Fabrication and characterization of gallium nitride electronic devices

MISSING IMAGE

Material Information

Title:
Fabrication and characterization of gallium nitride electronic devices
Physical Description:
ix, 268 leaves : ill. ; 29 cm.
Language:
English
Creator:
Johnson, Jerry Wayne, 1975-
Publication Date:

Subjects

Subjects / Keywords:
Chemical Engineering thesis, Ph. D   ( lcsh )
Dissertations, Academic -- Chemical Engineering -- UF   ( lcsh )
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 2001.
Bibliography:
Includes bibliographical references (leaves 250-267).
General Note:
Printout.
General Note:
Vita.
Statement of Responsibility:
by Jerry Wayne Johnson.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 028219551
oclc - 49323368
System ID:
AA00022657:00001

Full Text











FABRICATION AND CHARACTERIZATION OF GALLIUM NITRIDE
ELECTRONIC DEVICES











*' ** .. ..


I '
'O - i






By

JERRY WAYNE JOHNSON


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


2001
























To DSJ and MSJ, with loving memories of the past and dreams for a bright future.














ACKNOWLEDGMENTS

The completion of such a manuscript is in no way the sole accomplishment of the

author. I would first like to thank my research advisor, Fan Ren, for technical support

and guidance throughout every aspect of this work. I would also like to extend my

deepest gratitude to him for accepting me into his group and for the permanent mark he

has made on my life both professionally and personally. He has taught me more than he

could ever imagine. I am also indebted to the other members of my advisory committee:

Tim Anderson, Steve Pearton, and Cammy Abernathy. Their significant contributions to

this work are greatly appreciated. I am honored to have been associated with such an

eminent committee.

Collaboration in scientific research has become the norm, not the exception.

Certainly this work is a testament to the joint efforts of many individuals. First I would

like to thank my colleagues and co-workers in the Department of Chemical Engineering:

Ben Luo, Anping Zhang, Gerard Dang, Jeff LaRoche, Rishabh Mehandru, and Jihun

Kim. I specifically thank Ben for many enlightening discussions and Friday meetings.

May each of these continue in my absence. The excellent crystal growth, vacuum system

expertise, and friendship of Brent Gila are gratefully acknowledged. Thanks are also

extended to Kyu-Pil Lee for always finding time to assist me with ICP etching.

Several others have helped make the past 4 years more enjoyable and/or more

rewarding, and I thank them for it. Jay Lewis is recognized for his friendship, golf

tournaments, and 6-string wizardry. Mike Mastro has endured sharing office space with








me during our concurrent graduate careers. Olga Kryliouk has been a friend,

collaborator, and at times, a surrogate Mother. She is indeed one of the nicest people I

have ever met. Nancy and Shirley in the ChE office deserve special recognition for their

ever-smiling faces and friendly dispositions. I'd like to thank Mark George at Deposition

Sciences for lessons learned in the finer points of both LP-CVD and Red Zinfindel.

There truly is a necessary balance in life. I am grateful to Albert Baca for the opportunity

to spend 4 months at Sandia National Laboratories. Professionally, my time there was

one of the most productive of my life. Acknowledgement is also warranted for fellow

Sandians Ron Briggs, Cedric Monier, Randy Shul, Melissa Cavaliere, Gerry Lopez,

Andrea Ongstad, Joel Wendt, and Marce Armendariz.

To my family, I am eternally indebted. Since my earliest memories, they have

been a pillar of support and a source of confidence. They have motivated and allowed

me to strive for things I could have never accomplished on my own. I thank them for

their hard work, dedication to family, and their unwavering love. I will strive to model

my fathering style after the parenting skills they have always demonstrated. If I am

successful, Max will be very lucky-as I have been.

Finally, I thank my wife Traci. She is truly the most special person I have ever

met and the impact she has made on my life is immeasurable. She is my wife, the mother

of our son, and my very best friend. I feel this work is as much her achievement as it is

my own. I wish there was a way for me to express the deepness of my love and

appreciation.














TABLE OF CONTENTS

page

ACKNOW LEDGM ENTS................................................................................................. iii

ABSTRACT..................................................................................................................... viii

CHAPTERS

1 INTRODUCTION ........................................................................................................... 1

1.1 M otivation............................................................................................................ 1
1.2 Dissertation Outline............................................................................................... 10

2 BACKGROUND AND LITERATURE REVIEW ....................................................... 12

2.1 Historical Review ................................................................................................... 12
2.2 Crystal Structure .................................................................................................... 17
2.3 Thin Film Growth.................................................................................................. 18
2.3.1 Substrates for III-N Epitaxy............................................................................ 18
2.3.2 Doping............................................................................................................. 23
2.4 Etching................................................................................................................... 27
2.4.1 Photoenhanced W et Chemical Etching........................................................... 27
2.4.2 Dry Etching..................................................................................................... 29
2.5 M etallization.......................................................................................................... 33
2.5.1 Ohm ic Contacts............................................................................................... 37
2.5.1.1 n-type ....................................................................................................... 37
2.5.1.2 p-type ....................................................................................................... 40
2.5.2 Rectifying Contacts......................................................................................... 42
2.5.2.1 n-type ....................................................................................................... 42
2.5.2.2 p-type ....................................................................................................... 44
2.5.3 Surface Treatment........................................................................................... 44
2.6 III-N Optoelectronics............................................................................................. 45
2.6.1. Light Emitting Diodes.................................................................................... 46
2.6.2 Laser Diodes................................................................................................... 53
2.6.3 Detectors......................................................................................................... 57
2.7 III-N Electronics .................................................................................................... 58
2.7.1 Field Effect Transistors................................................................................... 59
2.7.2 Bipolar Transistors.......................................................................................... 65
2.7.3 Rectifiers......................................................................................................... 69









3 AlGaN / GaN HIGH ELECTRON MOBILITY TRANSISTORS................................ 72

3.1 Introduction........................................................................... ............................. 72
3.2 M OCVD-Grown HEM Ts on A1203 Substrates..................................................... 75
3.2.1 Device Processing........................................................................................ 75
3.2.2 Optical Gate Devices ...................................................................................... 81
3.2.3 Subm icron Gate Devices................................................................................. 86
3.2.4 Sm all Signal M odeling................................................................................. 104
3.3 Direct Comparison of HEMTs Grown on A1203 or AIN/SiC.............................. 111
3.4 AIGaN/GaN HEMTs on RF-assisted MBE-grown Epilayers............................. 143

4 GROWTH AND ELECTRICAL CHARACTERIZATION OF NOVEL GATE
OXIDES FOR GaN M OS DEVICES .......................................................................... 151

4.1 Introduction.......................................................................................................... 151
4.2 M olecular Beam Epitaxial Growth...................................................................... 152
4.3. Quantification of Oxide Quality......................................................................... 154
4.4. The M OS Capacitor............................................................................................ 157
4.4.1 Collection of Capacitance-Voltage Data...................................................... 159
4.4.2 Ideal M OS Capacitor.................................................................................... 160
4.4.2.1 Accum ulation......................................................................................... 161
4.4.2.2 Depletion................................................................................................ 164
4.4.2.3 Inversion ................................................................................................ 166
4.4.2.4 Flatband.................................................................................................. 168
4.4.2.5 Deep Depletion...................................................................................... 170
4.4.3. Non-Ideal M OS............................................................................................ 171
4.4.3.1 Effective Oxide Charge.......................................................................... 171
4.4.3.2 Interface Trapped Charge...................................................................... 173
4.5. Gadolinium Oxide............................................................................................... 176
4.5.1. High Temperature Gd203............................................................................. 179
4.5.2 Low Temperature Gd203 .............................................................................. 183
4.6. Scandium Oxide.................................................................................................. 191

5 GaN METAL OXIDE SEMICONDUCTOR FIELD EFFECT TRANSISTORS ......194

5.1 Introduction......... ........................................................................................... 194
5.2 M OSFET Processing........................................................................................... 196
5.3 Results and Discussion........................................................................................ 200
5.3.1 Gd203 / GaN Devices.................................................................................... 200
5.3.2 SiO2 / Gd203 / GaN Stacked Dielectric Devices.......................................... 203









6 GaN SCHOTTKY RECTIFIERS ON BULK GaN SUBSTRATES........................... 208

6.1 Introduction.......................................................................................................... 208
6.2 Material Characterization and Diode Fabrication................................................ 211
6.3 Current-Voltage Results and Discussion: Schottky Diodes................................. 217
6.4 Large- and Small-Area Bulk GaN Rectifiers with Implanted Guard Rings........ 224
6.5 Switching Behavior.............................................................................................. 237

7 SUMMARY, CONCLUSIONS, AND FUTURE WORK.......................................... 240

7.1. AIGaN / GaN HEMTs........................................................................................ 240
7.2. Gate Oxide Growth and Processing.................................................................... 243
7.3. MOSFETs and MOSHEMTs.............................................................................. 246
7.4. GaN Rectifiers on Bulk Substrates..................................................................... 248

REFERENCES ................................................................................................................ 250

BIOGRAPHICAL SKETCH ........................................................................................... 268














Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

FABRICATION AND CHARACTERIZATION OF GALLIUM NITRIDE
ELECTRONIC DEVICES

By

Jerry Wayne Johnson

December 2001

Chairman: Fan Ren
Major Department: Chemical Engineering

Gallium nitride (GaN)-based high electron mobility transistors (HEMTs), metal

oxide semiconductor field effect transistors (MOSFETs), and Schottky rectifiers were

fabricated and characterized. Novel dielectric materials Gd203 and ScO were evaluated

as potential gate dielectrics for GaN MOS applications. The devices presented herein

show tremendous potential for elevated temperature, high frequency, and/or high voltage

operation.

AlGaN/GaN HEMTs were grown by MOCVD on sapphire and SiC substrates and

by RF-MBE on sapphire substrates. Devices were fabricated with gate lengths from 100

nm to 1.2 prm. Drain current density approached 1 A/mm and extrinsic transconductance

exceeded 200 mS/mm for small gate periphery devices. For the shortest gate length, a

unity-gain cutoff frequency (fT) of 59 GHz and a maximum frequency of oscillation (fm)

of 90 GHz were extracted from measured scattering parameters. The experimental

s-parameters were in excellent agreement with simulated results from small-signal linear








modeling. Large signal characterization of 0.25 x 150 pm2 devices produced 2.75 W/mm

at 3 GHz and 1.7 W/mm at 10 GHz. Devices fabricated on high thermal conductivity SiC

substrates exhibited superior high temperature performance and a reduced density of

threading dislocations.

Novel gate dielectrics Gd203 and ScO were grown by gas source molecular beam

epitaxy (GSMBE). Current-voltage (I-V) and capacitance-voltage (C-V) data were

collected from MOS capacitors to evaluate the bulk and interfacial electrical properties of

the insulators. Single crystal Gd203 was demonstrated on GaN, but the resultant

MOSFET exhibited a large gate leakage attributed to defects and dislocations in the

oxide. MOSFETs with a stacked gate dielectric of Gd2O3/SiO2 were operational at a

drain source bias of 80 V and a gate bias of+7 V.

Bulk GaN templates grown by hydride vapor phase epitaxy (HVPE) were used to

fabricate vertical geometry Schottky rectifiers. Size- and temperature-dependent I-V

characteristics are reported. These devices show significant improvements in forward

turn-on voltage, on-state resistance, and reverse recovery characteristics relative to

previously reported devices fabricated on GaN layers grown on sapphire.













CHAPTER 1
INTRODUCTION


One might say, with tongue in cheek, perhaps, that once we
have developed diodes, transistors, tubes, resistors,
capacitors, and insulators to tolerate 500 []C, we will no
longer have thermal design problems.

Editorial, Electronics Design, p. 22, May 14, 1958


1.1 Motivation

The bulk of modem microelectronics rests upon the technology of silicon-based

transistors. Silicon is the 2nd most abundant element in the Earth's crust, and its

availability, attractive material properties, and stable native oxide have all contributed to

its position of prominence. Since shortly after the advent of the transfer resistor, or

"transistor," in the mid-20th century, the Si-based semiconductor electronics industry has

grown at an unprecedented pace. Moore's Law scaling has led to an exponential rate of

development never before witnessed in any other industry (Figure 1-1). Today, the Si

transistor is believed to be the most numerous man-made object on our planet [1],

totaling well over 1 million Si MOSFETs for each person in the Western world. While

silicon devices dominate over 90% of the -$250 billion solid state market and continue to

experience strong growth, alternative material systems are becoming increasingly

important.









As technology continues to advance, heightened performance demands are being

made on semiconductor devices for use in both niche and mainstream applications.

While silicon has proven to be the primary contestant in the integrated circuit (IC)


1970 1975 1980 1985
Year


1990 1995 2000


Figure 1-1. Transistors per chip for Intel CPUs from 4004 Microprocessor (1971) to
Pentium III (late 1999). Note the log scale of the vertical axis. Moore's Law has held
as the empirical economic and forecast basis for the entire silicon electronics industry for
over 25 years [2].



market, there is an ever-growing need for devices operating at conditions beyond the

limits of silicon. Satellite and radar communication systems, high temperature (>200C)

electronics, and high voltage solid state switching are examples of the numerous areas of

potential for appropriately designed non-silicon devices and circuitry. In addition,

traditional devices operating in harsh (e.g., chemical, radiation) environments are

desirable for many military and aerospace applications. For such devices, the limitations


Intel CPUs *









U
U


I I I I .


!


!








of silicon are largely due to its inherent material properties, such as its narrow indirect

band gap.

Group Im-V compound semiconductors represent an important class of alternative

materials for various electronic and optoelectronic applications. GaAs is by far the most

mature of the III-Vs, but still lags well behind Si in terms of development. GaAs

technology is attractive for many reasons, including a small electron effective mass, high

drift velocity, and the availability of AlGaAs/GaAs heterostructures. GaAs substrates

can be grown by typical crystal pulling techniques in large diameters at relatively low

cost. An additional benefit for device applications is the availability of semi-insulating

substrates, which reduce parasitic effects and enable high frequency devices. The wider

bandgap of GaAs (1.42 eV versus 1.12 eV for silicon) greatly enhances radiation

tolerance and high temperature operation. In addition, GaAs and most other III-V

materials have direct energy gaps, which allows fabrication of efficient light-emitting

devices such as light emitting diodes (LEDs) and laser diodes (LDs). Commercially

important Im-As electronic devices include GaAs MESFETs for RF power amplification.

GaAs monolithic microwave integrated circuits (MMICs) are helping to fuel the booming

market for wireless communications. Impressively, the number of GaAs FETs on a chip

has reached into the millions [3]. One of the most important optoelectronic devices is the

AIGaAs/GaAs LD, used extensively in the read/write heads of CD- and DVD-ROM

drives. Several other III-V materials, such as InP, are useful in niche markets but play

significantly lesser commercial roles than GaAs.

Wide bandgap semiconductors are--in many ways--a progression beyond GaAs,

just as GaAs evolved to overcome some of the limitations of Si. Wide bandgap materials








simultaneously provide attractive properties for electronic devices as well as the potential

for short wavelength (visible to deep UV) light emission and detection. The Group III-

nitrides (AIN, GaN, InN, and their alloys) were initially researched for their promise to

fill the void for a blue solid state light emitter. GaN, in particular, received considerable

attention because of its direct energy gap of 3.39 eV (365 nm).


- 0 1

0
O


diamond
diamond


Direct Gap
Indirect Gap


\1 ZnS
Sa-GaN UV
.... ..... .... \ .. . ....... .1 . . . . . . . . .
-SiC(4H) \\ZnSe
SiC(3C AlAs *CdS
GaP C
CdSe
.. .. .. a-InN ...... GaAs "I AISb
IR Si o InP
Ge o \GaSb
InAs, InSb


4.0 5.0

Lattice Constant (A)


3.0


6.0


VIOLET



--RED




7.0


Figure 1-2. Energy bandgaps of several III-V, II-VI, and elemental semiconductors as a
function of lattice constant. Note the limits of the visible portion of the electromagnetic
spectrum.


The energy gaps of several common semiconductors are given in Figure 1-2 as a

function of lattice constant. The approximate boundaries of the visible spectrum are


a-AIN


2.0


I II








shown as is the nature of the energy transition (direct or indirect) for each material. From

the figure, it is noted that most III-V materials have bandgaps from red to infrared.

Certain alloys, such as AlInGaP, are important for the 1.3 and 1.55 pm

telecommunications wavelengths, but none of the 'traditional' III-V materials can reach

the shorter green and blue wavelengths. InGaN/GaN-based blue LEDs and LDs have

recently been demonstrated and commercialized by Nichia Corporation. These light-

emitting devices remain an active area of worldwide research.

Electronic devices from III-nitrides have been a more recent phenomenon.

Although their potential has been realized for several decades, wide band gap

semiconductor electronic research is only beginning to flourish. The large energy gaps

responsible for short wavelength light emission in these materials also give rise to

extremely low thermal carrier generation rates. Figure 1-3 shows the dramatic decrease

in intrinsic carrier concentration (ni) for GaN relative to Si and GaAs. At room

temperature, ni for GaN is approximately 20 orders of magnitude lower than that of Si,

leading to significantly reduced leakage currents. This is further illustrated in Figure 1-4,

where the intrinsic temperature (Ti) is given as a function of semiconductor impurity

concentration. The intrinsic temperature is the temperature at which the concentration of

thermally generated intrinsic carriers equals the concentration of electrically active

impurities in the semiconductor (intentional + unintentional doping). This value can be

used as a metric of the high temperature operation limit for a given semiconductor. As

illustrated, GaN electronic devices are able to operate without cooling at temperatures

well in excess of the limits of Si or GaAs. This not only reduces the cost of a given












400 300 200


Temperature (C)
100


1020

E
C.) o
. 10



10-10
1,


u 10-1


Si


GaAs





GaN





1.5 2.0 2.5 3.0 3.5 4.0
1.5 2.0 2.5 3.0 3.5 4.0


1000 /T (K"')


Figure 1-3. Semilog plot of intrinsic carrier concentration versus inverse temperature for
Si, GaAs, and GaN.


1010 10" 1012 1013 1014 101 1016

Impurity Concentration (cm"3)


Figure 1-4. Intrinsic temperature of Si, GaAs, and GaN as a function of semiconductor
impurity concentration.


25 0








subsystem, but also makes bulky cooling equipment unnecessary, decreasing weight and

size.

The thermal conductivity of GaN is three times that of GaAs, as listed in Table

1-1. For high power or high temperature applications, good thermal conductivity is

imperative for heat removal or sustained operation at elevated temperatures. The

development of III-N and other wide bandgap technologies for high temperature

applications will likely take place at the expense of competing technologies, such as

silicon-on-insulator (SOI), at moderate temperatures [4]. At higher temperatures

(>300C), novel devices and components will become possible. The automotive industry

will likely be one of the largest markets for such high temperature electronics.

Automotive control components fabricated with wide bandgap materials could be

mounted directly to an engine block, reducing signal delay. Well logging equipmen-t,

military combat systems, and aerospace components are just a few of the additional

sectors in the global market for these devices. High temperature applications promise to

play an increasingly important role for electronic devices of the future, with a total world

market projected to reach over $1 billion by 2005 [4].

One of the most noteworthy advantages for III-N materials over other wide

bandgap semiconductors is the availability of AlGaN/GaN and InGaN/GaN

heterostructures. A 2-dimensional electron gas (2DEG) has been shown to exist at the

AlGaN/GaN interface, and heterostructure field effect transistors (HFETs) from these

materials can exhibit 2DEG mobilities approaching 2000 cm2 / V*s at 300K. Combined

with the polarization-enhanced sheet charge density of- 1013 cm"2 and large peak and

saturation velocities, AlGaN/GaN HFETs are capable of handling large DC and RF












Table 1-1. Selected 300 K material properties relevant to electronic device applications
for Si, GaAs, and wide bandgap semiconductors [5-7]
Property Si GaAs 4H-SiC Diamond InN a-GaN/ AIN
(AlGaN/GaN)
Energy Bandgap, Eg 1.12 1.42 3.25 5.45 1.89 3.4 6.2
(eV)
Saturated Electron Velocity, v,, 1.0 1.0 2.0 2.7 2.5 2.5 1.4
(xl07 cm/s) (2.7)
Peak Electron Velocity, v.m 2.1 4.3 3.1 1.7
(x107 cm/s)

Breakdown Field, EB 0.3 0.6 3 10 3 -
(MV/cm)

Electron Mobility, p, 1400 8000 800 2200 3200 900 135
(cm2/ V-s) (2000)

Hole Mobility, tp 500 400 50 1600 50 14
(cm2 / V.s)

Static Dielectric Constant, e, 11.8 12.8 9.7 5.5 15.3 9.5 8.5

Thermal Conductivity, K
(W/cm.K) 1.5 0.5 4.9 20-30 1.3 2.85



currents. HFETs have been by far the most heavily investigated III-N electronic device,

largely due to their promise in the radio frequency semiconductor market. Power

handling capabilities of -12 W/mm appear feasible, and extraordinary large signal

performance has already been demonstrated, with a current state-of-the-art of> 10W/mm

at X-band and 6.6 W/mm at 20 GHz for AlGaN/GaN HEMTs on SiC substrates. These

values are already far superior to the best power density achieved with GaAs devices,

which is on the order of 1.5 W/mm at X-band [8]. The microwave market for III-N

devices is expected to lie at moderate frequencies (X-band to Ku-band) and RF power

levels greater than can be achieved with competing technologies. At higher frequencies








(and lower powers) materials such as InP are expected to dominate because of an

extremely low electron effective mass.

Since the critical electric field for avalanche breakdown increases with the

semiconductor bandgap, the predicted breakdown field of GaN is very high (-3 x 106

V/cm). This presents exciting possibilities for devices required to block large voltages,

such as switches found in electrical transmission and distribution systems. Traditionally

these devices are Si-based, with limitations on the maximum voltage ratings and current

handling capabilities [9]. Wide bandgap solid state devices are expected to provide

"cleaner" switching, without the voltage or inductance spikes typically associated with

mechanical switches or p-i-n rectifiers. This will, in turn, allow the electric power grid to

operate closer to its rated value, increasing efficiency. Due to the large power dissipation

associated with switching, thermal management becomes a central issue in design and

implementation of such devices. In this regard, III-nitrides also offer advantages because

of their ability to operate at elevated temperatures without the need for external cooling.

A major short-term target for wide bandgap switches is 13.8 kV, a common residential

distribution mode. Future goals are 25 kV blocking voltage, 2 kA forward current, on-

resistance < 2% of the rated voltage, and switching speeds of ~50 kHz [10]. Other

applications include control circuitry for electronic motor drives, hybrid-electric

automobiles and military vehicles, and more-electric airplanes and nautical vessels.

To summarize many of the attractive features of the Ill-nitrides, selected material

properties of AIN, GaN, and InN relevant to electronic device applications are listed in

Table 1-1, along with those of Si, GaAs, SiC, and diamond. Properties of InN are

included for reference, although this material has not yet been applied extensively in








electronic devices. SiC and diamond are the primary competitors of the III-nitrides for

many of the electronic device applications previously mentioned in this chapter.

Currently, SiC technology is much more mature than that of GaN, with some RF devices

already having reached the market [11]. Power devices from SiC will almost certainly

make the first progress toward the commercial importance of wide bandgap electronics.

However, GaN offers a larger breakdown field than SiC, as well as the greatly increased

mobility afforded by heterostructures. It is believed that III-N devices will ultimately

provide performance superior to that achievable with SiC. Diamond offers an extremely

large bandgap and near-ideal material properties, but lags well behind due to difficulties

associated with crystal growth and n-type doping.

To place the tremendous potential of this material system into proper perspective,

Strategies Unlimited, a market research firm, recently released a market forecast

predicting an annual growth rate of 28% for GaN electronic and optoelectronic devices

for the period 1999 2009 [12]. Annual revenue is expected to skyrocket from $420

million in 1999 to >$4.8 billion by 2009. A majority of this revenue will stem from LED

and LD sales, since the III-N electronics market will not begin until 2002. However,

after 2002 the electronics sector is predicted to grow nearly 100% annually.


1.2 Dissertation Outline

This work deals exclusively with III-N electronic devices, with the only

discussion of optoelectronic or photonic devices occurring in the background information

of Chapter 2.

Chapter 2 reviews the technological progression of III-N materials and devices,

with particular emphasis on past challenges, key processing issues, and current status of








understanding and development. Chapter 3 presents details of device fabrication and

performance of AlGaN/GaN HEMTs on sapphire and SiC substrates. DC, small signal,

and large signal characteristics are given. Chapter 4 deals with thin film growth and

characterization of novel gate oxides for III-N MOS applications. The capacitance-

voltage technique is used extensively to investigate the bulk and interfacial electrical

characteristics of the gate dielectrics. This is followed in Chapter 5 by a demonstration of

Gd203 as the gate insulator of GaN depletion mode MOSFETs. Chapter 6 discusses the

fabrication and device performance of vertically-depleting Schottky diodes on high

quality free-standing GaN templates. Chapter 7 gives a brief summary, concluding

remarks, and suggestions for future work.













CHAPTER 2
BACKGROUND AND LITERATURE REVIEW

Group III-nitrides have been the most heavily researched compound

semiconductor material system of the past decade. Entire conferences or symposia [13],

books [14-16], and numerous review papers [17-20] have recently been devoted

exclusively to the III-nitrides. The promise of these materials, particularly GaN, for short

wavelength light emission was well understood more than 30 years ago. However, the

III-nitrides have proven to be very different from 'traditional' III-V semiconductors, and

several challenges unique to (Al,Ga,In)N were addressed before useful devices could be

realized. Presently, the nitrides remain a relatively immature technology compared to

other III-V semiconductors such as GaAs or InP. This chapter will introduce the

pertinent chemical and physical properties of III-nitrides, chronicle their technical

development, and present the current state of material growth, processing, and

optoelectronic and electronic devices.


2.1 Historical Review

Early reports of GaN synthesis in various forms were given by Johnson et al. [21],

Juza and Hahn [22], Grimeiss and Kowlmans [23], and Lorenz and Binkowski [24]. The

resulting material was usually small in size and polycrystalline. Marushka and Tietjen in

1969 [25] were the first to report thin film growth of GaN over large areas (-2 cm2

sapphire substrates). They used a hydride vapor phase technique with pure HCI and NH3

gases and a H2 carrier. The HCI reacted with metallic Ga to transport Ga-subchlorides to








the substrate for reaction with NH3, forming electronic quality hexagonal GaN. Room

temperature optical absorption measurements correctly estimated the GaN direct bandgap

to be 3.39 eV at room temperature. All as-grown films were characterized by strong

n-type conductivity, which was speculated to be caused by nitrogen vacancies. Initial

attempts at p-type doping with Zn, Mg, Hg, Si, and Ge proved unsuccessful, and n-type

behavior was retained in all cases.

Although the inability to produce p-type doping in GaN would continue for nearly

2 decades, successful light emitting devices were demonstrated by Pankove et al. in the

early 1970s using metal-(Zn-doped insulator)-(n-type semiconductor) structures [26].

Hot electron injection from the n-type GaN layer was speculated to cause impact

ionization of luminescent centers in the insulating film. These devices could be tuned

over a broad spectral range by varying the Zn content in the cathode. Other reports .

demonstrated luminescence from GaN/GaN:Mg and metal-Si3N4-GaN structures [27,28].

However, since these were MIS structures, very large electric fields were necessary to

produce radiative recombination, leading to devices much less efficient than conventional

p-n junction light emitters. Deep levels often introduced a yellow component (~550 nm)

to the light output, making the emission a greenish-blue color.

Despite the initial flurry of activity associated with the results of Marushka and

Pankove, the difficulty in achieving reproducible p-type doping led many groups to

abandon research in III-nitrides completely. Furthermore, the material quality was

generally very poor and adequate Ohmic contacts had not yet been developed for GaN.

Wide bandgap II-VI materials such as ZnS and ZnSe moved to the forefront of blue,

green, and UV optoelectronic device research. Much of the success achieved by today's









III-N researchers is due to the work of Professor I. Akasaki at Nayoga University in

Japan. Their group was one of the few that persisted with wide bandgap nitride research,

and in the 1980s made two profoundly important contributions. The first of these was the

development of a low temperature (LT) AIN buffer layer for GaN growth by

metalorganic chemical vapor deposition (MOCVD). The resulting 2-step growth

technique (Figure 2-1) greatly improved GaN epilayer quality, despite the 3-dimensional

island growth of the AIN buffer films. The large nucleation surface area was shown to

enhance lateral growth of the subsequent GaN, improving crystal quality, increasing

uniformity, enhancing luminescence, and decreasing the background doping to ~1017

cm"3. The success of the AIN buffer is highlighted by noting that some variation of this

technique is still employed by most modem III-N crystal growers.






I
>1000 C _______



^ /High High Quality
500 *C Temperature GaN Epilayer Growth
PBuffer
Low Temperature Temperature Anneal
AIN Buffer Growth Ramp

Time

Figure 2-1. Temperature profile during MOCVD growth illustrating use of LT-AIN
strain relief layer [29].



The donorlike nature of native defects and the lack of a shallow acceptor

precluded the growth of reproducible p-type material until 1988, nearly 20 years after the

first demonstration of thin GaN films. The "discovery" of p-type conductivity was made








somewhat accidentally by Amano and co-workers [30] while investigating

cathodoluminescence of MOCVD-grown GaN:Mg in a scanning electron microscope

(SEM). Their as-grown films were highly resistive, but showed a dramatic increase in

luminescence efficiency and decrease in resistivity after exposure to the low energy

electron beam (a technique they called low energy electron beam irradiation, or LEEBI).

Post-LEEBI Hall measurements indicated distinct p-type behavior in these films to a

depth corresponding to the penetration depth of the electron beam. With the availability

of high quality epilayers and both n- and p-type material, the first p-n junction LED was

demonstrated in 1989 [30]. This device employed 5000 A of LEEBI p-GaN grown on 3

Iunm of UID (2 x 1017 cmnf3) n-GaN with aluminum top and side contacts. The p-n junction

device showed a reduced turn-on voltage and a much greater near-bandedge

electroluminescence efficiency compared to similar devices fabricated with insulating

GaN:Mg layers.

In the 1990s, Nakamura and colleagues at Nichia Chemical Company advanced

the III-N light-emitting technology at a pace unmatched before or since by any academic,

industrial, or research group in the world. Although Akasaki's LEEBI technique was

successful in activating Mg-doped GaN, Nakamura et al. demonstrated that the same

effect could be realized by thermal annealing at temperatures >750C in a hydrogen-free

ambient-a much simpler process [31]. Similar anneals in NH3 showed the activation to

be reversible, confirming the role of hydrogen as the passivating agent. In 1994, Nichia

became the first company to take III-N LEDs to the commercial market, later followed by

Cree and Hewlett Packard. The Nichia devices had remarkably high brightness despite a

dislocation density >1010 cm"3 in the active region. Such a dislocation density would








easily preclude any operational light emitting devices from conventional III-V materials.

The mechanism for optical emission from such highly defective III-N material has still

not yet been explained in detail.

Nitride laser diodes (LDs) were also pioneered at Nichia by leveraging the

experience base attained during the development of III-N LEDs. These devices always

utilized an InGaN active layer due to certain difficulties associated with GaN active

region devices. Since Nakamura et al. demonstrated the first pulsed operation of a III-N

LD in 1996 [32], several other groups have reported fabrication of similar devices.

Again, commercialization was first achieved by Nichia. Continuous wave (CW)

operation has since been demonstrated, with recent estimated room temperature CW

operation of >10,000 hours at an output power of 2 mW [33]. It is interesting to note

that the Nichia structures are normally grown on a GaN buffer, unlike the LT buffer

proposed by Akasaki and co-workers.

The most recent evolution of III-N technology has been its application to

electronic devices. The amount of effort currently devoted to research-level exploration

of III-N electronics is staggering. Demonstrations of GaN-based BJTs, HBTs, Schottky

rectifiers, MESFETs, JFETs, HFETs, and MISFETs have appeared in the literature.

Although there are currently few (if any) commercially available products, this will

undoubtedly change in the near future. Silicon carbide MESFETs from Cree Research,

Inc. are beginning to infiltrate the low frequency RF market with 12 W output power and

11 dB gain at 48 V and -3 GHz. It is expected that III-N devices will eventually surpass

the power handling capabilities of SiC, and extend the frequency of high power solid-

state RF modules to Ku-band or beyond.








2.2 Crystal Structure

The III-nitrides can exist in a hexagonal (wurtzite) or a cubic (zincblende)

polytype, each shown in Figure 2-2. The wurtzite structure (a-GaN) consists of

interpenetrating hexagonal close packed (HCP) sublattices, each consisting solely or

gallium or nitrogen atoms. Likewise, the zincblende crystal structure (P-GaN) is formed

by interpenetrating body-centered cubic (BCC) cation and anion sublattices. The

bandgap of P-GaN is slightly smaller than that of hexagonal GaN. Wurtzite GaN is by

far the most common of the 2 polytypes, although the metastable zincblende phase has

been grown on certain cubic substrates [34]. Zincblende AIN has also been reported

from growth on P-SiC [35]. All subsequent discussions of GaN, AIN, or AlxGail.-xN

layers in this work specifically refer to wurtzite material due to the profusion of this

polytype compared to p-GaN.


Table 2-1. Material Properties of AIN, GaN, and InN [14]


GaN A1N InN
Wurtz. Zinc. Wurtz. Zinc. Wurtz. Zinc.
Lattice Constant a-= 3.189 3 a=3.112 a= 3.54
(A) c= 5.185 4.52 c= 4.982 438 c= 5.760 498

Energy Gap (eV) 3.39 3.2 6.28 5.11* 1.89 2.2*
* theoretical










a) b)














Figure 2-2. Crystal structures of III-nitrides: a) zincblende and b) wurtzite. The
hexagonal wurtzite structure is the most common for electronic quality (AI,Galn)N.


2.3 Thin Film Growth

2.3.1 Substrates for 111-N Epitaxy

Without question, the factor most responsible for hindering device development is

the unavailability of an appropriate substrate material with thermal compatibility and a

close lattice match to GaN. The lattice constants of the III-nitrides are, in general, much

smaller than typical III-V semiconductors such as phosphides or arsenides, as illustrated

in Figure 1-2 and Table 2-2. The hexagonal crystal structure ofa-GaN serves to

exacerbate this problem. The ideal substrate would be GaN itself, cut from a bulk

sample, which could be used for homoepitaxy with perfect lattice and thermal expansion

matching. However, bulk GaN growth is difficult to achieve in large scale due to the











Table 2-2. Material properties of GaN and candidate substrate materials for
heteroepitaxial III-N growth.
Coeff. of Thermal
Lattice Parameters
Material Crystal Structure Lattice Parameters Expansion
(A) (Aa/a, Ac/c x 106 K')


Hexagonal


Cubic
Hexagonal


Hexagonal


Hexagonal


Cubic
Hexagonal


Cubic
Hexagonal


Cubic
Cubic
Tetragonal


Orthorhombic


a-GaN


5.59
3.17
5.2
7.5
8.5



4.2
4.7


4.2
5.3
7.45
2.9
4.8
3.59
6.0
7.1
7.5
6.0
7.0


a = 3.189
c= 5.185
4.52
a =4.758
c= 12.991
a 3.08
c = 10.07
a 3.08
c = 15.12
4.3
a=3.104
c = 4.966
8.083
a =3.2496
c = 5.2065
5.4301
5.6533
a =5.1687
c = 6.2679
a =5.402
b = 6.372
c = 65.407


3-GaN
A1203


4H-SiC


6H-SiC


3C-SiC
A1N


MgAl204
ZnO


Si
GaAs
LiAIO2


LiGaO2








high dissociation pressure of nitrogen and necessary growth temperatures from 1200-

1600C. Sapphire (0001) is widely used as the substrate for heteroepitaxial GaN growth

due to its relatively low cost, availability in large diameters, and stability at typical

process temperatures [36]. However, GaN growth on A1203 suffers from a lattice

mismatch of about 14%, resulting in a high density (10-10 cm"2) of threading

dislocations in the epitaxial films, as shown in Figure 2-3. Also, the low thermal

conductivity of sapphire (K = 0.5 W/cm-K) is highly undesirable for devices operating at

elevated temperatures since heat cannot be effectively dissipated through the substrate.

Alternative substrate materials such as Si, GaAs, InP, GaP, SiC, ZnO, MgO,

MgAI204, NdAl204, NdGaO3, ScAlMgO4, LiGaO2, and LiAO102 have been investigated

[37-51], with varying levels of success. The lattice and thermal mismatch data for

several of these potential substrate materials are summarized in Table 2-2. Of these, SiC

has shown perhaps the most promise for electronic device applications. SiC, itself a wide

bandgap semiconductor, has a better lattice match to GaN than sapphire (0001). In

addition, SiC is an exceptional thermal conductor (K 3.3 4.9 W/cm-K), with the exact

K value depending on the polytype and doping. The most common SiC polytypes for

nitride growth are 4H and 6H. N-type, p-type, and semi-insulating substrates are

commercially available, with diameters up to 3 inches.

LiGaO2 (001) and LiA102 (100) are intriguing substrate candidates due to their

small lattice mismatch to GaN. LiGaO2 (LGO) is an orthorhombic crystal with

hexagonal mismatch of
between (1120) GaN and (100) LAO [48]. Although these are promising features, very

little work has yet been done to characterize nitride growth on these materials.








Bulk GaN growth by Czochralski or Bridgeman techniques from stoichiometric

melts has been largely unsuccessful due to the very high melting temperature and high

vapor pressure of N2 above GaN. Such growths have been conducted with a specialized

growth apparatus at temperatures of~1600C and nitrogen overpressures of-15 kbar

[52,53]. Small single crystals of GaN with dislocation densities -105 cm"3 have been

produced, but the dimensions of the largest samples are still < 1 cm2. The high

background doping in these samples has been decreased by Mg incorporation.

Interestingly, this has also been shown to improve crystal quality as measured by x-ray

diffraction [54]. The fundamental drawbacks to this technique are cost and sample size.

Both issues must be addressed for this method to gain widespread use for bulk Ifl-N

growth.

An interesting approach to bulk GaN growth involves hydride vapor-phase

epitaxy (HVPE) of GaN on A1203 and post-growth removal of the sapphire to create a

free-standing quasi-substrate [55-57]. This approach was used to grow bulk templates

described in greater detail in Chapter 6. For near-term availability of bulk GaN samples,

such VPE techniques appear promising. However, hydride growth techniques are

typically characterized by higher background impurity concentration and lower crystal

quality than films grown by MOCVD or MBE. Both of these aspects will need to be

improved before device-quality material is produced by this technique. Recent reports

[58,59], as well as the results presented in Chapter 6 of this work, indicate that the quality

of HVPE nitride layers is continually progressing.

It is far from clear if there is a specific "substrate of the future" for II-N

electronic and optoelectronic devices. Much work remains for those involved in the







crystal growth of these materials. Many creative approaches to dislocation reduction,
including epitaxial lateral overgrowth (ELO), pendeoepitaxy, and cantilever epitaxy, have
been demonstrated in the literature [60-68]. In GaN ELO, a patterned SiO2 mask is used
to regrow GaN material in regions above oxide window. The mechanism of coalescence
of the overgrown nitride is the topic of active research [69], and is a function of the mask
pattern size and shape as well as growth conditions. Threading dislocations tend to
propagate to the surface of the overgrown nitride in regions above the window, but are
greatly reduced in regions above the mask. Upon convergence of the lateral growth
fronts, material of improved electrical and optical quality can be obtained.


.j


Figure 2-3. Transmission electron microscopy (TEM) image of GaN film on sapphire
illustrating high density of threading and mixed dislocations.








Growth by conventional or specialized techniques on common, non-lattice

matched substrates (e.g. silicon) should not be discounted, since this would decrease

substrate cost, increase substrate size, leverage the silicon processing technology, and

allow direct integration of nitride devices with silicon ICs or other components.

2.3.2 Doping

The impetus for most of the early work in GaN was its potential as a short

wavelength light-emitter. Typically, semiconductor light-emitting devices rely on a

forward-biased p-n junction to create radiative electron-hole recombination, resulting in

the emission of a photon. Obviously, for such devices to be fabricated, both n- and p-

type conductivity must be possible from the host material. Bipolar electronic devices,

such as bipolar junction transistors (BJTs) or heterostructure bipolar transistors (HBTs)

also require (by definition) both n- and p-type layers.

In the earliest work on electronic quality GaN thin films, Marushka and Tietjen

noted strong n-type conductivity, with electron concentrations from 1 5 x 1019 cm"3

[25]. They speculated that nitrogen vacancies were probably responsible for the

unintentional doping. Later reports challenged this hypothesis, suggesting that the

nitrogen vacancy (while donorlike) possesses a rather high energy of formation,

precluding a sufficient concentration to account for the observed n-type conductivity

[70]. Gallium interstitials [71] as well as silicon substituting on a gallium site [72] and

oxygen substituting on a nitrogen site [72,73] have been suggested to play a role in this

behavior. Improvements in crystal quality and decreased impurity incorporation have led

to material with greatly reduced background doping.








Silicon doping has been demonstrated to provide reproducible intentionally n-type

films. The solubility of Si in GaN is excellent, various MOCVD precursors are readily

available, and films with electron concentrations from -10 (UID) to ~102 cm3 have

been demonstrated [74-76]. The Si concentration in MOCVD-grown films appears to be

proportional to the Si/Ga ratio of the source gases [77]. The activation energy for this

shallow donor is still under investigation, but appears to be in the range 10- 30 meV

[77,78]. Germanium doping with GeHI4 has also been investigated [79], but the

incorporation efficiency is much lower than for silicon.

P-type doping has proven especially difficult for the III-nitrides, partially due the

donorlike nature of native defects. Divalent species such as Zn, Cd, Mg, Be, and Ca [80-

83], as well as Group IV elements such as carbon [84], have been investigated as

potential p-dopants. Many of these effectively compensate the background electron

concentration, but low resistivity p-type behavior is usually not obtained. It has been

speculated that the strong ionic nature of the GaN crystal plays some role in the difficulty

with p-doping. As with ionic wide bandgap II-VI semiconductors such as ZnS, valence

band hybridization effects may lead to repulsion of certain acceptor species [85].

Magnesium has been the shallowest acceptor level and the most successful p-type

dopant in GaN. However, the properties of p-GaN are still far from ideal, and achieving

high hole concentrations with Mg as the dopant has proven difficult. The Mg acceptor is

located rather deep in the energy gap, approximately 170 meV above the valence band

edge. This leads to films with only ~1% of the incorporated acceptors actually

contributing to the p-type conduction (i.e. Mg incorporation needs to be approximately 2

orders of magnitude higher than the desired hole concentration). This effectively places








an upper limit on the obtainable hole concentrations in Mg-doped films. For highly

doped films, the Mg content can approach 1% of the concentration of the host species,

well above the typical dopant content of other III-V semiconductors. In addition, Mg-

doped GaN films are extremely sensitive to hydrogen exposure. It was shown that the

LEEBI technique resulted in the dissociation of hydrogen from Mg:H complexes in the

as-grown material [86]. Without the compensating hydrogen, the Mg acceptors are able

to contribute to the hole conduction. Nakamura, using activation by thermal annealing,

demonstrated the reversibility of this process by re-compensating GaN:Mg films upon

annealing in hydrogen ambients [87].

Aside from the deep acceptor level, one problem specific to magnesium doping

by MOCVD is the "memory" effect. This effect is characterized by the inability to

abruptly "turn off" the amount of Mg present in the vicinity of the growing film. In

MOCVD growth, it is typically possible to abruptly increase or decrease doping levels by

modulating the flow of the dopant source to the chamber. This allows the crystal grower

to accurately place p-n junctions, S-doped layers, or other features required by the device

structure. In the case of MOCVD Mg-doping, cyclopentadienly magnesium (Cp2Mg) is

typically used as the Mg source. Cp2Mg is a low vapor pressure crystalline solid at room

temperature, complicating its transport to the reaction chamber. Turn-on is usually

abrupt, but the turn-off is not well controlled, as illustrated in Figure 2-4. This is due to

the relatively low vapor pressure of Mg, making it easily "stick" to chamber walls during

growth. When Mg is desorbed from the chamber walls, it can be incorporated into the

film despite the fact that the Mg source may have already been turned off. The memory

effect is particularly troublesome in the growth of p-n heterostructures, such as for









AlGaN/GaN heterojunction bipolar transistors (HBTs). For an npn HBT, the memory

effect can place the p-n junction in the wide bandgap emitter, instead of at the

AlGaN/GaN interface. The resulting emitter homojunction eliminates the advantages

associated with the conduction band offset of npn HBTs. Growth techniques such as RF-

MBE are attractive alternatives for device structures where accurate p-n junction

placement is critical.






Snon-abrupt
\ turn-off:
"menmoy effect"





Mg Mg Tme
source source
on off

Figure 2-4. "Memory" effect of Mg-dopant during MOCVD growth of p-GaN.



Highly doped p-type films are necessary for Ohmic contact layers on devices such

as LEDs, LDs, and HBTs. Although Mg doping has allowed the fabrication of

commercially viable p-n light emitting III-N devices, the poor Ohmic contacts that result

from low concentrations of activated Mg acceptors has been perhaps the most limiting

factor in the development of III-N bipolar electronic devices. Continued efforts to obtain

a shallow acceptor level, increase Mg activation, or to demonstrate novel techniques such

as superlattice doping or co-doping will hopefully help overcome many of the limitations

currently associated with p-GaN films.
currently associated with p-CGaN films.








2.4 Etching

A commonly cited advantageous property of the III-N material system is chemical

stability. The ability of the nitrides to withstand harsh chemical environments is

attractive, for example, in many military applications. This property is due to the large

bond strength of the nitrides, which is approximately 11.5, 8.9,and 7.7 eV/atom for AIN,

GaN, and InN, respectively [9]. Unfortunately, this chemical stability also greatly

complicates wet etching. The most systematic study of wet chemical etching was

reported in 1997 [88]. In this work, the etch behavior of-30 potential etchants (acidic

and basic) was investigated at room temperature for AIN, GaN, InN, InAlN, and InGaN.

No wet etch was found for GaN, InN, AIGaN, or InGaN. It was shown that KOH-based

solutions were effective at etching AIN, with etch rates strongly dependent on crystal

quality and solution temperature [89]. Other reports indicate a very slow GaN etch rate

in -50% NaOH/H2O (-20 A/min.) although these samples were grown on GaAs

substrates [90]. Molten KOH or H3P04 has been shown to preferentially etch defects and

dislocations in GaN [59,91-93]. While this is useful for materials characterization, it

does very little to assist with device processing, where smooth etched morphology and

materials selectivity are needed and the formation of etch pits is undesirable. The general

lack of a suitable wet etch for the III-nitrides has led to a great deal of work to develop

dry etch techniques for electronic and optoelectronic device processing steps such as

mesa steps, recess etches, and facet formation. Alternative methods, such as

photoenhanced wet chemical etching and laser ablation have also been investigated.

2.4.1 Photoenhanced Wet Chemical Etching

The dissolution kinetics of an etching process may be affected by illumination

with light above the bandgap of the semiconductor. This type of illumination creates








electron hole pairs (EHP), which can participate in the overall etch mechanism. The first

report of photoelectrochemical (PEC) wet etching of GaN was given by Minsky et al.

[94]. Their experimental setup consisted of an electrochemical cell with front contact to

the GaN surface and 325 nm illumination with a He-Cd laser. Dilute KOH and HC1

solutions were investigated, with etch rates as high as several pm/min. for KOH. The

authors speculated that the hole-assisted oxidation of the GaN surface contributed to the

enhanced etching behavior, in a manner analogous to GaAs [95]. Similar reports using

different etchants and illumination sources were given by other groups [96-98]. Youtsey

et al. extended the work of Minsky and postulated the overall oxidation reaction:

2GaN + 6 -- 2Ga+ + N2t (2.1)

where represents a photogenerated hole. They support this mechanism by noting the

presence of bubbles during the etch process, consistent with the release of nitrogen gas.

Cho et al. investigated the UV illumination intensity, KOH concentration, and solution

temperature during PEC etching [99]. From their results, a maximum was found in the

etch rate vs. KOH molarity curve for 450 W Hg lamp illumination. An activation energy

of 0.8 0.3 kcal/mol was extracted from an Arrhenius plot, indicating diffusion-limited

etching. Also, Auger electron spectroscopy (AES) indicated that the etched films

remained near-stoichiometric as compared to control samples.

Nearly all the reported PEC etch studies have investigated n-type GaN, with little

work on intrinsic or p-type material. For p-GaN, Youtsey et al. reported no

photoenhanced etching [100]. This is usually explained by the inability to confine holes

at the surface of the semiconductor. Yang et al. have reported effective photoenhanced

etching of p-GaN by applying a negative external bias to the substrate during etching.








The bias voltage effectively causes downward band bending at the semiconductor

surface, allowing hole accumulation and photoenhanced etching in a manner analogous to

n-type material.

Other recent work involves the development of PEC techniques that do not

require physical contact to an external electrochemical cell. Bardwell et al. have

described a process using S208', added to conventional KOH solutions [101,102]. The

peroxydisulfate ion absorbs photons with X < 310 nm, and produces highly oxidizing ions

or radicals, depending on the pH of the solution. Photogenerated electrons are necessary

to sustain the reaction, and for this reason dislocations are not readily etched, since these

act as electron-hole recombination centers. Nonetheless, various device structures have

been fabricated using this technique, including HFETs [103].

2.4.2 Dry Etching

No wet etch (conventional or PEC) has proven capable of fabricating smooth,

well-controlled, anisotropic features with sufficiently fast etch rates for III-N device

processing. In addition, while wet chemical etching has excellent selectivity for etching

one material over another, it is generally unacceptable for maintaining good pattern

transfer at dimensions below -2 prn. These difficulties have led to extensive research in

the development of III-N dry etch processing. Dry etch processes are characterized by

chemical or physical mechanisms, or by a combination of both. For chemical etches, the

reactant gases typically adsorb onto the semiconductor surface and undergo a surface

reaction. The volatile etch products desorb and are swept away by the vacuum system.

This type of etch is many ways analogous to wet etch processes, and material selectivity

is often possible by an appropriate choice of reactant gases. However, as with wet etches,








anisotropic profiles are difficult to attain. Physical etches occur when energetic ions

bombard the surface of the semiconductor and transfer enough energy to cause bond

breaking. The sample is physically sputtered away, making selectivity more difficult to

attain. Furthermore, such etches can cause significant damage to the semiconductor via

the generation of point defects or preferential loss of certain species from the surface.

There are 3 major dry etch techniques that have been explored for the nitrides:

reactive ion etching (RIE), electron cyclotron resonance etching (ECR), and inductively

coupled plasma etching (ICP). The fundamental difference between RIE, ECR, and ICP

etching is the method of plasma generation. RIE is the simplest of the 3, with a simple

13.56 MHz RF power applied between parallel electrodes. The substrate is placed

directly on the powered electrode, and energized ions from the plasma gas(es) are

accelerated toward the etching surface. Typical ion energies are in the hundreds of eV.

Anisotropic profiles can be obtained with RIE due to the physical component of the etch,

but significantly more etch damage can also occur, for the same reason. Another problem

with RIE is the low ion density (typically <109 cm"3) and the associated difficulty with

initial bond breaking in III-N materials [104].

High-density plasma sources such as ECR and ICP are attractive due to their

effective decoupling of plasma density and ion energy. The high plasma density (2-4

orders of magnitude higher than typical RIE) assists with bond breaking and etch product

desorption [105]. ECR plasmas are formed by passing current through permanent

magnets to induce a strong magnetic field and create electron resonance at 2.45 GHz.

Inductively coupled plasmas are generated by applying an RF bias to inductive coils

encircling a dielectric vessel. The alternating electric field creates a magnetic field in the








vertical plane and confines electrons in a high-density plasma near the center of the

chamber. In both ECR and ICP, an RF bias is superimposed on the sample electrode to

control ion energy. Low pressures are typically used to increase the mean free path and

enhance anisotropic etching. Both ECR and ICP platforms have been extensively

investigated for III-N etch development. Chlorine-based chemistries have been most

widely studied, due to the higher volatility of III-chlorides compared to other III-halides

and hydrocarbons.

The earliest work on ECR etching of III-nitrides was conducted by Pearton and

co-workers using BCI3, CC12F2, and CH4/H2 chemistries [106]. Reasonable etch rates

were achieved for A1N, GaN, and InN at low pressures and moderate bias levels. Higher

substrate bias was necessary for AIN in CC12F2, likely due to the formation of AIFx. For

the Cl-containing chemistries, the Group III etch products were determined to be the

respective III-chlorides, while the nitrogen was speculated to form NCl3 and CC1N.

Vartuli et al. investigated alternative ICI/Ar and IBr/Ar chemistries in both ECR

and ICP plasmas and reported very smooth etch morphologies and little preferential loss

of nitrogen, but with etch rates slower than those achieved with Cl-based chemistries

[107,108]. Selectivities of 5-10 for GaN over InN, AIN, or InAIN were reported for

ICl/Ar, with selectivities much lower for IBr/Ar.

The effects of chamber pressure and Ar and BC13 additions on C12 ICP etch

characteristics were studied by Lee and co-workers [109]. They used 600W ICP power,

120 V DC self-bias, 70C substrate temperature, and chamber pressures from 10-30

mtorr. The etch rate of GaN increased for Ar and BCI3 levels to up 10%, but decreased

for higher additive content. An increase in chlorine radical density was speculated as the








cause for the increased etch rates, which were as high as 8500 A/min with selectivity of

3.7 over SiO2.

At present, Cl-based ICP seems to have prevailed as the preferred high-density

III-N etch technique for reasons including easier scale-up, improved plasma uniformity

over large areas, lower cost and power requirements, and lack of electromagnets and

waveguides necessary with ECR [5,105]. Other techniques, including chemically-

assisted ion beam etching (CAIBE) [110], reactive ion beam etching (RIBE), and laser

ablation have been applied to the III-nitrides, with varying levels of success. In the near-

term, ICP etching will likely remain the enabling technology for pattern transfer during

III-N device processing. However, development of a simple, controllable wet etch with

reasonable etch rate could greatly decrease the utilization of dry etch techniques for the

nitrides.

Despite the success of high density plasma etch techniques, dry etch-induced

lattice damage can severely degrade material and device properties. In GaN, the damage

may occur as lattice defects, unintentional passivation, or preferential N2 loss producing

surface non-stoichiometry. This is an especially troublesome issue for III-N bipolar

electronic devices, such as BJTs and HBTs, where typically a dry etch step is required to

uncover the p-base. Etch damage can decrease the hole concentration or completely

reverse the surface conductivity type, depending on the extent of the damage. A reduced

physical component of the etch, consistent with lower RF chuck power (ion energy), is

needed to help minimize damage. For this reason, the low ion energy and increased

plasma density of ECR and ICP are more attractive than RIE for such steps.








2.5 Metallization

Metallization schemes, or contacts, are critical to the performance of

semiconductor devices. A good contact not only provides the desired electrical

characteristics, but must also be mechanically and thermally stable during device

processing and operating conditions. There are 2 main categories of metallization

schemes: Ohmic and rectifying (Schottky). Ohmic contacts allow electrons or holes to

flow freely into or out of the underlying semiconductor--i.e. they should present a very

low impedance to current flow. The I-V characteristics of an Ohmic contact are linear,

thereby obeying Ohm's law. Ideal Schottky contacts present a very low impedance when

forward biased beyond the turn-on voltage (Von) and present a very high impedance when

reverse biased. Therefore, the I-V curve for an ideal Schottky contact is very steep (large

slope) for V > Vo, but very flat (near-zero slope) for reverse bias values up to

breakdown. Reverse breakdown occurs at the onset of avalanche multiplication and is

characterized by an abrupt increase in reverse current flow. Reverse breakdown is

usually accompanied by irreversible damage to the device. The behavior of ideal Ohmic

and Schottky contacts is illustrated in Figure 2-5.

The wide bandgap of GaN makes contact formation somewhat challenging. To

understand the fundamental mechanisms involved with contact formation, it is necessary

to understand the band alignment of a metal/semiconductor junction, as shown in Figure

2-6. This figure is drawn for an n-type semiconductor and illustrates the band bending at

the semiconductor surface that arises when the metal and semiconductor are brought into

intimate contact and allowed to reach equilibrium (equality of Fermi levels, EF). For the

n-Ohmic contact, a metal with a small work function (omn) is needed, where the work








function is defined as the energy from the metal Fermi level to vacuum. Depending on

the value of
Schottky barrier in the semiconductor, leaving majority electrons free to pass through the

metal/semiconductor interface. An Ohmic contact is formed when the current-voltage

characteristics are linear. A good Ohmic contact has a resistance much smaller than the

intrinsic device resistance.

For the Schottky contact of Figure 2-6(b), a large metal work function induces

negative band bending and creates a conduction band energy barrier, +B, known as the

Schottky barrier height of the contact. For an unpinned semiconductor Fermi level

without image force lowering, the barrier height is given by [111]:

8 = 0. -X (2.2)

where X is the electron affinity of the semiconductor. This barrier height is typically on

the order of 1 eV. Hence, the current-voltage behavior of the ideal Schottky contact in

Figure 2-6(b) can easily be understood as follows: For forward bias conditions (positive

potential applied to the Schottky metal), electrons in the conduction band initially

experience an energy barrier preventing them from entering the metal contact. As the

forward bias voltage induces sufficient band bending to overcome the built-in potential of

the Schottky contact, electrons are free to move from the



















4 _________

0.0



-2.0



-4.0 i
-4 -2 0 2 4
Voltage (arb. units)





2.0 b.)



1.0



U 0.0

-. breakdown

-1.0 J
-20 -15 -10 -5 0 5
Voltage (arb. units)




Figure 2-5. Current-voltage characteristics for (a) Ohmic and (b) Schottky contact. The
Ohmic contact provides a low resistance path for electrons or holes to enter or exit the
semiconductor. The Schottky contact allows large forward currents, but blocks reverse
current flow up to the breakdown voltage.











a.)


e
<-- ^ - - ^ -t -


e- tunneling


_ _-.


Figure 2-6. Energy band diagrams for (a) n-Ohmic, (b) n-Schottky, and (c.) n+-Ohmic
contacts.








semiconductor to the metal. In the reverse bias regime, the band bending becomes more

pronounced as the applied bias becomes more negative, and no current flows as the

depletion region widens. Reverse breakdown occurs when avalanche multiplication takes

place in the depletion region, leading to the onset of large current flow.

When the semiconductor doping density is sufficiently high, the EF lies very near

the conduction band. In this case, equilibrium band bending takes place very near the

semiconductor surface, leading to "thin" energy barriers as illustrated in Figure 2-6(c).

Such structures allow quantum mechanical tunneling of electrons through the energy

barrier and into the metal, providing the basis for a potential low-resistance Ohmic

contact. This metallization strategy is common in AlGaAs/GaAs HEMTs, where a thin

cap layer of n+-GaAs is typically grown to facilitate Ohmic contact formation.

2.5.1 Ohmic Contacts

2.5.1.1 n-type

Aluminum was one of the first metals investigated for Ohmic contact formation to

n-GaN, due to its low work function (+m = 4.3 eV) and widespread use in semiconductor

contacts and interconnects. While Al produced Ohmic behavior [112] at room

temperature, a Ti/Al bilayer scheme was shown to provide superior contact resistance.

Lin et al. reported a study of Au, Al, Ti/Au, and Ti/Al layers deposited on MBE-grown

GaN and annealed in a temperature range from 500 900C [113]. A minimum specific

contact resistivity (pc) of 8 x l0-6 Q-cm2 was obtained for Ti/Al annealed at 900C for 30

sec. These results suggested several features of the Ti/Al contact scheme. First, the work

functions of Ti and Al are nearly the same, implying that a chemical mechanism is likely

responsible for the improved contact resistance of the Ti/Al bilayer as compared to the








single-layer Al contact. Possible explanations are the gettering of nitrogen from the near-

surface region of the semiconductor, removal of the native oxide, or a combination of

both. For the former, Jenkins and Dow have shown that N vacancies exhibit donorlike

behavior in GaN [114]. Thus, nitrogen gettering via TiNx formation would serve to

increase the electron concentration at the surface, thereby enhancing tunneling

probability. In the latter case, thin native oxide layers at the metal/semiconductor

interface have been shown to increase the contact resistance, since these oxides typically

act as insulators. The Ti may diffuse through this native oxide to make intimate contact

with the underlying GaN. Alternatively, the strong oxygen affinity of Ti may help pull

the oxide away from the semiconductor surface during the formation of a discontinuous

Ti- or mixed metal-oxide phase. The results of Lin and co-workers also suggest the

importance of Al in the Ti/Al bilayer, since the Ti/Al contact exhibited electrical

properties superior to the Ti/Au contact. This implies the formation of a Ti-Al

intermetallic phase important to the electrical behavior of the contact. It should be noted

that numerous metals other than Ti have used in bilayer schemes with Al, including Pd,

Ta, Nd, Sc, and Hf [9]. Ohmic behavior typically results, but none have produced contact

resistance values as low as Ti/Al. Numerical values obtained for annealed Ti/Al contacts

deposited on properly cleaned n-GaN surfaces are R1 0.5 f.mm and Pc 10-6 0 cm2.

Of course, these values depend on the doping density in the underlying semiconductor

and the device structure.

It is desirable to cap the Ti/Al contacts with an oxidation-resistant metal such as

Au to avoid excessive oxidation of the contact, especially during elevated temperature

operation. To prevent excessive reaction between the Au layer and the Ti/Al bilayer, a








diffusion barrier is also included. Therefore, most modem n-ohmic metallization

schemes employ annealed Ti/Al/x/Au, where x may be Ti, Ni, Pt, or Pd. However, the

explicit roles of the diffusion barrier and cap layers are still not well understood. Cross-

sectional transmission electron microscopy (XTEM) has shown the existence of Au at the

GaN surface of Ti/Al/Pt/Au contacts, suggesting that the Ni diffusion barrier is not

effective at blocking Au diffusion into the contact during annealing. In fact, Ni, Pd, and

Pt have been shown to be permeable to Au during typical contact processing [115]. Their

role in contact formation is not yet well-understood. Thermodynamic calculations may

also be helpful in determining the phases present after high temperature annealing.

One issue unique to III-N n-ohmic contact formation is the extremely high

annealing temperatures required to minimize contact resistance. Typical annealing

conditions are -800C, 30 sec. in a N2 ambient or overpressure. This very high annealing

temperature is well above the melting point of Al (661 C), and typically leads to rough

ohmic contact morphology. The edge definition of the patterned contact is also affected

by the roughening, and edge acuity may become an issue for very small device features,

such as the channel of an FET. A tradeoff between electrical properties and surface

morphology seems to exist, with smoother morphologies available at lower annealing

temperatures while sacrificing RP. The Ti/Al ratio may play some role in this tradeoff,

with reports of optimum Ti/Al ratios having been given in the literature [116]. TEM

investigations of Ti/Al/Ni/Au contacts to AlGaN have shown that the mixed phases

present at any given spatial position of the post-annealed contact are a strong function of

the choice of materials (i.e. diffusion barrier) and the thermal history. Thus, a

simultaneous optimization of Ti/Al thicknesses, Ti/Al ratios, diffusion barrier metal,








diffusion barrier thickness, annealing time, annealing temperature, and annealing ambient

must be considered with respect to electrical properties and contact morphology.

2.5.1.2 -tvpe

Low resistance ohmic contacts to p-GaN have proven much more difficult to

fabricate than those on n-type material. This difficulty is related to the wide bandgap of

GaN and the large ionization energy of the Mg acceptor. For p-type semiconductors, the

Fermi level is near the valence band, and the work function is much larger than for n-

type. Alignment of metal and semiconductor Fermi levels therefore requires a metal with

a larger 4m. Due to the wide energy bandgap (X = 4.1 eV and Eg = 3.4 eV), typical p-

GaN workfunctions are w 7 eV. There are no metal work functions this large, resulting in

an equilibrium valence band energy barrier. The lack of highly-doped p-type films

exacerbates this problem, since tunnel barriers are not available to increase hole

conduction. In addition to these inherent materials-related limitations, certain device

processing steps (such as dry etching) may cause preferential loss of N from the GaN

surface. This is of particular importance in devices such as BJTs and HBTs, where a dry

etch step is usually required to uncover thin p-base layers.

The most common p-ohmic metallization for GaN has been Ni/Au [117]. As with

n-type contacts, a plethora of contact schemes have been investigated. A partial list of

these includes Pd, Au, Ti, Al, Zn, Cu, Mo, Ta, Nb, Mo, V, W, Pt/Au [118], Ru/Au [119],

Nb/Au [120], Pt/Ni/Au [121], Ni/Pd/Au [122], Ti/Pt/Au [123], Ni/AuZn [124], Cr/AuZn,

[124], Pd/Mg/Pd/Au [125], and Pd/Ag/Au/Ti/Au [126]. Ohmic behavior has been

achieved, but p-contact resistance is still a limiting feature of both electronic and

optoelectronic devices. In many cases, slight rectifying behavior is observed. Even if the








contact displays linear I-V characteristics, exceedingly high contact resistance can prove

detrimental to device operation. For unipolar electronic devices such as FETs, the source

and drain contact resistances are parasitic extrinsic circuit elements which increase self-

heating and decrease the speed of the extrinsic device. For III-N photonic devices such

as laser diodes (LDs), the situation can be even worse. Due to the resistive local heating

and vertical nature of threading dislocations in GaN, thermally-enhanced migration of

contact metal along dislocation lines can lead to device short-circuiting. For bipolar

electronic devices such as npn HBTs, high base contact resistance (combined with the

high sheet resistance) severely limits the output power of the device. The vertical current

flow in these devices also makes them susceptible to the lifetime issues mentioned for

LDs.

Although the Ni/Au contact has been widely discussed in the open literature, the

mechanism ofp-GaN/Ni/Au Ohmic contact formation is still far from clear. Generally, it

is noted that annealing in an oxygen-containing ambient is necessary to achieve Ohmic

behavior, implying the importance of a NiO phase. Ho et al. proposed a favorable band

alignment between p-GaN and a thin interfacial layer of NiO [127]. Kiode and co-

workers suggested a different mechanism, in which the oxygen anneal getters hydrogen

and increases the hole concentration in the near-surface region of the semiconductor

[128]. Qiao et al. recently reported a thorough study relating the electrical characteristics

to the contact formation mechanism by depositing films ofNi, Ni/Au, Au, and NiO/Au

[129]. They found that NiO/Au contacts (formed by annealing a thin Ni layer in air at

500C for 5 min. then redepositing Au) did not produce Ohmic behavior, which seems to

disprove the favorable band alignment mechanism proposed by Ho. Kiode's hypothesis








of 0 gettering was supported by the work of Qiao et al., and it seems clear that oxygen

plays some role in layer reversal of the as-deposited Ni/Au contact. Rutherford

Backscattering Spectroscopy (RBS) data indicated a p-GaN/Au/NiO structure after

annealing in air. The Au layer at the interface is not likely a continuous phase, with NiO,

Ni-Ga-Au intermetallics and mixed oxides also present.

2.5.2 Rectifying Contacts

2.5.2.1 n-type

Rectifying contacts on n-GaN have also been extensively investigated, due to

their importance in Schottky diodes and as FET gates. Experimental Schottky barrier

heights show a clear dependence on metal work function, indicating an unpinned surface

Fermi level. This is illustrated in Figure 2-7, where it is clear that metals with large work

functions should be chosen as n-GaN Schottky contacts to maximize barrier height.

These data are in sharp contrast to other common III-V semiconductors such as GaAs and

InP, where the barrier height is largely independent of
states. The degree of scatter in the Figure 2-7 data is likely due to variable GaN epilayer

quality, since the growth and processing technology in the III-N material system is still in

its early stages. Factors potentially responsible for the wide range of +B reported in the

literature include surface roughness, surface preparation, local stoichiometry variations,

or thermal history [130]. The presence of high concentrations of defects or dislocations

has been shown to affect the I-V behavior of GaN Schottky diodes and rectifiers [131].

To date the largest barrier heights of -~l.1 eV have been realized with Pt (+m = 5.6

eV) contacts. Near-ideal Pt/GaN and Au/GaN Schottky diodes have been fabricated with









n = 1.04 [132] and 1.03 [133], respectively, defined from the forward current density

expression for a Schottky contact:


S=AT2 ex ([B exp( qVi)_1] (2.3)
P( )[ nk


where JF is the forward current density, A* is Richardson's constant, q is the electronic

charge, +B is the barrier height, V is the applied bias, n is the ideality factor, k is

Boltzmann's constant, and T is the absolute temperature. For an ideal Schottky contact, n

= 1. Reported values for GaN Schottky diodes range from near-ideal to > 4.


1.2
>"


S0.8



S0.4
I-"


0.0
4.0


> Ti
SAu
0 Ni
V Pt
A Al
o Ag
o Cr
+ Pd
X Sn


4.5 5.0 5.5
Metal Workfunction,

Figure 2-7. Schottky barrier height versus metal workfimunction for various contacts
reported in the literature. The trend of increasing surface Fermi level is unpinned in GaN.



Although Pt has demonstrated a slightly larger, more consistent barrier height on

n-GaN, many state-of-the-art AlGaN/GaN HEMTs use Ni/Au as the gate contact. Nickel








Schottky contacts exhibit comparable electrical properties, and it has been found that Ni

adheres slightly better to the (Al)GaN surface than Pt. The improved adhesion facilitates

easier lift-off of the patterned gates and improves yield.

2.5.2.2 p-type

There have been very few reports of rectifying contacts to p-GaN. The well-

documented difficulties in achieving reproducible p-type crystal growth and Ohmic

contact formation have precluded extensive study of p-type contacts and devices,

especially FETs.

2.5.3 Surface Treatment

Even with optimized metallization recipes, proper surface cleaning is imperative

to achieve intimate contact between metal and semiconductor. It is well known that the

GaN surface forms a thin Ga203 native oxide when exposed to air. The presence of

native oxides or residual organic residues can increase the contact resistance of Ohmic

contacts and can lead to nonidealities in rectifying contacts. In addition, an atomically

smooth, clean surface is imperative for devices requiring regrowth steps. Examples

include LEDs and LDs employing epitaxial lateral overgrowth (ELO), HBTs with

regrown emitter, or MISFETs. For all of these, the success of the metallization or

regrown epilayers is largely determined by the quality and cleanliness or the

semiconductor surface. A typical surface clean may include organic solvent treatments,

acid etches, and rinsing in deionized water. It has been shown that ex-situ HC1 and HF

are successful in removing a large amount of the surface oxygen and that a UV/ozone

clean can greatly decrease the amount of surface hydrocarbon [134]. The amount of

oxygen reduction is much less for AIN surfaces, presumably due to the larger bond

strength of the AI-O bond relative to Ga-O. This is of practical importance for Ohmic








contact formation to AlGaN/GaN HEMTs, where the surface layer typically contains

~20-30% Al. In-situ thermal or plasma treatments are also viable options for

contaminant desorption prior to crystal growth, but are less feasible with typical metal

deposition equipment

Although a minor impact of surface treatments on the electrical properties of GaN

contacts has been reported by some groups [135], other groups have reported dramatic

improvements in contact properties, particularly p-type, with sulfur-based treatments

[136]. Ammonium sulfide, (NH4)2S, has been credited with effective passivation of the

p-GaN surface and modification of the Fermi level, reducing the valence band barrier

height Such surface treatments have previously been applied to AlGaAs/GaAs

microdisk lasers [137]. Specific contact resistivities in the low- to mid-10"5 Y'cm2 have

been reported for (NH4)2S-treated p-GaN samples [138]. These results are noteworthy,

but at this point require further study.

2.6 III-N Optoelectronics

Direct bandgap semiconductors allow efficient radiative recombination of

electrons and holes because their conduction band minimum and valence band maximum

lie at the same position in momentum space (k-space). The photons emitted during a

recombination process have an energy equal to the bandgap of the semiconductor and a

wavelength X = Eg / (2z.h), where Eg is the semiconductor bandgap and h is Planck's

constant. III-nitrides are wide direct bandgap semiconductors with energy gaps spanning

the electromagnetic spectral range from visible (InN: 1.9 eV) to far-UV (AIN: 6.2 eV)

(Figure 1-2). The wide direct bandgaps lead to exciting possibilities for short wavelength

solid state light emitting devices, such as light emitting diodes (LEDs) and laser diodes








(LDs). The super-bright blue II-N LEDs can be incorporated with standard III-V red

and yellow emitters for flat panel displays with color mixing across the visible spectrum.

Traffic lights from LEDs deliver reduced power consumption and longer lifetimes than

conventional bulbs. Also, by introducing a phosphor material above the LED, blue light

from InGaN LEDs can be absorbed and re-emitted as white light. Such technologies

promise to challenge incandescent sources for residential and outdoor lighting

applications in the foreseeable future. The market for non-communication LEDs is

currently -$2.1 billion and is expected to reach over $3 billion in less than 2 years.

Analogously, non-communication laser diodes currently enjoy a $1 billion market, which

is expected to double in the next 2 years [139]. Communications, printing, and data

storage are important commercial technologies that will benefit from continued

development of III-N LDs. Current solid-state lasers for such applications are AlGaAs or

InGaAsP alloys, emitting in the infrared. Since the minimum spot size for reading or

writing data scales with X2, the resolution of laser printers and storage capacity of CDs or

DVDs would be greatly increased by UV-emitting III-N LDs. This market alone is

expected to generate $87 million in 2003. Other optoelectronic applications for III-

nitrides include solar-blind UV detectors, which take advantage of the wide bandgap of

AlGaN alloys in detectors sensitive to wavelengths less than 300 nm while insensitive to

visible and IR solar background. Such detectors are useful, for example, in military and

defense applications.

2.6.1. Light Emitting Diodes

The demonstration, development, and commercialization of blue and ultraviolet

p-n junction LEDs have all taken place within the past 12 years. This rapid progress is a









testament to the both the uniqueness of the material system and the massive research

efforts that have been devoted to it by various academic, corporate, and governmental

research labs. Although available from relatively few vendors, the GaN LED market has

grown from -$0 before 1994 to -$420 million in 1999 (Figure 2-8). III-N LEDs offer

tremendous advantages in output power, quantum efficiency, and luminous intensity over

competing materials systems such as SiC, GaP, or (Zn,Cd,Mg)(S,Se,Te). The indirect

bandgaps of SiC and GaP result in devices with significantly reduced brightness (-10 -

20 mcd) and quantum efficiency (-0.1%) compared to direct bandgap III-N devices.

Although green Il-VI LEDs have been fabricated with reasonable brightness, the short

lifetimes and propagation of crystal defects in II-VI semiconductors have precluded their

widespread commercialization.


500-


400- GaN LED Market Revenue
(1995- 1999)
o 300
E

200


100-


0-
1995 1996 1997 1998 1999
Year


Figure 2-8. GaN LED revenue for 5-year period ending 1999. The annual compound
growth rate is nearly 80%.








The first demonstration of a InGaN p-n junction LED was given in 1989 [30].

Shortly thereafter, Nakamura and co-workers at Nichia Chemical Co. reported an LED

emitting at 430 nm with a power level approximately one order of magnitude higher than

comparable SiC devices [140,141]. In effect, this began Nichia's dominance ofnitride

LED and LD research and development in the 1990s. The structure of a Nichia high-

brightness blue LED, first fabricated in 1993, is given in Figure 2-9(a). The device

contains an AlGaN/InGaN/AIGaN double heterostructure (DH), with 15% Al content in

the cladding layers. The as-grown sample was annealed at 700C in N2 to activate the

Mg acceptor. Ti/Al and Ni/Au metallizations were employed as n-ohmic and p-ohmic

metallizations, respectively. The Ino.06Gao.94N active layer was co-doped with Si and Zn

and resulted in emission near 450 nm, probably due to impurity-assisted recombination.

The FWHM at peak wavelength was -70 nm. For such devices, a blue shift of 10 15

nm was typically observed by increasing the forward current from 100 pA to 20 mA.

The output power and external quantum efficiency of these devices were 3 mW and 5.4%

at a forward current of 20 mA [142,143]. The brightness of the AlGaN/InGaN/AlGaN

blue DH LEDs was ~2.5 cd, comparable to red AlGaAs LEDs.

Quantum well structures were developed by Nakamura and co-workers to extend

the III-N LED wavelength range to >500 nm. In order to increase the emission

wavelength of InGaN LEDs, increased In incorporation is necessary in the active region

of the device. However, due to the lattice and thermal mismatch between InN and

(Al,Ga)N, as well as thermodynamic considerations, high quality InGaN epilayers with

high indium content have been difficult to achieve in practice. Thin InGaN quantum

layers (-30A) allow elastic strain accommodation between the well(s) and barriers,








reducing misfit dislocations due to strain relaxation. It is believed that such dislocations

in the active region play a role in reducing the quantum efficiency of light-emitting

devices, such as LEDs. Nakamura et al. demonstrated single quantum well (SQW) p-

AIGaN/InGaN/n-InGaN/n-AlGaN LEDs and demonstrated electroluminescence (EL)

from 450 nm (blue) to 590 nm (yellow). A difference of -170 meV was observed

between the peak emission wavelength and the unstrained value of the energy gap. This

difference was attributed to quantum size and exciton effects of the active layer, although

the details of the mechanism were not elucidated. To improve the output power and

spectral width of these devices, a new p-AlGaN/InGaN/n-GaN SQW structure was

investigated, as shown in Figure 2-9(b). Ultimately, this became the standard structure

for high-brightness LEDs from blue to orange. Typical peak wavelengths and FWHM of

the EL signal were 450 and 20 nm (blue) and 520 and 30 nm (green). Although the

FWHM becomes larger at longer wavelengths, these devices offer much narrower

emission bands than previously reported DH LEDs. The output power and external

quantum efficiency of these devices at 20 mA forward current were 5.1 mW and 9.1%

(blue) and 3 mW and 6.3% (green). The green device displayed a record luminous

intensity of 12 cd with 10 cone viewing angle at 20 mA forward current. The output

power, quantum efficiency, and luminous intensity of the green device are higher than

any previous LED at a similar wavelength from AlInGaP, GaP, or ZnTeSe.

Red InGaN LEDs have been fabricated, but still suffer from poor InGaN crystal

quality and phase separation due to high In-content. Presently, AlGaAs LEDs offer

superior performance. The incorporation of InGaN blue, InGaN or GaP green or yellow,

and AlGaAs red LEDs allows near-complete coverage of the chromaticity diagram, as








illustrated in Figure 2-10. The Nasdaq MarketSite in Times Square, New York City-the

largest flat panel display in the world-is an impressive manifestation of such technology

(Figure 2-11). Continued advancements in InGaN blue and green LEDs are projected,

and white LED packages for general lighting applications are eagerly anticipated in the

near future. Modules capable of 100 Im/W luminous flux conversion efficacy with total

output of 50 Im have been predicted [139].

Corporate research and development dominates the current activity in III-N

LEDs, with Nichia, Cree, Toyota Gosei, OSRAM Opto, and LumiLeds among the major

participants. The current state-of-the-art is still set by Nichia, offering a full product line

of blue, green, red, and white LEDs, with output powers from 2 mW (red) to 6 mW

(blue) and luminous intensities as high as 10 cd (green) for devices operational to 85C.














p-contact

p-GaN


n-Al4, ,Gao ssN


Inno.o6GC*9NXZAS9

n-contact


n-GaN


GaN buffer

Sapphire


p-contact

5000 A p-GaN

1000 A p-Alo.2CGa'o.8N


'30 A lno,45Gao.55N


n-contact


- 4 pm n-GaN


GaN buffer

Sapphire


Figure 2-9. Cross-sectional schematic of (a) AlGaN/InGaN/AIGaN double
heterostructure blue LED and (b) single quantum well InGaN green LED.














0.8 3. Gromw
S green 5s4 4. YellowGreen
InGaN L Yellow
L. Orwyae
7. Red
sM 8. Red Puple
0 f GaP:N 9. Purple
0.6 ,10. rWhite

blue-green 3 4 InGaAIP
V r~InGaN ^' A
0.4 -6 Ioa Ao


AIGaAs
0.2


blue "
InGaN
0
0 0.2 0.4 0.6 0.8

X


Figure 2-10. Chromaticity diagram showing spectral coverage of currently available
LEDs from various III-V materials systems. The circular regions in the center represent
various shades of white light. Adapted from [144].































Figure 2-11. The NASDAQ MarketSite Tower-- the world's largest LED display--in
Times Square, New York. Millions of InGaN LEDs were used in the full-color 7-story
screen. Reprinted with the permission of The Nasdaq Stock Market, Inc.


2.6.2 Laser Diodes

The announcement of the first current-injection blue laser diode in 1996 [145]

was a monumental leap for III-N technology. Such a device has been coveted for nearly

30 years, and several reports of optically-pumped lasing from various III-N structures had

appeared [146-149]. However, Nakamura's InGaN multi-quantum well (MQW)

structure was the first to achieve stimulated emission by pulsed current injection. These

devices were grown by two-flow MOCVD and typical Ti/AI and Ni/Au ohmic

metallizations were deposited on n- and p-type layers, respectively. Mirror facet

formation was accomplished by RIE to define the stripe geometry lasers. Dry etching








was necessary due to the lack of cleavage planes along the (0001) sapphire face,

exacerbated by the fact that the nitride columns are slightly rotated with respect to the

hexagonal orientation of the substrate. Consequently, the morphology of the etched facet

of the InGaN device was not particularly good (-50 nm roughness), necessitating the use

of high reflection coatings. Still, the 417 nm output (FWHM = 1.6 nm) represented the

shortest wavelength ever obtained from a semiconductor laser. The output power and

threshold current density (Jt) of this device were 215 mW (at 2.3 A) and 4 kA-cm"2,

respectively, under pulsed current conditions at room temperature [145].

Subsequent reports described the fabrication of the first cleaved facet GaN-based

laser [150] and growth and fabrication of LDs on closely-lattice matched MgAl204

substrates [151]. In a surprising announcement, room temperature continuous wave

(CW) operation was reported less than 1 year after the first demonstration of pulsed

current InGaN LDs [152-154]. The threshold current, turn-on voltage, and threshold

current density were 80 mA, 5.5 V, and 3.6 kA.cm"2, respectively. A CW output power

of 1 mW was obtained at 405.83 nm. Estimated CW operation was >30 hours for a

device with structure given schematically in Figure 2-12 [155]. The spontaneous and

stimulated emission was attributed to excitons at deep levels originating from InGaN

quantum dots [156].

These early LDs were typically characterized by extremely short lifetimes due to

p-contact spiking along dislocation lines, shorting the p-n junction. This was, in part,

caused by the poor p-ohmic contacts, which led to significant local heating and high turn-

on voltage (34 V for the first Nichia device). The high threshold current density also

contributed to significant heat generation in the active layer. Improvements to CW








lifetime and operating voltages were addressed by 2 design changes, namely the

introduction of modulation-doped strained-layer superlattices (MD-SLS) and epitaxial

lateral overgrowth (ELO) [157]. The SLS structure allowed for thermal and lattice

mismatch stresses to be accommodated elastically, reducing cracking commonly

associated with thick AlGaN cladding layers. The modulation doping helped reduce the

operating voltage. Epitaxial lateral overgrowth (ELO) is a growth technique which uses

selectively masked regions for subsequent regrowth of low dislocation density GaN (see

Section 2.3.1). The regrown GaN nucleates homoepitaxially and grows laterally over the

masked regions, which do not allow nucleation. The growth conditions are chosen such

that lateral growth proceeds faster than vertical growth, and the film eventually coalesces

above the mask regions. Threading dislocations can be reduced below 106 cm2 in the

regions above the mask. The window regions typically retain a relatively high

dislocation density (107 109 cm2) In addition, the coalescence front may contain lattice

misregistry, tilt boundaries, and/or anti-phase domains. Namamura and co-workers

applied this growth technique to fabricate devices above both masked and window

regions. The devices above the window region had a threshold current density of 6-9

kA-cm2 while those above the mask exhibited values as low as 3 kA cm"2, clearly

suggesting the role of threading dislocations as leakage paths, increasing Jth. The

structure of a LD incorporating the InGaN MD-SLS structure and ELO substrate is given

in Figure 2-13. Room temperature CW operation of 2,200 hours was verified, and
dI
estimates of >10,000 hours were extracted from the degradation speed, (mA/100 h),
dt


where I is the operating current of the device [158].








Currently, Nichia retains its leadership role in commercially-available short

wavelength LDs, with a product line consisting of 30 mW single and multiple transverse

mode devices at X = 405 nm. These high power lasers are expected to find applications

in optical storage in the very near future. At least 20 groups have demonstrated

functional blue LDs, with Cree and Toyota Gosei currently among the closest to

challenging Nichia in the commercial arena.


p-AIGMGaO3N

I p-GaN


Energy


Figure 2-12. Structure of CW InGaN laser diode, with detailed band diagram of active
region quantum wells [155].











p-GaN S102
I-.Al ,.Ga.j N/GaN MD-SLS ,
p-AI 6.,,.GalU N
Inft.Ga&*/Iln ,.jGa#MN MQW
n.GAN n-electrode
n-Ai .,4GaUN/GaN MD-SLS 'ZM i _____
n-In &, GaoN n-GaN .
SiOC _ _ ._ ,
GaN buffer layer
(0001) sapphire substrate




Figure 2-13. Schematic ofInGaN MD-SLS laser diode grown on ELO substrate [158].


2.6.3 Detectors

In addition to light-emitting devices such as LEDs and LDs, III-N semiconductors

present exciting possibilities for use in visible and ultraviolet (UV) photodetectors. Much

of the interest for such devices stems from military and defense applications such as

missile detection. Other potential uses include ozone monitoring, flame detection,

chemical analysis, and aerospace communications. Many of these end-uses require

operation in extreme radiation, high temperature, or chemical environments. The III-

nitrides are well-suited for such harsh conditions, owing to their wide bandgap and

chemical stability. The potential exists for detectors with a large spectral coverage (200 -

650 nm) utilizing the entire (In,Ga,Al)N system. Devices capable of operation near 280

nm are especially attractive, due to an ozone absorption band in the solar spectrum. The

stray radiation at this wavelength (that would constitute background noise in a








photodetector) is effectively filtered out by the atmosphere. Devices operating in this

energy range are known as "solar blind." AlGaN alloys with -50% Al content absorb in

this region of the electromagnetic spectrum.

Numerous types of III-N photodetectors have been reported in the literature,

including p-n junction, metal-semiconductor-metal, p-i-n, photoconductive, and

transparent Schottky contact devices [159]. Avalanche photodiodes are not easily

implemented in the III-N system due to the materials' inherent high breakdown fields [9].

Of the reported devices, transmission cutoff is typically very abrupt, and noise and dark

current characteristics are in some cases even lower than for Si detectors [160]. Detector

response times have decreased to the nanosecond range for the best devices [161].


2.7 III-N Electronics

In addition to its 3.4 eV bandgap, GaN possesses attractive material properties

which warrant the tremendous attention recently given to nitride-based electronic devices

for high temperature, high power, and high frequency applications. Due to the early

research emphasis on light-emitting devices, the technology of GaN electronics lags

behind that of LEDs and LDs. However, a recent surge in interest has led to promising

results and vigorous active research in GaN bipolar and unipolar devices. In addition, III-

N electronics research has been able to leverage much of the processing technology

developed for LEDs and LDs. Demonstration of GaN-based MESFETs, JFETs, HFETs,

MISFETs, BJTs, HBTs, Schottky diodes, and p-i-n rectifiers have all appeared recently

in the literature.








2.7.1 Field Effect Transistors

The first III-N electronic device was a GaN MESFET fabricated by Khan an co-

workers in 1993 [162]. Although there have been numerous reports in the literature of

MESFETs with reasonable performance [163,164], the most noteworthy advantage of the

III-nitrides over other wide bandgap semiconductors, such as SiC, is the availability of

AlGaN/GaN heterostructures. High quality heterostructures with up to 50% Al content

have been demonstrated by a variety of growth techniques, including MOCVD and MBE.

Remarkable progress has been made in recent years in high performance heterostructure

field effect transistors (HFETs, also known as MODFETs or HEMTs) grown on a

variety of substrates including sapphire, SiC, and Si [165-172]. A typical HEMT

structure is given in Figure 2-14.

The extremely high current density achievable with AlGaN/GaN heterostructures

(up to 1.6 A/mm has been reported) is a result of the large polarization-induced field and

large conduction band offset in the AlGaN/GaN system. This polarization field has both

a piezoelectric, strain-induced, as well as a spontaneous polarization component. For the

AlGaN HEMTs most widely studied to date, the spontaneous polarization field

dominates. Even in the absence of modulation doping, this built-in field induces a two-

dimensional electron gas (2DEG) that is linearly proportional to the Al-mole fraction, x

(0 < x < 1) as ns = x-(5 x 1013 cm'2). For the -30% Al-comparison most widely studied, a


SModulation doped field effect transistors (MODFETs) are actually a subset of HFETs,
utilizing selective barrier doping to spatially separate ionized donors from the electrons in
the 2DEG. This spatial separation leads to an increase in channel mobility and, for this
reason, these devices are also known as high electron mobility transistors (HEMTs). Due
to polarization effects in the III-N system, undoped HEMTs are possible, and typically
the terms HFET, MODFET, and HEMT are used interchangeably in the nitride literature.
The convention adopted for this work will be to use the term 'HEMT' to describe all
heterostructure FET devices.









channel sheet electron density of ~1.5 x 1013cm"2 can be realized that is a factor of 5 10

times higher than typical GaAs or InP pHEMTs. The associated mobility at this high

current density is in the range of 1000 1500 cm2 / V s, which is well below that

achieved in the conventional III-V material. However, the combined pcAn product

remains competitive. It is the pNs product that enables this material system to be

applicable to low noise amplifiers as well as power amplifiers.



Source Gate Drain

/ UD cap layer
AJ~aN( F i
\ n donor layer UID
Sspacer
layer
2-D
electron
gas
UIDGaNA

GaN or AN buffer

i sapphire or SiC substrate



Figure 2-14. Generic HEMT device structure, illustrating modulation doping; 2-D
electron gas; source, gate, and drain contacts; and mesa isolation.



The high voltage capability of AIGaN/GaN HEMTs is the result of a critical

electric field of> 3 MV/cm. This value is 10 times that of Si and 5 times that of GaAs.

For typical AlGaN/GaN HEMT structures, this translates to gate-to-drain breakdown

voltages of approximately 100 V/pm. While record breakdown voltage over 500 V have

been demonstrated for large gate-to-drain spacing, the more important result is for a

device layout consistent with good microwave performance (i.e. gate-to-drain spacing of








1-2 pm). In this latter case, a breakdown voltage of -80 100 V can be maintained in a

microwave transistor that also has the high channel current mentioned previously. Figure

2-15 clearly illustrates these advantages compared to conventional high-speed compound

semiconductor technologies. Due to the high breakdown field, III-N HEMTs are able to

maintain large breakdown voltages in smaller geometry (i.e. short channel) devices that

retain excellent high-speed performance.

To make full use of the high voltage capability, a transistor must not be thermally

limited. GaN also has an advantage over GaAs in this regard, with thermal conductivity

up to 2.0 W/cm-K for GaN versus 0.46 W/cm.K for GaAs. Furthermore, GaN HEMTs

can be grown on semi-insulating SiC with a thermal conductivity of 3.3 W/cm-K. The

high breakdown field and good thermal conductivity allow these devices to be used in

high efficiency (theoretical efficiency of 78%) class B push/pull amplifiers at full power

rating. GaAs microwave devices, on the other hand, can only be implemented in a class

B push/pull amplifier by backing off the voltage bias (and hence the power level) to

accommodate the higher voltage swing in this configuration as compared to class A,

single-ended operation.

Over the past few years, dramatic progress has been made in understanding the

AIGaN HEMT device physics and in demonstrating record microwave power

performance. The potential of any microwave device technology is first realized in small

periphery devices with microwave power density being the relevant figure of merit.

These results are summarized in Figure 2-18, where state-of-the-art output power

densities at X-band are shown as a function of time. Note the steady progress in power





















GaN HEMT


GaA HBT


IlP DHBT


SiGe BJT


0 50 100 150 200 25


fT (GIz)
Figure 2-15. Breakdown voltage vs. unity-gain cutoff frequency for various compound
semiconductor device technologies, illustrating high frequency, high voltage capability of
the III-N material system. After [177].


IMP
SBHT








performance since the mid-1990s. The best results to date are 9.8 W/mm with 47%

power added efficiency (PAE) at 8 GHz [173]. This power density is close to seven

times the best result obtained with any GaAs technology and over twice that achieved

with SiC MESFETs at this frequency [174]. Compilations of reported power density and

PAE from AlGaN/GaN HEMTs and GaAs pHEMTs are given in Figures 2-17 and 2-18,

respectively. Power added efficiency values are comparable between the two

technologies, but the power densities achievable with AIGaN/GaN HEMTs are over an

order of magnitude higher. With continued improvements in material quality and device

design, a power density of 12 W/mm or higher appears plausible for AlGaN/GaN

HEMTs. Other impressive results include 6.6 W/mm with 35% PAE at 20 GHz [175].

The 20 GHz performance clearly illustrates the potential for high power III-N electronics

at K-band and beyond.

It has also been predicted and shown that AlGaN/GaN HEMTs can achieve low

microwave added noise figures (NF = 0.6 dB at 10 GHz) while maintaining a large

breakdown voltage (>60 V) and hence a large dynamic range [176]. These results imply

that these devices can be used to perform the active transmit and receive functions in

more robust, higher dynamic range modules.




























1995 1996 1997 1998
Date


1998 1999


Figure 2-16. Output power density of AlGaN/GaN HEMTs at X-band (8-12 GHz) as a
function of demonstration time. The data points represent values reported in the
literature. Note the superiority of III-N HEMTs over GaAs and SiC. Adapted from
[178].


Frequency (GHz)


Figure 2-17. Reported output power per unit gate periphery comparing GaN HEMT and
GaAs pHEMT technologies. Note that GaN HEMT power densities are more than an
order of magnitude higher for the frequency range 3 -20 GHz.


./

Best SiC -



best GaAs -


SI I I


--- -- .~~ '^ I -- - * .
A A A .6.6 W/mn
@ 20 GHz-
A A A
A t
A A




0
0 0
0 0 0 0 0

A GaN HEMT
SGaAspHEMT


.... I






65



80
00
A GaN HEMT
A o GaAspHEMT
60 A I
5 V 0 0
.41 A 0
Q A
W 40
A A 0
00
A
A A
| 20 A

of
0 I -- I I -- I
0 10 20 30 40 50
Frequency (GHz)


Figure 2-18. Reported power added efficiency for frequencies from 0 50 GHz
comparing GaN HEMT and GaAs pHEMT technologies. The GaN HEMT PAE values
remain comparable, despite much larger output power density.



2.7.2 Bipolar Transistors

Group llI-nitride bipolar junction transistors (BJTs) and heterostructure bipolar

transistors (HBTs) are attractive due to their potential for high current handling, high

breakdown voltage, and device linearity. Additional advantages for HBTs over FETs

include increased threshold voltage uniformity, lower phase noise, and higher

transconductance [179]. Applications include radar, satellite, and communications

applications in the 1-5 GHz range, with the potential for operation up to -30 GHz

predicted as the device technology matures [180]. Despite the many attractive features,

progress in III-N bipolar electronic devices has greatly lagged the tremendous successes

enjoyed by AlGaN/GaN HEMTs due to difficulties associated with both growth and

processing.








As discussed in Sections 2.3.2 and 2.5.1.2, the deep Mg acceptor leads to

relatively low p-type doping concentrations in GaN. Combined with the large bandgap of

3.39 eV, low resistance p-type Ohmic contact formation is difficult. In n-p-n BJTs and

HBTs, these difficulties are related with the p-base. In the base, it is desirable to have a

relatively high doping concentration to increase hole injection. Decreasing the hole

concentration leads directly to a decreased collector current, reducing gain. The poor

base contact resistance and conductivity introduces a large parasitic element and is

manifested by a common emitter offset voltage.

For n-p-n devices, the emitter mesa etch must stop on the buried p-base. In GaN,

point defects due to dry etch-induced lattice damage tend to be donorlike in nature. Type

conversion (i.e. p-type layers becoming n-type) of the exposed base surface is possible if

sufficient damage is introduced during the etch. Such a requirement presents a difficulty

specific to BJTs and HBTs, since LEDs and LDs are typically grown with buried n-layers

and p-type surfaces. For LEDs and LDs, etch damage is not as critical, since increased

donor doping actually aids n-ohmic contact formation. In contrast, low-damage dry etch

conditions are critical for successful HBT processing. Increased recombination at the

emitter-base junction may result from a damaging etch.

The first III-N HBT was demonstrated by Ren and co-workers in 1998 [181].

This device was grown by MBE on an MOCVD GaN buffer/sapphire substrate. The

device epilayer structure is given in Figure 2-19. The emitter mesa etch was stopped

-O~100A from the base to minimize damage. The remaining emitter material was oxidized

and stripped in acid. The low hole concentration in the p-base (~low 1017 cm"3) led to

strong series resistance effects and small low bias current at room temperature. Device








performance improved when tested at 300C, presumably due to increased thermionic

emission and Mg ionization. At 300C, a DC current gain of-10 was demonstrated, as

illustrated in the Gummel plot of Figure 2-20.

Extensive use of simulation and modeling has been employed to improve layer

structure design and predict device performance [182-186]. More recent experimental

efforts have focused on various approaches to reduce the undesirable effects of the p-

base. An n-p-n AIGaN/GaN HBT with selectively regrown GaAs:C was investigated to

reduce base contact resistance [187]. BJTs and HBTs with regrown emitter have been

investigated due to the lack of a need for an emitter mesa etch [188,189]. Superlattice

structures have been shown to increase the hole density of Mg-doped GaN by decreasing

the effective ionization energy [190,191]. A GaN BJT employing a superlattice base and

regrown emitter was demonstrated with a current gain of~-10 and operation at a power

density of 150 kW cm-2 [189]. Some p-n-p devices have been investigated to circumvent

the p-type base. However, this geometry suffers from the inferior transport properties of

holes, leading to reduced switching speeds.











500 nm GaN Si-doped (8E18)

20 nm AIGaN graded to GaN Si-doped (8E18)

80 nm AlGaN Si-doped (8E18 and Al = 0.2)

20 nm GaN nm undoped

220 nm GaN nm Mg-doped (5E17)

400 nm GaN Si-doped (1E16)

500 nm GaN Si-doped (8E18)

GaN Buffer Layer



Figure 2-19. Epilayer structure ofA1GaN/GaN npn HBT. The GaN buffer layer was
grown by MOCVD and all subsequent layers were grown by MBE [192].


0.01


1E-3


1E-4


1E-5


1E-6


1E-7


2 4 6 8 10
Base Voltage (V)


Figure 2-20. Gumnmel plot at 300C for first AIGaN/GaN HBT, illustrating current gain
of-10 [181].








2.7.3 Rectifiers

GaN electronic device research has been largely dominated by FETs. MESFETs

and HEMTs for high frequency, high power applications have been developed to exploit

the attractive material properties of the III-nitrides, as described in Section 2.7.1. HBTs

and BJTs are promising, but still suffer from the materials and processing-related issues

discussed in Section 2.7.2. A III-N device which has received considerably less attention

is the power rectifier. Applications for these devices are numerous and include radar,

satellite systems, hybrid-electric vehicles, utilities transmission and distribution, and

"smart" power modules.

One of the immediate applications for GaN power rectifiers stems from the

electric power utility industry. A major problem in the current grid is momentary voltage

sags, which affect motor drives, computers, and digital controls. Mechanical switches are

presently used to control electric current flow across utilities transmission and

distribution lines. Opening or closing these switches can lead to large voltage or

inductance spikes delivered to the load. Such spikes may be detrimental, for instance, to

major computing centers or other sensitive electronic equipment. An outage of less than

one cycle, or a voltage sag of 25% for two cycles can cause a microprocessor to

malfunction [9]. As a result of these potential fluctuations, the electric power grid must

be operated at capacities well below its rated value, leading to reduced energy efficiency.

A system for eliminating power sags and switching transients would dramatically

improve power quality [193]. Solid state devices, if available, are expected to show

"clean" switching and could potentially eliminate line transients and allow more efficient

operation of the grid. Since typical power devices are required to operate at elevated

temperatures due to the power dissipation associated with switching large currents and








voltages, wide bandgap materials are attractive due to their increased tolerance to

temperatures above the limits of silicon. Reduction of bulky, expensive cooling

equipment should be possible, leading to decreased system complexity and cost.

Motors consume up to 50% of the electricity produced in the United States each

year [9]. Repair costs for these motors are estimated at $5 10 billion annually. These

expenses are expected to be greatly reduced by high power electronic devices that permit

smoother switching and control. In addition, control electronics could dramatically

improve motor efficiency, reducing the amount of power consumption. Other end uses

for III-N high power switches include lighting and HVAC systems [193,194].

Wide bandgap materials are attractive for use in power devices due to their ability

to sustain extremely high electric fields, leading to large blocking voltages. The on-state

resistance of a diode rectifier is related to the breakdown voltage and critical electric field

of the semiconductor by:

RON C 4VB2 (2.4)


where VB is the reverse breakdown voltage, gn is the electron mobility, is the

semiconductor permittivity, and E, is the critical electric field for breakdown. From

Equation 2.4, it can be seen that for a fixed value of breakdown voltage, materials with

larger critical electric breakdown field have a smaller on-resistance (given comparable p.

and rs), leading to reduced power dissipation in the forward, conducting state.

Analogously, for a given value of Ro, large Ec corresponds to large breakdown voltage.

There have been a number of reports of mesa and lateral geometry GaN and

AIGaN Schottky and p-i-n rectifiers fabricated on heteroepitaxial layers on A1203

substrates [195-202]. However, these results still fall well below the theoretical





71


predictions for the III-N material system [8,195,203]. Bandic et al. have predicted that a

50 kV blocking voltage should be sustained by 20 pin of Alo.Gao.gN doped to 1 x 1016

cm'3. Such performance has not been achieved in practice, most likely due to the highly

defective nature of III-N material at present. It has been amply demonstrated in other

materials systems such as SiC that the presence of defects (dislocations, nanopipes) leads

to premature breakdown in diodes and rectifiers [204-206].













CHAPTER 3
AlGaN / GaN HIGH ELECTRON MOBILITY TRANSISTORS

3.1 Introduction

One of the most noteworthy advantages of the III-nitrides over other wide

bandgap semiconductors, such as SiC, is the availability of AlGaN/GaN heterostructures.

The Type I band alignment between AIGaN and GaN has been shown to form a potential

well and a 2-dimensional electron gas (2DEG) at the heterointerface [165,207]. When

these materials are brought into contact, thermal equilibrium requires alignment of their

respective Fermi levels (EF). This induces conduction (Ec) and valence (Ev) band

bending in both the AIGaN and GaN layers and can cause the GaN conduction band at

the interface to drop below EF, as illustrated in Figure 3-1. Since (for n-type material) the

Fermi level can be viewed as an electrochemical potential for electrons, majority

electrons will accumulate in the narrow gap material just below the heterointerface to fill

the quasi-triangular potential well between Ec and EF. These electrons are confined by a

distance shorter than their deBroglie wavelength, causing quantization of the allowed

energy levels in the potential well. Depending on the structure, there may be more than

one allowed energy level below the Fermi level, although only the lowest allowed level

will be substantially populated at room temperature. With the heterointerface on one side

and a potential barrier on the other, electrons in the 2DEG are only free to move in along

the plane of the interface. A thin 'sheet' of negative charge (electrons) results.

















S ....... ... I _

S2-dimensional
i electron gas







-E
AIGaN A GaN
>- x


Figure 3-1. Energy band diagram illustrating formation of 2-dimensional electron gas at
AlGaN/GaN heterointerface.



Modulation doped field effect transistors (MODFETs) are a class of

heterostructure FET that use selective barrier doping to spatially separate ionized donors

from the electrons in the 2DEG, leading to an increase in channel mobility. For this

reason, these devices are also known as high electron mobility transistors, or HEMTs. A

typical AlGaN/GaN HEMT structure is given in Figure 2-14. The AlGaN barrier layer is

typically grown with the same Al content throughout, but with varying doping levels. A

heavily doped donor layer supplies electrons to the 2DEG. The ionized donors remain in

the AlGaN, while the electrons are transferred as mobile carriers into the GaN. An

unintentionally doped (UID) AlGaN layer between the donor layer and the gate metal








allows formation of a Schottky contact to the wide bandgap barrier. A thin AlGaN spacer

layer between the donor layer and channel (-30 A) is used to screen the Coulomb

potential of the ionized impurities. Without this spacer, Coulomb scattering from

positively charged impurities in the donor layer would lead to reduced electron mobility

in the 2DEG.

AIN, GaN, and their alloys are polar crystals. Polarization occurs due to the lack

of inversion symmetry for typical growth along the polar (0001) axis of the wurtzite

crystal structure. The strong spontaneous and piezoelectric polarization effects in the III-

nitrides can strongly influence the electron density and potential profile of heterostructure

devices. [208,209]. The polarization field is such that sheet electron densities on the

order of 1013 cm"2 can be realized even in undoped HEMTs [210].

In this chapter, a thorough discussion of all aspects of AlGaN/GaN HEMT device

fabrication, characterization, and small-signal modeling will be given. In Section 3.2, the

details of device processing, as well as DC, small signal, and large signal device

characterization will be presented for MOCVD-grown HEMTs on sapphire substrates. In

addition, s-parameter modeling will be used to model the small signal performance and

extract information on the intrinsic elements of the FET. In Section 3.3, a direct

comparison of AIGaN/GaN HEMTs of identical structure grown on A1203 and SiC

substrates will be given. In Section 3.4, the device performance of HEMTs grown on

sapphire substrates by reactive molecular beam epitaxy will be discussed.








3.2 MOCVD-Grown HEMTs on AlOi Substrates

3.2.1 Device Processing

All HEMTs in this study were fabricated by a 4-step process sequence consisting

of dry etch mesa isolation, Ohmic metallization, Ohmic anneal, and gate metallization.

The mesa etch was performed in a load-locked Plasma-Therm SLR 770 with a 2 MHz, 3

turn coil ICP source. A conductive adhesive was used to mount the samples to an

anodized Al carrier plate which was cooled from the backside by He gas. A 13.56 MHz

rf bias was superimposed on the substrate. The clear field etch mask was photoresist

AZP 4330, which was spun at 4000 rpm for 30 sec., baked at 90C for 90 sec., exposed

for 6.5 sec. at 20 mW/cm2, and developed for -125 sec. in MF319. Surface profilometry

indicated a photoresist (PR) thickness of -3.75 pin, which was highly reproducible across

the surface of the sample. The ICP process gases and flowrates were 8.0 sccm BCl3, 32.0

sccm C12, 5.0 sccm Ar. The addition of BC13 to the CI/Ar chemistry was important for

maintaining smooth surface morphology of the etched surface [211]. The chamber

temperature and pressure were maintained at 2 mTorr and 25C, respectively. An ICP

source power of 500 W and a substrate RF power of 40 W was used to minimize surface

damage. By using high ICP and low RF powers, a high density plasma is created which

maintains fast etch rates with reduced ion energy. The DC self bias was below -90 V

under these conditions. Typical etch rates were -1300 A/min. Etch depths (mesa step

heights) were 1300 1700 A. The stability of the 4330 photoresist was excellent.

Ohmic lithography was performed with AZ 5214 PR in both positive tone (dark

field mask) and image reversal (clear field mask). The linewidth resolution was

approximately the same for both processes, while the thickness of the positive tone PR








was slightly larger (1.46 jym vs. 1.35 psm for 5000 rpm spin speed). A hexamethyl

disilazane (HMDS) treatment and dehydration bake was used prior to PR application to

improve PR adhesion. The PR procedure for positive tone was: 5000 rpm spin for 30

sec., bake 90C for 90 sec. on hotplate, pattern expose -6.5 sec. at 20 mW/cm2, and

develop -110 sec. in 1:1.4 MF312:H20. Immediately before loading the patterned wafers

for metallization, a 60 sec. 1:10 HCl:H20 dip was used for removal of the native oxide.

Electron beam evaporation was initiated at chamber pressures of-3 x 10-7 torr. An

Ohmic metallization scheme of Ti/Al/Pt/Au (250/1000/450/1500A) was deposited at

rates from 2 10 A/sec, depending on the metal. This quad-layer scheme was annealed in

a rapid thermal annealing (RTA) furnace at 850C for 30 sec. under a N2 ambient. Four-

probe transmission line method (TLM) analysis resulted in specific contact resistances

from high 10-7 to low 10-6 0-cm2 with transfer resistances of 0.3 0.4 f-.mm. A typical

TLM plot of resistance vs. contact spacing is given in Figure 3-2. From such a plot, sheet

resistance, transfer resistance, and specific contact resistivity can be simultaneously

determined. Figure 3-3 (bottom) illustrates the very rough morphology of the Ohmic

metallization after the high temperature anneal. The top picture is unannealed

Ti/Al/Pt/Au for comparison. Note, however, that the edge acuity of the contacts (which

affects the channel length) appears to be quite good. This will be further illustrated by

scanning electron micrographs in Section 3.2.3. Some reports claim a tradeoff between

contact resistance and surface roughness [212], which can be tailored by varying layer

thicknesses or annealing conditions. For this study, an Ohmic metallization experiment

was performed in which Ti/Al/x/Au (250/1000/450/1500A) was deposited on

AlGaN/GaN HEMT structure, where x = Ni, Pt, or Ti. Each of the metallizations was








annealed at 800,825,850, 875, and 900C for 30 sec. in N2. Circular TLM patterns were

used to construct Figs. 3-4 and 3-5 by averaging 3 sets of TLM data for each

experimental data point shown. Note that the general trend is for decreasing contact

resistance at higher annealing temperature. However, as temperatures were increased

above -850C, the contact morphology significantly degraded. An 850C-annealed

Ti/AI/Pt/Au Ohmic metallization was chosen for all devices presented in this work,

unless otherwise noted.

Optical (contact) gate lithography was carried out with AZ 5214 photoresist, in a

process analogous to that used for the Ohmic level. Gate lengths for the optically

patterned devices were 0.8 and 1.2 pmr. Electron beam lithography was used to define

submicron gates using a bilevel PR scheme of 6% PMMA (bottom) and 9% P(MMA-

MAA) copolymer (top). The copolymer is more sensitive than the PMMA and therefore

develops out wider than the PMMA resulting in a T-shaped cross section. Additionally,

the copolymer may be selectively developed with respect to the PMMA by choosing an

appropriately low dose that is sufficient to expose the copolymer but not the PMMA.

Therefore, the 'wing' portion of the mushroom gate may be extended by placing lightly-

dosed boundaries on either side of the main gate dose. The main gate boundaries were

weighted in the CAD file and doses were optimized to achieve the desired gate lengths.

The 6% PMMA was spun at 6600 rpm for 30sec. and baked a minimum of 15

min. on a hotplate at 170C, which produced -4000 A resist. The 9%/ P(MMA-MAA),

was spun at 4400 rpm for 30sec. and baked a minimum of 15 min. on hotplate at 170C

for ~5500A resist. Patterns were exposed at 50 kV, -4nA beam current in 480 x 320 pm

fields, with a 25 nm step distance. A typical large area dose was 225 iC/cm2. Typical








gate doses ranged from 375 ILC/cm2 for 0.5 pm gates to 800 C/cm2 for the 0.1 pm gates.

For the Cascade Probe FET mask design used in this study, the write time was 27

minutes per unit cell (1 cm2). The majority of the write time was for the pads and device

labels. The exposed PR was developed for 2 minutes in a 1:2 mixture of MIBK:IPA

(methyl isobutyl ketone: isopropyl alcohol) and dried with N2. Prior to metallization, a

1:10::HCI:H20 oxide strip was performed. For optimum liftoff using this recipe, the

thickness of the deposited metal should be between 4000 and 5500 A. For all submicron

devices presented in this study, gate metallization consisted of e-beam evaporated Ni/Au

(200/4500 A). A gentle acetone rinse peeled away the PR in one thin sheet. The gate

yield was close to 100% (i.e. no gate metal adhesion problems).




30i

25
TIM
S 20-

I 15

10-

5


-5 0 5 10 15 20 25 30


Figure 3-2. Plot used to determine contact resistance by the 4-probe )M method.

Figure 3-2. Plot used to determine contact resistance by the 4-probe TLM method.




79



F ,




















Figure 3-3. Ti/Al/Pt/Au Ohmic metallization before (top) and after (bottom) 850C
RTA.















~1


800 825 850
Temperature (C)


875 900


Figure 3-4. Specific contact resistivity as a function of annealing temperature for
Ti/Al/x/Au, where x = Ti, Pt, or Ni.


g0o 825 850 875 900
Temperature (C)


Figure 3-5. Transfer resistance as a function of annealing temperature for Ti/Al/x/Au,
where x = Ti, Pt, or Ni.


-o- Ti
-0- Pt
-A-Ni


10 -6L


I I I


-A- Ti
-0- Pt
-o-Ni


I I I 1 1 i B m m








3.2.2 Optical Gate Devices

The device structures were grown with conventional precursors at 1040C in an

impinging jet MOCVD system with rotating substrate. The growth sequence consisted of

a thin AIN nucleation/strain relief layer followed by ~-1.6 pmrn of unintentionally doped

GaN, deposited at a growth rate of-300 A/min. The modulation doped cap layer

consisted of30A UID Alo.2Gao.gN, 200 A Alo.2Gao.gN:Si (nM-l x 10" cm'3), and 100 A

UID Al0.2Gao.sN. A schematic of the device structure is given in Figure 3-6 and an

optical micrograph of a completed HEMT is shown in Figure 3-7.

The DC characteristics were measured in common source mode with an HP

4145A Parameter Analyzer. Maximum drain current densities of 0.5 A/mm were

obtained for 0.8 x 100 Wm2 gate dimension devices, as shown in Figure 3-8. For the same

devices, the extrinsic transconductance was ~135 mS/mm at Vo = -1.5 V and VDS = 3 5

V (Figure 3-9). Slight self-heating effects were evident for the higher power levels

associated with wider gate devices (i.e. 200 pm devices) biased to VDS =10 V. The

Schottky gate contacts exhibited leakage current on the order of 0.3 3mA.

Scattering parameters were measured with an HP 8510 Vector Network Analyzer

calibrated via the short-open-load-thru (SOLT) method to --40 GHz. The small signal

current gain (h21) and Unilateral gain (IUI) are shown in Figure 3-10. From this plot, a

cutoff frequency (fT) of 9.7 GHz and maximum frequency of oscillation (fm,) of 23.5

GHz were determined using -20 dB/decade extrapolation to the frequency axis.

Figure 3-11 is a summary of the microwave performance of 0.8 and 1.2 um gate

length devices with gate widths of 100, 150, and 200 pm. These data were taken from

immediately adjacent devices in an effort to minimize the effects of material non-








uniformity. In general, measurements at various positions on the wafer provided similar

results. Both fT and f.. were larger for the 0.8 gim gate length devices than for 1.2 pin

devices, although the magnitude of this difference was smaller for fma. However, the

cutoff frequency-gate length product (fT.L-) was -20% lower for the 0.8 mm gate length

devices. The fT-LG product of-10 GHz- jm is comparable to results from similar devices

in the literature [213,214], although values as high as 18.3 GHz-jun have been reported

[215]. From simple charge control theory:

Veff = 2nfTLG (3.1)

where veff is the effective electron velocity in the 2DEG and LG is the gate length.

Equation 3.1 provides a convenient estimate of electron velocity if both fT and L are

known. Using the 1.2 pm gate length device data from Figure 3-11, ve is calculated to

be (6.2 0.1) x 106 cm-s1. Still, this is roughly a factor of 2 below the theoretical value

predicted for this material system, clearly pointing toward the importance of reducing

scattering mechanisms (such as interfacial roughness) during crystal growth. A possible

explanation for the discrepancy between the fT-rLG product of the 0.8 and 1.2 pm devices

is the value of LG used in the calculation. Dimensions of 0.8 pm are very near the

resolution limit for contact lithography using a standard Karl Suss MJB-3 aligner. It is

possible the actual gate lengths were slightly larger than the nominal values, which would

lead to both a smaller percentage difference between the 2 gate lengths and a larger

calculated effective electron velocity (closer to the theoretical value).

































Figure 3-6. Layer structure of MOCVD-grown AlGaN/GaN HEMT.


Figure 3-7. CCD image of optical HEMT with gate dimension given in device label (0.8
x 100 pm2). Source, drain, and gate contact pads are indicated.


O A UD Aio.2Gao.8N
100 A Alo .2GaoNSi, -1E18 cm"3
30 A UD Ao.2Gao.08N

-200 A 'transition' GaN



-1.7 gm UIDGaN



-200 A AiN


(0001) Sapphire































2 4 6 8 10


VDS (V)


Figure 3-8. Low voltage DC output characteristics of 0.8 x 100 jin2 gate dimension
HEMT


-4 -3 -2 -1 0 1
Vo (V)


Figure 3-9. DC output characteristics of HEMT at 25C.


0.0 M
0


50 F


i



100 4

0
o

50

CO


0.8 x 100 gm2
VDS = 5V
















40. . . .



30



20
~.-.

0 . ..
a B

10
h 21


0
0.1 1 10 o100
Frequency (GHz)



Figure 3-10. Small signal characteristics of 0.8 x 100 pnm2 device illustrating 23.5 GHz
fmax.


U


10 g


8 |
A..


Gate Width, Wa (AM)


Figure 3-11. Gate length and gate width dependence of fT, fmx, and fT-Lo product.


SGate Length:
0 -e- 0.8 gm
f -o- 1.2 gm
max :S__--o-L2pm


o0

0
0 0 0
0 ----- o -0- -


I*4








3.2.3 Submicron Gate Devices

Submicron gate devices were fabricated using the processing sequence detailed in

Section 3.2.1 on samples with 230 A (sample NF1214D, Figure 3-6) and 430 A (sample

NF 1110A, Figure 3-12) AlGaN layers (spacer + donor + cap). Scanning electron

microscopy (SEM) was used to examine the quality of the lithographic processing steps

and estimate the gate lengths of the submicron devices. Figure 3-13 shows an SEM

micrograph of a 0.1 pmn gate crossing the mesa. Note the use of an anchoring technique

at the end of the gate to improve adhesion. The anchors were necessary for the 0.1 pm

devices, but not for the longer gate lengths. A cross-sectional view of the same gate

finger (0.1 x 100 un2) is shown in Figure 3-14. Note the excellent edge definition on

both the gate and the source/drain contacts. The surface morphology of the MOCVD-

grown epilayers is also excellent, with no evidence of defects or tilted growth boundaries.

Figure 3-15 shows an SEM micrograph of a 0.25 pm gate length device with the mesa

step in the background. Once again the edge acuity of the metal is excellent. This

micrograph gives a good view of the T-gate and shows that the lithography and lift-off

processes produced well-defined gate contacts. From detailed SEM analysis, gate lengths

of 0.12, 0.23, and 0.52 un were measured for the nominal 0.1, 0.25, and 0.5 un values,

respectively.































Figure 3-12. Layer structure of MOCVD-grown HEMT.



The maximum drain current density for the NF 1110A devices approached 1

A/mm before self-heating effects caused a decrease in the output current. Characteristic

drain I-V curves are given in Figures 3-16 (top and bottom) and 3-17 for 0.1,0.25, and

0.5 pm gate length devices, respectively. Note the excellent pinch-off of the 0.25 and 0.5

pm devices. The 0.1 pin devices do not maintain complete pinch-off and exhibit

significant output conductance due to short-channel effects. These are caused by the 2-

dimensional nature of the electric field for short gate lengths and large drain bias, and are

common in GaAs- and InP-based HEMTs. For the shortest gate length devices, the

extrinsic transconductance exceeded 200 mS/mm. The transfer characteristics for each

gate length are given in Figures 3-18, 3-19, and 3-20. The peak of the extrinsic

transconductance curve shifted toward more positive voltage with increasing gate length


200 A UID Ao.2Gao.8N cap

200 A Alo 2Gao 8N:Si, -E18 cm"3
30 A UID Al .2GaO .gN spacer

-200 A 'transition' GaN



1.0 pm UID GaN buffer



-200 A LT AN


(0001) Sapphire








and the magnitude ofgm scaled with inverse gate length (Figures 3-21 and 3-22). For

these devices, high temperature operation with excellent pinch-off characteristics has

been verified to 400C, the temperature limit of the probe station heater. Drain current

density decreases monotonically with increasing temperature over the range 25 400C,

as shown in Figure 3-23.

Maximum gate-source breakdown voltages were -25 V for NF 1110A and > 60 V

for NF1214 devices with 0.25 x 150 pm2 gates measured in 3-terminal mode with source

and drain shorted to ground (Figure 3-24). The gate leakage current for 1214D was -200

pA at a reverse bias of-60 V. The leakage from 1110A devices was much higher (-1

mA prior to hard breakdown). The reverse current density of the 0.25 and 0.5 pm gate

length devices was similar, but leakage from the 0.1 pm device was significantly higher,

as shown in Figure 3.25. This may be due to the higher dose (-800 AC/cm2) during e-

beam writing of the shorter gate lengths. This effect should be investigated in additional

detail to optimize the submicron lithography process.

Gate current-voltage-temperature characteristics from 1214D were measured on a

heated chuck from 25C to 395C. The gate leakage current decreased by nearly 3 orders

of magnitude upon heating the device (in ambient air) from room temperature to 395C

(Figure 3-26). The turn-on voltage (Vo.) also increased with temperature, as shown in

Figure 3-27. Note that from 100C 150C, V. and the on-resistance (Rm) increases

slightly. From 150C 200C, the on-resistance is restored to approximately its initial

value. From 200C -300C, there is very little change in the forward I-V characteristics.

Above 300, Vo. begins to increase with no change in Ron. This behavior is likely caused

by the hypothesized e-beam-induced damage to the AIGaN surface. The damaged








regions under the gate contact are partially repaired upon heating. It is possible that the

anomalous forward I-V characteristics are caused by a characteristic activation energy

required for damage repair. This phenomenon emphasizes the sensitivity of the (Al)GaN

surface to disruption by energetic particle-enhanced processes (others include dry etching

or plasma enhanced chemical vapor deposition). In the early days of GaN device

processing it was thought that the material was relatively impervious to these problems.

Part of the reason for the dramatic improvement in device performance over the past few

years has been the development of processing conditions that minimize surface damage.

Continued progress in this area is paramount to the realization of the full potential of

these devices.




Table 3-1. Summary of experimental DC and RF characteristics of MOCVD-grown
AlGaN/GaN HEMTs.
Maximum Maximum
Nominal Current Extrinsic Cutoff Oscillation
Gate Length, Gate Width, Density, Transconductance, Frequency, Frequency,
Lg (gm) Wg (pm) IDS (A/mm) gm (mS/mm) fT (GHz) f., (GHz)
0.1 100 0.89 207 58 90
0.25 150 0.79 186 29 63
0.5 200 0.73 172 19 39


The cutoff frequency and maximum frequency of oscillation were extracted from

measured s-parameters and are given in Figures 3-28, 3-29, and 3-30 for gate lengths of

0.1,0.25, and 0.5 pm devices, respectively. These values are also listed in the Table 3-1,

which summarizes the key DC and RF parameters of the 11 lOA devices. Maximum fT

and fmax of 58 GHz and 90 GHz, respectively, were achieved for the 0.1 x 100 pm2

devices, clearly illustrating the potential of AlGaN/GaN HEMTs for high frequency








operation. The 0.25 upm devices exhibited an fT of 29 GHz and f. of 63 GHz, while the

same parameters for the 0.5 pm device were measured to be 19 and 39 GHz, respectively.

The high fm values are encouraging, since this is one of the most important metrics of

device performance for applications such as satellites where overall gain is critical.

Figures 3-31 and 3-32 give a graphical representation of the gate length

dependence of fT and fn. A linear fT vs. LG"1 relationship is observed, as predicted from

Equation 3.1. Calculation of the effective electron velocity from Figure 3-32 and

Equation 3.1 gives veff = 6.5 x 106 cm/s, consistent with the value estimated in Section

3.2.2, but below the theoretical value of-2 x 107 cm/s derived from Monte Carlo

simulations.

Microwave power characteristics were measured with a load pull system at

frequencies up to 10 GHz. The output power of a 0.25 x 150 Pn device at 3 GHz is

shown in Figure 3-33 as a function of DC input power. The output power increased

monotonically with input power in the range 100- 800 mW. At 1 dB compression, this

device produced 22.5 dBm output power with a gain of 19.2 dB. The associated power-

added efficiency (PAE) was 20.6%. At the 3 dB compression point, the same device

produced 26.2 dBm output power, 17.2 dB gain, and 56.3% PAE, corresponding to an

output power density of 2.75 W/mm (Figure 3-34). These power values at 3 GHz are

approximately a factor of 5 higher than can be obtained with GaAs pHEMTS. At 10

GHz, an output power density of 1.7 W/mm was measured. This is roughly a factor of 3

higher than state-of-the-art GaAs pHEMTs at this frequency. As has been well

documented in the AIGaN/GaN HEMT literature, the output power of devices grown on

sapphire is severely limited by the poor thermal conductivity of the substrate. Flip-chip




91

bonding or other means of heat sinking would allow higher power densities to be
realized.








.. ;; . ^^ il^
.. In. .r..' *S. . ..







i, -., .. ... ., ,* "R. saG


IAN




mesa step, and gate profile.