The efficiency of transmitters is critical in extending the life of batteries for
millimeter wave portable wireless applications. Using InP-based power HEMT
technology we can easily obtain electron sheet charge density of more than 5x10|: cm' 2
leading to transistors with current densities of I A/mm. The high electron density
combined with the high electron mobility in the channel also leads to a low source
resistance and. hence, a low knee voltage. The combination of high current density, low
knee voltage, and high gain make the InP-based power HEMT technology ideal for low
voltage applications.
68
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 2. Comparison of power performance of InP-based FETs. In this table, Lg and Wg
are the gate length and width, respectively. Pou, represents the high frequency output
power.
Channel
f
{GHz)
Wg
Pout
Pout/Wg
(micro- meter)
(mW)
(mW/mm)
Lg
Delta-Doped
4
0.25
2000
Composite
HJFET
Composite +
Regrown Contact
Delta-Doped
Composite
4
4
7
0.25
0.6
0.2
500
130
300
10
10
0.25
0.15
1000
450
Double-Doped
12
0.22
InGaAs MISFET
Double-Doped
lnGaAs(ln 53%)
18
20
20
0.7
0.2
0.15
Composite
InP
20
30
0.15
0.3
InP MISFET
Delta-Doped
Delta-Doped
Delta-Doped
Double-Doped
Composite
lnGaAs(ln 53%)
30
44
60
60
0.3
0.2
0.15
0.15
0.2
0.15
0.1
lnGaAs(ln 67%)
60
0.1
inGaAs(ln 53%)
60
0.22
InGaAsdn 53%)
lnGaAs(ln 67%)
94
94
0.15
0.1
60
1100
600
560
130
270
450
350
380
110
150
300
288
184
200
800
516
20.5
50
39
280
450
100
145
120
75
135
600
225
448
155
2x448
288*
448
145
170
450
50
20.6
14.9
200
80
400
192
150
56
450
150
50
15
100
26
200
50
PAE
w
550
300
1120
1000
900
50
60
63
40
76
450
780
840
730
960
920
645
410
780
620
1450
1200
1800
375
350
320
320
380
412
300
400
480
370
330
300
260
250
59
60
58
50
40
29
47
52
44
46
13
23
20
39
30
20.4
24
30
45
49
36
30
20.4
20
21
33
26
Gain Reference
(dB)
14.1
13.4
15
[861
24.3
[88]
11
12
12.7
11.1
11
3.2
7.1
10.5
10.2
9
[96]
[51]
[83]
[871
[84]
[89]
[90]
[91]
[92]
[58]
5.2
5
4.9
3.6
4.2
4
8
8.6
6.6
4.4
4
2.6
4
6
5
69
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[93]
[94]
[57]
[95]
[3]
[3]
[96]
[74]
[97]
CHAPTER III
GROWTH OF CONVENTIONAL InP-B ASED HEMTs
Since the 2DEG characteristics in MD structures depend on growth method and
conditions, and material quality and structural design, we have studied the SSMBE
growth of conventional InP-based MD structures in order to obtain the optimum growth
conditions and to investigate the influence of material design factors on the 2DEG
characteristics. These optimized growth conditions have also been utilized to confirm the
machine status and test the preparation procedures of InP substrates. In addition, these
baseline 2DEG transport data will be used to analyze the influence of epitaxial layer
design changes on the 2DEG properties in InGaAs/InP and InAsP/InP composite-channel
MD structures grown by the SSMBE with the same growth conditions. The maximum
2DEG density achieved without a degradation of mobility will be compared with the
theoretical equilibrium 2DEG density calculated in section 2.2.3. Furthermore, an impact
of the channel depth on the channel conductivity will be investigated. Finally, the 2DEG
properties of conventional InP-based power HEMT structure, such as doped-channel
InAIAs/InGaAs HEMT structures and double heterojunction InAIAs/InGaAs HEMT
structures will be reported
70
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3.1 Calibration Procedures for InP-Based HEMT Growths
Before HEMT structure growth we must calibrate several parameters for each
material in the structure, including the composition, growth rate, surface morphology,
doping activity, and residual impurity concentration. X-ray diffraction is used to
determine the compositions of InGaAs and InAIAs. The x-ray diffraction peak of these
epitaxial alloys can be shifted away from the InP substrate peak by a lattice-mismatch
between the epitaxial films. Through the x-ray diffraction measurement of as-grown
epitaxial material, we can quantify the lattice match. With this information, Group III
source temperatures are adjusted. This procedure is continuously iterated until a latticematched condition (< 0.3 %) is achieved. In our Riber 2300 MBE system, 500 nm thick
InAIAs and InGaAs samples are usually grown for these material calibrations.
The growth rate of InAIAs and InGaAs epitaxial films is determined through
thickness measurements of calibration samples grown on semi-insulating InP (100) epiready substrate. In order to measure the thicknesses of InAIAs and InGaAs on InP,
approximately half of each sample is covered with black wax and then etched by dipping
in a citric acid-based selective etchant that does not etch the InP substrate. After
removing the black wax with solvents, the thickness of the step is measured by using the
Dektak profilometer machine. In order to obtain good accuracy, 25 measurements are
made and averaged. With this measurement procedure, a variance in InAIAs and InGaAs
thickness of less than
10
% is typically achieved.
71
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The elecirical properties of epitaxial layers, including MD structures, are typically
measured by the Van der Pauw method at 300 K and 77 K. To avoid effects of sample
size in the Hall measurement, all Hall samples (typical size of approximately
1
cm x 1
cm) are patterned into a clover-leaf shape by using a specially designed metal mask and
sand blaster. Ohmic contacts are made via indium with 10 min anneal steps on a hot plate
at 300 °C. A commercial system (Bio Rad 1000) was used for the Hall measurements.
This system was calibrated with standard InAIAs/InGaAs HEMT samples grown and
measured at the Hughes Research Laboratory (HRL). Undoped InGaAs and InAIAs
epitaxial films grown in our MBE system are n-type. In undoped InGaAs layers, the
residual electron concentration is around mid-1 0 15 cm' 3 with a mobility of approximately
6000 cm2/Vs at room temperature. Most as-grown undoped InAIAs layers exhibit high
resistivity. In a few undoped InAIAs samples, residual electron concentrations of mid1014 cm4 and mobilities of about 1000 cm:/Vs at room temperature were measured. The
Si doping conditions are also determined by the Hall measurements of doped samples in
the same system.
72
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3.2 Standard InAIAs/InGaAs MD Structures
Figure
20 shows a schematic diagram of the standard lattice-matched
InAIAs/InGaAs MD structure, which was optimized by Brown et al. [98] at HRL. As
shown in this figure, a standard M D structure consists of a Si doped InGaAs cap (7 nm),
an undoped InAIAs Schottky (20 nm), a Si uniformly-doped InAIAs donor
(8
nm), an
undoped InAIAs spacer (2 nm), an undoped InGaAs channel (40 nm), and an undoped
InAIAs buffer (250 nm). The low field electron mobility of the 2DEG is the most
important figure of merit of as-grown MD structures and is affected by various growth
parameters, such as the growth temperature, m /V flux ratio, growth rate, and
heterointerface roughness.
To optimize the electrical properties of the 2DEG in the standard lattice-matched
InAIAs/InGaAs MD structure, various experiments were designed to assess the influence
of As4 flux, growth rate, superlattice buffer and buffer thickness, growth temperature,
interface interruption, planar doping scheme, and channel depth (a summation of film
thicknesses grown on the channel) on the 2DEG conductivity.
The MD structures were grown on epi-ready semi-insulating lnP:Fe (100)
substrates, indium bonded or nonindium bonded, in a Riber 2300 equipped with the
valved-cracker arsenic source (EPI RB500VP). An As4 over pressure of 1.5 x 10' 5 Torr
with the cracker temperature of 600 °C was used. A growth rate and substrate
temperature were 1 pm/hr and 500 °C, respectively. There was no any intentional
interface interruption, i.e., no pausing at heterointerfaces. The substrate temperature was
73
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
referenced to the (2 x 4) - (4 x 2) InP surface transition as determined by reflected highenergy electron diffraction (RHEED) patterns, where the transition is assigned a
temperature of 540 °C, with all lower temperatures (about 60 °C) measured by
thermocouple with respect to this transition temperature.
To reduce oval defects
generated by indium droplets, a dual filament effusion cell for the indium source was
used instead of a conventional effusion cell. The oval defect density was improved from
103
to
10
* —1 0 “ cm'2.
74
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
InGaAs:Si(7xl0ls c m )
InAIAs
InAlAsiSKSxlO11* cm'5)
InAIAs
InGaAs
InAIAs
InP substrate
7 nm
nm
8 nm
2 nm
40 nm
250 nm
20
Cap
Schottky barrier
Donor layer
Spacer
Channel
Buffer
Figure 20. Schematic diagram of standard lattice-matched InAIAs/InGaAs MD structure.
3.2.1 Optimization of 2DEG Conductivity
To improve the conductivity of the 2DEG in a standard lattice-matched
InAIAs/InGaAs M D structures (Figure 20), several experiments were carried out. First of
all, the influence of As4 flux on the 2DEG conductivity was investigated.
Figure 21 shows Hall data measured at 300 K and 77 K. respectively, for a
standard InAIAs/InGaAs MD structure grown with various As4 fluxes monitored as
balance equivalent pressure (BEP). As shown in these figures, variation in As4 flux does
not significantly change the 2DEG mobility of InAIAs/InGaAs MD structures.
In addition, the impact of a growth rate and the use of a superlattice buffer on the
2DEG conductivity of the InAIAs/InGaAs MD structure have been investigated. Table 3
presents a comparison of Hall data for a standard InAIAs/InGaAs MD structure with a
250 nm InAIAs buffer and a four-period InAIAs
(8
nm)/InGaAs (2 nm) superlattice (SL)
buffer and half the normal growth rate. The 2DEG mobility is slightly decreased by - 10
75
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
9c by introducing the SL buffer and lower growth rate in comparison to that of the
conventional InAIAs/InGaAs MD structure.
Figure 22 demonstrates the typical 2DEG properties of a standard lattice-matched
InAIAs/InGaAs M D structure (Figure 20). In Figure 22, the variation of the 2DEG
density stems from intentional changes in the Si doping density. When a Si doping
density of 8 x 10
Io
^
cm' is used in an uniformly-doped InAIAs donor layer of
8
nm, the
300 K 2DEG density and mobility of the standard InAIAs/InGaAs MD structure was
typically in the range of 2.5 to 3 x 1012 cm' 2 and 9500 to 10500 cm2/Vs, respectively. At
77 K. the 2DEG mobility was typically three times that of the room temperature mobility.
Structures grown at Hughes Research Labs (HRL) are compared. As previously discussed
in the section 2.3 (see Figure 3 and 4). the equilibrium 2DEG density calculated in the
standard InAIAs/InGaAs MD structure with No =
8
x 1018 cm 3 and the spacer thickness
d, of 2 nm was 3.1 x 1012 cm'2. This result indicates that the 2DEG density of 3 x 1012
cm' measured in the standard InAIAs/InGaAs MD structure has been optimized to the
theoretically estimated ns.
76
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 3. Comparison of Hall data of conventional InAIAs/InGaAs MD structure with 250
nm InAIAs buffer and InAIAs/InGaAs MD structure grown with half growth rate ( - 0.5
pm/hr) and InGaAs/InAIAs superlattice buffer.
Sample
A 180
AI833
T Si
(°C)
1208
1185
BEP
torr)
1.5
0.85
As4
(x
1 0 5
Hall mobility
(cm2/Vs)
300 K 77 K
9586 28211
8741
25226
Rs
@ 300 K
(ti/sq)
322
384
2DEG density
(xlO 12 cm’2)
300 K
77 K
2.0
2.3
1.9
2.2
a InAIAs/InGaAs MD structure grown with half growth rate and superlattice buffer.
77
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
20
15
10
i
o
1
06
15
Ai
2
26
3
B£P i i iO *to fr)
50
77 K
40
30
20
10
0
0.5
1
1.5
2
3
A t, B E P fH O 'to r r )
Figure 21. Hall data for the standard InAIAs/InGaAs M D structures versus As4 BEP flux:
(a) 300 K and (b) 77 K Hall data.
78
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2DEG tftn n ty (« 10,; cm
77 1C
40
>
I
"o
(b)
3
1
1.5
2
25
3
3.5
4
4.5
20EG dtntity ( i 10,? cm')
Figure 22. 2DEG properties of standard MD structure (Figure 20): (a) 300 K and (b) 77 K
Hall data.
79
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3.2.2 Impact of Channel Depth on 2DEG Conductivity
The channel depth is defined as the sum of the Schottky layer thickness, donor
layer thickness, and spacer layer thickness. Table 4 shows Hall data for standard
InAIAs/InGaAs MD structures with different channel depths. A delta or planar doping
(PD, Nd = 5.1 x !012 cm'2) scheme, with an interruption of five seconds before and after
the doping step, is used. Hall data for a standard structure (uniform doping scheme with
18
3
Np = 8 x 10 cm ) were also compared. All samples were grown with the same growth
conditions.
In both the planar-doping and uniform-doping scheme, the two dimensional
doping density is same, i.e.. the Si two-dimensional doping density, N;u in the planardoping scheme, was determined by the relation of N
:d
=
N
jd x
td o n o r x
R where N
id
is the
bulk doping concentration, td0n0r is the doping layer thickness, and R is the ratio of the
planar doping time to the uniform-doping time. In these doping schemes, the same Si cell
temperature was used.
As the channel thickness decreases from 40 nm to 20 nm, the conduction band
profile in the channel changes from a triangular well to a square quantum well. Since the
confined carriers in the channel are close to the inverted heterointerface, the 2DEG
mobility can be affected by the roughness of inverted interface. In general, in a single
heterojunction MD structure, the peak of the 2DEG distribution is located close, 95
A
(26], to the growth front interface, i.e., the spacer/channel interface. The 2DEG
conductivity of A946 is very similar to that of the A950 sample with the quantum well
80
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
channel. This means that our spacer/channel interface quality is good. If the channel
thickness further decreases, the effect of the interface roughness will increase and thus
reduce the low field mobility of 2DEG. The 2DEG density of the sample A951, grown
with a planar-doping scheme of N:d = 5.1
x
1012 cm'2 and a spacer thickness of 4 nm was
slightly decreased in comparison to that of the uniformly-doped sample A950 with a
spacer thickness of 2 nm. In the uniform-doping scheme, the calculated two-dimensional
doping density N^d = 6.4
x
1012 cm'2 is larger than the planar-doping density and the
spacer thickness is a half of that of the sample A951.
The reduced doping density and increased spacer thickness decrease the 2DEG
density in the same 20 nm thick channel. However, as the channel depth decreases, the
2DEG density and mobility decrease. This result indicates that the channel depth must be
optimized to maximize the doping efficiency and the 2DEG mobility. According to a
simple charge control model, a free electron density generated in a thin planar-doped
donor layer is inversely proportional to an undoped Schottky layer thickness. Therefore,
as the Schottky layer thickness decreases, the 2DEG density formed in the channel
decreases due to surface depletion from the donor layer. A screening effect is a possible
mechanism to explain the decrease of the 2DEG mobility with a thinner Schottky layer in
MD structures, i.e.. a higher 2DEG density effectively reduces interactions with impurity
atoms by screening, leading to the enhancement of 2DEG mobility.
81
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 4. Hall data for conventional InAlAs/InGaAs MD structures grown with different
channel depths and a planar-doping (PD, ND = 5.1 x 1012cm 2) scheme.
Sample
A946b
A950b
A951
A955
A954
A953
Channel
thickness
(nm)
40
20
20
20
20
20
Channel depth3
and spacer
thickness (nm)
30,2
30,2
30,4
30,4
22.5, 3.5
19,4
Hall mobility
(cm2/Vs)
300 K
77 K
35190
10526
10395
33113
10134
30875
10295
33134
9439
25761
9887
29059
2DEG density
(xlO12 cm'2)
300 K
77 K
2.7
3.0
2.6
2.7
2.2
2.5
2.3
2.6
2.2
2.5
1.7
1.9
3 The channel depth is defined as the sum of the Schottky barrier thickness, donor layer
thickness, and spacer thickness.
bThese samples were doped by the uniform-doping scheme with No = 8 x 10l!t cm'3.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3.3
Doped-Channel InAlAs/InGaAs MD Structures
Doped-channel HEMTs are viewed as promising devices and have attracted
considerable attention for power applications [68, 99-100], Heavily doping the channel,
of course, degrades the low field electron mobility of the structure. As device sizes
continue to shrink into the deep submicron regime, however, the importance of low field
transport is minimized due to a decrease in mobility-limited transport under the gate
[101]. The effects of various doping profiles in quantum-well channels on 2DEG
transport have been evaluated [102, 103].
In our work, planar-doping at the center of quantum well channel has been
performed during the growth of MD structure. As the doping density increases, the
achievable 2DEG conductivity has been investigated for this doping scheme. To
investigate the influence of channel doping on the 2DEG conductivity in a standard
InAlAs/lnGaAs MD structure, we have grown uniformly-doped InAIAs donor and
planar-doped InGaAs single channel MD structure as shown in Figure 23. In this
structure, the thickness of the InGaAs channel was 20 nm or 30 nm, and doping density
of 1.6 x I012 cm'2 was supplied at the center of the InGaAs quantum well channel. The
same growth conditions as that of the standard InAlAs/InGaAs MD structures were used
with the exception of the channel doping step.
Figure 24 shows the 2DEG conductivity of the planar-doped InGaAs channel MD
structure measured at 300 K and 77 K. The Hall data of the undoped InGaAs channel (20
nm) MD structure was also compared in this figure. The 2DEG density in the doped-
83
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
channel MD structure was increased as expected, but, the 2DEG mobility was
dramatically decreased by introducing channel doping from 10000 to 7000 cm2/Vs at 300
K, and from 35000 to 10000 cm2/Vs, at 77 K, as compared to that of undoped InGaAs
channel MD structure. These results indicate that the 2DEG mobility in the doped
channel MD structure is limited by the impurity scattering in the channel, and thus the
decrease of the 2DEG mobility is pronounced at 77 K. With a planar channel doping of
1.6 x 1012 cm'2, a 2DEG density of 4.1 to 5.1 x 1012 cm 2 was achieved, and agreed well
with the sum of the channel doping density (Nip = 1.6 x 1012 cm ') and the 2DEG density
(2.5 to 3 x 1012 cm'2) obtained in the undoped InGaAs channel MD structure.
84
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
InGaAs:Si(7xl01!1cm ')
Ir>oaAlo 6As
InAIAs:Si(8xl0ls cm''1)
InAIAs
InGaAs
PD(N2d = 1-6 x 10'- cm ')
InGaAs
InAIAs
InP substrate
7 nm
20 nm
8 nm
2 nm
15, 10 nm
15, 10 nm
250 nm
Figure 23. Schematic diagram of the uniformly-doped donor and PD InGaAs single
channel MD structure.
85
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
40
77 K
30
77 K Hall mobility (x 10* cm2/Vs)
300 K Hall mobility (x 10 cm /Vs)
35
Undoped InGaAs channel MD
25
20
15
Planar-doped InGaAs channel MD
300 K
10
5
0
0
2.5
3
3.5
4
4.5
5
5.5
2DEG density (x 1012 cm'2)
Figure 24. Hall data of planar-doped InGaAs channel MD structure (see Figure 23).
86
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3.4
Double-Heteroiunction InAlAs/InGaAs MD Structures
In double-heterojunction (DH) power HEMTs, carriers are introduced into the
InGaAs channel by doping the InAIAs on the both sides of the InGaAs. This results in a
high channel current density and hence a high power-handling capability. A 0.15 pmlong gate DH InAlAs/InGaAs HEMTs has been demonstrated high cutoff frequencies (f,
= 102 GHz, fmax = 345 GHz), but mediocre current density (700 mA/mm) [104]. Usually,
a planar-doping scheme is used for DH structures with spacer thicknesses of 3 - 5 nm on
the top and bottom to obtain high sheet carrier density and aspect ratio. The bottom-side
planar-doping density, one third of the total target doping density (N^d)* is less than the
top-side doping density (two third of N?d) to achieve good pinch-off, since the bottomside doping is located greater distance from the gale.
Generally, in a DH structure, the 2DEG distribution in the quantum well channel
is different from that of a single-heterojunction (SH) structure. While the 2DEG
distribution in the SH structure is asymmetric (close to the top interface of InGaAs
channel), that of DH structure is more symmetric, and thus the 2DEG mobility may be
degraded due to inverted interface roughness and the consequent increase in scattering.
The relation between the planar-doping density and the 2DEG characteristics has been
investigated in DH InAlAs/InGaAs MD structures.
Figure 25 shows a schematic diagram of the double heterojunction (DH)
InAlAs/InGaAs quantum well MD structure. In this structure, the 2DEG is supplied from
both top side planar-doping and bottom side planar-doping layer. The bottom side planar-
87
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
doping density, one third of the total target doping density (Nsd) is less than the top side
doping density (two third of N:d) to obtain good pinch-off. To avoid Si segregation, the
growth temperature was reduced from 500 °C to 420 °C during the planar doping. For
both top and bottom planar-doping, the same spacer thickness of 5 nm was used. To
increase the Schottky barrier height, a lattice-mismatched Ino.aAloeAs Schottky layer of
20
nm was grown.
Figure 26 shows the 300 K 2DEG Hall data for the DH InAlAs/InGaAs MD
structures grown using different planar-doping densities, Njd- The doping density was
calibrated with the uniform-doping conditions and controlled by the planar-doping time
at the same Si cell temperature. As the planar-doping density increases, the 2DEG density
increases up to 6.2 x 1012 cm' 2 and the 2DEG mobility decreases to about 7000 cm2/Vs.
Therefore, there exists a trade-off between the 2DEG density and mobility.
The decrease of the 2DEG mobility with an increase of planar-doping density can
be explained by two reasons as follows. First, as the 2DEG density is increased, the
2DEG has more interaction with any inverted interface roughness, resulting in the
decrease of the mobility in the channel. Second, as the planar-doping density increases,
more 2DEG carriers populate a higher subband state up to E3 , leading to a slight decrease
of 2DEG mobility [103]. In Figure 26, the 2DEG Hall data for a single heterojunction
(SH) MD structure grown with the same growth condition and structure are compared
with those of DH MD structures. In this comparison, the 2DEG mobility of the DH MD
structure grown with the similar planar-doping density is smaller than that of 2DEG
88
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
mobility in the SH MD structure. This phenomenon can be explained by the difference in
the 2DEG distribution in the channel.
In the SH channel MD structure, the 2DEG distribution within the channel is
close to the top InAlAs/InGaAs interface and consequently leads a relatively smaller
interaction with the inverted interface in comparison to the DH channel MD structure.
However, in the DH MD structure, the 2DEG distribution is enhanced near at the center
of the quantum well channel or close to the inverted interface according to the doping
density. As a result, this will increase the interaction with the rough inverted interface and
can reduce the 2DEG mobility in DH M D structures.
89
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
InGaAs:Si(7xl018 cm'3)
Ino.;Alo6As
PD1 (2/3 N2n)
InAIAs
InGaAs
InAIAs
P D 2(1/3N 2u)
InAIAs
InP substrate
4
7 nm
20 nm
T g = 420 °C
5 nm
20 nm
5 nm
T g = 420 °C
250 nm
Figure 25. Double Planar-doped (PD1 and PD2) InGaAs single channel MD structure.
During the planar-doping, the substrate temperature was reduced from 500 to 420 °C to
avoid the segregation of Si.
90
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
12
300 K
11
(/)
£
I
■b
o
Single planar doping
10
E
o
*b
CM
9
X
S
o
5
5
c
Q
)
"O
O
LU
Q
8
E
m
X
CM
7
6
5
1
4
5
6
8
7
9
10
11
Planar doping density, N2D (x 1012 cm'2)
Figure 26. 300 K Hall data of DH InAlAs/InGaAs MD structure grown with various
planar-doping density, in which Hall data of SH MD structure (open circle and square)
was also compared.
91
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER IV
GROWTH AND DEVICE PERFORMANCE OF InGaAs/InP
COMPOSITE-CHANNEL HEMTs
As outlined in chapter I, InGaAs/InP composite-channel HEMTs have been
considered as an approach to overcoming the impact ionization in the InGaAs channel.
This is one of the primary limitations of InP-based HEMT due to low breakdown and
resultant low output power, etc. To develop this composite-channel structure, optimal
growth conditions for InP epitaxial layers are required. SSMBE has been comparatively
little utilized for the development of P-based heterostructures for microwave electronic
devices. In this thesis, we report the growth and performance of various InGaAs/InP
composite-channel MD structures grown on InP (100) semi-insulating substrates in a
solid source Riber 2300 MBE system equipped with conventional group m cells and
valved-cracker sources for arsenic (EPI RB500V) and phosphorus (Riber KPC40).
4.1
Preliminary Experiments
Before the growth of arsenide/phosphide heterostructures for devices, the growth
conditions and material quality of InP grown by the SSMBE have to be evaluated.
92
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Baillargeon et al.[18] reported that the electrical properties of the undoped InP films were
strongly dependent upon the cracking zone temperature of the cell. With adjustment of
the cracker temperature alone, a free electron concentration range of 2.9 x 1015 cm' 3 to
4.0 x
1 0 15
cm' 3 was obtained. In our experiments, a P4 cracking temperature of 920 °C
was used. A typical n-type background concentration for undopcd-InP films was 1 x 1016
cm'3, which is similar to their results ( 8 x 1015 cm' 3 at a cracking temperature of 920 °C).
In our experiments, an optimum P: beam equivalent pressure (BEP) of 4.5 x 10'6 torr was
determined via a study on the electrical properties and the surface morphology of
SSMBE-grown InP films.
Table 5. Hall data of InP single channel MD structures.
Sample
A217a
A2l8b
A274c
1
As4. P: BEP
(x 10' 5 torr)
1.5.0.45
1.5,0.45
1.5,0.45
Rs
Hall mobility
@ 300 K
(cm2/Vs)
(Q/sq)
300 K
77 K
940
2568
5432
908
2145
3873
586_______2302
5335
2DEG density
(xlO 12 cm'2)
300 K 77 K
2.6
2.4
3.2
2.8
4.6
3.9
1nP channel thickness of 40 nm.
b InP channel thickness of 200 nm.
c Planar-doped InP M D structure: InP (15 nm)/PD(1.6 x 1012 cm'2)/InP (35 nm).
93
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 5 presents Hall data of InP single channel MD structures, which have been
grown with the standard InAlAs/InGaAs MD structure (see Figure 20) except that an InP
channel was used to replace the InGaAs channel. In all samples, the 2DEG mobility was
relatively low (similar to that of bulk InP films) and the 2DEG density had a comparable
value to that of standard InAlAs/InGaAs MD structures. In sample A218, an InP buffer
was used instead of the InAIAs buffer and the mobility was decreased, but the 2DEG
density was increased as compared to that of A217 sample with InAIAs buffer. This
indicates that InP buffer has relatively low resistivity and thus is not good choice. The
planar doped-channel (sample A274) with a doping density of 1.6 x 1012 cm'2, which will
be used to dope the subchannel InP in the growth of InGaAs/InP composite-channel MD
structure, did not degrade the 2DEG mobility and increased the carrier density as
compared to that of sample A2I4.
4.2
InGaAs/InP Composite-Channel MD Structures
As previously stated, the InGaAs/InP composite-channel HEMT structure has
advantages from the physical properties of both channel material, i.e., the high electron
mobility of LM InGaAs at low electric fields, and the high breakdown and velocity of InP
at high electric fields. At low drain bias, most of electron populate in InGaAs channel as
shown in Figure 44. At high drain bias, electrons spatially transfer into InP channel and
thus improve breakdown.
94
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
To optimize the 2DEG properties in the InGaAs/InP composite-channel M D
structures, a systemic study of various structural designs and doping strategies must be
made. The thickness of the InGaAs channel and the doping profile of the InP sub-channel
are considered to be the key design issues in this MD structure. As the channel size of the
quantum well decreases, the quantized subbands increase, and thus effectively increase
the bandgap of the channel material. This concept is considered one of the techniques that
suppresses impact ionization in the narrow bandgap channel. This channel size effect and
the planar-doping strategy, which are quite an attractive technology for power HEMT
applications, has been investigated. In this research, a study on the effect of structural
parameters and doping profiling on the 2DEG property has been performed with the
InGaAs/InP composite-channel MD structures, which are grown with the optimized
conditions achieved in the previous research on the conventional InAlAs/InGaAs MD
structures.
The InGaP hole barrier approach to reducing the hole current generated via the
impact ionization in the channel has been carried out with the optimized InGaAs/InP MD
structure. The Ga composition in the InGaP alloy is determined by x-ray diffraction
analysis with InGaP/InP MQWs. In addition, the impact of the planar-doped InGaAs
channel in the InGaAs/InP composite-channel structures on the 2DEG properties has
been investigated and compared with that of conventional InAlAs/InGaAs MD structures.
To further enhance the breakdown voltage, a strained IncuAlo^As Schottky layer has been
grown.
95
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
As shown in Figure 28, an InGaAs/InP multi-composite-channel MD structure
has been investigated to maximize the 2DEG density. In this structure, the channel
consists of two InGaAs channels with thicknesses of 5 nm and two InP subchannels of S
nm and 25 nm thickness. To achieve a high 2DEG density in the channel with a total
thickness of 40 nm, a double uniform-doping concept was used. Two donor layers of 8
nm were uniformly doped with doping densities of 8 x 1018 cm'3 at the top-side and with
doping densities of 5 x 1018cm'3 at the bottom-side.
InGaAs:Si(7xlO 8 cm'3)
Ino.4Alo.6 As
InAIAs:Si(8x 1018 cm'3)
InAIAs
InGaAs
InP
InGaAs
InP
InAsP
InAIAs
InAlAs:Si(5x 1018 cm'3)
InAIAs
InP substrate
7 nm
20 nm
8 nm
2 nm
5 nm
5 nm
5 nm
25 nm
2 nm
2 nm
8 nm
250 nm
Figure 28. Multi-channel InGaAs/InP M D structures.
97
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 6. Hall data of multi-channel InGaAs/InP M D structure.
Sample
A284
A286
Rs
@ 300 K
(Q/sq)
133
138
Hall mobility
(cm2/Vs)
300 K
77 K
9038
36553
8882
37035
2DEG density
(xlOl2cm2)
300 K 77 K
5.2
4.6
5.1
4.6
Table 6 shows the Hall data of multi-channel InGaAs/InP MD structure measured
at 300 and 77 K. As seen in this table, the 2DEG mobility was comparable with that of
standard InAIAs/lnGaAs M D structures but, the 2DEG confinement was enhanced by
approximately two times as compared to that for standard InAlAs/InGaAs MD structures.
This consequently leads to the lowest sheet resistance (133 i2/sq at 300 K) in comparison
to that of any InP-based MD structure as shown in Figure 30.
Table 7 shows the influence of the InP spacer thickness, which is defined as an
undoped InP layer between InGaAs channel and planar-doping placement, for planardoping InP subchannels on the 2DEG conductivity of the InGaAs/InP composite-channel
MD structure (Figure 27). As the InP spacer thickness for the planar-doping of InP
subchannel decreases from 15 nm to 4 nm, the 2DEG density is increased due to
improved doping efficiency for the InP layer and the mobility is decreased due to an
increase in remote impurity scattering or in interface roughness scattering. However,
98
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
according to our results (2DEG mobility of 10300 cm2/Vs at 300 K and 31000 cm2/Vs at
77 K) for a very thin InGaAs (5 nm) composite-channel MD structure, the influence of
InP spacer thickness on the 2DEG mobility seems to be governed by the remote impurity
scattering. This good mobility, achieved in the thin InGaAs channel (5 nm) MD structure,
indicates that our SSMBE-grown InGaAs/InP interface is comparably smooth and does
not impact significantly the 2DEG mobility.
Table 7. Influence of InP spacer thickness for planar-doping InP subchannel on the 2DEG
conductivity of InGaAs/PD (N;d = l.6 x 1012 cm'2) InP composite-channel MD structure.
A uniform-doping of N d = 8 x 1018 cm'3 for the top donor layer was utilized.
Sample
A250
A366
A365a
InP spacer
thickness
(nm)
15
4
Undoped InP
Hall mobility
(cm2/Vs)
300 K
77 K
10267 30960
6837
18284
8963
21877
R>
@ 300 K
(fi/sq)
140
185
233
2DEG density
(xlO12 cm'2)
300 K 77 K
4.3
4.4
4.9
4.0
3.0
3.7
aA planar-doping of N: d = 4.6 x I0 12cm'2 for top donor layer was utilized.
As seen in Table 7, the 2DEG mobility and density of an undoped InGaAs/InP
composite-channel MD structure (sample A365) was compared with those of M D
99
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
structures grown with planar-doping in the InP subchannel. In the undoped InGaAs/InP
composite-channel MD structure, the 2DEG density was decreased to 3 x 1012 cm'2 due
to reduced doping density. The mobility was also slightly lower due to the change of the
2DEG distribution in the InGaAs channel resulting from different potential distributions
between a single- and double-modulation-doped quantum well channel.
Figure 29 shows a schematic diagram of a planar-doped InGaAs channel/undoped
InP composite-channel MD structure grown with various total planar-doping densities
(N:u) from 4.8 to 7.5 x 1012 cm 2. The top planar-doping (PD1) density is two thirds of
N:d and the bottom planar-doping (PD2) density for the InGaAs channel of 15 nm is
carried out at the center of channel with a doping density of one third of N^dThe room temperature 2DEG conductivities of the InGaAs/InP composite-channel
MD structures grown with various doping schemes are compared in Figure 30. As shown
in this figure, the best 2DEG mobility is achieved in the InGaAs/InP composite-channel
MD structure grown by combining planar-doping in the InP channel and a 15 nm thick
InP spacer. For the case of the planar-doped InP channel with a thinner InP spacer (4
nm), the 2DEG density was comparable with that of the same doping scheme, except
with the 15 nm thick InP spacer, but the mobility was dramatically decreased due to
increased remote impurity scattering, and strongly dependent on 2DEG density. While
the 2DEG mobility in the mutl-channel InGaAs/InP MD structure using the double
doping scheme for the InAIAs donor layer was slightly decreased in comparison to that of
the InGaAs/InP composite-channel MD structure with the planar-doped InP and the 15
100
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
nm thick InP spacer, the 2DEG density was 5 x 1012 cm'2 without a sacrificing in
mobility, consequently leading to the lowest sheet resistivity of 133 Q/sq.
In a planar-doped InGaAs channel and InP subchannel M D structure, the highest
2DEG density of greater than 6 x 1012 cm'2 can be obtained but the mobility is
dramatically degraded, as expected, due to the overlap of the 2DEG and impurity
distribution in the channel. The 2DEG conductivity of InGaAs/InP composite-channel
MD structure with the InGaP hole barrier layer located above the donor layer was
comparable with that of MD structures without the InGaP hole barrier. Compositechannel MD structures with InGaP hole barrier will be discussed later.
In Figure 31, 300 K Hall data for various InGaAs/InP composite-channel MD
structures with various epitaxial structures and doping schemes were compared to
investigate trends in sheet resistance. These Hall data clearly demonstrate that an
optimized 2DEG conductivity is achieved at a 2DEG density around 4 x I012 cm'2 and
mobility around 10000 cm2/Vs. This corresponds to a minimum sheet resistance for the
channel at a low electric field.
101
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
InGaAsiSKTxlO18 cm'3)
In()jAlo6As
PD1 (2/3 N2D)
InAIAs
InGaAs
PD2 (1/3 N:o)
InGaAs
InP
InAsP
InAIAs
InP substrate
4
7 nm
20 nm
T g = 420 °C
5 nm
10 nm
To = 420 °C
10 nm
30 nm
2 nm
250 nm
Figure 29. Schematic diagram of planar-doped InGaAs channel/undoped InP MD
structure with various doping density (N:d = 7.54 - 4.8 x 1012cm'2).
102
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
r
.
1
|
1
!
1! TI !
T
.
-r-T-T-r
T- r
T-
12 _
•
•
300 K
•
•
•
10
.
♦ •♦ I
(/)
£
i
•b
•
♦♦
•
■
1—
*
■ l
-
□
i*
C ■
1
!5
o
E
n
X
-
□C
C
□
•
Undoped InGaAs/InP channel
Planar-doped InGaAs channel
Planar-doped InP channel
with 15 nm InP spacer
A
♦
Multi-channel InGaAs/InP HEM T
Planar-doped InP channel
and InGaP hole barrier
■
Planar-doped InP channel
with 4 nm InP spacer
-
... 1. . . .
1
2
3
4
5
6
1...
7
.
8
2DEG density (x 1012 cm'2)
Figure 30. Comparison of 300 K 2DEG conductivity for InGaAs/InP composite-channel
MD structures with various doping schemes.
103
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
I
12
-
10
-
I
I
I
I
I
I
I " p i
T
1
I
|
1
I
I
1' ■J
I
I
I
I
|
I
I
I
i
|
I
I
I
1000
I
InGaAs/InP composite-channel HEMTs
800
300 K
••
E
600
u
X
g
5
o
E
400
••
m
X
o 0
- O& ~
O'g
Qd
oo
Sheet resistance (Ohm/sq)
• •
(0
$
cm
200
0
1
2
3
4
5
6
7
8
2DEG density (x 1012 cm'2)
Figure 31. Comparison of the 300 K Hall mobility and sheet resistance versus 2DEG
density for various InGaAs/InP composite-channel M D structures.
104
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4.2.2 Quantum Size Effects of InGaAs Channel
As the channel size of the quantum well decreases the quantized subbands
increase, and thus, effectively increase the bandgap of the channel material. This concept
has been considered as one technique to suppress the impact ionization in the narrow
bandgap channel. We have investigated the influence of the InGaAs channel thickness in
the InGaAs/InP composite-channel MD structures on the 2DEG transport.
InGaAs:Si(7xl0' cm '1)
lno.4 Alo.6 As
lnAlAs:Si(8 x 1018 cm'3)
IAs
InGaAs
InP
PD (N;d = 1. 6 x 101' cm'2)
InP
InAsP
InAIAs
InP substrate
7 nm
nm
8 nm
2 nm
20,15, 10,5, 2.5, Onm
15 nm
20
15,20, 25, 30, 32.5,35 nm
2 nm
250 nm
Figure 32. Uniform-doped donor and thickness varied InGaAs/15 nm InP spacer/PD InP
MD structures grown to investigate the channel-size quantization effects.
Figure 32 shows the schematic diagram of an InGaAs/InP composite-channel MD
structures with different InGaAs thickness from 0 nm to 20 nm, while maintaining a
constant thickness (SO nm) for the combined InGaAs and InP channel. As the InGaAs
105
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
thickness decreases from 20 nm to 0 nm, the 2DEG mobility is dramatically decreased
(Figure 33). As seen in Figure 33, the 2DEG mobility is slightly changed in the rage of 5
to 20 nm, but greatly decreased for thicknesses of less than 5 nm. For channel thickness
greater than 5 m, the 2DEG mostly resides in the InGaAs channel but, as the channel is
made narrower (< 5 nm), transfers into the InP subchannel, leading to a decrease in the
2DEG mobility due to the large electron effective mass in InP. Finally, a very thin
InGaAs channel shows 2DEG mobilities like InP-channel HEMTs.
Although most of the 2DEG resides in the thin InGaAs channel,generally,
a
thinner channel thickness decreases the 2DEG mobility, because the 2DEG suffers more
interactions with both heterointerfaces of the InGaAs channel. In our results, the best
mobility at both 300 K and 77 K was obtained in composite-channel MD structures with
a 5 nm thick InGaAs channel.
As the thickness of the InGaAs channel decreases, the effective conduction band
discontinuity also decreases due to an effective increase of the bandgap of quantum well
channel as previously explained. This results in a decrease of the 2DEG density inthe
InGaAs channel at a low electric field. Figure 33 captures this trend.
106
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
t5
900 K
10
E
■b
5
0
2
0
5
to
15
20
25
(a)
30
inGiAs Ctannti tnickntss (nm>
77 K
2
“g
o
3
20
-
(b)
0
S
2
10
15
20
25
30
MGAM Ctwnai v ic k n tti (nm)
Figure 33. Hall data versus InGaAs channel thickness for InGaAs/InP composite-channel
MD structures: (a) 300 K and (b) 77 K Hall data.
107
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
As seen in Figure 33 (a), the 2DEG density of the MD structure grown with only
the InP subchannel instead of the InGaAs/InP composite-channel was higher than that of
the InGaAs/InP composite-channel MD structures. The interface roughness at the
InAIAs/InP interface can be considered as the main reason of the higher 2DEG density
since the arsenic overpressure typically roughens the surface of the phosphide alloy
[105]. Very recently, Meneghesso et al. [106] reported on the breakdown properties of
InGaAs/InP composite-channel HEMTs as a function of InGaAs channel thickness. Their
2DEG Hall data are consistent with our results (Figure 34).
The
influence
of
the
InGaAs
channel
thickness
on
the
maximum
transconductance, gm, and the maximum drain current, Imax, of a 0.15 |im-gate
InGaAs/InP composite-channel HEMTs. has been investigated as shown in Figure 35. As
the InGaAs channel thickness decreases from 20 nm to 5 nm, both the maximum
transconductance and maximum drain current are decreased due to the decrease in both
2DEG density and mobility demonstrated in Figure 34. In fact, the maximum drain
current is proportional to the 2DEG conductivity in the channel, i.e., 2DEG density times
mobility. Figure 36 (a) shows the impact of channel size on the 2DEG conductivity
(2DEG density x mobility), which was calculated with the Hall data shown in Figure 34.
This trend is consistent with that of Inux shown in Figure 35.
The intrinsic maximum transconductance is given by the following equation [30]:
( gmW = (q H nso/L) x [ 1 + ( 0 p nso( d + Ad)/e vsL)2 ]‘l/2
108
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(24)
where ^ is the 2DEG mobility, nso is the 2DEG density in the channel without bias
voltage, L is the gate length, d + Ad is the channel depth from the gate, £ is the
permitivity of InAlAs, and vs is the saturation velocity. To investigate the impact of the
channel size effect on the intrinsic maximum transconductance, the normalized intrinsic
maximum transconductance was calculated with the above Eq. (24) and the 2DEG
conductivity data (see Figure 36 (b)). For channel thickness from 5 nm to 20 nm, the
normalized intrinsic maximum transconductance is not significantly changed. However,
as shown in Figure 35, the extrinsic maximum transconductance of a 0.15 |im-gate
InGaAs/InP composite-channel HEMTs strongly depends on the InGaAs channel
thickness because the extrinsic transconductance is inversely proportional to the sheet
contact resistance. The extrinsic transconductance is. in turn, proportional to the channel
conductivity. An InGaAs channel of 20 nm yields optimized combination of current and
transconductance. Reducing the InGaAs channel thickness may improve the breakdown
due to the quantum size effect. However, it can degrade the maximum channel current.
109
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
15
Hall mobility (x 1CP cm 2/Vs)
300 K
o
Meneghesso et al. (1999)
10
E
u
©
CM
i
3
o
HI
o
5
C\J
0
2
0
5
10
15
20
25
InGaAs channel thickness (nm)
Figure 34. Comparison of the impact of InGaAs channel thickness on 2DEG mobility and
density of InGaAs composite-channel MD structures with reference data [106].
110
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1000
1000
800
800
600
600
400
400
200
200
0
5
10
15
20
25
(mA/mm)
InGaAs/InP composite-channel HEMTs
(0.15 micron Gate length)
30
InGaAs channel thickness (nm)
Figure 35. Impact of InGaAs channel size on maximum transconductance and drain
current density in the InGaAs/InP composite-channel MD structure.
Ill
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1005
Y • mote My 1 2 0 € O darw ty
IQ J ^ - A Y / S 0 RT{1 * BY') A
(a)
009
•5
0
5
10
IS
20
25
inOiA * d w r m thcfrn#** i m i
60
50
40
30
(b)
5
0
5
10
IS
20
2S
inO oA* cftonnoi tfu ckn o M (nm)
Figure 36. Influence of InGaAs channel thickness on 2DEG density and mobility product (a) and
on normalized intrinsic (gm)™ (b).
112
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4.2.3 Growth of InGaP Hole Barrier
As described in section 2.4.2.6, hole barriers of various materials such as InGaP
[77], AlAs [79], lattice-mismatched InAlAs [78], and mixed InAlAs/InP [80] have been
inserted between the InAlAs spacer layer and the InGaAs channel to improve the
breakdown voltage in InP-based power HEMTs.
InGaAs:Si(7xl0 cm )
Inn aAlo (,As
InAlAs:Si(8x 1018cm ’)
InAlAs
Ini.vGavP (y = 0-0.5) hole barrier
InGaAs
InP
PD(N;,)= 1.6 x IO'Jcm -)
InP
InAsP
InAlAs
InP substrate
7 nm
20 nm
8 nm
2 nm
2.5 nm
5 nm
15 nm
30 nm
2 nm
250 nm
Figure 37. InGaAs/InP composite-channel MD structures with an Ini.yGayP (2.5 nm) hole
barrier.
Figure 37 shows a schematic diagram of the InGaAs/InP composite-channel MD
structure combined with a Ini yGayP (0 < y < 0.5) layer grown to block holes generated
113
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
via the impact ionization process in the narrow bandgap channel of InGaAs from the
gate. The thin Ini.yGayP alloy was calibrated via the growth of InGaP/InP MQWs and xray diffraction analysis. Before the growth of the InGaP hole barrier, a growth pause was
carried out to obtain a stabilized P: flux. Figure 38 shows the influence of the 2.5 nm Ini.
yGayP (0 < y < 0.5) hole barrrier on the 2DEG Hall data measured at 300 K and 77 K,
respectively.
The 2DEG density and mobility of the InGaAs/InP composite-channel MD
structure with a InGaP hole barrier both decreased in comparison to that of the same
structure without the InGaP hole barrier. As seen in Figure 38 (b), the 2DEG mobility at
77 K is considerably lower. This decrease of mobility can be attributed to interlace
roughness and/or strain at the InGaP/InGaAs interface. Since the 2DEG mobility in the
MD structure with the lattice-matched ln|.yGayP (y = 0) hole barrier is also lower, the
interface roughness of InGaP/InGaAs is probably the dominant mechanism responsible
for the mobility reduction.
The relationship between the 2DEG mobility and the Ga composition, y, in the
Ini.yGayP hole barrier shows a parabolic trend as can be seen in Figure 38 (a). In the MD
structure with the 2.5 nm Ini.yGayP (0 < y < 0.5) hole barrier, the thickness of the spacer
layer is effectively increased to 4.5 nm. This effective spacer thickness increase reduces
the efficiency of the modulation doping and thus decreases the 2DEG density in the
InGaAs channel. This fact explains the observed decrease of the 2DEG density in all the
MD structures with the Ini.yGayP hole barrier (see Figure 38 (a)).
114
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
In general, the 2DEG density in the channel is proportional to the conduction
band discontinuity, AEC, at the hetrojunction interface. As the composition, y, of the Ini.
yGayP increases, AEC is larger and better confines the 2DEG in the channel. However,
Figure 38 (a) shows a contradictory trend. To date, it has been assumed that the bandgap
of the strained In|.yGayP alloy would monotonically increase with an increase in the Ga
composition, y. However, the bandgap of strained Ini.yGayP calculated with the
deformation theory in chapter II is much different from that of the unstrained In iyGayP
alloy. The tensile strain present in In|.yGayP/InP moves the light hole valence band edge
close to the conduction band edge and the heavy hole valence band edge away from the
conduction band edge. Consequently, the bandgap of strained InGaP is effectively
narrowed. This means that the energy bandgap of strained InGaP is determined by the
shifted light hole valence band edge. Indeed, the calculated light hole valence band edge
has a convex form as the Ga composition of InGaP alloy varies from 0 to 1 as shown in
Figure 17.
In order to investigate the modification of the 2DEG characteristics resulting from
the position of the InGaP/InGaAs interface, the InGaP hole barrier was moved above the
InAlAs donor layer. This resulted in greatly improved characteristics as shown in Figure
39. This result indicates that the 2DEG conductivity strongly depends on the quality of
the InGaP/InGaAs interface.
115
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
12
KOK
10
Without InGaP hot* barrar
I
6
4
0
0
2
01
03
02
04
05
06
Compoaitton ot In G« P hot* bamar tyi
50
40
Without InGaP hoi* barrar
30
„
20
I
(b)
0
0
25
01
03
Compoaitton of In
04
0.6
not* b*m *r (y)
Figure 38. Hall data for the InGaAs/InP composite-channel MD structure with Ini.yGayP
hole barrier (0 < y < 0.5): (a) 300 K and (b) 77 K Hall data.
116
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
300 K
12
2DEG density (x 1012 cm
10
8
6
4
2
A: InGaP hole barrier above the donor layer
B: InGaP hole barrier under the InAlAs spacer
0
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
Composition of In^Ga^P hole barrier (y)
Figure 39. Comparison of 300 K Hall data of the InGaAs/InP composite-channel MD structures
with two different layer positions of the InGaP hole barrier.
117
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4.3 Device Performance
With a 20 nm InGaAs/15 nm InP spacer/delta-doping/15 nm InP channel (see
Figure 27), we have achieved an optimized channel conductivity (a 2DEG density of 4.3
x 1012 cm'2 and mobility of 10,300 cm2/Vs at 300 K). To the best of our knowledge this
result is the highest channel conductivity ever achieved in a composite channel structure,
as shown in Figure 40. In Figure 41, the maximum transconductance and maximum drain
current density for 0.15 pm-gate InGaAs/InP composite-channel HEMTs with various
doping schemes are compared to those of TRW baseline InGaAs-InAIAs InP-based
HEMTs.
In a 0.15 pm x 200 (4 x 50 pm) T-gate InGaAs/InP composite-channel HEMT
(see Figure 42) fabricated using TRW InP-based HEMT process, a drain current density
of 830 mA/mm was measured at a 4.4 V. which corresponds to a DC power of 3.6
W/mm. A maximum transconductance of 730 mS/mm was obtained at the gate voltage of
0.25 V as shown in Figure 43. This optimized 0.15 pm T-gate device demonstrated stateof-the-art of maximum transconductance-maximum drain current combination for
InGaAs/InP composite-channel HEMTs.
At low drain bias, electrons flow in the InGaAs channel. The 2DEG distribution
simulated by the ATLAS (commercial device simulation software, SELVACO) is shown
in Figure 44. At low electric field, the device performs as well as the conventional one.
At high Vds, electrons spatially transfer to InP channel, resulting in improved breakdown
voltage. As can see in Figure 45, RF performance such as device gain, ft, and fma* did not
118
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
degrade as compared with that of conventional InAIAs/InGaAs HEMT fabricated in the
same process. This InGaAs/InP composite-channel HEMT also showed a 1.5 V
improvement in on-state and off-state breakdown over the conventional InAIAs/InGaAs
HEMT without degrading the RF performance (Figure 46).
Using an optimized InGaAs/InP composite-channel structure, a two-stage W-band
MMIC power amplifier has fabricated and characterized in TRW. The amplifier exhibited
17 dB linear gain at 94 GHz, which is about 4 dB higher than that measured from a
conventional
InP-based HEMT
MMIC power amplifier. On-wafer pulse power
measurements were carried out to evaluate the power performance. An excellent output
power of 25 dBm (Figure 47) was obtained. To our best knowledge, this is the first
demonstration of InGaAs/InP composite-channel HEMT at 94 GHz with excellent power
performance. These results confirm that the InGaAs/InP composite-channel structure can
improve the breakdown of In-based HEMT without sacrificing RF performance and thus
is suitable for millimeter-wave high power applications.
119
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Hall data of InGaAs/InP composite-channel HEMTs
12000
300 K
SSMBE (GT)
10000
• GSMBE (HRL)
</)
MOVPE (FT)
8000
o
MOVPE (NTT)
■O
O
E
n
6000 -
r
#
InP/lnGaAs/lnP HEMTs
GSMBE [Ref. 107]
4 000
«
SSMBE [Ref. 108]
2000
1 ' 1 *- 1 1 1 ‘ 1 1 ‘ 1 1 * 1 1 1 ' ‘ 1 ' ‘ ‘ ‘
2
3
4
5
6
7
2DEG density ( x 1012 cm'2)
Figure 40. Comparison of 300 K Hall data of InGaAs/InP composite-channel HEMTs
and InP/lnGaAs/lnP HEMTs grown by various growth methods.
120
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
InP P 1-27
1200
InP P1-Z0
InP P1-Z10
InP P1-Z12
1000
TRW baseline
E
E
A245
c/5
E
0
)
o
c
m
o
■3o
c
o
800
A364
A265
A363
A361
600
o
CO
c
(Q
w.
400
£
200
300
400
500
600
700
800
Maximum channel current (mA/mm)
Figure 41. Maximum transconductance versus maximum drain current density of 0.1S
(im-gate InGaAs/InP composite-channel HEMTs with various doping schemes.
121
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ISMSHSWai
Figure 42. 0.15 |im x 200 (4 x 50) pm InGaAs/InP composite-channel HEMTs (TRW)
[131].
122
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(b)
Figure 43. Current-voltage characteristics (a) and transconductance-gate voltage
characteristics (b) of InGaAs/InP composite-channel HEMT (0.15 Jim x 40 |im) [131].
123
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1
18
0.5
17
>
o>
>
ui
o
lii
-0.5
C
InAlAs
InGaAs
1.5
10
200
400
o
£
M
InP
InAlAs
E
600
800
16
&
15
9
o
v
S
ID
14
1000
Distance From Surface (A)
Figure 44. Energy band diagram and electron distribution of InGaAs/InP compositechannel HEMT (Figure 27) simulated by ATLAS (Device simulation software,
SILVACO) [131].
124
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
A: Conventional
B: Composite-Channel
Legend: (ft [GHz], fmax [GHz], MSG@60GHz [dB])
500
A:(100, 240,11.7)
B:(100, 220,11.8)
400
?
|
A:(100, 2 1 0 ,11.4)A:(85, 200,11.68
300
A:(110,185; 8.94)
B:(100,185,8.82)
£,
J 200
B:(100’ 220’ H-«)B:(80, 205,11.6
^
•
•
A:(110,192, 9.42)
B : ( 1 0 0 ,W 9.30)
•
A:(100, 225,10.8) A:(8 5 ,190, 9.8
B:(95, 210,10.9) B;(80t 190>
A:(110,175,9.03)
B:( 95,160, 9.02)
100
0
0
1
2
3
4
Vds (V)
Figure 45. Comparison of f,, f ^ , and gain of the conventional InAIAs/InGaAs HEMT
and InGaAs/InP composite-channel HEMT (Figure 42) fabricated with the same process
(TRW’s InP-based HEMT process line) [ 13 1].
125
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
■ Conventional
Composite Channel
1000
_
750 “
!
•
|
500 —
!
i
m
250
1
2
8
Burnout Vds (V)
Figure 46. Burnout breakdown voltages of the conventional InP-based HEMT and the
InGaAs/InP composite-channel HEMT measured at various drain current levels. The
composite-channel structure improves the breakdown voltage by about 1.5 V [132].
126
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Frequency: 94 GHz
28
CD
2,
c
Pout
24
'5
0
«o
20
?
16
CD
2
12
Gain
1
CL
3
&
5
O
8
4
0
4
Input Power (dBm)
Figure 47. Output power and gain as functions of power measured from two-stage
composite-channel HEMTs MMIC amplifiers under on-wafer pulse power conditions. A
saturated output power of 25 dBm (316 mW) achieved at 94 GHz [132].
127
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
channel MD structure grown by SSMBE. However, there has been not enough work to
understand the achievable 2DEG transport properties in the InAIAs/InAsxP|.x/InP
composite-channel MD structures.
5.1 In A sJW In P Multi-Quantum Wells
Determining the composition of strained InAsP is not as simple as for latticematched films since a film thickness less than the critical thickness for the coherent
growth is hard to measure by x-ray diffraction and, for thick films greater than the critical
thickness, the composition is shifted due to strain [112]. In order to determine the
relationship between the As4 and P2 beam equivalent pressure (BEP) and the As
composition, x. in the InAsxP|.x alloy, InP/lnAsxP|.x/InP MQW structures have to be
grown and analyzed with x-ray diffraction measurements and simulations. The optical
quality of these MQWs is evaluated by photoluminescence measurements. The
composition and well thickness as determined by x-ray analysis can be confirmed
through the measured PL emission peaks and compared to corresponding transition
energy levels calculated with the x-ray data.
The InP/InAsxPi.x/lnP MQW samples were grown on epi-ready InP:Fe (100)
semi-insulating substrates in a solid source MBE system equipped with conventional
Group 01 cells and valved-cracker sources for arsenic and phosphorus. Phosphorus was
evaporated at a temperature of 92 °C and subsequently cracked at 920 °C, producing
129
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
dimeric P;. For the arsenic source, As4 was used to allow greater control of As/P flux
ratio in material [113]. A Pi/ln BEP ratio of 17-18 was used for the growth of InP/InAsP
MQWs. The growth rates and temperature of InAsP/lnP was 0.53 pm/hr and 500 °C,
respectively.
Three-period InP/InAsP MQWs were grown in order to relate the arsenic flux to
the As composition in the InAsP. The As percentage was varied up to a maximum of 74
% corresponding to a strain of 2.4%. Figure 48 (a) shows a schematic of the InP/InAsP
MQW structure used for this study. To prevent substantial relaxation due to strain relief,
quantum well thicknesses of 4.5 nm-5.6 nm, which are less than the critical thickness, 6
nm, of strained InAso7jPo26 [114], were used. The InP barrier thickness was 20 nm-26
nm. All InP/InAsP MQWs were grown without interruption at the interfaces.
130
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
24 nm
InP or InAlAs
5.9 nm
InAsxP|.x (xi)
24 nm
InP or InAlAs
5.9 nm
InAsxP|.x (x2)
24 nm
InP or InAlAs
InAsxP|.x (x3)
5.9 nm
200 nm
InP or InAlAs
InP(100):Fe Substrate
InP
20-26 nm
InAsxP|.x (xi) 4.5-5.6 nm
InP
20-26 nm
InAsxP|.x (x2) 4.5-5.6 nm
InP
20-26 nm
InAsxP|.x (x3) 4.5-5.6 nm
InP
200 nm
InP(100):Fe Substrate
(b) xi<x:<x3
(a) X|=X:=X3
Figure 48. Schematic structures of the InP/InAsP MQW’s and InAlAs/InAsP MQWs.
Double crystal x-Ray diffraction was used to determine the compositions and well
thicknesses of InAsP in the InP/InAsP MQWs. The compositions were determined by
matching experimental X-ray rocking curves to simulated curves with a dynamic
diffraction theory model [110]. The PL measurements were performed in at 300K, 77 K,
and 4.2 K with the samples mounted on a Cu block. The PL was analyzed using a 0.85 m
SPEX double monochromator equipped with a liquid N2 cooled Ge detector. The power
density of Ar laser was set to 80 mW/cm2.
In order to identify the observed PL spectral peaks in the InAsP/InP and
InAsP/InAIAs MQWs, calculations of transition energies, based on the envelope-function
approximation for an abrupt square well and deformation potential theory in the strained
131
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
QW structure [43] have been carried out. The InAsxPt.x layers grown on the (100) InP
substrates experience a tetragonal distortion due to the built-in biaxial compressive strain.
The hydrostatic component of the strain changes the energy band gap of the InAsP. The
uniaxial shear component of the strain introduces a splitting between the heavy-hole
valence band and the light-hole valence band. Furthermore, it introduces mixing of the
light-hole band and the split-off valence band, which results in a nonlinear correction to
the light-hole valence band by moving the edge towards the heavy-hole band.
The 4.2 K band gap for the strain-free InAsxP|.x was determined by an
interpolation between the Varshni equations for InAs [115] and InP [116] as a function of
temperature. A 4.2 K band gap of 1.508 eV [117] and material parameters presented in
Watanabe et al.’s theoretical work [118] for lattice-matched InAlAs were used in the
calculation of the transition energies for InAlAs/InAsP MQWs.
All other material parameters, such as the lattice constants, elastic stiffness, band
gap deformation potentials, and strain-free effective masses for different carriers, were
evaluated using a linear interpolation between values for InAs and InP [118]. The
temperature dependence of these parameters is neglected. The strain-dependent band gap
of bulk InAsxP|.x on a (100) InP substrate is shown in Figure 49 as a function of x at
T=4.2 K. The compressive strain in InAsP increases the bandgap.
Kane’s three-band model and the strain effects deduced from deformation theory
were used to determine the bound states for electrons and light-holes. Band nonparabolicity is inherent to this model. The heavy-hole band was treated separately with a
132
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
standard single-band Kronig-Penney model since it is decoupled from other states. In
order to obtain the transition energy, the exciton binding energy was calculated using a
hydrogen model. A valence band offset of 25% [119] was assumed.
The As composition in the InAsP alloy layers was determined by matching the
measured double-crystal x-ray rocking curve (DCRC) to that simulated by the dynamic
diffraction theory. Abrupt interfaces were assumed. Figure 50 shows a measured DCRC
and the corresponding simulation for a sample with x = 0.47. This measured result was
matched to the theoretically predicted curve for x = 0.47, a well thickness of 5.4 nm, and
a barrier thickness of 20.8 nm. From a sensitivity analysis (see Figure 51-55), the
composition and quantum well thicknesses of InAsP could be determined within 2% and
0.2 nm, respectively.
In order to investigate the relationship between the incident As4 BEP and the As
composition, x. incorporated in the InAsxPi.x well, InAsxP|.x/InP MQWs were grown.
The incident arsenic BEP was varied from 0 to 5.5 x 10'6 torr, without any interruption at
interfaces. The structures were analyzed using the x-ray diffraction. As seen in Figure 56,
the measured x-ray rocking curves clearly show changes in the Pendellosung fringes of
InAsxP|.x/InP MQWs resulting from the variation of arsenic incorporation in the well.
Figure 57 shows the relationship between the incident arsenic flux and the arsenic
composition, x, incorporated in the InAsxP|.x well as determined by matching x-ray
diffraction measurements and simulations of InAsxP|.x/InP MQWs. As shown in Figure
57, there exists a nonlinearity in the incorporation of arsenic for the growth InAsxP|.x
133
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
alloy, i.e., at relatively low As4 flux, near unity incorporation efficiency is obtained, but
for InP/InAsxPi.x/InP MQWs with higher As compositions, the incorporation efficiency
of arsenic decreases. For InP/lnAsxP|.x/InP quantum wells grown by gas-source MBE
[112, 120, 121] and SSMBE [109] this nonlinear incorporation efficiency of arsenic has
been reported. We obtain a 28 % lower incorporation of arsenic in high composition
(InAso 74P0 .26 ) samples in comparison to the lower composition samples. The mechanism
governing nonlinear
incorporation
rate
in
the growth
of these mixed-anion
heterostructures is not currently clear. Our results in the InAs0 7 4 Po.26 MQW sample can
be compared to a 20 % lower incorporation of dimer arsenic incorporated in gas source
MBE-grown InAsmPo* epitaxial layers [112].
Room temperature PL emission (Figure 58) of InP/InAs047 Po.53 quantum wells,
emitting at approximately 1.3 pm, is intense with full widths at half-maximum (FWHM)
in the range of 19-30 meV. The FWHMs of 77 K PL in three InP/InAsP MQWs with
different As mole fractions (0.3<x<0.54) vary from 12 meV to 16 meV (see Figure 59).
Each peak of the PL spectra observed in the InP/InAsP MQWs corresponds to each E lHH1 (the first electron bound-state and heavy-hole bound-state) transition for the three
quantum wells each with a different As composition at 300K, 77 K, and 4.2 K, as shown
in Figure 60. Figure 61 shows a comparison of the PL peak energy measured for
InP/InAsxP|.x MQWs with the theoretical calculations at 4.2 K. A well width of 5.9 nm
and barrier width of 24 nm as determined the double-crystal x-ray diffraction
measurement was used in this calculation. As this figure shows, the experimental PL
134
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
spectra versus As composition in InP/InAsP MQWs agrees well with the theoretical
prediction. This agreement was used as a check of our composition and thickness
determinations by the x-ray diffraction measurements.
The impact of an anion flux interruption at InAsP/InP interfaces on x-ray
diffraction of InP/InAs0 .6 Po.4 MQWs grown on InP (100) was investigated. As can see in
Figure 62, an InP/InAs06Po•* MQW grown with a pause (1 min) under a Pi flux showed
the same x-ray diffraction pattern with that of InP/InAso 6Po ■» MQW grown without any
interruption. However, in InP/InAs0 6 Po4 MQWs grown with a pause (1 min) under an
As4 flux, a clear change of x-ray diffraction pattern was obtained.
135
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4.2 K
1.8
1.2
SPLIT-OFF
0.8
LH
HH
UNSTRAINED -
0.4
0.2
0.4
0.6
0.8
As composition in InAsP
Figure
49. Energy band gap of InAsxP|.x on an (100) InP substrate versus As
composition x at 4.2 K. The compressive strain in InAsxP|.xincreases the band gap for the
heavy-hole (HH), light-hole (LH), and split-off valence band, respectively.
136
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1000
X-ray diffraction intensity (a.u.)
(a) measurement
0.001
(b) simulation
-2000
0
-1000
1000
2000
Angle(arcsec)
Figure 50. X-ray rocking curve (a) experimental and (b) theoretical of a three period
InP(20.8 nm)/InAso.47Po.53(5.4 nm) MQW.
137
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
InP/lnAs1P1*1 MQW
X-ray DCRC simulation
L./L = 5 nm/25 nm
X-ray diffraction
Intensity (a.u.)
x = O.i
x = 0.7
x = 0 .6
= 0.5
1000
0.1
x = 0.3
0.001
x = 0.2
-2000
0
-1000
1000
2000
Angle (arcsec)
Figure 51. X-ray DCRCs simulation of InP (5 nm)/InAsxP|.x (25 nm) (0.2 < x < 0.8)
MQWs.
138
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
10000
A313 (x = 0.47)
— A315 (x = 0.49)
InP/lnAsI P1*1 MQW
X-ray DCRC
X-ray diffraction Intensity (a.u.)
1000
100
0.1
•2000
0
•1000
1000
2000
Angle (arcsec)
Figure 52. X-ray DCRCs measured in two InP/InAsxP|.x MQWs with different
compositions (x = 0.47 and 0.49).
139
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1
•
0.1
i
i
i
i
|
i
i
i
I
1 I'
i
1
I"
[
InP/lnAsi Pi>i MQW
'
i
i
i
i
x = 0.45
x = 0.43
x = 0.44
X-ray simulation
LJLe = 5.6/20.8 nm
0.01
maximum peak shift
(15-35 arcsec)
0.001
C
c
o
o
(0
c
■o
>.
(0
0.0001
X
10'6
10'7
-1000
-500
500
1000
Angle (arcsec)
Figure 53. X-ray DCRCs simulated in InP (20.8 nm) /lnAsxPi-x (5.6 nm) MQW with
different compositions (x = 0.43,0.44, and 0.45).
140
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
0.1
InP/lnAsXP1-X MQW
x = 0.45, L = 5.6 nm
X-ray simulation
L s 20.8 nm
x = 0.45, L a 5.5 nm
x = 0.45, L = 5.7 nm
0.01
maximum peak shift
(15 arcsec)
0.001
0.0001
>.
-1000
0
-500
500
1000
Angle (arcsec)
Figure 54. X-ray DCRCs simulated in InP (20.8 nm) /InAso^Po.ss MQW with different
well thicknesses (Lw - 5.5,5.6, and 5.7 nm).
141
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
:
InP/lnAs P
X
1-*
MQW
X-ray simulation
- x = 0.45, L = 5.6 nm. L = 20 8 nm
«
- x = 0.45,
B
= 5.6 nm, Lg = 20.7 nm
>• x = 0.45, Lw = 5.6 nm, Lg = 20.9 nm
0.01
>.
’35
c
0)
0.001
c
c
maximum peak shift
(5 arcsec)
o
0.0001
x
-2400
-2200
-2000
-1800
-1600
Angle (arcsec)
Figure 55. X-ray DCRCs simulated in InP/InAso45 Po.55 (5.6 nm) MQW with different
»
barrier thicknesses (L b = 20.7,20.8, and 20.9 nm).
142
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1 1 - 1 1
10'
InP/lnAsXP1«x MQW
X-ray DCRC
10s
3
(0
C/3
c
107
0)
c
o
u
<n
■o
>>
(5
10"
10J
10'
10''
-2000
-1000
1000
2000
Angle (arcsec)
Figure 56. X-ray double crystal rocking curves (DCRCs) of a three-period InP/lnAsxP|.
x/InP MQW (Figure 48) grown with various arsenic fluxes.
143
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
0.6
0.5
0.4
0.3
0.2
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
As composition in InAsP
Figure 57. The ratio of arsenic flux and V element flux (arsenic flux + phosphorous flux)
versus As composition determined by x-ray diffraction measurement and simulation in
the abrupt InP/InAsxPi.* MQWs.
144
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
10000
300 K
PL intensity (a.u.)
8000
InP/lnAs P
6000
MQW
A315 (X = 0.49)
A313 (x = 0.47)
4000
2000
0.8
0.85
0.9
1
0.95
1.05
1.1
1.15
1.2
Energy (eV)
Figure 58. 300 K PL emission of InP/InAsxP|.,i (x = 0.47 and 0.49) MQWs (Figure 48
145
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
12000
InP/lnAs P
I
77 K
MQW
Photoluminsscsncs
11000
6000
PL intensity (a.u.)
1-1
A312
A314
A317
6000
0.3 < x < 0.54
FWHM: 12- 16meV
4000
2000
0
0.8
0.9
1
1.1
1.2
1.3
Energy (eV)
Figure 59. 77 K PL emission of InP/InAs*P|.x (0.3 < x < 0.54) MQWs (Figure 48 (b)).
146
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
11000
InP/lnAsI P1-M MQW (A317)
'
77
'
Photoluminescence
PL intensity (a.u.)
8000
6000
300 K
4000
4.2 K
2000
0.8
0.9
1
1.1
1.2
1.3
1.4
Energy (eV)
Figure 60. Temperature dependence of PL emission measured in InP/InAsP MQW, A317
sample.
147
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4.2 K
InP/InAsP MQW(24 nm/5.9 nm)
Calculation
>
£■
>.
O)
v
c
V
c
o
«<n
c
<
0
0.9
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
As composition in InAsP well
Figure 61. The relationship between the calculated transition energies and As
composition x in InP (24 nm) / InAsxP|.x (5.9 nm) MQWs. A heavy-hole valence band
offset of 259c was assumed in this calculation. The circles represent the experimental
data.
148
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
I i r i—i—| i i i—i—[-
107
InP
106
InAsP
InP/InAsP MQW
As /P BEP
4
InAsP
X-ray diffraction Intensity (a.u.)
105
2
(4 x 10 /4.5 x 10 torr)
InAsP
104
. InP (100) substrate
i: :i L1 min pause under As flux
103
102
interruption
10
10°
1 min pause under P flux
10' ’
-3000
._ A
1
t
-2000
_i_
1
i
-1000
I
0
I
I
I
i
i
1000
4
lit
2000
t. i n t
3000
Angle (arcsec)
Figure 62. Comparison of x-ray DCRCs of InP/InAso.ePo4 MQWs grown on an InP (100)
substrate with an anion flux interruption at InP/InAsP interfaces.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5.2 InAs«Pi../lnAlAs Multi-Quantum Wells
For the growth of InAlAs/lnAsP heterostructures, a large difference in As4 flux is
required for the individual alloys. This As flux change must be made during the transition
from one material to the other at the interface. For InAIAs, a high As4 beam equivalent
pressure on the order 105 torr is required, whereas, for the growth of lnAsxP|.* a lower
A s4 beam equivalent pressure of 10'6 to 10'7 torr is used. We chose to use As4 instead of
As; for better control of the flux, and therefore composition, during the growth of InAsP.
As4 incorporates less efficiently and thus a higher, more controllable flux, can be
reproduced and measured prior to growth [112].
The influence of the switching conditions between the As4 and P: fluxes at the
interfaces of strained InAlAs/lnAsP is assessed by comparing the low temperature PL
properties of ln0 s’Alo 48As/lnAsP MQWs with those of InP/InAsP MQWs. The observed
PL spectra are compared to calculations of electron and hole energy levels in the QWs
based on the envelope-function approximation and deformation potential theory. The
band non-parabolicity and strain-induced valence band mixing are included in the
calculations.
InAlAs/lnAsP MQWs with the same quantum well thickness and compositions as
in the reference InP-barrier samples (see Figure 48 (b)) were grown to investigate the
effects of InAIAs growth conditions on the inverted and normal interface of the wells.
150
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
InAlAs/InAsP/InAlAs Interface
InAIAs
InAIAs
InAsP
a
3
C
in
In & Al
t|
t2
I3
t4
ts
Time(s)
Figure 63. The schematic diagram of the growth sequence at the interface of
InAlAs/lnAsP MQWs. t3-t2 and ts-u are an interruption time under P flux at the top
interface of InAsP and under As flux at the bottom interface, respectively.
Figure 63 shows a schematic of the growth sequence at the interface of
InAlAs/lnAsP MQWs. As stated earlier, a high As beam equivalent pressure on an order
of jo -5 j orr an£j a relatively low As beam equivalent pressure of 10"6 Torr were used for
the growth of the InAIAs barrier and InAsP well, respectively.
151
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
To assess the affect of the P flux impinging on the surface of InAIAs, the growth
of the inverted interfaces was changed in the following manner. A pause of either t3-t;=
15 sec or 60 sec under the P2 flux was used after a delay time of ti-ti=120 sec under
arsenic flux. To study the influence of an impinging As flux on the surface of the InAsP,
a pause of either ts-t4 = 60 sec or 120 sec under the arsenic overpressure was used.
Samples A318 and A319 were grown with t5-L»=120 sec and t3 -t:= 15 sec and 60 sec,
respectively. Samples A320 and A321 were grown with ts-t4=60 sec and t3 -t2= 15 sec and
60 sec, respectively.
Figure 64 shows a comparison of the 4.2 K PL spectra for the InAlAs/lnAsP
MQWs and the InP/InAsP MQWs (A317). In this figure, the peaks numbered 1, 2, and 3
correspond to As compositions of 0.32, 0.48, and 0.54, respectively, in the InAsP of the
InP/InAsP reference MQW. As see in this Figure, four PL peaks in the InAlAs/lnAsP
MQW were observed. The fourth PL peak, which is different from the three peaks
corresponding to each quantum well, as expected in the InP/InAsP MQWs, is explained
below. Generally, this comparison shows that the emission peaks in the InAIAsAnAsP
MQW are shifted and broadened as compared to those of the InP/InAsP MQW.
In order to interpret the 4.2 K PL spectra of the InAIAsAnAsP MQWs, the
transition energies were calculated using the same well and barrier thicknesses as the
InPAnAsP MQWs.
Figure 65 shows the calculated transition energies versus the As
composition, x, and the PL peak energies of the InAlAsAnAsxPi-x MQW. The
corresponding composition in the reference InPAnAsP MQW was used for the
152
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
composition, or x value. There is a large difference in the experimental data and
calculated results. This probably results from the presence of a thin InAsP layer produced
at the InAlAs/lnAsP interface by the As-P exchange reaction that occurs during the
growth interruption.
In all the InAIAsAnAsP MQWs, the PL linewidths were broadened significantly
(by ~ 100%) as compared to those in the InPAnAsP MQW. At 4.2 K, the FWHMs of the
InAIAsAnAsP MQWs varied from 22 meV to 50 meV. In the Figure 66, a comparison of
the FWHM of the InPAnAsP MQWs and the InAIAsAnAsP MQWs versus As
composition is made. This shows two trends. First, the FWHMs are more broadened for
quantum wells with the lower As composition. They are also broader for the samples
grown with a longer InAsP surface exposure to arsenic flux.
Table 8 shows the relative position of the 4.2 K PL peak energy for the
InAIAsAnAsP MQWs as compared to the InPAnAsP MQW with the same composition.
Each peak shift was obtained by subtracting the peak energy of InPAnAsP MQW from
the corresponding peak energy for the InAIAsAnAsP MQWs. The FWHM and the
emission energy for the peaks are highly dependent on the growth conditions of the top,
or normal interface, in particular, the time under which the surface was exposed to the
A s 4 flu x.
This is consistent with observations that an As flux roughens the surface of P-
containing alloys [105]. The effect of the As4 overpressure was dependent on the
composition of As in InAsP. A lower percentage of As in the InAsP alloy increased the
sensitivity of the surface to the roughening induced by the As4 overpressure. We believe
153
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
this is due to the increase probability of substitution of As atoms for P since the
concentration of P is greater.
Two PL peaks
(numbered 3 and 4) for the InAsP well with x=0.54 in the
InAlAs/lnAsP MQWs were observed. In order to assign an origin to these transitions, a
calculation of the E2-HH1 transition and E1-LH1 (the first light-hole bound-state)
transition was performed for the abrupt Ino.s:Alo.48 As (24 nm)/lnAsxPi.*(5.9 nm) quantum
well. The energy difference of the HH-LH and E2-E1 has a larger value (by a factor of
three) than the energy difference of the observed peaks (see Figure 67). The calculation
of transition energies in the InAlAs/lnAsP MQW indicates that all 4.2 K PL peaks
measured in the InAlAs/lnAsP MQWs result from the E1-HH1 transition.
The largest energy difference between the peaks numbered 3 and 4 corresponds to
an -5 monolayer (M L) thickness of the quantum well. Yang and co-workers [122]
demonstrated in their study on the InAsP layer that is formed by As-P exchange on an
(100) InP surface that as much as 90-95% of the P could be exchanged by As. The
average depth of this reaction increases with an increase in As exposure time, being as
deep as 5.5 ML for an exposure longer than about 100 seconds. Our peak-splitting is
consistent with a fluctuation in the well thickness originating from the formation of InAs
islands during the As-P exchange reaction.
At the bottom, or inverted InAsP interface, a longer exposure under P2 flux results
in a smaller shift of peak energy. This may result from the better stabilization of the P2
flux after opening the shutter, i.e. the composition is more uniform in the well.
154
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
In summary, we have studied the growth of InP/InAsxP|.x and InAIAs/InAsxP|.x
heterostructures using solid P and As sources. The As.» flux is incorporated much less
efficiently into higher As percentage InAsP due to the higher strain in the grown InAsP
film. The InP/InAsP MQWs showed photoluminescence FWHMs of 12 meV-19meV at
4.2 K. Each peak energy corresponding to three quantum wells with different As
compositions
agreed
well
with
calculations
based
on
the
envelope-function
approximation and deformation potential theory at 4.2 K. The broadening and shift of PL
peaks observed in the InAlAs/lnAsP MQWs as compared to those of the InP/InAsP
MQWs indicate that the photoluminescence properties of the InAlAs/lnAsP MQWs are
highly dependent on growth conditions, in particular the exposure time of the InAsP
surface to As flux during growth interruption. This results from As-P exchange at the
interface.
155
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4.2 K
Peak 3
PL intensity (mV)
8000
Peak 4 • •
InP/InAsP MQW(A317)
InAlAs/lnAsP MQW: A318
A319
A320
A321
6000
4000
Peak 2
2000
Peak 1
0*
0.9
1
1.1
1.2
1.3
Energy (eV)
Figure 64. A comparison of the 4.2 K PL spectra for InAIAs/ InAsxP|.* MQWs and InP/
InAsxPi.x MQW.
156
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4.2 K
InAlAs/lnAsP MQW(24 nm/5.9 nm)
Calculation
A321
A320
A319
A318
•
■
a
♦
>
QJ.
>.
OJ
0
)
c
0)
c
o
55
c
re
0.9
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
As composition in InAsP well
Figure 65. The relationship between the calculated transition energies and As
composition x for Ino^AlojsAs (24 nm)/InAsxP|.*
(5.9 nni) MQWs. A heavy-hole
valence band offset of 25% was assumed in this calculation. The various marks represent
the experimental PL data at 4.2 K.
157
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
70
'
I
I
!
!
“ i—
•
o
□
4.2 K
60
X
?
t— r
i
■
i
r
i
InP/lnAsP MQW(A317)
InAIAs/lnAsP MQW:A318
A319
A320
A321
50
>
40
E
2
X
30
20
10
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
X
Figure 66. A comparison of the FWHM (full width at half-maximum) of PL spectra for
InAIAs/ InAsxP|.x MQWs and for InP/ InAsxP|.x MQW with the same As composition x
in InAsP.
158
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 8. The relative peak position of the 4.2 K PL spectra for InAlAs/ InAsxPi.x MQWs
as compared to InP/ InAsxP|.x MQW grown with the same growth conditions and
structure without interruptions.
Sampl
e
A318
A319
A320
A321
Interruption
time(sec)
trt:
ts-L
15
60
15
60
120
120
60
60
Shifted energy(meV)
Peak
1
-117
-110
-81
-78
Peak 2
-31
-13
-41
-31
Peak
3
37
2
-10
-6
Peak 4
86
25
38
40
* This value was determined by Gaussian-deconvolution.
159
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
300
4.2 K
InAIAs/lnAsP MQW(24 nm/5.9 nm)
250 -
200
A
o
□
x
150
HH-LH
E2-E1
A321
A320
A319
A318
100
0.45
0.5
0.55
0.6
0.65
0.7
0.75
0.8
As composition in InAsP well
Figure 67. A comparison of the measured energy differences in the two peaks (numbered
3 and 4) and the calculated HH (heavy-hole)-LH (light-hole) and E2-E1 (Eland E2 are
electron bound-states) energy difference as a function of As composition x for InAlAs/
InAsJV* MQWs at 4.2 K.
160
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5.3 InASvPi^AnPComposite-Channel MD Structures
Using the optimized conditions obtained in the study of InGaAs/InP compositechannel MD structures and InAsP/InP or InAsP/lnAIAs MQWs, InAsP/InP compositechannel MD structures have been grown on InP (100) substrates by SSMBE. In order to
investigate the achievable 2DEG properties in this pseudomorphic MD system, the
impact of various parameters such as As composition (0 < x < 0.74), InAsP channel
thickness, InAIAs spacer thickness, and doping scheme on 2DEG conductivity has been
studied. In addition, the influence of the group V element flux switching scheme at the
arsenic/phosphide heterointerfaces on the 2DEG conductivity has been characterized. To
further improve the 2DEG mobility in InAsosPu-i/InP composite-channel MD structures,
a strain compensation approach has been applied to this compressively strained MD
structure.
In InA1As/InAsxPi.x/InP composite-channel MD structures, the low field 2DEG
mobility and concentration have been investigated as a function of various parameters
including the As composition in the InAsxP|.x channel, exposure time of As4 fluxes at the
InAIAs/InAsxP|.x interface, doping thickness product, and spacer and channel thickness.
The InAlAs/InAsxPi.x/InP composite-channel MD structures were first grown at
500 °C to investigate the influence of As composition in the InAsxP|.x channel on the
2DEG conductivity. In these M D structures (Figure 68 (a)), the thickness of the InAsxP|.x
and InP channel was 12 and 38 nm, respectively. Si was used as the n-type dopant source.
The doping level of the doped InGaAs cap layer and InAIAs donor layer was 7 x 1018 and
161
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
8 x 1018cm'3, respectively. In general, to obtain the good 2DEG conductivity, the channel
material must have a low as-grown residual carrier concentration.
Our undoped InP
epilayers grown on the InAIAs buffer usually have a residual electron concentration of 1
x 1016 cm'3 at 300 K, larger than that of as-grown InGaAs (1 x 1015 cm'3) and InAIAs
film (mid-1014 cm'3). This is most likely due to the incorporation of impurities from the
source materials and the MBE background ambient. A thin graded InAsP buffer (2 nm),
which was grown on the InAIAs buffer using Pi flux and an exponentially decreasing the
Asj flux with As cracker valve closure, was inserted in order to reduce the residual
electron concentration in the channel of the InAlAs/InAsxP|.x/InP composite-channel MD
structures.
162
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
lnGaAs:Si(7x!018 cm'1)
InAIAs
InAlAs:Si(8 x 1018 cm ’)
InAIAs
InAsxPi.x
InP
InAsP
InAIAs
InP substrate
7 nm
2 0 nm
8 nm
InGaAs:Si(7xl018 cm’3)
lno.4 Alo.6 As
InAlAs:Si(8xl0 18 cm'3)
InAIAs
lnAso6Po4
InP
InAsP
InAIAs
InP substrate
nm
1 2 nm
38 nm
2 nm
2
250 nm
(a)
Figure
68
7 nm
2 0 nm
ti
t2
t3
50 - t.i nm
2 nm
250 nm
(b)
. The schematic diagram of the InAl As/In AsxP |x/InP composite-channel MD
structure: (a) an InAsxP ix channel M D structure with As composition variation and (b) a
strained InAIAs/InAs0 ()Po4 composite-channel MD structure with the InAs0 6 Po4 channel
grown at 420 °C. The thickness t|, ti, and tj were changed from 4 to
8
nm. from 2 to
8
nm. and from 12 to 4 nm. respectively.
Figure
68
(b) shows the schematic diagram of strained InAlAs/InAsofiPoo/lnP
composite-channel MD structures prepared for the transport studies. A 20 nm lno.4 Alo&As
Schottky layer was used in order to achieve a higher Schottky barrier height, thereby
reducing the gate leakage current. As stated above, the lnAsxP|.x/InP composite-channel
was used to take advantage of the relatively high mobility of InAsxP).x at low electric
fields, and the high breakdown and saturation velocity of InP at high electric fields in
163
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
analogy to the InGaAs/InP composite-channel HEMT. The thickness of the InAsxPi.x
channel was varied from 4 to 12 nm to investigate channel quantization effects. All
samples had smooth surface morphologies with no evidence of cross-hatching, as
confirmed using an optical microscope and atomic force microscope (AFM). The
electrical properties of the 2DEG in the InAlAs/InAsxP|.x/InP composite-channel MD
structures have been investigated using Hall measurements.
In order to investigate the influence of As composition, x, on the 2DEG transport
properties in the InAIAs/InAsxP|.x/InP composite-channel MD structure (Figure
68
(a)),
the As composition was varied from x = 0.14 to x = 0.72. In Table 7, 300 K and 77 K
Hall data for the InAl As/InAsxP(x/lnP composite-channel MD structure are summarized
and compared to that of the lattice-matched InAlAs/InGaAs/InP composite-channel MD
structure grown with the same epilayer design (shown in Figure
68
(a)) except for an
InGaAs channel. As the As composition, x, in the InAsxP|.x increases, the electron
mobility increases as expected, due to the lower electron effective mass of InAs as
compared to InP. The 2DEG concentration in the InAsxP|.x quantum well channel also
increases with increasing of As composition in the alloy, due to the increased conduction
band discontinuity realized by the decreasing bandgap of InAsxPi.x with higher x.
At room temperature, the energy band gap (0.756 eV) [10] of strained InAs0 6 Po4
is comparable to that of Ino.53Gao.47 As (0.74 eV) [10], but the conduction band
discontinuity of the InAIAs/InAsP heterostructure (0.75AEg) is much higher than that of
lattice-matched InAIAs/InGaAs heterostructure (0.65AEg) and thus can confine more
164
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
electrons in the channel with the same doping scheme and layer design, consequently
leading to an improved power capability.
Table 9. Hall data of lnAlAs/lnAsxP|.x/lnP composite-channel MD structure grown at
500 °C with various As compositions in the InAsxP).x channel. Hall data of latticematched InAlAs/InGaAs/InP composite-channel MD structure grown with the same
epilayer design (Figure 68 (a)) except the channel material was compared.
Sample
(x)
0.14
0.34
0.53
0.72
InAlAs/lnGaAs/lnP
Hall mobility
(cm2/Vs)
300 K 77 K
4200
11400
5600
16000
5700
14000
6300
16000
8900
22800
Carrier density
( 1012 cm'2)
77 K
300 K
2.9
2.9
3.4
3.1
3.8
3.5
3.9
4.0
2.7
2.9
As expected. Table 9 shows that the InAlAs/InAsxP|.x/lnP composite-channel
MD structure can confine a greater 2DEG density (3.5 x 10i: cm'2 at 300 K) in the
channel with a comparable As composition of x = 0.53 than the lattice-matched
InAlAs/lnGaAs/lnP composite-channel MD structure (2.7 x 1012 cm'2 at 300 K), due to
the difference of conduction band discontinuity. As summarized in Table 9, however, the
165
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2DEG mobility of the InAIAs/InAso.^Po-n/InP composite-channel MD structure (5700
cm2/Vs at 300 K and 14000 cm2/Vs at 77 K) is much less than that of the lattice-matched
InAIAs/lnGaAs/InP composite-channel MD structure (8900 cm2/Vs at 300 K and 23000
cm:/Vs at 77 K). Although strain affects the electron effective mass in the pseudomorphic
MD structures, the electron effective mass interpolated over binary materials [123] for
the unstrained InAso.53 Po.47 is 0.044mo, is comparable with that of the lno5 .1Gao.47 As
(0.046m„) [118]. According to a theoretical study, these two materials have similar bulk
transport properties, i.e., velocity versus electric field characteristics. This fact means that
the 2DEG mobility in the InAlA.s/InAs,)5 iPu47/InP composite-channel M D structure may
be improved and may be a comparable to that of the lattice-matched InAIAs/InGaAs/lnP
composite-channel MD structures. Thus, it is necessary to optimize the growth conditions
of InAl As/In AsP heterostructures in order to improve the 2DEG mobility.
Although the InP channel power HEMT [58] has high breakdown performance,
InAIAs/lnP/InAlAs quantum well heterostructures have relatively large sheet resistance
as compared to that of conventional lattice-matched InAIAs/InGaAs HEMT structure
(approximately 210 Q/cm2), leading to limitations for high frequency power device
applications. To overcome the high sheet resistance in the InAIAs/lnP/InAlAs MD
structures without sacrificing their good high field properties, the InAl As/In AsxP|.*/InP
composite-channel MD structures have been considered.
166
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
600
500
400
300
200
100
InAlAs/lnGaAs/lnP MD structure
0.2
0.6
0.4
0.8
As composition in InAsP
Figure 69. The 300 K sheet resistance of InAlAs/InAsxPi.x/InP composite-channel MD
structure (circle) shown in Figure 68 (a) versus As composition in the InAsxPi.x channel.
The sheet resistance of lattice-matched InAIAs/lnGaAs/InP composite-channel MD
structure, which was grown in the same batch and with the same epilayer design except
the channel material, was compared.
167
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure 69 shows a comparison of the sheet resistance of InAl As/In AsxP|.x/InP
composite-channel MD structures and lattice-matched InAlAs/lnGaAs/lnP compositechannel MD structure with the same layer design and doping scheme except the channel
material. The room temperature sheet resistance of the former structure with x = 0.53 is
290 Q/cm2, which is slightly higher than that of the lattice-matched InAlAs/lnGaAs/lnP
composite-channel MD structure (260 Q/cm2).
In general, an As4 overpressure roughens the surface of the P-containing alloys
due to the As-P exchange reaction. In the growth of InAlAs/InAsxP|.x/InP compositechannel MD structures, the 2DEG conductivity can be modified as a result of roughness
resulting from various interruption schemes.
In order to investigate the influence of roughness at the InAIAs/InAsxP|.x
interface on the 2DEG conductivity, the exposure time of As4 flux at the InAIAs/
InAs,)bPu4 interface was varied from 0 to 60 sec during the growth of the
InAIAs/lnAsubPoj/InP composite-channel MD structures. The growth temperature of the
InAs()flP04 channel was decreased to 420 °C to reduce the tendency towards threedimensional growth for strained materials. An As4 flux impinged on the surface of the
InAsxP|.x during the growth interruption at the InAlAs/InAso6Po4 interface, leading to
interface roughness via the As-P exchange. By Neave and Joyce’s model of the one­
dimensional disorder boundary [124], the surface of InAsxP|.x grown on InP (100)
substrate using an anion-stabilized scheme has an As-P face. Therefore, a longer
exposure of As4 fluxes results in the more exchange of As-P atoms at these interfaces.
168
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 10. Hall data for the InAlAs/lnAsobPoVInP composite-channel M D structure
grown with various interface interruption times.
Sample
A429
A428
A427
A420
Interruption
time(sec)
0
20
40
60
Hall mobility
(cm2/Vs)
77 K
300 K
14500
5500
11900
5500
4000
8200
8800
5000
As summarized in Table
10
Carrier density
( 1 0 12 cm'2)
300 K
77 K
3.9
4.1
4.0
4.2
4.1
4.2
3.9
4.0
, the growth interruption at the InAlAs/lnAsu6 Po4
interface does affect on the 2DEG conductivity of the InAIAs/InAs^Pfij/lnP compositechannel MD structures. The 2DEG concentration in the channel was not significantly
changed but the mobility, particularly at 77 K, was seriously decreased with increasing
interruption time. This result indicates that a longer pause under AS4 overpressure
roughens the InAs*P|.* surface more, resulting in a decrease of the electron mobility in
the channel. In the SSMBE-grown InAl As/In AsojPo^/InP composite-channel MD
structures reported by Hoke and coworkers [1I I ] , the 2DEG density and mobility were:
3.39 x 1012 cm 2 and 4300 cm2/Vs at 300 K and 3.31 x 1012 cm' 2 and 8520 cm2/Vs at 77
K.
These relatively low mobilities for an InAlAs/InAsojPo.r/InP MD structure as
169
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
compared to those of the InAlAs/InAs0 .34 Po66 /InP described in Table 9 (3.1 x 1012 cm'2
and 5600 cm2/Vs at 300 K and 3.4 x 1012 cm'2 and 16000 cm2/Vs at 77 K), in particular
77 K mobilities, may result from interface roughness.
The influence of the InAs06Po4 thickness on the 2DEG conductivity was
investigated by changing the InAso6 Po4 thickness from 12 to 4 nm. In the compositechannel M D structure, a thin channel is utilized to increase the effective band gap via
quantum size effects. As the thickness of the quantum well decreases, the quantized
energy levels in both the conduction band and the valence band increase, and the
distribution of the 2DEG concentration in the channel is closer to channel interfaces,
resulting in a decrease of the 2DEG density and mobility. Thus, a trade-off between
breakdown voltage and achievable current density is required for high frequency power
applications.
As shown in Figure 70. the 2DEG concentration is approximately unchanged
(from 3.9 x 101" to 3.75 x 1012 cm'2) and the 2DEG mobility was slightly reduced from
5500 to 5000 cm2/Vs as the InAsP channel thickness was decreased. Recently,
Meneghesso and coworkers [106] reported on InGaAs/InP composite-channel HEMTs
with variable InGaAs channel thickness, in which the 2DEG concentration decreased
from 4.2 x 1012 to 3.7 x 1012 cm'2 and the mobility decreased from 9100 to 7000 cm2/Vs
at 300 K as the InGaAs channel thickness was changed from 10 to 3 nm.
170
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
6000
300 K
4.5
5000
2DEG density ( x 10,z cm
5500
4500
4000
2
4
6
6
10
12
3.5
14
InAsP channel thickness (nm)
Figure 70. The influence of InAs0 .6 Po.4 channel thickness on 300 K 2DEG mobility and
density of the InAl As/In AsoePoVInP composite-channel MD structure.
171
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 11. Hall data for the InAIAs/InAs0.6 Po.4/InP composite-channel M D structure with
different spacer thickness of 2 to 8 nm.
Sample
A420
A431
A432
Spacer
thickness(nm)
2
4
8
Hall mobility
(cm2/Vs)
300 K
77 K
5000
8800
5700
16000
7200
27000
Carrier density
(1012 cm'2)
300 K
77 K
3.9
4.0
3.3
3.3
2.5
2.4
Table 11 shows the influence of InAIAs spacer thickness (2 - 8 nm) on the 2DEG
mobility and concentration. As the thickness of spacer increases, ionized impurity
scattering decreases, leading to an increase of the mobility. Moreover, the modulation
doping efficiency decreases, resulting in a reduced concentration of electrons in the
channel. As the spacer thickness increases from 2 to 8 nm, the 2DEG concentration was
decreased from 3.9 x 1012 to 2.5 x 1012 cm'2 at 300 K. The highest mobility of 7200
cm2/Vs at 300 K was obtained in the InA1As/InAso.ePoVInP composite-channel M D
structure with a spacer thickness of 8 nm. To the best of our knowledge, this is the
highest value measured in the InAs06Po.-i M D structures as shown in Figure 71.
172
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
A decrease in doping thickness product in the donor layer results in a decrease of
ionized impurity scattering leading to an increase of mobility, and a reduction of electron
concentration in the channel. Figure 72 shows the influence of doping thickness product
(Npx t|) on the 2DEG conductivity in the InAlAs/lnAso^PoVInP composite-channel MD
structure shown in Figure 68 (b). As the thickness of donor layer decreases from 8 to 4
nm with the same doping condition (No = 8 x 1018 cm'3), i.e., the doping thickness
product decreases from 6.4 x 1012 to 3.2 x 1012 cm'2, the 2DEG concentration and
mobility were changed from 3.9 x 1012 to 2.5 x 1012 cm'2 and from 5000 to 6000 cm2/Vs
at 300 K. respectively. The decrease of doping thickness product has relatively little
effect on mobility. These results indicate that possible parallel conduction in the InAIAs
donor layer is negligible in our MD structure design and doping scheme.
In summary, we have studied the transport properties of the 2DEG formed in
InAIAs/InAsxP|.,/InP composite-channel MD structures. As the As composition in the
InAsxP|.x channel increases, the 2DEG mobility and concentration increase due to the
change of electron effective mass and band gap in the channel, respectively. In the
InAl As/In Aso^Poa/lnP composite-channel MD structures, the influence of interruption
time during the formation of InAlAs/InAso6 Po4 interfaces, channel thickness, spacer
thickness, and doping thickness product on the 2DEG conductivity have been
investigated. The exposure time of As4 flux during interruption at the InAIAs/InAsoePo-i
interface considerably degrades the 2DEG mobility. The 2DEG mobility and
concentration, and the sheet resistance of the InAlAs/InAso.-jiPms/InP composite-channel
173
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
MD structure were 6300 cnr/Vs and 3.9 x 1012 cm'2, and 250 fl/cm2 at 300 K,
respectively. These results show the great potential of InAlAs/lnAsxP|.x/InP material
system for microwave and millimeter wave device applications.
174
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4
Bell Labs [Ref. 110]
MOCVD-grown InAsP/InP HEMT
Ratheon [Ref. 111]
SSMBE-grown InAIAs/lnAsP/lnP composite-channel HEMT
8000
8 nm spacer
7000
300 k
:
6000
2 nm spacer
(A
:>
"fe
u
E
u
*o
5000
<V
4000
10 nm spacer
n
0
CO
c
■0o)
O
O
CJ
E
15
1
3000
111
2000
1000
0
0.2
0.4
0.6
0.8
1
As composition in InAsP channel
Figure 71. Comparison of 300 K Hall data of InAsP channel HEMTs (The closed and
open marks are a mobility and 2DEG density, respectively.).
175
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
7000
300 K
6500
4.5
5500
5000
3.5
4500
4000
2DEG density ( x 10 * cm
6000
2.5
3500
3000
2
3
4
5
6
7
8
2
Doping thickness product (x 1012 cm'2)
Figure 72. The influence of doping thickness product (ND x ti) on 300 K 2DEG mobility
and density of the InAIAs/lnAs0 ePo.-j/InP composite-channel M D structure.
176
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5.4 Strain Compensation of InAs,Pi.,/InPComposite-ChanneI
The impact of strain-compensation in the InAlAs/InAsxP|.x/lnP material system
on the channel conductivity has been investigated. Theoretically, the electrical properties
of bulk InAso6 Po4 are similar to that of bulk InGaAs [125], Therefore, the 2DEG
mobility in InAso 6Po4/InP MD structures may be enhanced. Although the InGaP alloy
has been successfully utilized to compensate strain in the growth of InAsxP|.x/InP MQWs
[126], the Ga-rich InGaAs alloy may be a good choice due to the better electrical
characteristics and closer lattice-match with the InGaAs system. In addition, previous
results for the InGaP hole barrier indicate that the InGaP/InAsP interface roughness can
also greatly degrade the mobility in the InAsP channel. The strain-compensation
technology has been primarily utilized in order to improve mechanical stability or to
increase the critical thicknesses of strained epitaxial layers.
Figure 73 shows the schematic diagram of InAso 6 P0 4/InP composite-channel MD
structure strain-compensated with 10 nm Ini.xGaxAs (x = 0.55. 0.6, 0.7, 0.75) grown by
SSMBE. The growth temperature of 500 °C, which is the same temperature previously
used for the InAsP/InP MD structures. At the interfaces of InGaAs/InP and InAsP/InP, an
interruption of 1 min was used to achieve a stabilized Group V flux.
177
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
InGaAs: Si (7 x 10|H)
InAIAs
InAIAs: Si ( 8 x 10l8)
InAIAs
Ini.xGaxAs ( x = 0.55,0.6,0.7,0.75)
InAso.6Po4
InP
InAsP
InAIAs
InP (100): Fe substrate
7 nm
20 nm
8 nm
2 nm
10 nm
12 nm
38 nm
2 nm
250 nm
Figure 73. The schematic diagram of an InAs06P04/lnP composite-channel MD structure
strain-compensated with 10 nm Ini.xGaxAs (x = 0.55,0.6, 0.7,0.75) grown by SSMBE.
In the strain-compensated epitaxial structure, the degree of strain-compensation
can characterized by the compensation index defined as follows:
t2( Aa) J a _
t2( a - a : )
r,(Afl),/a
r,(a -a ,)
where r, and a, are the thickness and lattice constant of layer 1 and 2, respectively. In our
structure, the layer 1 and 2 correspond to an InAso.6 Po.4 (12 nm) epitaxial layer and Ini.
xGaxAs (10 nm) epitaxial layer, respectively. The compensation index for straincompensated 10 nm Ini.xGaxAs /InAso^PoVInP composite-channel M D structure was
calculated as shown in Table 12. The lattice constant values of the InGaAs and InAsP
178
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
alloys were interpolated from the lattice constant values of binary materials listed in
Table 1. The Ga percentage was varied from 55 to 75 %, which corresponds to a variation
in compensation index from 24.4 % to 83.7 %.
Table 12. Calculated compensation index of Ini.xGaxAs (10 nm)/InAso6Po.4 (12 nm)
epilayers grown on InP.
Material
Lattice Constant
(A)
Strain
(%)
InAs()6Po4
(compressive)
5.9826
-1.938
5.8356
5.8153
5.7748
5.7546
+0.566
+0.911
+ 1.601
+ 1.946
In ].xGaxAs
(tensile)
x = 0.55
x = 0.60
x = 0.70
x = 0.75
Strain
compensation
Index
179
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
0.244
0.391
0.689
0.837
Table 13. 300 K Hall data of the strain-compensated In|.xGaxAsAnAs0 6 Po4 /InP
composite-channel, pseudomorphic Ino.4 Gao.6 As single channel, and pseudomorphic
InAs0 6 Po4/InP composite-channel HEMT structures.
Sample
In() jG ao 6 As channel
(12 nm) HEMT
InAsofiPo 4/lnP
Mobility
(cm2/Vs)
7506
2DEG density
(x 1 0 12 cm*2)
2.7
Sheet Resistance
(ohm-cm)
305
6353
3.3
298
3.2
3.2
4.2
4.2
271
287
501
650
Composite-Channel
HEMT
Strain Compensation Ini ,xGaxAs/In As0 6Po 4 HEMT
6999
x = 0.55
x = 0.60
6839
2970
x = 0.70
x = 0.75
2273
Table
13 presents 300 K Hall data of strain-compensation Ini.xGaxAs
/lnAso6 Po4/lnP composite-channel M D structures as shown in Figure 73. Before the
growth of the strain-compensation composite-channel M D structures, a HEMT structure
with a single channel of
12
nm Ino.4Gao.6 As and with the same layer design except the
channel was grown to compare its channel conductivity. At 300 K, the 2DEG density of
2.7 x 1012 cm' 2 and mobility of 7506 cm2/Vs were achieved. This 2DEG density was
slightly reduced by -1 0 % as compared to that of standard InGaAs HEMT (3 x 1012 cm*2)
180
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
due to lower conduction band discontinuity resulting from larger bandgap of
Ino4Gao6 As. The reduction of mobility from 10,000 to 7506 cnr/Vs results from larger
effective mass of Ino.4Gao6As.
In addition, a pseudomorphic InAso^PoVInP composite-channel M D structure
without an Ini.*GaxAs layer for the strain-compensation was grown. The 300 K 2DEG
density of 3.3 x 10|: cm'2 and mobility of 6353 cm2/Vs were achieved. As shown in
Table 13, two strain-compensation structures with low strain compensation index (LSC,
24.4 % and 39.1 %), which correspond to Ga percentage of 55 % and 60 % for the
InGaAs compensation layer, showed a slightly increased mobility by - 10 9c with
approximately the same 2DEG density. In another set with two strain-compensation
structures with high strain compensation index (HSC, 68.9 % and 83.7 %), the mobility
dramatically decreased by greater than 50 % as compared to that of pseudomorphic
InAs06PoVInP composite-channel MD structure. A 2DEG density of 4.2 x 1012 cm'2 was
obtained. These HSC samples were difficult to measure due to the anisotropy of Hall
voltages at the four contacts.
Recently, Letartre et al.[128], reported the influence of the strain-compensation
on structural and electrical properties of InAlAs/InGaAs HEMT structures grown on InP.
They grew four HEMT structures: a lattice-matched InGaAs HEMT structure, a classical
pseudomorphic Ino.73Gao.27 As (approximately 2 % strained) HEMT structure, and two
compensated structures with LSC of 35 % and HSC of 80 % by a growing tensilely
strained InAIAs layer. They demonstrated that the 300 K mobility of the LSC and HSC
181
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
HEMT structure increased by ~ 10 % and degraded by ~ 50 %, respectively, as compared
to that of pseudomorphic HEMT and the mobility. They showed that the large reduction
in mobility for the HSC samples resulted from strain relaxation using Spectrally
Integrated Scanning Photoluminescence (SI-SPL), which clearly demonstrated the crosshatch generated on the surface of HSC samples. Despite this SI-SPL evidence, no
macroscopic variation of the lattice parameters has been detected by High Resolution Xray Diffraction (HRXRD) measurements.
The trends in channel conductivity obtained in our strain-compensation
InGaAs/InAsP/InP HEMT growths are similar to that of Letartre group’s. For all
samples, no macroscopic variation of surface morphologies was detected by microscopy
investigation. In our results, a low strain-compensation can only slightly improve the
mobility with an increase of the critical thickness of the strained channel and an improved
mechanical stability. At 300 K, the best channel conductivity (2DEG density of 3.2 x 10i:
cm 2 and mobility of 7000 cm:/Vs) was achieved in the strain-compensation
In()asGao ssAs/InAso6 Po4/InP composite-channel HEM T structure.
Very recently, Yoon et al. [127] demonstrated that the electrical properties of
SSMBE-grown InP films are strongly dependent on the growth parameters such as the
V/m ratio, cracker temperature, and growth temperature. In order to enhance the 2DEG
property InAsP/InP MD structures, further optimization of phosphide growth conditions
is required.
182
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER VI
CONCLUSIONS
In this thesis, the growth conditions and material design factors related to
optimizing the 2DEG properties of InP-based composite-channel HEMT structures have
been investigated through a comparative study of the 2DEG properties of various InPbased MD structures, such as conventional InAlAs/InGaAs M D structures, and
InAlAs/InGaAs/InP and InAlAs/InAsP/InP composite-channel MD structures grown with
the same SSMBE system.
The SSMBE growth of conventional InP-based MD structures have been studied
in order to obtain the optimum growth conditions and to investigate the influence of
material design factors on the 2DEG characteristics. To optimize the electrical properties
of the 2DEG in the standard lattice-matched InAlAs/InGaAs M D structure, various
experiments were designed to assess the influence of As4 flux, growth rate, superlattice
buffer and buffer thickness, growth temperature, interface interruption, planar doping
scheme, and channel depth (a summation of film thicknesses grown on the channel) on
the 2DEG conductivity. In the standard InAlAs/InGaAs HEMT structure, a 2DEG
mobility and density were optimized at approximately 10,000 cm2/Vs and 3 x 1012 cm' 2 at
183
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
300 K, respectively. This 2DEG density is consistent well with the equilibrium 2DEG
density theoretically calculated for the same structure.
Conventional InP-based power HEMT structures, such as doped-channel and
double-heterojunction InAlAs/InGaAs HEMT were grown by using SSMBE. In the
doped-channel MD structure, the 2DEG density was increased to 5.1 x I0 |: cm :,
however, the mobility was dramatically decreased by -30 % due to the impurity
scattering in the channel. With the double-heterojunction InAlAs/InGaAs HEMT
structure, as the planar-doping density increases, the 2DEG density increased up to 6.2 x
I01' cm' 2 and the 2DEG mobility decreased to approximately 7000 cm2/Vs. Therefore,
there exists a trade-off between the 2DEG density and mobility.
The impact of structural design parameters, such as channel thickness, the
inclusion of hole barriers. Al-rich Schottky barriers, and various doping profiles, on the
properties of InGaAs/InP composite-channel modulation-doped structures grown by
SSMBE have been investigated. The best combination of 2DEG density and mobility
(4.3 x lO12 cm' 2 and 10,300 cm2/Vs at 300 K) was achieved in the 20 nm InGaAs/InP
composite-channel HEMT structure with 30 nm InP subchannel delta-doped at the center.
The mobility of the structures with delta-doped subchannels depends critically on the
spacer layer thickness. As the InP spacer thickness is reduced from 15 nm to 4 nm, the
electron mobility degraded by ~ 50 % due to an increase in remote impurity scattering.
The InGaAs channel thickness was varied from 0 to 20 nm, while maintaining a constant
total channel thickness of 50 nm. The electron mobility is fairly insensitive to the InGaAs
184
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
channel thickness in the range of 5 to 20 nm, with a very high mobility of 11,016 cm2/Vs
achieved for the thinnest channel; however, when the InGaAs channel thickness is less
than 5 nm, the 300K mobility is degraded by > 30 % . In our results, the best mobility
was obtained with a 3 nm thick InGaAs channel. This results from a design trade-off
between the InGaAs thickness and the 2DEG conductivity in exploiting quantum size
effects. The optimum thickness, placement and composition of an InGaP hole barrier
were investigated. The conductivity of a structure with a 2.5 nm Ini.xGaxP (0 < x < 0.5)
hole barrier inserted between the channel and donor layer decreased in comparison to a
control. This can be attributed to interface roughness at the InGaP/lnGaAs interface.
The 0.15 pm-gate InGaAs/InP composite-channel HEMT showed comparable
DC and f, and fmax, and improved breakdown by approximately 1.5 V as compared to
conventional InGaAs HEMT fabricated with the same process (TRW InP-based HEMT
fabrication line). In addition, this composite-channel HEMT has demonstrated state-ofthe-art RF power performance at W-band.
The growth of InP/lnAsxP|.x and InAIAs/lnAsxP|.x multi-quantum wells using
solid P and As sources has been studied. The As4 flux is incorporated much less
efficiently into higher As percentage InAsP due to the higher strain in the grown InAsP
film. The InP/InAsP MQWs showed photoluminescence FWHMs of 12 meV-l9meV at
4.2 K. Each peak energy corresponding to three quantum wells with different As
compositions
agreed
well
with calculations
based
on
the
envelope-function
approximation and deformation potential theory at 4.2 K. The broadening and shift of PL
185
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
peaks observed in the InA1As/InAsP MQWs as compared to those of the InP/InAsP
MQWs indicate that the photoluminescence properties of the InAlAs/InAsP MQWs are
highly dependent on growth conditions, in particular the exposure time of the InAsP
surface to As flux during growth interruption. This results from As-P exchange at the
interface.
The channel conductivity of the 2DEG formed in InAlAs/InAsxP |x/InP
composite-channel MD structures has been investigated. As the As composition in the
InAsxP|.x channel increases, the 2DEG mobility and concentration increased due to the
change of electron effective mass and band gap in the channel, respectively. In the
InAlAs/InAsoePo j/InP composite-channel M D structures, the influence of interruption
time during the formation of InAIAs/InAso6 Po4 interfaces, channel thickness, spacer
thickness, and doping thickness product on the 2DEG conductivity have been
investigated. The exposure time of As4 flux during interruption at the InAlAs/InAso 6 P0 4
interface considerably degrades the 2DEG
mobility. The 2DEG mobility and
concentration, and the sheet resistance of the InAlAs/InAs0 72 Po.:8/InP composite-channel
MD structure were 6300 cm2/Vs and 3.9 x 1012 cm 2, and 250 fl/cm 2 at 300 K,
respectively. These results show the great potential of InAlAs/InAsxPi.x/InP material
system for microwave and millimeter wave device applications.
The impact of a strain-compensation in the InAIAs/InAsxP|.x/InP material system
on the channel conductivity was investigated. InAsoePo.-t/InP composite-channel MD
structure strain-compensated with 10 nm Ini.xGaxAs (x = 0.55, 0.6, 0.7, 0.75) grown by
186
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
SSMBE. A low strain-compensation (compensation index, |i = 24.4 % and 39.1 %)
slightly improved the mobility by -
10
% with an increase of the critical thickness of the
strained channel and an improved mechanical stability. However, a high straincompensation (p = 68.9 % and 83.7 %) dramatically degraded the mobility by greater
than 50 %. At 300 K, the best channel conductivity (2DEG density of 3.2 x 1012 cm' 2 and
mobility of 7000 cm2/Vs) was achieved in the strain-compensation (p = 24.4 %)
Ino.45Gao.s5 As/InAso6 P0 4 /InP composite-channel HEMT structure. In order to further
enhance the 2DEG mobility for InAsP/InP composite-channel HEMT structures, further
optimization of the phosphide growth conditions, such as the V/ID ratio, cracker
temperature, and growth temperature is required.
187
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
REFERENCES
[1] P.M. Smith, S.M.J. Liu, M.Y. Kao, P. Ho, S.C. Wang, K.H.G. Duh, S.T. Fu, and P.C.
Chao, “W-band high efficiency InP-based power HEMT with 600GHz f™X,"IEEE
Microwave Guided Wave Lett., vol. 5, no. 7, pp. 230-232, July 1995.
[2] L.D. Nguyen, A.S. Brown, M.A. Thompson, and L.M. Jelloian, "50nm self-aligned
gate pseudomorphic AlInAs/GalnAs high electron mobility transistor,” IEEE Trans.
Electron Devices, vol. 39, pp. 2007-2014, 1992.
[3] P. Ho, P.M. Smith. K.C. Wang, M.Y. Kao, P.C. Chao, and S.M.J. Liu, “60GHz power
performance of 0.1pm gate-length InAlAs/InGaAs HEMT’s," Proc. 6th Int. Con}. InP
and Related Materials, pp. 411-414, 1994.
[4] T. Enoki, T. Kobayashi. and Y. Ishii, “Device technologies for InP-based HEMT’s
and their applications to IC’s," GaAs IC Symp. Dig., pp. 337-339, 1994.
[5] P.M. Smith, "Status of InP HEMT technology for microwave receiver applications,"
IEEE Trans. Microwave Theory' Tech., vol. 44, no. 12, pp. 2328-2333, Dec. 1996.
[6 ] J. Dickmann. S. Schildberg, A. Geyer, B.E. Maile, A. Schurr, S. Heuthe, and P.
Narozny, “Breakdown mechanisms in the on-state mode of operation of InAIAs/InxGai.
*As pseudomorphic HEMT’s," Proc. 6th Int. Conf. InP and Related Materials, pp. 335338. 1994.
[7] T. Enoki, K. Arai, A. Kohzen, and Y. Ishii, “InGaAs/InP double-channel HEMT on
InP", in IEEE Proc. 4th Int. Conf. on InP and Related Materials, pp. 371-374, 1992.
[8 ] T. Enoki, K. Arai, A. Kohzen, and Y. Ishii, “Design and Characteristics of
InGaAs/InP Composite-Channel HFET’s”, IEEE Trans. Electron Devices, vol. 42, pp.
1413-1418, 1995.
[9] M. Matloubian, L.M. Jelloian, M. Lui, T. Liu, L.E. Larson, M. Le, D. Jang, and R.A.
Rhodes, “GalnP/InP composite channel HEMTs for millimeter wave power
applications,” Proc. Int. Conf. On Millimeter and Sub-Millimeter Waves and
Applications, pp. 579-580, 1993.
[10] A. B. Chen and A. Sher, in Semiconductor Alloys: Physics and Materials
Engineering (Plenum Press, New York, 1992), P280.
188
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[11] D. Yang, P. K. Bhattacharya, W. P. Hong, R. Bhat, and J. R. Hayes,”High-field
transport properties of InAsxPi.x/InP (0.3 < x < 1.0) modulation doped heterostructures at
300 and 77 K”, J. Appl. Phys. 72, pp. 174-178, 1992.
[12] R. P. Schneider, Jr and B. W. Wessels, “Photoluminescence excitation spectroscopy
of InAs0 67 P0 33/InP strained single quantum wells”, J. Electron. Mater. 20, pp. 1117-1123,
1991.
[13] S. Tiwari and D. J. Frank, ‘ Empirical fit to band discontinuities and barrier heights
in m -V alloy systems”, Appl. Phys. Lett. 60, pp. 630-632, 1992.
[14] J. J. Brown, A. S. Brown, S. E. Rosenbaum, A. S. Schmitz, M. Matloubian, L. E.
Larson, M. A. Melendes, and M. A. Thompson, “Study of the dependence of
Gao4 7 lno5 .iAs/AlxIni.xAs power HEMT breakdown voltage on Schottky layer design and
device layout”, IEEE Trans. Electron Devices, vol. 40, pp. 2111-2112, 1993.
[15] G. W. Wicks. M. W. Koch. J. A. Varriano, F. G. Johnson, C. R. Wie. H. M. Kim.
and P. Colombo, “ Use of a valved solid phosphorus source for the growth of GalnP and
AsInP by MBE", Appl. Phys. Lett. 59, pp. 342-344, 1991.
[16] D. L. Miller, S. S. Bose, and G. J. Sullivan, “Design and operation of a valved solidsource As: oven for molecular beam epitaxy", J. Vac. Sci. Technol. B8 , pp. 311-315,
1990.
[17] M. L. Dotor, D. Golmayo, and F. Briones, J. Cryst. Growth 127, 619, (1993).
[18] J. N. Baillargeon. A. Y. Cho, and R. J. Fischer, “Evaluation of the performance and
operating characteristics of a solid phosphorus source valved cracking cell for molecular
beam epitaxy growth of m-V compounds”, J. Vac. Sci. Technol. BI3, pp. 64-68, 1995.
[19] M. Toivonen, M. Jalonen, A. Salokatve, J. Nappi, P. Savolainen, M. Pessa, and H.
Asonen, “All solid source molecular beam epitaxy growth of strained-layer
InGaAs/GalnAsP/GalnP quantum well lasers (K = 980 nm)”, Appl. Phys. Lett. 67, pp.
2332-2334, 1995.
[20] W. H. Loh, D. Atkinson, P. R. Morkel, M. Hjopkinson, A. Rivers, A. J. Seeds, and
D. N. Payne, IEEE Photonics Technol. Lett. 5, 35 (1993).
[21] L. Esaki, “Semiconductor superlattices and quantum wells”, Proc. 17th Int. Conf.
Physics o f semiconductors, Springer-verlag, New York, P. 473, 1984.
189
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[22] R. Dingle. H. L. Stormer, A. C. Gossard, and W. Wiegmann, “Electron mobilities in
modulation doped semiconductor heterojunction and superlattices”, Appl. Phxs. Lett. 33,
pp. 665-667, 1987.
[23] J. Dickmann, K. Riepe, A. Geyer, B. E. Maile, A. Schurr, M. Berg, and H.
Daembkes. “Ino.5 :Al048 As/InxGai.xas (0.53 < x < 1.0) pseudomorphic high electron
mobility transistor with high breakdown voltages: design and performances”, Jpn. J.
Appl. Phys. 35, pp. 10-15, 1996.
[24] D. Delagebeaudeuf and T. Linh, “metal/(n) AlGaAs, GaAs two-dimensional electron
gas FET”, IEEE Trans. Electron Device. 29, pp. 955-960, 1982.
[25] T. J. Drummond. H. Morkoc, K. Lee, and M. Shur, “Model for modulation doped
field effect transistors", IEEE Trans. Electron Device.Lett. 3, pp. 338-341, 1982.
[26] T. Itoh, T. Griem, G. W. Wicks, and L. F. Eastman, “Sheet electron consentration at
the heterointerface in Al0asIno.szAs/Gao.47 ^ 0 .53 As modulation-doped structures". Electron
Lett. 21, pp. 373-374, 1985.
[ 27 ] K. S. Yoon, s. B. Stringfellow, and R. J. Huber, "Two-dimensional electron gas
density
calculation
in
Gao.47Ino.53As /Akuslno s:As,
Gao.47Ino5.3As/InP,
and
Gao.47Ino.53As/lnP/ Alo48 lno5 :As heterostructures", J. Appl. Phys 66, pp.5915-5919, 1989.
[28] T. Ando, A. B. Fowler, and f. Stem,”Electronic properties of Two-dimensional
systems". Rev. Mod. Phys. 54, pp. 437-672, 1982.
[29] J. Yoshida, "Classical versus quantum mechanical calculation of the electron
distribution at the n-AlGaAs/GaAs heterointerface”, IEEE Trans. Electron Device. 33.
pp. 154-156. 1986.
[30] K Lee, and M. Shur, "Electron density of the two-dimensional electron gas in
modulation doped layers”, J. Appl. Phy. 54, pp. 2093-2096, 1983.
[31] Y. Nakata, S. Sasa, Y. Sugiyama, T. Fujii, and S. Hiyamizu, “ Extremely high 2DEG
concentration in selectively doped Ino.53Gao.47As/n-Ino52Alo.48As heterostructurcs grown
by MBE” , Jpn. J. Appl. Phys. 26, pp. L59-L61, 1987.
[32] E. F. Schuber and K. Ploog, “Electron subband structure in selectively doped nAlxGai.xAs/GaAs heterojunctions”, IEEE Trans. Electron Devices 32, pp. 1868-1873,
1995.
190
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[33] Y. Ando and T. Itoh, “DC, samall-signal, and noise modeling for two-dimensional
electron gas field-effect transistors based on accurate charge-control characteristics”,
IEEE Trans. Electron Devices 37, pp. 67-78, 1999.
[34] D. Chattopadhyay, “Electron mobility in InGaAs quantum wells”, Phxs. Rev. B38,
pp. 13429-13431, 1988.
[35] T. Matsuoka, E. Kobayashi, K. Taniguichi, C. Hamaguchi, and S. Sasa,
“Temperature dependence of electron mobility in InGaAs/InAIAs heterostructures”, Jpn.
J. Appl. Phys. 29, pp. 2017-2025, 1990.
[36] J. W. Matthews and A. Blakeslee, "Defects in epitaxial multilayers”,
Growth 27, pp. 118-125, 1974.
J. cnst.
[37] R. People and J. C. Bean, “Calculation of critical layer thickness versus lattice
mismatch for GexSi|.x strained layer heterostructures”, Appl. Phys. Lett. 47, pp. 322-324,
1985.
[38] F. M. Poliak and M. Cardona, Phys. Rev. B72, p. 816, 1968.
[39] K. Nishi. K. Hirose. and T. Mizutani, Appl. Phys. Lett. 49,794 (1986).
[40] G. Bastard, “Superlattice band structure in the envelope-function approximation",
Phys. Rev. B24. pp. 5693-5697, 1981.
[41] G. Bastard."Theoretical investigations of superlattice band structure in the envelopefunction approximation", Phys. Rev. B25, pp. 7584-7597, 1982.
[42] G. Bastard. NATO school on MBE and heterostructures, Erice, March 1983
(Martinus Nijhoff. the Netherlands. 1983).
[43] G. Ji, D. Huang, U. K. Reddy, T. S. Handerson, R. Houdre, and H. Morkoc,
‘'Optical investigation of highly strained InGaAs-GaAs multiple quantum wells”, J.
Appl. Phys. 62, pp. 3366-3373, 1987.
[44] F. Ali and A. Gupta, HEMTs and HBTs: Devices, Fabrication, and Circuits, Boston:
Artech House, 1991.
[45] L.D. Nguyen, L.E. Larson, and U.K. Mishra, “Ultra-high-speed modulation-doped
field-effect transistor: a tutorial review,” Proceedings o f the IEEE, vol. 80, no. 4, pp. 494518.1992.
191
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[46] S.R. Bahl, W.J. Azzam, J.A.del Alamo, J. Dickmann, and S. Schildberg, “Off-state
breakdown in InAlAs/InGaAs MODFEET’s,” IEEE Trans. Electron Devices, vol. 42, no.
1, pp. 15-22, 1995.
[47] C. Heedt, F. Fuchali, W. Prost, F.J. Tegude, D. Fritzsche, and H. Nickel,
“Characterization of impact-ionization in InAlAs/InGaAs/InP structures using a novel
photocurrent-measurement technique,” Proc. 5lh Int. Conf. InP and Related Materials,
pp. 243, 1993.
[48] M. Matloubian, “InP-based HEMT’s for millimeter wave and submillimeter wave
power applications," Proc. Int. Conf. Millimeter and Submillimeter and Applications III,
pp. 22-32, 1996.
[49] J.J. Brown, M. Matloubian, T.K. Liu,L.M. Jelloian, A.E. Schmitz. R.G. Willson, M.
Lui, L.E. Larson, M.A. Melendes, and M.A. Thompson, "InP-based HEMT’s with AlxIni.
*P Schottky barrier layers grown by gas-source MBE,” Proc. 6lh Int. Conf. InP and
Related Materials, pp. 419-422, 1994.
[50]
K.B.
Chough,
C.
Caneau,
W.P.
Hong,
and J.I.Song,
“Alo.:s
Ino.75P/Alo.48ln<)52As/Gao..i5lNo.65As graded channel pseudomorphic HEMT’s high
channel-breakdown voltage", IEEE Electron Device Lett., vol. 15, no. 1, pp. 33-35, Jan.
1994.
[51] L.M. Jelloian. M. Matloubian, T.K. Liu, M. Lui, and M.A. Thompson, “InP-based
HEMT’s with Al(MnIn(>?;;As*P|.x Schottky layers,” IEEE Electron Device Lett., vol. 15.
no. 5, pp. 172-174, May 1994.
[52] M. Amano. S. Fujita, S. Hosoi, T. Noda, A. Sasaki, and Y. Ashizawa, “InAlAs
/InGaAs HEMT’s using InGaP Schottky contact layer,” Proc. 7th Int. Conf. InP and
Related Materials, pp. 416-419, 1995.
[53] S. Fujita, T. Noda, C. Nozaki, and Y. Ashizawa, “InGaAs/InAIAs HEMT with a
strained InGaP Schottky contact layer,” IEEE Electron Device Lett., vol. 14, no. 5, pp.
259-261. May 1993.
[54] A. Kusters, C. Puls, R. Wuller, Behres, A.Kohl, V. Sommer, and K. Heime, “Heighperformace Al-free Ino75Gao2sP/InP/InxGai.xAs/InP(x>53%) backside-doped splitchannel HFET’s with 0.25pm T-gate,” IEE Electronics Lett.., vol. 31, no. 5, pp. 409-410,
Mar. 1995.
192
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[55] J.B. Shealy, M.M. Hashemi, K. Kiziloglu, S.P. DenBaars, U.K. Mishra, T. Liu, J.J.
Brown and M. Lui, “High-breakdown-voltage AlInAs/GalnAs junction-modulated
HEMT’s(JHEMT’s) with regrown ohmic contacts by MOCVD,” IEEE Electron Device
Lett., vol. 14, no. 12, pp. 545-547, Dec. 1993.
[56] A. Fricke, G. Stareer, T. Kummetz, D. Sowarda, J. Mahnss, W. Kowalsky, and K.
Ebelling, “ 1.09eV Schottky barrier height of nearly ideal Pt-Au contacts directly
deposited on n' and p' n-Alo48 lno52 As layers,” Appiled Physics Lett., vol. 65, pp. 755-757,
1994.
[57] M. Matloubian, L.M. Jelloian, A.S. Brown, L.D. Nguyen, L.E. Larson, M.T.
Delaney, M.A. Thompson, R.A. Rhodes, and J.E. Pence, ‘‘V-band high-efficiency highpower AlInAs/GalnAs/InP HEMT’s,” IEEE Trans. MTT pp. , Dec. 1993.
[58] L. Aina, M. Burgess, M. Mattingly, A Meerschaert, J.M. O’Connor, M. Tong, A.
Ketterson, and I. Adesida, “A 1.45-W/mm, 30GHz InP-channel power HEMT,” IEEE
Electron Device Lett., vol. 13, no. 4. pp. 300-302. May 1992.
[59] M. Matloubian, L.M. Jelloian, M. Lui, T. Liu. L.E. Larson, L.D. Nguyen, and M.V.
Le, "GalnAs/InP composite channel HEMT’s,” 5I s' Device research conf., June 1993.
[60] S.R. Bahl, and J.A. del Alamo, "Breakdown voltage enhancement from channel
quantization in InAIAs/nMnGaAs HFET’s," IEEE Electron Device Lett., vol. 13, no. 2,
pp. 123-125, Feb. 1992.
[61] M. Matloubian, T. Liu, L.M. Jelloian, M.A. Thompson, and R.A. Rhodes. "K-band
GalnAs/lnP channel power HEMT’s," IEE Electronics Lett.., vol. 31, no. 5. pp. 409-410,
Mar. 1995.
[62] R.L. Ross, S.P. Svensson, and P. Lugli, Pseudomorphic HEMT Technology and
Applications, Series E, Applied Sciences, vol. 309, Boston: Kluwer Academic, 1994.
[63] A. El-Sabbathy, A.R. Adams, and M.L. Young, “Pressure and composition
dependence of high Field instability in InAsP alloys,” Solid-state Electron., vol. 21, pp.
83-90, 1978.
[64] W-P. Hong, J.R. Hayes, R. Bhat, P.S.D. Lin, C. Nguyen, H.P. Lee. D. Yang, and
P.K. Bhattacharya, “Novel pseudomorphic InP/InAs0 6 Po4 quantum-well HEMT’s.”
IEDM Technical Digest, pp. 243-246, 1991.
193
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[65] T-H. Kim, A.S. Brown, and R.A. Metzger, “Solid source MBE growth of
InAlAs/InAsP heterostructures for HEMT applications,” Electronic Materials Conf.
Digest, Fort Collins, CO, pp. 34, June, 1997.
[6 6 ] W-P. Hong, R. Bhat, J R. Hayes, C. Nguyen, M. Koza, and G.K. Chang, “Highbreakdown, high-gain InAlAs/InGaAsP quantum-well HEMT’s,” IEEE Electron Device
Lett., vol. 12, no. 10, pp. 559-561, Oct. 1991.
[67] W-P. Hong, R. Bhat, F. DeRosa, J.R. Hayes, and G.K. Chang, “OMCVD-grown
Ino4 AI0 6 As/InP quantum-well HEMT," IEEE Electron Device Lett., vol. 12, no. 6 , pp.
284-286, June, 1991.
[6 8 ] S.R. Bahl, B.R. Bennett, and J.A. del Alamo, "Doubly strained Ino4 iAl0 59 As/n+lno.65Gao.35As HFET with high breakdown voltage,” IEEE Electron Device Lett., vol. 14,
no. I, pp. 22-24. Jan. 1993.
[69] Y.C. Pao and J.S. Harris, “Low-conductance drain(LCD) design of
InAlAs/lnGaAs/InP HEMT’s,” IEEE Electron Device Lett., vol. 13, no. 10, pp. 535-537.
Oct. 1992.
[70] J.B. Boss and W. Kruppa, “InAlAs/lnGaAs/InP HEMT’s with high breakdown
voltage using double-recess gate process," IEE Electronics Lett.., vol. 27, no. 21, pp.
1909-1910. Oct. 1991.
[71] K. C. Hwang. P. Ho. M.Y.Kao, S.T. Fu, J. Liu, P.C. Chao, P.M. Smith, and A.W.
Swanson. "W-band high power passivated 0.15pm InAlAs/InGaAs HEMT device," Proc.
6th Int. Conf. InP and Related Materials, pp. 18-20, 1994.
[72] J.B. Shealy. M. Matloubian, T.Y. Liu, M.A. Thompson, M.M. Hashemi, S.P.
Denbaars, and U.K. Mishra, "High-performance submicrometer gatelength GalnAs/lnP
composite channel HEMT’s with regrown ohmic contacts,” IEEE Electron Device Lett.,
vol. 17, no. 11, pp. 540-542, Nov. 1996.
[73] J.B. Kuang, P.J. Tasker, S. Ratanaphanyarat, W.J. Schaff, L.F. Eastman, G.W.
Wang, Y.K. Chen, O.A. Aina, H. Hier, and A. Fathimulla, "Low frequency microwave
characterization
of
submicron-gate
Ino.52 Alo48 As/Ino.s3 Gao47 As/Ino.s2 Alo48 As
heterojunction metal-semiconductor field-effect transistor grown by molecular-beam
epitaxy,”/ Appl. Phys.,vo\. 6 6 , no. 12, pp. 6168, 1989.
194
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[74] S.R. Bahl and J.A. del Alamo, “Elimination of mesa-sidewall gate leakage in
InAlAs/InGaAs heterostructures by selective sidewall recessing,” IEEE Electron Device
Lett., vol. 13, no. 4, pp. 195-197, Apr. 1992.
[75] Y.C. Pao and J.S. Harris, “Physical origin of the high output conductance in
Ino.52Alo4 gAs/Ino53 Gao4 7 As/InP HEMT’s,” Proc. 3rd Int. Conf. InP and Related
Materials, pp. 344-348, 1991.
[76] U. Auer, R. Reuter, C. Heedt, H. Kunzel, W. Prost, and F.J. Tegude,
“InAlAs/InGaAs HFET with extremely high device breakdown using an optimized buffer
layer structure,” Proc. 6,h Int. Conf. InP and Related Materials, pp. 443-446, 1994.
[77] F. Sheffer. C. Heedt, R. Reuter , A. Lindner, Q. Liu, W. Prost, and F.J. Tegude.
“High breakdown voltage InGaAs/InAIAs HFET using InosGaosP spacer layer," IEE
Electronics Lett.., vol. 30, no. 2, pp. 169-170, 1994.
[78] C. Heedt, F. Buchali. W. Prost, W. Brockerhoff. and D. Fritzsche, “Drastic reduction
of gate leakage in InAlAs/InGaAs HEMT's using a pseudomorphic InAlAs/InGaAs hole
barrier layer,” IEEE Trans. Electron Devices, vol. 41. no. 10, pp. 1685-1690, Oct. 1994.
[79] U. Auer, R. Reuter, P. Ellrodt, C. Heedt, W. Prost. and F.J. Tegude, “The impact of
pseudomorphic AlAs spacer layers on the gate leakage current of InAlAs/InGaAs
heterostructures field-effect transistors," Proc. 7ih Int. Conf. InP and Related Materials,
pp. 424-427, 1995.
[80] J. Tardy, X. Letartre, P. Viktrorovitch, M. Gendry,
D. A. Thompson, and J.G.
Simmons, “Spacer design
to improve the breakdown voltage of InAlAs/InGaAs
HEMT’s," Proc. 7 Int. Conf. InP and Related Materials, pp. 420-423, 1995.
[81] J. Zahurak. A.A. Iliadis, S.A. Rishtou, and W. T. Masselink, "Effect of high field
electron transport characteristics on transistor performance in InAlAs/InGaAs
heterostructures," IEEE Proc. Cornell Conf. On Advanced Concepts in High Speed
Semiconductor Devices and Circuits, pp. 270-278, 1993.
[82] C.S. Putnam, M.H. Somerville, J.A. del Alamo, P.C. Chao, and K. G. Duh,
“Temperature dependence of breakdown voltage in InAlAs/InGaAs HEM T’s,” Proc. 9"'
Int. Conf. InP and Related Materials, pp. 197-200, 1997.
[83] M. Matloubian, L.M. Jelloian, M. Lui, T. Liu, and M. Thompson, “Ultra-high
breakdown high-performance AlInAs/GalnAs/InP power HEMT’s,” IEEE IEDM
Technical Digest, pp. 915-917,1993.
195
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[84] M. Matloubian, L.D. Nguyen, A.S. Brown, L.E. Larson, M.A. Meleudes, and M.
Thompson, “High power and high efficiency AlInAs/GalnAs on InP HEMT’s," IEEE
MTT-S Digest, pp. 721-724, 1991.
[85] L.E. Larson, M. Matloubian, J.J. Brown, A.S. Brown, R. Rhodes, D. Crampton, and
M. Thompson, “ AlInAs/GalnAs on InP HEMT’s for low power supply voltage operation
of high power-added efficiency microwave applifiers,” IEE Electronics Lett., vol. 29, no.
15, pp. 15-16, 1993.
[86] J.J. Brown, A.S. Brown, S.E. Rosenbaum, A.S. Schmitz, M. Matloubian, L.E.
Larson, M.A. Melendes, and M.A. Thompson, “Study of the dependence of
Gai47 lno5 ?As/AlxIn|.xAs power HEMT breakdown voltage on Schottky layer designed
device layout, Device Research Conf. Digest, 1993.
[87] M.M. Hashemi, J.B. Shealy, S.P. Denbaars, and U.K. Mishra, “High-speed p+
GalnAs-n InP heterojunction JFET’s(HJFET’s) grown by MOCVD,” IEEE Electron
Device Lett., vol. 14, no. 2. pp. 60-62, Feb. 1993 and Proc. 5lh Int. Conf. InP and Related
Materials, pp. 375-378, 1993.
[88] J.B. Shealy, M. Matloubian, T.Y. Liu, W. Lam, and C. Ngo, “0.9W/mm, 767c
P.A.E.(7GHz) GalnAs/InP composite channel HEM T’s,” Proc. 9? Int. Conf. InP and
Related Materials, pp. 20-22. 1997.
[89] G.A. Johnson. M.D. Biedenbender, and V.J. Kapoor, “InGaAs/InP submicron gate
microwave power transistor for 20GHz applications,” Proc. 3rd Int. Conf. InP and
Related Materials, pp. 423-426, 1991.
[90] M. Matloubian, A.S. Brown, L.D. Nguyen, M.A. Melendes, L.E. Larson, M.J.
Delaney. M. Thompson, R.A. Rhodes, and J.E. Pense, “20GHz high-efficiency AlInAsGalnAs on InP power HEMT,” IEEE Microwave and Guided Wave Lett., vol. 3, no. 5,
pp. 142-144, May 1993.
[91] M.A. Kao, P.M. Smith, P.C. Chao, and P. Ho, “Millimeter wave performance of
InAlAs/lnGaAs/InP HEMT’s," IEEE/Cornell Conf. On Advance Consepts in High Speed
Semiconductor Devices and Circuits, pp. 469-477, 1991.
[92] M. Matloubian, T. Liu. L.M. Jelloian, M. Thompson, and R.A. Rhodes, “K-band
GalnAs/InP channel power HEMT’s,” IEE Electronics Lett., vol. 31, no. 9, pp. 761-762,
1995.
196
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[93] P. Saunier, R. Nguyen, L.J. Messick, and M.A. Khatibzadeh, “An InP MISFET with
a power density of 1.8W/mm at 30GHz,” IEEE Electron Device Lett., vol. 11, no. 1, pp.
48-49, Jan. 1990.
[94] K.Y. Hur, R.A. McTaggart, M.P. Ventresca, R. Wohlert, W.E. Hoke, P.J. Lemonias,
T.E. Kazior, and L.M. Aucoin, “High efficiency single pulse doped
Aloeino-lAs/GalnAs/InP HEMTs for Q-band power applications,” IEE Electronics Lett.,
vol. 31, no. 7, pp. 585-586, 1995.
[95] M. Matloubian, L.M. Jelloian, M. Lui, T. Liu, L.E. Larson, M. Le, D. Jang, and R.A.
Rhodes, “GalnP/InP composite channel HEMTs for millimeter wave power
applications,” Proc. Int. Conf. On Millimeter and Sub-Millimeter Waves and
Applications, pp. 579-580, 1993.
[96] M. Matloubian, A.S. Brown, L.D. Nguyen, M.A. Melendes, L.E. Larson, M.J.
Delaney. J.E. Pense. R.A. Rhodes, M.A. Thompson, and J.A. Henige, “High-power Vband AlInAs/GalnAs on InP HEMT’s," IEEE Electron Device Lett., vol. 14, no. 4, pp.
188-189, Apr. 1993.
[97] S.C. Wang. M.Y. Kao, S.M.J. Liu, P. Ho, and K.G. Duh, “High W-band
pseudomorphic InAlAs/lnGaAs/InP power HEMT’s," Device Research Conf. Digest,
1994.
[98] L. Nguyen. A. Brown. M. Delaney, U. Mishira, L. Larson, L. Jelloian, M. Melendes,
C. Hooper, and M. Thompson, “Vertical Scaling of Ultra-High-Speed AlInAs-GalnAs
HEMTs”, IEEE IEDM, 5.3.1, 1989.
[99] J. Dickmann, “Influence of the Delta Doping Position in the Channel on the Device
Performance of AIGaAs/InGaAs Modulation-Doped Field-Effect Transistors”, Appl.
Phys. Lett., voi 60, pp. 88-90, 1992.
[100] Y. -J. Chan and D. Pavlidis, “Single and Dual p-Doped Channel InAlAs/InGaAs
FETs and the Role of Doping”, IEEE Trans. Electron Devices, vol. ED-39, pp. 466-472,
1992.
[101] J. Dickmann, H. Dambkes, H. Nickel, W. Schlapp, and R. Losch, IEEE Electron
Device Lett., vol. 12, 327,1991.
[102] J. Zahurak, A.A. Iliadis, S.A. Rishtou, and W. T. Masselink, “Effect of high field
electron transport characteristics on transistor performance in InAIAS/InGaAs
197
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
heterostructures,” IEEE Proc. Cornell Conf. On Advanced Concepts in High Speed
Semiconductor Devices and Circuits, pp. 270-278, 1993.
[103] J. Dickmann, “The Backside Pulse Doped Channel Heterostructure Field-Effect
Transistor", Jpn. J. Appl. Phys., vol. 32, pp. 17-25, 1993.
[104] N. Iwata, M. Tomita, and M. Kuzuhara, “High performance double-doped
InAlAs/lnGaAs/InP heterojunction FET with potential for millimeter-wave power
applications”, IEE Electron. Lett., vol. 29, pp. 628-629, 1993.
[105] T. Anan, S. Sugou, K. Nishi, and T. Ichihashi, “Improvement of InP/InGaAs
heterointerfaces grown by gas source molecular beam epitaxy”, Appl. Phvs. Lett. 63, pp.
1047-1049. 1993.
[106] G. Meneghesso, A. Neviani, R. Oesterholt, M. Matloubian, T. Liu, J. J. Brown, C.
Canali. E. Zanoni, “On-state and off-state breakdown in GalnAs/InP composite-channel
HEMTs with variable GalnAs channel thickness”, IEEE Trans. Electron Devices vol. 46,
pp. 2-9. 1999.
[107] S. Loualiche, A. Ginudi, A.L. Corre, D. Lecrosnier, C. Vaudry, L. Henry. C.
Guillemot.
“Low-temperature
DC
characteristics
of
pseudomorphic
Gao i8ln()8;As/InP/Ga« 4 7 lno 53 As HEMT", IEEE Electron Device. Lett. 4, pp.153-155,
1990.
[108] K. Radhakrishnan, T. H. K. Patrick. H. Q. Zheng, P. H. Zhang, and S. F. Yoon.
"Inp/In*Ga|.xAs (0.53 < x < 0.81) high electron mobility transistor structures grown by
solid source molecular beam epitaxy”, J. Cryst. Growth 207, pp. 8-14, 1999.
[109] J. P. R. David. M. Hopkinson, P. N. Stavrinou, and S. K. Haywood, “Growth of
InAsxP|.x/lnP multi-quantum well structures by solid source molecular beam epitaxy", J.
Appl. Phys. 78. pp. 3331-3334, 1995.
[110] W. -P. Hong, R. Bhat, J. Hayes, F. DeRosa, M. Leadbeater, and M. Koza, "Novel
strained InP/InAsxP|.x quantum-well modulation-doped heterostructures”, Appl. Phvs.
Lett. 60, pp. 109-1I I , 1992.
[ I l l ] W. E. Hoke, P. J. Lemonias, D. G. Weir, and H. T. Hendriks, “InAsxP|.
x/InAlAs/InGaAs and InAIAs/InAsojPo? high-clectron mobility transistor structures
grown by solid source molecular beam epitaxy”, J. Vac. Sci. Technol. B 14, pp. 22332235,1996.
198
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[112] C. Qiu, R. V. Kruzelecky, D. A. Thompson, D. Comedi, G. Balcaitis, B. J.
Robinson, and R. W. Streater, “Strain-induced compositional shift in the growth of
InAsyPi.y onto (100) InP by gas-source molecular beam epitaxy”, Can. J. Phys. 70, pp.
886-892, 1992.
[113] T. -H . Kim, A. S. Brown, and R. A. Metzger, “Optical and structural properties of
strained InAlAs/InAsxP|.x multi-quantum wells grown by solid source molecular beam
epitaxy", J. Appl. Phys. 86, pp. 2622-2627, 1999.
[114] W. Q. Chen and S. K. Hark, “Strain-induced effects in (111 )-oriented InAsP/InP,
InGaAs/InP, and InGaAs/InAlAs quantum wells on InP substrates”, J. Appl. Phys. 77, pp.
5747-5750, 1995.
[115] J. I. Pankove, in Optical Processes in Semiconductors (Dover, New York, 1971), P.
27.
[116] H. Morkoq, H. Unlu, and G. Ji, in Principles and Technology o f MODFETs (John
Wiley & Sons, Chichester. 1991), P. 88.
[117] O. Madelung. in Semiconductor-Basic Data (Springer, Berlin, 1996), P. 153.
[118] I. Watanabe. T. Torikai, and K. Taguchi, “Monte Carlo simulation of impact
ionization rates in InAIAs-lnGaAs square and graded barrier superlattice", IEEE J.
Quantum Electron. 31, pp. 1826-1834, 1995.
[119] M. Beaudoin, A. Bensaada. R. Leonelli, P. Desjardins, R. A. Masut, L. Isnard. A.
Chennouf. and G. L’Esperance, “Self-consistent determination of the band offsets in
InAsxP|.x/InP strained-layer quantum wells and the bowing parameters of bulk InAsxP|.
x". Phys. Rev. B 53, pp. 1990-1996, 1996.
[120] H. Q. Hou and C. W. Tu, “In situ control of As composition in InAsP and InGaAsP
grown by gas-source molecular beam epitaxy”, Appl. Phys. Lett. 60, pp. 1872-1874,
1992.
[121] J. E. Cunningham, M. D. Williams, R. N. Pathak, W. Jan, “Non-linear As (P)
incorporation in GaAs|.yPy on GaAs and InAsi.vPy on InP", J. Cryst. Growth 150, pp.
492-496. 1995.
[122] B. X. Yang, L. He, and H. Hasegawa, “Properties of InAsxP|.x layer formed by PAs exchange reaction on (001) InP surface exposed to As4 beam”, J. Electron. Mater. 25,
pp. 379-384, 1996.
199
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[123] S. Adachi, “Material parameters of Ini.xGaxAsyP|.y and related binaries”, J. Appl.
Phys. 53, pp. 8775-8792, 1982.
[124] J. H. Neave and B. A. Joyce, “Dynamics of film growth of GaAs by MBE from
Rheed observations”, Appl. Phys. A 31, pp. 1-8, 1983.
[125] A. Sher, S. Krishnamurthy, A. -B. Chen, “Transport in Submicron Devices”,
Microelectronic Engineering, vol. 9, pp. 377-380, 1989.
[126] A. Kasukawa, N. Yokouchi, N. Yamanaka, N. Iwai and T. matsuda, “InAsP/InGaP
All-Ternary Strain-Compensated Multiple Quantum Wells and Their Application to
Long-Wavelength Lasers”, Jpn. J. Appl. Phys., vol. 34, pp. 965-967, 1995.
[127] S. F. Yoon, H. Q. Zheng, P. H. Zhang, K. W. Mah, and G. I. Ng, “Molecular Beam
Epitaxial Gowth of InP Using a Valved Phosphorus Cracker Cell: Optimization of
Electrical, Optical and Surface Morphology Characteristics”, in IEEE Proc. l( f h Int.
Conf. on InP and Related Materials, pp. 108-111,1998.
[128] X. Letarte, P. Rojo-Romeo, J. Tardy, M. Bejar, M. Gendry, M. A. Py, M. Beck, H.
J. Biihlmann. L. Ren, C. Villar, A. Sanz-Hervas, J. J. Serrano, J. M. Blanco, M. Aguilar,
O. Marty, V. Souliere, and Y. Monteil, "Influence of strain compensation on structural
and electrical properties of InAlAs/InGaAs HEMT structures grown on InP”, in IEEE
Proc. I f f 1' Int. Conf. on InP and Related Materials, pp. 215-218, 1998.
[129] Tong-Ho Kim, Robert A. Metzger, and April S. Brown, “Electrical Properties of
InAIAs/InAsxP|.x /InP Composite Channel Modulation-Doped Structures Grown by Solid
Source Molecular Beam Epitaxy”, J. Electron. Mater. 29, pp. 215-221, 2000.
[130] J. Cowles, R. A. Metzger, A. Gutierrez-Aitken, A. S. Brown, D. Streit, A. Oki, T.
Kim. and A. Doolittle, “Double Heterojunction Bipolar Transistors with InP Epitaxial
Layers Grown by Solid-Source MBE”, IEEE International Conference on Indium
Phosphide and Related Materials, 1996.
[131] Yaochung Chen, April A. Brown, Tonp-Ho Kim. Robert A. Metzger, Richard Lai,
Yeong-Chang Chou, Huan-Chun Yen, and Dwight C. Streit, “Composite-Channel InPInGaAs HEMT Material for Optimized Millimeter Wave Power Performance”, 4Cfh
Electronics Materials Conference, 1998.
[132] Y. C. Chen, P. Chin, D. Ingram, R. Lai, R. Grundbacher, M . Barsky, T. Block, M.
Wojtowicz, L. Tran, V. Medvedev, H. C. Yen, D.C. Streit, and A. Brown, “Composite-
200
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Channe InP HEMT for W-Band Power Amplifiers”, IEEE International Conference on
Indium Phosphide and Related Materials, 1999.
JOURNAL PUBLICATIONS
1. T-H Kim. R. A. Metzger, and A. S. Brown, “Optical and Structural Properties of
Strained InAIAs/InAsP Multi-Quantum Wells Grown by Solid Source Molecular Beam
Epitaxy”, J. Appl. Phys. 86, pp.2622-2627, 1999.
2. Tone-Ho Kim. Robert A. Metzger, and April S. Brown, “Electrical Properties of
InAlAs/InAs*P|.* /InP Composite Channel Modulation-Doped Structures Grown by Solid
Source Molecular Beam Epitaxy”, J. Electron. Mater. 29, pp. 215-221, 2000.
3. Georgiana Dagnall. Jeng-Jung Shen, Tong-Ho Kim. Robert A. Metzger. April S.
Brown, and Stuart R. Stock, “Solid Source MBE Growth of InAsP/InP Quantum Wells",
J. Electron. Mater. 28, pp. 933-938, 1999.
4. Tong-Ho Kim. April S. Brown, Robert A. Metzger, and Y. C. Chen, “Electrical
Properties of InGaAs/InP Composite-Channel Modulation-Doped Structures Grown by
Solid Source Molecular Beam Epitaxy", to be submitted to J. Appl. Phys.
CONFERENCE PAPERS AND PRESENTATIONS
1. J. Cowles, R. A. Metzger, A. Gutierrez-Aitken, A. S. Brown, D. Streit, A. Oki, T.
Kim, and A. Doolittle, “Double Heterojunction Bipolar Transistors with InP Epitaxial
Layers Grown by Solid-Source MBE”, IEEE International Conference on Indium
Phosphide and Related Materials, 1996.
201
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2. T-H Kim. A. S. Brown, and R. A. Metzger, “Solid Source MBE Growth of
InAlAs/InAsP Heterostructures for HEMT Applications". 39'h Electronics Materials
Conference, 1997.
3. Stephen E. Ralph, B. R. Washburn, S. S. Prabhu, G. Dagnall, T. Kim. C. Yi, K. Lee,
E. Roberts, T. Brown, G. May, R. Metzger, and A. Brown, “Ultrafast Carrier Lifetimes
and Electronic Compensation in Low Temperature Grown Be-Doped InGaAs”, 4(fh
Electronics Materials Conference, 1998.
4. Yaochung Chen, April A. Brown, Tong-Ho Kim. Robert A. Metzger, Richard Lai,
Yeong-Chang Chou, Huan-Chun Yen, and Dwight C. Streit, “Composite-Channel InPInGaAs HEMT Material for Optimized Millimeter Wave Power Performance”, 4(fh
Electronics Materials Conference, 1998.
5. Georgiana Dagnall, Tong-Ho Kim. Robert A. Metzger, Stuart R. Stock, and April S.
Brown, "Solid Source MBE Growth of InAsP/InP Quantum Wells", 4(fh Electronics
Materials Conference, 1998.
6. Kyeong K. Lee, William A. Doolittle. Tong-Ho Kim. April A. Brown, Gary S. May,
and Stuart R. Stock, “The Role of the Initial Growth Conditions on GaN Epilayers Grown
by Molecular Beam Epitaxy", North American Molecular Beam Epitaxy’ Conference,
1999.
7. Tong-Ho Kim. April S. Brown, and Robert A. Metzger, “Electrical Properties of
InGaAs/InP Composite-Channel Modulation-Doped Structures Grown by Solid Source
Molecular Beam Epitaxy", 42"' Electronics Materials Conference, 2000.
202
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
VITA
Tong-Ho Kim was bom in Cheju, Korea, on March 1, 1962. He graduated Cheju
Jeil High School on February 1981. He is the First son of Kye-Hwa Kim and Ki-Soon Ko.
He is married to Hyeon-Ja Choi and has two sons, Sung-Hoon and Sang-Hyun.
From 1981 to 1987, he attended Hanyang University, Seoul, Korea, where he
received a Bachelor of Science degree and a Master degree in Electronic engineering. He
worked as a research engineer of Material Research Center in Agency for Defense
Development (ADD), Taejon, Korea, from 1987 to 1993.
He entered graduate school in Georgia Institute of Technology to pursue Ph. D.
degree on January 1995. He joined compound semiconductor group, headed by Dr. April
S. Brown on June 1995. He has received Ph.D. degree in electrical and computer
engineering of Georgia Institute of Technology on July 2000. Currently, he is a
postdoctoral research engineer in Georgia Institute of Technology.
203
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Документ
Категория
Без категории
Просмотров
0
Размер файла
6 201 Кб
Теги
sdewsdweddes
1/--страниц
Пожаловаться на содержимое документа