|Table of Contents|
|2. OHMIC contacts: An overview|
|3. Epitaxial growth principles...|
|4. Sample characterization|
|5. InN epitaxial growth|
|6. Contact structures and...|
|7. Nitride device development|
|8. Conclusions, implications, and...|
|Table of Contents|
Table of Contents
2. OHMIC contacts: An overview
3. Epitaxial growth principles and apparatus
4. Sample characterization
5. InN epitaxial growth
6. Contact structures and performance
7. Nitride device development
8. Conclusions, implications, and future work
IMPROVED NITRIDE-BASED DEVICE PERFORMANCE: AN
EPITAXIAL GROWTH APPROACH
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
Were it not for the seemingly limitless patience and understanding of Prof.
Cammy Abernathy it is unlikely that I would have persisted as a graduate student, much
less completed this PhD dissertation. It is extremely rare to encounter an advisor as
thoughtful, insightful, inspiring and generous as she is. I feel extraordinarily fortunate to
have had the privilege of working for Prof. Abernathy and only hope that I can continue
my scientific career in a manner that honors her renowned reputation. It is impossible to
overestimate my gratitude to her.
In order to successfully negotiate a project like this, one requires the integration of
talent and equipment from several groups. In particular, the efforts of Prof. Fan Ren,
Prof. Stephen Pearton, and the staff at the Major Analytical Instrumentation Center
(MAIC) should be recognized as they are deeply appreciated. Additionally, assistance
from fellow students Devin MacKenzie, Brent Gila, Mark Overberg and K. N. Lee has
been invaluable. Thanks also go to Prof. Stan Bates, Prof. Fred Sharifi, and Prof. Rajiv
Singh for serving on my committee.
TABLE OF CONTENTS
ACKNOWLEDGMENTS ................................................................ ii
ABSTRACT ................................................................................. v
1 INTRODUCTION .................................................................... 1
1.1 Motivation ................................. ....... ....................... 1
1.2 Group III-Nitrides......................................................... ....... 3
1.3 Scope of Current Work ............................................................. 5
2 OHMIC CONTACTS: AN OVERVIEW .......................................... 8
2.1 Description of Ohmic Contacts ............................................................. 8
2.2 The Work Function, D ......................................................... .. 9
2.3 Models for Contact Resistance ................................................... 11
2.4 History of Contacts to III-Nitrides .............................................. 14
2.4.1 N-Type ............................... ....................... ................ 14
2.4.2 P-T ype ................ ........... ....... ............................... ..... 16
2.5 The Intermediate Semiconductor Layer ......................................... 17
3 EPITAXIAL GROWTH PRINCIPLES AND APPARATUS ..................... 20
3.1 Principles of Epitaxy .............................................................. 21
3.2 Models of Nucleation .............................................................. 26
3.3 Heteroepitaxy ....................................................................... 28
3.4 Graded Layers ....................................................................... .... 31
3.5 Comparison of Growth Techniques ............................................... 32
3.5.1 MBE ................................................ .......... ........ 33
3.5.2 MOCVD ..................................................................... 38
3.5.3 MOMBE ................................................................... 41
3.6 The Varian Gas-Source Gen II ...... .............................................. 42
3.6.1 Vacuum Pumping Systems ................................................. 42
3.6.2 Sources ....................................................................... 43
3.6.3 MOMBE Safety ........................................... ............. 49
3.6.4 MOMBE Maintenance ... ............................................. ...... 51
4 SAMPLE CHARACTERIZATION..................................................... 53
4.1 Electrical Characterization ...................................................... 53
4.1.1 The Hall Measurement ..................................................... 54
4.1.2 Capacitance-Voltage (C-V) Measurement ............................... 56
4.2 Structural Characterization ......................................... ........... 57
4.2.1 X-ray Diffraction (XRD) .................................................. 58
4.2.2 Transmission Electron Microscopy (TEM) ............................... 61
4.3 Morphological Characterization ........................................ ........ 64
4.3.1 SEM .......................... .............. ..... ................ .... 64
4.3.2 AFM .................... ................................. ........... .. 67
4.3.3 Profilometry .............................................................. 70
4.4 Compositional Analysis ........................................... .............. 71
4.4.1 AES ................................................ ...................... 71
4.4.2 SIMS .......................................................... .............. 75
5 InN EPITAXIAL GROWTH .......................... ........... ................... 76
5.1 History of InN Growth .............................................................. 76
5.2 Role of Nitrogen Plasma and Substrate Temperature .......................... 80
5.2.1 Sample Loading ........................................... .............. 80
5.2.2 Sample Growth .. .. ............................................................ 81
5.3 Substrate Effects .................................................................. 91
5.4 Indium Precursor ...................................................................... 100
5.4.1 TMI Derived InN ............................................................ 102
5.4.2 Solution TMI Derived InN ................................................. 105
5.4.3 Solid In Derived InN ......................................................... 106
5.5 Summary of InN Growth .......... ............................................. 109
6 CONTACT STRUCTURES AND PERFORMANCE .............................. 113
6.1 Contacts to n-Type Nitride Films................................................. 113
6.2 Thermal Stability of n-Type Contact Structures ................................ 121
7 NITRIDE DEVICE DEVELOPMENT ................................................ 126
7.1 GaN/AlGaN HBT Growth ......................................................... 126
7.1.1 Comparison of Aluminum Precursors ................................. 127
7.1.2 HBT Fabrication and Performance ..................................... 136
7.2 P-Type Contact Layers Using GaAs ............................................... 147
8 CONCLUSIONS, IMPLICATIONS, AND FUTURE WORK .................... 154
REFERENCES ................................................................................... 157
BIOGRAPHICAL SKETCH .................................................................. 163
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
IMPROVED NITRIDE-BASED DEVICE PERFORMANCE: AN EPITAXIAL
Sean M. Donovan
Chairperson: Professor Cammy R. Abernathy
Major Department: Materials Science and Engineering
An exploration of methods for improving nitride-based device performance using
epitaxial growth techniques is made. Particular emphasis is placed on lowering the
specific contact resistance, p,, of devices using an intermediate semiconductor layer
(ISL). Analogous to what has been done with GaAs, where contact layers fabricated
from InAs are used to improve contact performance, an ISL based on InN has been
developed for wide-gap nitride devices. Additional work addressed the issue of impurity
contamination during MOCVD growth of GaN/AlGaN heterostructure bipolar transistors
(HBTs) by examining the use of an alternative aluminum precursor.
InN layers were deposited with ultra-high vacuum epitaxial techniques using
metalorganic, solid and gaseous sources. Growth conditions were optimized to produce
single-crystal films with smooth surface morphologies. This optimization included the
use of a nitridation step prior to growth initiation. Nitridation was accomplished by
exposing the heated substrate (>8750 C) to reactive nitrogen supplied by a vacuum-
compatible plasma cell. All InN films, regardless of substrate type or temperature,
nitrogen plasma condition, or indium precursor, exhibit as-grown background electron
concentrations on the order of 1020/cm3. This characteristic precluded the use of InN
ISLs for p-type contacts.
Specific contact resistance was measured using the transmission line method with
sputter deposited WSix metallization. InN contact layers graded from InAIN were found
to be thermally stable to anneals of 6000 C, with pc 6 x 10"6 Q-cm2. Intermixing of W
with the underlying contact layer degraded contact performance at higher temperatures.
HBT performance is limited by the large resistance in the p-type base region. A
solution utilizing p-GaAs contact layers was explored. Unfortunately, hole transport was
restricted due to a valence band offset of -1.8 eV. This led to the investigation of a digital
alloy composed of p-GaAs and p-GaN to simulate a graded layer.
Parasitic reactions in the gas phase leads to low MOCVD growth rates of AlGaN
relative to its binary components and to increased impurity incorporation that is
detrimental to device performance. Examination of an alternative Al precursor, tri-
tertiarybutylaluminum, suggests that adducts formed between the Al precursors and NH3
interact with trimethylgallium, scavenging out these growth nutrients.
This dissertation examines methods to improve nitride-based device performance
using epitaxial growth techniques. In order to maximize the potential of this emerging
technology, several issues must be resolved including minimizing contact resistance,
minimizing impurity incorporation and achieving abrupt dopant profiles in
heterostructure devices. The first two of these issues are addressed in this work.
Simple empirical observation suggests that humankind has been subject to an
inexorable urge to elucidate the nature of the universe, how best to control it, and the role
of humanity therein. These tendencies have given rise to the theory of evolution,
identification of DNA as the code of life, and the advent of space exploration, for
example. Incidental to this craving for understanding and control has been the
development of complex tools such as computers and radio telescopes. These tools
consist in part of interconnected semiconductor devices that precisely regulate electrical
current. This technology has been extended to include an array of consumer electronic
products. The combination of an innate desire for greater scientific understanding and
market driven economics creates a demand for, and leads to ever-improved products
based on semiconductor devices. Further impetus for improvement is provided by a
yearning for military superiority. Since device performance is a function of the material
it is composed of, new semiconductor devices derived from novel materials must be
developed in order to satisfy higher performance requirements. Concomitantly, there is a
circular relationship between basic research and product development. As technological
advances are introduced in the form of better analytical tools, improvements in research
are possible which lead to further advances.
Silicon was the first semiconductor material to gain prominence for integrated
circuits. It possesses several outstanding features that make it amenable to device
applications including dopability, ease of processing, and relative abundance. However,
Si is limited by having an indirect bandgap, which eliminates it as a candidate for light
emitting devices (LEDs). Also, clock speeds are limited by the mobility of majority
carriers (1500 cm2/ s for electrons).' These limitations led to the development of GaAs
based devices. With its direct bandgap and proper alloying with In, Al, and P, light
emitters useful to the fiber-optic communications and data storage industries have been
achieved. Faster devices are also possible owing to improved electron mobilities (8500
cm2V.s ).' For most applications, the capabilities afforded by Si and GaAs-based
devices are more than adequate. Nonetheless there are areas of performance unreachable
by these traditional semiconductor materials. Areas of special interest include high-
power (100-2000 kVA) 2 and high-temperature (>3000C) 3 operation and light emission
over the entire visible spectrum into UV. Researchers have turned to large bandgap
semiconductors such as SiC, II-VI compounds such as ZnSe and the III-nitrides to
address these issues.4'5'6 This dissertation examines the latter group of semiconductors.
1.2 The Group III-Nitrides
Elements from group III in the periodic table, especially Al, Ga and In, combined
with nitrogen constitute a family of compound semiconductors commonly known as III-
N semiconductors, or more simply, nitrides. Table I lists some important materials
properties of A1N, GaN and InN. Silicon, GaAs and SiC are included for comparison.
Table I. Important Material Parameters for Electronic Devices.
Band-Gap Lattice Electron Saturation Breakdown
(eV) Contstant Mobility Velocity Field
(A) (cm2/Vs) (cm/s x 107) (V/cm x107)
Si 1.12 5.4 1500 1.0 3.0
GaAs 1.42 5.65 8500 1.0 6.0
4-H SiC 3.26 10.05 (c) 1140 (6-H) 2.0 30
GaN 3.4 5.19 (c) 1000 2.5 50
AIN 6.2 4.98 (c) 135 1.4
InN 1.9 5.7 (c) 3200 (calc) 2.5
Of principal interest are the large band-gap, high saturation velocities and high
breakdown fields of the nitrides which can lead to superior device performance over Si
and GaAs based technologies. This class of semiconductors possesses excellent thermal
stability as well. Further, the direct band-gap of the III-N's permits fabrication of light
emitting devices. To date, long lifetime continuous wave operation blue laser diodes
have been introduced.7 Work has also led to the development of UV LED's and laser
diodes as well.8 Proper selection of alloy composition of InxGal-xN can result in optical
emission over the full visible spectrum. Figure 1-1 illustrates the possible alloys and
band gaps of the wurtzitic nitrides. Included for comparison are other relevant
S......--------- .-- ----: --- - -..
i- .i r '_i
L 1 C -
I 1 I I
GaP p^AlA,:. 1-
_1 _^_^.L. -..^
i..rernical bond length (A)
Figure 1-1. Bandgap energy versus bond length for important
semiconductors illustrating range of alloy compositions and related
Owing to their outstanding physical properties, III-nitrides have been intensely
studied in recent years as candidates for high power, high frequency and high temperature
electronic devices, with the first GaN MESFET reported by Khan et al.9 Such a high
level of interest exists that many conference symposia proceedings and other
compilations of recent work are routinely published.'10, 1, 12
As is the case with any emerging technology, the nitrides require improvement in
critical areas, especially with ohmic metallization. Metal contacts are an integral part of
electronic and optoelectronic devices that allow electrical current to flow into and out of a
device with a minimum of power loss. Suitable ohmic contact formation to large band-
gap materials such as the III-nitrides is particularly troublesome because high p-dopant
activation levels are difficult to achieve. Moreover, interfacial energy barriers to carrier
transport arise which are exacerbated as the band-gap increases. A deeper discussion is
presented in chapter two. Currently, device performance is hampered by high contact
resistance that causes degradation of the contact and reduced operation lifetimes. This
degradation arises from I2R heating at the contact region which promotes spiking of
metal through the active region of the device. In order to take full advantage of the high
temperature capabilities of the III-Nitrides, thermally stable ohmic contacts must be
1.3 Scope of Current Work
This dissertation details research done to improve nitride-based device
performance using epitaxial growth techniques. Special emphasis was placed on
developing thermally stable ohmic contacts for large bandgap nitrides. Thin
semiconductor films of InN were developed for use as an intermediate semiconductor
layer (ISL) as an aid to reducing contact resistance. The bulk of this effort was directed
toward optimizing InN film growth using metalorganic molecular beam epitaxy
A survey of ohmic contact physical principles is presented in Chapter 2. Models
of specific contact resistance are reviewed along with the history of contacts for large
bandgap nitrides. The basis for using an InN ISL to improve contact resistance to nitride
semiconductors is developed.
Chapter 3 discusses issues pertinent to epitaxial film growth with a description of
nucleation and growth models. Comparisons of MOMBE, metalorganic chemical vapor
deposition (MOCVD), and molecular beam epitaxy (MBE) growth equipment and
techniques are made. A detailed description of the MOMBE system utilized for this
work is presented.
Chapter 4 reviews characterization techniques used to analyze samples for this
work. Samples were analyzed using scanning electron microscopy (SEM), atomic force
microscopy (AFM), secondary ion mass spectroscopy (SIMS), Auger electron
spectroscopy (AES), transmission electron microscopy (TEM), x-ray diffraction (XRD)
and Hall measurements.
Chapter 5 starts with a review of efforts to develop InN films to date. An
investigation of the origin of n-type behavior in as-grown InN is then presented in three
parts. First, the effect of varying nitrogen plasma characteristics on electrical properties
of InN films is studied. Next, substrate effects such as substrate choice and pregrowth
treatment are examined. Finally, the role of In precursor on film characteristics is
Contact performance results based on InN ISLs are presented in Chapter 6.
Thermal stability of contact structures is examined for n-type material.
Chapter 7 presents work done to develop nitride-based heterojunction bipolar
transistors (HBTs). Growth conditions for the GaN/AlGaN heterostructure are given and
an analysis of certain gas phase interactions during growth of AlGaN is made. These
interactions reduce the growth rate, allowing increased impurity incorporation. HBT
device performance is presented, which is limited by high specific contact resistance in
the p-type base region. Experiments with p-GaAs contact layers were conducted in an
effort to ameliorate this problem.
Chapter 8 offers conclusions drawn from and implications of this body of work,
with suggestions for future research .
OHMIC CONTACTS: AN OVERVIEW
This chapter will examine the physical principles of metal-semiconductor
interfaces and the basis for constructing ohmic contacts to the large-gap nitrides. A
review of work done to date on contacts to GaN is presented to act as a foundation for the
current path of investigation.
2.1 Description of Ohmic Contacts
In order for a semiconductor device to operate it must be able to accept and emit
electrical current. Current is brought into and out of the device by metal wires which are
physically attached to the semiconductor device at special contact regions. A successful
ohmic contact is one which has a small resistance relative to the device itself and follows
the familiar Ohm's Law, V=IR. Contacts of this type will exhibit linear current density
(J) vs V behavior around the operating voltage of the device as seen in figure 2-1, curve
A. This seemingly straightforward idea is complicated by energy barriers that may exist
when a metal and semiconductor are brought into intimate contact and results in
rectifying behavior, exhibited in curve B of figure 2-1.
In order to reach thermodynamic equilibrium, the Fermi levels of the metal and
semiconductor must come into alignment. This causes the conduction and valence bands
to bend at the interface as shown in figure 2-2 for the case of an n-type semiconductor.
Note that this is the case for an n-type semiconductor contacted to a metal whose work
function is greater than that of the semiconductor, the typical situation for nitrides. There
exists an energy barrier to carrier transport across the interface, OB, defined as the
difference in energy between the Fermi level in the metal and the bottom of the
conduction band in the semiconductor at the interface. This value is related to the
difference in work function between the semiconductor and metal.
.......... .. ......V
Figure 2-1. Current density versus voltage curves showing Ohmic behavior (A)
and rectifying behavior (B).
2.2 The Work Function, c
The work function of a material is defined as the difference between that
material's Fermi energy and ionization energy. This is the energy required to remove an
electron from the material and place it an infinite distance away, figure 2-3 is a schematic
When selecting metals for ohmic contacts to n-type semiconductors, the desired
case is cM< Os in order to avoid formation of an energy barrier, OB, to electron transport.
Figure 2-2. Band diagram of metal/semiconductor interface showing
energy barrier to carrier transport,OB.
-^7-- ---- ----- - --
............ EF (B= M-X
Figure 2-3. Energy bands for a metal with work function, (M, and n-type semiconductor
with work function, Os, (a) before and (b) after intimate contact. X is the electron affinity.
The energy barrier to electron transport, OB, is outlined with bold lines.
EF .... ....
Table II. Work functions for some typical contact metals. [after references
13 and 14]
Metal: Ag Ti Cr Au Pd Ni Pt Al W
Function(eV): 4.3 4.33 4.5 4.9 5.13 5.15 5.65 4.08 4.55
For p-type semiconductors, FM>DS is desired. Table II lists some typical contact metals
and their work functions. The work function of n-GaN is around 4.1 eV. Unfortunately,
in the case of GaN there does not exist a suitable contact metal that avoids barrier
formation. Although Al has a smaller work function than n-GaN, contacts directly to Al
have been shown to be rectifying after annealing.13 This is likely due to the formation of
A1N at the metal-semiconductor interface. Since no metal exists that would allow
unimpeded carrier flow across the contact interface, carriers must be thermally excited
over the barrier (thermionic emission), quantum mechanically tunnel through it (field
emission) or move via a combination of the two (thermionic field emission).
2.3 Models for Contact Resistance
Murakami and Koide have explained the fundamental models describing specific
contact resistance, pc.14 There are three cases, for light, medium, or heavy doping. Figure
2-4 illustrates the band-diagrams for these three cases. The dominant conduction
mechanism will be a strong function of the doping concentration and temperature.
Thermionic emission will occur for the case of light doping, ND<1017cm3. Here the
depletion width in the semiconductor at the interface is large and carriers are unable to
tunnel through the barrier. The current flow is a function of temperature and the contact
-A. Light doping
B. Medium doping
C. Heavy doping
Figure 2-4. Band diagrams for an n-type semiconductor with different doping levels
and depletions widths, W. The block arrows indicate the predominant conduction
mechanism. See text for details
resistance, pc, can be described, after the Wentzel-Kramers-Brillouin (WKB)
Pc = CI expQ -
where C1 is a function of temperature, q is the electronic charge, k is Boltzmann's
constant and T is temperature.
Thermionic field-emission occurs when doping is at an intermediate level, 1017 to
1018 cm-3. The depletion region becomes narrower and charge carriers can breach the
interfacial barrier with some thermal activation. The specific contact resistance here is
Pc = C2 exp bO -)
ND coth U
where E0o = -
C2 is a function of (DB and T, h is Planck's constant, e is the dielectric constant of the
semiconductor, m* is the effective mass of the tunneling electron and ND is the doping
Finally, when the doping level is high, >1018 cm-3, the depletion width becomes
narrow and the carriers are able to tunnel through the interfacial barrier. Here the contact
resistance is governed by field emission and given by
Pc = C3 exp
where C3 is a weak function of temperature.
2.4 History of Contacts to III-Nitrides
Now that a description of the physical principles governing contact performance
has been made, a review of work on contacts to III-nitrides will be rendered. Since most
of the important optical and electronic devices based on III-N materials rely on GaN, this
section will examine the current state of the art for that material. The figure of merit for
contacts is the specific contact resistance, Pc, with units of C-cm2.
The first reported contacts to GaN were made by Foresi and Moustakas.
employing a Au or Al contact directly deposited on GaN yielding pc 10-3 -cm2.15
Subsequently, Lin et al. used an Al/Ti bilayer metallization scheme to achieve pc = 8 X
10-6 f-cm2.16 Their approach utilized a 20 nm layer of Ti e-beam deposited on GaN
followed by a 100 nm capping layer of Al. A 9000 C anneal for 30 s was needed to
obtain the best result. Using Pd/Al metallization and a 6500 C, 30 s anneal, Ping et al.
measured pc = 1.2 X 10-5 f-cm2.17 Guo et al. deposited 15 nm of Ti followed by 150 nm
of Ag on GaN to obtain an unannealed contact resistance of 6.5 X 10'5 K-cm2.18 A
multilayer structure was studied by Fan et al. that yielded contact resistances as low as
8.9 X 10-' -cm2.19 Their approach starts with an RIE etch of the GaN surface followed
by deposition of Ti (15 nm), Al (220 nm), Ni (40 nm), and Au (50 nm). This was
followed by an RTA of 9000 C for 30 s. Cole et al. studied the thermal stability of W
ohmic contacts to GaN.20 The 50 nm W contacts were sputtered onto etched GaN
samples and annealed at 10000 C for one minute. An interfacial W-N phase was
observed that was considered responsible for the thermal integrity of the contacts, which
had an pc = 8.0 X 10-5 Q-cm2 Miller and Holloway studied several metallization
schemes including Ag, Au, TiN, Au/Ti, Au/Mo/Ti, and Au/Si/Ti.21 Their results showed
Ag and Au diffuse across the GaN interface for anneals >5000 C but the other contact
schemes were thermally robust even with 30 minute anneals at 5000 C. Their results
indicate the formation of an interfacial TiN is essential to obtaining a useful contact.
Smith et al. examined Al contacts at room temperature and 5000 C.22 Values of 8.6 X 10-
5 Q-cm2 were obtained at room temperature with 6.2 X 10-5 _-cm2 at 5000 C. These
values rose after samples received 60 s annealing treatments at 550 or 6500 C. It was
discerned through high-resolution TEM and electron energy loss spectroscopy that an
interfacial Al nitride or Al oxynitride had formed. Further investigation of the
importance of an interfacial TiN layer was conducted by Wu et al.23 Their study used a
2-step Ti deposition capped by a 200 nm layer of Au. First, 20 nm of Ti was deposited
and annealed at 9750 C for 30 s followed by another 20 nm of Ti. This scheme yielded
contact resistances of 3 X 10-6 K-cm2. SIMS data showed the presence of an interfacial
TiN layer. These samples also proved resistant to HC1 and HF.
Work has also been done with alloys of In, Ga, and Al nitride. Vartuli et al. has
examined the conduction mechanism for W and WSix contacts to InGaN and InN.24 The
dominant mechanism for annealed samples was concluded to be field emission as
determined by temperature dependent contact resistance measurements. The exception
was with as deposited metal on InN which exhibited thermionic emission.
It should be noted that the contact resistances for p-GaN are several orders of
magnitude higher than n-GaN because of the difficulty in achieving high carrier
concentrations. This stems from the low activation fraction of acceptor impurities
(-10%) and the presence of compensating defects like N vacancies. The approach to
forming contacts to p-type GaN has thus far been largely identical to that used for n-type
material. That is, a metal such as Ni, Au or Pt is deposited and then annealed, forming an
interfacial metal-nitride layer.
Trexler et al. compared the efficacy of Ni/Au, Pd/Au and Cr/Au contacts to p-
GaN having a carrier concentration of -1017/cm3.25 The structures consisted of 50 nm of
Ni, Pd or Cr capped with 100 nm of Au. All contacts were rectifying as deposited with
the Cr/Au contacts becoming ohmic after an RTA of 9000 C for 15 s. The specific
contact resistance for these contacts was ~ 4.3 X 10'1Q-cm2. Koide et al. confirmed the
relationship between contact resistance and metal work function for contacts to p-GaN.2
However, they concluded that the non-alloyed contacts (Ni and Ta) used for their study
did not provide low enough resistance for blue laser diodes. This work was corroborated
by that done by Ishikawa et al.27and Mori et al.28 The temperature dependence of Pt/Au
contact resistance was elucidated by King and coworkers.29 They found rectifying
behavior at room temperature changed to Ohmic above 2000 C. A minimum pc of 4.2 X
10-4 g-cm2 was obtained at 3500 C. The dopant was not activated until after metal
deposition when an anneal at 750 C for 10 min in N2 was performed. It is believed that
dopant activation may damage the p-GaN surface, creating compensating defects like N
vacancies. By using only one thermal treatment to both activate dopant species and
anneal the contact, compensation is minimized. Kim et al. used e-beam evaporated
Pd/Au contacts to achieve pc = 9.1 X 10-3 Q-cm2 after annealing.30 Diffusion of Cr into
p-GaN is believed by Yoo et al. to explain a pc = 1.2 X 10-4 9-cm2 for Cr/Au annealed at
500 C for 1 min.31 A tri-layer scheme consisting of Ni/Cr/Au was found by Kim et al. to
have pc = 8.3 X 10-2 O-cm2 to p-GaN and pc = 2.6 X 10-4 Q-cm2 to n-GaN, demonstrating
the ability to make ohmic contacts to both n- and p-GaN simultaneously.32
The literature indicates that ohmic contact research has been largely limited to
studying the effect of various metal deposition and annealing techniques. The use of
refractory metallization schemes useful for high temperature applications has been
ignored. Some consensus has developed that the formation of an interfacial nitride is
required for thermally stable n-type contacts.
2.5 The Intermediate Semiconductor Layer
A re-examination of the equations describing specific contact resistance presented
above reveals that the contact resistance can be lowered by reducing the interfacial
energy barrier that exists at the metal/semiconductor (M/S) interface or by creating a
highly doped layer in the semiconductor. The best approach to achieve either condition
is by creating an intermediate semiconductor layer (ISL) between the underlying active
layer and the contact metal, as illustrated in figure 2-5. The ISL must have a smaller
band-gap than the active layer in order to reduce the interfacial energy barrier. Further, to
promote tunneling of carriers, the ISL must have the capacity for being highly doped.
Figure 2-5. Schematic of ohmic contact formation to large bandgap semiconductor
using an intermediate semiconductor layer (ISL).
This approach has been successfully employed with GaAs devices utilizing a GaInAs
ISL.33 Analogous to what has been done with GaAs, InN based contact layers will be
explored for the nitrides. In addition to achieving low specific contact resistance,
thermally stable contacts are required to fully exploit the high-power and high-
temperature applications accessible by III-nitride based devices. This dissertation will
discuss fabrication of an ISL that reduces the energy barrier at the M/S interface, as this
method is required for large band gap semiconductors where dopant activation is low.
There are several techniques available to fabricate thin film semiconductor layers
suitable for ISLs. Metalorganic molecular beam epitaxy (MOMBE), metalorganic
chemical vapor deposition (MOCVD), and molecular beam epitaxy (MBE) are three
common approaches. The principles of these crystal growth techniques and their relative
merits will be examined in the following chapter.
EPITAXIAL GROWTH PRINCIPLES AND APPARATUS
This chapter will present the basic principles of growing high quality
semiconductor films, the basis for forming low resistance ohmic contacts using the ISL
approach. The bulk of the work done for this dissertation centered on selecting and
refining film growth conditions leading to epitaxial layers using molecular beam sources.
The term "epitaxy" is used to denote deposited layers that are in crystallographic registry
with the underlying single crystal substrate. That is, the deposited film adopts the long-
range atomic order associated with the position of atoms in the substrate. It is essential to
select film deposition parameters that result in epitaxial growth in order to realize optimal
semiconductor carrier transport properties. Crystalline defects such as dislocations,
antiphase domains, and grain boundaries degrade carrier mobility and contact
performance and should therefore be minimized. The degree to which epitaxial growth is
achieved is a function of the film growth conditions used and substrate choice.
The latter sections of this chapter will cover the three primary types of equipment
used to develop epitaxial films. Epitaxial growth equipment is generally very complex,
expensive, and difficult to maintain. Finally, a description of the hybrid apparatus
employed for this work is made.
3.1 Principles of Epitaxy
This section will present a qualitative development of the fundamental processes
leading to epitaxial film growth. A rigorous quantitative discussion is omitted because the
focus of this work is with experimental results, and the complexities of nitride
semiconductor growth are not amenable to precise modeling. Consequently, predictions
of film attributes derived from mathematical formulae are of little practical value. This
being the case, a qualitative, rather than quantitative treatment of growth theory is more
consistent with the realities of nitride epitaxy.
Epitaxial growth requires the confluence of several kinematic processes involving
the arrival, interaction and departure of atomic or molecular species at a heated surface.
Epitaxial films are grown under non-equilibrium conditions where an overpressure of
precursor species exists above the growing film. This gives rise to the formation and
growth of nuclei on the substrate and subsequent film growth. The growth rate is
determined primarily by the substrate temperature and the flux of film constituents
reaching the surface. An important point to emphasize is that final film characteristics are
the result of overall system energy minimization during growth. Hence, epitaxial quality
will be constrained by kinematic processes. The five salient features are as follows:
1. The arrival and adsorption of molecular or atomic species to the substrate or
2. Decomposition of molecular species.
3. Surface diffusion of adatoms.
4. Desorption of adatoms away from the surface.
5. Incorporation of adatoms into the growing film.
These five processes are illustrated in figure 3-1.
The arrival rate of precursors is determined by the temperature of effusion ovens
when solid sources are used, and the gas flow rate when gaseous sources are used.
Section 3.5 will provide a description of precursors used for film growth. When growth
takes place under conditions where the total pressure is greater than a few torr it is
possible to encounter prereactions in the gas phase that can reduce the arrival rate of film
constituents.34 This discussion will be limited to the case of high-vacuum deposition
conditions, as found with molecular beam growth techniques, therefore precursor
interactions can be neglected. However, parasitic prereactions must be considered when
higher pressure techniques such as metalorganic chemical vapor deposition (MOCVD)
2. 3. 4. 5.
Figure 3-1. Kinetic processes associated with film growth. 1. Arrival and
adsorption, 2. Decomposition, 3. Diffusion, 4. Desorption, 5. Incorporation.
Adsorption of chemical species on a surface will occur to reduce the free energy
of the adsorbate. As a molecule moves toward a surface, attractive forces arising from
coulombic, covalent, and van der Waals interactions are balanced by repulsive forces
arising from the overlap of filled electron orbitals. Figure 3-2 shows the resultant energy
curve. The force balance results in energy minimization when the adsorbate is a few
angstroms from the surface.
Decomposition of adsorbed molecular species will occur if the thermal energy
imparted by the heated substrate is sufficient to break chemical bonds comprising the
adsorbate. When metalorganic precursors are developed, care is taken to synthesize
molecules that will readily decompose at normal substrate temperatures (400-900 C). It
is possible for multi-component species to only partially decompose. This increases the
possibility of incorporating unwanted elements into the growing film, especially carbon
and oxygen. Usually, substrate temperatures are selected to maximize decomposition of
the precursor while maintaining a sufficient growth rate.
Figure 3-2. Energy curve resulting from summation of attractive and repulsive forces on
an adatom at a surface. The minimum energy position indicates the equilibrium adatom
Surface diffusion of adatoms occurs when adatoms acquire sufficient thermal
energy to breach energy barriers imposed by the periodic array of atoms in the substrate.
Figure 3-3 illustrates schematically the activation energy required for an adatom to
translate to neighboring lattice sites. Usually this energy is much less than that needed for
desorption of the adatom, which requires bond breaking. Diffusivity is a function of
substrate temperature and the distance an adatom migrates will increase with temperature
(up to the point when desorption becomes dominant). For epitaxy to occur the adatom
must have sufficient surface diffusivity to incorporate at low energy surface sites. In
Figure 3-3. Schematic representation of adatom diffusion illustrating energy barrier to
some cases, surfactants may be utilized to lower the activation energy associated with
surface diffusion. This results in an increased diffusion length, thereby increasing the
probability of adatom incorporation at low energy lattice positions. Otherwise the
adatom may incorporate at sites resulting in amorphous or polycrystalline material.
Low energy sites correspond to crystallographic lattice positions that occur at
ledges or kinks on the surface, as illustrated in figure 3-4. The lower energy is a
consequence of the adatom retaining fewer dangling bonds at ledges or kinks as
compared to terrace positions. Once the adatom encounters a ledge or kink position, it is
no longer energetically favorable to continue migrating and it will incorporate into the
Figure 3-4. Schematic of growth surface showing possible incorporation
sites for the adatom.
Desorption of adatoms can be described by a desorption frequency, Vdes that is a
function of the atomic vibrational frequency, vo ( 10I3/s) and an activation energy which
is characteristic of the escape process, AGdes. The relationship is of the form:
Vdes = VO exp Gd
where k is Boltzmann's constant. The growth rate of the epilayer will begin to decrease
when Ts becomes high enough to promote desorption of adatoms over surface diffusion.
3.2 Models of Nucleation
The most important stage of film growth is nucleation. Nucleation determines the
resulting structural characteristics of a film and whether epitaxy will result. There are
three widely accepted modes of film nucleation leading to epitaxial growth, shown
schematically in figure 3-5. Three-dimensional island formation on a bare substrate
(Volmer-Weber or 3D), layer-by-layer deposition (Frank-van der Merwe or 2D), and
island formation on a thin uniform layer (Stranski-Krastanov). The mode of nucleation is
determined by surface energy minimization criteria, as discussed qualitatively below.
The variables a and oe represent the surface energy of the substrate and the
epilayer, respectively. Generally, Volmer-Weber (VW) growth will dominate when
much of the low energy substrate exposed as possible while minimizing the surface area
of the high energy epilayer. Crystal surface energy is typically anisotropic, meaning
Volmer-Weber (VW), islanding
Frank- van der Merwe (FM), layer by layer
Stranski-Krastanov (SK), layer-island
Figure 3-5. The three nucleation mechanisms leading to epitaxial growth.
. .. .. .. ..
some crystal planes have higher surface energy than others. The shape of an island will
be one that maximizes the area of low energy facets while minimizing high energy ones.
Frank-van der Merwe (FM) growth is favored if as>7e. Energy minimization is
accomplished by complete coverage of the substrate by the epilayer. In some cases,
interfacial energy differences between substrate and epilayer and between monolayers in
the film will favor Stranski-Krastanov (SK) growth. This situation implies there is
relatively high interfacial energy between the first and second monolayers of the film.
Energy is minimized by island formation on top of a thin uniform layer.
Ideally, 2-D (FM) growth is desired for thin film semiconductor layers. This
results in smooth, uniform layers with fewer antiphase domains, dislocations, or grain
boundaries. Epilayer smoothness is especially important when multi-layer structures are
utilized as film roughness can be translated to, and exacerbated by, subsequent
deposition. High surface roughness can contribute to irregular metallization behavior.
Microscopic thickness variations in metal contacts cause current density variations
leading to spiking of metal into the active region of a device.
Additional issues must be considered when film growth is attempted on a
substrate that is unlike the film, i. e. GaN/sapphire. It is not always possible to find a
substrate that has a lattice dimension matching that of the epilayer, af. The atoms in the
epilayer will attempt to match the atom spacing of the substrate, as, via bond distortion,
creating strain energy. The strain energy is a function of film thickness and can be
0rs =- dff2
Young's modulus is Y, d is film thickness and f =(laf as )/aavg is the degree of
lattice mismatch. The critical thickness, dc, occurs when the strain energy matches the
energy required to form a misfit dislocation in the film. When d, is exceeded, a misfit
dislocation will form that lowers the total energy of the system and the film is considered
relaxed. Strained films that are less than dc will adopt the in-plane lattice parameter of
the substrate with a corresponding deformation of the film perpendicular to the interface.
Such films are qualified as pseudomorphic, as figure 3-6 illustrates.
The critical thickness, dc, is given by Mathews and Blakeslee as
b(1 cos2 0) d +
c 2f(1 + ) cos b b
where b is a Burger's vector on the order of as and af, [t is Poisson's ratio, 0 is the angle
between the dislocation line and b, ( is the angle between the slip direction and the
direction in the film plane which is perpendicular to the line of intersection of the slip
plane and the interface.35
By way of example, a film with f=1.0% will start generating misfit dislocations at
dc = 30nm. This corresponds to a misfit dislocation every 30 to 50 nm along the interface
between film and substrate. Dislocations are known to be deleterious to device
performance, as they act as non-radiative recombination centers for photons and
scattering sites for current carriers. Therefore, it is desired to minimize the opportunity
for these defects to form, usually by finding substrates that offer f< 1.0%. Given fGaN/sapp
__j1 -- Substrate and film with lattice
af dimension as and af, respectively
}afp af Mismatch accommodated by strain,
Mismatch accommodated by misfit
dislocations, film thickness >dc,
Figure 3-6. Illustration of substrate and free standing (unstrained) film (top).
Heteroepitaxial growth leading to a pseudomorphic film (middle). Film thickness
exceeding critical thickness, dc, causing misfit dislocations (bottom).
S14% would seem to eliminate sapphire as a useful substrate for GaN growth. However,
nitride devices have shown a unique ability to withstand defect densities orders of
magnitude greater than devices fabricated from traditional semiconductors.
Heteroepitaxial films may face the problem of thermal mismatch as well.
Differences in thermal expansion coefficients between the film and substrate can generate
strain when the film is cooled from the growth temperature. In some cases the strain
developed during cool-down can result in cracking of the epilayer.
3.4 Graded Layers
Heteroepitaxy can also refer to film growth on a dissimilar epilayer that has been
previously deposited. Compound semiconductor devices rely on the ability to construct
unique band structures through multiple epilayer deposition. The conduction band and
valence band positions at a heterointerface in part determine the operating characteristics
of a device. Figure 3-7 illustrates the three types of band offsets that can occur at a
heterointerface. Note that this schematic represents offsets that occur before thermal
equilibrium has been established between the two films. The quantities AEc and AEv
represent the conduction band and valence band energy differences, respectively. After
thermal equilibrium is established, band bending will occur at the interface. The final
shape of the band diagram will depend on the initial position of the Fermi energy in each
layer. For certain systems, the heterointerface will contain an energy barrier to charge
transport across the interface. It is possible to ameliorate the effects of this barrier
through the use of a graded layer.
A B A
Figure 3-7. Types of band offsets at a heterointerface. Straddling (left), staggered
(middle), and broken-gap (right). [after reference 36]
A graded layer involves continuously varying the composition of an intermediate
layer to form a smooth transition between two films having different components. For
example, to grade between films consisting of AQ and BQ, one would grow a layer of Ai.
xBxQ with 0< x < 1 (see figure 3-8). The graded layer thickness is on the order of 10nm.
The graded layer removes or reduces spikes in the band structure at the interface,
augmenting carrier transport across the junction. This technique will be employed for
improved ohmic contact performance to nitrides, as discussed in Chapter 6.
3.5 Comparison of Growth Techniques
Thus far, the basic concepts of epitaxial film growth have been presented. This
section will introduce and discuss the apparatus utilized for epitaxial film growth. A
comparison of the three most common growth systems: MBE, MOCVD, and MOMBE is
Figure 3-8. Diagram of band structure at a heterointerface without (a) and with (b)
a graded layer.
Molecular beam growth implies that the ambient pressure above the growth
surface is low enough that gas phase interactions are essentially nonexistent. Source
atoms or molecules entering the growth chamber can be directed toward the substrate in a
near collimated beam and arrive at the substrate unimpeded. This condition is considered
the molecular flow regime. Background chamber pressure (not during growth) can
approach 10-" torr. During growth, pressures of 10-6 torr are typical. Because of the
extremely low pressures involved, MBE growth is highly sensitive to contamination and
great lengths are extended to avoid introducing unwanted substances into the growth
Molecular beam epitaxial growth systems typically consist of an ultra-high
vacuum (UHV) deposition chamber, a buffer chamber and load-lock. Vacuum levels are
achieved with several types of pumps including mechanical, turbo, cryo, ion, effusion and
turbomolecular. The use of a load-lock and buffer chamber allows substrates to be
introduced to the growth chamber without contaminating it. The growth chamber is
continually maintained at UHV conditions unless sources need replenishing or repairs
must be made.
Solid sources are contained in an effusion cell that is directed toward the heated
substrate, see figure 3-9. The effusion cell assembly is bolted to a stainless steel source
flange on the growth chamber. An inert ceramic insert holds the source and is heated with
a resistive coil. Temperature is monitored with a thermocouple that is part of a feedback
loop attached to a temperature control unit. This way, a temperature setpoint can be
entered and the control unit will automatically adjust the voltage and current to the heater
coil to reach and maintain the prescribed setpoint. When not in use, a setpoint of -1000 C
is maintained to prevent adsorption of unwanted contaminants onto the cell. The effusion
cell is outfitted with a shutter that can be opened or closed rapidly, starting or stopping
the flux of atoms as needed. This permits the fabrication of abrupt interfaces when
growing multilayer structures. The flux, Fs of atoms of molecular weight M, impinging
on a substrate a distance I from the aperture with area A, at temperature T and cell
pressure P is given by
F, = 1.12x1022 PA
S12 -T molecules/cms.
Figure 3-9. Schematic of MBE effusion cell and substrate arrangement.
If solid Ga (M = 70) is used as a source, for example, the vapor pressure will be 2.2 x10-3
torr at a cell temperature of 9700 C. Given a source to substrate distance of 15 cm and an
opening of 4 cm2 on the effusion cell, Fs will be w 1.5 x 1015/cm2s. This corresponds to a
growth rate of -1.5 monolayers, (8A) per second. Commonly, an overpressure of the
group V (i. e. nitrogen) element is maintained because it will adhere only to an existing
layer of group III atoms and is therefore self limiting. The arrival rate of the group III
species is usually rate limiting.
It is common to find gaseous sources used in conjunction with solid sources for
some semiconductor growth systems. For example, hydrides such as arsine, AsH3 and
phosphine, PH3 are used as sources for arsenic and phosphorous. Ultrapure N2 is
generally used for growth of nitride semiconductors. The gas is passed through a plasma
..............-- ............. ............
generator prior to injection into the growth chamber to provide active monatomic
nitrogen species. More on this will be presented in section 3.6. Gas flows are controlled
using a mass flow controller (MFC), with typical flow rates ranging from 1.0 to 20
standard cubic centimeters/s (sccm).
Even though great pains are taken to ensure a pristine growth environment,
invariably impurities will persist. Unwanted contaminants such as carbon and oxygen
may be incorporated into the growing film. The likelihood of contaminant incorporation
will be a function not only of their relative abundance but also of the gettering ability of
the arriving growth constituents. As the bond strength of the source atoms increases, the
probability of their bonding with and incorporating impurities increases. Table III
compares bond strengths of some common constituents. Included for comparison are
methyl bonded precursors that are typical of MOCVD and MOMBE. The relatively high
bond strength of nitrogen can lead to high impurity incorporation levels making nitride
growth especially difficult.
Because of the possibility of contamination, the growth rate should be as high as
possible without sacrificing material quality. Typical MBE growth rates range from
several hundred to several thousand angstroms per hour, depending on the layer to be
The UHV environment associated with MBE growth is amenable to a variety of
in-situ characterization techniques. A residual gas analyzer (RGA) or quadrapole mass
spectrometer (QMS) can be attached to the growth chamber allowing for analysis of
constituents present during growth. This is possible because there are unincorporated
species that will remain in the growth chamber long enough to be detected by the analysis
equipment. Another powerful analytical tool is reflective high-energy electron
diffraction (RHEED). Here, an electron gun is positioned on one side of the growth
chamber to send a beam of electrons to the substrate or film surface at a glancing angle.
The diffracted beam is collected on a phosphor screen on the opposite side of the growth
chamber. Information about surface condition can be discerned such as degree of
crystallinity. Further, spot intensity variations over time can be used to determine growth
rate during the early stages of film growth.
Table III. Relative
MOCVD and MBE.
[after reference 37]
bond strengths of constituents commonly found in MBE,
The stronger the bond, the easier the impurity incorporation.
MBE is an excellent method of producing complex multilayer structures where
precise control of dopant concentration, layer thickness and interface abruptness is
required. Unfortunately, source replenishment can be a long and difficult process usually
Bond Bond Strength (kcal/mole)
involving venting the entire growth chamber. Once the growth chamber has been
exposed to air, the inside walls and components will be coated with moisture and
everything else found in the ambient. The removal of such unwanted adsorbates requires
heating the evacuated growth chamber over a period of days. Usually, a bake-out will be
done prior to venting as well in order to reduce the chances of introducing hazardous
adsorbed material into the atmosphere. The whole process can take up to a week, making
source replenishment a significant drawback to MBE growth.
Metalorganic chemical vapor deposition, sometimes referred to as organometallic
vapor phase epitaxy (OMVPE), has emerged as the most common apparatus for nitride
growth. The technique relies on the decomposition and reaction of a metalorganic group
III species (i. e. trimethylgallium, TMG) with a hydride group V species (i. e. ammonia,
NH3). Chemical decomposition and reaction take place on a heated substrate. For GaN
growth, a typical reaction might be
Ga(CH3)3 + NH3 = GaN + 3CH4.
As in MBE growth, the group V species will be in excess over the group III species. The
substrate temperature will be well above that required for pyrolisis of the precursors,
making the arrival rate of group III constituents rate limiting. The operating pressures are
orders of magnitude greater than that found with MBE systems, typically 10 to several
Metalorganic sources are available as solids or liquids. The chemical is contained
in a stainless steel bubbler having an inlet and outlet manual valve. The bubbler is
immersed in a chilled water bath that maintains a constant temperature. A temperature is
selected that will give a partial pressure from a few tenths to several torr. Material is
transported to the growth chamber via a carrier gas, usually H2, that flows through the
inlet and outlet of the bubbler. The carrier gas flow is controlled using an MFC and line
pressure is commonly controlled with a capacitance manometer. Carrier gas flows on the
order of 100 seem are common. Hydride flows of up to a few thousand seem are
possible. The relatively high flow of gases into the reactor tends to reduce the
incorporation of unwanted impurities by having a gettering effect. The flux of
metalorganic, FMo, injected into the reactor is given by
FMO = CGMO
where FCG is the carrier gas flux, PMO is the partial pressure of the metalorganic, and PT is
the total pressure. The flow of gas above the substrate creates a stagnant boundary layer
through which constituents must diffuse.
The key to MOCVD growth is the complex gas handling equipment that allows
for switching of sources into the reaction chamber in a precise manner. In addition to the
manual valves located on the bubbler, it is common to have electronically controlled
inlet, outlet and bypass valves on a bubbler manifold. These are supplemented with
carrier gas on/off valves, vent and run valves. Figure 3-10 illustrates the arrangement
used in a typical metalorganic delivery system. This arrangement is also found on
MOMBE systems including the Varian Gen II used for this study.
MOCVD growth offers the advantage of easy source replenishment. Since the
bubblers can be valved off from the rest of the system, it is possible to replace them
Carrier gas in
4- Manual Valve
Figure 3-10. Schematic of metalorganic delivery system showing bubbler, carrier gas
transport and control elements.
A. Carrier gas inlet valve
B. Bubbler inlet
C. Bubbler bypass
D. Bubbler outlet
E. Vent valve
quickly without risk of contaminating the reactor. Moreover, since UHV conditions are
not required, massive pumping capacity is unnecessary.
The high pressures employed during growth preclude the use of most in-situ
analytical devices such as RHEED and mass spectrometry. The high flow rates can lead
to parasitic gas phase prereactions that have a deleterious effect on growth rate. Care
must be taken to separate overly reactive precursors until just before reaching the
substrate, making hardware design of critical importance. Sometimes it may be
necessary to choose alternative precursors to lessen the chance of gas phase reactions.
Another major disadvantage is that high substrate temperatures are generally required.
For nitride growth, the substrate must be hot enough to decompose a sufficient quantity
of NH3 to provide reactive nitrogen. Substrate temperatures of >7000 C are usually
required. This makes growing films consisting of high vapor pressure elements such as
indium very difficult.
Metalorganic molecular beam epitaxy is a blending of the two techniques
previously discussed. It employs the UHV environment of MBE, making in-situ analysis
possible, along with the ease of source replenishment associated with the use of bubblers.
This flexibility comes at a price, however, in that the challenges of maintaining a UHV
environment are combined with the complexities of gas handling equipment. The
pressure regime during growth is higher than standard MBE but is still orders of
magnitude less than MOCVD. Metalorganic sources used for MOCVD also work well
for MOMBE, although the gas flow rates are much less with the latter system. Figure 3-
11 summarizes the differences in constituent arrival and surface interactions for the three
3.6 The Varian Gas-Source Gen II
This section will describe the specific MOMBE system used for fabricating
samples for this work. The equipment is housed at the Microfabritech facility on the UF
campus. Figure 3-12 shows the layout of the growth chamber, buffer chamber, load lock,
metalorganic bubbler manifold cabinet and pumping systems. The system can
accommodate up to six metalorganic sources, four solid sources, and three hydrides. In
addition, reactive nitrogen is supplied by either an electron cyclotron resonance (ECR) or
radio frequency (RF) plasma generator. A quadropole mass analyzer allows identification
and monitoring of growth species artifacts. A schematic of the growth chamber showing
the source flange and substrate heater is provided in figure 3-13.
3.6.1 Vacuum Pumping Systems
As with MBE systems, the Varian has a host of pumps to handle the high-vacuum
needs of the system. Pressure is monitored in the load-lock, buffer and growth chambers
with ion gauges. Additional pressure gauges are present on the vent line, fore line and
upstream of the molecular drag pump located by the load lock. The load-lock is roughed
using a high-vacuum compatible mechanical pump. When the pressure has reached a few
torr, a molecular drag pump (MDP) is initiated that will bring the pressure to the 10-4 torr
regime. At this point, a gate valve isolating a CTI 100 cryopump is opened allowing the
pressure to reach <10-6 torr. The buffer chamber is pumped with an ion pump and reaches
a base pressure of 10-8 torr. The growth chamber is pumped with a dual rotor, 2200
liter/s Balzers turbopump. This pump is excellent at removing most species including
He, and maintains a background pressure of ~107 torr. The turbopump is backed by a
second mechanical pump that is lubricated with special perfluorinated oils that are
hydride compatible. During growth, a CTI-8 cryopump is brought into service to assist
the turbo with the high pumping capacity required when decomposition products are
present in the growth chamber. When not in service, the CTI-8 is isolated with a gate
valve. Growth chamber pumping is further assisted by the use of a cryo-shroud. The
stainless steel growth chamber is actually a double walled vessel that can be filled with
liquid nitrogen prior to growth. The cold inner walls can remove additional
decomposition products and other species by condensation. Condensed gasses are then
pumped out with the turbo pump after growth is completed and the chamber walls are
allowed to return to room temperature.
Aside from the pumps required for achieving and maintaining high vacuum on the
load-lock, buffer and growth chambers, there is a second TMP used to evacuate the vent
line. This is backed by a third mechanical pump.
As mentioned above, the Varian possesses a variety of sources for epitaxial
semiconductor growth and doping. The four solid metal sources are housed in effusion
cells exactly as described for use by MBE growth. Gallium and indium are routinely kept
charged while the other two sources may be changed periodically, depending on the
needs of the users.
* Group HI Alkyls
S_._ Stagnant Boundary
S.VI 0 0 Kinetic Process
.. .. < 4.^...... . ..; .. i e i
S4-- Group III Alkyl
I) Surface Pyrolysis
Figure 3-11. Schematic depicting the differences in constituent arrival and surface
interaction for three common epitaxial techniques.
Drag Pump ii]
to exhaust 4
Figure 3-12. Schematic of Abernathy MOMBE lab. [after reference 38]
Figure 3-13. Schematic of MOMBE growth chamber. [after reference 39]
Metalorganic sources include trimethylindium (TMI), trimethylgallium (TMG),
dimethylethylaminealane (DMEAA), and carbontetrabromide (CBr4) among others. A
discussion of the pertinent features of specific metalorganic precursors will be made in
chapter five. The chemicals are contained in bubblers, just as described for use in
MOCVD (section 3.5.2). Each source is equipped with a full complement of valving as
shown in figure 3-10 that is housed inside an evacuated manifold cabinet. Care must be
taken when handling these sources as they can be pyrophoric and in some cases toxic.
These sources can be changed as needed with relatively little effort.
The Varian has three hydrides, arsine (AsH3), phosphine (PH3) and silane (SiH4).
Owing to their extreme toxicity, these gaseous sources are contained in pressurized
cylinders housed in a special toxic gas cabinet outside the lab. The gasses are conducted
to the growth chamber via double-walled stainless steel tubing especially designed for the
task. This tubing is itself enclosed in conduit that is connected to a scrubbed exhaust
system. The hydrides pass through a cracker which decomposes them before entering the
growth chamber. The cracker consists of a tantalum catalyst with resistively heated coils
and a thermal control system similar to that used for solid sources. During growth, a set-
point of 1000 C is maintained in order to adequately decompose the hydrides and
provide a reasonable flux of As, P or Si. When not in use, the cracker is sustained at -
650 C to avoid buildup of adsorbates.
Since N2 is extremely difficult to thermally decompose, and rather unreactive,
temperature independent techniques must be adopted to provide a usable nitrogen flux.
One way is by forming a plasma from a flow of nitrogen gas. Among other things, a
nitrogen plasma will contain levels of neutral atomic N. It is this highly reactive species
that is believed to be important for nitride growth. The Varian has been fitted with either
of two types of plasma sources, ECR or RF. These compact plasma generators can be
bolted to the source flange in a way similar to a solid source effusion cell and are
sometimes referred to as "remote" plasma sources. Initially a Wavemat MPDR 610 ECR
plasma source was used for sample development, this was later switched to an SVT
Assoc. RF source.
The ECR source consists of a cell that is connected to an ultra-pure, mass flow
controlled N2 supply. The cell is surrounded by a series of permanent magnets that
generate a field along its axis. Microwave energy at 2.45 GHz is coupled into the source
gas (N2) creating a complex mixture of electrons, charged and neutral species. The
magnetic field will cause electrons to accelerate in a helical path, promoting further
ionization and neutral species generation. Power can be independently controlled with a
typical range of 150 to 300 W. Varying the power allows reactive species to escape the
cell aperture with various energies. The aperture is covered by a biasable grid that is
attached to an independent power supply. This allows charged species to be
preferentially selected or excluded from the cell.
The RF plasma source was produced by SVT associates and uses energy at 13.56
MHz. An MFC is used to carefully regulate gas flow into the cell, where energy is
coupled to create a plasma. The RF power supply can be varied to control the energy of
species effusing out, usually 200 to 400 W.
Using plasma generated reactive nitrogen obviates the need for high substrate
temperatures, as would be found with MOCVD growth. This affords the Varian a
distinct advantage over other growth systems by allowing access to much lower substrate
temperatures. It would be virtually impossible to grow nitride samples containing
significant quantities of high vapor pressure elements (i. e. indium) without a plasma
The ability to utilize solid, metalorganic and hydride sources under UHV
conditions make the Varian an exceptional semiconductor research tool. The instrument
permits the investigation of a variety of unique compounds and complex structures
inaccessible by other techniques. It is not unusual to select a path of experimental
endeavor based on the research equipment available to the investigator. The
development of improved ohmic contacts for nitride based devices using an intermediate
semiconductor layer is well suited to the capabilities of the Abernathy MOMBE lab.
3.6.3 MOMBE Safety
Considerable effort has been taken to ensure that the toxic and pyrophoric
substances associated with MOMBE growth will not pose a threat to the lab and other
building personnel. Any areas that could possibly be exposed to hazardous gasses are
housed in special enclosures that are evacuated through a facility scrubber designed to
decontaminate the gas before being introduced to the environment. This includes the
source flange, metalorganic manifold cabinet, and the turbo backing mechanical pump.
After passing through the turbopump, residual growth chamber gasses are pumped
through Carusorb scrubber material before reaching the mechanical pump. The
mechanical pump exhaust is in turn scrubbed by the facility scrubber. Additional
scrubber material is used on the vent line.
Toxic gas monitoring points are located in the manifold cabinet, turbo backing
mechanical pump covering, injection flange shroud, and other critical areas. Any leaks
detected will sound a special toxic gas alarm and emergency personnel are automatically
contacted. The toxic gas monitor is interlocked with the Varian gas handling valves
which are closed automatically in the event of an alarm. In case of a power failure, a
diesel generator is automatically brought on line to maintain monitoring service.
In addition to being interlocked to the toxic gas alarm, all process valves will be
automatically closed in the event of a power outage. Because the Microfabritech facility
is subject to occasional power dips and transients that may last less than a second, it was
necessary to install an uninteruptable power supply (UPS). This battery array conditions
all the power going into the various MOMBE components and accommodates any
transients in service. When all systems (pumps, power supplies, etc...) are in operation,
the UPS can sustain them for up to 20 minutes before the batteries become drained.
When the growth chamber or any other toxic gas exposed equipment is vented,
self-contained breathing apparatus (SCBA) is worn to protect the worker from any
residual toxic substances that might be inhaled. Also, all others working at the
Microfabritech facility are notified of the event and are kept clear of the MOMBE area.
Aside from hazards to the environment and humanity, care must be taken to
ensure the service life of MOMBE components. Because high-vacuum components and
pumps are extremely costly, it is important to follow proper procedures when working
with the MOMBE. Opening valves in the incorrect order, for example, can be
detrimental to the functioning of pumps or ion gauges.
3.6.4 MOMBE Maintenance
Owing to the many systems (pumping, power supply, chilled water baths) the
MOMBE requires frequent checks and servicing. Some maintenance duties simply
involve visual inspection to ensure proper functioning. The recirculating chilled water
baths used to maintain bubblers at constant temperature must be kept filled and any ice
buildup removed. Occasional addition of ethylene glycol may be necessary. The water
levels the bubblers are submerged in must be kept filled.
The two cryopumps must be regenerated on a regular basis. Regeneration entails
allowing the pump to warm to room temperature and pumping out all adsorbed species.
The frequency depends on how much gas load the pumps have seen, usually once or
twice a week. The compressors must be checked to ensure proper He pressure.
Mechanical and turbo pump oil levels and quality must be maintained. If pump
oil starts to become cloudy, it must be replaced. This requires the use of SCBA. Used
pump oil must be properly labeled and disposed of. All pump oil is considered
contaminated and must be handled accordingly.
Once every six months or so, the Carusorb scrubber material both downstream
from the turbopump and on the vent line must be replaced. Used Carusorb is considered
contaminated and SCBA must be used.
Less frequently, solid sources must be recharged. This requires venting the
growth chamber. Prior to venting, the growth chamber bake panels are installed and the
system is baked for -72 hours. This helps remove any volatile species that may have
adsorbed onto inner surfaces. Viewports can become coated with residue from
metalorganics or solid sources and must be removed and cleaned. Ion-guage filaments
may need replacing. After all work has been completed, the system is pumped down and
baked for another 72 hours prior to use.
This chapter discusses the methods, principles and rationale for the various
characterization techniques used to discern the properties of epitaxial thin films grown by
MOMBE. It may be the investigators first impulse to perform a full battery of analysis
techniques for every sample. However, time and money can be conserved if some
thought is given prior to examining a sample.
There are four areas of sample analysis that may be examined: electrical,
structural, morphological, and compositional. Within these four areas are techniques that
are destructive and non-destructive. Ideally, nondestructive analysis is done first and as
little material as possible is consumed by destructive techniques.
It is important to confirm results by making sure the total data set for a given
sample is self-consistent. Electrical and compositional information should be
complementary, for example. Also, if there is some doubt as to the validity of a
measurement, it should be verified by another technique.
4.1 Electrical Characterization
In some ways, the electrical properties of a sample (carrier concentration,
resistivity, and mobility) are the most relevant for semiconductor research. Ultimately, it
is these attributes that determine the utility of a material for use in an electronic device
structure. The development of ISLs for improved ohmic contact performance intimately
relies on knowledge of the electrical properties of thin epitaxial films. Therefore it is
imperative to have a reliable way of determining these properties.
4.1.1 The Hall Measurement
The resistivity, carrier mobility, and carrier concentration can be determined using
a technique based on the Hall effect. Figure 4-1 shows a schematic of a sample
illustrating the Hall effect. Here, an n-type semiconductor has current flowing in the + x-
direction with a magnetic field in the + z-direction. Electrons will be subjected to a force,
known as the Lorentz force, that causes charge accumulation on one side of the sample.
This charge separation induces an electric field known as the Hall field or Hall voltage,
The sample geometry shown in figure 4-2 is known as the van der Pauw
geometry, and gives a resistivity, p, of
7Jd (RAB,CD + RBCDA) AB,CD
In 2 2 RBC,DA
where R is resistance (=V/I), d is film thickness, and f is an ideality factor based on the
symmetry of the contacts and is unity when RAB,CD=RBC,DA. For a sample with magnetic
field normal to the surface, the Hall voltage is given as
AV = PHIiP
where [AH is the Hall mobility and j is current flowing from contact A to C. The Hall
mobility can be computed by
Figure 4-1. Schematic of n-type semiconductor with current, jx, and magnetic
field, B. Electrons accumulate on one side of the sample creating an electric field,
E, the Hall field.
Figure 4-2. Contact geometry of van der Pauw method for performing a Hall
where AR is the change in resistance caused by the magnetic field. The carrier
concentration, n, is then calculated as
with q=1.6x 10-19 coulombs.
Usually, contacts consist of an Hg/In amalgam that has been annealed at 4000 C
for three minutes under flowing nitrogen. A typical sample size is on the order of lxl
cm2. This technique is semi-destructive in that the area covered by metal contacts is not
recoverable but the rest of the surface can still be examined. In practice, a change in
voltage is measured from which AR is calculated, hence it is important that the current
used falls within the ohmic region for the sample. Hall measurements are limited by the
ability to measure a change in voltage. Usually a variation of several millivolts is
required for an accurate measurement. The sample carrier concentration and quality of
the contacts will determine the ability to obtain useful data. Samples with less than 1016
carriers/cm3 will generally not be good candidates for a Hall measurement. In order to
avoid carrier depletion problems, films should be at least 200 nm thick (depending on
4.1.2 Capacitance-Voltage (C-V) Measurement
Another method for determining carrier concentration is based on the fact that
semiconductors will acquire a depletion layer capacitance when subject to a reverse bias.
The magnitude of the capacitance is a function of the number of carriers. For a one-sided
abrupt junctions. An incremental change in voltage, dV, will result in an incremental
increase in charge per unit volume, giving
where q is the electron charge, N is the carrier concentration, and sE is the permitivity.40
A plot of 1/C2 vs applied voltage will yield a line whose slope gives the carrier
concentration. A Hg probe station can be used to quickly and nondestructively form an
electrical contact to the sample. No annealing of contacts is required. Commercial
systems are available that will interface with the probe station and automatically perform
the measurement and calculate a carrier concentration. Sample size can be anywhere
from -1/6 wafer (2") to a full wafer. This method of measurement tends to break down
when the semiconductor is degenerate (-10'8/cm3)and depletion widths become very
4.2 Structural Characterization
Structural analysis will reveal the overall crystal type of a sample, for example if the
film is wurtzitic or zinc blende, and what its lattice constants are. Structural
characterization is also performed in order to elucidate the types and number of
crystalline defects such as dislocations. Structural imperfections and crystal type will in
part determine the electrical and optical characteristics of a semiconductor film.
Therefore, crystalline structure information can help develop an understanding of the
observed electrical behavior of a sample. Usually, a relationship between the growth
conditions used, the resulting crystalline structure, and electrical behavior can be
established. This information can then be used to direct future research.
4.2.1 X-ray Diffraction (XRD)
This technique is based on the diffraction behavior of photons in a crystalline lattice.
Bragg established the relationship between wavelength, k, and lattice dimension, d, as
n = 2d sin 0, where theta is as shown in figure 4-3. In this construction, it is assumed
that specular reflection of the incident radiation occurs. Then, constructive interference
of photons will occur when the path difference, 2dsinO, is an integer multiple, n, of the
wavelength. An x-ray diffractometer (see figure 4-4) will scan through all angles theta
and detect any reflected energy. When the Bragg condition is satisfied for a family of
planes (i.e. (0002) or (0004)), the intensity of diffracted energy collected by the x-ray
detector will increase. A plot of detected intensity versus angle can be generated. The
instrument used for this study was a Phillips Electronic Instruments APD 3720, located in
the Major Analytical Instrumentation Center (MAIC)
The peak positions will be characteristic of a sample with a certain structure and
lattice constant. The composition of ternary alloys can be estimated from peak position
as well. For example, if a sample of InxAll-xN has a peak between the (0002) InN and
(0002) AIN peak, it may be inferred from Vegard's law and the relative position of the
intermediate peak, what the value of x is.
Intensity peaks will have some broadening from variations in lattice constant due to
thermal energy vibration. Peak width can give an indication of the number of linear or
planar defects in the sample. This results from variations in the lattice constant around a
defective region. In this case, the Bragg condition will be met for more than one narrow
set of angles. The more defective the sample, the broader the characteristic peak.
The signal detected will be a function of the film thickness. Usually, samples
should be on the order of 400 nm or more to get useful information. XRD is a completely
nondestructive technique that can be done relatively quickly (an hour or less). Samples
are mounted onto a glass slide using double-sided tape. This assembly is placed in an
enclosed stage that allows x-rays from the source to diffract into the detector. Since x-
rays are extremely dangerous, safety protocol must be followed when operating the
"""""-S-\-- S -1 ""
d 0""- ./ ...
Figure 4-3. Schematic of photon diffraction off of a crystal lattice satisfying the
Bragg condition, leading to constructive interference at angle of incidence, 0.
Figure 4-4. Geometry of sample, x-ray source, and detector used for x-ray
4.2.2 Transmission Electron Microscopy (TEM)
TEM is a powerful tool that relies on the ability of a high-energy electron beam to
transmit through a sample. A filament is used to generate a stream of electrons that is
collimated by a series of electromagnetic lenses. This beam is then brought to bear on a
specimen. More lenses are used to collect the diffracted and transmitted electrons.
Contrast is generated by variations in scattering of the electron beam by the sample.
Magnification up to 1,000,000X is possible for certain samples when the instrument is in
the hands of a competent operator. Information on crystal structure including type, lattice
constant, and defect density is obtainable. Figure 4-5 illustrates how contrast formation
occurs allowing imaging of line defects. Exceptional imaging of heterostructure
interfaces is also possible showing threading dislocations and any residual surface
Unfortunately, TEM requires samples that are thin enough to allow electrons with
energies of 100 KeV to be transmitted. This requires samples to be thinned to
approximately 100 nm or less, an often difficult task. Sample prep for cross-sectional
imaging starts by gluing a stack of layers together that contains the sample of interest
bonded face to face as seen in figure 4-6. This makes handling of the specimen easier for
the subsequent thinning steps. Typically, thinning is initiated with sandpaper of
increasingly fine grade. Special lapping films and grinders are then used to polish the
sample further. A specially designed instrument known as a dimpler is used to etch a
concave depression in the sample. Final thinning is done with an ion-mill. Sample
preparation is an art that can take years to perfect.
\ \\ 1
\ \\ I 1
\ \ I
II I II
\ \ A /
I \ / \ I
---I __ --I __ ___ ___ ___ __ ^- __ ^ __ __
,/ \ Lattice planes
I I \ \
I I \ \
1 \ \
II \ \
SI \ \
iI i \
II I \ \
I i \ \
Figure 4-5. Schematic of electron/ sample interactions leading to contrast formation
around a dislocation.
I i I
i i I
I I i
I I i
I I I
I i I
i i I
I I I
Dummy layer from scrap Si wafer bonded to
Sbacksides of sample for ease of handling
I Epitaxial film bonded face to face
Material removed from this surface
Figure 4-6. Sample stack glued for cross sectional TEM sample prep. Each layer is cut
from a wafer and is ~ 5x5 mm.
Along with sample preparation, sample imaging can be a skill that takes many
months, if not years to develop. For this reason, only samples perceived as having
exceptional qualities are selected for TEM analysis. Sample prep and imaging for this
work was conducted by fellow Abernathy lab mate Brent Gila. The instrument used was
a JEOL 200CX at MAIC.
4.3 Morphological Characterization
Information on surface roughness can be invaluable, especially when film
smoothness is critical, as is the case when developing multilayer structures. An
examination of the surface of a sample can give clues about growth evolution and film
structure. Establishing a relationship between epilayer growth conditions and surface
morphology is the desired goal.
Often, surface features of interest may not be revealed with standard optical
microscopy. In such cases, scanning electron microscopy (SEM) or atomic force
microscopy (AFM) can be used to assess film surfaces.
Most of the SEM images for this work were generated at MAIC on a JEOL 6400
instrument. This machine can usually obtain excellent images up to 25000X for most
semiconductor samples studied. Figure 4-7 shows a generalized schematic illustrating
the principle of operation for SEM. An electron beam (10 to 20 KeV) is rastered across
the sample surface using electronically controlled deflection coils. The electron beam
rastering is synchronized with a cathode ray tube so the x and y positions for the electron
beam and the CRT are in phase. A secondary electron detector generates a signal that
modulates the brightness of the CRT. The intensity of secondary electrons detected will
be a function of the sample topography, as illustrated in figure 4-8. Secondary electrons
will be deflected toward or away from the detector depending on the slope encountered
by the incident electron beam.
r r 1 B Detector giving
Secondary Cathode ray
S electron tube
ir .,:- Screen
Figure 4-7. Schematic illustration of SEM operation. An electron beam is
rastered across the sample in phase with the CRT signal. A secondary electron
detector modulates the signal, generating the image on the screen.
Figure 4-8. Schematic of contrast formation from surface topography. Secondary electrons
scattered toward the detector contribute to bright areas on the CRT image.
A sample (as small as can be easily handled, usually 1 cm2) is mounted to an
aluminum stub with carbon paint. Carbon paint is required to form electrical contact
between the sample and the stub. If the sample is not grounded in this manner, charge
will accumulate and degrade image formation. If the sample is insulating it must be
coated with a thin film of conductive metal, usually gold/palladium, to avoid charging.
Most semiconductor samples examined were conductive enough to bypass this
requirement. The properly prepared sample is then loaded into the evacuated SEM
column. The internal componentry including electron source, detector, and sample must
be housed in an evacuated column to avoid gas phase interactions with the electron beam.
Usually a plan view image is easily acquired in less than 30 minutes. It is
sometimes possible to image a cross-section of the sample when the film-substrate
interface is oriented normal to the electron beam. When done properly, film thickness
can then be accurately evaluated. Having an accurate image of an epilayer surface is the
cornerstone of sample analysis and SEM is a relatively operator insensitive and reliable
method of obtaining such information.
Atomic force microscopy is a method of acquiring a three-dimensional image of a
surface with up to atomic resolution. The primary advantages this technique has over
SEM is that it can quantify a root-mean-square (rms) surface roughness and does not
require electrical grounding. The technique relies on piezo-driven positioning equipment
that can raster an atomically sharp probe tip across a surface as illustrated in figure 4-9.
When the probe tip is scanned along a surface, it will deflect the cantilever as it traces
over surface roughness. This will cause variations in reflected beam intensity reaching
the quad-photodetector array. Computer software interfaced with the piezo-positioners
and photodetectors is able to use this information to develop a three dimensional image of
The software can also analyze the data and derive numerical descriptions of the
surface such as RMS roughness. This is calculated as
RMS = Z Z
where Zi is the current height, and Zavg is the average of N values.
The MAIC facility contains a Digital Instruments Nanoscope III AFM. With this
device, it is possible to select a range of scan sizes, usually on the order of 10 x 10
microns. This area is usually sufficient for extracting the most useful information from a
sample. Drawbacks of this technique include a sensitivity to operator supplied hardware
settings. Large variations in RMS value can be obtained from small changes in the
settings used to take data. Additionally, probe tips are extremely fragile and damage to
them can yield erroneous images and RMS values. The instrument seems to give
differing results for the same specimen on different days or when operated by different
users. It is therefore a good idea to support the data derived from AFM with SEM images.
This ensures that the AFM is functioning properly and providing useful data. One should
also take care in how much meaning is placed on RMS values. For example, vastly
Si cantilever tip
XYZ piezo-positioned stage
Figure 4-9. Sample and hardware arrangement for AFM
morphologies seen in the SEM might yield very similar RMS roughness values. Also, if
surface roughness can be experimentally controlled only to a very crude degree, i. e.
rough, medium, or smooth, it makes little sense to report RMS roughness values to a
tenth of a nanometer.
One of the easiest and most reliable ways of obtaining sample growth rates is with
stylus profilometry. The technique involves measuring film thickness at a step edge using
a scanning stylus. A step can be formed in two ways: by masking a small region of the
substrate before growth is initiated, or using a selective etch to remove a small area of
film down to the substrate.
Substrate masking is done by indium bonding a tiny (- 2 x 2 mm) scrap piece of
Si wafer to the substrate prior to placing it into the load-lock. The key is to use enough
indium to bond the Si chip while leaving some surface under the chip uncoated. This
allows the chip to act as a shadow mask, leaving a sharp edge that can be traced by the
profilometer. Additional care must be taken to ensure the chip is flush against the
substrate so growth will not occur beneath it. When growth is completed, the chip is
removed by gently heating the substrate to melt the indium. Disadvantages of this
approach are that some substrate surface area is sacrificed and the Si chip may de-bond
When using the selective etch method, a small piece (< 0.25 cm2) of the sample is
cleaved out and masked with an etch resistant polymer such as black wax. For nitrides, a
plasma etch is usually necessary to remove the film without affecting the substrate. After
etching, the wax is removed with a solvent, revealing a sharp step. This method is more
time consuming than the Si chip method, but tends to be more reliable.
After a suitable step has been formed, the sample can be positioned on the
profilometer stage for measurement. The hardware is computer controlled and a film
thickness can be determined in just a few minutes.
4.4 Compositional Analysis
It is often important to have an accurate assessment of the quantities and types of
elements comprising a sample. When constructing ternary compounds, i. e. InxAll-xN,
one selects growth conditions that should yield a film with a prescribed value of x.
However, precise compositional control can be difficult, especially in the early stages of
investigation, and sample analysis must be conducted to evaluate the relative amounts of
its constituents. Moreover, the concentration and types of electrically active impurities
must be discerned if a full understanding of the electrical properties of a sample is sought.
The two types of information, composition and impurity concentration, may be derived
by several means including x-ray photoelectron spectroscopy (XPS), energy dispersive
spectroscopy (EDS), wavelength dispersive spectroscopy (WDS), and electron probe
microanalysis (EPMA). This section will discuss the two methods most widely
employed for this work: Auger electron spectroscopy (AES) and secondary ion mass
Auger electron spectroscopy relies on an electron beam probe (3-5 KeV) of a
sample that generates Auger electrons possessing characteristic energies which are
collected by an energy analyzer. One example of Auger electron formation is illustrated
in figure 4-10. The Auger electron energy will be a function of the binding energy of the
intermediate electrons involved and the work function, 4, of the atom. Since 4 is usually
much less than the electron binding energy, it can safely be neglected. For a Si atom, as
an example, the K shell electrons have a binding energy of ~ 1.8 KeV. A vacancy here
will typically be filled by the decay of an L shell electron, binding energy 0.1 KeV.
Sometimes, the remaining 1.7 KeV of energy will be relieved by the emission of an X-
ray. Usually, the energy will be relieved by the ejection of another L shell electron,
whose kinetic energy will be 1.6 KeV after subtracting its binding energy. This process
is known as the KLL Auger transition.
0 ( Electron from probe
beam ejects core level
SEL3 Electron relaxation from
SEL2 less tightly bound level
g Auger electron emission
S with characteristic
2 \ energy:
i EKLL=EK- ELI-EL3 -
Figure 4-10. Energy diagram showing process leading to Auger electron emission.
A plot of number of electrons detected versus electron energy can be generated.
Peak positions and intensities can be used in conjunction with sensitivity factors to obtain
quantitative information about composition. The approach is useful for determining the
gross composition of a sample to within 5 atomic %. The method samples the first few
atomic layers of a film (see figure 4-11) and can be used to detect all elements except H
and He. Figure 4-11 is a useful guide for judging the volume of a sample probed, and
hence resolution attainable, by different electron beam techniques.
The instrument at MAIC used for this work is a Perkin-Elmer PHI 660 Scanning
Auger Multiprobe, operated by Mr. Eric Lambers. This device is equipped with
secondary electron imaging capabilities and has a lateral resolution of down to 0.5
microns. In addition, an argon ion gun is attached and can be used to sputter samples to
obtain composition as a function of depth. This feature is especially useful when
analyzing multi-layer structures. When the sputter rate of a sample is known, a plot of
composition versus depth can be constructed. Minimum sample size is roughly 3 mm in
diameter. No sample prep is necessary but the technique is destructive if sputtering is
used. The primary disadvantage of AES is that it can be difficult to analyze samples
whose constituents emit Auger electrons with similar energies. This is the case for InN.
Here, the In MNN transition is at 410 eV and the nitrogen KLL transition is at 395
Secondary. .. .:. .....+
electrons (ShM;! s.
e t n :.. .. ..el ectrons :"
EPMA. W.DS,.......- C
Continuum x-.rys ... .
'" r'*. S s *"**.* ** :
o; ..-" ..'., '" '.-.': t .." .. ., r, ,
EPM A, DS,.i:.:i.+i. i ::/: w,':+ : :". :: <: .." ,xv. 'P:
~~~~~~~~~ "" ... %' ;: '
Up to one
Figure 4-11. Cross-sectional view of interaction volumes for various electron
beam probe techniques. The width of the interaction volume represents the
resolution of the technique
Secondary ion mass spectroscopy is an extremely sensitive technique used to
quantitatively measure concentrations as low as 1012 atoms/cm3, making it ideal for
analyzing dopant and impurity levels. The principle of operation is illustrated in figure 4-
12. A primary beam of ions, usually oxygen or cesium, with energies from 1 to 20 KeV
is impacted on a sample. The transfer of momentum will eject neutral and charged
species. Ejected ions can be analyzed in a mass spectrometer where they are separated
according to their charge to mass ratio and counted.
SIMS can be used to measure any element from H to U but requires standards to
obtain quantitative information. A plot of concentration versus depth can be derived
showing matrix and impurity element levels for any structure. Samples as small as a few
mm in diameter can be examined. All SIMS analysis done for the Abernathy group was
performed by R.Wilson.
ion beam Ejected ions to mass
Figure 4-12. Diagram of SIMS sputtering and analysis process.
InN EPITAXIAL GROWTH
The primary candidate for an ISL to large-gap nitrides is InN. The relatively small
bandgap and potential for grading from other III-N's make it an excellent choice. Figure
5-1 shows an approximate band diagram for an ISL using n-InN. The relative band
offsets used to construct the diagram are adapted from Wang et al.41 In addition, the
further development of red light emitting devices from the III-nitrides will require the use
of InN or very In-rich InGaN active layers. Consequently, understanding the electrical
and optical behavior of InN is an important task. A survey of past work on InN thin film
growth will be presented. The current chapter examines past methods used to deposit
epitaxial InN films and introduces the specific details for InN sample development in the
Abernathy lab. Most work focused on optimizing the InN crystal quality and developing
an understanding of the origin for the observed n-type behavior of as-grown films.
5.1 History of InN Growth
The development of InN is not as mature as GaN in part due to the difficulty of
preparing In containing films (owing to indium's low sticking coefficient and the high
nitrogen vapor pressure above InN). This section will review the status of InN, focusing
particularly on reported mobilities and carrier concentrations. All of the InN films
reported in the literature consist of wurtzitic oriented crystallites unless otherwise noted.
Figure 5-1. Schematic band diagram of ISL using InN. The bottom
diagram shows the effect of using a continuously graded region going from
GaN to InN.
Indium nitride films were first investigated by Hovel and Cuomo in 1972.42 They
used RF sputtering to obtain polycrystalline films having mobilities on the order of 200
cm2NVs with electron concentrations of 7 x 1018 cm-3. Tansley and Foley43 reported RF
sputtered films with mobilities of >1000 cm2/Vs at 150 K having an electron
concentration as low as 3 x 1016 cm-3. Sullivan et al. used a magnetron sputtering system
to deposit poly-InN on glass or fused quartz.44 They obtained films with 3 to 6 x 1020
electrons/cm3 having mobilities of 6 to 10 cm2/Vs. MOCVD was used by Matsuoka et
al. to deposit nominally single crystal films, as determined by RHEED and XRD,
epitaxially on sapphire substrates.45 Mobilities of 300 to 400 cm2/'s were achieved with
carrier concentrations of 1018 cm3. Kistenmacher et al. studied the effect of film growth
and AIN buffer layer growth temperature on electrical properties of reactively sputtered
InN films on sapphire.46 Samples had -2 x 1020 cm-3 carriers with a maximum mobility
of 60 cm2/Vs, largely independent of buffer and film growth temperatures. Strite et al.
reported the observation of zincblende InN grown on GaAs with MBE.47 These films
appeared to be highly defective single crystal in XTEM micrographs. Mobilities of 220
cm2NVs with carrier concentrations of 1020 cm3 were measured. A modified MBE
system was used to deposit films by Sato and Sato.48 Samples had mobilities of 35
cm2NVs with electron concentrations of 3.2 x 1020 cm-3. Figure 5-2 summarizes the work
done to date reported on InN film growth.
One feature common to all InN films is their n-type autodoping. All reported
samples, regardless of preparation technique, have very high background electron
concentrations as grown. If p-type ohmic contacts to large gap nitrides using an InN ISL
are to be fabricated, electron concentration reduction to levels that can readily be
compensated by Mg is essential. To illuminate the origin of n-type autodoping, a series
of experiments were conducted to determine if the predominant contributor was point
defects, structural defects, or impurities. The variables examined were substrate
temperature, nitrogen plasma condition, substrate type and pregrowth treatment, and
O Hovel and Cuomo
* Tansley and Foley (43)
* Sullivan et. al44)
A Kistenmacher et. al.(46)
Matsuoka (MOCVD) (45)
Strite (MBE) (47)
Sato (-MBE) (48)
Figure 5-2. Summary of InN film properties reported in literature.
Carrier Concentration (cm 3)
. - -I A o i
5.2 Role of Nitrogen Plasma and Substrate Temperature
The effects of varying nitrogen plasma condition and substrate temperature on the
resulting electrical and structural properties of epitaxial InN films have been
investigated.49 Two nitrogen plasma sources were utilized, ECR and RF. Samples were
grown on epi-ready (100) GaAs substrates.
5.2.1 Sample Loading
Two-inch GaAs wafers were cleaved into quarters and mounted on two-inch Si
backing wafers using indium. The Si wafer was heated on a standard hot-plate to melt a
small amount of indium. The GaAs piece was carefully placed over the melted indium to
avoid thermal shock. Using two sets of stainless steal tweezers, the '4 wafer was
manipulated over the Si until complete indium wetting of both surfaces occurred. Care
must be taken to avoid contact with the top surface of the GaAs. Any excess molten
indium is scraped away with a razor blade. When growth rate information is needed, a
small chip (< 1 cm2) of scrap Si is indium mounted for post-growth profilometry. Once
the substrate has been indium mounted, the backing wafer is loaded onto a moly block.
The block is specially designed to interface with a trolley that transports blocks between
the load-lock and buffer chamber. The trolley is placed into the load-lock which is then
pumped down. Once the pressure has reached < 10-6 torr, the gate-valve separating the
load-lock from the buffer is opened and the trolley is moved to the buffer. When the
buffer chamber pressure is below 10-7 torr, blocks are transferred from the buffer to the
growth chamber using a magnetically driven transfer rod.
Once in the growth chamber, reactive group V species are introduced, in this case
nitrogen from the plasma source. The sample is then rotated from the transfer position to
the growth position. Substrate rotation is set to 15 rpm to ensure uniformity across the
sample surface. The above mentioned procedure is standard for all work done in the
Abernathy MOMBE lab.
5.2.2 Sample Growth
Since no effort is made to clean substrates ex situ it is necessary to prepare the
GaAs surface by exposing it to nitrogen plasma at -6250 C before growth is initiated.
This effectively removes the oxide that resides on the GaAs surface. Substrate
temperature is monitored with a thermocouple that is connected to a Eurotherm
temperature controller. Temperature set-point is maintained by the automatic adjustment
of current or voltage fed to the substrate heater. Differences in set-point and actual
substrate temperature can be determined by observing the melting point of InSb (m. p. =
5250 C) or GaSb (m. p. = 712 C) chips In mounted to the substrate. When establishing
such temperature off-sets, it is essential to maintain a group V flux to avoid
decomposition of the calibration chips.
A He carrier gas flow of 7.5 seem was used to transport trimethylindium (TMI) to
the reactor. The TMI bubbler was maintained at 11.4 C and 20 torr. The ECR plasma
source (Wavemat MPDR 610) operated at 2.45 GHz and 200 W forward power with
ultra-pure nitrogen flowing at 20 sccm. This source is fitted with a TiN coated Mo grid
connected to an independently controllable power supply. During growth, a floating bias
of -15V is reached, as measured by a portable multimeter. Errors in measurement of this
value may occur due to electromagnetic interference from the plasma source. The rf
plasma source (SVT Associates) was operated at 13.56 MHz and 400 W forward power
with 5 seem nitrogen flow.
Figure 5-3 shows SEMs of ECR derived InN grown at 525, 575 and 6250 C.
Surface roughness, as measured by AFM rms values, also increases with substrate
temperature, as plotted in figure 5-4. Caution must be exercised when examining rms
roughness values, however. The surface morphologies of samples grown at 525 and 5750
C are quite different as seen in the SEMs, but their rms roughness values are very similar
(18.1 vs 20.4 nm). Conversely, morphologies of material grown at 575 and 6250 C
appear similar while having a large difference in rms values (20.4 vs 67.2 nm).
Crystallinity, as measured by the hexagonal (0002) FWHM X-ray peak, increases
with substrate temperature, depicted in figure 5-4. The maximum growth temperature is
limited, however, by the desorption of indium from the surface. In fact, at 6750 C,
virtually no growth takes place. A marked increase in electron mobility was observed
when the substrate temperature increased from 525 to 5750 C, diminishing slightly when
increased to 6250 C, see figure 5-5. This improvement in mobility is most likely due to
the improved crystallinity obtained at higher growth temperatures. The electron
concentration varied by less than a factor of two for this temperature range, going from
3.6 x 1020 at 5250 C to a minimum of 1.6 x 1020 at 5750 C. It is clear that while the
crystallinity, surface roughness and mobility of InN grown by ECR plasma are functions
of substrate temperature, the electron concentration is largely invariant.
As with growth temperature, plasma biasing strongly affected the structural
quality of InN films. X-ray scans show a decrease in FWHM of the hexagonal (002)
peak as the ECR grid was biased more negatively (figure 5-6). RMS surface roughness
Figure 5-3. Scanning electron micrographs of InN/GaAs derived from
ECR nitrogen plasma at different substrate temperatures.
520 540 560 580
Figure 5-4. RMS roughness and hexagonal InN
temperature for ECR derived
(002) FWHM vs growth
70- D A -
60- 20 c
E 3.0x10 .
o -0--n n
-A- Mobility a
0 20 0
2 40- 2.0x10 0
500 550 600 650
Growth Temp (C)
Figure 5-5. Mobility and electron concentration vs growth
temperature for ECR derived InN.
40 ii i- 1050
E 35- --fwhm 1000
15 I I 850
-40 -30 -20 -10 0 10
Grid Bias (V)
Figure 5-6. RMS surface roughness and hexagonal InN (002) peak FWHM vs grid
bias for ECR derived InN.
values do not appear to follow any trend but are of the same order measured for the
substrate temperature study. Surface morphologies seen in SEM micrographs are similar
for InN grown at 40 V bias (5250 C) and no bias at 6250 C (figure 5-7). These samples
also share similar crystal quality, having comparable hexagonal (002) FWHM values.
The surface morphology of InN grown at +10 V bias is similar to that grown with
a floating bias although there is a reduced density and increase in size of the nodular
features. The FWHM of the hexagonal (002) peak is nearly identical for the two films.
Electron concentration does not vary significantly with applied grid bias, although the
mobility increases by roughly a factor of two for the sample grown under the highest
negative bias, as shown in figure 5-8. Carrier concentrations are similar to those found in
InN grown at various temperatures with no applied bias.
The fact that the smoothest surfaces, narrowest FWHM hexagonal (002) peaks
and highest mobilities are obtained from the ECR plasma biased at 40 V suggests that
increasing neutral N density is beneficial to growth. One way to further examine the role
of ions is to compare films derived from ECR plasma with those from rf plasma. The rf
source produces higher energy species than the ECR source. RMS roughness values for
rf derived InN are similar to those found for samples grown at 6250 C using the ECR
source. However, the morphology of the two samples is quite different as shown by the
SEM micrograph of figure 5-9. The InN grown using rf plasma appears to have
hexagonal grains in contrast to the pyramidal grains of ECR derived films. X-ray scans
of rf derived material exhibit indications of the zinc-blende phase in addition to the
wurtzitic phase prevalent in ECR derived samples. GaN films grown by various methods
+ 10 V
(top), floating (middle), and + 10 V (bottom).
600 i 1 I I 1 1 6.0x1 0m
-0- ECR n
-*- RFn --- n
500 -- ECR Ibbility 5.0x10m
---- RF Mbility
100 2.0x10m 0
-40 -30 -20 -10 0 10
Applied Grid Bias (V)
Figure 5-8. Mobility and electron concentration vs applied grid bias for ECR derived InN
grown at 5250 C. Included are mobility and electron concentration values for rf derived InN
grown at 5250 C.
Figure 5-9. SEM micrograph at 10kX of rf derived InN grown on GaAs at 5250 C.
on several types of substrates have resulted in mixed phases as well.50-54 In fact, the rf
derived InN did not yield a strong hexagonal (002) peak so such a comparison cannot be
made with ECR derived InN. RF derived samples have electron concentrations on the
same order as ECR produced samples (figure 5-8) but have significantly higher
mobilities, > 500 cm2/Vs. This may indicate that cubic InN has a higher mobility than
5.3 Substrate Effects
Auger electron spectroscopy (AES) was used to examine the nitridation behavior of
GaAs, sapphire and lithium aluminate (LAO) substrates exposed to an RF nitrogen
Four substrates were compared for this study, epi-ready (100) GaAs, c-plane
sapphire, LAO (100), and LAO (100) miscut 180. The miscut LAO is believed to posses
matching symmetry as well as lattice dimension for nitride epitaxy. Lithium aluminate
has an improved lattice match to III-nitrides with LAO (a= 3.13A) having a near perfect
match to AIN (a=3.11 A).56 This is an important consideration when growing low
temperature AIN buffers. The substrates were indium mounted to a Mo block without
any ex-situ cleaning steps and loaded into the MOMBE system. In order to investigate
their nitridation behavior, the substrates were exposed to the RF nitrogen plasma prior to
growth initiation. The plasma source was operated at 400 W forward power with 5 sccm
N2 flow. Substrates were exposed for five minutes at temperatures of 7250, 7750, 8250
and 8750C. Samples were exposed to the plasma during heat-up and cool-down as well.
These samples were then removed from the system and analyzed using depth-profile
auger electron spectroscopy (AES) and atomic force microscopy.
InN films were grown on the various substrates using trimethylindium (TMI)
transported by a He carrier gas. Reactive nitrogen was provided by the RF plasma source
operated under the conditions mentioned above. Substrates were maintained at -5250 C
giving a growth rate of -250 nm/hr. Three cases were examined for this study: 1) InN
films grown on as received substrates with no prior plasma treatment, 2) InN films grown
on substrates that received a plasma exposure of 8750 C for five minutes and 3) InN films
grown on substrates that received a plasma exposure of 8750 C for five minutes and an
AIN buffer. The 30 nm AIN buffers were grown at 4250 C using dimethylethylamine
alane (DMEAA) and RF nitrogen plasma. The DMEAA bubbler is maintained at 9.20 C
and 7 torr. A carrier gas of He is used at 10 sccm. Nitrogen plasma is operated at 400 W
forward power and 5 seem nitrogen flow.
Electrical transport properties were obtained from Van der Pauw geometry Hall
measurements at 300K using alloyed (400 C, 2 min) HgIn ohmic contacts. Surface
morphology was examined by SEM and contact mode AFM.
SEM micrographs of InN grown on sapphire and GaAs are shown in figure 5-10.
The top images show InN films grown on as received substrates while the bottom images
are of films deposited on RF nitrogen plasma exposed substrates. The surface
morphologies of the "no anneal" films are quite rough. The InN surface morphology
shows significant improvement in the case of the GaAs substrate that received plasma
exposure prior to growth. However, for InN grown on plasma exposed sapphire there is
no morphological improvement. The differences in morphology between the two
substrates can be traced to differences in nitridation behavior as measured by AES. GaAs
samples that received a nitrogen plasma exposure at 8750 C for five minutes yield a
Figure 5-10. SEM micrographs of InN films grown on sapphire (left) and GaAs (right).
The top samples were grown on as-received substrates while the bottom samples were
grown on nitrogen plasma exposed substrates.
strong nitrogen signal at 385 eV, indicating nitridation of the surface. No such signal is
discernible for sapphire substrates given the same treatment. This would suggest that no
nitride template is being formed on sapphire under the given nitrogen plasma exposure,
resulting in no improvement in morphology. However, work done by Yamamoto et al.
indicates that nitridation, as measured by electron spectroscopy for chemical analysis, of
sapphire can occur under an NH3 flow over a heated (-8000 C) substrate.57 They report a
marked improvement in epitaxial InN films grown on nitridated sapphire as measured by
in-situ reflection high-energy electron diffraction. Although no nitrogen signal was
discernible by AES for sapphire exposed to rf nitrogen plasma, it is evident that plasma
exposure prior to growth is important to the final film morphology when an AIN buffer is
used. Figure 5-11 shows InN grown over AIN buffers that were deposited on plasma
exposed substrates at different temperatures. The importance of substrate temperature
during plasma exposure is illustrated by the marked improvement in morphology for
samples grown on substrates receiving an 8750 C pretreatment.
Next examined were InN films grown on LAO (100). Figure 5-12 shows SEMs
of films grown using the three conditions discussed above. The film grown on the as
received substrate is seen to be discontinuous and has poor morphology. Notable
morphological improvement is seen when the substrate is exposed to nitrogen plasma
with even further improvement obtained from the use of an AIN buffer. Figure 5-13
shows SEMs of InN films grown on LAO (100) miscut 180. Again, the as received
substrates yield discontinuous, poorly formed films. Improvements in film continuity are
gained by nitrogen plasma exposure prior to growth with best results obtained using an