Development of Algan/Gan High Electron Mobility Transistors (Hemts) On Diamond Substrates

Calhoun: The NPS Institutional Archive
DSpace Repository
Theses and Dissertations 1. Thesis and Dissertation Collection, all items
2006-06
Development of AlGaN/GaN High Electron

Mobility Transistors (HEMTs) on diamond substrates
Newham, Wesley Scott.

Monterey California. Naval Postgraduate School
http://hdl.handle.net/10945/2820
Downloaded from NPS Archive: Calhoun

NAVAL
POSTGRADUATE
SCHOOL
MONTEREY, CALIFORNIA
THESIS
DEVELOPMENT OF AlGaN/GaN HIGH ELECTRON

MOBILITY TRANSISTORS (HEMTS) ON DIAMOND
SUBSTRATES
by
Wesley Scott Newham
June 2006
Thesis Advisor: Todd R. Weatherford

Second Reader: Andrew A. Parker
Approved for public release; distribution is unlimited

THIS PAGE INTENTIONALLY LEFT BLANK
REPORT DOCUMENTATION PAGE Form Approved OMB No. 0704-0188
Public reporting burden for this collection of information is estimated to average 1 hour per response, including
the time for reviewing instruction, searching existing data sources, gathering and maintaining the data needed, and
completing and reviewing the collection of information. Send comments regarding this burden estimate or any
other aspect of this collection of information, including suggestions for reducing this burden, to Washington
headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite
1204, Arlington, VA 22202-4302, and to the Office of Management and Budget, Paperwork Reduction Project
(0704-0188) Washington DC 20503.
1. AGENCY USE ONLY (Leave blank) 2. REPORT DATE 3. REPORT TYPE AND DATES COVERED
June 2006 Master’s Thesis
4. TITLE AND SUBTITLE: 5. FUNDING NUMBERS
Development of AlGaN/GaN High Electron Mobility Transistors (HEMTs) on
Diamond Substrates
6. AUTHOR(S) Wesley Scott Newham
7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING
Naval Postgraduate School ORGANIZATION REPORT
Monterey, CA 93943-5000 NUMBER
9. SPONSORING /MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSORING/MONITORING
N/A AGENCY REPORT NUMBER
11. SUPPLEMENTARY NOTES The views expressed in this thesis are those of the author and do not reflect the official
policy or position of the Department of Defense or the U.S. Government.
12a. DISTRIBUTION / AVAILABILITY STATEMENT 12b. DISTRIBUTION CODE
13. ABSTRACT (maximum 200 words)
Silicon based semiconductor devices are rapidly approaching the theoretical limit of operation and are
becoming unsuitable for future military requirements. The scope of semiconductor devices has been expanded by
wide bandgap devices such as gallium nitride (GaN) to include the possibility for high power and high frequency
operation. A new generation of high speed – high frequency devices is required to meet current and future
military needs. The Gallium Nitride High Electron Mobility Transistor (HEMT) is showing great promise as the
enabling technology in the development of military radar systems, electronic surveillance systems,
communications systems and high voltage power systems. Typically, sapphire or silicon carbide is utilized as the
substrate material in most HEMT designs. This thesis explores the possibility of utilizing a diamond substrate to
increase the power handling capability of the AlGaN/GaN HEMT. Diamond offers increased thermal property
parameters that can be simulated in the commercially available Silvaco software package. A complete electrical
and thermal analysis of the model was conducted and compared to actual device characteristics. The results of the
software simulation and measurements on the test devices indicate diamond substrates will enable the HEMT to be
operated at a higher power than traditional sapphire substrate HEMTS.
14. SUBJECT TERMS 15. NUMBER OF

High Electron Mobility Transistor, HEMT, Gallium Nitride, Heterostructure, Sapphire Substrate, PAGES
Diamond Substrate, Silvaco, ATLAS, BLAZE, GIGA 75
16. PRICE CODE
17. SECURITY 18. SECURITY 19. SECURITY 20. LIMITATION
CLASSIFICATION OF CLASSIFICATION OF THIS CLASSIFICATION OF OF ABSTRACT
REPORT PAGE ABSTRACT
Unclassified Unclassified Unclassified UL
NSN 7540-01-280-5500 Standard Form 298 (Rev. 2-89)
Prescribed by ANSI Std. 239-18
i
ii
DEVELOPMENT OF AlGaN/GaN HIGH ELECTRON MOBILITY

TRANSISTORS ON DIAMOND SUBSTRATES
Wesley S. Newham
Lieutenant, United States Navy
BSEE, Auburn University, 2000
Submitted in partial fulfillment of the

requirements for the degree of
MASTER OF SCIENCE IN ELECTRICAL ENGINEERING
from the
NAVAL POSTGRADUATE SCHOOL

June 2006
Author: Wesley Scott Newham
Approved by: Todd R. Weatherford

Thesis Advisor
Andrew A. Parker
Second Reader
Jeffrey B. Knorr
Chairman, Department of Electrical and Computer Engineering
iii
iv
ABSTRACT
Silicon based semiconductor devices are rapidly approaching the theoretical limit
of operation and are becoming unsuitable for future military requirements. The scope of
semiconductor devices has been expanded by wide bandgap devices such as gallium
nitride (GaN) to include the possibility for high power and high frequency operation. A
new generation of high speed – high frequency devices is required to meet current and
future military needs. The Gallium Nitride High Electron Mobility Transistor (HEMT) is
showing great promise as the enabling technology in the development of military radar
systems, electronic surveillance systems, communications systems and high voltage
power systems. Typically, sapphire or silicon carbide is utilized as the substrate material
in most HEMT designs. This thesis explores the possibility of utilizing a diamond
substrate to increase the power handling capability of the AlGaN/GaN HEMT. Diamond
offers increased thermal property parameters that can be simulated in the commercially
available Silvaco software package. A complete electrical and thermal analysis of the
model was conducted and compared to actual device characteristics. The results of the
software simulation and measurements on the test devices indicate diamond substrates
will enable the HEMT to be operated at a higher power than traditional sapphire substrate
HEMTS.
v
vi
TABLE OF CONTENTS
I. INTRODUCTION........................................................................................................1
A. BACKGROUND ..............................................................................................1
B. ALGAN/GAN HEMT RESEARCH...............................................................3
1. Defense Advanced Research Projects Agency
(DARPA)/Microsystems Technology Office (MTO).........................3
2. DARPA Wide Bandgap Semiconductor Technology Initiative
(WBGSTI).............................................................................................4
C. SUMMARY OF PREVIOUS RESEARCH...................................................5
II. HIGH ELECTRON MOBILITY TRANSISTORS (HEMT) ..................................7
A. HEMT FUNDAMENTALS.............................................................................7
1. HEMT Structure..................................................................................7
2. AlGaN/GaN HEMT Fundamentals....................................................9
a. HEMT Equations....................................................................10
b. HEMT Material Composition Issues .....................................12
c. HEMT Operational Parameters .............................................13
3. Two Dimensional Electron Gas (2DEG) ..........................................15
a. Spontaneous Polarization .......................................................15
b. Piezoelectric Polarization .......................................................17
B. BERKELEY HEMT DESIGN......................................................................18
III. AlGaN/GaN HEMT SILVACO MODEL .............................................................21
A. SILVACO SEMICONDUCTOR MODELING EQUATIONS .................23
1. Poisson’s Equation .............................................................................24
2. Carrier Continuity Equations...........................................................24
3. Transport Equations..........................................................................25
4. Polarization Effects............................................................................26
B. SILVACO MODEL .......................................................................................28
1. Model Definition.................................................................................28
2. Gate Characterization .......................................................................29
3. Heterojunction Bandgap Results......................................................30
4. Electron Concentration due to Polarization....................................32
5. I-V Curves...........................................................................................33
C. THERMAL CHARACTERIZATION.........................................................34
1. Physical Device Thermal Testing .....................................................35
2. NPS HEMT Model Thermal Testing ...............................................37
D. FINAL THERMAL SIMULATION ............................................................41
1. Sapphire vs. Diamond Substrate Comparison ................................41
E. CHAPTER CONCLUSION..........................................................................45
IV. CONCLUSIONS AND RECOMMENDATIONS...................................................47
A. CONCLUSIONS ............................................................................................47
B. RECOMMENDATIONS..............................................................................48
vii
LIST OF REFERENCES ......................................................................................................51
APPENDIX: ATLAS™ INPUT DECK ..............................................................................53
INITIAL DISTRIBUTION LIST .........................................................................................57
viii
LIST OF FIGURES
Figure 1: Tri-Service Team Interaction [4]. ......................................................................5

Figure 2: AlGaN/GaN HEMT Layer Structure. ................................................................7
Figure 3: HEMT Device Dimensions [10]. .......................................................................8
Figure 4: SEM Image of Berkeley GAN HEMT device. [9]. ...........................................9
Figure 5: DC Output characteristic of AlGaN/GaN HEMT............................................10
Figure 6: 2DEG electron density distribution for different Al(x) composition [14] ......13
Figure 7: Bandgap diagram of AlGaN/GaN HEMT [9]..................................................14
Figure 8: Schematic drawing of the crystal structure of wurtzite Ga-face[15]. ..............16
Figure 9: Ga-Faced polarization induced sheet charge density and directions for the
spontaneous and piezoelectric polarization effects [15]. .................................17
Figure 10: Silvaco Simulation Flowchart [16]. .................................................................23
Figure 11: Sheet charge as a function of alloy composition [15]......................................27
Figure 12: 2D Representation of the Modeled AlGaN/GaN HEMT.................................29
Figure 13: (a) ATLAS™ Generated Gate Characterization and (b) Measured. .................29
Figure 14: Band diagram of HEMT showing (a) Conduction Band and (b) Valence
Band. ................................................................................................................31
Figure 15: Band Diagram of modeled AlGaN/GaN HEMT. ............................................31
Figure 16: Electron Concentration Contour View (a) and Cutline View (b). ...................32
Figure 17: NPS HEMT Model I-V Characteristic.............................................................33
Figure 18: Measured Berkeley HEMT Device I-V Characteristic....................................34
Figure 19: Topmost Curve (Vgs= -1 V) I-V Measurement of Physical Device over a
range of temperatures.......................................................................................35
Figure 20: I-V Characteristic Room Temperature vs. 100º C. ..........................................36
Figure 23: NPS HEMT Room Temperature vs. 100º C. ...................................................38
Figure 26: Decreasing Current trend of Actual Device and Modeled Device. .................40
Figure 27: NPS HEMT Gate Leakage Over Temperature. ...............................................40
Figure 28: Contour Plot of (a) Sapphire Substrate Material and (b) Diamond
Substrate...........................................................................................................41
Figure 29: Thermal Horizontal Cutline of 2DEG with Diamond Thermal
Conductivity set to 10 and 20. Sapphire (brown) Diamond 10 (green) and
Diamond 20 (red).............................................................................................43
Figure 30: Drain Current as a Function of Substrate Material. .........................................43
Figure 31: Gate Leakage as a Function of Substrate Material. .........................................44
ix
x
LIST OF TABLES
Table 1: Electronic Properties of Semiconductor Materials [2].......................................2

Table 2: Spontaneous polarization, piezoelectric and dielectric constants of AlN,
GaN [15]. .........................................................................................................16
Table 3 Al Composition and layer thickness of AlxGa1-xN barrier layer in AlxGa1-
xN/GaN HEMT structure [9]............................................................................19
Table 4: Percent Reduction in Current with Temperature. ...........................................39
Table 5: Summary of Testing Results............................................................................45
xi
xii
ACKNOWLEDGMENTS
I can never express my gratitude enough to the people who have helped me
complete this thesis. First I would like to thank Professor Todd Weatherford who
provided me with necessary tools to complete this thesis. Your ability to make complex
theories understandable was remarkable and made the design process easier. Thank you
for keeping me focused and pulling me out of the weeds when I needed it. I would also
like to thank Andrew Parker who was the second reader of this thesis. Your suggestions
were thought provoking and added to the quality of this thesis.
I would like to thank Dr. Petra Specht from UC Berkeley for providing the GaN
HEMT devices for testing. Your willingness to provide suggestions and guidance has
enhanced the quality of this thesis.
I would like to thank my three children Josh, Sydney, and Chase. You guys have
been a blessing to me and helped me in ways you will never know. You three always
make me realize what the true meaning of life is. Thank you for being you.
To my wife, Rachel, thank you for always being the calm in the storm I have put
us through during this process. You were always there and willing to assist every step of
the way. I count my blessings every day and thank God that he has provided such a
wonderful person for me to spend the rest of my life with. You inspire me everyday!
Thank you!
xiii
xiv
EXECUTIVE SUMMARY
Current AlGaN/GaN High Electron Mobility Transistors (HEMTs) grown on

sapphire substrates have a low reliability due to thermal effects. The need for a reliable
high-frequency, high-power HEMT device is necessary for future naval communication
and radar needs. A computer model of an AlGaN/GaN HEMT has been developed for
the purpose of studying the thermal effects due to self heating. The thermal effects were
examined on a device grown on a sapphire substrate and then on a diamond substrate to
determine if the high thermal conductivity parameter of diamond would provide a
reduced device operating temperature. The model developed in this thesis was created
with the commercially available Silvaco software package and includes the BLAZÉ™
heterojunction routine, the GIGA™ thermal modeling routine, as well as various models
to simulate the low field mobility as a function of doping and lattice temperature.
The ATLAS™ model was created based on the parameters extracted from an
actual AlGaN/GaN HEMT grown on a sapphire substrate. The parameters extracted
from the actual device were utilized as baseline for the model development.
One of the critical parameters necessary to simulate heterojunction devices is
polarization. The polarization effects present in heterojunction devices are spontaneous
polarization and the piezoelectric effect and are crucial in the formation of the unique two
dimensional electron gas (2DEG) channel present in heterojunction devices. Although
ATLAS™ provides models to simulate the effects individually, the results were less than
desirable. In the NPS model, an interface charge was placed on the AlGaN side of the
heterojunction to simulate the polarization effects. This interface charge provided the
necessary electron concentration to form the 2DEG which forms an abrupt bangap shift at
the interface. This abrupt change forms a quantum well which induces the 2DEG to
form. All parameters extracted from the physical model were able to be re-created in the
software model to a high degree of accuracy.
Simulations were conducted utilizing a sapphire substrate to match the operational
parameters of the physical device. Once an accurate model was developed, the substrate
material was changed to diamond, and the thermal characteristics were studied. A
xv
simulation of the sapphire substrate device and the diamond substrate device was
conducted to determine if a change in the thermal parameters affected the device heating.
The sapphire substrate HEMT model was validated with measurements over a
temperature range of 270C (room temperature) to 3000C. The validated device modeling
then showed a dramatic change in the lattice temperature as well as the channel
temperature of the device when the substrate was changed to diamond. The diamond
substrate reduced the peak temperature of the device by 1/2 while the average channel
temperature was reduced by 1/3.
The work presented in this thesis represents the most accurate NPS HEMT model
developed to date and can be utilized in follow-on efforts to study reliability and high
frequency operations. AlGaN/GaN HEMT is showing great promise in becoming the
enabling technology in high-power, high-frequency semiconductor amplifiers, and the
results of this thesis indicates that utilizing a diamond substrate will enhance the
operation of the HEMT.
xvi
I. INTRODUCTION
The purpose of this thesis was to create an accurate electrical and thermal
computer model of an AlGaN/GaN HEMT device in order to investigate device operating
parameters as the substrate thickness and material are changed. To develop an accurate
model, an actual AlGaN/GaN HEMT device was examined and parameters extracted
utilizing Silvaco’s UTMOST™ program. The biggest departure of this thesis compared
to previous work in this area is that this work was based and verified with physical
devices and correlated with simulated models. The data extracted from the physical
device served as a benchmark in the development of the device model. The AlGaN/GaN
HEMT model was created utilizing Silvaco’s ATLAS™, BLAZE™, and GIGA™
routines.
This chapter will discuss the background that motivated this thesis, current uses of
HEMT devices, the problems associated with HEMT devices, ongoing research and
funding associated with HEMT research, and will conclude with a summary of previous
results and problems encountered. Chapter two will describe HEMT device physics and
operation. Chapter three will outline the procedures for testing the actual HEMT device,
the Silvaco model design as well as the physics models utilized to match the model with
the actual device, and compare the physical device characteristics with the model results.
Finally, chapter four will conclude the thesis with conclusions and recommendations and
discuss possible future research.
A. BACKGROUND
To achieve the vision of Sea Power 21, the United States Navy requires new
technology in the field of high-power, high-frequency electronics to field new state of the
art systems. Current silicon technology is not capable of achieving the performance
required by future systems. The technology that is poised to fulfill the future needs of
industry and the Navy is wide bandgap (WBG) semiconductor devices. In particular, the
most promising device to satisfy many of the future needs of the Navy is the GaN High
Electron Mobility Transistor (HEMT). Some desirable attributes of next generation
WBG electronics include the ability to withstand currents in excess of 5 kA and voltages
1
in excess of 50 kV, provide rapid switching, maintain good thermal stability while
operating at temperatures above 250° C, have small size and light weight, and be able to
function without bulky heat-dissipating systems [1]. Table 1 provides a comparison of
the various semiconductor technologies available today.
Table 1: Electronic Properties of Semiconductor Materials [2].
In addition to becoming the enabling technology for future systems, GaN HEMTs
have the potential to replace vacuum tubes in high power millimeter wave radar and
electronic warfare systems. They have the potential to deliver high power at high voltage
and do it at a fraction of the cost of current vacuum tube technology. The wide bandgap
of GaN increases the breakdown field by five times and the power density by a factor of
10 to 20 compared with GaAs based devices. This results in smaller components with
less capacitance for the same operating power, which results in a wider range of
bandwidth.
GaN HEMT devices appear to be a great technology on paper, and great progress
is being made toward technology development and process improvement. There have
been demonstrations of very high output power densities in GaN HEMTS, but significant
developmental work remains. One of the remaining issues is reliability. The output
power from these devices can permanently degrade over a relatively short period of time.
A phenomenon known as current collapse, a reduction in the dc drain current, after the
application of a high drain–source voltage is prevalent in many GaN HEMT devices.
2
The high voltage leads to the injection of hot carriers into regions of the device adjacent
to the conduction channel that contains deep traps [3]. One reliability issue this thesis
attempts to address is how increased thermal stress affects the operation of the GaN
HEMT. It has been shown that the DC characteristics are better for HEMTs fabricated on
SiC substrate than on a sapphire substrate, possibly because of the increased thermal
conductivity of SiC. The model developed in this thesis will examine the DC
characteristics of the HEMT utilizing a diamond substrate. Diamond’s thermal
conductivity is about twice that of SiC and, therefore, will reduce the channel
temperature of the HEMT, which will increase the power density potential and reduce
thermally induced reliability issues.
B. ALGAN/GAN HEMT RESEARCH

1. Defense Advanced Research Projects Agency (DARPA)/Microsystems
Technology Office (MTO)
DARPA is currently funding many research projects dealing with WBG
development and implementation. One DARPA program currently underway is the Wide
Bandgap Semiconductor-Radio Frequency (WBGS-RF) program. The program
objectives are:
• Develop reliable, high performance devices and MMICs with
o Higher power
o Superior thermal performance
o Higher efficiency
• Rapid insertion into DoD RF systems.
The WBGS-RF program is organized into three phases. Phase I is complete and was
focused on WBG semiconductor materials. The output of Phase I results in high quality
Si and SiC substrates and highly uniform GaN epitaxy, utilizing both MBE and MOCVD.
Phase II is scheduled for three years and is currently underway. Phase II is concerned
with device technologies such as X and Q band discrete components. The output of
Phase II will be device demonstrations. Phase III is a two-year phase concerned with
electronic integration and circuit technologies.
3
The DARPA WBGS-RF program has been awarding contracts in the fields of
wide bandgap materials optimization, wide bandgap device fabrication and modeling,
high yield wide bandgap MMICs, and packaging and thermal management of wide
bandgap devices. There were multiple contract awards starting in calendar year 2005
totaling $75-$110 million dollars.
2. DARPA Wide Bandgap Semiconductor Technology Initiative
(WBGSTI)
Another project sponsored by DARPA that is dedicated to advancing WBG
technology is the WBGSTI. The objectives of the WBGSTI program are to scale up high
quality SiC substrates, develop alternative substrates, develop uniform AlGaN/GaN
HEMTs, and to examine materials/device correlations. This project is unique because it
has developed a Tri-Service team comprised of the Air Force Research Laboratory
(AFRL), Army Research Laboratory (ARL), and the Naval Research Laboratory (NRL).
The Tri-Service team is composed of a Materials Panel and a Device Panel. Figure 1
depicts the interaction of the Tri-Service team.
4
Figure 1: Tri-Service Team Interaction [4].
The WBGSTI is a multi-phase, multi-year project that is poised to revolutionize

the development of WBG devices. This project has the potential to change the way in
which high-power, high-frequency systems are designed.
C. SUMMARY OF PREVIOUS RESEARCH

This section summarizes previous Naval Postgraduate School modeling attempts
relating to AlGaN/GaN HEMT modeling research. The previous efforts by Eimers [5],
Holmes [6], and Salm [7] concentrated on correctly modeling the HEMT device.
5
Previous versions of the Silvaco software did not provide adequate models for the surface
charge or piezoelectric effects contained in AlGaN/GaN HEMTS. Therefore, many of
the earlier attempts to correctly model the HEMT device were focused on independently
developing these models to include in the Silvaco routine.
In [6], Holmes attempted to incorporate the polarization, piezoelectric and

spontaneous effects into the HEMT model utilizing a C-interpreter method. Holmes was
able to get an adequate I-V characteristic of a HEMT device utilizing a C-interpreter
routine. However, his model needed to be refined to more closely match the overall
operation of the HEMT device.
Since the completion of Holmes’ thesis in 2002, Jon C. Freeman [8] developed a
method to model GaN HEMTs with the Silvaco software package. In particular, he
utilized ATLAS™ to add the effects of strain, piezoelectricity, and spontaneous
polarization at the interfaces. The objectives of Holmes’ thesis seem to have been met by
Silvaco’s software upgrade and Freeman’s research. The model developed in this thesis
incorporated the new Silvaco models and resulted in a more accurate device model.
Salm’s [7] thesis, was the first Naval Postgraduate School attempt to utilize the
piezoelectric model developed by Silvaco in a HEMT device simulation. Although he
was not able to get accurate results with the piezoelectric model, he modeled the effect
with an interface charge, and his results were adequate and more accurately modeled the
desired characteristics of a HEMT device. His thesis also studied the thermal
characteristics of the device utilizing the GIGA™ module. The GIGA™ module is
capable of giving a numerical, as well as illustrative, representation of device heating.
Salm’s results indicated a remarkable difference in device heating due to a change in
substrate material. The specifics of the structure and device physics will be discussed in
chapter three of this thesis.
6
II. HIGH ELECTRON MOBILITY TRANSISTORS (HEMT)
This chapter presents an introduction to the High Electron Mobility Transistor

(HEMT). The theory of operation for a HEMT device will be discussed to include
equations, bandgap calculations, and the unique feature of the HEMT, the two
dimensional electron gas (2-DEG). This chapter will conclude with an overview of the
AlGaN/GaN structure designed and fabricated by the University of California at
Berkeley.
A. HEMT FUNDAMENTALS
1. HEMT Structure
The GaN HEMT is similar to a typical MESFET in that it is a three terminal
device that utilizes the gate to control current flow. The Naval Postgraduate School was
provided with an array of devices designed and tested by Susie Tzeng [9] from the
University of California at Berkeley. The exact structure dimensions shown in Figure 3
selected for this thesis includes a gate length (Lg) of 2µm, a gate width (Wg) of 100 µm,
and a source to gate (Lgs) and gate to drain (Lgd) distance of 2 µm. Additionally, the Al
(x) composition of the Berkeley HEMT was varied between 28% and 35%. The Al
composition selected for this thesis was 28%. Figure 2 illustrates the layer structure of
the device and Figure 3 defines the regions for the dimensions described above.
25 nm AlxGa1-xN
1.5µm GaN
Buffer Layer
Sapphire
Figure 2: AlGaN/GaN HEMT Layer Structure.
7
Figure 3: HEMT Device Dimensions [10].
Ti/Al/Ni/Au ohmic metallization was deposited by e-beam evaporation and

patterned by lift-off [9]. This process resulted in a contact resistance of
approximately 5 × 10−5 Ω-cm 2 . The physical characteristics of the Berkeley device were
utilized as an input to the ATLAS™ model and are discussed in chapter three of this
thesis. Figure 4 is a scanning electron microscope (SEM) image of the HEMT device
structure where the gate contact can be seen positioned between the source and drain
ohmic contacts [9].
Although a complete strain analysis of the device is beyond the scope of this
thesis, an overview of the strains and polarization effects in a HEMT device will be
discussed in section two of this chapter. The polarization effects inherent in HEMT
devices have been modeled by Dr. Jon Freeman [8] and can be incorporated utilizing the
Silvaco software package. The modeling efforts of this thesis will be detailed in chapter
three of this thesis to include the polarization effects that occur in HEMT devices.
8
Figure 4: SEM Image of Berkeley GAN HEMT device. [9].
2. AlGaN/GaN HEMT Fundamentals

The AlGaN/GaN HEMT also known as the modulation doped field effect
transistor (MODFET) has similar DC characteristics to the MESFET. The unique feature
of the HEMT is channel formation from carriers accumulated along a grossly asymmetric
heterojunction, i.e. a junction between a heavily doped high bandgap and a lightly doped
low bandgap region [1]. Instead of the channel being formed as in a standard
homojunction device, a two dimensional electron gas (2DEG) is formed at the
heterointerface. To achieve proper operation of the device, the barrier layer AlxGa1-xN
must be at a higher energy level than the conduction band of the GaN channel layer. This
conduction band offset transfers electrons from the barrier layer to the channel layer. The
electrons that are transferred are confined to a small region in the channel layer near the
heterointerface. This layer is called the 2DEG and a defining characteristic of the
HEMT. There are many factors that determine the quality of the 2DEG. The factors
involved in the development of the 2DEG are type of substrate, growing method, and
level of doping of the carrier supply layer [10]. Electron density and mobility are
important factors when dealing with a 2DEG and a more detailed discussion of the 2DEG
will be presented later in this chapter.
9
a. HEMT Equations
This section will outline basic HEMT parameters. The primary operating
parameters dealing with high power HEMT operations are the transconductance (gm), the
2DEG sheet charge concentration (ns), the drain current (ID), the threshold voltage (VT),
and the contact barrier height ( φb ).
The typical DC measurement to determine the operation of a HEMT is the

drain current versus the voltage drain to source (ID vs. VDS) while the gate is swept from
below cutoff to saturation. Figure 5 is an ID vs. VDS curve extracted from one of the
Berkeley devices.
Figure 5: DC Output characteristic of AlGaN/GaN HEMT.
10
Unlike traditional MOS devices, the charge flowing in a HEMT device is
normally confined to the 2DEG and changed by the gate voltage. When a gate voltage is
applied to the device, the sheet charge concentration (ns) is changed proportionally to
gate voltage and is given by equation 2.1 [11]:
εi
ns =
q(di + ∆d )
(Vg − VT ) (2.1)
where ε i and di are the dielectric permeability and the thickness of the wide bandgap
semiconductor, respectively, and ∆d can be interpreted as the effective thickness of the

2DEG [11]. A typical ns value for an AlGaN/GaN heterojunction is ~ 1013/cm2.
The threshold voltage is an important parameter because it is a measure of

when the HEMT device will begin conducting. VT can be expressed as:
qN D 2
VT = φB − ∆EC − d (2.2)
2ε i
where φB is the gate Schottky-barrier height, and ∆EC is the change in the conduction
band at the heterojunction, and ND is the background doping of the GaN layer.
Saturation current and transconductance are important measures of a

HEMT device when utilized as a power amplifier. For long channel devices (Lg>
0.25µm), such as the ones presented in this thesis, IDSat can be expressed the same as a
standard MOSFET [9]:
µ CW
I DSat = (Vg − Vt ) 2 (2.3)
2L
where µ is the carrier mobility, C is the gate capacitance, W is the gate width, and L is
the gate length.
One of the measures of quality for a HEMT device is transconductance. It

is a measure of how the drain current will change with a change in the gate voltage.
Mathematically, transconductance is defined as [12]:
11
∂I D µCW
gM = = (Vg − Vt ) . (2.4)
∂VG VD=Constant
L
Additionally, the output conductance known as the drain or channel

conductance is an important measure in the HEMT device and is defined by [12]:
∂I D
gd = . (2.5)
∂Vd Vg=Constant
The output conductance plays a crucial role in determining the output impedance
matching properties of different stages of an amplifier.
The equations presented above are a good starting point for modeling the
Berkeley HEMT device. However, there are many variables within the equations that
will be modified to achieve a proper model of the device. There are a variety of models
included in the Silvaco package that can be incorporated into the device definition, and
when necessary a more detailed description of the chosen model will be provided.
b. HEMT Material Composition Issues
As mentioned above, the HEMT is a three terminal device where

conduction is controlled by the gate voltage. The biggest operational difference between
a HEMT and a traditional MOS device is that instead of creating a channel between the
source and drain with a doped material, a wider bandgap material barrier layer
(AlxGa1-xN) is grown on top of the channel and serves as a carrier supply layer [9].
Therefore, the channel is formed as a 2DEG below the doped layer which increases
mobility due to a reduction in impurity scattering. A technique utilized to increase the
electron density in the channel is to increase the Al mole fraction (x) in the AlxGa1-xN
layer. An increase in the Al mole fraction results in a higher bandgap and a larger
conduction band discontinuity ∆EC relative to the GaN [13]. The higher bandgap and
∆EC results in an increase in the composite breakdown field and improves carrier
confinement. Additionally, the piezoelectric and spontaneous polarization charges
increase and serve to confine the carriers to the 2DEG. The effects of increasing the Al
12
mole concentration are attractive, but there is a limit to benefit gained. Figure 6
illustrates the electron density as a function of Al composition.
Figure 6: 2DEG electron density distribution for different Al(x) composition

[14] .
Increasing the Al mole fraction also results in a reduced carrier mobility due to different
effects including, intersubband scattering, increased alloy disordering in AlGaN,
increased density of interface charges and larger potential fluctuations at the interface due
to the surface roughness [14]. The Al composition for the device tested in this thesis is
28%, which roughly correlates to an electron density of 2e19 cm-3.
c. HEMT Operational Parameters
This section will provide an overview of the device structure, where the
2DEG is created, and the energy bandgap diagrams of an AlGaN/GaN HEMT device.
The HEMT operational characteristics are based on the MESFET. The

2DEG is formed just below the heterojunction where the defects are lower and a higher
electron mobility is possible. Figure 3 gives a visual estimate of where the 2DEG will be
within the structure. The conductivity of the two dimensional channel is given by [10]:
σ = qns µ (2.6)
where q is the electron charge, ns is the sheet carrier concentration given by equation 2.1,
and µ is electron mobility.
13
Figure 7 depicts the typical bandgap diagram of an AlGaN/GaN HEMT
device.
EF
Figure 7: Bandgap diagram of AlGaN/GaN HEMT [9].
One of the defining characteristic of a heterojunction device is the abrupt bandgap shift
where the AlGaN layer meets the GaN Layer. The polarization effects that occur at this
interface allows the 2DEG to be formed and become a conducting channel when properly
biased.
The Id vs. Vds curve generated in Figure 5 will serve as the basis for
discussion of the HEMT operating conditions. The initial values for the measurement
were:
• Source voltage (Vs) =0
• Drain voltage (Vd) = 0
• Gate voltage(Vg)=-5
The source voltage was held constant and the drain voltage was swept from 0-7 V. The
gate voltage began at -5 V and was incremented by 1 V per sweep to a final value of -1V.
With the Vs and Vg = 0V, Vd was incremented in steps up to 7 V. It is apparent from
14
Figure 5 that a small current began to flow in the 2DEG channel with no gate voltage.
By applying a positive voltage to the drain, current will flow due to the potential
difference between the source and drain, but the magnitude of the current flow is
controlled by the gate. Each branch in the diagram represents a 1 V increase in Vg which
causes the drain-source current to increase linearly until the saturation point of the device
is reached. The value of IDsat is dependant upon the ns concentration. By increasing the
concentration of ns , the 2DEG is capable of transferring more charge from source to

drain.
Unlike a typical MOS Id vs. Vds curve, GaN HEMTs on a sapphire

substrate show a small negative slope in the saturation region. This anomaly is known as
negative resistance and is attributed to self-heating of the device due to the poor thermal
conductivity of sapphire[1].
3. Two Dimensional Electron Gas (2DEG)
The 2DEG in AlxGa1-xN/GaN based HEMT is induced by strong polarization
effects [15]. Ambacher et al., utilizing piezoelectric constants obtained from previous
research, calculated the induced sheet charge bound at the heterojunction interface. His
conclusion was that the formation of the 2DEG in AlGaN/GaN structures was the result
of both spontaneous polarization (PSP) and the strain induced piezoelectric effects (PPE)
[15]. The Silvaco software package includes new modeling parameters that will enable a
more realistic HEMT device model to be developed. The following sub-sections address
the polarization effects of a GaN HEMT and will provide some of the variables that
affect the 2DEG.
a. Spontaneous Polarization
The two distinct polarization mechanisms found in GaN HEMTs enable a
high carrier concentration even though there may be no intentional doping. Spontaneous
polarization is the effect that is present at the heterojunction even though there is no
strain present. The spontaneous polarization (PSP) is developed due to the wurtzite group
III-nitrides being tetrahedrally coordinated with a lack of symmetry along the [0001] axis
or c-direction [10]. The polarization effects vary according to the how the structure was
formed. The structure will have either N-face or Ga-face surfaces. The Berkeley device
15
provided was grown on sapphire by MOCVD which results in Ga-faced structures only.
Figure 8 shows the crystal structure of a wurtzite Ga-face.
Figure 8: Schematic drawing of the crystal structure of wurtzite Ga-face[15].
In the case of the wurtzite crystal grown in the manner previously

discussed, the PSP will be given as a constant determined by the method of growth. Table
3 provides the spontaneous polarization, piezoelectric and dielectric constants of AlN and
GaN. Ambacher et al. [15] developed this table from many sources and utilized the
constants to determine the PSP and PPE for an AlGaN/GaN structure.
Table 2: Spontaneous polarization, piezoelectric and dielectric constants of

AlN, GaN [15].
The ability to achieve a high carrier concentration at the 2DEG is determined by the
ability of the structure to confine carriers through the polarization effects. To achieve a
high carrier confinement, the polarization directions should be the same. The
polarization induced sheet charge density ( σ ) should be positive. This positive sheet
charge attracts electrons and forms the 2DEG. Figure 9 illustrates the polarization charge
16
effects for the different types of strain. The first case (a) where AlGaN (relaxed) is
grown on top of GaN (relaxed), a positive sheet charge ( +σ ) is developed at the interface
and a 2DEG forms. To enhance the polarization effects, example (b) grew tensile
strained AlGaN over GaN which resulted in all polarization effects acting in the same
direction. This also results in a positive sheet charge at the interface. The piezoelectric
(PPE) effects will be illustrated and calculated in the next section.
Figure 9: Ga-Faced polarization induced sheet charge density and directions for
the spontaneous and piezoelectric polarization effects [15].
b. Piezoelectric Polarization
The total polarization in the structure is given by PSP + PPE=PTOTAL.
Utilizing Table 2, the piezoelectric polarization can be calculated by the following
equation given by [15]:
PPE = e33ε z + e31 ( ε x + ε y ) (2.7)
17
⎛ c − c0 ⎞
where a0 and c0 are the equilibrium values of the lattice parameters, ε z = ⎜ ⎟ is the
⎝ c0 ⎠
⎛ a − a0 ⎞
strain along the c-axis, and the in-plane strain ε x = ε y = ⎜ ⎟ is assumed to be
⎝ a0 ⎠
isotropic. The third independent component of the piezoelectric tensor e15 is related to
the polarization induced by shear strain, which is not applicable in this device. The
relation between the lattice constants of the hexagonal GaN is given by Ambacher et al.
[15] as:
⎛ c − c0 ⎞ ⎛ C13 ⎞⎛ a − a0 ⎞
⎜ ⎟ = −2 ⎜ ⎟⎜ ⎟ (2.8)
⎝ c0 ⎠ ⎝ C33 ⎠⎝ a0 ⎠
where C13 and C33 are elastic constants provided in previous literature. The total
piezoelectric polarization in the direction of the c-axis can be determined by equations
2.7 and 2.8 [15]:
⎛ a − a0 ⎞ ⎛ ⎛ C13 ⎞ ⎞
PPE = 2 ⎜ ⎟ ⎜⎜ e31 − e33 ⎜ ⎟ ⎟⎟ . (2.9)
⎝ a0 ⎠ ⎝ ⎝ C33 ⎠ ⎠
⎡ ⎛ C ⎞⎤
Since the value of ⎢ e31 − e33 ⎜ 13 ⎟ ⎥ < 0 for AlGaN over the entire range of compositions,
⎣ ⎝ C33 ⎠ ⎦
the piezoelectric polarization is negative for tensile and positive for compressive strained
barriers, respectively [15]. GaN based semiconductors are attractive in HEMT designs
due to the high piezoelectric constants. The piezoelectric fields developed in GaN
devices can create a vertical electric field in the MV/cm range which increases the
channel conductivity. Many of the polarization effects outlined above have been modeled
into ATLAS™ as an upgrade to heterojunction modeling.
B. BERKELEY HEMT DESIGN

All HEMT structures provided to NPS were metal-organic chemical vapor
deposition (MOCVD) grown by Emcore Corporation [9]. The structure was grown on
[0001] c-plane sapphire substrate.
18
Figure 2 depicts the layer by layer structure of the device tested. From bottom to
top, it consists of the sapphire layer, a buffer layer, 1.5 µm of unintentionally doped
(UID) GaN, and approximately 25 nm of UID AlxGa1-xN. The Al composition, (x), in the
Berkeley devices was varied between 28% and 35% to assess the impact of Al
composition on the 2DEG transport properties and the device performance [9]. The GaN
HEMT was fabricated by a 3-step process: dry-etch mesa isolation, ohmic contact
metallization, and gate metallization. Figure 4 provides a SEM image of the device.
Table 3 [9] shows the AlxGa1-xN thickness as a function of the percent of Al.
Sample Al% AlxGa1-xN

Number thickness (Å)
1 28 267
2 29 230
3 30.2 230
4 32 221
5 34 242
6 35 206
Table 3 Al Composition and layer thickness of AlxGa1-xN barrier layer in

AlxGa1-xN/GaN HEMT structure [9].
19
20
III. AlGaN/GaN HEMT SILVACO MODEL
The model presented in this thesis was generated with Silvaco International’s
software package. An input deck was generated in the DECKBUILD™ VWF interactive
tool, solved through the ATLAS™ routine, and analyzed utilizing the TONYPLOT™ tool.
Additionally, the BLAZE™ routine, a routine specifically designed for group III-V
materials and devices with position dependent band structures was utilized to simulate the
AlGaN/GaN HEMT.
The desired result of this thesis was to generate an accurate HEMT model and
determine the thermal characteristics of the device as the substrate thickness and material
are changed. The GIGA™ routine was utilized to determine general thermal
characteristics of the device. GIGA™ extends the ATLAS™ program to account for
lattice heat flow and general thermal environments. The routing implements Wachutka’s
thermodynamically model of lattice heating which accounts for Joule heating, heating
and cooling due to carrier generation and recombination, and the Peltier and Thomson
effects [16].
The Silvaco software suite is a powerful simulation software that uses physics
based simulations. Physically-based device simulators predict the electrical
characteristics that are associated with specified physical structures and bias conditions.
ATLAS™ achieves this simulation by applying a two or three dimensional grid that forms
nodes throughout the structure. The nodes are solved in a matrix with charge transport
equations from Maxwell’s continuity and drift-diffusion equations [16].
The ATLAS™ simulator is a command driven simulator that requires an input in a

specific sequence. The order in which the statements occur is important to the proper
simulation of a device. There are five groups of statements, and if not correctly stated
will result in errors or erroneous simulations. The required groups and statements are as
follows [16]:
21
• Structure specification
o Mesh
o Region
o Electrode
o Doping
• Material Models Specification
o Material
o Models
o Contact Interface
• Numerical Method Selection
o Method
• Solution Specification
o Log
o Solve
o Load
o Save
• Results Analysis
o Extract
o TONYPLOT.
The above commands represent the entire structure definition and solution method. It is
critical to have a firm understanding of the simulation variables prior to defining a
structure. A small change in one variable or solution method can have undesirable
consequences on device performance. Figure 10 depicts a flow chart of the device
simulation utilizing Silvaco International’s simulators and modules.
22
Figure 10: Silvaco Simulation Flowchart [16].
A. SILVACO SEMICONDUCTOR MODELING EQUATIONS

The models developed to simulate the operation of semiconductor devices consist
of equations derived from Maxwell’s laws, Poisson’s equation, the continuity equations
and the drift-diffusion transport equations [16]. The following sections discuss the
equations in broad terms, and when necessary, a more detailed explanation of the
equations will be presented.
The method in which the above equations will be solved is a prime consideration
when developing a device model in ATLAS™. Several different numerical methods can
be used for calculating the solutions to various device structures. Different solution
methods are employed depending on the situation. The model developed in this thesis
incorporated the lattice heating parameter simulated through GIGA™. When the lattice
heating model is added to drift-diffusion equations an extra equation is added. The
BLOCK algorithm solves the three drift diffusion equations as a Newton solution and
follows this with a decoupled solution of the heat flow equations [16]. The NEWTON
23
algorithm solves all four equations in a coupled manner. NEWTON is preferred once the
temperature is high, however BLOCK is quicker for low temperature gradients. The
combination utilized for this thesis was: METHOD BLOCK NEWTON.
1. Poisson’s Equation
Poisson’s equation is a well known partial differential equation that has functions
in electrostatics and theoretical physics. In semiconductor modeling it often serves as the
starting point in obtaining quantitative solutions for electrostatic variables [12]. In
electrostatics, Poisson’s equation relates the electrostatic potential to the space charge
density and is given by [16]:
div ( ε∇Ψ ) = − ρ (3.1)
where Ψ is the electrostatic potential, ε is the local permittivity, and ρ is the local space
charge density. The local space charge density is a function of all mobile and fixed
charges including electrons, holes, and impurities. The electric field is obtained from the
gradient of the potential and is given by [16]:
→
E = ∇Ψ (3.2)
2. Carrier Continuity Equations

The carrier continuity equations for electrons and holes are defined by [16]:
→
∂n 1
= div J + Gn − Rn (3.3)
∂t q n
→
∂p 1
= div J p + G p − Rp (3.4)
∂t q
→ →
where n and p are the electron and hole concentration, Jn and Jp are the electron and
hole current densities, Gn and Gp are the generation rates for electrons and holes, Rn and
Rp are the recombination rates for electrons and holes and q is the magnitude of the
charge on the electron. It is possible to solve for only the holes or electrons in the
equations, and the model presented in this thesis solves the equations for electrons only.
Hole transport in the HEMT is insignificant.
24
3. Transport Equations
Electrons in thermal equilibrium at temperature TL within a semiconductor lattice
obey Fermi-Dirac statistics. That is the probability f( ε ) that an available electron state
with energy ε is occupied by an electron is given by [16]:
1
f (ε ) = (3.5)
⎛ ε -E F ⎞
1 + exp ⎜ ⎟
⎝ kTL ⎠
where EF is a spatially independent reference energy known as the Fermi level and k is
the Boltzmann’s constant. If ε - EF >> kTL, equation 4.5 can be approximated as:
⎛ E −ε ⎞
f (ε ) = exp ⎜ F ⎟ (3.6)
⎝ KTL ⎠
Statistics based on equation 3.6 are know as the Boltzmann statistics and were
utilized in the formulation of the HEMT model. The current density equations know as
the charge transport models are obtained by applying simplifications to the Boltzmann
transport equation. These assumptions can result in a number of different transport
models such as the drift-diffusion model, the energy balance model or the hydrodynamic
model [16]. The choice of charge transport model will then have a major influence on
the choice of generation and recombination model. The drift diffusion transport model
was utilized in the development of this model. The following equations were derived in
chapter three of [16], and show the conventional form of the drift-diffusion equation:
→ →
J n = qnµ En + qD ∇n
n n (3.7)
→ →
J p = qpµ E p + qD ∇p
p p
(3.8)
where Dn and Dp, in the case of the Boltzmann statistics corresponds to:
kTL
Dn = µn (3.9)
q
kTL
Dp = µp . (3.10)
q
25
4. Polarization Effects
Perhaps one of the most challenging aspects of modeling heterojunction devices
in ATLAS™ is implementing the polarization effects present at the interface of the
heterojunction. It has been widely published [15, 17] that surface donor like traps are the
source of electrons in the channel and the polarization effects in AlGaN/GaN structures
force the electrons into the channel. Previous modeling efforts have modeled the
⎛ e ⎞
polarization effects as an interface charge of approximate 1e13 ⎜ 2 ⎟ at the
⎝ cm ⎠
heterojunction interface.
Recent upgrades to ATLAS™ include the ability to model polarization effects

directly. The ATLAS™ polarization model was designed to simulate polarization in
wurtzite materials and include spontaneous polarization and piezoelectric polarization.
The model allows the designer to input the piezoelectric constants e31 and e33 as well as
the elastic constants C13 and C33. Various methods to implement the polarization model
into the device model were attempted however, all attempts were unsuccessful. When
incorporated into the model, the polarization effects were evident on the contour plots,
but I was unable to get a representative bandgap diagram or Id vs. Vd characteristic.
The polarization effects were modeled as an interface charge of approximately

⎛ e ⎞
1e13 ⎜ 2 ⎟ on the GaN side of the heterojunction. Ambacher et al. [15] utilizes the
⎝ cm ⎠
lattice constant (a), the piezoelectric constants (e), and the elastic constants (C) to derive
the sheet concentration density in wurtzite materials. The sheet concentration density
σ ( x) is given by :
σ ( x) = PPE ( Alx Ga1− x N ) + PSP ( Alx Ga1− x N ) − PSP (GaN ) . (3.11)
Substituting equation 2.9 for the PPE variable gives:
⎛ a(0) − a0 ( x) ⎞ ⎪⎧ ⎛ C13 ( x) ⎞ ⎪⎫
σ ( x) = 2 ⎜ ⎟ ⎨e31 ( x) − e33 ( x) ⎜ ⎟ ⎬ + PSP ( x) − PSP (0) .
⎝ a0 ( x) ⎠ ⎪⎩ ⎝ C33 ( x) ⎠ ⎪⎭ (3.12)
26
When the sheet charge density σ is divided by e, the result is the polarization
induced sheet charge density. Figure 11 illustrates the induced sheet charge as a function
of alloy composition (x).
Figure 11: Sheet charge as a function of alloy composition [15].
The device modeled in this thesis has an Al alloy content of 0.28, and a sheet charge of
0.99e13 cm-2 was introduced at the GaN side of the heterojunction to simulate the
polarization effects.
Modeling the device in this manner resulted in a more realistic device model and
allowed for continued thermal studies. When modeled with the interface charge, the
HEMT device was characterized well, and with small discrepancies, matched the
Berkeley devices. A comparison of the physical device in relation to the simulated
device will be presented later in this chapter.
27
B. SILVACO MODEL
The model developed in ATLAS™ was designed based on the work of Tzeng [11].
The purpose of her dissertation was to study noise parameters in the HEMT device. The
Naval Postgraduate School’s purpose is to study the thermal characteristics of the device
through measurements and a model developed in Silvaco to determine if a diamond
substrate will enhance the thermal characteristics of the HEMT device. The following
sections will contain a discussion of the modeling steps taken to simulate the HEMT
device.
1. Model Definition
To develop the software model of the HEMT, ATLAS™ requires an input in the
form of geometric statements to define the dimensions of the device to be modeled. The
first step is to define the structure dimensionally and then assign a mesh to enclose the
structure. The purpose of the mesh is to define solution points throughout the structure.
Defining the mesh correctly is one of the most important factors in the model
development. If the mesh is too coarse, the solutions will not be correct, and if the mesh
is too finely defined, the computation time can be prohibitive. The rest of the program
file is devoted to modifying statements and output parameters that will be discussed
individually when necessary. The entire model definition file, referred to as a *.in file is
provided in Appendix A. Figure 12 is a TONYPLOT™ structure view of the device
modeled. The structure definition is saved in ATLAS™ as a *.str file. The graphic
representing the HEMT device is a close up view in order to see the AlGaN layer. The
actual device is 20µm thick. The structure file view in TONYPLOT™ provides a visual
representation of the device performance. There are drop down menus that allow for a
selection of the desired plot.
28
Figure 12: 2D Representation of the Modeled AlGaN/GaN HEMT.
2. Gate Characterization
The characterization of the gate was accomplished by measuring the gate leakage
of the Berkeley device and then simulating the device. The gate on the HEMT is a
Schottky barrier, and the characterization was accomplished by adjusting the gate
workfunction until the leakage current on the simulation matched the measured values.
Figure 13 (a) shows the simulated gate characteristic of the tested device and Figure 13
(b) shows the measured characteristic of the tested device.
(a) (b)
™
Figure 13: (a) ATLAS Generated Gate Characterization and (b) Measured.
29
To test the device, the source and drain were set to 0V and the gate was swept from -6V
to 6V. The simulated version matches the tested version with the exception of how low
the current goes at VGS=0. The difference is most likely due to the meshing technique
used. Due to the magnitude of the current, is not a concern in the simulation. The
simulated workfunction value of the Schottky gate that provided the best characterization
was 4.3eV.
3. Heterojunction Bandgap Results
Simulating heterojunction devices in ATLAS™ is accomplished through
BLAZE™, a compound material 2D simulator. BLAZE™ accounts for the effects of
positionally dependent band structure by modifications to the charge transport equations
[16]. The difference in the two bandgaps creates an abrupt discontinuity in the valence
and conduction band. ATLAS™, through BLAZE™ provides three methods for defining
the conduction band alignment for a heterojunction. The method utilized in this model is
the ALIGN™ parameter. BLAZE™ allows the difference in the conduction band ∆E C to
be calculated by specifying the ALIGN parameter and the desired value of offset on the
material statement. BLAZE™ creates the desired conduction band offset by modifying
the electron affinity of the material for which the parameter is specified. The ALIGN
parameter was set to 0.8 on the AlGaN layer which allowed 80% of the bandgap
difference to appear as the conduction band discontinuity.
The 0.8 ALIGN parameter was established through previous HEMT design
attempts [5, 7] and trial and error. Figure 14(a) shows the valence band characteristic
with no bias, and Figure 14(b) shows the conduction band. The diagrams in Figure 14
are broken down into three regions. A cutline was utilized to extract the electrical
properties from the compiled structure. The left region represents the gate electrode, the
center region represents the AlGaN layer, and the right region is the GaN layer. The
unique bandgap feature of the HEMT is evident at the AlGaN/GaN interface.
30
(a) (b)
Figure 14: Band diagram of HEMT showing (a) Conduction Band and (b)
Valence Band.
Figure 15 is the full view of the bandgap diagram showing the conduction band in red,
the Fermi level in blue and the valence band in green. The bandgap of the simulated
structure was measured and found to be 3.42eV, the approximate value of GaN from
Table 1. The value of this figure is that it shows the structure design is correct and that
the 2DEG is formed on the GaN side of the heterojunction.
Gate
AlGaN 3.42eV
GaN
Figure 15: Band Diagram of modeled AlGaN/GaN HEMT.
31
4. Electron Concentration due to Polarization
Figure 15 clearly shows the effects of polarization at the heterojunction interface
by the formation of the triangular shaped quantum well. The polarization charge causes
the electron concentration to increase and a 2DEG is formed. As discussed in chapter II,
a net positive polarization charge is present at the AlGaN/GaN junction in Ga-faced
structures. Donor electrons will attempt to compensate this charge and will form the
2DEG that causes the conduction band to drop below the Fermi level at the interface.
Figure 16(a) is a contour view of the structure that shows the electron concentration. The
contour plot shows that there is a distinct 2DEG that has formed just below the AlGaN
layer (the thin layer just below the gate electrode) as predicted. Figure 16(b) shows the
cutline view of the electron concentration. The blue line is the electron concentration and
the red line is the donor concentration. The increase of electrons to approximately
⎛ e ⎞ ⎛ e ⎞
1e19 ⎜ 2 ⎟ from the background doping level of 1e15 ⎜ 2 ⎟ illustrates the effects of
⎝ cm ⎠ ⎝ cm ⎠
polarization and agrees with Ambacher’s predictions[15].
AlGaN
AlGaN
2DEG GaN
GaN
(a) (b)
Figure 16: Electron Concentration Contour View (a) and Cutline View (b).
32
5. I-V Curves
Once the model was complete and analyzed, a subroutine in ATLAS™ was
created to generate the Ids vs. Vds (I-V) curves. The physical device provided by Berkeley
was tested utilizing the UTMOST™ program and a Micromanipulator microscope probe
system. The data extracted from the device was imported to TONYPLOT™ for analysis
and comparison. The test performed on the model was performed with the same DC
parameters that were utilized in the physical device testing. Figure 17 is the I-V
characteristic of the HEMT model and Figure 18 is the I-V characteristic of the measured
device.
Figure 17: NPS HEMT Model I-V Characteristic.
33
Figure 18: Measured Berkeley HEMT Device I-V Characteristic.
The two I-V characteristic curves match closely with a few exceptions. The topmost
curve on the modeled device peaks at about 14 mA while the topmost curve on the
physical device peaks at about 12 mA. The topmost curve on the physical device has a
negative slope in the saturation region attributed to a negative resistance due to the poor
thermal qualities of sapphire. I was unable to match the negative going slope on the
modeled device. Although thermal effects were added to the device model, the
combination used in this effort did not produce the exact results. The I-V characteristics
match closely enough to give me confidence that the device was modeled correctly and it
can be used to conduct follow on thermal measurements.
C. THERMAL CHARACTERIZATION
The majority of the modeling efforts in this thesis have been dedicated to ensuring
that there was a proper HEMT device model to be used for thermal testing. The previous
sections have demonstrated that the HEMT device modeled is a good representation of
the Berkeley device when modeled at room temperature. This section will analyze the
34
model characteristics over a range of temperatures and compare the simulation results to
the measured results of the physical device over the same range.
1. Physical Device Thermal Testing
The thermal testing for the physical device was conducted utilizing the
UTMOST™ program and a hot chuck on the Micromanipulator testing microscope. The
physical device was tested at room temperature, 100º C, 200º C, and 300º C. Figure 19 is
a compilation of the I-V characteristics over the range of temperatures.
Figure 19: Topmost Curve (Vgs= -1 V) I-V Measurement of Physical Device over
a range of temperatures.
35
Figure 20, Figure 21, and Figure 22 represent the I-V characteristic of the room
temperature measurement against 100ºC, 200ºC, and 300ºC respectively. The figures
clearly show the current degradation due to the temperature increase.
Figure 20: I-V Characteristic Room Temperature vs. 100º C.
36
Testing the physical device over a range of temperatures served as the starting point for
the thermal model development. Utilizing the data gained from the physical devices, the
simulated device was modified to simulate lattice heating. The GIGA™ routine was
utilized in the thermal testing of the device model and will be discussed in section two of
this chapter.
2. NPS HEMT Model Thermal Testing
Introducing the thermal characteristics of a HEMT device is accomplished
through the GIGA™ routine. GIGA™ extends ATLAS™ to account for lattice heat flow
and general thermal environments [16]. The models included in this device (see
Appendix A) were the lat.temp model and the joule.heat model. These two models
enable ATLAS™ to apply heat flow equations to the solution of the device parameters.
To simulate the hot chuck in the physical device testing, a thermal contact was placed in
the substrate and set to the desired temperature. This applied the same temperature to the
model as the hot chuck applied to the actual device. The purpose of this type of testing
was to ensure that the HEMT model would react in the same manner as the physical
device did when heated. The results, although not exactly the same, were remarkably
similar and indicated that the modeled device reacted similar to the physical device under
thermal stress. The following figures are of the HEMT model I-V characteristics under
37
the same thermal environment the physical devices were subjected to. Figure 23, Figure
24, and Figure 25 represent the model I-V characteristics over temperature.
Figure 23: NPS HEMT Room Temperature vs. 100º C.
38
The I-V characteristics of the model show the same general trend as the physical device.
Table 4 shows the differences in the model and physical device. The error analysis is
conducted on the topmost curve which represents a gate voltage of -1 V.
Room Temperature 100º C 200º C 300º C
Physical Device 0.012 A 0.01 A 0.007 A 0.002 A

Percent Change 16.6% 41.6% 83%
Modeled Device 0.014 A 0.009 A 0.004 A 0.002 A

Percent Change 35.7% 71.4% 83%
Table 4: Percent Reduction in Current with Temperature.
Although the percent change is approximately double in the model design, the same
general decreasing current trend is observed in the modeled device. The trend indicates
that the thermal models incorporated in the HEMT model are within a factor of 2 of
degradation, and should be sufficient to investigate thermal testing. Figure 26 illustrates
the decreasing current trend as the lattice temperature of the device is increased from
room temperature to 300° C.
39
0.016
0.014
0.012
0.01
0.008
0.006
0.004 Measured
Sapphire Device
0.002
0 NPGS HEMT
Room Temp 100C 200C 300C Model
Figure 26: Decreasing Current trend of Actual Device and Modeled Device.
An important measure utilized in high power electronic FET devices is gate

leakage. When the device as previously described, there is a significant shift in the gate
leakage. Figure 27 illustrates an approximate increase of four orders of magnitude in
gate leakage per 100° C.
Figure 27: NPS HEMT Gate Leakage Over Temperature.

40
D. FINAL THERMAL SIMULATION
The final thermal simulation will consist of running the HEMT model at a high
DC bias which will result in high temperatures required to analyze a change in substrate
materials. To accomplish the shift in from sapphire to diamond substrate materials, the
thermal conductivity constant tc.a was set to 0.1 for a thermal conductivity of 10 and 0.05
for a thermal conductivity of 20. The values for thermal conductivity were taken from
[18], the SP3 corporation website. SP3 is working with NPS and UC Berkeley to
develop diamond substrates suitable for AlGaN/GaN HEMT development. The device
will be biased as described in the previous section except the drain will be swept from 0
to 55V.
1. Sapphire vs. Diamond Substrate Comparison
The point of testing the HEMT device on different substrates was to determine if
a change in substrate material will reduce the channel temperature. Once a suitable
HEMT device was modeled, it was swept from zero to 55 V on the drain creating a high
current situation. The result of the higher drive current was increased channel/device
temperature. Figure 28 shows a contour plot of the device on sapphire and diamond
substrates indicating the lattice temperature throughout the device. The thermal
conductivity of diamond ranges between 10 and 20, Figure 28 represents diamond
thermal conductivity set to 10 and sapphire thermal conductivity set to 1.7.
(a) (b)
Figure 28: Contour Plot of (a) Sapphire Substrate Material and (b) Diamond
Substrate.
41
The plot indicates that the peak temperature in the device changes dramatically with a
change in substrate material. Of particular note, the diamond substrate device was able to
dissipate more heat per unit area, and the size of the “hot spot” was reduced by
approximately 75%. The sapphire substrate device under the 55 V drain bias reached a
maximum temperature of 157ºC while the diamond substrate device reached a
temperature of 84º C. Figure 30 shows a drain current of approximately 0.06 A/mm for
the diamond substrate device with a TC constant of 20. With 55 V applied to the drain of
the device, the output power of the modeled device was calculated to be 3.3 W/mm.
Figure 29 is a plot taken from a horizontal cutline along the 2DEG of the sapphire
substrate device, the diamond substrate device with a thermal conductivity constant of 10
and 20. Although the simulation was conducted with varying diamond thermal
conductivities, the results indicate that the model is relatively insensitive to the thermal
conductivity change of diamond. There was a 2% change in channel temperature and an
8% change in drain current. The peak temperature of the device remained the same when
the thermal conductivity was changed from 10 to 20.
Figure 29 better illustrates the advantage the diamond substrate provides. The
area under each of the thermal curves represents the average temperature of the channel.
The average channel temperature of the sapphire substrate device is 137ºC, the diamond
substrate with a thermal conductivity parameter of 10 reduced the average channel
temperature to 47ºC, and the diamond substrate with a thermal conductivity parameter of
20 reduced the average channel temperature to 46ºC. The average channel temperature
was calculated utilizing the integrate function of TONYPLOT™. The simulation
indicates that a dramatic reduction in lattice temperature and channel temperature occurs
by changing the substrate material from sapphire to diamond.
42
Figure 29: Thermal Horizontal Cutline of 2DEG with Diamond Thermal
Conductivity set to 10 and 20. Sapphire (brown) Diamond 10 (green) and
Diamond 20 (red).
Figure 30 shows the device drain current utilizing the three different substrates.
Figure 30: Drain Current as a Function of Substrate Material.
43
The final parameter to be analyzed is the change in the gate leakage characteristic with a
change in substrate material. Figure 31 illustrates the gate leakage characteristic as a
function of substrate material.
Figure 31: Gate Leakage as a Function of Substrate Material.
There is a reduction of five orders of magnitude in the amount of gate leakage on the
diamond substrate device compared to the sapphire substrate device. Every measure of
effectiveness of a HEMT device improved with a change in substrate material. The
simulation indicates that HEMT devices grown on diamond substrates will improve the
overall effectiveness of the device when high power applications are required. Although
it is beyond the scope of this thesis, it is logical to assume that the reliability of the
devices will be enhanced if the temperature can be reduced.
Table 5 shows the results of the various parameters tested. The change was
calculated with the diamond TC=10 parameter.
44
Diamond TC=10 Diamond TC=20 Sapphire Change
Peak Temperature 84 C 84 C 150 C 56%
Channel Temperature 47 C 46 C 137 C 34%
Gate Leakage 7.00E-20 2.00E-20 1.00E-15 1E5
Drain Current .57 A .62 A .55 A 0.04%
Table 5: Summary of Testing Results.
E. CHAPTER CONCLUSION
This chapter introduced the modeling techniques utilized to simulate the
AlGaN/GaN HEMT device. To ensure proper thermal modeling a HEMT model was
developed and tested both at room temperature and at various temperatures. The results
of the model were compared with the physical device for accuracy. Once a suitable
device structure was developed, it was subjected to a high bias condition to generate
internal heating. The device parameters were then extracted from a sapphire based
substrate and then a diamond substrate to determine if the substrate material enhanced the
thermal capacity of the device. It was shown that a diamond substrate has a dramatic
effect on channel temperature. There was a 54% reduction in peak temperature and a
34% reduction in the channel temperature. The following chapter briefly summarizes the
work completed on this thesis and will make recommendations for follow-on simulations.
45
46
IV. CONCLUSIONS AND RECOMMENDATIONS
A. CONCLUSIONS
The AlGaN/GaN HEMT model developed in this thesis provides a representation
of an actual HEMT device that was tested over a range of temperatures from 270C to
3000C and validated against measured data. The thermal effects that occur by device self
heating effects were correctly modeled through the Silvaco’s GIGA™ routine. The
HEMT model was able to duplicate the actual transistor operation both electrically and
thermally. The I-V curves generated by the software model represented the actual device
I-V curves well with only a 13% difference at the topmost I-V curve.
The difficult part of this thesis was ensuring that a proper HEMT device was
modeled. A tremendous effort was made ensuring all operational aspects of the HEMT
device were addressed. The focus of the thesis however, was to examine the thermal
effects of the device when the substrate material was changed from sapphire to diamond.
Changing the substrate and examining the thermal properties was accomplished by
changing the material name in the ATLAS™ file and adjusting the thermal conductivity
constants. Sapphire’s thermal conductivity was set to 1.7 W/cm·K, and diamond’s
thermal conductivity was set to both 10 and 20 W/cm·K to test over the range given by
SP3 incorporated. The effects of the substrate change were drastic. There was a 1/2
reduction in the peak temperature of the device when diamond was used see Table 5.
The average temperature in the channel was reduced by 1/3. Thermal analysis is
important in the calculation of device lifetime. One approach to predicting device
median lifetimes is the Arrhenius model for temperature dependence [4] and is given by:
EA
tm (T ) = Ae kT
(4.1)
where EA is the activation energy (eV), A is a constant acceleration factor

(dimensionless) and k is the Boltzmann’s constant (eV/T). Activation energies range
from 0.3 eV to 1.7 eV for GaN HEMTs [7]. Utilizing an acceleration factor of 1 and
47
activation energy of 0.3 eV and 1.7 eV, the median lifetime of a HEMT device can
theoretically be increased to a range of 5 to 1e6 times the median lifetime of a sapphire
substrate HEMT.
Initially, I attempted to utilize the new polarization models available in

ATLAS™. Many attempts to model the polarization effects by utilizing the PSP and
POLARIZATION statements failed. I was unable to get a suitable bangap diagram or
2DEG utilizing the new BLAZE™ models. The interface statement on the AlGaN side
of the heterojunction provided the best method to simulate the polarization effects
experienced in wurtzite materials. The interface command places a sheet charge at the
interface and provides the necessary electron concentration to form the 2DEG.
The model designed in this thesis proved to be relatively insensitive to the thermal
conductivity constants given for diamond. There was very little change in any of the
parameters analyzed when the constants were varied through the published values.
Although many factors in the ATLAS™ code had to be optimized to get a representative
HEMT device, one of the enabling parameters to simulate a geometrically correct device
was the Albrecht model. Albrecht et al. [19] developed a set of transport parameters for
devices with electron conduction GaN (position of the 2DEG). The Albrecht parameter
models the carrier mobility as a function of doping and lattice temperature which are
critical parameters in the device modeled for this thesis. In [19] Albrecht provides initial
experimental values for the variables available in the ATLAS™ model. These values
were implemented in the HEMT model and then optimized to match the measured device
characteristics. The variables in the Albrecht model proved to be the most sensitive
parameters affecting the I-V curves and further analysis and optimization efforts should
provide a better I-V representation.
B. RECOMMENDATIONS
Follow-on efforts should examine the possibility of implementing the polarization
and piezoelectric models available in ATLAS™ to determine if a suitable model can be
developed utilizing the new polarization models.
48
The parameters examined in this thesis were DC measurements only. One of the
advantages of a HEMT is that it has the ability to operate at high frequencies. It would
be beneficial to examine the RF characteristics of the model developed in this thesis to
see if the modeled device can accurately simulate the RF parameters of an actual HEMT
device. Once the basic characteristics are extracted, the device should be simulated at
high-frequency, high-power conditions to determine if the benefits of a diamond substrate
are as dramatic in the AC domain.
The HEMT design presented utilized an 18µm thick sapphire or diamond

substrate. The purpose of simulating the model with a thick substrate was to analyze the
thermal parameters when the device was not saturated thermally. The assumption in this
model is that the device is capable of dissipating all of the heat applied. In reality, the
substrate thickness will be much smaller and a more comprehensive heat flow study
should be conducted in order to simulate a more realistic environment.
The two dimensional analysis of the device provided and adequate simulation of
the HEMT device, however Silvaco has a three dimensional capability that can provide a
more comprehensive analysis of the device operational parameters. The model
developed in this thesis can be expanded to a three dimensional model in order to re-
examine the electrical and thermal properties of the device. An initial attempt was made
to extend the model to three dimensions, but I was unable to get the model to converge at
a high bias condition. The probable cause of the failure was meshing techniques. Great
care must be taken to properly mesh the device for simulation. In three dimensions, the
number of nodes required for a solution increases dramatically. If there are too many
nodes, the computation time is prohibitively long, and if the mesh is reduced too much,
the results will not be accurate.
The physical device grown on a sapphire substrate tested for this thesis was
designed and fabricated by S. Tzeng at UC Berkeley. The Weber material science group
at UC Berkeley, NPS, and SP3 are currently working to grow AlGaN/GaN HEMT
devices on diamond substrates. Once a viable diamond substrate device is created, a
comprehensive test should be preformed on the device to see if the ATLAS™ model
accurately predicts the thermal effects of the diamond substrate device. Once the
49
diamond HEMTs are available and tested to ensure model accuracy, electrical defect
densities could be modeled to increase with temperature. A study could be performed to
predict degradation due to temperature and time. The initial results gained from this
thesis indicate that AlGaN/GaN HEMTs grown on a diamond substrate have the
possibility to be the enabling technology for high-power high-frequency operations.
50
LIST OF REFERENCES
1. F. Ren and J. C. Zolper, Wide Energy Bandgap Electronic Devices. River Edge,
NJ: World Scientific Publishing Co. Pte. Ltd., 2003, pp. 514.
2. E. J. Martinez, "Gallium Nitride & Related Wide Bandgap Materials and

Devices," pp. 17, 2005.
3. J. A. Mittereder, S. C. Binari, P. B. Klein, J. A. Roussos, D. S. Katzer, D. F.

Storm, D. D. Koleske, A. E. Wickenden and R. L. Henry, "Current collapse
induced in AlGaN/GaN HEMTs by short-term DC bias stress," in 2003, pp. 320-
323.
4. F. Jensen. 1995, Electronic Component Reliability.
5. K. P. Eimers, "2-D Modeling of GaN HEMTs Incorporating the Piezoelectric

effect." Master's Thesis, Naval Postgraduate School, Monterey, California, 2001.
6. K. L. Holmes. 2002, Two-dimensional modeling of aluminum gallium nitride.

Master's Thesis, Naval Postgraduate School, Monterey, California
7. R. P. Salm. 2005, Thermal modeling of GaN HEMTs on sapphire and diamond.

Master's Thesis, Naval Postgraduate School, Monterey, California.
8. J. C. Freeman. 2003, Basic equations fro the modeling of gallium nitride (GaN)
high electron mobility transistors (HEMTs). NASA/TM--2003-211983 Available:
http://gltrs.grc.nasa.gov/reports/2003/TM-2003-211983.pdf
9. S. Tzeng, "Low-Frequency Noise Sources in III-V Semiconductor

Heterostructures," 2004.
10. P. Javorka, "Fabrication and Characterization of AlGaN/GaN High Electron

Mobility Transistors." 2004.
11. S. M. Sze, Modern Semiconductor Device Physics. New York: John Wiley &
Sons, Inc., 1998, pp. 556.
12. R. F. Pierret, Semiconductor Device Fundamentals. Massachusetts: Addison

Wesley Longman, 1996, pp. 792.
13. M. Lundstrom, Fundamental of Carrier Transport. ,2nd ed.New York:

Cambridge University Press, 2000, pp. 418.
14. P. Valizadeh and D. Pavlidis, "Investigation of the impact of Al mole-fraction on

the consequences of RF stress on Al/sub x/Ga/sub 1-x/N/GaN MODFETs,"
Electron Devices, IEEE Transactions on, vol. 52, pp. 1933-1939, 2005.
51
15. O. Ambacher, B. Foutz, J. Smart, J. R. Shealy, N. G. Weimann, K. Chu, M.
Murphy, A. J. Sierakowski, W. J. Schaff, L. F. Eastman, R. Dimitrov, A. Mitchell
and M. Stutzmann, "Two dimensional electron gases induced by spontaneous and
piezoelectric polarization in undoped and doped AlGaN/GaN heterostructures,"
J. Appl. Phys., vol. 87, pp. 334-344, 01/01/. 2000.
16. Silvaco International, "Silvaco International, ATLAS User's Manual, February

2000." 2000.
17. I. P. Smorchkova, C. R. Elsass, J. P. Ibbetson, R. Vetury, B. Heying, P. Fini, E.

Haus, S. P. DenBaars, J. S. Speck and U. K. Mishra, "Polarization-induced charge
and electron mobility in AlGaN/GaN heterostructures grown by plasma-assisted
molecular-beam epitaxy," J. Appl. Phys., vol. 86, pp. 4520-4526, 10/15/. 1999.
18. SP3 Incorporated. Available: http://www.sp3inc.com/
19. J. D. Albrecht, R. P. Wang, P. P. Ruden, M. Farahmand and K. F. Brennan,

"Electron transport characteristics of GaN for high temperature device modeling,"
J. Appl. Phys., vol. 83, pp. 4777-4781, 05/01/. 1998.
52
APPENDIX: ATLAS™ INPUT DECK
######################################################################
######################################################################
#LT Scott Newham
#NPS AlGaN/GaN HEMT Design
#
#267 Angstrom AlGaN layer
# Schottky WF=4.3
# 1e15 doping GaN AlGaN=1e10
#Interface charge=.99e13
# no doping in AlGaN
# Albrct mobility model an.albrct=5e-3 bn.albrct=3e-3 cn.albrct=2e-3
#######################################################################
go atlas
set devthk=20
set sourcegatespace=2
set gatedrainspace=3
set devwidth=4+$sourcegatespace+$gatedrainspace
set reg3xmin=$devwidth-1
set reg4xmin=1+$sourcegatespace
set reg4xmax=2+$reg4xmin
set reg5xmax=$devwidth-.5
set drainxmin=$devwidth-1
set drainxmax=$devwidth-.6
set ifchargemax=$devwidth-1.05
set WF=4.3
# Sweep increment
set vstart = 0
set vstop = 7
set vinc = .25
#################### Mesh Construction #############
####Mesing and width = 100 microns#################
mesh auto width=100
x.m l=0.0 s=0.1
x.m l=0.5 s=0.1
x.m l=1.0 s=0.1
x.m l=$reg4xmin s=0.1
x.m l=$reg4xmax s=0.1
x.m l=$reg3xmin s=0.1
x.m l=$reg5xmax s=0.1
x.m l=$devwidth s=0.1
y.m l=0 s=0.5
y.m l=0.4 s=0.1
y.m l=0.49 s=0.005
y.m l=0.5 s=0.005
y.m l=0.525 s=0.005
y.m l=0.5267 s=0.00005
y.m l=0.5269 s=0.00005
y.m l=0.5270 s=0.005
y.m l=1.0 s=0.1
y.m l=2 s=0.5
y.m l=$devthk s=2.0
######################## Region Definitions ################
region num=1 mat=AlGaN donors=1e10 x.comp=0.28 x.min=0.5 x.max=$reg5xmax

y.min=0.5 y.max=0.5267
region num=2 mat=oxide x.min=0 x.max=$devwidth y.min=0 y.max=0.5
region num=3 mat=GaN donors=1e15 x.min=0.5 x.max=$reg5xmax y.min=0.5267
y.max=0.5269
53
region num=4 mat=GaN donors=1e15 x.min=0.5 x.max=$reg5xmax y.min=0.5269
y.max=1.0
region num=5 mat=GaN donors=1e15 x.min=0 x.max=$devwidth y.min=1 y.max=2
region num=6 mat=Sapphire x.min=0 x.max=$devwidth y.min=2 y.max=$devthk
elec num=1 name=source x.min=0 x.max=0.5 y.min=0.5 y.max=1
elec num=2 name=drain x.min=$drainxmax x.max=$devwidth y.min=0.5 y.max=1
elec num=3 name=gate x.min=$reg4xmin x.max=$reg4xmax y.min=0 y.max=0.5
elec num=4 substrate
############## Modifying Statements ##################

#######Inserts polarization effects#######
interface charge=0.99e13 y.min=.52 y.max=.5267 s.s
material mat=AlGaN align=.8

mobility albrct.n an.albrct=5e-3 bn.albrct=3e-3 cn.albrct=2e-3 vsatn=2e7
material mat=GaN tcon.polyn
material mat=AlGaN tcon.polyn
material mat=Sapphire tcon.polyn
models k.p print lat.temp joule.heat srh albrct
contact name=gate work=$WF
#thermcontact num=1 x.min=0 x.max=$devwidth y.min=2 y.max=5 temp=300 ^boundary

alpha=1.7
thermcontact num=1 x.min=0 x.max=$devwidth y.min=14 y.max=$devthk temp=300
^boundary alpha=1.7
output con.band val.band charge
method gumits=300 clim.dd=1e5 autonr block carr=1
###################### Output Statements ##########
# idvd curves
solve
save outf=NPGS_SAP_g0.str
log outf=NPGS_SAP_d_0.log
solve name=drain vdrain=$vstart vfinal=$vstop vstep=$vinc
save outf=NPGS_2D_SAP_g0d7.str
log off
tonyplot NPGS_SAP_d_0.log -set IDVD.set
solve vdrain=0 vgate= -5
save outf=solve_vgate-5.str
log off
log off
log off
log off
log off
#solve vdrain=0 vgate= 0
#save outf=solve_vgate_0.str
54
#log outf=NPGS_SAP_d_6.log
#solve name=drain vdrain=$vstart vfinal=$vstop vstep=$vinc
#log off
#solve vdrain=0 vgate= 1
#save outf=solve_vgate_1.str
#log outf=NPGS_SAP_d_7.log
#solve name=drain vdrain=$vstart vfinal=$vstop vstep=$vinc
#log off
tonyplot NPGS_SAP_d_0.log -overlay NPGS_SAP_d_1.log -overlay NPGS_SAP_d_2.log -
overlay NPGS_SAP_d_3.log -overlay NPGS_SAP_d_4.log -overlay NPGS_SAP_d_5.log -\
overlay NPGS_SAP_d_6.log -overlay NPGS_SAP_d_7.log -set IDVD.set
# idvg curve
#log outf=NPGS_GaNsapp-gate_sweep.log
#solve vdrain=0
#solve name=gate vgate=0 vfinal=-4.2 vstep=.6
quit
55
56
INITIAL DISTRIBUTION LIST
1. Defense Technical Information Center

Ft. Belvoir, Virginia
2. Dudley Knox Library

Naval Postgraduate School
Monterey, California
3. Engineering and Technology Curricula Office, Code 34

4. Chairman, Code EC (590)

Department of Electrical and Computer Engineering
5. Professor Todd R. Weatherford, Code EC(590)/Wt

6. Andrew A. Parker, Code EC(590)/Pk

Monterey, California\
7. Dr. Petra Specht

Materials Science and Engineering
University of California, Berkeley at LBNL
Berkeley, California
8. Jerry Zimmer
SP3 Diamond Technologies
Santa Clara, California
9. LT Wesley S. Newham, United States Navy

Salinas, California
57

Development of Algan/Gan High Electron Mobility Transistors (Hemts) On Diamond Substrates

Uploaded by

Document Informationclick to expand document informationbien

Document Informationclick to expand document information

Copyright:

Available Formats

Development of Algan/Gan High Electron Mobility Transistors (Hemts) On Diamond Substrates

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Development of Algan/Gan High Electron Mobility Transistors (Hemts) On Diamond Substrates

Uploaded by

Copyright:

Available Formats

Calhoun: The NPS Institutional Archive

Theses and Dissertations 1. Thesis and Dissertation Collection, all items

Development of AlGaN/GaN High Electron

Newham, Wesley Scott.

Downloaded from NPS Archive: Calhoun

DEVELOPMENT OF AlGaN/GaN HIGH ELECTRON

Wesley Scott Newham

Thesis Advisor: Todd R. Weatherford

Approved for public release; distribution is unlimited

14. SUBJECT TERMS 15. NUMBER OF

DEVELOPMENT OF AlGaN/GaN HIGH ELECTRON MOBILITY

Submitted in partial fulfillment of the

MASTER OF SCIENCE IN ELECTRICAL ENGINEERING

NAVAL POSTGRADUATE SCHOOL

Author: Wesley Scott Newham

Approved by: Todd R. Weatherford

Figure 1: Tri-Service Team Interaction [4]. ......................................................................5

Table 1: Electronic Properties of Semiconductor Materials [2].......................................2

Current AlGaN/GaN High Electron Mobility Transistors (HEMTs) grown on

Table 1: Electronic Properties of Semiconductor Materials [2].

B. ALGAN/GAN HEMT RESEARCH

• Develop reliable, high performance devices and MMICs with

o Superior thermal performance

• Rapid insertion into DoD RF systems.

The WBGSTI is a multi-phase, multi-year project that is poised to revolutionize

C. SUMMARY OF PREVIOUS RESEARCH

In [6], Holmes attempted to incorporate the polarization, piezoelectric and

This chapter presents an introduction to the High Electron Mobility Transistor

Ti/Al/Ni/Au ohmic metallization was deposited by e-beam evaporation and

2. AlGaN/GaN HEMT Fundamentals

The typical DC measurement to determine the operation of a HEMT is the

Figure 5: DC Output characteristic of AlGaN/GaN HEMT.

semiconductor, respectively, and ∆d can be interpreted as the effective thickness of the

The threshold voltage is an important parameter because it is a measure of

Saturation current and transconductance are important measures of a

One of the measures of quality for a HEMT device is transconductance. It

Additionally, the output conductance known as the drain or channel

As mentioned above, the HEMT is a three terminal device where

Figure 6: 2DEG electron density distribution for different Al(x) composition

The HEMT operational characteristics are based on the MESFET. The

and µ is electron mobility.

Figure 7: Bandgap diagram of AlGaN/GaN HEMT [9].

• Source voltage (Vs) =0

• Drain voltage (Vd) = 0

concentration of ns , the 2DEG is capable of transferring more charge from source to

Unlike a typical MOS Id vs. Vds curve, GaN HEMTs on a sapphire

Figure 8: Schematic drawing of the crystal structure of wurtzite Ga-face[15].

In the case of the wurtzite crystal grown in the manner previously

Table 2: Spontaneous polarization, piezoelectric and dielectric constants of

PPE = e33ε z + e31 ( ε x + ε y ) (2.7)

B. BERKELEY HEMT DESIGN

Sample Al% AlxGa1-xN

Table 3 Al Composition and layer thickness of AlxGa1-xN barrier layer in

The ATLAS™ simulator is a command driven simulator that requires an input in a

• Material Models Specification

• Numerical Method Selection

A. SILVACO SEMICONDUCTOR MODELING EQUATIONS