Ohmic contacts to P-type gallium nitride

MISSING IMAGE

Material Information

Title:
Ohmic contacts to P-type gallium nitride
Physical Description:
xii, 175 leaves : ill. ; 29 cm.
Language:
English
Creator:
Liu, Bo, 1969-
Publication Date:

Subjects

Subjects / Keywords:
Materials Science and Engineering thesis, Ph. D   ( lcsh )
Dissertations, Academic -- Materials Science and 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 164-174).
Statement of Responsibility:
by Bo Liu.
General Note:
Printout.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 026988398
oclc - 47421294
System ID:
AA00020450:00001

Full Text










OHMIC CONTACTS TO P-TYPE GALLIUM NITRIDE


By

BOLIU






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















ACKNOWLEDGMENTS

I would like to thank my supervisor, Dr. Paul H. Holloway, for the opportunity

and financial aid to study GaN, a promising semiconductor material. I am also grateful

for his philosophy on the relation between research and knowledge, for his desire to

understand students, and his devotion to improving my communication skills. I treasure

the memory of his guidance and help with my research and personal affairs.

I thank my committee members: Drs. RolfE. Hummel, Kelvin S. Jones, Fan Ren

and Wolfgang Sigmund for their valuable contributions. The assistance from the staff of

MAIC, Drs. Ren's group, Dr. Jones's group, Dr. Hummel's group and collaborators in

the clean room in the Department of Electrical Engineering is highly appreciated.

I thank all of the members in Dr. Holloway's group, both past and present. I

appreciate their friendship and patience over the past few years. I enjoyed working with

each of them.

I also thank all of the friends I made during my time in Gainesville, including

those who favor me as their barber. I thank each of them for making my life in

Gainesville so beautiful.

I especially thank my family for their patience, support and encouragement

through all my professional endeavors.

This work was supported by EPRI/DARPA under agreement No.W-08069-07.















TABLE OF CONTENTS
page


ACKNOWLEDGMENTS .................................................1ii

LIST OF TABLES........................................................................... ...... ........................ vi

LIST OF FIGURES .................................................................................. ..................... vii

ABSTRA CT.......................................................................................................................xi

CHAPTERS

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

2. REVIEW OF LITERATURE ................ ................................ ..................................... 6

2.1 Growth of GaN and D efects................................. .. ................................. ........ 6
2.2 M echanism s of Ohm ic Contact ........................................................................ 9
2.3 Ohmic Contact to p-GaN: Present Research Status................................................ 20
2.3.1 Ferm i Level Pinning.................................................................. ..... ......... .. ... 22
2.3.2 Surface Preparation of GaN ....................................................................... ..... 23
2.3.2.1 H ydrogen..................................... ................................................. ............. 24
2.3.2.2 Carbon ......................................... .......... ............................................. ....... 25
2.3.2.3 Oxygen......................................................... ............................................. 26
2.3.2.4 Cleaning of Surface ................................................................ .. ............ 27
2.4 Interfacial M etallurgical Reactions......................................................................... 28
2.4.1 Gallide-form ing M etals ....................................................................... ....... 29
2.4.2 Nitride-form ing M etals..................................................................... ......... 34
2.4.3 Neutral M etals............................................................................... ................... 37
2.4.4 Therm al Stability.................................................................................... .... 38
2.5 M etallization Schem es and Analysis.................................................................. 40
2.5.1 Conventional Contact Schem es....................................................................... 40
2.5.2 Non-conventional Contact Schem es.................................... ............................ 43

3. EXPERIM ENTAL PROCEDURES ............................................................... .......... 46

3.1 Introduction .......................................................................................... .. ................. 46
3.2 Contact Preparation........................................................ .................................... 46
3.3 Characterization............................................................................ ....................... 50









4. EFFECTS OF H202 SOLUTION TREATMENT ON p-GaN ...................................... 54

4.1 Introduction............................................................................................................. 54
4.2 M odification of Electrical Conductivity ....................................... ......... ........... 55
4.2.1 Effects of H202 Concentration.................................................................... 55
4.2.2 Effects of Extended Imm version Tim e........................................... ..... ......... 58
4.2.3 Stability of the Increased Electrical Conductivity....................................... 60
4.3 Structural Characterization ....................................................................... ...... ..... 63
4.3.1AES .................................... .......................................................................... 63
4.3.2 ESCA .................................................................. ........ ...... ........................... 64
4.3.3 SIM S.............................................................................................. ......... 64
4.3.4AFM .......................................................................................... ............... 64
4.4. Application in Formation of Ohmic Contact to p-GaN......................................... 71
4.5 Discussion................................................................................................................... 71
4.6 Summ ary................................................................................................................. 83

5. "NOG" SCHEME FOR OHMIC CONTACT TO p-GaN............................................. 84

5.1 Introduction...................................................................................... ....................... 84
5.2 Principles of"NOG" Schem e ............................................................................. .... 85
5.3 Comparison with Published Contact Results .......................................................... 86
5.4 Experim ental Studies............. .......................................................... .. ......... .. ....... 94
5.4.1 Effects of Ti and Al as Nitride-Form ing M etals.............................................. 95
5.4.2 Effects of Si and M g as Nitride-Forming M etals ............................................. 97
5.4.3 Neutral M etals............... ............. ....................................................... ............... 105
5.5 Discussion ........................................................................................ ..................... 107
5.6 Summ ary ................................................................................................................ 110

6. EFFECTS OF Ni CAP LAYER ON THIN Ni/Au CONTACTS TO p-GaN ........... 112

6.1 Introduction........................................................................................................... 112
6.2 Contact Electrical Properties................................................................................. 114
6.2.1 Annealing Temperature ................................................................................. 115
6.2.2 Effects of Annealing Tim e............................................. ................................ 117
6.2.3 Effects of 02 Flow Rate......................................... ........................................ 118
6.3 Light Transm ittance.............................................................................................. 119
6.3.1 Effects of Annealing Temperature............................... .................................. 120
6.3.2 Effects of Annealing Tim e............................................................................. 121
6.3.3 Effects of 02 Flow Rates............................ .................................................... 122
6.4 M icrostructure Characterization................................................. .......................... 123
6.4.1 SEM ............................................................................................................... 123
6.4.2 AES Survey and Depth Profiling.................................... ............................... 131
6.4.3 XPS Analysis................................................................................... .............. 136
6.5 Discussion .................................................................................................................. 145
6.6 Summ ary ...................................................................................... .. ....................... 157

7. CONCLUSION S............................................................................ .............................. 159


iv









8. FUTURE W ORK ......................................................................................................... 162

LIST OF REFEREN CES................................................................................................. 164

BIO GRAPH ICAL SKETCH ........................................................................................... 175











LIST OF TABLES


Table Page

1.1 GaN-based electric devices .......................................................................................3.......3

2.1 Comparison of experimental and calculated values for the heat of formation AH fr of
related gallides and nitrides ............................................................... .......... 32

2.2 Current metallization schemes of ohmic contact to p-GaN ........................................ 41

4.1 Atomic concentration of elements from AES surface survey analysis in MBE p-GaN..63

4.2 XPS results from a 1:1, 300sec H202 cleaned p-GaN sample ......................................... 66

4.3 Relation of hydrogen incorporation and processing steps............................................... 78

5.1 Enthalpy and entropy of hydride formation................................. .................................... 89

5.2 Electrical conductivity of selected nitrides...................................................................... 92

5.3 Calculated driving forces of metal reactions to Ga and GaN.......................................... 108

5.4 Calculation of diffusion characteristic distance of selected metals in nickel.................. 110

6.1 Lattice constants (A) of components in oxidized Ni/Au contacts to p-GaN................... 147

6.2 Values of surface tension of Ni and Au] .......................................................................... 150














LIST OF FIGURES


Figure Page

1.1 Relation ofbandgaps and lattice constants...................................................................... 2

2.1 A nanopipe in GaN imaged by HRTEM ...................................................................8......8

2.2 Energy diagrams of a metal contact to a semiconductor ................................................. 10

2.3 Change of the interfacial behavior factor S........................................... ..................... 12

2.4 Mechanisms of ohmic contact formation ......................................................................... 14

2.5 Possible patterns used to measure contact resistance......................................................17

2.6 Plot of measured resistance vs. contact separation .......................................................... 18

2.7 Pressure-temperature projection of a three phase equilibra............................................. 19

2.8 Dopant locations in bandgap of GaN............................................................................21

2.9 Barrier height vs. metal work functions on n-type GaN .................................................23

2.10 Calculated Ni-Ga-N diagram at 600C .......................................................................30

2.11 Calculated (a)Zr-Ga-N and (b) La-Ga-N isothermal diagram at 298K ......................... 35

2.12 Calculated phase diagram for Ti-Ga-N at 800C........................................................... 36

2.13 Calculated (a) W-Ga-N and (b) Re-Ga-N diagram at 6000C...................................38

3.1 Configuration of contacts used in Hall measurement ............................................. 49

3.2 Schematic of light transmittance measurement ............................................................... 51

4.1 Effects of H202 solution treatment on the I-V curves of25A Ni/500A Ti/500A Au to
M B E p-G aN .......................................................... ........................... .................... 56

4.2 Effects of H20z2 solution treatment on the I-V curves of 100OA Ni/500A Ti/500A Au to
M B E p-G aN ......................................................................................... ................ 57










4.3 Hall measurement results ofMBE p-GaN ................................................................ 58

4.4 Effects of H202 solution treatment with extended time on the I-V curves of 100A
Ni/500A Ti/500A Au to MBE p-GaN ................................................................. 59

4.5 Effects of H202 solution treatment with extended time.......................... ......................... 60

4.6 Hall measurement results of H202 solution treated MBE n-GaN.................................... 61

4.7 Effects of H202 treatment on MOCVD p-GaN......................... ....................................... 61

4.8 Hall measurement on stability of H202 treated samples.................................................. 62

4.9 Comparison of XPS peaks from as-cleaned and 1:1, 300sec H202 treated GaN......... 65

4.10 Negative SIMS depth profile for 5:1 H202 solution treated p-GaN........................ 66

4.11 AFM images of GaN................................................................................. ... ...............68

4.12 Microstructure of MBE-GaN........................................................................ ........ 69

4.13 Effects of 5:1, 60sec H202 treatment on the I-V of 500A Ni/500A Au to p-GaN......... 70

4.14 Relation among width of depletion region, carrier concentration and built-in
potential in GaN ............................................................................................... 73

4.15 Relation between pH values and H202 concentration.................................................... 76

4.16 Relation of carrier concentration decrease and nanopipe density in GaN ..................... 81

5.1 Principle of "NOG" scheme................................................................... .......... .8........86

5.2 I-V ofNi/Au, Ni/Ti/Au and Ni/Al/Au on p-GaN...................................................... 96

5.3 Effects of Ni thickness on the I-V of Ni/Ti/Au contact to p-GaN ................................... 97

5.4 Effects of thermal annealing on I-V data................................... ...................................... 98

5.5 I-V curves of500A Pt/500A Au, 100oA Pt/50A Si/500A Pt/500A Au andlOOA Pt/50A
Mg/500A Pt/500A Au contact on p-GaN, as-deposited...................................... 99

5.6 I-V curves of 500A Pt/500A Au, o100A Pt/50A Si/500ooA Pt/500A Au and ioo100A
Pt/50A Mg/500A Pt/500A Au contact on p-GaN, 600C for Imin annealing.....99

5.7 I-V curves of 500A Pt/500A Au, o100A Pt/50A Si/500A Pt/500A Au and o100A
Pt/50A Mg/500A Pt/500A Au contact on p-GaN, 800C for Imin annealing..... 100

5.8 AES depth profile of Pt/Au contact on MOCVD p-GaN................................................ 102









5.9 AES depth profile of Pt/Si/Pt/Au contact on MOCVD p-GaN........................................ 103

5.10 AES depth profile of Pt/Mg/Pt/Au contact on MOCVD p-GaN ................................... 104

5.11 I-V of 500ooA Ni/500A Au, 100A Ni/looo1000A Ti/looo1000A Au and 100oA Ag/1000A
Ti/1000A Au on p-GaN .......................................................................................105

5.12 Comparison of In metal and 100oA Ni/500A Ti/500A Au as ohmic contact for Hall
m easurem ent........................................................................................................ 106

6.1 Effects of anneal temperature on I-V of 50/50 contact.................................................... .115

6.2 Effects of annealing temperature on the specific contact resistance of the 50/50,
50/50/50 and 50/100/50 contacts ................................................................. ........ 117

6.3 Effects of annealing time on the specific contact resistance of 50/50, 50/50/50 and
50/100/50 schemes annealed at 600C ............................................ ..................... 119

6.4 Effects of anneal time on resistance of contact pads at 600C ........................................ 120

6.5 Comparison of light transmittence at X = 450 nm........................................................... 121

6.6 Effects of annealing time on light transmittance at X = 450 nm ...................................... 122

6.7 Effects of 02 flow rates on light transmittance at 500C................................................. 123

6.8 Microstructure of the 50/50 contact after annealing ........................................................ 124

6.9 Microstructure of the 50/50/50 contact after annealing ................................................... 125

6.10. SEM backscattering electron image of same sample as in Figure 6.9-(d) but at a
higher magnification showing the Au film is still continuous ............................. 126

6.11 Microstructure of the 50/100/50 contact after annealing............................................... 127

6.12 EDS analysis of the light region in Figure 6.16-c .......................................................... 128

6.13 EDS analysis of the dark region in Figure 6.16-c ............................... ........................... 128

6.14 EDS analysis of spherical particle in Figure 6.16-d ...................................................... 129

6.15 Annealed Au film on GaN and sapphire ........................................................................ 130

6.16 Annealed (600C, 10min) Ni film on GaN................................................................ .... 131

6.17 AES surface spectra from 50/50 contacts......................................................................132

6.18 AES surface spectra from 50/50/50 contacts................................................................. 133









6.19 AES surface spectra from 50/100/50 contacts............................................................... 134

6.20 AES depth profile of 50/50 contacts.............................................................................. 135

6.21 AES depth profile of 50/50/50 contact .......................................................................... 136

6.22 AES depth profile from 50/100/50 contacts .................................................................. 137

6.23 XPS spectra of Ni2p from 50/50 contact after annealing at 600C ............................... 139

6.24 XPS spectra of Ni2p from 50/50/50 contact after annealing at 600C .......................... 140

6.25 XPS spectra of Ols from 50/50 contact after annealing at 600C................................. 141

6.26 XPS spectra of Ols from 50/50/50 contact after annealing at 600C............................ 142

6.27 XPS spectra of Ga2p from 50/50 contact after annealing at 600C............................... 143

6.28 XPS spectra of Ga2p from 50/50/50 contact after annealing at 600C.......................... 144

6.29 Energy diagram of oxidized thin Ni/Au contact to p-GaN............................................ 146

6.30. Schematic diagram of interface equilibrium between three phases.............................. 148

6.31 AE/p (= rp/R) for (3 = 42 and 10/5 calculated for 50/50 and 50/50/50........................... 152

6.32 Schematic diagram of contact microstructure................................................................ 155














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
OHMIC CONTACTS TO P-TYPE GALLIUM NITRIDE
By

Bo Liu

May 2001
Chairman: Dr. Paul H. Holloway
Major Department: Materials Science and The effects of H202 treatment, multi-layer

metallization, and Ni cap-layer on Ni/Au have been studied for ohmic contacts to p-GaN.

First, surface H202 treatments are found to increase the hole concentration by up to 100% in

p-GaN grown by molecular beam epitaxy (MBE), while causing no change in n-GaN or p-

GaN grown by metalorganic chemical vapor deposition (MOCVD). Treatment of 20 min

increased, while treatment >60 min decreased the hole concentration in MBE p-GaN. With

this treatment, the current in Ni/Au contacts increased. The increased hole concentration was

attributed to reduction of nitrogen vacancies or H-Mg complexes in GaN. The decrease of

carrier concentration was attributed to recompensation of shallow acceptors by oxygen

serving as hole traps.

Second, general principles were defined for selecting metals for ohmic contacts to

GaN, a scheme called "NOG" for Nitride-forming metal Over Gallide-forming metal. In

"NOG" a gallide-forming metal dissociates GaN and a nitride-forming metal increases the

nitrogen thermodynamic activity at the interface. Literature data were compared to these

ideas, and experimental data on Ni/Ti/Au, Ni/Al/Au, Pt/Si/Pt/Au, Pt/Mg/Pt/Au were

collected and compared with data from Ni/Au and Pt/Au contacts. Higher currents were









found for schemes based on the "NOG" principles. However, the contact resistivity was still

high and thermal stability became a limiting factor.

Last, ohmic contacts to p-GaN were obtained after oxidizing Ni/Au and Ni/Au/Ni

contacts. Both Ni/Au/Ni and Ni/Au were shown to have resistivities of-10"4 Qf-cm2.

Transparent NiO was obtained and thin Au film formed pores which led to optical

transparencies at X = 450nm of>85%. The porosity in Au was demonstrated to result from

interface and grain boundary energies. Addition of the Ni cap-layer was shown to increase

the thermal stability of thin Ni/Au ohmic contacts and increase the light transmittance to

93%, while keeping contact resistivities of low 10-4 Q-cm2.














CHAPTER 1
INTRODUCTION

As members of the III-V nitrides family, InN, GaN, A1N and their alloys are all

wide band gap materials, and can crystallize in both hcp wurtzite (a) and cubic

zincblende (P) crystal structures. As shown in Figure 1.1, the wurtzite InN, GaN and A1N

have direct bandgaps of 1.9 eV, 3.4 eV and 6.2 eV, respectively, at room temperature

[Mor97]. In the cubic form, GaN and InN have direct bandgaps, and A1N has an indirect

bandgap. The GaN alloyed with InN and AIN can form a continuous (AlGaln)N alloy

system spanning a continuous range of direct bandgap energies throughout the visible to

near UV region of the electromagnetic spectrum. This makes the nitride system very

attractive for optoelectronic device applications, such as light emitting diodes (LEDs),

laser diodes (LDs) and optical detectors [Mor96]. Various GaN-based electric devices

also have been demonstrated for applications in high power/high temperature electronic

devices [Table 1.1] because of their intrinsic properties of wide bandgap and high

breakdown fields [Cho94, Kha94, She99b Gas98].

In particular, GaN with its band gap of 3.4 eV plays a central role in the alloy

system of (AlGaIn)N. Other advantageous properties of GaN include high mechanical

and thermal stability, large piezoelectric constant, good thermal properties [Bin97] and

the possibility of passivation by forming thin layers of Ga203 or A1203 with bandgaps of

4.3eV and 9.2eV respectively.
















-P-AIN


4.0

I3-GaN.
3.0 P- aaN


2.0 t k
a-InNN
300 K
1.0 *I*' 1'' '' ... j: i* ....* ... j..... .
3.0 32 3.4 3.6 3.8 4.2 4.4 4.6 4.3 5.0
Lattice Constant (A)


Figure 1.1 Relation ofbandgaps and lattice constants of hexagonal (a) and cubic (P) InN,
GaN, AIN and their alloys [Mor97]



Because of its wide bandgap and chemical inertness, GaN and related materials

are challenging materials researchers. Examples of some challenges include but not

limited to growth ofheteroepitaxial films [Amb98], the low level of p-type doping

because of deep acceptor levels [Mol93] and Mg-H complex [Got95], difficulties in

achieving low-resistance ohmic contact to p-GaN [Liu98], slow wet and dry etch rate

[Ade93, Pea94], and lack of good passivation films [Uza95], etc. This work focuses on

formation and improvement of ohmic contacts to p-type GaN.










Table 1.1 GaN-based electric devices [Shu99]

Device Status
Schottky barrier diode Demonstrated for a variety of metals

p-n junction Demonstrated on both regular and
lateral epitaxial overgrowth
material

GaN MESFET Demonstrated


GaN MISFET


AlGaN/GaN HFET


AlGaN/GaN and
GaN/SiC HBTs

GaN-based piezoelectric
and piezoelectric
sensors
GaN pyroelectric sensor


Demonstrated


Demonstrated on sapphire and SiC
substrates with record power levels

Demonstrated


Demonstrated for GaN and
AlGaN/GaN

Demonstrated for primary and
secondary pyroelectric effect


Progress has been made in the past several years in developing reliable ohmic

contact to p-GaN [Tre96, Tre97, Hol97, Liu98, Ho99a, Ho99b, Koi99 and PalOO], both

for lower specific contact resistance and better thermal stability. Nevertheless, the goals

are far from being achieved. It has been proven difficult to obtain sufficiently low contact

resistance (<10-6 Q-cm2) to p-GaN. In GaN based LEDs, Ni/Au are commonly used as

ohmic contact on p-type GaN top layer. The low doping levels of the p-GaN usually

result in non-ohmic contact, and thereby degrade the device performance. Also, due to

the high resistivity of the p-GaN layer, the current from the top electrode can not be

spread effectively through the entire device chip, leading to current crowding. Large


Possible Applications
Switch, FET building block

Photodetector, switch, BJT
building block

High-temperature digital
circuits
High-temperature digital
circuits, non-volatile
memories
High-temperature, high power
microwave, high temperature
digital circuits
High-temperature, high power
microwave, high power
switches
Pressure sensors, especially for
high temperature
applications
Temperature sensors, especially
for high temperature
applications









Joule heating caused by high operating voltages has been reported to limit the lifetime of

continuous wave (CW) laser diodes [Nak97]. For metal semiconductor field effect

transistors (MESFETs), an ohmic contact resistivity of mid -10-6 Q-cm2 may be

acceptable for a channel length of 1 pm [Mur90]. For p-GaN, the best stable ohmic

contact resistivity is only 10.1 10-3 Q-cm2. Ohmic contacts to n-GaN have been

developed with resistivities of 10-6 Q-cm2 [Ren97].

The high specific contact resistances to p-type GaN can be attributed to several

factors, including 1) absence of a metal with a sufficiently high work function (The

bandgap of GaN is 3.4 eV and the electron affinity is 4.1eV, so a work function of 7.5 eV

is needed for a good ohmic contact. Metal work functions are typically < 5.5 eV); 2)

difficulty in achieving high hole concentrations in p-GaN because of the deep ionization

level of the acceptors (Mg is -170 meV, others are deeper [Str91]); and 3) the tendency

for preferential loss of nitrogen from the GaN surface during processing, which probably

result in surface conversion to n-type conductivity. To decrease the contact resistance to

p-GaN, high p-type electrical conductivity or lower barrier height at the contacts interface

would be helpful.

Specific contact resistivities as low as 10-6 Q-cm2 have been reported for thin

oxidized Ni/Au films [Ho99b]. This contact scheme also showed high (>60%) light

transmittance, which is desired in optoelectronics. The reported contact thickness (only

100 A) was comparable to the GaN surface roughness. Lower contact resistivity and

higher light transmittance are obviously desirable.

In this dissertation, H202 solutions were demonstrated to increase the carrier

concentration of MBE grown p-GaN. A "NOG" scheme (Nitride-forming Over Gallide-









forming metals) offers guidance in development of contacts to both n- and p- type GaN

epilayers. Low resistance and high transparency contacts to p-type GaN are also studied.

It was demonstrated that adding a thin Ni surface cap-layer, to the thin Ni/Au layered

contacts improved light transmission and enhanced contact stability.

Thus the scope of this dissertation is as follows.

Chapter 2 presents a review of the physical mechanisms of ohmic contact

formation and discusses current literature on ohmic contact to p-GaN.

Chapter 3 summarizes the experimental procedures used in this study, including

contact processing parameters and characterization techniques.

Chapter 4 summarizes the use of H202 solutions to increase the free hole

concentration. Possible explanations for increased hole concentration are discussed.

A methodology to choose materials to test for better GaN ohmic contact

formation is proposed and discussed in Chapter 5. This procedure ("NOG") is based on

the properties of different metal groups. Various contact schemes to p-GaN are tested and

discussed.

In Chapter 6, the microstructural evolution of transparent Ni/Au ohmic contacts is

reported. The consequence of annealing these layers in N2 or 02 ambients are described.

Addition of a Ni cap-layer to form GaN/Ni/Au/Ni contacts is shown to increase rather

than decrease the optical transparency.

Finally, Chapter 7 is a summary, and chapter 8 discusses future work to improve

the understanding ofp-GaN ohmic contacts.














CHAPTER 2
REVIEW OF LITERATURE

Many GaN-based electronic and photonic devices have been developed, and all

these devices need ohmic contacts to external current and voltage sources. Because of

high conductivity, metals are normally used for this purpose. Typically, interfaces

between metals and semiconductors have been found to exhibit rectifying properties

[Rho88, Bri93]. Although these rectifying properties are useful in some devices, many

applications require efficient transport of current across the contact interface. Ohmic

contacts are generally required for such applications. Creating low-resistance ohmic

contacts to GaN, especially for p-type material, has proved to be one of the major

challenges faced in the development of GaN semiconductor technology [Pea97a].



2.1 Growth of GaN and Defects

Because sufficiently large (> 1 cm in diameter) single GaN crystals are generally

unavailable for use as substrate for homoepitaxial growth, heteroepitaxial growth of GaN

has been used in practice and the choice of substrate is critical to the structural, electrical

and morphological properties of the obtained epilayers [Li96, Kob97, Hamn98, Kun96,

Kur95, Geo96, Sun96, Pop97]. Possible substrate materials typically need to have good

matches in thermal expansion coefficients and lattice constants with GaN, and should be

resistant to the growth chemistries at high growth temperatures (over 1000C in certain

cases). Sapphire [Kai98, Pop97] and SiC [Pop97, Abe97] are the most popular substrates

currently used due to their adequate thermal and chemical stability at high growth









temperatures, excellent structural and surface morphology and availability in large

quantities. Other substrates include Si [Vis95], MgA1204 [Kur95], LiGaO2 [Kun96],

NdGaO3 [Kur95], quartz glass [Iwa97] and ZnO [Pop97, Dav97].

Metalorganic chemical vapor deposition (MOCVD) [Ama86, Kat94, Kel96,

Mor81 and Nak94], molecular beam epitaxy (MBE) [Mou93, Van97] and their

derivatives are the techniques most extensively used to grow GaN epilayers. Mg and Si

are widely used as p and n type dopants. Because of a large activation energy (270

meV) and passivation of acceptors with hydrogen, in which the activation energy is

dependent on dielectric constant (GaN, 9.5) and effective mass of the material (GaN, me

= 0.2 mo, mhh = 0.75 mo) [Den97], the ionization ofMg acceptors is less than 1% at room

temperature. Typical hole concentrations are -1017 cm-3 for p-GaN (although Mg

concentration can be as high as 1020 cm'3). For GaN grown with the MOCVD method,

the as-grown materials typically is unintentionally n-type, which is widely believed to be

due to intrinsic nitrogen vacancies. Intentional n-type doping can be easily

accomplished using silicon as the donor.

In the MOCVD method, the Ga source materials generally used are GaCI3

(produced by passing HC1 vapor over molten gallium), trimethylgallium (TMGa),

triethylgallium et al, and the nitrogen source is mostly of ammonia [Dav88, Wal97].

Biscyclopentadienyl (Cp2Mg) is used as a source of Mg and methyl silane (Si (CHaSiH3)

as a source of Si [Sun97, Her97]. The MOCVD method has been the leading technique

for production of III-nitrides optoelectronic and microelectronic devices. Characteristics

of this method include high purity chemical sources, easy composition and uniformity

control, high growth rates and abrupt junctions. Since partial pressure of nitrogen in the








CVD reactor is always less than the equilibrium partial pressure of nitrogen over GaN at

the high substrate growth temperatures, the samples may contain high concentrations of

intrinsic n-type carriers, commonly believed to be nitrogen vacancies [Pan97].

In the MBE process, the fluxes of Ga, Mg and Si are generated by heating high

purity elements in the Knudsen or effusion cells [Dav88]. The substrate temperature for

MBE growth is typically operated at relatively low temperature at 898K or even lower to

663K [Got81]. Nitrogen is typically supplied as an atomic species using electron

cyclotron resonance (ECR) or radio frequency (rf) plasmas [Pop98, Abe97, Mou93].


Figure 2.1 A nanopipe in GaN imaged by HRTEM [Kan99]








The GaN-based III-nitride heterostructures are found typically to contain characteristic

one-dimensional (edge, mixed and screw dislocations) and two-dimensional (stacking

faults and domain boundaries) extended defects [JaiOO]. Although the dislocation density

is high, the dislocations are usually clustered in local regions of the epilayer, so large

volumes of the epilayer are defect free [Pon97]. Nanopipes are also found in GaN with

diameters ranged from 5nm to 0.5um and densities as high as 108 cm2, as identified

with high-resolution transmission electron microscopy (HRTEM). These nanopipes were

reported to be parallel to the c-axis of GaN unit cell [Ven99, Kan99], as shown in Figure

2.1. The main constituents inside these nanopipes were Ga, C and 0 [Kan99].

A model was constructed to explain the effects of dislocations on minority

carriers in GaN epilayers [Jai98], and was found to be consistent with the experimental

results. Dislocations and nanopipes are postulated to help explain changes in the carrier

densities of MBE grown GaN after room temperature treatment with H202 solutions in

this work.



2.2 Mechanisms of Ohmic Contact

A contact is said to be ohmic when the ratio of the potential drop V across a

contact versus the current I flowing through the same contact is linear with a constant,

low contact resistance (Re). Ohmic contacts are characterized by a parameter called

specific contact resistance (or resistivity), pc, which is expressed as [She92]


Pc= (')0. (2.1)
aV












YT


.. Em Ef
~E~E
> % E91\1-w
E- ........... Ef ......--..--- i - ----- ....
E E ---- "- ---------- ^^,I4bi-

S Semiconductor Metal Semiconductor Contact

yi

(a)I (b)


Figure 2.2 Energy diagrams of a metal contact to and semiconductor,
explaining formation of ohmic contact to p-GaN (a) Before charge
equilibrium; (b) After charge equilibrium.


In general, when a semiconductor of a given electrochemical potential is brought

into contact with any phase with a different electrochemical potential, charge will flow

automatically across the semiconductor/contact junction (Figure 2.2). For an ideal

semiconductor/metal contact (Schottky limit), all the voltage drops across the

semiconductor. The contact barrier height for p-type semiconductor, ,, p, is calculated

using the equation [Kum93]:

b ,p -(Z + Eg) (2.2)
q

where Ef, is the Fermi level of the isolated metal (before contact) and X is the electron

affinity of the isolated semiconductor. In the literature, the work function of a metal, 4p (in









electron volts), is often used to estimate the barrier height limiting charge transfer at

semiconductor/ metal junctions. The work function is used instead of Ef m because the

Fermi level is more difficult to determine experimentally, whereas q7 is readily accessible

using photoemission or other data. Ideally, a metal with a lower work function than that

of an n-type semiconductor or a higher work function than that of a p-type semiconductor

can be used to form ohmic contacts to a semiconductor. Unfortunately, it has been shown

experimentally that most semiconductor/metal contacts do not obey the predictions of

this ideal Schottky limit [Rho88, Kum93]. In a simple model, the Schottky barrier height,

4,, can be expressed as follows [Sze81]:

q'-b = q(S- + 0) (2.3)

where Xm is the metal electronegativity (Note: not to be confused with &s, the electron

affinity of the semiconductor), and 4 represents the contribution of surface states from

the semiconductors. The interface index

S = dob (2.4)


is a function of the electronegativity difference A4 between the cation and anion

components of a compound semiconductor, as shown in Figure 2.3* [Ren98]. The reason

for the name of"S" factor is just from the shape of the graph. Note the sharp transition

around AX = 1. This S factor should be equal to unity for an ideal semiconductor/metal

junction in the Schottky limit. When S = 1, the measured barrier height equals the initial

contact potential difference obtained from the Schottky limit. The unity of S value means


The electronegativity difference of GaN was labled mistakenly at 1.8 eV in the literature, instead of the
true value of 1.23eV as shown in Figure 2.3.






12



1.2
ZnS AIN SO, ZnO SrTiO3
U) 1- *.4.... A^-^
'*GaA6 ,10 KTaO3
S/ GaS Expected
S0.8 Value
&CdSI
| 0.6 ,^Ga~e
0.8 Case

0.4 SiC
16 4 -Aadse
X GaP Cse
4 n-type
S0.2 -Go Gakp/ Actual
IGnSb *inP Values p-type
0 -,-te

0 0.5 1 1.5 2 2.5
BElectronegativity Difference


Figure 2.3 Change of the interfacial behavior factor S with different
semiconductors [Ren98].



absolutely no pinning of the Fermi level. When S =0, the system is in the regime of

strong "Fermi level pinning." This terminology indicates that the Fermi level position at

the surface of the semiconductor, measured relative to the vacuum level, does not vary

when either the work function or the Fermi level of the contacting phase is changed. For

covalent semiconductors with AX < 1, S is small, and 4, is typically affected by a high

density of surface states from dangling bonds, so that it depends weakly on the metal

work function. On the other hand, for ionic compound semiconductors, where AX%> 1, the

index S approaches unity, and depends strongly on the metal work function. GaN has

an electronegativity difference of 1.23 eV(Ga: 1.81eV, N: 3.04eV) [HsuOO], which

suggests that the Schottky barrier height should be a function of the metal work function.

The origin of this non-ideal interfacial behavior (Fermi level pinning) is a topic of

intense controversy. Bardeen originally proposed [Bar47] that surface states at the









semiconductor/metal interface are the source of this pinning. Chemical reactions, such as

stoichiometry changes [Spi86] and alloy formation [Fre81], have also been proposed to

explain Fermi level pinning.

For either an ohmic or rectifying contact, current is transported across the

interface by mechanisms shown in Figure 2.4. Current transport can be principally

ascribed to the following three mechanisms [Sze81]:

1. Thermionic emission (TE) is dominant in low and moderately doped

semiconductors with Ne(h) < 1017 cm"3. At low to moderate carrier densities,

the wide depletion region prevents tunneling through the barrier. When the

barrier height is small, the electrons can be thermally excited over the top of

the barrier (thermionic emission, Figure 2.4a). On the other hand, for a high

barrier, the vast majority of electrons are unable to overcome this barrier,

resulting in non-ohmic (rectifying) contacts.

2. Thermionic-field emission (TFE) is applicable for intermediate doping

densities, _1017 cm-3 < Ne(h) < _1018 cm-3. Both thermionic and tunnel emission

are important (Figure 2.4b).

3. Field emission (FE) is effective in heavily doped semiconductors, NO) > -10"8

cm"3. In this case the depletion region is narrow, and electron or hole tunnel

easily from metal to semiconductor or vice versa (Figure 2.4c).

A very useful parameter, KT/Eoo, can be used [Yan71] to calculate Pc for each of

these three mechanisms, where Eo00 is the tunneling parameter equal to:

EOO- q.h ,Neh. (2.5)
400 (e2m



























(b)









(c)


e e- e e'e










K
- - - - - - - - - --'_-






-- - - - -


Figure 2.4 Mechanisms of ohmic contact formation. (a) thermionic
emission; (b) thermionic field emission; (c) field emission.









where q is electronic charge, h is the Planck's constant, Ne(h) is electron (hole)

concentration, e is dielectric constant of the semiconductor, and m* is electron (hole)

effective mass.

For kT/E00oo << 1 (heavy doping concentrations), the specific contact resistance is

given by

oc exp(qA). (2.6)


In this case, the field emission mechanism (FE) dominates current transport. The Pc

depends strongly on doping concentration. At high doping concentration, the depletion

width of the Schottky junction is decreased, resulting in high tunneling transmission

coefficients. Hence, even a metal with a low metal work function can still form an ohmic

contact. With this method, as in GaAs, ohmic contacts can be formed on a semiconductor

with a pinned Fermi level [Hol97].

For kT/E00oo ~ 1 (intermediate doping concentrations), a mixture of both thermionic

and tunneling (TFE) transport is observed, and the specific contact resistance becomes



pc, exp q 1--q E (2.7)
I E0.coth( K )

The specific contact resistance in this case depends on both temperature and transmission

coefficient for tunneling.

For kT/E00 >> 1 (moderate doping concentrations), the TE mechanism dominates

the current conduction and the specific contact resistance is


Pc oc exp(q'A) (2.8)
A. 1






16


The specific contact resistance is clearly dependent on temperature. At higher

temperatures, the thermionic emission current increases and results in a smaller Pc.

The term ohmic contact in practice does not necessarily require a linear current

voltage characteristic [Rid75]. A metal-semiconductor contact is associated with a space-

charge region in which the current-voltage behavior eventually becomes nonlinear, as

bias increases. Ideally the contact resistance of the space-charge layer would be

negligible relative to the bulk or spreading resistance of the semiconductor contacted by

the metal, but this is rarely achieved in practice. A contact is usually acceptable when the

voltage drop is very small compared to the drop across the active region of the device,

even though the current-voltage behavior of the contact is not strictly linear. This is

certainly true for current contacts to p-GaN devices, due to the unavailability of low

resistance ohmic contact to p-GaN (see later in Table 2.2).

In theory, the contact resistance can be defined completely if physical and

operating parameters are known [Rid75]. The physical parameters are mainly of contact

area and thickness, while operating parameters are predominantly temperature and bias.

In practice, the contact resistance can be affected seriously by a number of other factors,

such as interfacial layers (oxide formation or contamination), surface damage, minority

carrier injection, and energetically deep impurity levels or traps.

The most widely used method for determining the specific contact resistance is

transfer length method, also often called the transmission line model (TLM) [Ber72,















(a)


Figure 2.5 Possible patterns used to measure contact resistance: (a)
linear array; (b) circular contacts [Wil84]. The grey region is
contact metal and the white region presents areas where the
metal has been etched or lifted off leaving bare semiconductor
or an etched mesa.


Wi184]. An array of metal contacts (Figure 2.5) is fabricated with different spacings

between the metal areas. The resistance is measured as a function of the gap spacing.

Extrapolation of the resistance to zero gap spacing gives a value equal to twice the

contact resistance R, (Figure 2.6). The x-intercept is equal to twice the transfer length, Lt,






18







I / dR RS
dL W

2,Y

s/


SContact spacing
2L,




Figure 2.6 Plot of measured resistance vs. contact separation [Wil84]


where L, = (pI/R) 1/2, Pc is the specific contact resistance and Rs is sheet resistance of the

semiconductor epilayer. The transfer length is defined as the distance from the edge to

where the current in the semiconductor falls to 1/e (e being the base of natural logarithm)

of its original value, pc can be calculated using the equation:

P, = R. R L (2.9)
R,

where L, should be much smaller than the gap spacing between contacts. However, as the

spacing become very small, irregular edges may cause a problem. The test pattern should

be isolated so that current flow occurs only in the space between pads (no leakage current

path), therefore a mesa structure may be needed [Wil84].











11 ,'--- ----------~-'-*



0 a^ 0^e P(N2)



A-- Assessed
-10 (9) \ ^
o (22)
(9)(24)
(25) I
-|5 (26) P(Ga)
o (27)
(31)
-20 1-_
4 6 8 10 12 14
10000/T(K)

Figure 2.7 Pressure-temperature projection of the three phase equilibra: GaN +
liquid + gas. The area inside each envelope represents the two-phase
equilibrium of GaN + gas and corresponds to GaN stability [Dav99].



For the pattern in Figure 2.5-(b) with a circular TLM (CTLM), device isolation

can be omitted because there is no leakage path for the current flow. The total resistance,

R1, calculation is complex, but when r/Lt >> 1, it can be calculated using [Ho99b]:
R R ln()+L-s-c1d











R,= ,. .. (-+ -)] (2.10)
2-in r R r

where Rs represents the sheet resistance ofp-GaN, R and r denote the radii of the outer

and inner circular contacts, respectively, and L, is the transfer length. The total resistance

is measured for different spacings and plotted as a function of ln(R/r). The least square

curve-fitting method can be used to obtain a straight linear plot of R, vs ln(R/r). The slope

gives R,, and the intercept at ln(R/r) = 0 is RsL/r.;r, leading to L', so the specific contact

resistance, pc, can be expressed as

L, = (,1 (2.11)
t R 26








In practice, linear I-V curves cannot always be obtained. To quantify the degree

of linearity conveniently, an arbitrary figure of merit [Piq98] is defined as follows:

dV
LM= dIW (2.12)
dVI


where the derivatives are the resistance at 5V and OV, respectively, and LM ranges from

zero to unity, approaching unity for samples with nearly linear I-V curves.

A good ohmic contact should have a low specific contact resistance, high

stability, smooth surface morphology and good edge definition. Because contact

resistance usually constitutes a small portion of the total measured resistance, caution

must be taken to avoid errors.

Metallic contact layers usually are prepared by vacuum deposition (electron beam

or thermal evaporation) followed by heat treatment. The most commonly used method of

heat treatment for the metal-semiconductor system is furnace annealing in H2, N2, or N2 +

H2 forming gas to protect the contact metals from oxidation. For p-GaN, the annealing

environment should not contain H2, due to the compensation of Mg acceptors with H

element [Sug98].



2.3 Ohmic Contact to p-GaN: Present Research Status

Initially, as grown GaN films were unintentionally n-type with a high carrier

concentration. This is widely believed to be due to nitrogen vacancies being intrinsic

donors [Neu94]. Vacancies result from the large vapor difference between the Ga and N

components as shown in Figure 2.7 [Dav99] The nitrogen vacancy lies just below the












Conduction Band N site
Ga site N site
-30 meV --
Native Defect Level




750 meV Li
700 meV Be

550 meV Cd
410meV Hg
340 meV Zn 225meV Si
___225 meV Si
250 meV Mg

Valence band



Figure 2.8 Dopant locations in bandgap of GaN [Den97]



conduction band (- 30 meV), which makes it an efficient donor (Figure 2.8) [Str9l,

Ben97].

In contrast to n-type, a critical breakthrough was achieved when p-type GaN was

reported using Mg dopants followed by low electron beam irradiation (LEEBI) [Aka89].

It has been shown that interstitial hydrogen is incorporated into GaN to form an H-Mg

acceptor complex, thus passivating the Mg acceptors. The H-Mg bonds can be broken

with LEEBI or high temperature annealing in an inert ambient [Pea96]. Acceptor doping

using Be has also been predicted [Ber97]. To understand ohmic contact formation, the

subjects of pinned Fermi level, surface cleaning, interfacial metallurgical reactions and

data on multilayer metallization schemes are discussed in this section.











2.3.1 Fermi Level Pinning

Semiconductors can be classified into two groups based on the dependence of

Schottky barrier height on metal work function. The ionic materials have a direct

dependence of barrier height on metal work function whereas covalent materials have a

weak dependence, as discussed in section 2.2 and shown in Figure 2.3.

From the ionic nature of GaN (an electron negativity difference of 1.23eV

between Ga and N atoms), the Schottky barrier height of GaN is expected to depend

directly on the metal work function. The Schottky barrier heights change with the metal

work function, but the changes are much less than expected for both doping types of

GaN. Mori et al [Mor96a] found that the change in Schottky heights is smaller than the

difference between the metal work function and the work function ofp-GaN for Pt, Ni,

Au and Ti metals. The Schottky barrier heights on n-type GaN are also found to depend

weakly on the metal work function [Guo95]. In both cases, deposition was made by

electron beam evaporation and the contacts were not subjected to heat treatment after

deposition. The metal work function was not the dominant factor affecting the Schottky

characteristics of Pt and Pd on n-GaN. Rennie et al studied the electrical properties of

various metal (Ti, Al, Sn, Cu, Zn, Mo, Ni and Pd) contacts to n-type and p-type GaN in

the hope of determining the relationship between metal work functions and barrier

heights [Ren98]. Contrary to the expected trend, the Fermi level was calculated to be

strongly pinned, with the effect being greater in the p-type material (Figure 2.9). The S

factor is shown to be only 0.01 for p-type and 0.21 for n-type GaN (in Figure 2.3), in

contrast to the expected S = 1 for a completely unpinned Fermi level (Schottky limit).










1.4

1.2-

0 -.t
A A


0.8 Pd
m5 Au
= Pb
0.6 A
^ Cr
m +
0.4
TI
+
0.2-
4.2 4.4 4.6 4.8 5 5.2 5.4 5.6 5.8
Work Function, eV


Figure 2.9 Barrier height vs. metal work functions on n-type GaN from various
researchers [CaoOO]



Another basic consideration is the sum of the p- and n-type Schottky barrier

heights for the same metal, which is found to be significantly less than the bandgap of

GaN. The measured values of Schottky barrier heights on p-GaN are likely to be affected

by the presence of interface states, damage and contamination. More research is

necessary to understand the interfacial conditions and properties of metal/GaN contacts.



2.3.2 Surface Preparation of GaN

The preparation of the GaN surface before metallization influences the electrical

characteristics of metal/GaN contacts [Kin96]. Cleaning the surface in solvents and

common acids or bases is effective in removing a significant fraction of the surface

oxides and other contaminants, but these procedures can not produce atomically clean









surfaces. Surface roughness, which can vary depending on the techniques and conditions

used to grow GaN epilayers, can influence the uniformity of contacts. One important

issue in the processing of ohmic contact to GaN is to understand the semiconductor

surface and to find appropriate chemical treatments to clean it before contact metal

deposition. Ideally, the metal/semiconductor contact should be oxide- and defect-free,

atomically smooth, uniform, and thermally stable.

The most common impurities on the surface after preparation of ohmic contact to

GaN include C, H and 0 from the ambient air, alcohol, methanol, acetone, water,

photoresist residues and sample handling, etc. [Abe96]. Their effects are discussed

below.



2.3.2.1 Hydrogen

A small amount of H2 in the carrier gas can passivate the electrical activities of

Mg and C acceptors in p-GaN during cooling-down from MOCVD growth [Sug98]. This

reduces the p-type doping levels unless an annealing step is performed. Annealing at 450

~ 500C can restore the free hole concentration but hydrogen atoms do not physically

leave the films until a higher (>800C) temperature is reached [Pea96]. Low resistivity p-

type GaN could be obtained by H2-free MBE growth without any post-treatment [Sug98].

Hydrogen is predicted to act as a donor (H) in p-type GaN, and as an acceptor

(H) in n-type material [Neu96]. Implantation of 2 H creates high resistivity materials

from both n- and p- GaN [Pea98].

For ohmic contacts to p-GaN, the Mg-H complex may be dissociated at any

period during processing. In principle, dissociation of the Mg-H complexes before or









after metallization should have no impact on final resistance. But for Pt/Au, Pt, Pd/Pt/Au

and Ni/Au contacts to p-GaN [Kin97], lower Pc values were obtained in the absence of

premetallization annealing. The post-metallization anneal simultaneously activates the p-

dopant and anneals the contact and/or contact interface. Annealing in N2/02 mixture

environment decreased the contact specific resistance for ohmic contact to p-GaN by a

factor of three [Suz99]. This is attributed to 02 reacting with H in GaN to form H20 and

uncompensated Mg acceptors. Some researchers disagree with this postulated

mechanism.



2.3.2.2 Carbon

Although carbon has been shown to produce p-type GaN, the hole concentration

obtained has been limited to 1017 cm"3 even though the carbon concentration was 1020

cm"3 or higher [Pan76]. It has been found in other HI-V materials that the maximum hole

concentration from carbon doping is related to the difference in bond strength between

the group III- and group V-carbon sites. In the case of InP, the carbon actually sits on the

group III site and acts as a donor resulting in an n-type material. This is also expected to

occur in InN and high In concentration alloys of InxGai-xN and InxAl.-.N grown by

MOMBE. Carbon has been shown to be a strongly n-type doping element for x > 0.15 in

InGal.xN and x 0.3 in InxAll.xN. As the In concentration is reduced, the tendency

increases for carbon to act as an acceptor rather than a donor. It has also been proposed

[Abe96] that C displays amphoteric doping behaviors in the nitride with acceptor

formation under some conditions (MOMBE grown GaN) and donor formation in other

cases (implantation in GaN and growth of In containing alloys).











2.3.2.3 Oxygen

From photoluminescence data [Sei83], it was concluded that oxygen is neither a

shallow nor a very deep donor. It has a moderately deep level (about 78 meV below the

conduction band edge at 4.2 K) which could form an impurity band near the conduction

band edge.

When 0 is implanted into GaN and annealed at 1100C [Abe96], it creates n-type

doping with an ionization level of-29 meV. Seifert et al. [Sei83] proposed that

substitutional incorporation of oxygen onto nitrogen sites could be the origin of intrinsic

free carriers in the growth of highly conductive n-type GaN. Because of the similarity of

their atomic radii, both substitutional oxygen and nitrogen vacancies could cause donor

defects.

Using an (Al, Ga, In) bubbler to purify the NH3 during the growth of GaN, a

dramatic reduction of carrier concentration in n-GaN is found due to the removal of

H20/02 from the NH3 [Chu92]. Seifert et al. [Sei83] used Mg3N2 to purify the NH3, and

also observed the reduction of as-grown carrier concentrations. Further verification of the

contribution of H20/02 to the increase of the carrier concentration comes from

experiments where water is introduced intentionally during growth. The carrier

concentrations of the water-injected samples are always above 10 2cm"3.

Chemisorption of oxygen on atomically clean and ordered GaN(0001l) surfaces

showed that saturation occurs at coverage of 0.4 monolayers [Ber96]. Low energy

electron diffraction (LEED) indicates an ordered adsorbate layer, and x-ray photoelectron

spectroscopy (XPS) peak of 0 Is suggests a single chemically distinct adsorption site.








The oxygen also is reported to react with GaN to form monoclinic P-Ga203 [Wol97] at

900C when GaN (both film and powder) is exposed to dry air. An interfacial reaction

mechanism is identified as the rate limiting step for oxidation, with an apparent activation

energy of -72 kcal/mole. The oxidation resulted in roughening of the oxide/GaN

interface and oxide surface.



2.3.2.4 Cleaning of Surface

As stated earlier, the metal/semiconductor interface should be inert, oxide- and

defect- free, atomically smooth and covered by epitaxial metal. For GaAs and other III-V

semiconductors, a thin oxide layer (-3 to 20 A) grows rapidly on the surface when

exposed to air, necessitating in-situ cleaning of the GaAs surface under ultra-high

vacuum (UHV) for the epitaxial growth of metal films.

For GaN, in-situ deposition of Ga metal followed by thermal desorption under

ultra-high vacuum is found to yield atomically clean surfaces using Auger electron

spectroscopy (AES) [Kha93]. This has been used in the study of Ni [Ber93] and Al

[Ber96b] films on GaN. In-situ nitrogen ion sputtering and annealing can also produce

atomically clean GaN surfaces [Ber96a, Hun93].

Ex-situ cleaning is usually used in practical metal deposition. Although solvent

cleaning and wet etching with common acids or bases cannot produce atomically clean

surfaces, they are effective in removing a significant fraction of the surface oxides and

other contamination. This results in relatively intimate metal/GaN interfaces. The effects

of aqua regia (HNO3: HC1 = 1:3), HCL: H20, HF: H20 and NH4OH and NaOH for

cleaning the surface of GaN are also investigated [Kin96]. A HC1 based solution is found








to be more effective in removing oxides and leaving less oxygen residue, but HF is more

effective in removing carbon and/or hydrocarbon contamination. The HCl and HF based

solution should be equally effective in removing the total contamination.

The importance of surface preparation is exemplified by cleaning in boiling aqua

regia [Kim98] for 10 min and then depositing a layered structure of 200 A Pd/5000 A Au

in a vacuum of 10-7 Torr on p-GaN (Nh = 2.98 x 1017 cm'3). A good contact resistance, Pc

= 4.3 x 10-4 2-cm2, was obtained on as-deposited samples. With no surface treatment,

samples deposited with the same metallization exhibited a high resistance of 2. Ix 10-2 Q-

cm2. The lower contact resistivity is attributed to removal of the surface oxide from the p-

GaN surface. Similarly, with a short time interval between the GaN film growth and

metal film deposition, a good ohmic contact is also possible as obtained in Ref. [Jan99]

even if the samples are only ultrasonically degreased with trichloroethylene, acetone and

methanol for 5 min each step. A contact resistivity of-3x10-3 0-cm2 is found for as

deposited p-GaN/Pt/Ni/Au samples,and a value of 5.1 x 10-4 Q-cm2 is reported after

annealing at 350C for 1 min in an inert ambient.



2.4 Interfacial Metallurgical Reactions

Interfacial reactions are critical to the formation of ohmic contacts to

semiconductors, whether they have a large or a small bandgap. Interfacial reactions can

lead to disruption of interfacial contamination layers consisting of native oxides,

hydroxides, and hydrocarbon/organic residue layers due to reaction or adsorption

[Hol97]. Although some work has been reported on contact schemes to p-GaN, little

information is available about the metallurgical reactions on the metal/p-GaN system.









Calculations about the transition metal-Ga-N systems have been performed on the

metallurgical reactions of metals with GaN in the absence of experimental studies

[Moh96]. According to the enthalpy of reactions to form gallides and nitrides, all

transition metals can be classified into three groups, the late, early and middle transition

metals. They correspond to the gallide-forming, nitride-forming and neutral metals

discussed for the "NOG" contact scheme in Chapter 4:



2.4.1 Gallide-forming&Metals

These are mainly group VIII metals. The group VIII metals are characterized by

the absence of intermediate phases in the metal-N binary systems with either positive or

small negative enthalpies of formation (Table 2.1). For Ru, Rh, Pd, Ir and Pt, no metal

nitrides have been reported, although these systems have not been investigated

thoroughly. The nickel nitrides are believed to be metastable under 1 atmosphere or lower

N2 pressures at and above room temperature, and osmium nitrides are not considered

because no thermodynamic data are available. In contrast to nitrides, group VIII metals

form many metal gallides, like NiGa, Ni2Ga3, etc.

Two types of tie line configurations are predicted for metal-Ga-N phase diagrams

for these gallide-forming (late transition) metals. Figure 2.10 shows the Ni-Ga-N tie-line

configurations with different N2 pressures as an example. Changing the pressure of N2

represented at the top comer of the isothermal phase diagram can alter the tie-line

configuration. In an annealing environment, N2 is also predicted to have effects on how

far the metal/GaN reaction can be driven in these systems.














N2 (1 atm)


N2(102 atm)


Ga(1)


NiGa


Ni2Ga3
(b)


N2(2 x 10-4 atm)


Ga(l)


Figure 2.10 Calculated Ni-Ga-N diagram at 600C. The nitrogen comer of the
diagram represents (a) N2 at 1 atm; (b) N2 at 10-2 atm and (c) N2 at 2 x
10- atm [Moh96]


Ga(1)









Higher reaction temperatures increase driving forces for reactions between these

metals and GaN. At 600C, a tie-triangle between NiGa, GaN and N2 gas at 1 atmosphere

is observed (Figure 2.10-a). Assuming that a Ni contact is much thinner than the

underlying GaN layer, Ni/GaN is favored to react under 1 atmosphere of N2 at

thermodynamic equilibrium. The entire contact should be converted to NiGa/GaN, and

nitrogen gas is released during the reaction. In the calculation, it was also shown that a tie

line exists between Ni2Ga3 and GaN in the phase diagram. However, the equilibrium

partial pressure of N2 over a Ni2Ga3-GaN contact would be less than 1 atmosphere. Thus,

there would be a driving force for a Ni2Ga3 contact to react with N2 at 1 atmosphere to

form NiGa and GaN. Of course, the reaction would be too slow to be observed, just as in

the case of GaN, which does not grow appreciably on liquid Ga at 600C under 1

atmosphere of N2, even though the reaction is very thermodynamically favorable.

The prediction of reactions between Ni and GaN is supported by experimental

data. The growth of thin Ni films on GaN [Ber93] is examined and a pronounced reaction

occurrs upon annealing above 600C in vacuum. Ga4Ni3 is identified by x-ray diffraction

(XRD) along with Ni in the as-deposited film [Guo96]. Although no reaction is found for

Ni on GaN in as-deposited state, after annealing at various temperatures between 400 and

900C, a trend of increasing Ga in the reacted films is observed with increasing

temperature [Ven97]. New phases consistent with Ni3Ga and NiGa are found.

The tie-line configuration predicted for these metal-Ga-N diagrams is strongly

affected by the stability of the metal gallides, the temperature of the isothermal section

and the N2 pressure of interest. Thermodynamic data shows that Pd and Pt gallides are










Table 2.1 Comparison of experimental and calculated values for the heat of formation
AHf"r of related gallides and nitrides. Published experimental values for the
entropy of formation, ASfOr, have been added when available. The units for AHfr
and ASfOr are kJ/mol and J/(K.mol) respectively [Boe88]


V Ga V2Ga5
V6Gas
V3Ga
V-N VN
V2N
Cr Ga CrGa4
Cr5Ga6
Cr3Ga
Cr-N CrN
Cr2N
Mn-Ga MnGa
Mn3Ga2
Mn N MnsN2
Mn4N
Fe Ga FeGa3
FegGa1
FeTGa6
Fe-N Fe2N
Fe4N



Co Ga CoGa3


CoGa


System
Sc Ga
Sc-N
Ti-Ga
Ti-N


Compound
ScGa
ScN
TiGa
TiN


ASfor


A frexp


- 157 (T unknown)


-169 (298 K)
-173 (298 K)
-11 (763- 953 K)
- 16 (763 953 K)
- 19 (763 953 K)
-109 (298 K)
-90 (298 K)
7.6 (850 K)
-4.7 (850 K)
-3.3 (850 K)
-53 (298 1800 K)
-31 (298 1800K)



-34 (298 K)
-26 (298 K)
20 (298 K)
33 (298 K)
-35 (298 K)
1.3 (298 K)
-16 (298 K)
2.2 (298 K)
2.4 (298 860 K)
-45 (298 K)
-29 (1100 K)
-41 (298 K)


AHcaic
-68
-184
-51
-146


-18
-28
-20
-76
-74
-8
-15
-12
-22
-35
-34
-34
-44
-31
-10
-16
-18
-11
-7


-18


-31


-48


-3.8 (763 953 K)
-5.1 (763 953 K)
- 13.8 (763 953 K)
-45 (298 K)
-33
+ 2.9 (850 K)
+ 4.8 (850 K)
+ 2.7 (850 K)
- 35 (298 1800 K)
- 17 (298- 1800 K)










Table 2.1 Continued


Co-N
Ni Ga


Co3N
Ni3Ga7
Ni2Ga3
NiGa


Ni3Ga2


Ni-N Ni3N
Ru Ga RuGa
Ru-N RuN
Rh Ga RhGa
Rh-N RhN
Rh5N
Pd-Ga Pd3Ga
Pd88Ga12
Pd-N PdN
PdsN
Os- Ga OsGa
Os-N OssN
Pt Ga Pt3Ga
Pt94Ga6
Pt-N PtN
PtsN


- 32 (298 K)
- 36 (1173 K)
- 38 (1173 K)
+ 2 (298 K)
- 34 (300 K)
-45 (300 K)
- 38 (300 K)
-47 (298 K)
-43 (1023 K)
-43 (1223 K)
- 36 (300 K)
-45 (298 K)
+ 0.2 (292 K)








- 59 (1000 K)
- 27 (1000 K)




-28 (298 K)


-44 (1000K)
6.7 (1000K)


5.4 (1173 K)


+1
-37
-33
-37



2.9 (1223 K)
-37


+6
-34
+49
-53
+46
+11
-51
-24
+48
+12


+12
-46
-11
+62
-18









particularly stable among the group VIII metal gallides. The reaction of these metals with

GaN is thermodynamically favorable. They form metal gallides and release N2 gas, even

under 1 atmosphere of N2 at room temperature, although these reactions may be

extremely slow at this temperature.



2.4.2 Nitride-forming Metals

Be, Mg, Ca, Sr, Ba, Al and transition metals like Sc, Ti, V, Y, Zr, Nb, La, Hfand

Ta are nitride-forming metals. Si can also be classified into this group because of the

large enthalpy of the Si-N reaction. The majority of the non-transitional metal-nitrides

have low electrical conductivity, hence only the transition metals are considered [Table

2.1]. In contrast to the late transition metals, these early transition metals form metal

nitrides of considerable thermodynamic stability. These metal-N binary systems all

contain a refractory metal mono-nitride (MN phase). Some of them actually exist over a

wide range of compositions and some systems contain nitride phases besides mono-

nitride.

In all of the systems for which no ternary phases have been reported (Sc, Y, Zr,

La and Hf metals), the common feature in the calculated phase diagrams is a tie line

between the MN phase and GaN. Such a tie line would also be expected for the (Nb, Ta,

V, and Ti)-Ga-N systems, as long as tie lines to any ternary phases do not alter this

situation. At room temperature, the predicted MN-GaN tie lines are stable against

competing tie lines by at least 40kJ/g-atom, due to the enhanced thermodynamic stability

of the MN phases. Even at 600C, these tie lines still represent a negative enthalpy of

reaction of more than 15kJ/g-atom. Figures 2.11-a and b. show the calculated Zr-Ga-N










N2 ( atm) N2 (I atm)



Ga _____ J GaN /______ a

Ga(l) Zr GaNLaN

Ga(I) Ga(I) LaGa La

Figure 2.11 Calculated (a)Zr-Ga-N and (b) La-Ga-N isothermal diagram at 298K
[Moh96]


and La-Ga-N diagrams. To simplify the diagrams, the ranges of homogeneity of the

binary phases are neglected. Taking ZrN as an example, there is not enough data to

predict the exact compositional range in equilibrium with GaN. In contrast to the results

of gallide-forming metals, the temperature and N2 pressure represented in the nitrogen

comer of the diagram have less dramatic effects on the calculated phase diagrams, at least

over the range of temperature and pressure normally encountered in contact processing.

The phase diagram for the Ti-Ga-N system at 800C is shown in Figure 2.12

[Moh97]. According to this diagram, if a N2 pressure of 1 atm is maintained continuously

over the contact, the Ti/GaN contact would be thermodynamically favorable to extract

nitrogen from the annealing environment, ultimately resulting in a TiN/GaN contact. This

result is expected because only TiN and GaN are simultaneously in equilibrium with N2

gas at 1 atm. The contact would therefore come to equilibrium with GaN without any net

consumption of GaN through an interfacial reaction. However, there may be competition













N2(latinm)





GaN ................. TiN
Ti2N

/ Ti2GaN



Ga(l) -- g Ti



Figure 2.12 Calculated phase diagram for Ti-Ga-N at 800C [Moh97]



between nitrogen incorporation into the film and metallurgical reaction at the non-

equilibrium Ti/GaN interface. Because the Ti contact surface is usually covered by a

protective metal layer (like Au), the reaction of metals with the annealing environment is

usually hampered. A Ti contact would react with GaN to ultimately form TiNx and leave

liquid Ga remaining on the GaN. This situation is more favored when contacts are

annealed under a lower N2 partial pressure.

Reactions of metals with GaN to form nitrides plus Ga is expected to decrease the

hole concentration at the metal/p-GaN interfacial region due to a subsequent increase of

N vacancies in the interfacial region. However, before the formation of stable nitride,

these metals could allow more nitrogen atoms to diffuse into the semiconductor/metal

interface. This can decrease the nitrogen vacancy concentration and is expected to









increase the hole concentration. So, the contact could be expected to exhibit low

resistance at short time, but degrade to higher resistance at long time after the formation

of stable nitrides. Literature data are discussed relative to this concept in Chapter 5.

Specifically, Ta/Ti metallization forms low resistance contact to p-GaN, but degrades

after a few days [Suz99]. This is also believed to be the reason that Ti, Al and Ti/Al

forms good ohmic contact to n-GaN [Les96, Cor98, Ruv96]. Interfacial reactions should

also contribute to ohmic contact of W and WSix to n-InGaN and n-InN [Var97a, Var96a,

Var97b].



2.4.3 Neutral Metals

Neutral metals include Cr, Mn, Mo, Tc, W and Re. Cr and Mn can form

compounds to both Ga and N (see appendix A). The Mo-Ga, W-Ga and Re-Ga binary

phase diagrams are characterized by an absence of intermediate phases under

atmospheric pressure and negligible miscibility between liquid Ga and the metals. Figure

2.13 shows the W-Ga-N and Re-Ga-N diagrams at 600C. Both W and Re are expected

to be in thermodynamic equilibrium with GaN at room temperature and 600*C. Although

W2N, WN and ReNo.43 are reported to form, none of these nitrides are expected to be

stable at 600C under 1 atmosphere or lower N2 pressures. Mo is predicted to be more

like a gallide-forming metal because a tie-triangle is predicted at 600C between a metal

gallide, GaN and N2 gas at 1 atmosphere in the Mo-Ga-N diagram.

The Cr is found to form a contact with linear I-V curves to p-GaN after annealing

at 900C for 15 sec [Tre97]. This is speculated to result from the reaction of Cr with both

Ga and N. This dual reaction may improve the carrier concentration in the metal/p-GaN









N2 (I atm) N2 (I atm)




GaN O GaN





Ga(l) W Ga(l) Re
(a) (b)

Figure 2.13 Calculated (a) W-Ga-N and (b) Re-Ga-N diagram at 600C [Moh96]



contact region under some condition, but this improvement may vary in performance

with the change of temperature and time.



2.4.4 Thermal Stability

An increase of operating or processing temperature would allow Mg-H complexes

to dissociate (to reactivate the p-type dopants), thus the carrier transport via thermionic

emission could increase as temperature increased. The kinetics of interfacial reaction may

be also increased. All these effects would improve the ohmic contact quality.

The temperature is shown experimentally to have a strong influence on the

contact properties. Detailed studies of the electrical properties of the Pt/Au contacts to p-

GaN revealed that the I-V linearity improves significantly with measurement at higher

temperatures [Kin97]. At room temperature, a slightly rectifying I-V curve is observed,

while at 200C and above the I-V curve is linear. When temperature is increased from

25C to 350C, the specific contact resistance is found to decrease by nearly one order of








magnitude. A minimum Pc of 4.2x 10-4 Q.cm2 is obtained for a Pt/Au contact at 350C.

The behavior ofPt, Pd, Ni is also studied on n-GaN as a function of annealing

temperature [Dux97]. The Pt film began to form submicron spheres and islands after

annealing above 600C, and the Pd and Ni films changed their morphology to islands for

Pd/GaN and Ni/GaN interfaces, respectively. Delamination occurred at Pd/GaN

interfaces upon annealing above 700C, but no delamination occurred at Ni/GaN

interfaces because of a Si02 cap layer on the Ni film. No structural changes are observed

below these temperatures using XRD and RBS analysis.

The delamilation of metal contacts on GaN was explained with the concept of

thermal stress generated by differences in thermal expansion coefficients between metals

and GaN. The surface energy and linear coefficient of expansion (6x10-6 K'I) of GaN are

lower than most metals. Since the thermal expansion coefficient of Pd (11.7x 10 -6 k') is

almost twice that of GaN, the change between room temperature and the annealing

temperature generates an extension stress causing delamination. It may be necessary to

use metals with low surface energies and low thermal expansion coefficients to avoid this

delamination problem. Stable, non-reactive contacts that can withstand high temperatures

are desired.

It is believed that the current is governed by thermionic emission in current p-

GaN samples with carrier concentration level of 1017 cm"3. The improvement in I-V

linearity at high temperature is attributed to the increase in thermal energy of carriers

which enables them to activate over the barrier. On the other hand, because the large

ionization energy for Mg, -250 meV as shown in Figure 2.8, the percentage of activated

Mg acceptors is low. The increase of operating energy would increase the activation









percentage of Mg acceptors, and improve the hole concentration. Increased carrier

density could lead to field emission or thermionic field emission as the dominant

mechanism of charge transport. As temperature increases, the I-V linearity improves and

the contact resistance decreases significantly. At temperature above 200C, the I-V curve

exhibits ideal ohmic behavior and the resistance is constant with current.



2.5 Metallization Schemes and Analysis

Of significant interest in the improvement of GaN devices is a p-type ohmic

contact of low resistance. Common reported values are around 10-2 ohm-cm2 but a lower

limit has been reported of 3x10-5 Q-cm2 for a doping level of7x 1017cm"3 [Suz99].

Recently, contact schemes with better performance have been demonstrated. These

include oxidized thin Ni/Au [Ho99b, Koi99] and polarization-charge-based contacts

[LiOO]. In this section, the current published contact schemes are compared and the

mechanisms analyzed. Based on these contact mechanisms, the contact schemes are

classified into conventional or non-conventional methods. Conventional methods mean

formation of contacts via interfacial reactions to improve doping levels in the contact

region in the GaN epilayer. Non-conventional methods used innovative designs to

increase the carrier concentrations other than by interfacial metallurgical reactions.



2.5.1 Conventional Contact Schemes

These contacts fall into the scheme of the "NOG" concept developed below in

Chapter 5. They will be analyzed in the discussion section in that chapter. In Table 2.2,

the contact schemes, processing condition and contact resistivities are summarized.










Table 2.2 Current metallization schemes of ohmic contact to p-GaN


Metalization (*)
Au250
Aul00
Au32/Mg32/Aul70


Co80/Aul00
Cr20O/Au300
Cr20/Au300
Cr20O/Au300
Cr50/Au100
In200/AulOO
lnl5/Zn20/In55/AulOO
Mgl2O0/Au100
Mn50/AuO 100
NiI00
Ni8O/Aul 00
Ni80O/Au.lOO
Ni20/Au500
NilO 0/Au4O0
NiiO/AuS5
Ni5/Au5
Ni80O/AulOO
Ni?/Au?
Nil 5/Crl5/Au500O
Nil 5/CrI 5/Au500
Ni25/Mg5/Ni25/Si240


Ni25/Mgl I/Ni25/Si240


Ni25/Mg8/Ni25/Si240


6.6x10-3


2.2x. 10-3


pc, Q-cm2
53
0.026
214


8.1x10-3
1.2x104
2x10-'
lxl0-3
<4.3x100'
1.5x10-'
1.7x10"'
Bad
Bad
0.015
1.8x10'
3.1xl10-2
3.4x 10-'
_10-2
<10.4

4x 10-6
3.8x10-'
1.2x10-2
-8.3x10-'
8.3x 10.2
9.6x 10-4


Nd(cm3)
lxl018
Mg Ix 1020
lxl018


1.6x 1017
1.4x 1020
2.7x1018
1.4x1020
9.8x 1016
1.6x1017
3.1x1017
1.6x1017
1.6x 1017
Mg Ixl020
1.6x1017
3.1x10'7
0.1-1xl017

8.4x 1017
2* 1017
2* 101'7
2.2x1017
5.5x1016


Comment
800C,10 min in N2
AD, ImA bias
Linear AD, more resistive
w/750C,15 sec in N2
AD
500C, 1 min in N2
700C, Imin in N2
AD
900C, 15 sec in N2
500C, 1.5 min
RT-500C, 1.5 min
Up to 850C
Up to 850C
AD, ImA bias
500C, 1.5 min
500C, 1.5 min
500C, 30 sec
500C,10 min in vac
400-500C, 10min in 02
500C, 10mmin in air
500C, 1.5 min
ImA bias,700C, 10 min


Refer
[Sim96]
[Mor96]
[Smi96b]


[Fun99]
[Yoo96]
[Yoo96]
[Yoo96]
[Yoo96]
[Fun99]
[Fun99]
[Fun99]
[Fun99]
[Mor96]
[Fun99]
[Fun99]
[Kim97a]
[Ish97]
[Ho99a]
[Ho99b]
[Fun99]
[Kin97]
[Kim97a]
[Kim97a]
[Kam98]


[Kam98]


[Kam98]


0.1 -Ix.1017 500C, 30 sec inN2
0.1 -lx.1017 AD
3x 10'7 400C, 10min PR/30 min
RTA in N2
3x1017 400C, lOmin PR/30 min
RTA in N2
3x 1017 400C, lOmin PR/30 min
RTA in N2









Table 2.2- continued


Ni25/Mg8/Ni25/Si240 2.5x 103
Ni25/Mg8/Ni25/Si240 4.5x10"3
Ni8/Zn132/AulOO 2.5x10-'
Ni45/Zn-Au46 3.6x 10-3
Ni2.5/ZnlO/IlnlO 2.3x10-'
/Ni47.5/AulOO0
Pd20/Au500 9.1 x 10-3
Pd20/Au500 9.1 xl 0-3
Pd20/Au500 2.9x 10.2
Pd20O/Au500 4.3x 104
Pd20/Au20/Pd20/Au500 -9xl0-'
Pd20/Au20/Pd20/Au50O0 -9x 10-3
Pd20/Au20/Pt2O/Au500 -9x 10-3
Pd20/Au20/Pt20/Au500 Mid 10.2
Pd?/Pt?/Au? 1.Ox 10.2
Pd?/Pt?/Au? 1.7x10"2
Pt? 3.4x10.2
PtlOO 1.3x10-2
Pt?/Au? 5.7x 10-3
Pt2O/Au30_0 -1.8x 10-3
Pt20O/Au300 2x10-'
Pt/Au 4.2x104
Pt20/Ni30/Au80 -3x l03
Pt20/Ni30/Au80 5.1 x10-4
Ta50 >103
Ta60/Ti40 3x 10'
Ti50 >10-3
TilOO 3.5x10.2
Ti20/Pt80/Au30Q 4x 10-4
Ti20/Pt80/Au300 2.5x 10-3
Zn50/AulOO 4.9x 10.2
Metallization thickness unit: nm
?: Not reported;
AD: As Deposited.


3x1017
3x1017
3.1x1017
4.4x 101'7
2.2x 1017


450C, 30 min in N2
500C, 30 min in N2
400C, 1.5 min
600C, 2 min in N2
600C, 1.5 min


9x 1016 500C, 30 sec inN2
9x 1016 500C, 30 sec in N2
2.98x 10'7 AD, no clean, 3x107 Torr
2.98x 1017 AD,
9x1016 AD
9x10'6 500C, 30 sec inN2
9x1016 AD
9x1016 500C, 30 sec in N2
5-6x10'6 lmAbias, 700C, 10min
5-6x1016 lmA bias, AD
5.5x10'6 lmAbias, AD
Mg lxl020 AD, lmA bias
5.5x10'6 750C, 10min, cap/GaN
1.4x1020 AD
2.7x108 700C, Imin in N2
5-6x1016 Measured at 350C
3x1017 AD
3x1017 350C, Imin in N2
7x 1017 800C, 20miniVac
7x 1017 800C, 20minVac
7x10"7 8000C, 20min/Vac
Mg lxl020 AD, ImAbias
1.4x1020 AD
2.7xl018 700C, Imin in N2
1.6x 1017 7500C, 1.5 min


[Kam98]
[Kam98]
[Fun99]
[You98]
[Fun99]


[Kim97b]
[Kim97b]
[Kim98]
[Kim98]
[Kim97b]
[Kim97b]
[Kim97b]
[Kim97b]
[Kin97]
[Kin97]
[Kin97]
[Mor96]
[Kin97]
[Yoo96]
[Yoo96]
[Kin97]
[Jan99]
[Jan99]
[Suz99]
[Suz99]
[Suz99]
[Mor96]
[Yoo96]
[Yoo96]
[Fun99]










2.5.2 Non-conventional Contact Schemes

Recently the use of thin Ni/Au to p-GaN for both low resistance and high-

transparency ohmic contact has been extensively studied [She99b, Koi99, Ho99a, Ho99b,

Che99, Mae99]. Transparent NiO is a p-type semiconductor with a wide bandgap, which

varies from 3.6 to 4.0eV [Sat93]. Lower specific contact resistance values (
cm2 for 100OA Ni/50AAu and later of 4x10-6 Q-cm2 for 50A Ni/50A Au) are obtained

[Ho99a, Ho99b]. Some Au-based contacts (Ni/Au, Co/Au, Cu/Au, Pd/Au and Pt/Au)

annealed in an 02 partial pressure are also found to reduce the contact resistance, Pc, and

the sheet resistivity ofp-GaN epilayers (Ps) [Koi99].

While significant progress has been made on ohmic contacts to p-GaN by

oxidizing the thin Ni/Au, different mechanisms have been suggested to explain the

results. Murakami's group [Koi99] ascribed the low contact resistance to formation of an

intermediate semiconductor layer (ISL) with high hole concentration caused by removal

of hydrogen atoms which were bonded with either Mg or N atoms in the p-GaN epilayer.

Ho et al [Ho99a, Ho99b, and Che99] explained the results as being due to the formation

of a p-NiO layer directly on the p-GaN layer surface. This NiO layer is believed to act as

a low barrier ISL and small islands of Au bridged the current path via electron tunneling.

This microstructure resulted in the reduction of the Schottky barrier height (4) at the p-

GaN/metal interface. However, Murakami's group [MaeOO] further studied a variety of

Ni/Au based contacts, such as 500A NiO/500A Au, 500ooA NiO (Li)/500A Au, 50A

Ni/500A NiO (Li)/50A Au, 50A Ni/200A Li20/500A NiO/500A Au and 50A Ni/200A

Li20/50A Ni/500A NiO/500A Au (much thicker than the 50A Ni/50A Au in Ho's work),









using a method of sputter deposition and annealing in an 02 ambient for 5min in the

temperature range of 300 to 500C. They found that all contact schemes with NiO mid-

layer did not reduce the pc to values lower than the conventional Ni/Au contact annealed

in N2 ambient. From these results, they concluded the p-NiO did not act as the ISL to

reduce the Schottky barrier height at the p-GaN/Au interface.

Although, the resistivity of NiO can be decreased by an increase in Ni3+ ions, or

addition ofmonovalent atoms, such as lithium, or increased nickel vacancies and/or

interstitial oxygen concentration [Ant92], NiO usually remain very resistive [Sat93].

Stoichiometric NiO is an insulator with a resistivity of the order of 1013 Q-cm at room

temperature [Adl70]. Because of the insulation of NiO, the poor contact performance in

the work of [MaeOO] might be due to the large thickness of NiO film and non-epitaxial

sputtered NiO in the contacts, resulting in larger resistance in the contact. The data from

those contacts with thick NiO layers may be insufficient to evaluate the role ofp-NiO, as

proposed by Ho [Ho99b].

An interesting scheme was proposed to form ohmic contacts to p-GaN using

internal electric field caused by polarization effects [LiOO]. Due to spontaneous and

piezoelectric polarization effects, sheet charges can be induced in the AlxGal-xN/GaN

materials system, and low specific contact resistance can be attained in contacts based on

polarization fields at lower doping levels. In this way, a polarization-charge-based

contact becomes a viable alternative to the formation of ohmic contacts to p-GaN.

Experimental work showed that a spontaneous polarization vector could be formed and

pointed to the substrate in the Alo.2Gao.gN/GaN superlattice structure. Ni contacts to this






45


structure resulted in a specific resistivity of 9.3x 104 Q-cm2 after annealing at 400C for

300sec.













CHAPTER 3
EXPERIMENTAL PROCEDURES

3.1 Introduction

This chapter describes the experimental procedures that were followed for sample

cleaning and preparation of electrical contact to p-GaN. The procedures consisted of an

initial cleaning of the samples followed by metal contact deposition. These contacts were

then heat treated and characterized in terms of their electrical properties, surface

composition, surface morphology and interfacial reaction products.



3.2 Contact Preparation

The p-GaN films used in these experiments were either purchased from SVT

Associates grown with a Molecular Beam Epitaxy technique (referred to as MBE-GaN,

1 ptm-thick) or grown by low pressure Metalorganic Chemical Vapor Deposition

(referred to as MOCVD-GaN, 2.5p.m-thick). Mg and Si were used as doping elements in

p- and n-type GaN, respectively. The SVT 1 .tm-thick p-GaN epitaxy film was grown on

the c-plane of sapphire substrates with an rfN2 plasma and a solid Ga source. The

substrate temperature was of 700C and the growth rate was of 0.5pm/hr. A 350A-thick

A1N was used as a buffer layer before the deposition of a 1p m thick p-type GaN film

using elemental Mg as the p-type dopant. The GaN film was as-grown p-type with no

post-growth activation. SIMS data showed about 1020 cm"3 for the actual level of Mg in

the film. A hole concentration (p-type) of 1.1 3.5 x 1017 cm-3, and an electron









concentration of 7 9 x 1018 cm3 (n-type) were measured by Hall measurement at room

temperature for the MBE- and MOCVD- GaN.

For the H202 treatment experiments, samples (5 mm x 5 mm) were cleaned with

ultrasonically agitated acetone (5 min), methanol (5 min) and boiling aqua regia (10 min)

sequentially before being flushed with DI water. All samples were blown dry with N2 gas

between each step. These samples were then quickly placed into an electron beam

evaporator system for contact deposition (using the "NOG" metallization principles as

developed in Chapter 5), through a van der Pauw shadow mask [10] at a base pressure of

<10-6 Torr. One sample was not treated further, i.e. kept in the as-cleaned state with

contacts deposited. Other samples (with contacts on the surface) were immersed in

H202/H20 solutions (1:1 and 1:5) for 30 sec or 300 sec followed by DI water rinse, N2

blow dry, and a short time (<60sec) dry at 80C in air.

The contact metals used in the "NOG" scheme consisted ofNi/Au, Ni/Ti/Au,

Ag/Ti/Au, Ni/Al/Au, Pt/Au, Pt/Si/Pt/Au and Pt/Mg/Pt/Au with the first listed metal being

deposited first and adjacent to the GaN. The Ni/Au and Pt/Au were mainly used for

comparison, and the first layer of either gallide-forming metal (Ni, Pt), or neutral metal

(Ag) was used to study the effects of different gallide-forming metals on the contact

resistance. Ti, Al, Mg, Si were used as the second layer because they are nitride-forming

metals. It is difficult for Al to absorb H from the GaN lattice, therefore metallization with

Al were evaluated to test whether the contact resistance was decreased when using an Al

layer. If it is, this would argue against H extraction postulated by Suzuki et al [Suz99]. Al

reacts readily with 0 to form an insulator, A1203, which is expected to increase the

specific contact resistance. Mg was used because it is the p-type doping element in p-








GaN. It was postulated that Mg in the metallization would increase the hole concentration

in the contact interface region. Si was tested because of its strong tendency to form

nitrides (Deh93).

All samples were degreased prior to deposition using ultrasonicated acetone

followed by methanol, each for 5min., and blown dry by N2. Any native oxide was then

removed using boiling aqua regia for 10min followed by a 5min DI rinse and N2 blow

dry. Either regular (square) TLM patterns or circular TLM (CTLM) patterns, discussed in

Chapter 2, was used for contact resistivity measurements. For square TLM patterns, the

leakage path was isolated with a technique of inductively coupled plasma (ICP) dry

etching using Ar + Cl2 + N2 plasma under a pressure of 5 mTorr. For the CTLM patterns,

the outer ring contact of the mask has a diameter (2 R) of 350/rm and the inner dot

contact's diameters vary (2 r) between 340 to 310 /anm to result in contact distances of 5,

10, 15 and 20p/m. Ohmic contacts are characterized by plotting the total resistance versus

the contact spacings. The specific contact resistance and sheet resistance were derived

from this plot as discussed in Chapter 2. The CTLM pattern was transferred to the

multilayer metal contacts by a photolithographic lift-off process. Positive photoresist

(AZ1529) were used in these experiments. Prior to metal deposition the samples were

etched with 10% diluted HF for 30sec to remove native oxides. The samples were then

immediately introduced into the vacuum chamber for contact deposition.

For contacts used in Hall measurements, the Van der Pauw configuration [Van58]

was employed to determine both the Hall coefficient and the resistivity of the GaN films,

as shown in Figure 3.1. Dot contacts of multiple metal layers of25A Ni/500oA Ti/500A

Au and 100oA Ni/500A Ti/500A Au were used to form the ohmic contacts to p-GaN.


























Figure 3.1 Configuration of contacts used in Hall measurement.


These contacts were deposited through a shadow mask with the dot size of <0.4mm, and

the distance between two contacts of 4mm.

All contacts were deposited in an electron beam evaporation system with a glass

bell jar. The system was pumped with a Varian oil diffusion pump backed by a two-stage

mechanical rotary vane pump, providing a base pressure of 10-6 low 10-5 Torr. The

metal charge in the electron beam well consisted of metal pellets with the following

purities: Ni (Target Materials Inc., 99.98%); Ti (Cerac, 99.95%); Au (Materials Research

Corporation, 99.95%); Pt(99.95%, Johnson Matthey, Inc.), Al (Cerac, 99.99%), Mg

(Cerac, 99.99%), Si (semiconductor grade) and Ag (Cerac, 99.99%). The metal layer

thickness was monitored using a quartz crystal oscillator. The contact thickness varied

with different experiments and will be noted in the results sections.

For RTA (Rapid Thermal Annealing) experiments, all metallization schemes were

heat treated in a custom 50cm quartz tube with a 25cm hot zone and flowing N2 or 02









(high purity, 99.995%) as the ambient. The gas flow rate was monitored with a

MANOSTAT flow meter with a typical setting of 110 standard cubic centimeter per

minute (sccm).



3.3 Characterization

Contacts on GaN were characterized in the as-deposited state and following each

of the above described heat treatments. Their electrical properties (I-V) and surface

composition, surface morphology and interfacial reaction products were also

characterized. A group of characterization techniques of I-V, light transmittance

measurement, Hall measurement, Auger electron spectroscopy (AES) [Bru92], X-ray

photoelectron spectroscopy (XPS) [Sib96], scanning electron microscopy (SEM) [Lee93]

and transmission electron microscopy (TEM) were used in this study.

The electrical properties of all contacts were investigated using room temperature

current-voltage (I-V) measurements between two front surface dot contacts or the

concentric dot/ring pattern described above. The I-V data were obtained by measuring the

current flow between two adjacent top contacts under an applied bias. The ohmic or

rectifying nature of the contacts could be determined by the linearity of the I-V curves

and total resistance. The reverse-bias breakdown voltage of rectifying contacts could also

be determined from their I-V characteristics.

To evaluate the possibility of increased electrical conductivity from surface

leakage current after the H202 treatment in Chapter 4, all of these samples were measured

after contact deposition, and then cleaned with acid (1HC1 + 2H20) followed by base

(1KOH + 2H20) solutions for 10 min in each step, and rinsed with DI water after each







cleaning step. The I-V data were collected again without additional treatment. As shown
below, the I-V data were unchanged by this treatment.
The light-transmission characteristics of the Ni/Au and Ni/Au/Ni contacts (in
Chapter 5) were measured with a commercial Zeiss UV grating monochrometer over the
wavelength range of 300 to 700nrim. The light transmission through the GaN film with
and without the metal contact were measured on each sample at either a fixed or variable

wavelength (Figure 3.2). The ratio of these two transmission values (Ii/12) at X = 450nm

was reported as the light transmittance through the Ni/Au "transparent" ohmic contacts.
The repeatability of the data were checked with three sets of data obtained from each
sample. The data reported are generally the average of these three measurements.

For Hall measurements, a computer controlled MMR commercial measurement
system was used. The Van der Pauw method [Van58] was employed to determine both
the Hall coefficient and the resistivity of the films. Dot contacts of 100OA Ni/500A Ti/
500A Au were used for p-GaN. All were deposited with an electron beam evaporator





Metal contact 12

GaN on sapphire



t 0 t


Figure 3.2 Schematic of light transmittance measurement









through a shadow mask. The dot diameter was of 0.4 mm, and the distance between two

contacts was 4 mm. A magnetic field of 3 kilo-Gauss was chosen automatically by the

computer program to improve the accuracy of the Hall coefficient and resistivity values.

Scanning electron microscopy (SEM, JEM 6400) was used in the secondary

eletron or back scattering eletron mode to characterize the surface morphology of the

grown films as well as the deposited layers in an attempt to determine the microstructure

evolution and possible flaws that could cause high contact resistance.

Transmission electron microscopy (TEM, JEOL 200CX) was performed with 200

keV acceleration voltage for plane view analysis after H202 solution treatment to detect

any defects in MBE-GaN

The surface of the GaN film was characterized by atomic force microscopy using

a Nanoscope III system from the Digital Instruments, Inc. The AFM was operated in the

tapping mode, and the height data was used in this work.

Auger electron spectroscopy (AES) [Bru92] was used to measure the elemental

composition of the atoms in the surface region using a Perkin-Elmer PHI Model 660

scanning Auger microprobe (SAM). AES surface N(E) survey spectra from these GaN

samples were recorded over the energy range of 50 to 2050 eV using a 5 keV, 30 nA

electron beam with a diameter of-1 jm. Depth profiles were also collected. With this

technique, all elements of the layers can be determined except for H and He and generally

down to a value of 0.1- lat%. The interfaces of the layers can be analyzed with a

resolution approaching 100OA to determine possible compound formation.

X-ray photoelectron spectroscopy (XPS) [Sib96] or electron spectroscopy for

chemical analysis (ESCA) is an analytical technique similar to AES except the incident





53


energetical beam is x-rays. This technique is usually used for chemical state identification

of surface species. XPS was used in this study to provide information on chemical

bonding and compound formation that were not apparent from AES profiles. These

measurements were performed with a Perkin-Elmer PHI Model 5100 ESCA system. Two

large samples were used (10 x 10 mm) in this analysis. The ESCA data were collected

using an Al X-ray source (Ka, hv = 1486.6 eV) and a hemispherical analyzer set at a pass

energy of 35.75 eV. For these studies the ESCA source area was 4 mm x 6 mm.

Secondary ion mass spectrometry (SIMS) was used to determine impurity levels

at the surface or in the films through depth profiling and was used to measure trace

dopant profiles through the structure. SIMS is capable of quantifying the impurity levels

for many elements depending on their detection sensitivity. However, SIMS has a limited

possibility of providing useful information following heat treatments when diffusion

distance are large. In this work, SIMS was used to detect the oxygen and hydrogen

impurity levels in the H202 treated samples. This analysis were performed using a Cs

primary ion beam and negative secondary ion detection. The raster size was 250 x 250

2
pm and the counts were taken only from the central 75 x 75 pr.m .













CHAPTER 4
EFFECTS OF H202 SOLUTION TREATMENT ON p-GaN



4.1 Introduction

As described in Chapter 2, the high vapor pressure difference between gallium

and nitrogen in GaN could lead to preferential loss of nitrogen and a gallium-rich surface.

In this Chapter, stabilization of the surface of MBE grown GaN with H202 solutions is

demonstrated. It is found that the H202 treatments are able to increase the carrier

concentration by a factor of two. and lead to higher current levels through contacts for

MBE p-GaN.

As discussed in Chapter 2, oxygen is reported to react with GaN to form

monoclinic beta-Ga203, with many polytypes being observed [Wol97, Roy52]. Annealing

GaN in a 02 ambient is reported to dissociate the Mg-H complex and reactivate the Mg

acceptors [Suz99]. Also, cleaning in common acids or bases or solvent is effective in

reducing the amount of surface oxides and other contamination. The effects of aqua regia

(IHNO3 + 3HC1) [Kim98], HCl [Kin96], HF [Kin96], KOH [Lee99], and (NH4)2S

[Cao00] on removing of native oxides on GaN have been reported.

This chapter describes how H202 solution was used to clean/passivate the p-GaN

film after contact deposition. Hydrogen peroxide was shown to increase the epilayer

conductivity. Hydrogen peroxide is an active oxidant, and may be expected to react with

GaN to form gallium oxide and/or gallium hydroxide to break up the Mg-H complexes

and reactivate the Mg acceptors. These effects will be discussed.









For I-V and Hall measurements, two types of metallization were used. For the

treatments using 1H202:5H20 (volume ratios hereafter referred as 1:5), and 1:1 solutions,

a contact scheme of25A Ni/500A Ti/500A Au was used, where the first metal layer is

deposited onto the GaN epilayer. For treatments with 5:1 and "pure" (37%) H202, the

contacts are 100OA Ni/500A Ti/500A Au, because this contact scheme led to higher

current and more linear I-V curves, as discussed in Chapter 5.



4.2 Modification of Electrical Conductivity

4.2.1 Effects ofH2O2 Concentration

Remember, for these experiments, dot contacts were deposited onto GaN epilayer

as described in Chapter 3 using a shadow mask., then the entire sample, including the

contacts were treated with H202 solution. The I-V results after immersing the sample in

1:5 or 1:1 solutions are shown in Figure 5.1. The current transported in the samples was

found to increase as the soaking time and H202 concentration increased from 30 to 300

sec and from 1:5 to 1:1, respectively. Compared to the as-cleaned state, an increase of

~100% in the magnitude of current was found after immersion in the 1HzO22:lH20

solution for 300sec as shown in Figure 4.1 a.

The I-V curves after immersion in the 5:1 and "pure" H202 treatment are shown

in Figure 4.2. Straight, linear I-V curves were found for as-deposited contacts of 100A

Ni/500A Ti/500A Au treated with concentrated H202 solution (Figure 4.2), as compared

to those with the 25A Ni/500A Ti/500A Au contacts and treated with a solution with

lower H202 concentration (Figure 4.1). The highest currents were achieved in the 5:1

treated samples. In the as-deposited state, after a "pure" H202 treatment, the highest










50-
40 (a)

30
20
=L10-
0o


-20

-30 As cleaned
-40 ol 1H202:5H20 for 30sec
-50 4 - 0 .. .
-5 -4 -3 -2 -1 0 1 2 3 4 5


Voltage, Volt


60

40

,:20
::L
S20

I,
0 -20

-40

-60


-5 -4 -3 -2 -1 0 1


2 3 4 5


Voltage, Volt


Figure 4.1 Effects of H202 solution treatment on the I-V curves of25A Ni/500A
Ti/500A Au to MBE p-GaN. (a) 1:5 solution; (b) 1:1 solution









200
(a)
150




100-
<50-

C -

0-501

-1001*Aclae
a 51-202+1 H20 for 30sec
-150
A 5H202+1HR20 for 300sec
-200. .., ,
-5 -4 -3 -2 -1 0 1 2 3 4 5
Voltage, Volt


100
(b)

50-



C 0 -

50 As cleaned
a pure H202 for 30sec
A pure h2o2 for 300sec
-100 1 1 1 1 ---
-5 -4 -3 -2 -1 0 1 2 3 4 5
Voltage, Volt



Figure 4.2 Effects of H202solution treatment on the I-V curves of 100A Ni/500A
Ti/500A Au to MBE p-GaN. (a) 5:1 solution; (b) "pure" (37%) H202












20 15




, ~--- "
_10 0
C E
00C

10 -o- Hole Concentration
x0
-o---Mobility

5 1 0
0 50 100 150 200 250 300 350
Immersition Time (sec)


Figure 4.3 Hall measurement results of MBE p-GaN after immersion in 1:1
solution.



current increase after 300sec was -50%, while the increase after a 5:1 solution treatment

was -200%. For samples treated in 1:5 and 1:1 solutions, the increases in current at 5V

are 41% and 89% respectively with a treatment of 300sec.

The Hall data for the samples treated with 1:1 solutions are shown in Figure 4.3,

An increase of about 70% was found for the carrier concentration after immersion for

300sec, consistent with the increased current found for the treatment. The decrease in

hole mobility is within experimental errors.



4.2.2 Effects of Extended Immersion Time

Using the 5:1 solution, the effects of 20 to 60 min extended treatment times were

studied. The I-V data are shown in Figure 4.4 and the Hall data are shown in Figure








200
150
100,
S50-
o i

S-50 As cleaned
Oz 50 '., 0 5H202:1H20 45sec
-1001 A 5H202:1H20 5min
-150 X 5H202:1 H20 for 20min
5H202:1 H20 for 60min
-200 1 1 1 1 1 ,;,,,---,
-5 -4 -3 -2 -1 0 1 2 3 4 5
Voltage, Volt

Figure 4.4 Effects of H202 solution treatment with extended time on the I-V
curves of 100OA Ni/500A Ti/500A Au to MBE p-GaN.


4.5. Compared to as cleaned (no H202 treatment), an immersion time of 20min resulted in
a 200% increase in the current level. As the immersion time was further increased to
60min, the current level dropped to -50% of the as cleaned state.These Hall data show
that the hole concentration and mobility first increase and then decrease as the immersion
time increase, consistent with the I-V data.
For MBE-GaN, n-type GaN was also treated with H202 for times of 10,30,300,
1200 and 3600sec using the 5:1 solution. The Hall data are shown in Figure 4.6. Short
treatment times did not change the electron concentration or mobility within experimental
error. At 30 and 300sec, the electron concentration increased (-12%), but these changes
and the mobility changes are still in experimental errors (-15%).





60





8 15

7- 13

6-
0 6 11 ?
o

U. 9
004
| 3- M-^.
000
0 0
Z 2

1 -0--- Hole Concentration 3
----- Mobility
0 11
10 100 1000 10000
Immersition Time (sec)





Figure 4.5 Effects of 5H202: 1H20 solution treatment with extended time on Hall
measurement results of 100A Ni/500A Ti/500A Au to MBE p-GaN.



All these results were obtained on epilayers of p- or n-GaN grown on sapphire by

MBE, as described in Chapter 3. For comparison, epilayers grown by MOCVD were also

treated with the same solutions and methods. The I-V and Hall data as shown in Figure

4.7, and no changes above the experimental noise levels were detected, even after

treatment time up to 60min. Thus the effects of peroxide treatment depend upon the

growth history of the p-GaN




4.2.3 Stability of the Increased Electrical Conductivity

The stability with time of these modified electrical properties was monitored with

I-V and Hall measurements every two days for two weeks. The improved electrical







61





15 --20


13

S'15
qy 11 4
C4EE

9-
-10~
olO0
z
7 -o-- Electron Concentration
--- Mobility
5 -- 5
1 10 100 1000 10000
Immersition Time (sec)



Figure 4.6 Hall measurement results of 5H202: 1H20 solution treated MBE n-
GaN




8 14

7 12

6
c "10
oo

00

0 3
S4- 4

2 z
x 2-4 5
-o-- Hole Concentration

1 ---Mobility 2

0 1 0

1 10 100 1000 10000
Immersion Time (sec)

Figure 4.7 Effects of H202 treatment on MOCVD p-GaN.












10- -15
(a)

47
E -8-0--Hole Concentration 13
8
o
e Mobility
6*
6-11 6
o
5 E
Uo u

j 4- 9 I
_ee
C
C

o 2-



0 15
0 2 4 6 8 10 12 14
Immersion Time, day


12 200
(b)

E 10 180
-18


8 160o
E
E 6- u
a *
-140
04
C
o

S-a- Electron Concentration
S120
_2 2 -e--Mobility


0 ] 100
0 2 4 6 8 10 12 14
Immersion Time, day


Figure 4.8 Hall measurement on stability ofH202 treated samples. (a) n-GaN; (b)
p-GaN. Both were treated with 5:1 solution for 20min.












Table 4.1 Atomic concentration of elements from AES surface survey analysis in MBE p-
GaN after a xH202:yH20 treatment for either 30 or 300sec


As cleaned 1:5 for 30sec 1:5 for 300sec 1:1 for 30sec 1:1 for 300sec
C 4.9 6.3 5.2 5.9 3.5
C1 1.9 1.0 1.1 0.8 1.0
Ga 52.0 48.5 47.6 47.2 49.0
N 33.7 36.4 39.1 37.5 39.4
0 7.6 7.8 7.0 6.7 7.1


conductivity was found to be very stable. Figure 4.8 shows the Hall data for (a) p-type,

and (b) n-type GaN treated with 5:1 solution for 5min and 20min respectively.



4.3 Structural Characterization

4.3.1 AES

Auger peaks from Ga, N and 0 plus C and C1 were detected from samples after

aqua regia cleaning and minimum exposure to air. The surface composition changed after

the H202 solution treatment, as shown in table 4.1. Compared to the as-cleaned state, a

slight increase of N (3 6%) and a slight decrease of Ga (3 5%) were found. Small

increases of C (except for the 1:1 for 300sec treatment) and decreased Cl also resulted

from the H202 soaking. The atomic concentrations of oxygen remained fairly constant

instead of increasing with the H202 immersion. This result was surprising because oxide

or hydroxide formation was expected from this treatment of the GaN surface due to the

reaction with H202.











4.3.2 ESCA

The ESCA spectra for the Ga 2p, Ga 3d and Ols photoelectron peaks are shown

in Figure 4.9 for the as-cleaned (boiling aqua regia, no H202 treatment) and soaked

(1H202:1 H20 solution for 300sec) surfaces. The Ga2p1"2 and Ga2p3a peaks are at

binding energies of 1118.5 and 1145.4 eV, similar to standard data [Mou95]. Comparing

the peaks of samples before and after H202 immersion, no binding energy shift in the

Ga2p1/2 and Ga 2p3/2 spectra (Figure 4.9-a) were found. However, energy shifts of 0.85

and 1.1 eV to higher binding energies were measured for the Ga 3d (Figure 4.9-b) and 0

Is (figure Figure 4.9-c), respectively, as shown in Table 4.2.



4.3.3 SIMS

Secondary ion mass spectrometry (SIMS) depth profiling was performed to

determine any changes in the Ga, N, 0 and H levels. Two different sites were analyzed in

each sample. While no significant variation in Ga and N levels were found, large

differences in the levels of H and 0 were found between the two sites on each samples.

The results are shown in Figure 4.10 for each sample. These results suggested that the H

and 0 levels were not uniform and had large variations even over the same sample.

Because the impurities were not uniform, systematic comparison of 0 and H levels with

the change of immersion time was impossible by SIMS.



4.3.4 AFM

The AFM technique was used to characterize the surface morphology of both

MBE- and MOCVD-GaN epilayers with and without H202 treatments.













(0
.- 11 ,,-,

I 9-

7

5-
1150


26 24 22 20 18
Binding Energy, eV
15 -
(c)


cc10

S1:1 for
0 5- 300se

As
cleaneOA".
0 ___ z


545


540 535 530 525


16 14


520


Binding Energy, eV

Figure 4.9 Comparison of XPS peaks from as-cleaned and 1:1, 300sec H202
treated GaN. (a)Ga 2p; (b) Ga 3d; (c)O Is.


1140 1130 1120
Binding Energy, eV


1110









Table 4.2 XPS results from a 1:1,300sec H202 cleaned p-GaN sample


Surface Treatment Ga2p3/2 Ga2pl/2 Ga3d Ols
As-cleaned 1118.5 1145.3 19.4 532.9
1H202:lH20 for 300sec 1118.5 1145.3 20.25 534.0


6


6 4



0
0


0 200


6

(0
?4
C
2


400 600 800
Sputter Time (sec)


0 200 400 600 800
Sputter Time, sec


1000


Figure 4.10 Negative SIMS depth profile for 5:1 H202 solution treated p-GaN
film (a) as-cleaned state; (b) immersed for 45sec; (c) immersed for 5min;
(d) immersed for 20min and (e) immersed for 60min. Data from 2
different points (a, b) on each sample are shown


1000










6

CO
<4


c 2


0





6


$4
C
aI
5

c2


0


0 200 400 600 800
Sputter Time, sec


0 200 400 600 800


Sputter Time, sec


6


*1
,4


C2


0


200 400 600 800


Sputter Time, sec


Figure 4.10 O-Continued


1000


1000


1000



























0 250 500 rN
NP









(b) 1."00



-0.75



-0.50





-0


0 0.25 0.50 0.7 5 1.O0 N


Figure 4.11 AFM images ofGaN. (a) MBE-GaN, RMS roughness is of 2.6nm; (b)
MOCVD-GaN, RMS roughness is 0.7nm.The points on the MBE-GaN
surface labeled NP are potentially nanopipes as discussed in the text.


















Defects ,













Figure 4.12 Microsturcture of MBE-GaN (TEM, plane-view)


The AFM images for both types of samples are shown in Figure 4.11. The as-

cleaned MBE-GaN and MOCVD-GaN had RMS surface roughness of 2.6 nm and 0.7

nm, respectively. Not only is the MBE epilayer rougher, but they also exhibited

numerous areas labeled "NP" on Figure 4.11 a. As discussed below, these are possibly

nanopipes which allow quick transport of atoms throughout the epilayers. The density of

the "NP" points was _1010 cm"2. Microstructure obtained with TEM plane view

observation of the MBE GaN showed the presence of defects, as in Figure 4.12. While

the character of the defects were not studied in detail, the larger dark defects are












3
(a) ,
2


E


-2 0
-2 *j?0Untreated
-3 15:1 for 60sec
-3 r-''I
-5 -4 -3 -2 -1 0 1 2 3 4 5
Voltage, Volt



0.01 -(
(b)
0.006

S0.002
C
| -0.002

-0.006- / -0- Untreated
S--- 5:1 for 60sec
-0.01 --I-I-I
-1 -0.6 -0.2 0.2 0.6 1
Voltage, Volt

Figure 4.13 Effects of 5:1, 60sec H202 treatment on the I-V of 500A Ni/500A Au
to MBE p-GaN. (b) is the same to (a) but in a smaller scale.



consistent with nanopipes. Their density, while still large, is only 108 cm"2 compared to

the density of 1010 cm"2 reported from Figure 4.1 l-(a).








4.4. Application in Formation of Ohmic Contact to p-GaN

As shown above, the peroxide treatment of MBE p-type GaN can increase the

hole concentration although the immersion time and solution concentration must be

controlled.Increased hole concentration would allow formation of a lower resistance

ohmic contact, so this technique was combined with the Ni/Au contact scheme for both

MBE and MOCVD p-GaN. Figure 4.13-(a) shows the I-V data for Ni/Au contacts on

MBE p-GaN. Increased currents at the same voltage was found for voltages > 2V after

treatment with a 5:1 solution for 60sec. However, at voltages < 1 V, the current was

lower than for an untreated surface, as shown in Figure 4.13-(b). Lower currents are

attributed to incomplete removal of oxide from the contact interface. The results from the

MOCVD p-GaN show similar results. As a consequence, the increased conductivity of

MBE-GaN did not result in reduced contact resistance as expected, presumably due to the

dominance of interfacial reaction layers on current transport.



4.5 Discussion

Considering both hole and electron carriers, the electrical conductivity of a

semiconductor material is usually described by [May90]:

a = q-(n-p. L+ p.- p) (4.1)

where a is electrical conductivity, q is the electron charge, n andp and /, and pp are

electron and hole concentrations and mobilities, respectively. In a p-type semiconductor,

the majority carriers are holes, so the effects of electrons could be omitted because their

concentration is so small. The electrical conductivity is only related to the hole

concentration and mobility. The opposite is true for the case ofn-GaN. From Figure 4.3,









the hole mobility is shown to be constant within experimental error, but the carrier

concentration increased after immersion in the 1H202:1 H20 solution.

As the Hall contacts were deposited before the treatment in H202 solutions, the

effects of H202 on the metal/GaN interface can be omitted. Possible reasons to account

for the increased electrical conductivity after the H202:H20 immersions then might

include: (a) a surface layer (e.g. gallium oxide) could form which if more conductive than

GaN could lead to surface leakage currents, (b) etching of GaN by the H202 solution to

reduce the epilayer thickness, (c) reduced hydrogen passivation of Mg, or (d) formation

of Ga-O compounds to stabilize the excess Ga atoms and decrease the concentration of N

vacancy donors.

A reaction of the type

yH202 + xGa = Ga.Oy + yH20 (4.2)

was possible in this experiment, resulting in a GaxOy oxide. As no data is available to

calculate the gallium oxide thickness in this reaction, the thickness was assumed to be

very thin. If this oxide layer (or other possible reaction products) was more conductive

than the GaN, then the increased conductivity after H202 immersion could be explained

by a surface leakage current between the contact structures. However, literature data

show that Ga203 is normally an insulating dielectric layer on the GaN [Ren99]. Various

acid and base solutions [Sam73] were reported to be effective for removal of the Ga203

layer.

If the peroxide treatment created a conductive surface layer, then the conductivity

should decrease after re-cleaning the surface with an acid or base solution. In contrast, the
















< *"" \ \ \;
2000 \

1500 . : .

1000

500

0
1-1016 I .1017 1.10is 11 10t9
Carrier concentration, cm^-3
Vbi= IeV
---- Vbi = 2eV
- Vbi = 3eV

Figure 4.14 Relation among width of depletion region, carrier concentration and
built-in potential in GaN


I-V data after the acid and base solution cleaning showed a slightly increased (rather than

a decreased) conductivity, so surface leakage current is discounted.

If this Ga203 layer can create a depletion region in p-GaN due to the band
bending effects, the carrier concentration might be changed. To consider this effect, the

depletion width in p-GaN was calculated. The equation used in this calculation is

[Sze81]:

W =.2/-6-Vi (4.3)
qA
FqN~





74


where W is the depletion width, s = es & is semiconductor permitivity, r is

semiconductor dielectric constant, go is vacuum permitivity, Vi, is built in potential, NA is

acceptor concentration, and q is the magnitude of electronic charge. Using the value of ,

= 10.4 for GaN [Ho99b] and q = 1.6 x 10-19 C, the width of the depletion region in GaN

related with carrier concentration was calculated for different built-in potentials as shown

in Figure 4.14. At the hole concentration of 2-3 x 1017 cm-3, the depletion region is

-1000 A. Compared to the thickness of 1 pm, this depletion region is narrow, and is not

expected to cause carrier change of 50% or higher.

Oxygen impurities incorporated into p-GaN from the H202:H20 solution could

cause a surface conversion to n-GaN and result in increased I-V conductivity. When 0

was implanted into GaN and annealed at 1100C [Abe96], it created n-type doping with

an ionization level of-29 meV. Seifert et al [Sei83] proposed that oxygen

substitutionally incorporated into nitrogen sites could be the origin of the free electron

carriers in highly conductive n-type GaN. In the present case, oxygen incorporation

should increase the electron concentrations in n-GaN, but decrease the hole concentration

in p-GaN because of compensation. This might be used to explain any increased

electrical conductivity in n-GaN, but not in p-GaN. Thus, an increased oxygen level from

H202 treatment is also discounted.

The etching of GaN can also lead to the change of carrier concentration in the

measured results. Pearton et al [Pea93] have reported on the chemical etching of GaN

films in aqueous 30-50% NaOH solutions at an etch rate of-2nm/min, although other

authors [You97, Min96, Var96b] did not observe any open-circuit etching in the dark in

alkaline solutions. Weyher et al [Wey97] found the etch rate of [0001] oriented bulk GaN








crystals and epitaxial GaN layers in aqueous KOH to depend upon the face exposed to

the etching solution. The N-terminated face is etched in alkaline solutions but the Ga-

terminated face remains unaffected. Similar crystallographic dependency of wet etching

was also reported by Stocker et al [Sto98]. In addition to open-circuit etching in the dark,

studies were also performed on open circuit photoetching of GaN. Minsky et al [Min96]

noted photoetching of GaN in aqueous KOH and HC1 solutions. The etch rate seemed to

be an order of magnitude higher in alkaline solutions. Extended study of the photoetching

behavior of GaN in KOH solutions [You97] found that the etch rate was dependent on

the incident light intensity, the doping of the material, and the KOH concentration (pH

values of the solutions). Peng et al [Pen98a, Pen98b] also demonstrated the open-circuit

photoetching in aqueous KOH and H3PO4 solutions to be pH dependent.When

illuminated from a 253.7-nm mercury line source, etching of GaN was observed in

solutions with pH values ranging from 11 to 15. The GaN etch rate was 90nm/min at pH

= 14.25, but drops rapidly for pH values above or below 14.25.

The relation ofpH value with the H202 concentration in the H202/H20 solutions

is shown in Figure 4.15. For the solution used in this work, the pH value of the 5:1

solution is around 5.5, of the 1:5 solution is about 4.7 and of the "pure" H202 solution is

around 4.6 4.7 (Note the concentration of H202 in "pure" H202 is about 37%.), which

are significantly smaller than the reported peak values of 12 or over.

If the GaN has been etched by the H202 solution, the GaN film thickness would

decrease and the apparent carrier concentration would decrease the original values of film

thickness was used. If only the etching effect was involved, the measured carrier

concentration would decrease monotonically. This prediction is inconsistent with the





76



7.5
7
6.5
6
C L 5 .5 -

5
4.5
4
0 20 40 60 80 100
H202 Concentration (%, wlw)


Figure 4.15 Relation between pH values and H202 concentration
in H202/H20 solutions [PerOO]



results obtained in this study where the carrier concentration increased with short time

treatment and decreased with longer time of treatment. Besides, the measurement of GaN

epilayer thickness with profilometer and depth profiles showed no significant difference

in the film thickness.

Another possible explanation of increased electrical conductivity may be related

to the reaction of

OH- + H = H20 (4.4)

i.e. the removal of H from the GaN film. It is well known that H compensates Mg

acceptors in p-GaN [Sug98, Ama89]. In fact, a similar mechanism of H removal from

GaN was used to explain the low resistance of Ta/Ti contacts to p-GaN [Suz99]. It was

speculated that Ta/Ti can withdraw (absorb) H from Mg-H complexes, thus reactivating

the Mg acceptor near the contact interface. If this argument was correct, the I-V data

would be expected to continually increase with extended treatment time because more









hydrogen could be removed and more Mg acceptors reactivated. The I-V data in this

work showed the conductivity decreased as the immersion time was extended. So the H

removal cannot explain all of the results obtained. Additional physical or chemical

processes must be involved during the immersion of GaN epilayers in H202 solutions, as

discussed below for surface injection of isoelectronic oxygen donors.

Finally, formation of gallium oxides can reduce the excess Ga atoms, resulting in

a decreased nitrogen vacancy concentration, and the hole concentration should increase

due to less compensation by nitrogen vacancies, as discussed in Chapter 2.

In addition to the formation of Ga oxides, dissociation of H202 to H20 to yield 0 could

result in the formation ofchemisorption bonds with GaN. This would explain the increase

in binding energies for the Ga3d due to an increased net charge on the oxygen versus

nitrogen atoms resulting from the larger electronegativity of oxygen [Deh93]. The

oxygen electronegativity (3.44 on the Pauling scale) is larger than that of nitrogen (3.04),

so higher gallium bonding energies were expected for the Ga-O versus Ga-N bonds.

Oxygen bonds decrease the electron concentration in the Ga valence orbitals which

results in a weaker screening effect of the nuclear charges by the inner electrons, so the

binding energy of Ga3d should increase as observed [Cza75]. If only GaxOy formed,

these arguments predict a lower binding energy for the Ol s, whereas an increased

binding energy was measured. This suggests that Ga(OH)3 formed rather than GaxOy. The

argument of Ga(OH)3 formation is consistent with increased binding energies of both the

Ga3d and Ols in the results shown above. It also could be consistent with removal of H

from the Mg acceptors to form hydroxyl bonds. The typical depths of hydrogen

incorporation and maximum concentration measured in various processing steps were









Table 4.3 Relation of hydrogen incorporation and processing steps [Pea97b]

Process Temp, C Maximum [H] (cm-') Incorporation Depth (gim)
H20 boil 100 1020 1.0
PECVD SiNx 125 3 x 1019 0.6
Dry Etch 170 i0o19 10-o20 >0.2
Implant isolation 25 Dose dependent 2.0
Wet etch 85 2 x 1017 0.6


compiled in Table 4.3 [Pea97b]. Based on these data, it is clear that even at room

temperature, such large incorporation distances have not been demonstrated for nitrogen

vacancies or changes in excess Ga profiles.

A model of the amphoteric roles of 0 is presented here. One effect of 0 could be

the combination of excess Ga and 0 (or OH') to form stable compounds with reduction of

donor N vacancies, but a second effect could be the incorporation of donor 0 impurities

into the GaN lattice which should compensate Mg acceptors.

Formation of Ga(OH)3 could also modify the concentration of excess Ga and

effectively reduce the N vacancies in the GaN film. The high vapor pressure of nitrogen

above GaN could lead to desorption of N, excess Ga and high concentration of

compensating donors in the surface region. Immersion of GaN samples in H202 solutions

could potentially reduce the concentration of excess Ga atoms by formation of oxides or

hydroxides. In the literature, oxidation of Ga with 0 leads to several polytypes for

oxidation products [Wol97].

Theoretical [Neu995] and experimental [Wet97] studies have also shown that

oxygen is a relatively shallow donor which can be present at high concentrations (1016

1019 cm3) in undoped GaN films. High concentration of oxygen resulted in high

background free electron concentrations commonly observed in as-grown films. The








formation energy for substitution of 0 onto the N sites is much lower than onto the Ga

sites, and the covalent radius and ionicity of 0 and N are similar. The group VI element

O would contribute one extra electron to the conduction band and increase the free

electron concentration after incorporation into the GaN lattice. The electrical conductivity

should decrease for the p-GaN, as observed. In this experiment, the decrease of electrical

conductivity after a long time (60min) immersion is believed to result from this

mechanism.

As the immersion time increased, two processes could compete to influence the

change of carrier concentration. Reduction of compensation by removal of H, excess Ga

or N vacancies should increase the hole concentration and conductivity. As more oxygen

atoms were incorporated substitutionally into the GaN films decreased hole concentration

and mobility would be expected at long times.

While the competing effects at short versus long time on excess Ga and VN

qualitatively explain the experimental results, the question remains as to whether there is

sufficient atomic mobility at room temperature to make this a viable mechanism.

The diffusion of oxygen in GaN has been studied using the 170 in SiO2/GaN

interface [Pea99] and the diffusion depth was found to be -400 500A for as-deposited

films. It was also found dislocation pipe diffusion led to larger regions of diffusion

around threading dislocations. The overall transport of the oxygen is dominated by the

pipe diffusion which is along the dislocation axis only. The square of the diffusion

distance in the dislocation is given by

X2 =2*Dd t (4.5)








where Dd is the diffiusivity in the dislocation, and t is the diffusion time. The least square

fit to the data led to the equation [Pea99]:

D = (4.5 2.2). 10-2.exp(- 0.23 0.12e) (4.6)
k.T

where k is the Boltzmann's constant. The activation is approximately half of the expected

value of diffusion in bulk material.

Using these two equations (4.5 and 4.6), the diffusivity was calculated to be

between 2.38 x 1018 and 8.8x 1014 cm2/sec (using the upper and lower limit of the data in

equation 4.6) and the diffusion distance at which the oxygen concentration equals to the

background concentration of 1017cm3 was between 18.5 A and 0.355pm for a time

period of 60 min. With the experimental times used in this study, it is difficult to believe

that the oxygen could diffuse deep enough into the GaN film at room temperature to react

with excess Ga over a l1 m epilayer by a uniform transport mechanism. It is much more

reasonable to expect rapid diffusion down to a defect perpendicular to the surface, such

as a dislocation core or nanopipe, with diffusion out of the core or pipe, parallel to the

surface, over much shorter distances.

Defects like nanopipes, screw dislocations and others have been reported in GaN

and are known to be paths of enhanced mobility for oxygen and hydrogen. Nanopipes

with diameters ranging from 5nm to 0.5pm were reported at a densities of-108 cm2.

They were reported to be parallel to the c-axis of GaN unit cell [Ven99, Kan99] as shown

in Figure 2.1. The main composition of nanopipes was Ga, C and 0 [Kan99], as

discussed in Chapter 2. Considering the high chemical activity of 0 in the H202 solution,

diffusion of 0 or H along these nanopipes and out of the GaN or nanopipes into the






81


surface or GaN lattice would be logical and would explain the results from H202

treatments. With the help of these nanopipes, the point defects are expected to be

extracted or to reach the deep site of the epilayer easily.

Based on this model of enhanced atomic diffusion along nanopipes, the expected

increase of carrier concentration from passivation of these defects (using the nanopipe as

an example) was predicted as following.

For the perfect crystals with no nanopipe, the total concentration of holes would

be


N=Nh S't


(4.7)


where Nh is the average hole concentration in this perfect crystal, S is the sample surface

area, and t is the sample thickness. With the incorporation of nanopipes, the amount of


Figure 4.16 Relation of carrier concentration decrease and nanopipe density in
GaN








carriers will be changed. If the dislocation/nanopipes are perpendicular to the sample

surface, extending through the thickness of epilayer film, and the hole carriers are

completely removed in the dislocation/nanopipes and over the depletion distance around

the dislocation/nanopipes without effecting the carrier density in the "perfect" region

beyond the depletion distance, the reduced carriers concentration Nh' would be:

Nh"S't =Nh 2[1- (-+x)2 "N ]"S't (4.8)

where x is the average depletion distance around the nanopipes, d is the average nanopipe

diameter, and Np is the nanopipe density with a unit of cm2.

The reduced carrier concentration compared to the perfect crystal will be:

ANh (A-(+x)2 N, =l-7'y2*'N, (4.9)

where since x is of the same order of magnitude as d and exerts the same influence on the

change of hole concentration, the values are added together as defined asy. Figure 4.16

shows the relationship between carrier concentration decrease and the influence of

dislocation/nanopipe size and density. At a density Np of 109 cm"2, the carrier

concentration would be reduced by 50% with ay distance of 126nm. At a nanopipe

density of 1010 cm'3, the y distance of only 40nm would result in the same reduction.

In the AFM image shown in Figure 4.11-a, the "cavities" labeled "NP" are

suspected to be nanopipe or dislocation sites and their density is calculated to be 4x 1010

cm"2 with a size of 10 -50nm. They distance for p-GaN with our measured carrier

concentration increase of -100% is calculated to be 40nm. These values are in good

agreement with the calculated depletion of holes assuming the nanopipe surface cause

depletion of carriers. The variation in 0 and H from SIMS data (Figure 4.10) would be









consistent with the literature data [Kan99] showing that nanopipes were rich in Ga, 0 and

C. The result that 0 and H at one point was continually higher than another one was

consistent with the published results, and is also consistent with the assumption that the

direction of these nanopipes were parallel to the c-axis of GaN unit cell. It should be

noted that analysis size of SIMS is 75 x 75 utm2, which means each analysis covered an

area containing many nanopipes. The results from the SIMS analysis in this work suggest

the nanopipe/dislocation sizes and/or density are not uniform.



4.6 Summary

The effects of H202 treatment on GaN had been studied using the I-V, Hall, AES,

ESCA, SIMS, AFM and TEM. It was found that the electrical conductivity increases after

short (< 20 min.) H202 solution treatments with contact already deposited. The

magnitude of the current change was related to the H202 concentration, with 5H202:

1H20 solutions giving best results. The Hall data showed that while the carrier mobility

exhibited changes <10%, the average hole concentrations could be changed by up to

100%. According to AES data, the Ga atomic concentration was decreased, N increased

and 0 not changed after the H202 treatment. Using the ESCA analysis, formation of

Ga(OH)3 was shown to result from H202 treatments. With SIMS depth profiling, big

variations of H and 0 were found across the samples, consistent with published work and

AFM data from this study. Based on the possible reactions between GaN and H202,

reasons for the increased conductivity of MBE grown p-GaN were discussed, and a

model of reaction with oxygen to deplete nitrogen vacancies or compensating H at short

time, and injection of 0 donors at long times, was proposed.













CHAPTER 5
"NOG" SCHEME FOR OHMIC CONTACT TO p-GaN



5.1 Introduction

When Maruska and Tietjen [Mar69] used a chemical vapor deposition technique

to make GaN layer in the late 1960's, the GaN was highly conductive n-type even when

not deliberately doped. Two dominant mechanisms were speculated to explain intrinsic n-

type as-grown materials. One was nitrogen vacancies and another was of unintentional

oxygen doping. The N vacancy was considered to be a donor because a N vacancy would

form a void surrounded by four Ga atoms contributing three electrons. Two of these three

electrons could reconstruct and leave a single electron that could be donated to the

conduction band. This model was later questioned, and instead unintentional oxygen was

proposed to be the donor in as-grown GaN [Pan73]. Oxygen with its six valence electrons

on a N site (N has five valence electrons) would be a single donor.

The proposed "Nitride-forming metal Over Gallide-forming metal" ("NOG")

scheme in this chapter is based on interfacial reactions to control the Ga and N to

decrease N vacancies. Based on the review above and in Chapter 2, N vacancies would

be at least one of the reasons for a high concentration of donors in GaN

Metallurgical reactions of transition metals with GaN have been reviewed in

Chapter 2. Only the principles of the "NOG" scheme and its applications are presented in

this chapter. Applications of"NOG" include interpreting published literature results and

designing/testing new contact schemes in this study.











5.2 Principles of"NOG" Scheme

Since there is a large difference in the vapor pressures of Ga and N, there may be

a high concentration of N vacancies, VN, in as-grown epilayers. A high VN condition is

equivalent to a Ga-rich condition. An opposite situation could be postulated: if a N-rich

condition could be created in as-grown GaN films, which is equivalent to creating Ga

vacancies, the as-grown GaN should be intrinsic p-type. This condition has not been

achieved in bulk GaN films, probably because of the high vapor pressure of N in

equilibrium with GaN. Instead, this postulated condition leading to p-GaN might be

achieved by interfacial reactions in the contact region. If extra N atoms could be kept

between the contact metal layer and the bulk p-GaN film, a N-rich condition could be

formed at the metal/GaN interface. The extra N atoms could fill the VN positions and

create Ga vacancies acceptors. If such Ga vacancy acceptors were shallow and reached a

sufficient concentration (>1018 cm'3), the interfacial region could become p+-GaN and

current transport could be dominated by field emission. A low resistance ohmic contact

could be obtained as a result. Even a decrease in the VN concentration should increase the

free hole concentration in the contact region. These are the postulates upon which the

"NOG" scheme is based. While these reasons seem sound, the "NOG" scheme has not

led to new and improved contacts. It is worth a while however to review the progress

made using these ideas.










Protection metal (Au)
Nitride-forming Metal
Gallide-forming metal








Figure 5.1 Principle of "NOG" scheme.


According to the enthalpy of the metallurgical reactions, all transition metals were

classified into three groups in Chapter 2: gallide-forming, nitride-forming and neutral

metals with respect to reactions with GaN. Metals for forming "NOG" contacts were

selected based upon this classification.

The structure of a "NOG" contact is illustrated in Figure 5.1. A gallide-forming

metal adjacent to GaN is followed by a nitride-forming metal, which is covered with a

layer of protective metal (such as Au). Under suitable annealing conditions, the gallide-

forming metal reacts with GaN to form stable gallides and release N atoms. This first

metal layer must both dissociate the GaN lattice and prevent or slow down the nitrogen

out-diffusion. The nitride-forming metal layer would help to keep the released N atoms in

the contact interfacial region and create a high N chemical potential.



5.3 Comparison with Published Contact Results

The achievement of low resistance ohmic contact to p-GaN is of great importance

to GaN device performance. Many studies have been reported along with postulated









mechanisms to explain reduced contact resistance, such as interfacial reactions

eliminating barriers [Hol97] and doping the surface region [Tre96, Tre97], GaN re-

growth [You98, Tre96], H extraction [Suz99] and Ni oxidation [Ho99a, Ho99b]. A few

representative contact schemes are discussed here based on the principle of "NOG" to

show the possible applications of this contact scheme. The discussion remains strictly

speculative.

The contact scheme of Ni/Au [Fun99, Kim97a, Ish97, Kin97, Tre97] has been

widely used for GaN device fabrication. Based on the "NOG" principles, this lower

resistance for these contacts results from the reaction between Ni and GaN[Ber93,

Guo96, Ven97]. The Ga would react with Ni to form stable gallides and reduce excess Ga

atoms. Reduced excess Ga is expected to result in reduced concentrations of VN and less

compensation of acceptors. This would result in higher free hole concentrations in the

interfacial region. The success of some other contact schemes, like Pd/Au[Kim98,

Kim97b], Pt/Au [Kin97, Mor96, Yoo96], Pd/Au/Pt/Au [Kim97b], Pd/Pt/Au [Kin97] and

Pt/Ni/Au [Jan99], could be explained by the same mechanism. As discussed in Chapter 2,

Pd and Pt formed more stable gallides than Ni, so it is not surprising that contacts with

lower specific resistance were obtained with Pd/Au and Pt/Au contacts

A relatively low specific contact resistance (3.6x 103 Q-cm2) was obtained with

Ni/Zn-Au [You98] on p-GaN with a carrier concentration ofNh = 4.4x1017cm3. The

authors reduced the time between p-GaN film growth and contact metal evaporation in a

high vacuum system. They postulated that Zn was an acceptor and that the Zn-Au alloy

layer increased the interface carrier concentration. Zn is an acceptor in GaN, but the

energy level is deep (EA = 570 meV)[Pan97] and therefore should not be ionized at room









temperature. Based on the "NOG" scheme, the mechanism should be the same as the

Ni/Au scheme discussed above, although the Zn might help increase the carrier

concentration in the contact region at high temperatures. The main reason for improved

contact performance with Ni/Zn-Au probably came from the limited time for native oxide

to grow on the GaN and the use of high vacuum for metallization. Optimum contact

resistance would not be predicted for this Ni/Zn-Au metallization because no reaction

was found between Zn and GaN and no nitride forming component exist in the contact.

A low resistivity (3.2x 105 Q-cm2) ohmic contact to p-GaN was produced with

Ta/Ti metallization after a high temperature anneal (800C for 20 min) [Suz99]. The

authors postulated that Ta and Ti were able to remove hydrogen from Mg-H complexes

and therefore reduced compensation of the acceptors. It was also found that a dual layer

structure of both Ta and Ti formed better contacts than a single layer of either Ta or Ti,

although both Ti and Ta were reported to have stronger binding energies with hydrogen

than Mg. Although more hydrides are possible, it was reported in the literature that the

common hydrides to Mg, Ti and Ta were MgH, TiH2 and TaHo.5, and the enthalpy for

MgH (-0.77 eV/atom) was more negative than for TiH2 (-0.68 eV/atom) or TaHo.5 (-0.417

eV/atom) as shown in Table 5.1 [Fuk93]. This means that MgH is more energetically

favored than the TiH2 or TaH0.5. For these reaction products, Ta and Ti might not be able

to reduce the amount of H in MgH. After a few days, the resistance of the Ta/Ti ohmic

contacts increased to a much higher value. This was attributed by Suzuki et al to a

reverse transport of compensating hydrogen from the Ti/Ta layers back to the interface

region and recompensation of Mg acceptors. Contrary to the H mechanism, the "NOG"

scheme would be consistent with a postulate that Ta and Ti would dissociate the GaN