Electron stimulated oxidation and rhenium electrical contacts on 6H-SiC

MISSING IMAGE

Material Information

Title:
Electron stimulated oxidation and rhenium electrical contacts on 6H-SiC
Physical Description:
viii, 123 leaves : ill. ; 29 cm.
Language:
English
Creator:
McDaniel, Gavin, 1973-
Publication Date:

Subjects

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

Notes

Thesis:
Thesis (Ph.D.)--University of Florida, 2002.
Bibliography:
Includes bibliographical references (leaves 115-121).
General Note:
Printout.
General Note:
Vita.
Statement of Responsibility:
by Gavin Y. McDaniel.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 029221803
oclc - 51020756
System ID:
AA00022868:00001

Full Text










ELECTRON STIMULATED OXIDATION AND RHENIUM ELECTRICAL
CONTACTS ON 6H-SIC















By

GAVIN Y. McDANIEL


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


2002














ACKNOWLEDGMENTS

I would like to express my deepest gratitude to Dr. Paul H. Holloway and Dr.

William V. Lampert for all of their patience and guidance during the last 5 years. Both

worked with me long before I arrived at the University of Florida and only through their

vision was I able to pursue a Ph.D. while remaining on active duty with the Air Force.

As my co-advisors, they gave me countless hours of instruction and nudged me along the

way when I was stuck. They kept a careful guard over my time constraints and always

allowed me to "stub my toes" in order to learn but were always willing to give helpful

advice so I could pick up the pieces and drive forward. I would also like to thank the

members of my committee, Dr. Gijs Bosman, Dr. Rolf Hummel, Dr. Kevin Jones, and

Dr. Wolfgang Sigmund. Each member of my committee has been very patient with my

completion in absentia from the university.

I would like to express my deepest appreciation for the patience and advice given

by Steven Fenstermaker. He introduced me to the operation of each major piece of

equipment used in this study. He was always willing to stay late and come in early when

the equipment didn't want to work "as planned." He continued to support this research

after I departed Wright Patterson Air Force Base by answering questions and providing

information vital to completion. He was an invaluable asset, without which, I would not

be getting this degree.

I would also like to thank Jim Soloman for always being there with AES and

XRD advice. He provided useful information on data collection for the designed set of

ii








experiments. When Steve and I were stuck, he could be counted on for advice and

direction with equipment and theory.

Without Don Thomas, all of my computer troubles would still be computer

troubles! He was a lifesaver, especially with the XRD. Don's help with countless hours

devoted to fixing what seemed the unfixable is greatly appreciated.

My other coworkers, Dr. Kurt Eyink, David Tomich, and Dr. Matthew Seaford,

were all excellent sounding boards. They would listen to my procedures and keep me on

track. I would like to thank them for the guidance.

To all of my friends, I can't repay what you have given me along the way. Tyler,

Kim, Kristie, and Robert were with me from the beginning. I'm glad we've all made it to

the end together. Angela, Tim, Chuck, Rene, Doug, and Erika provided an outside voice

that helped me keep my sanity during the long hours at Wright Patterson. They always

supported me through thick and thin. I'll never forget how they have helped me get here.

Tim, Shawn, Blake, and Jen shared sympathetic ears while going through graduate

programs of their own at AFIT. We supported each other, never willing to let one

another fall short. I want to thank them for always being there, even if we didn't get the

chances to be there in person all the time. I would be remiss without thanking my dear

friend Brian. Without his constant reminders "are ya done yet," Ijust might not be. In

his own way, he has made the completion of this work possible. The NASA co-ops, in

particular Laura, have been an inspiration, which has fueled my dreams for the future.

They have touched my heart in a special way giving me the drive to complete this work

and the determination to tackle the world. I owe them more than my gratitude. I owe

Ludie many thanks. Without her, my coordination efforts from Dayton to Utah to Florida








to wherever I was with the University would never have worked. She kept me updated

with schedules and fought the local battles that I couldn't fight in person. Ludie is a gem

and now I can finally answer her question, "When are you gonna graduate?" And what

would I do without Mrs. Alva "Fran" Johnson and Mr. Mortenous "Marty" Johnson. I

promised I would finish and if there was anyone who kept the fire burning in me, it was

them! They are wonderful friends and I thank them for their undying support!

Finally, I'd like to thank my family, Wyatt, Sandy, and Gardner. They have been

my ultimate support structure through all aspects of my life including this degree. They

never fail to send an encouraging word, an uplifting e-mail, or a simple "I know where

you're coming from." It's the little things that you can't put your finger on that make

family the most important thing in your life. For this, I dedicate this dissertation each of

them. I would also like to dedicate this work to my grandfather who passed away during

its inception and my living grandparents who never fail to remind me of their pride and

their love. I'm sorry my Pa Pa couldn't be here to celebrate with me but I know he would

be proud!














TABLE OF CONTENTS

page

ACKNOWLEDGMENTS...........................................................................ii

A B ST R A C T ................................ ...... ....... ............. ...............................vii

CHAPTER

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

2 REVIEW OF LITERATURE................................................................... 8

SiC Growth ........................................................................................ 8
Surface Preparation ................................................ ............................. 11
C contact Physics ............................................. .. ...... ............... ............ 18
Contact Properties and Characterization..................................................... 27
Ohmic Contacts To n-Type 6H-SiC: Present Demonstrated Research................. 32
Impurity Concerns ....................................................................... 32
SiC and Fermi-Level Pinning.......................................................... 34
Metallic Reactions With n-Type 6H-SiC............................................. 36
Rhenium As a Contact.................................................................. 41

3 EXPERIMENTAL PROCEDURES ......................................................... 47

Electron Stimulated Oxidation................................................................ 47
Re Contact Deposition......................................................................... 48
Sample Preparation...................................................................... 48
Deposition ................................................................................. 53
General procedure.................................................................53
Specialized conditions............................................................ 55
C haracterization ........................................ .................................. 56
Dektak stylus profilometry...................................................... 56
X-Ray diffraction.................................................................. 57
Electrical measurements......................................................... 58
Auger electron spectroscopy..................................................... 59






v








4 RESULTS AND DISCUSSION............................................................... 60

Electron Stimulated Oxidation (ESO)....................................................... 60
R results ........................................................................... .... .... 60
D discussion ...................................... ......... ......... .............. .. ......... 67
Dependence on chamber pressure............................................... 68
Dependence on beam energy.................................................... 70
Dependence on beam current.................................................... 71
Re contacts ....................................................................................... 72
R results ................................................ .. .................................. 72
D iscussion................................... ........... .... ............................. 100

5
CONCLUSIONS.................................................................................. 108

Electron Stimulated Oxidation............................................................... 108
R henium C ontacts............................................................................. 109
Significance To Literature................................................................... 112

6 FUTURE WORK.............................................................................. 114

R EFER E N C E S ........................................... ......................................... 115

BIOGRAPHICAL SKETCH..................................................................... 122


























vi












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

ELECTRON STIMULATED OXIDATION AND RHENIUM ELECTRICAL
CONTACTS ON 6H-SIC

By

Gavin Y. McDaniel

August 2002

Chair: Dr. Paul H. Holloway
Major Department: Materials Science and Engineering

Electron stimulated oxidation (ESO) and rhenium (Re) metallizations on a-silicon

carbide (6H-SiC) were studied. ESO was quantified versus electron beam exposure, total

vacuum pressure, beam energy Ep (3 to 6 keV), and current Ip(25 to 500 nA), while

monitoring chemical change using Auger electron spectroscopy (AES). Vacuum

chamber pressures below 2.6 x 108 Torr required beam irradiation to induce oxidation.

A positive linear correlation between oxidation rate and total chamber pressure was

observed. Rate did not correlate with concentration of a particular ambient species (H20,

CO, or C02). Oxidation rate decreased with increased Ep suggesting surface secondary

electrons stimulate oxidation. Dependence of oxidation rate on Ip indicated current

limited dissociation below 200 nA.

Rhenium contacts (1000 angstroms thick) were deposited on carbon-rich,

stoichiometric, and silicon-rich 6H-SiC surfaces. Morphology (Dektak), texture (X-ray

diffraction (XRD)), chemistry (AES), and electrical properties (current/voltage (I-V))

vii







were characterized for as-deposited and annealed (120 minute, 1000 C, <1 x 10-6 Tonrr)

contacts.

As-deposited films were non-ohmic with total contact resistances of 59, 1620 and

110 ohms for carbon-rich, stoichiometric, and silicon-rich surfaces, respectively. Films

grown on carbon-rich surfaces were non-specular, granular, and often delaminated during

characterization. Island-growth was observed on stoichiometric surfaces with thickness

variations of-800 angstroms. Films remained specular for 3 hours, but then became

hazy from oxidation. Textured (101) growth was observed on silicon-rich surfaces.

Thickness varied by -250 angstroms and films resisted ex-situ oxidation for more than 24

hours.

Annealed samples remained specular without visual signs of oxidation. Films

were smoothed with thickness variance less than 100 angstroms. Phase separation was

observed based on formation of interfacial Re clusters and -100 angstrom graphite

surface layers. Auger confirmed as-deposited Si layers (50 to 100 angstroms) were

consumed by reaction during annealing and Re/Si and Si/SiC interfaces were diffused

~500 angstroms more compared to as-deposited interfaces. Annealed contacts were

largely ohmic with averaged total contact resistances reduced to 11, 3.2, and 1.4 Q for

carbon-rich, stoichiometric, and silicon-rich samples, respectively. Average specific

contact resistances of 7.0 x 10-5 Qcm2 for stoichiometric and 1.6 x 10-5 f2cm2 for

silicon-rich samples were observed.















CHAPTER 1
INTRODUCTION

SiC is a member of the IV-IV family of indirect wideband gap semiconductors.

While SiC shares similar hexagonal wurtzite (x) and zincblende cubic (13) crystalline

structures to the III-V Nitride family, this material exhibits a phenomenon known as

polytypism which allows for more than 200 mixed forms of these basic structures.1 The

range of band gaps (2.2 eV to 3.3 eV) available from different polytype structures gives

SiC a materials advantage over other conventional Ill-V, II-V and mineral

semiconductors particularly in applications where high-power, high-temperature,

high-frequency, radiation and chemical resistance may be required.23" Table 1

provides a short list of possible applications where SiC devices would provide a

considerable improvement for operations compared to current materials.

Table 1. Current and projected operating temperatures, lifetime and reliability of
electronics components for various applications.9
Application field Current operating Future operating Requested reliability
temperature (C) temperature (C) (Hours)
Automotive 125 to 140 165 to 250 10,000
Aircraft 125 200 10,000
Spacecraft 300 500 1,000 to 30,000
Oil Logging 175 175 10,000 to 30,000
Geothermal 200 250 to 260 10,000 to 30,000
Power Electronics 125 250 to 500 10,000 to 30,000

Table I shows that space applications and power electronics require the greatest

improvement in high-temperature operations, justifying the need for improved studies on

SiC in this environment. From a technology thrust area at the Army Research







2

Laboratory, Katulka et al. 10 reported that "high-power and high-temperature materials are

of great interest to the electric combat systems community. The technical interest in SiC

is mainly due to its ability to operate at greatly elevated power and temperature (>300 C)

levels compared to its Si-based counterpart. Advanced high-power electric weapon

concepts have the ability to improve ballistic performance by as much as 30% over

conventional guns with respect to projectile muzzle kinetic energy in

electrothermal-chemical guns; or up to as much as 7 km/s muzzle velocity with

electro-magnetic rail guns"."'"2 Space-based radar, digital engine control, Airborne

Warning and Control Systems (AWACS) (airborne high-power radars), and additional

avionics applications are a few military areas in which elevated temperature and

radiation-resistant performance could improve with SiC development.

SiC technology is not limited to military application. Common commercial

devices such as diodes, FETs, bipolar transistors, dinistors, short wave LEDs, and UV

Table 2. High-temperature operation of silicon carbide based devices and circuits.9
Device Material Working T Remarks
JFET 6H-SiC 873 (K) Siemens AG, Germany, CVD
UV photo detector 6H-SiC 773 (K) AMETEK
JFET 6H-SiC 573 (K) US Amy Research Lab., USA, amplifier
JFET 6H-SiC 873 (K) NASA, USA, 30 h operation in the air
MOSFET 6H-SiC 673 (K) CEA-LETI, France
p-n-diodes 4H-SiC 623 (K) Cree Research, USA
Vertical 4H-SiC 473 (K) Cree Research, USA
Thyristor 6H-SiC 773 (K) Cree Research, USA
Op amp IC 6H-SiC 573 (K) GE Corp. Res. Center, USA
Schottky diodes 6H-SiC 673 (K) Kyoto University, Japan,

photodiodes have already been demonstrated (Table 2) using SiC. Applications in power

microwave devices and integrated circuits, however, remain limited and both industry






3

and the military would benefit from development efforts to improve the properties of SiC

that are discussed below.13

As previously mentioned, SiC has nearly 200 polytype structures which give rise

to variability in physical, chemical, thermal, and electrical properties. The two main

crystal structures, cubic 13-SiC and hexagonal a-SiC are the crystal basis for all

polytypes. Bulk material is comprised from a periodic sequence of Si-C, bilayer,

stacking sequences. SiC crystals are more commonly referenced by stacking sequence

where the number of atomic layers required to repeat structure in the lattice and the

overall crystal symmetry is defined. a-SiC, for example, is denoted 2H-SiC because two

layers (AB) repeat within the bulk resulting in a hexagonal close-packed symmetry, while

3-SiC is denoted 3C-SiC because three layers (ABC) are required for repetition to form a

zinc-blende cubic structure. Mixtures of the two pure structures form the most

commonly studied SiC crystals, 2H, 4H, 6H, and 3C. 4H is 1/2 cubic and 1/2 hexagonal

in nearest neighbor bonding, while 6H is 2/3 cubic and 1/3 hexagonal.214

3.4
3.2
> 3
ta 2.8
-1 2.8 4H-SiC
2.6 6H-SiC
S2.4 Pure Pure
2.2 Cubic Hexagonal
2 -------
3C-SiC 2H-SiC
0 0.2 0.4 0.6 0.8 1
Fraction of Hexagonal Bonding
Figure 1. SiC band gap (eV) as a function of the fraction of hexagonal nearest neighbor
bonding.







4

Variable stacking sequences between the Si-C bilayers result in different intrinsic

properties. The band gap, for example, changes nonlinearly from 2.2 to 3.3 eV as a

function of hexagonal nearest neighbor bonding.2,'14 In addition to band gap variance,

other properties such as melting point (Tm), Debye temperature (TD), working

temperature (Tw), thermal conductivity (kt), dielectric constant (E), electron and hole

mobility (p, Rh), saturated electron drift velocity (%s), and critical voltage field (Ecr)

change notably from one polytype to another.2'7'9,1314 Table 3 shows that SiC's

properties make it a good candidate for applications where other semiconductors are

likely to fail. Si and GaAs, for example, are limited to operations near room temperature

with lower power. High electron mobility allows SiC devices to operate at high speeds

and its wide band gap and large breakdown voltage allow operation at larger biases and

greater power (higher temperature) than Si and GaAs. A melting/sublimation point of

Table 3. Electrophysical parameters of semiconductor materials.0
Material Eg Tm TD Tw k, e ute h Ecr Vs
(eV) (K) (K) (K) (W/(cm-K)) (cm2/(V-s)) (cm2/(V-s)) (10 V/cm)(107cm/s)
Ge 0.7 1210 374 280 0.60 16 3800 1800 1.0 0.6
Si 1.1 1690 645 410 1.35 12 1350 480 2.0 1.0
GaAs 1.4 1510 400 570 0.45 12 6000 320 2.6 1.0
GaAs, 1.6 1550 270 650 0.20 11 4000 300 1.0 1.0
-xPx
BP 2.0 2300 1140 720 2.50 11 1500 300 4.0 2.0
GaP 2.2 1740 520 800 0.70 11 250 150 4.5 1.5
CdS 2.3 2020 290 840 0.20 11 350 15 10.0 1.0
3C-SiC 2.3 3100 1430 840 3-5.00 9.8 750 40 7.0 2.5
6H-SiC 3.0 3100 1200 1200 5-7.00 9.8 370 90 21.0 2.0
4H-SiC 3.2 3100 1200 5-7.00 9.8 800 115 20.0 2.0
GaN 3.3 1800 710 1250 1.00 9.5 400 14.0 2.0
ZnS 3.6 2200 340 1300 0.03 10 140 5
C 5.6 3300 2230 2100 14.00 5.71800 1500 7.5 3.0
AIN 6.0 3100 1100 2100 2.00 9.1 1.5
BN 10.0 3300 1700 3300 3.00 7.1 1.5







5

>3100 K suggests that devices using this material should be easily capable of operating

over the range of 600-1000 'C.5'1,014 To further illustrate the advantage of SiC vs. other

semiconductors, Porter et al.14 compiled data on selected figures of merit (FM) for a

variety of materials as compared to Si. Johnson's FM (JFM) is based on high-frequency

and high-power operation in discrete devices. Keyes' FM (KFM) compares high-speed

operation in ICs. Baliga's FM (BFM) considers the need to minimize conduction losses

in FETs. Finally, Baliga's high-frequency FM (BHFFM) looks at high-frequency

operations but extends the analysis to include switching losses.14 Table 4 shows the

results from these FMs.

While SiC has several material advantages over conventional semiconductors for

harsh environment application, processing and development of the semiconductor is not

without challenge. Efforts to advance processing technology for SiC are focused on but

not limited to growth of ultra-pure bulk and epitaxial single crystal polytypes,2 high

concentration p-type doping which remains stable at elevated temperature

Table 4. Figures of merit (FM) for various semiconductors as compared to Si. Johnson's
(JFM), Keyes' (KFM), Baliga's (BFM), and Baliga's high-frequency (BHFFM).'4
Material JFM KFM BFM BHFFM
Si 1.0 1.0 1.0 1.0
GaAs 7 0.456 13 10
InP 16 0.608 10 7
GaN 282 1.76 910 100
6H-SiC 695 5.12 106 13
3C-SiC 1137 5.8 40 12
Diamond (C) 8206 32.2 8574 454

conditions,214'15 unintentional doping/contamination during processing,2"14'15 and

formation of stable and durable metallizations which operate in the harsh environments

previously described. This dissertation focuses on control of contamination during









real time in-situ metal deposition analysis as well as formation and characterization of a

refractory high-temperature ohmic metallization on 6H-SiC.

Achieving stable/reliable, low specific contact resistance, ohmic contacts to each

polytype of SiC has been a continual goal for researchers. Specific contact resistances as

low as 1 x 10-6 &cm2 have been reported for contacts at room temperature on an n+

6H-SiC layer with a carrier concentration of 4.5x 1020 cm-3 after thermal annealing for 5

min at 1000 C.16 While this specific contact resistance is acceptable for most device

requirements, several challenges remain, particularly with elevated temperature

operation. A few obstacles include but are not limited to: contact oxidation, melting,

peeling, metal inter-diffusion, formation of silicides, silicates and carbides, and

generation of interfacial defects due to lattice mismatch and surface reconstruction.15

In this work, it is shown that 6H-SiC surfaces are susceptible to electron

stimulated adsorption (ESA) and electron stimulated oxidation (ESO) as artifacts of

in-situ electron beam process control during surface analysis and metallizations.

Conditions that contribute to ESO and ESA are quantified for control. Rhenium ohmic

contacts are demonstrated for the first time on 6H-SiC with varied surface chemistry:

carbon-rich, stoichiometric, or silicon-rich. It is shown that reliable, low specific contact

resistance, ohmic contacts are generated from a single layer of rhenium on silicon-rich

6H-SiC surfaces after heat treatment at 1000 C for 120 minutes.

Thus the scope of this dissertation is as follows. Chapter 2 provides a review of

the "physics" for the development and present state of knowledge for ohmic contacts to

n-type SiC. Chapter 3 reviews the experimental procedures including in-situ analysis,

deposition source material preparation, deposition techniques, and characterization







7

methods used for this work. Chapter 4 presents environments and conditions, which

contribute to surface contamination and oxidation during in-situ analysis of 6H-SiC with

an electron beam. ESO is quantified vs. chamber pressure, electron beam current, and

electron beam energy. Environments and parameters, which mitigate ESO are discussed.

Chapter 4 also reports on Re contacts to 6H-SiC. Surface preparations and their affect on

the electrical character of the as-deposited contact are discussed. The affect of heat

treatment at 1000 C for 120 minutes on contact performance is reported for each surface

preparation. Chapter 5 provides a summary of the key findings reported in this study and

the significance of the results to contact literature. Finally, Chapter 6 suggests future

work in the areas of contamination reduction, thin-film interfacial interaction, and

refractory metal ohmic contacts to 6H-SiC.















CHAPTER 2
REVIEW OF LITERATURE

Space and harsh environment microelectronics applications require materials,

which can endure high-power, high-temperature, and harmful radiation fluxes. In these

environments, traditional silicon based devices require bulky, expensive shielding, and

cooling for operation. Materials, which intrinsically resist radiation and thermal effects,

provide cost savings and are more attractive solutions than shielding. Devices made from

SiC can operate in these demanding environments. Interconnects which make contact to

external current and voltage sources must also be able to withstand the extremes and must

exhibit efficient current transport, i.e. low specific contact resistance ohmic character.

Many metals used in traditional applications may form good contacts to SiC but will

degrade with higher temperature/power applications. Development of a

high-temperature, contamination resistant, ohmic contact would be a significant stride in

the area of SiC device technology.

2.1 SiC Growth

Because more than 175 varieties of SiC polytypes can grow in nearly identical

conditions, development of single crystal bulk substrates and epitaxial films of SiC has

been a challenge to materials science researchers for many years."''17 The first and

arguably the most difficult step for growth of any semiconductor is suitable choice of

substrate. The unique structure of many SiC polytypes creates mismatched lattice

constants of as much as 20% vs. substrates. In addition, the high temperatures required







9

for growth make it difficult to find substrates which have a less than 8% mismatch in

coefficient of thermal expansion. These two properties alone limit substrate choice. ,93

In addition, SiC also tends to form small tube-like "holes" in its lattice during growth.217

The density of these 'micropipes' limits carrier mobility and degrades device

performance. Despite these difficulties, some success has been demonstrated using Si13 1

and cL-SiC915 as substrate material.

Modified Lely Sublimation,213'15 Liquid Phase Epitaxy (LPE),2"'1315 Chemical

Vapor Deposition (CVD),2,913'15 Gas Source Molecular Beam Epitaxy (GSMBE),2'15 and

Solid Source MBE18 are techniques most used for bulk and epilayer growth of SiC.

Nitrogen is a dopant for n-type and aluminum and boron are used to create p-type

material.19 While doping can occur during growth phases with the addition of N2, NH3,

or trimethylaluminum (TMA)2 '9, due to low diffusion constants in SiC, it is traditionally

accomplished through ion implantation after growth.2"13'17 Densities on the order of -101

cm-3 have been realized for both n and p-type materials with doping during growth, while

densities through implantation have reached -10 19-1020 cm3.219 Obtaining n-type

material is considerably easier than p-type due to SiC's ability to unintentionally dope

with N during growth and during high-temperature processing. It has been demonstrated

that donor carrier concentration continuously rose by up to 80% in only 0.5 hours of heat

treatment between 1100 and 1300 C.19 Achieving stable p-type material is still a

challenge today.

Several growth processes exist to obtain bulk, single crystal, SiC, but the most

successful employs a modified Lely sublimation procedure.2"'13'17 In this procedure, SiC

sublimes from source material held at 2300-2500 C. The gas phase SiC clusters pass







10

through porous high purity graphite enroute to the SiC seed crystal, which is held at

2100-2200 C in a low-pressure argon atmosphere. Boules of up to 20 mm in length, 30

mm wide, have been achieved with this process.13 Reduction of micropipes and

minimization of substrate defect propagation have occurred, but still provide continual

areas for research.

Liquid Phase Epitaxy (LPE) can be used for bulk growth of SiC at lower

temperatures (-1500-1700 C), than Lely sublimation techniques.2,13 SiC dissolved in a

Si melt is held suspended within an electromagnetic field. This procedure eliminates

carbon contamination that was evident from earlier procedures where carbon crucibles

were used. The Si melt is saturated with SiC at which point SiC begins to crystallize

from the melt. Epitaxial layers can be grown in excess of 150 microns per hour with this

technique. 17

Until the mid 1980's Lely and LPE techniques were the only source of SiC

crystals. At that time, large surface area growth of cubic SiC was accomplished by CVD

on Si substrates at temperatures much lower than even LPE techniques (1300 C).2,9'13,17

Various gas chemistries including Si, Sil4, SiCI4, CH4, C2H2, C3H8, C3H3SiCI3 and H2

have proven successful as CVD sources.13 Mismatch threading dislocations, stacking

faults (SF), microtwins, and antiphase boundaries (APB) are the most common structural

defects that occur with CVD growth of SiC. Orienting the substrate (0001) plane off axis

3-5 degrees toward the (1120) direction and etching the substrate surface seems to

significantly reduce propagation of these defects.2'9'13"17

A final technique for achieving thin epitaxial layers of SiC is gas source

molecular beam epitaxy (GSMBE). The advantage over other techniques is a lower







11

growth temperature (<1100 C) and cleaner ultra high vacuum ambients, however, the

lower temperature limits diffusion of dopants introduced during growth phases.2 "7 This

growth process shows promise for heteroepitaxial growth of 6H-SiC using a variety of

substrates as common as Si to those as unique as AIN films.17

2.2 Surface Preparation

Surface preparation of SiC is non-trivial. Wafer surfaces often require a

combination of physical and chemical modifications followed by cleaning steps in order

to produce a surface that is atomically smooth with a desired chemistry that is free of

contaminants. Figure 2 shows the complexity of this process with a surface preparation

decision tree for SiC. While various approaches exist to obtain a desired surface before

metallization, most will involve a mixture of the techniques shown.

Purchased wafers of SiC are typically received with stoichiometric crystal

surfaces terminated on either the carbon or silicon face. Factory polishing of these

wafers has been shown to leave rough surfaces or scratches on the order of 50-100

angstroms in depth.1520 Due to the physical strength of the Si-C bonds, processors are

largely limited to physical/mechanical polishing techniques using diamond embedded

pastes and polishing wheels to reduce large surface features. A cleaning phase is

required after using these techniques to remove residues from paste, washes, and tools.

An alternative, more finely controlled, technique for achieving atomically flat surfaces

uses hydrogen reactive etching at elevated temperatures (1600-1700 C). After 15

minutes of exposure, 4H-SiC wafers show several thousand angstroms of atomically flat

surfaces and 6H-SiC wafers have shallow undulating surfaces.2'22 Surfaces treated by

this technique are contaminated with hydrogen. Several studies have shown that














CHEMICAL






/ I
????9
No Documented
Techniques


Mechanical
Polish

H2 Etch
(1600-1700 C)


4,


Low Angle
flash (1000

Si Flux (90<


Ar+ Ion Etch

Ar/N2 Ion Etch


RIE
(CHF3/CF4/NF3/SF6)
(w/ 02 or H-2)


CLEANING .

Ar+ Acetone/Met
'C) 10-50%HF
DIN2 Silicon-rich


SL)


(/uu to 1IiUL )
KOH (500 C)


WET


Ar4 and N





Amorphous


HEAT
(1025 C 10min)
(425 C 5keV Ar+ 5 min)
(425 C 3keV Ar+ 10 min)
(425 C lkeV Ar+ 10 min)
(425 C 10 min)


Figure 2. A surface preparation decision tree that shows processes to achieve an
atomically smooth physical surface with a desired chemistry that is free of contamination.


Native Oxide
and Organics


Molten KOH
(500 C)

Heat
Treatment




Carbon-rich







13

contaminant hydrogen can diffuse to interstitial sites in the SiC matrix and cause

uncontrolled changes in the near surface resistivity because of passivation of dopants.23'24

Annealing in He and/or other cleaning steps must be taken to remove the adsorbed

species before metallization.

A smooth surface is the first step in the preparation process. Additional physical

modifications to the wafer may include patterning steps prior to metallization. The most

effective techniques involve inert-gas sputtering and reactive ion etching (RIE). Due to

their non-reactive nature and reasonable mass for energy transfer, Ar and or N2' are

commonly used as sputtering agents for cleaning oxide rich surfaces or for small surface

feature patterning.15'20'25'26 Due to the strength of the SiC bonds, however, long sputter

times with high power are required. The resultant surface is often disordered, and

embedded with Ar and N2. Additional treatment, usually with an anneal temperature in

excess of 1273 K, is required to remove contaminants and restore crystal order. 15,25

Alternatively, RIE with fluorinated plasmas (CHF3, CF4, NF3, SF6) produces

faster etch rates at lower power and can be used for more extensive patterning. 5,20,25,26

RIE drawbacks include possible anisotropy of surface features, preferential etching of Si

or C, and residues from reaction products or embedded species. In all cases, fluorine and

carrier gas (typically Ar, N2, 02, or H2) remain on or in the sputtered surface. Heat

treatment (above 500 C) can often reduce the concentration of the residue but in many

cases, it cannot be fully removed.25

Figure 2 shows the chemical surface of the SiC wafer is an important

consideration in addition to achieving an atomically smooth physical surface prior to

metallization. As previously mentioned, SiC is typically produced with a stoichiometric







14

crystalline surface that is terminated on either the carbon or the silicon face. Processing

techniques can be used to vary this surface to become more silicon-rich, more

carbon-rich, or non-crystalline and amorphous in structure. Choice of metal, interfacial

reaction, and final contact properties will dictate which surface is required prior to

deposition.

Carbon-rich surfaces may be desirable if a chosen contact metal readily forms

stable carbides. Surfaces with carbon-rich chemistry are easily produced. Many cleaning

techniques (sputtering, heating, wet chemical etching) used for achieving an oxide free

SiC surface produce a carbon-rich surface layer as a by-product.14152527 Heat treatment is

the most documented method for creating a reproducible carbon-rich SiC surface.5,15'5'2527

Bozack et al.15 describe four distinct temperature regions that affect the surface structure

of SiC. Below 900 K, there appears to be no preferential migration of C or volatility of

Si from the surface. From 900 to 1100 K, there is preferential carbon migration in C face

terminated structures. Si face terminated structures remain quasi-stable. LEED patterns

confirm graphite formation for both faces near 1100 K. From 1100 to 1300 K stability of

both faces is recorded with reduced carbon migration. Finally, above 1300 K there is

massive C migration to the surface of both crystal faces due to preferential volatility of

Si. The intermediate temperature regions are explained by temperature dependent crystal

surface reconstructions. These temperature regions can be exploited to achieve moderate

or thick carbon surface layers depending on desired chemistry.

Achieving a disordered, amorphous SiC surface is also well documented in the

literature as a result of ion bombardment.15,27 Non-reactive as well as RIE techniques can

contribute to disorder in the crystalline surface. Disordering the surface is not difficult








15

but doing so in a reproducible, controllable fashion is extremely challenging. Variable

concentrations of implanted sputtering agents are inherent with the process and the

broken Si and C bonds are quick to react with ambient oxygen. The resultant surface

makes a reproducible as-deposited metallization nearly impossible.

Silicon-rich surface chemistries may be desirable if a chosen contact metal readily

forms stable silicides. To date, however, there appears to be no documented procedures

for producing a silicon-rich surface from a stoichiometric starting surface. As a result,

the only way to create such a surface is by depositing Si on a clean, contaminant free,

stoichiometric SiC surface.

Attainment of an atomically smooth chemically desired surface has been

described above. The techniques that produce these surfaces, however, often require

subsequent steps to achieve surfaces free of contaminants. At the center of Figure 2 and

key to treatment of both as received samples and all samples treated for desired surfaces

is the cleaning process. There is no single treatment that will provide a SiC surface

completely free of contamination. Typically a combination of wet, dry, and heat

processing techniques are used to achieve a surface that is adequate for metal deposition.

A few of the more common approaches to these processes are described below along with

their limitations.

SiC, much like pure Si, readily forms a native oxide in atmospheric conditions.

This oxide and physiadsorbed hydrocarbons are the most prevalent contaminants in as

received samples and treated samples that are removed from vacuum prior to deposition

steps. Ex-situ, wet processes and vapor streams are the only methods for removing these

contaminants. Wet processing usually involves multiple steps to remove the adsorbates








16

and oxide. Acetone and methanol remove organic from the surface.15 An ex-situ

technique known as CO2 snow cleaning can also be applied to surfaces to remove more

tightly bound thick layer hydrocarbons. The process involves spraying adiabatically

cooled CO2 onto the substrate surface. The stream contains CO2 vapor, liquid, and

crystallized particles, creating a mixture of momentum transfer and solubility at the

surface. The process requires simple, but specialized equipment for fine control.

The next step in wet processing is oxide removal. Most techniques have limited

success with fully removing the native oxide. The two most practiced involve dipping

the native substrate in buffered aqueous HF acid or heating the substrate in molten KOH

at 500 C.15,25'27 With both treatments, the native oxide is reduced but not completely

eliminated. Enhanced results have been reported with HF dipping when the SiC has been

pre-oxidized beyond the native oxide. Unlike treatment with KOH, the resultant surface

after HF dipping preserves the C/Si ratio.'5"25 With KOH treatment, the surface is

25
carbon-rich.25 The remaining surfaces are typically rinsed again with acetone or

methanol to remove hydrocarbon byproducts and then with de-ionized water to remove

residue. The final surface is blown dry with a compressed N2 stream and placed in to

vacuum for further processing.'5'27 Additional treatments involve heat and/or dry

processing techniques to remove adsorbed H, N, F, and/or C from surfaces.

In-situ, there are a variety of dry processes which can be used to enhance wet

processes previously performed or which can be used with heat treatment to remove

adsorbed species and oxide. The most simple involves alternating low angle (<80

degrees) Ar' sputtering with 2 minutes of flash heating to 1000 C. This process

successfully produces an atomically clean SiC surface but may not preserve the crystal








17

structure.15 Bozack et al.15 present an alternative to this technique which is less

destructive and is described in more detail with heat treatment discussions below.

Less damaging in-situ techniques for oxide removal take advantage of the

reactivity of oxygen with other gaseous species at temperatures above 1100 K.

Annealing SiC in a gallium flux at 1250 K produces volatile Ga20 and effectively

removes oxide from the surface.27 Contaminant Ga and the volatility of Si at this

temperature, however makes this process less attractive. Substituting Si for the Ga flux

with a substrate temperature of 1170 K has proven to be a successful technique. SiO

volatilizes removing the native oxide. Excess Si replaces the volatilized species and

reacts with C contamination at the elevated temperatures to reform pure SiC.15'27 These

processes are good follow-on steps to ex-situ wet cleaning procedures, but are not likely

to be possible in most commercial metallization process lines.

Finally, in-situ heat treatment is a vital step for success in nearly all forms of

cleaning. All wet processing techniques must be complemented with flash heating or

anneals to remove physiadsorbed residues. In addition, all in-situ dry processing

techniques must occur at elevated temperatures or require annealing phases to restore

crystal structure and remove adsorbed or implanted species. Noting that heating is a

necessary step, Bozack et al.15 have documented a simple, fully in-situ procedure, which

claims success at removing native oxide, adsorbed species, and retains stoichiometry.

The authors do not verify crystal structure after performing this procedure and admit that

it may be compromised in the sputtering process. The procedure follows.

1. Heat the substrate to 1025 C for 10 min removing oxide and organic but also
graphitizing the surface.
2. Heat the substrate to 425 C while sputtering with 5keV Ar' ion beam for 5 min.
3. Heat the substrate to 425 C while sputtering with 3keV Ar+ ion beam for 10 min.








18

4. Heat the substrate to 425 C while sputtering with lkeV Ar' ion beam for 10 min.
5. Heat the substrate to 425 C for 10 min.

There are clearly a wide variety of preparation techniques which can be applied to

SiC in order to achieve contaminant free, atomically smooth physical surfaces with a

particular desired chemistry. The key is applying the appropriate techniques to reduce

surface roughness, maintain crystal order, and remove oxides and other contaminants for

the desired application. Performing these controlled steps allows the process to have the

most influence on Schottky barrier height and reproducibility of contact performance

after metallization.

2.3 Contact Physics

A metal to semiconductor contact is a complex system whose properties can be

difficult to explain and even more challenging to model. A stepwise approach to the

development of a contact helps to characterize the key parameters that may affect the

final properties observed in the system. First consider an ideal, clean, n-type substrate

(semiconductor) surface prior to metal deposition. An energy diagram characteristic of

this surface is shown in Figure 3 where Ef is the Fermi energy level, Os is the

semiconductor work function, and Xs is the semiconductor electron affinity. Without

modification, the semiconductor surface will maintain this energy configuration.

Consider the addition of static charge to the surface of this system. The carriers,

electrons in the case of an n-type substrate, will either be attracted to or repelled from the

surface of the system based on the polarity of the static charge. The bands compensate

for this distribution by bending upward or downward to form a depletion region (W) in

the semiconductor that neutralizes the surface charge (see Figure 4).7 The magnitude of







19

W is proportional to the intensity of the surface charge and the density or number of

carriers that have been displaced, and can be expressed by Equation 1.28 In theory,


W= 20- -11 (1)


V is the static surface charge and qND is the dopant charge density. In practice, V is the

summation of all applied voltages and surface voltages created via contact between a

metal and the semiconductor surface. The permittivity of free space and the dielectric

constant of the semiconductor are expressed with the constant FoF0s. 72

Energy" --- --------acuum

Os Xs Conduction Band
Ef
No electrons
Valence Band Electron filled



Distance from surface
Figure 3. Energy band diagram for an ideal, clean, n-type semiconductor surface.

Instead of applying a static charge or bias to the semiconductor surface, consider

the addition of a few adsorbate atoms (metal or contaminant). The adatoms will polarize

or ionize as a result of differences in electronegativity with respect to the semiconductor

surface. A local surface dipole will be established which interacts with the energy bands

of the semiconductor in the same fashion as the static charge previously discussed. 15'29,30

With increasing adsorbate coverage, a sheet of surface dipoles will develop that will

either increase or decrease the surface potential in the semiconductor. From this








20

perspective, it is clear why contaminant free deposition of material is so important in

contact development. A contaminant adsorbate could adversely influence the surface

potential of the semiconductor causing a permanent change to the electrical structure of

the interface. Care must be taken for this not to occur. As adsorbate coverage increases,


energy + Energy + _
++ -
S + S +

u + u + Conduction Band
r-=- r+ Ef
f Ef f Ef
a a+ +
c c +
e e + Valence Band

S Dl+ +Depletion Region (W) -Depletion Region (W)

Distance from surface Distance from surface
Figure 4. Band Bending for an ideal, clean, n-type semiconductor in the presence of a
static surface charge: a) negative and b) positive.7

the growing potential, AO, between the added species and substrate can be approximated

as a capacitor until the individual dipoles begin to interact and can be expressed as a

function of adsorbate coverage (Equation 2).15,29 0 is the fractional monolayer of


AO = 4r90co,


adsorbate coverage (0 to 1), cTo is the number of adsorption sites available on the surface,

and p, is the effective dipole moment.15 je becomes dependent on adsorbate coverage

when dipoles in close proximity begin to interact and effectively reduce the overall dipole

with the surface.'5 p, is expressed as a function of adsorbate coverage (Equation 3)

where Plo is the dipole moment per adsorbed particle when coverage is zero, and ca is its

p(0) = Po[l+9a(000O)3/2]-I (3)


E


I







21

effective polarizability.5 With the addition of multiple monolayers, changes in

AO become negligible. At this point, a thin film has formed on the semiconductor and

AO represents a measure of the work function differences between the bulk

semiconductor and the film. This condition is shown graphically in

Figure 5 for an n-type semiconductor with work function Os < m both before and after

contact with the metal adsorbate.

Vacuum
.. ....... ..... ..... .... .






Metal n-type semiconductor

Figure 5. Simple contact between metal and an n-type semiconductor with work function
s < OM both before and after contact. This condition is known as a Schottky contact.
The Schottky barrier height, O,, is shown to be equal to OM- Xs31


Before the thin film was applied to the surface, the electrons above the Fermi

level in the semiconductor were at a higher energy state than those in the metal. For

energy minimization, electrons flow out of the semiconductor into the metal until the

Fermi energies are equal. The metal then contained a higher density of electrons giving it

a negative polarity at the interface compared to the semiconductor. The bands curve

upward as predicted with Figure 4 by an amount of energy equal to AO (Equation 4)

giving rise to a positive space charge region. A potential barrier to electron flow from the

metal to the semiconductor is created in the conduction band and is defined by b.32 This

condition (Equation 5) is known as the Schottky-Mott Limit and the contact is a







22

Schottky or rectifying contact.32 The magnitude of W, in the case of a rectifying contact,

limits the ability for thermionic electron flow. In a device, for example, when bias is

AO= OM OS (4)

b OM XS (5)

applied to the contact it will either widen W as in the case of negative or reverse bias

preventing current flow or shrink W as in the case of forward or posit, bias enhancing


Ef

E+ n-type
+ E+ semiconductor
S+ ,n-type
M + semiconductor Mt +
Metal + Metal-
T + T Depletion Region + UtDepletion Region
Reverse Biased Forward Biased
Figure 6. The response of the Schottky contact space charge region, W, to reverse and
forward bias.7

current flow (Figure 6).732 While rectifying nature is desirable in some devices, for

others, it is important for current to flow with either bias condition. A contact capable of
Vacuum



E f ------------------ E f




Metal n-type semiconductor
Figure 7. Simple contact between metal and an n-type semiconductor with work function
M < Os both before and after contact. This condition is known as an ohmic contact.








23

this kind of performance can ideally be achieved by simply choosing a metal whose work

function is less than that of the n-type semiconductor. When the metal is deposited,

electrons with higher energy will flow from the metal into the semiconductor creating a

negative space charge region. According to Figure 4 the bands will bend down without

creating a barrier to current flow. This condition, shown in Figure 7, is defined as an

ohmic contact, because the interface allows the magnitude and direction of current flow

to be linear with respect to applied bias obeying Ohm's Law.

Contacts that obey Schottky-Mott rules are said to be "ideal" because prediction

of their characteristics involves a simple comparison of the properties (4M, 4s, and Xs) of

the starting materials. Contact schemes that obey this "ideal" situation are very limited in

number.7,32

To understand why most contacts deviate from Schottky-Mott rules, consider the

difference between bonding character in ionic and covalent materials. Ionic materials

completely transfer electron density from the cation to the anion and they are bonded by

electrostatic attraction. Electrons are not "shared" between atoms in the bulk material.

As a result, surface termination, where a contact film would be applied, is energetically

stable.

In contrast, covalently bonded materials "share" electron density between atomic

orbitals of different species in the bulk material. Termination of the bulk in these

materials leaves empty or partially filled atomic orbital states known as "dangling bonds"

on the surface. These surfaces are not energetically stable. Electronic states exist which

can serve as traps for carrier density between an applied metal and the semiconductor

surface. As a result, covalently bonded materials are more likely to allow voltage drop








24

from an applied thin film to occur across this layer of surface states rather than creating a

space charge within the semiconductor as predicted by Schottky-Mott. The Fermi level

in cases such as these is said to be "pinned" because it will remain fixed regardless of the

difference in work function between the metal and the semiconductor.

The interface index parameter (Equation 6), S, has been developed to measure the

dependency of barrier height, b, on a deposited metal's electron affinity, XM.15'32'33

o pb
S- 69b
5 M (6)
6XM

As S approaches 0 the barrier is independent of the metal and the Fermi level exhibits

strong pinning. As S approaches 1, the barrier height begins to follow Schottky-Mott

prediction and the Fermi level exhibits no pinning. Trends for S, like those depicted in

Figure 8, are typically shown versus the difference in electronegativity, AX, or ionicity of

the constituent components for a semiconductor. Compounds with low AX are covalent

while those with large AX tend to be ionic. Figure 8 depicts the reality of the theory

previously presented. The degree of covalency affects the extent of Fermi level pinning

and the dependence of barrier height on surface states. SiC has an electronegativity

difference of 0.65 eV (Si: 1.90 eV, C: 2.55eV) indicating a high probability that surfaces

states will affect the Fermi level.34 Complete pinning, however is not predicted.

Figure 8 only shows S as a function of constituent attraction. A different

approach, as shown in Figure 9, compares the heat of formation for the compound and

measures the resistance of constituents to reaction with other materials. Covalent

semiconductors exhibit lower heats of formation traditionally indicating less resistance to









25


reaction. Interfacial reaction with an applied thin film (contaminant or intended metal)

will make the barrier less dependent on the electrical properties of the parent


I .. .
1.0 A3s 7 Z ~Alp, ST'0
@Gas sno'
50?
0.8 C-.
Covalent Ionic
0.6 s.
S .
0.4 sc
GaTe. CUse
CITea
0.2 GOP
Si 9Ge
*InSb I~
0.0 1-- - L- -
0.0 0.5 1.0 1.5 2.0 2.5
AX

Figure 8. Interface index as a function of ionicity between constituent elements for
various semiconductors.'


Covalent


1.0 .,s *ZMO *OS, C .

0.8 Ionic
; Ionic
0.6 ,S
SS
0.4
GOT.. 6
0.2 A' "
eG GAM~

*n mp___ /
0.0 "

0 50 100 150 200 380 400 420

AHf (kcal/mol)

Figure 9. Interface index as a function of heat of formation, AHf, for various
semiconductors.

semiconductor and metal. Figure 9 verifies this assumption to be true.35 The more

resistant ionic semiconductors have smooth, well-defined interfaces that continue to








26

follow the Schottky-Mott prediction. AHf for SiC is -15 kcal/mol, but AHf is -217

kcal/mol for SiO2, indicating that a surface reaction to form SiO2 is predicted. The

prevalence of native surface oxide shows this to be true.36 The strength of this

passivation layer explains the difficulty with removing it during surface preparation steps

prior to metallization.36

A model has been proposed for the non-ideal metal/semiconductor contact. In his

theory, Bardeen37 suggests that "dangling bond" electronic states can be treated as a

semi-insulating layer, 6, on the surface of the semiconductor. This layer has an effective

energy level Oo that represents allowed electronic energy states within the band gap.37'32

When metal is deposited on the surface, charge compensation takes place within this thin

insulating layer and as previously described for covalent semiconductors, the energy

bands remain unaltered.15'15'37 The Fermi level remains pinned at 0. This condition

known as the Bardeen Limit is expressed by Equation 7 and shown graphically in




{M E\,,\M 4,,= Ega 0 O, M

-;;- b g M i------ ---
^ Eg Eg s






A B



C
Figure 10. Formation of the semi-insulating layer, 8, from surface states and its affect on
band bending structure: a) before Fermi level pinning, b) after pinning before contact
with metal, c) after contact with a thin metal layer.32








27

Figure 10.15,37 O8 represents a quasi-work function for the layer 6, Efm is the Fermi level

b=Eg- 00 (7)

of the metal, Efs is the Fermi level of the semiconductor, and Eg is the band gap of the

semiconductor. The change in workfunction, A0, with application of the thin metal film

represents the energy difference between 08 and OM and the barrier height is unaffected.

2.4 Contact Properties and Characterization

Section 2.3 explained contact theory from the perspective of the metal, the

physics of the semiconductor surface, and the stepwise addition/interaction of atoms of

metal or contaminants with the semiconductor surface. The Schottky-Mott rules were

derived from this discussion. While these rules can be applied to an ideal metallization,

in principle most semiconductors do exhibit some degree of Fermi level pinning and

processing must take into account surface states and their effect on contact performance.

This section describes the physical properties of interest that affect the success or failure

of a contact for device application, how to potentially control those properties, and how

to characterize/quantify them.

As previously described, in theory, a contact is ohmic if it follows Ohm's Law

and allows current flow that is linear in magnitude and direction with applied bias. In

practice, an ohmic contact is only useful if the amount of voltage required to push current

through the interface is minimized. The total contact resistance, Rc, is a measure of this

impedance to current flow. Since Rc (Q) is an extrinsic property that scales with contact

size, a more useful parameter for measuring ohmic contacts is specific contact resistance,

Pc (Q2cm2). Equation 8 shows that Pc is the intrinsic property of the contact







28

corresponding to resistance that has geometry included with its value.38'39 V is the

applied bias, J is the current density.

PC=dV (8)
Pc V v=0


In order to reduce Pc, a large current density must be achieved with minimal

applied voltage. This situation is inherent for the case of an ideal ohmic contact. As

shown in Figure 7, no potential barrier exists to prevent electron flow with the application

of bias. In the case of an ideal rectifying contact, or more generally, when a potential

barrier exists from Fermi level pinning, the current density is calculated with


J=AT2 e, kT][ e( 1 (9)



where A is the Richardson constant, T is temperature, q is the charge on an electron, k is

Boltzman's constant, Ob, is the barrier height, and n is an ideality factor.,5 For ideal

rectifying contacts, the ideality factor is equal to 1. According to this equation, the only

interfacial property that can affect current transport is barrier height. A minor reduction

in Ob will generate an exponential change in current density. As a result, any

modification that can reduce barrier height will increase current density indicating an

overall reduction in specific contact resistance.

Due to Fermi-level pinning, reducing the barrier height in non-ideal systems may

be very difficult. In order for these systems to behave like ohmic contacts, carriers must

either receive enough energy to "step" over the barrier, or tunnel through it. These two

forms of current transport are respectively known as Thermionic Emission (TE) and Field

Emission (FE).6'738"40 TE dominates in thick depletion layer (low barrier height)








29

conditions when electrons are unable to tunnel through the barrier. FE on the other hand

dominates when the depletion layer is narrow and bias provides enough energy for

electrons to tunnel through in either direction. In reality, both transport mechanisms

contribute to current density.

Since TE requires elevated temperatures, in practice, it is more advantageous to

maximize FE through depletion layer control. Equation 1 shows that this can be achieved

with high concentrations of doping.7'71'32 In general, for carrier densities less than 1017

cm-3, TE will be the majority transport mechanism while above 108 cm-3, FE will

dominate.3239'41' For intermediate doping density, a mixture of the two mechanisms

known as Thermionic-Field Emission (TFE) dominates.32'39'41 As a result, high

concentrations of doping during growth may provide an effective method for enhancing

ohmic behavior in materials that are predicted to exhibit Fermi-level pinning, such as

SiC.

The end measure of a good ohmic contact is its ability to exhibit low specific

contact resistance. In principle, Pc could be calculated from the characteristics of the

semiconductor and metal contact that have been discussed. In reality, external influences

such as interfacial reaction, surface damage, and oxidation affect Pc making it difficult to

model.41 More often, techniques must be use to empirically measure its value. Table 5

gives a short list of the more commonly used techniques to experimentally determine Pc.

The most commonly used technique proposed by Berger in 1972 is the Linear

Transmission Line Model or the Linear Transfer Length Method (TLM).38' 39,42 With this

technique, equally sized rectangular metal mesas are applied to a semiconductor surface

with varied spacing as shown in Figure 11. The total contact resistance, RT, between











Table 5. Summary of methods for experimentally determining pc. The most widely used
method for SiC contact work has been the Linear Transmission Line Model.39
Method Test Structure Advantage Disadvantage
Cox and Strack Varying diameter dots Easy to fabricate; Difficult to measure
on top surface with gives vertical small values of
large contact on back measurement of contact resistance
of substrate contact resistance


Four co-linear dots
with the same
diameter


Diffused path and
metal layer in a
four-point structure


Linear TLM






Circular TLM


Easy to fabricate




Capable of measuring
very small contact
resistance


Equal area rectangular East to extract
contacts with varying moderate to small
spaces between values of lateral
contacts on an isolated contact resistance
area


Varying radii
concentric circles on
epitaxial or implanted
material


Inaccurate, does not
account for current
crowding at the
contact edge

Test structure requires
diffused layer


Test structure requires
two masking levels


Capable of extracting Requires extremely
moderate to small low metal sheet
values of lateral resistance and the
contact resistance with analysis of modified
only one masking Bessel functions
level


adjacent contacts is plotted vs. contact spacing, d, such that the y-intercept is equal to

twice the total contact resistance of a single contact, R,. The x-intercept yields twice the

distance of a value known as the transfer length, Lt. This distance refers to the space

through which current transmission has dropped the voltage across the interface to a

value of l/e of its original value. The slope of this graph is ART/Ad or more appropriately


Pc = LTRs


(10)


R,/Z, the sheet resistance of the semiconductor under the contact width Z. Using the

relationship defined with Equation 10, specific contact resistance can be calculated from


Kuphal


Kelvin








31

the values obtained with this type of plot. To obtain RT measurements for this plot, the

slope from I-V curves are obtained for adjacent contacts.






I L I d |ILI| d, | L |d, L |dA L I




R ^^Slope== Rs/Z
RR
-" 2Rc
K
2LTr 0 d
Figure 11. A TLM test structure and a plot of total resistance as a function of contact
spacing, d. Typical values might be: L = 50 (tm, W = 100 gm, Z-W = 5 .tm (should be
as small as possible), and d = 5 to 50 n. '38

In theory an ohmic contact would exhibit a linear current-voltage (I-V) trace for

all applied voltages. In reality, this is rarely achieved. As a result, a figure of merit is

often applied that reveals the degree of linearity (LM) in a near-linear I-V trace. It is

dV
dl V-=5
LM = -=(11)
dl V=0


simply the ratio of the slope of the curve at V = 5 to the value at V = 0 volts and is

expressed by Equation I I1.44 As LM approaches 1, the curve indicates a truly linear

ohmic contact. The value of LM can be considered a measure of accuracy when

determining RT values for a plot like the one in Figure 11.








32

2.5 Ohmic Contacts To n-Type 6H-SiC: Present Demonstrated Research

Sections 2.1 to 2.4 explained the growth of SiC substrates, the methods and

concerns for preparing substrate surfaces, general contact physics theory, and techniques

to measure the properties that typically define success or failure for an ohmic contact.

Section 2.5 will discuss actual demonstrated results for the preparation and

characterization of contacts made to n-type 6H-SiC.

2.5.1 Impurity Concerns

As discussed in Section 2.3, the physics that governs contact formation during

atomic additions of metal to the semiconductor surface will also apply to atomic

additions of impurities. Minute adsorbate contamination during metallization has been

shown to negatively influence the surface potential of a semiconductor causing

detrimental changes to the electrical structure of the interface.31,44"45 46'474S As a result,

contaminant free deposition of material is extremely important during contact

development.

An understanding of impurities, their source, and their affect on a

metallization/semiconductor interface is a large step in contact design. For SiC, oxygen

is the most prevalent impurity and consequently, the most studied. SiC, like Si, readily

forms a thin, insulating, surface oxide in room temperature air. For some applications,

such as metal-oxide-semiconductor (MOS) devices,4950'51'52 this property is an advantage

for the semiconductor. For contacts, however, interfacial oxides may contribute to

rectifying character by creating barriers to charge transport and limiting interfacial

reaction. Using surface preparation techniques discussed in Section 2.2 to reduce the








33

impact of oxidation prior to deposition is the first step. Limiting oxidation during the

deposition process is also of concern.


ft C G o 0 0 0 ##a*
Og o- .A? ." s.7 "
0 *** *, Y
20 ... ..

15 'a A A, e S

A 11-A AA A A'
A & &A& AA
** A51 5
C'A A 0N

CAi* # A A
Aa a AA

4 A
AA Aj A
k.5 A. JA- I


0 500 1000 1500 2000 2500
Beam Irradiation Time (sec)

Figure 12. Unintentional incorporation of oxygen into a Ni thin film during deposition
from a 99.9945% pure Ni filament in a vacuum base chamber pressure of 6.2 x 10-9 Torr
(filament off) while monitoring with a 500 nA, 3 keV incident primary Auger electron
beam. Ni (LMM) and 0 (KLL) transitions normalized for concentration are plotted as a
function of beam irradiation time59

With in-situ analysis, it is common practice to monitor step-by-step growth during

contact formation. 15 29.3053 Electron beam techniques such as Auger Electron

Spectroscopy (AES) are some of the traditionally used analytical methods. It has been

shown that exposure to electron beam irradiation in some environments causes electron

stimulated adsorption (ESA) and oxidation (ESO) on Si. This phenomenon results in

very thin (10 to 20A) Si02 layers. 51.54,55.56,57.58 Initial findings for similar processes

occurring on 6H-SiC have been reported.59,60 Figure 12 shows how substantial levels of

oxygen were unintentionally incorporated into a Ni thin film during deposition from a

99.9945% pure Ni filament in a vacuum base chamber pressure of 6.2 x 10-9 Torr while

monitoring with a 500 nA, 3 keV incident primary Auger electron beam. The oxygen








34

signal is saturated in less than 500 sec while nearly 2500 sec of deposition are required to

achieve an equivalent concentration of Ni. While the mechanisms and rates for oxygen

adsorption and oxide formation are well studied for Si, there is little or no

characterization of these phenomena for SiC.61'62'63"64 One of the purposes of this work is

to report quantified studies of this phenomenon as a function of analysis parameters

(chamber pressure, beam energy, beam current) and present solutions to reduce the

oxidizing effects of in-situ real time analysis with an electron beam during deposition of

metal contacts.

2.5.2 SiC and Fermi-Level Pinning

As previously discussed in Section 2.3, the interface index parameter, S, is a

measure of a semiconductor's tendency toward ideal behavior. Because covalent

materials are more likely to have allowable electron states with abrupt surface

termination, they tend to exhibit Fermi-level pinning. Figure 8 predicts SiC to have an

index of 0.4 indicating that it will exhibit partial Fermi-level pinning.'15"33 Figure 9 on the

other hand predicts SiC's reactivity will reduce the S parameter to 0.05 making the

barrier height independent of the parent metal's work function and interface properties

will solely depend on reaction products.'5,35 Using the same data, however, the large S

parameter of Si02 indicates if the native oxide for SiC is not completely removed prior to

deposition, reaction is unlikely and the contact character may be solely dependent on the

relationship between metal and surface oxide work functions.

In practice, results have been shown to vary. Porter et al.65 report the barrier

height for as-deposited contacts of Ti, Pt, and Hf on n-type 6H-SiC varied by less than

0.2eV while predicted values varied by up to 1.75 eV. The lack of variance suggests










66
complete pinning of the Fermi-level for these results. Rastegaeva et al. report a similar

existence of increased surface state density for as-deposited Ni contacts especially on the

carbon face. With elevated temperature, however, low specific contact resistance ohmic

contacts result from reaction of Ni with Si in the SiC matrix. The authors report that the

presence of carbon is key to contact behavior. This conclusion supports earlier claims by

Porter et al.14 that Ef can be artificially pinned after crystal heating due to carbon-rich

surfaces and may be related to intrinsic carbon vacancies which vary with polytype.

With more recent reports of metal dependent barrier height, Porter suggests in this work

that partial pinning is possible and older literature reports of complete pinning may have

been due to poorer substrate quality and preparation techniques.14

67
Yakimova et al.67 support Porter's claim of partial pinning as a result of surface

treatment by demonstrating moderate dependence of barrier height with respect to the

work function ofCr, Mo, Ta, W, Au, and Ni contacts. Still, other authors again report

pinned behavior in SiC, but are able to promote ohmic character contacts with TFE and

FE through excessive doping.68"69 Finally, Waldrop et al.70 report completely unpinned

behavior for contacts made to n-type 6H-SiC with Pd, Au, Ag, Tb, Er, Mn, Al, and Mg.

Barrier height was reportedly crystal face dependent but varied as predicted by

Schottky-Mott for the applied metals.

It is clear from the literature that a definitive understanding of the SiC surface has

not been achieved. Most reports agree however, that surface preparation processes before

deposition are the key to limited or full un-pinning of the Fermi level.








36

2.5.3 Metallic Reactions With n-Type 6H-SiC

Traditional metallization contact schemes have been reported in the literature for

Si61.62,63,64 GaAs31,71.72.73.74,75.76.,77 GaN78,79.80,81,82.83.84,85,86,87,88,89,90,91 and other popular

semiconductors. Choosing appropriate candidate elemental or compound contact

schemes for SiC is not as trivial since device application requires exposure to harsh

environments. As previously described, SiC's advantage over traditional, more mature,

semiconductor technologies is in the areas of high-power and high-temperature

environment operations. Metals must be chosen which will not limit operation in these

environments. Since most documented stable metal reactions with SiC occur at

temperatures in excess of 1000 C, it is important to consider metals that can be exposed

to these extremes in the contact fabrication process. 10,14,15,19,39,92 The titanium through

copper families of transition elements are acceptable choices with nominal melting

temperatures ranging from 1246 to 3422 C.36 In particular, elements which are stable

near or above the sublimation temperature of SiC (3100 C) are Ta, W, Re, and Os.

A standard approach to modeling M-SiC (metal-silicon carbide) contact reaction

considers the interfacial layer as a ternary phase M-Si-C system. Schmid-Fetzer and

Goesmann mathematically modeled the behavior of Ni, W, Cr, and Ti with 6H-SiC and

then generated compressed powder pellets to examine experimental data via XRD after

various temperature anneals.92 The resulting ternary phase diagrams shown in Figure 13

indicated Ni reaction with Si, tie lines for two compounds of W with Si and C, and

extremely complicated systems for Cr and Ti with compounds both to Si and C.

Porter et al.14 report that clean SiC surfaces readily react with most metal contact

schemes as Figure 13 would indicate. These reactions may be dissociation limited,









37

however, and the reaction kinetics can be extremely different from reactions with

elemental Si or C. For example, between 570 to 1200 C, Ti will react with SiC to form

TiC, Ti5Si3, or both TiC ,, and TisSi3. Mo, annealed at 1200 C and W annealed at 950


1C0
1300 C ^
o


850 C


Si W


1000 C \

2 0


r20~ -


Cr e, I, V: --L7 r^
Cr Cr CrS3 CTS S
6a t at%SI


-lSi
*tL%SI


T=Ti3SiC2

SiC


i Si
It. Si


Figure 13. Isothermal sections of ternary M-Si-C phase diagrams for Ni (850 C), W
(1300 C), Cr (1000 C), and Ti (1000 C).92

to 1100 C revealed similar M2C (M=metal) and M5Si3 structures. In all three cases, the

resultant phases are different than would be observed if reacted with elemental Si and C

alone. While a few metals favor reaction to one component such as Ni, Co, and Pd for Si,











most such as Ti, Cr, Mo, Fe, Nb, Pt, and W show reaction products with both Si and C

and are not easy to model.4'14'92

Ni has been the most widely studied ohmic contact to SiC because it is easy to

fabricate and its simple reaction kinetics with only Si (as shown in Figure 13) are readily

modeled.13' 14.15'19'39'6667'69'92 The as-deposited Ni contact is rectifying. With increased

Table 6. Ohmic contacts on n-type a(6H or mixed)-SiC. NR = not reported; RT = room
temperature; b.p. = base pressure. Surface preparations which consisted of at least a
surface oxide etching step and a hydrocarbon removal step by heating in high or
ultra-high vacuum were rated as "very good". Those which consisted of at least a
chemical clean and a surface oxide etching step but lacked the capability for heating in
high vacuum were rated as "good". This rating system is based on analyses of the SiC
surface by XPS.14'39
Metal Method Temp Pressure Anneal o Carriers Surface


Cr melting NR
Ni e-beam evap NR
TiN ion-assisted 350
e-beam evap
TiW sputtering RT

Ni e-beam evap RT
Ni-Cr sputtering RT
60-40%
W thermal evap NR

W thermal evap NR


TiW sputtering
Ti thermal evap

Mo sputtering
Ta sputtering
Ni/ resistive evap
3C-SiC
Ni e-beam evap
Ni sputtering
WTiNi sputtering
Ni thermal evap
Ti-Al thermal evap
TiC CVD


RT
RT

107 b.p.
10-7 b.p.
NR

RT
NR
NR
NR
NR
1260


(Torr)
10-3 b.p.
NR
10 1 b.p.


(C)
>2130
1000/20 s
600/30 m


106 b.p. 02 Plasma
600/5 m
10-6 950/5 m
10,6 b.p. 950/5 m


NR

NR

10.6 b.p.
NR

NR
NR
NR

10-6
10-5 b.p.
10-5 b.p.
NR
NR
NR


(fQcm2) (cm"3) Prep
NR NR NR
1.7x10-4 4.5xl017 NR
4x102 lxlO18 very
good
7.8x104 4.7x10'8 02
plasma
mid 10.2 4.7x10'8 good
1.8x10-3 4.7x1018 02


1200 to 1600 5x103 to
Ixl 0-4
1200 to 1600 1xl02 to


750/5 m
none

none
none
1000/30 s

950/2 m
1050/5 m
1050/5 m
1000/5 m
1000/5 m
1300/15 mH,


5x104
8x10-4
1xl102 to
<2x105
Ixl 0-4
lxl 0-4
6x 105 to
<1.7x105
<5x 106
10-3_ 104
103_10-4
lxl 06
1.3x105


plasma
3xRl08 tovaried
lxl019
lxl0'17 togood
lxl01'9
7 to 8xl0108good
2x10R togood
lxl020
>lxl1019 good
>lxl019 good
I to 2xl018very
good
7 to 9xl0t8good
9.8x1017 good
9.8x 1017 good
4.5x1020 good
4.5x1020 good
4xl019 NR


(0c:


)_











Table 7. Ohmic contacts on n-type b-SiC. Multi-layer contacts are listed in sequence
(left to right) from the topmost layer to the layer in contact with the SiC. See Figure 6 for


a description of the rating system for surface preparations."'
Metal Method Temp Pressure Anneal pc
(C) (Torr) (C) (Qcm2)
Ni NR RT NR 930/3m NR


RT NR


e-beam evap NR


thermal evap
thermal evap
thermal evap
thermal evap


sputtering RT
MBE RT


106 b.p.
10-6 b.p.
10-6 b.p.
106 b.p.

NR
UHV


none


700/30m NR


none
1250/5m
1250/5m
1250/5m

850/5m
1000


sputtering RT 10-7 850/30m
sputtering RT 10-7 b.p. 300/30 to
90m
sputtering RT 10-7 b.p. none to
600/10m
sputtering RT 10-7 b.p. 1000/10s+
450/390m
sputtering RT 10-7 b.p. 1000/1Os +
450/390m
e-beam evap NR 106 b.p. none or
1200/60m
sputtering NR NR 900/3 to 5m
soutterin2 NR 10-7 b.o. 650/1 h


Au/Pt/W sputtering NR 107 b.p.


Au/Pt/TiN/ sputtering
Ti
Pt/Ti/W/Ti sputtering


NR 107 b.p.

NR 10-7 b.p.


sputtering NR 107 b.p.


1.6x10-'
1.4x10'
7.0xl 0-2
3.0x 102

2.0x10-2
NR


2.4x10-I
7.6x10-3 to
9.2x 10-3
1.5x10-2 to
2.3x103
l.lx104

3.9x104

6.2x10-2 to
4.0xl0 2
NR
l.lxl 04


650/8h 2.0x104


650/31 h


1.4x104


650/3h 2.6x10-4


none


sputtering NR 10-7 b.p. none to
1000/60m
sputtering NR 10-7 b.p. none to
900/30m
sputtering NR 10-7 b.p. none to
500/30m


103

7x10-7to
4.3x106
xl0-4 to
lxl05
6x10-6to
IxlO5


Carriers
(cm3)
BxOto
1xc10'"
6x 1016 to
lxl017
6x 1016 to
lxl018

5x1016 t
5x10'6

5x1016
5x I016

5XI016
NR


2x1017
10" to
1018
1017 to
1028
10187 to

1028
1017 to
lO'1 to
1018


NR

1018
1017 to
1018
lOto
1018
lolto


1017 to
1018
1017 to
1018
5x10'9

5xl019

5xl019


Al
Ni
Cr
Au-Ta
(97-3 at.%)
TaSi2
Ag


w

TiSi2

WSi2


Ni
Au/Pt/Ti


Surface
Prep
fair


fair

good

good
good
good
good

good
Sputter
w/500
eVAr
sputter
good

good

good

good

NR

good
sputter

sputter

sputter

sputter

good

fair

fair

fair








40

anneal temperature (570 K) Ni begins to react with Si forming a disordered NixSiy layer.

Above 770 K, Ni2Si phases form with carbon uniformly distributed in the layer but the

rectifying character remains. With increased temperature (870 K) carbon migrates to the

surface of the contact to form a layer of graphite. The contact becomes ohmic upon

annealing to 1220 K. The properties are attributed to a fully passivated reaction zone of

Ni2Si free of carbon contamination. The contact has been shown to remain electrically

stable after 300 hours in a 923 K environment, however the reaction of Ni with Si

continues.15 Over time, the Ni is fully consumed by Si reaction and the metal contact is

destroyed. A diffusion barrier is required for this contact scheme to allow extended

application in high-temperature environments.

Other contact schemes take advantage of tunneling through FE and TFE.14'68'69

Metallizations of this type form a Metal-Insulator-Semiconductor (MIS) structure at the

SiC interface. The MIS structure passivates dangling electronic states with an extremely

thin (5 to 15 angstrom) insulating layer.14'93 Low work function elemental or compound

metal schemes favorable for ohmic contact formation can supply electron density via

tunneling through the thin insulating film. Ti, which has in many cases been shown to

exhibit rectifying character, can form an ohmic contact when reacted with N during ion

assisted reactive evaporation. Si-N forms the MIS structure.93 Ti-W contacts have also

shown ohmic behavior on SiC when the substrate surface was first reacted with an

oxygen plasma to form an Si02 MIS structure.94

While great strides have been made in explanation of data from demonstrated

contacts, property and performance prediction models are lacking. Because of the

difficulty in modeling, an only moderate understanding of SiC surface kinetics, and








41

possibility of unique reaction products, in the end, there is no substitute to date for pure

empirical results to determine new contact schemes. Table 6 and 7 summarize the most

promising elemental and compound ohmic metallizations demonstrated on n-type SiC.

2.5.4 Rhenium As a Contact

As previously described, a limited number of transition elements (Ta, W, Re, and

Os) have ideal temperature stability (above 3000 C) for SiC contact metallization

schemes. Of these metals, only rhenium and osmium are hexagonal in structure, making

them likely to have better lattice match to 6H-SiC.36 Osmium metal is stable in

atmosphere, however powders and spongy forms of the material give off a tetra-oxide

vapor which is highly toxic. Concentrations as low as [107g/m3] can cause lung

congestion, skin and eye damage so the material is considered hazardous and is not

practical for metallization studies.36 Rhenium, however, is stable in atmosphere, has

limited reaction with oxygen, and all forms of oxide become volatile with heat treatments

in excess of 1000 C. With a melting temperature of 3186 C, it also has nearly the

highest elemental thermal stability, rivaled only by carbon and tungsten.36 In addition,

possible reaction to form a stable diffusion barrier at high temperature is predicted by

Re-Si phase equilibria64 (Figure 14) and several reports of Re contacts on Si substrates

have reported favorable reaction to form silicides.62.63'95 These properties suggest

investigation of Re as a stable high-temperature ohmic contact could be promising.

To date, a very limited number of reports on the behavior of Re with SiC surfaces
96
have been published and the results are varied. Chen et al. reported the first Re

thin-film contact on n-type [3-SiC in 1993. In this study, -40 nm thick films of Re were rf

sputtered onto 5gtm thick CVD grown (3-SiC, unintentionally doped n-type (1017 cm-3)











Weight Percent Silicon
10 20 30 40 50 70100


b.5.

31(

29(

27(

25(

23(

21(

19(

17(

15(

13(


20 30 40 50 60
Atomic Percent Silicon


70 80 90 100
Si


50 60 70 80 90 11
C Atomic Percent Rhenium
Figure 14. Phase diagrams for Re-Si64 and Re-C equilibrium. The Re-C data were
generated from JCPDS files in the JADE XRD software package.


-.- r0*0


)00I,
"o0 Single Phasec
S, (Metallographic)
S)0 '. A DTA
i a o icroprobe
)0 Liquid
i %
i I
*
)0 ;
)0
)0 \
t S


0 1")\ \\ I


)0 10* \ 3t28^ a "1^
\0 w In .i ^_


_,+n
01 ; F-.^c


0 10
Re








43

epilayers on Si substrates. Films were exposed to 30 minute anneals at 700, 800, 900,

1000, or 1100 C in a vacuum (<5x107 Torr) tube furnace. MeV He" backscattering

spectrometry, X-ray Diffraction (XRD), Secondary Ion Mass Spectrometry (SIMS), and

Transmission Electron Microscopy (TEM) showed that there was no evidence of reaction

between Re and SiC for any heat treatment and the interface remained distinct and highly

ordered. A growth in Re grain size and a fibrous texture after annealing was reported.

The results supported preliminary calculated thermodynamic data. Figure 15

shows the ternary phase diagram expected for Si-C-Re at 1100 C. Coupled with the

heats of formation presented in Table 8, these calculations clearly show that the Re-SiC

tie line (no reaction) was favored. Electrical characterization was not performed.


T = 1100 C

X51/ ReSi2SC
X/ -- SIC
Re5SiS/ \

Re C
XT
Figure 15. Isothermal section of the Re-Si-C phase diagram calculated for 1100 C.96

Table 8. Enthalpy and entropy of formation at 298 K, AHf,298 and S0298 for different
species of Re-Si-C system.96
Species AHff298 S298
(kJ/g atom) (J/K g atom)
Re 0 36.5263
Si 0 18.8196
C 0 5.74
Re5Si3 -19.66 31.98
ReSib -20.56 27.68
ReSi2 -30.12 24.68
SiC -36.61 8.30








44

The same authors in 1995 reported values for specific contact resistance of Re, Pt,

and Ta films on n-type 13-SiC. They found that as-deposited Re contacts on

unintentionally doped n-type P3-SiC were ohmic with a specific contact resistance of 4 x

10-4 Qcm2. After annealing at 500 C for 30 min, the specific contact resistance

decreased, however annealing for 30 min at 900 C made the contact non-ohmic. When

nitrogen implanted SiC was annealed with similar parameters, the as-deposited specific

contact resistance was 1 x 10-4 Lcm2 and improved to 1 x 10-5 jcm2 upon annealing to

900 C. They conclude that Re did not react with SiC up to 1100 C, leaving the changes

in electrical character unexplained.97
98
In 1997, Kennou et al.8 reported the first studies of Re contacts on n-type

6H-SiC. Up to 2 nm thin films of Re were electron beam evaporated from a degassed tip

in 5 x 1010 mbar vacuum onto n-type (3 x 1017 to 3 x 1018 cm-3) 3 to 4 off axis grown

6H-SiC polar (0001) C and Si faced wafers. Films were characterized with X-ray

photoelectron spectroscopy (XPS), low energy electron diffraction (LEED), and work

function measurements. Regardless of C or Si-face, the as-deposited contacts exhibited

Schottky barriers of 0.7 +/- 0.2eV. Samples were annealed up to 1100 K for 2 minutes.

While no chemical changes were confirmed and the contacts remained Schottky with no

change in the barrier height, minor shifts in the C Is and Si 2p XPS peaks indicated

possible free surface carbon and formation of free Si or traces of ReSi2.

In 2000, Shalish and Shapira99 performed the most recent electrical

characterization of Re contacts to SiC. 100 nm Re films were sputter deposited onto

6H-SiC and studied by backscattering spectrometry and I-V measurements. The contacts

were reported to be rectifying as-deposited with a barrier height of 0.71 eV. The barrier








45

increased to 1.04 eV after a 120 min anneal at 700 C. An additional 120 min anneal did

not indicate further change. Annealing beyond 900 C caused instability in the barrier

although reaction cannot be confirmed to be the cause. All depth profiling revealed

stable interfaces regardless of heat treatment.

No additional empirical characterization of Re contact properties has been

published. In the same year, however, Bryant and Bozack studied growth modes of Re

and Nb on the polar faces of 4H-SiC. A single monolayer forms followed by additional

simultaneous monolayers (MSM mode) on the Si-face while layer by layer (Frank-van

der Merwe, FM mode) growth is demonstrated on the C-face.' This was the last series

of experimental work that has been published for Re-SiC structures.

In 2001, Wiff et al.101 completed a quantum mechanical CASTEP modeling of Re

and its reaction kinetics with a silicon-rich 6H-SiC surface. The modeling predicted Re

would bond with the silicon atoms in the overlayer but would not interact with underlying

carbon atoms. The quantum modeling predicted s-bonding only with a total bond energy

of 846 kJ/mol. When applied to a carbon-rich surface, the Re would only bond with the

underlying Si atoms in the matrix. This prediction is consistent with thermodynamic data

97
earlier presented by Chen et al.97 The work presented within this dissertation was

conducted simultaneously but without knowledge of the modeling investigation being

conducted by Wiff et al.'01 As will be presented, empirical data collected in this study

verifies the accuracy of the predicted models. Subsequently, supporting results from this

study were in part published with that work.

With the exception of studies by Shalish and Shapira, empirical results to date

have concentrated on thin-film <400 angstrom interfacial studies on stoichiometric








46

surfaces of various SiC polytypes in as-deposited and relatively short term <30 min heat

treatment conditions to no more than 1100 C (only 830 C for n-type 6H-SiC). The

conclusions for these reports are in conflict. On moderately doped 13-SiC, as-deposited

contacts appeared ohmic with degraded behavior after heat treatment. Highly doped

samples, however, experienced improved ohmic behavior under similar conditions. The

only electrical characterization of Re on 6H-SiC indicates rectifying behavior

as-deposited and after heat treatment on moderately doped samples. Both researchers

report no reaction at the interface while electrical character changes with temperature and

XPS data indicates minute changes in surface chemistry. Clearly, the mechanism for

contact formation and reaction is poorly understood.

To date, little research has been performed on Re contacts -1000 angstroms or

more thick. The chemistry and mechanisms of the contact are poorly understood.

Finally, durability after long-term exposure to elevated temperature has not been

investigated. The study within this dissertation takes a detailed approach to investigating

thick Re contacts on moderately doped n-type 6H-SiC. Carbon-rich, silicon-rich, and

stoichiometric SiC surfaces were generated to provide ideal reaction zone conditions to

investigate preferential formation (if any) of carbides, silicides, or both at room

(as-deposited) and elevated (1000 C) temperatures. Heat treatment of 120 minutes has

been selected to test durability of contact properties and encourage reaction (if any) to

progress further than previously reported in the current literature. Contacts were

characterized by physical morphology, texture, chemistry, and electrical character in both

as-deposited and heat-treated conditions as a function surface preparation.
















CHAPTER 3
EXPERIMENTAL PROCEDURES

3.1 Electron Stimulated Oxidation

An n-type 6H-SiC wafer 34.9 0.5 mm in diameter and 0.33 0.13 mm thick

doped with nitrogen (-1018 cm3) was purchased from Cree Research Incorporated. The

SiC wafer was cut 3.5 0.5 off the {0001 } axis toward (lo20) 10 on a Si surface

and polished on the silicon terminated face using patented techniques (Cree Research).

The wafer was sectioned with a diamond saw into -3.0 x 1.0 cm strips. These strips were

further diamond scribed into -1.0 cm2 pieces for analysis.

Experiments were performed in a stainless steel chamber using a PHI Model

25-270AR DPCMA (double pass cylindrical mirror analyzer) system. Data were

collected in the dN(E)/dE mode. Depending on conditions of interest, the UHV chamber

achieved a pressure of between 2.2 x 10-7 Torr unbaked and 3.8 x 1010 Torr after bake

out. Residual gas analysis (RGA) before and after AES analysis was performed using a

UTI Model 100C quadrupole mass spectrometer. The RGA filament was off during

oxidation experiments.

Samples were introduced into the Auger electron spectrometer (AES) chamber

after CO2 snow cleaning102 and exposure to a compressed air stream. Auger survey scans

verified that the samples had native oxide and surface carbon contamination. Additional

cleaning was performed by in-situ sputtering. The ion gun was operated at 1.5 keV with

-1100 nA of Ar ion current with a stationary beam (-2 mm2) striking the sample at an








48

angle of -60 degrees with respect to the surface normal. Sputtering was continued until

the oxygen signal disappeared and Si and C signals were stable. No heating was

performed after sputtering.

Clean surfaces were irradiated by a continuous primary electron beam. The Si,

C, and 0 KLL transitions were monitored in 60 eV windows for the duration of each

experiment. Primary beam energies and currents ranged from 3 to 6 keV and from 25 to

500 nA respectively with a beam size of 0.008 mm2. Data were collected while rastering

the primary electron beam.

Secondary electron yield was measured for sputter cleaned SiC, SiC oxidized by

dipping in 65% HN03 for 30 min followed by a DI water rinse, and a pure sputter

cleaned polycrystalline molybdenum surface. Yields were calculated from the area under

the electron energy curve in the N(E) pulse count mode. Samples were biased by -20 V

to shift the energy of the true secondary electrons to a value above zero kinetic energy.

Since the majority of secondary electrons are emitted with energies in the 0 to 50 eV

range, the area under the curve from 18.0 to 68.0 eV was determined and equated to the

secondary electron yield. Yield was determined by irradiating samples with primary

energies from 600 to 5000 eV with a current of 10 nA.

3.2 Re Contact Deposition

3.2.1 Sample Preparation

Sample preparation for each of the three SiC wafers used in these experiments

followed the same procedures previously described for ESO studies with the following

exceptions: the 6H-SiC wafers (nitrogen doped, 1.28 x 1018 cm-3) were








49

chemical-mechanical polished and the samples were scribed into strips approximately 0.5

x 2.0 cm.

Samples were bathed in a buffered aqueous 49% Buffered HF solution for 10

minutes. When removed from the acid bath, samples were rinsed with de-ionized water.

Each sample was then flushed with acetone to remove organic solvents and dried with a

pure N2 stream. Samples were placed into a holder constructed of tungsten blocks,

tantalum foil, threaded molybdenum rods and nuts, and a Macor'" support (Figure 16).


fl]Macor
SHFragment SiC
THeat Sink Mask Guide
SiC

DMo


Leads Ow f
PadC B
TLM Mask
Figure 16. In-situ Macor' sample support holder. The leads and mask were made of
tungsten and the heat sinks were tantalum.

For each experiment, the SiC sample was cleaved after preparation to form two

samples denoted as "main" and "fragment". These samples were placed in the holder as

shown in Figure 16. Because AES depth profiling is destructive, it was important to have

a sample for pre-anneal AES survey and depth profiling analysis only. Since the

fragment originated from the main piece, it was exposed to identical ex-situ treatment.

In-situ, the fragment could not be flash annealed like the main piece because only one

sample could make electrical connectivity to both leads. The fragment sample, however,








50

was only used for pre-anneal AES survey and depth profile characterization. The main

sample was use for all other pre-anneal and post-anneal characterization.

The molybdenum rods were used as leads for resistive heating of the sample. The

Ta mask guide was used to ensure flat uniform contact between the mask and the sample.

The addition of Ta metal to the right side required a similar addition to the left in order to

keep uniform resistive heating. The holder was attached to a rotary feed through that

allowed for single axis rotation into and out of the metal deposition flux. Contact

patterns were designed to use the Transmission Line Model, or Transfer Length Method

(TLM). The TLM shadow mask was attached to an additional insertion rod on the feed

through allowing for positioning and removal of the mask in-situ. The mask had three

slits labled A, B, and C, as shown in Figure 16. The resultant Re contact pattern after a

deposition through the TLM shadow mask is shown in Figure 17. The distance between

contacts A and B was 910 jtm, between B and C was 2170 gtm, and between C and the

fully exposed SiC contact, Pad, was 3400 gim.


4j1 m 430m 460m "_
910 E 2170 3400


A B C Pad

Figure 17. Re pattern after deposition through the TLM shadow mask. Contact regions
are labeled A, B, C, and Pad. Width for each contact and spacing are shown in
micrometers.

After sample mounting, the mask was positioned over the SiC and the sample

manipulator was attached to the vacuum chamber. The system was then evacuated using

a removable turbo pumping station to a pressure of approximately 5x 10-4 Torr. The








51

system was isolated and opened to the ion pump for further evacuation. The system was

allowed to reach a pressure of <5x 10-6 Tonrr before baking into the ion pump. Windows

were covered with aluminum foil to prevent cracking due to non-uniform heating. A heat

shroud around the ion pump was turned on and the chamber was surrounded by four heat

tapes and covered with an insulating blanket to achieve a temperature of 180 C for

approximately 24 hours. The bake was continued until a pressure of less than 3x107 Torr

was achieved at temperature (~ 72 hours). All heat sources were then turned off

simultaneously and the resultant pressure after cooling was <8xl0'10 Torr. Titanium

sublimation pumps helped further reduce the total chamber pressure.

Deposition source materials were prepared and introduced before baking the

chamber. These included Re, Si, and Au. The Re source was > 99.99% pure slugs from

Alpha Aesar. A carbon crucible was used for holding the material. The first depositions

of Re were contaminated with C from the crucible when the slugs were electron beam

heated for deposition. 99.99% pure Re foil 0.25 mm thick was molded to fit inside and

around the edges of the carbon crucible to prevent evaporation of C. This eliminated

contamination of the Re contacts.

The Re slug was outgassed prior to deposition on the sample using a Thermionics

electron beam evaporation system powered by a 3kW constant voltage Perma Beam

150-0030 electron gun. The crucibles were cooled by contact with a water-chilled hearth.

An Inficon XTC Quartz Crystal Oscillator was set to 21.03 g/cm3 with a z-ratio of 0.15 to

monitor the Re flux. Carbon monoxide (28 AMU) signals were monitored with a

Spectromass Selector residual gas analyzer to indicate when out gassing was completed.

The electron beam power was incrementally raised and the CO signal was allowed to










pass through a maximum and return to near zero conditions before subsequent increases

of power. When the CO signal did not change with increases in power and a flux of 5

A/min or greater was indicated on the oscillator, the outgassing of the slug was

considered complete. The power was returned to zero and the slug allowed to cool.

After this treatment, the Re source was ready for deposition.

To test the properties of Re/SiC contacts with a silicon-rich interface, Si was

deposited onto the SiC from a crushed (in a mortar with a pestle) high purity silicon

wafer loaded into a carbon crucible. The crystal oscillator control was set to 2.33 g/cm3

with a z-ratio of 0.712 to monitor the Si flux. The Si powder was heated with small

incremental steps in power with the electron beam rastered to avoid localized heating and

evaporation from one region. The Si mass was allowed to melt slightly and then cool to

reform a more solid Si source. This procedure was performed several times until a

constant flux of Si (1.5 A/min or greater) was measured and the region under the electron

beam appeared uniform and solid when cooled.

A thin film of Au was initially used to cap the Re after deposition on SiC. This

step prevented oxidation and contamination after removal from the deposition chamber

and prior to characterization. This practice was discontinued with later experiments

because the focus of this study was Re contact performance in atmospheric conditions

with exposure to elevated temperatures. Addition of Au introduced an unnecessary

variable for experiments where high-temperature heat treatment was involved. The Au

source was prepared by placing several Au ingots into a carbon crucible in the deposition

chamber. The ingots were melted completely to form one solid slug of Au. No

outgassing was required as evident from RGA data.











3.3.2 Deposition

3.3.2.1 General procedure

To ensure consistent spatial orientation, the samples were always placed at 2.2 (a
T.
permanent measurement) on the insertion feed through with the Macorr side of the

sample holder facing the deposition source (SiC sample facing upward). The mask was

removed and the sample was flash heated to approximately 700 C for less than 30

seconds to remove volatile compounds. After cooling, the mask was returned to its

position over the sample.

The Perma Beam electron evaporation source was turned on, with minimal

rastering, to heat the Re. While each deposition was different, in general, a constant Re

flux was achieved when the current reached 125 to 140 mA. The power was increased up

to this value over a period of approximately 20 min to ensure uniform heating and flux.

Near 120 mA, the quartz crystal oscillator reading was erratic, but it gradually achieved a

constant deposition rate of 0.2 A/sec. When this rate was constant for several minutes,

the sample was rotated into the flux of Re and the oscillator total deposition reset to track

the thickness of the contact.

During each deposition, chamber pressure, power setting, current, and deposition

rate were monitored. In most cases, the current and the deposition rate would slowly

increase. It was necessary to monitor the rate increase to avoid "spitting" and

non-uniform deposition. When the rate reached 0.4 A/sec, the power was reduced to a

rate of 0.2 A/sec. Rates faster than 0.4 A/sec over sustained periods of time generally

resulted in non-uniform films. When the total deposition thickness reached 900 to 950 A,

the sample was rotated out of the Re flux and the power quickly reduced. Metal







54

deposition after this point was considered to be negligible because the source material

temperature was quickly reduced. Heat radiating from the deposition source during

deposition, however, affected the crystal oscillator output. Therefore, when all in-situ

systems were cooled, another reading was taken and recorded as the final thickness of the

film. The mask was withdrawn from the sample in-situ to avoid scratching of the film

when the feed through was removed from the chamber. The sample was then removed

from the stage and the as-deposited condition characterized using X-ray Diffraction,

Dektak profilometry, Auger Electron Spectroscopy, and current/voltage measurements.

After characterization was completed, each sample was placed into another

sample heater system like the one shown in Figure 16 and again placed in vacuum.

When the pressure was lower than lxl0-6 Torr, the sample was ramped to 1000 C within

1 minute by resistive heating. The temperature was monitored with a two-color optical

pyrometer to ensure uniform/constant heating for 120 minutes. The sample was then

returned to room temperature over a 1 minute cool down. Samples were removed from

vacuum and once again characterized using X-ray Diffraction, Dektak profilometry,

Auger electron spectroscopy, and I-V measurements.

The process described is the generic procedure for depositing Re on a

stoichiometric SiC sample. Because AES depth profiling was destructive, it was

important for the fragment piece of SiC to be added to the sample heater system as shown

in Figure 16. It was placed above the main piece already depicted, and was held in place

by only one of the resistive leads. These samples were subjected to all treatment

encountered by the main SiC piece with the exception of the in-situ flash anneal. After







55

deposition, these pieces were used only for AES survey scans and depth profiling for the

as-deposited condition. All other characterization was performed on the main piece.

3.3.2.2 Specialized conditions

Specialized procedures were required to create non-stoichiometric SiC surfaces

for study. Graphitized and silicon-rich surfaces were of interest to see if Re had a

preferential reaction for a constituent of SiC rather than the stoichiometric surface.

Reasonably thick (50 to 100,A) films were desired to ensure reaction, if any, would be

predominantly with the surface layer rather than the SiC underlayer. Procedures for

preparing these surfaces follow.

L. Muehlhoff and others report that heat treatments below 900 K (-630 C) do not

change the SiC surface.52 Between 900 and 1300 K (-630 to 1030 C) both the Si and C

faces of 6H-SiC become graphite terminated.52 While the C face graphitizes faster,

above 1300 K both C and Si faces become heavily graphitized due to Si sublimation.

This procedure was used to graphitize SiC surfaces. In the flash heating step of the

procedure above, the sample was flash heated to 1300 C (270 C higher than in

Muehlhoff's experiments) rather than only to 700 C. The mask was retracted during this

heating process so the SiC was only in contact with the W and Ta leads. The temperature

was monitored with a dual wavelength optical pyrometer and held constant for 10 min to

ensure graphitization. After cooling of the sample, the mask was returned and the

procedure for Re deposition was followed as previously described. A few samples

prepared with this technique were also heated at 1000 C for 120 min in-situ after Re

deposition and then capped with Au before being removed from the chamber. These







56

particular samples will be described in further detail in the results and discussion

sections.

In other experiments, a silicon-rich surface was created by depositing a thin Si

layer on the stoichiometric SiC surface prior to Re deposition, as described above. The

mask was withdrawn and the substrate flash heated as previously described to -700 C

for a period of less than 30 sec. After cooling, the mask remained withdrawn and the Si

film was evaporated from an electron beam heated Si source. It was often necessary to

slowly increase the power to maintain an appropriate flux rate. When the total deposition

thickness reached 50 A or 100 A, depending on the sample, the stage was rotated out of

the flux and the electron beam reduced to zero power. The sample was allowed to cool

and then the mask was returned to position on top of the sample and the generic

procedure for Re deposition was followed, as previously described.

3.3.3 Characterization

3.3.3.1 Dektak stylus profilometry

Contact thickness was measured in-situ using a quartz crystal oscillator. This

value was verified ex-situ using a Veeco Instruments Inc. Sloan Dektak IIA. The 90

diamond tip stylus with a radius of 10 nm traced from the "Pad" region to "A" at a

medium speed as shown in Figure 18. Feature height, contact thickness, and spacing

were recorded. Measurements were made on as-deposited samples and again after heat

treatment.

F

Contact C B A

Figure 18. Dektak procedure.









3.3.3.2 X-Ray diffraction

Glancing angle XRD103 was used to analyze the structure of the contacts

as-deposited and after heating. The samples were mounted on an aluminum disk with

double-sided tape. In order to ensure the sample was flush with the holder, it was pressed

against a glass slide and mounted with setscrews as shown in Figure 19. The "Pad"

region of the sample was centered on the Al plate to ensure maximum irradiation of a Re

covered surface with the glancing X-ray beam. After mounting, the sample was placed

on a continuously rotating stage.


Sample Mounting Sample Mounting Apled Pressure
(Viewed from top) (Viewed from side) le pressure

Set Icrew




Glass Slide

Figure 19. Procedure for XRD sample setup. The glass slide was use to ensure the
sample was flush with the mount.

The sample was irradiated with X-rays produced from a Cu cathode

(Xave=1.541871A) in a Rigaku Geigerflex. The tube voltage and emission current were

40 kV and 25 mA, respectively. The system was run in continuous scanning 20 mode

from 10 to 90 with 0.1 increments at a scan speed of 6 seconds per step or 1 degree per

minute. The glancing angle was chosen by maximizing the intensity of the reflectance

signal from a Re film at a set 20. After performing this step, a 6 angle was chosen for

the remainder of all experiments. Data were collected in counts/second. Re and SiC









powder samples were used as standards in conjunction with spectra defined in the JADE

analysis software package.

3.3.3.3 Electrical measurements

TLM42,38104,105 electrical measurements were made using a Keithley 196 System

DMM and a Keithley 228A Voltage/Current source. A four point probe configuration

was chosen to eliminate error contributed from the total contact resistance, Re, between

the probe and the thin film.38 The measurement orientation is shown in Figure 20 and

data were collected by incrementally ramping current from zero to maximum positive and

negative values based on a limiting source voltage of +/- 10 V. Initial I-V data were

collected between "A" and "B", "B" and "C", "A" and "C", and finally "C" and the

"Pad". In order to get more valuable data from each deposition, the measurements were

Current Source
Voltage Limited








Voltage Measured

Figure 20. Measurement set up for 4-point probe I-V curves.

later extended to include: "AB", "AC", "APad", "BC", "BPad", and "CPad". These

orientations cover all possible measurements with the given TLM pattern and allow for a

more accurate correction for the intrinsic resistance of the SiC. Data were graphed in the

form of I-V curves and a total resistance for each measurement was extrapolated from a

linear best fit. Total resistance was plotted versus probe spacing and the total contact

resistance, Re, determined by the y-intercept divided by 2 as shown in Figure 11.38105







59

This value corresponds to the total contact resistance of the contact at zero spacing,

eliminating the intrinsic contribution of SiC. The specific contact resistance, Pc, was

calculated using Equation 10 with the transfer length and Ps as described in Section 2.4.

The x-intercept is equal to 2 times the transfer length when plotted like Figure 11 and Ps

was calculated from the slope divided by the width of the Re contacts. This procedure

was followed for both as-deposited and the heat-treated samples.

3.3.3.4 Auger electron spectroscopy

Auger Electron Spectroscopy (AES)6'15'106 was used to determine surface

composition after each deposition and after heat treatment. Data for the as-deposited

samples on stoichiometric and silicon-terminated surfaces were taken from the additional

piece of SiC described at the end of Section 3.3.2.1. Data for as-deposited contacts on

graphitized samples and from heated contacts for all three surface conditions were taken

from the main SiC sample. Surveys were collected using a 500 nA, 5keV electron beam

(0.008 mm2) rastered over an area of 0.001 mm2 using a Phi Model 1 1-045A Electron

Gun and a Phi Model 20-085 High Voltage Supply and the DPCMA. The primary

electron beam current density was 6.25 mA/cm2. Depth profiling was accomplished with

2keV Ar' ions emitted from a differentially pumped Phi 04-303 ion gun. The ion gun

emission current was 20 mA and the gas pressure ranged from 2x 105 Torr to 5x 10-4 Torr

for variable non-rastered point sputter rates. Sputter rate was assumed to be constant

throughout each analysis and the data are reported as Angstroms removed. All data were

collected with small/small aperture settings using a Phi Model 25-270AR DPCM

analyzer. Si (LMM-92eV), C (KLL-272eV), Re (MNN-1799eV), and 0 (KLL-512eV)

signals were plotted for each depth profile.















CHAPTER 4
RESULTS AND DISCUSSION

4.1 Electron Stimulated Oxidation (ESO)

4.1.1 Results

SiC surfaces remained oxygen free, as detected by AES, for as long as 17 hours in

a UHV chamber with a background pressure of 2.6 x 10-8 Torr or less if filaments

remained off and if the electron beam for AES was not operated continuously. Exposure

to an electron beam under the same conditions stimulated growth of surface oxides in less

than 25 minutes. Figure 21 shows data that are representative of the surface chemistry


45 -
40 -
35 1
30 -
25

20 -


15 -
10


-Carbon (KLL)


Oxygen (KLL)


--Silicon (KLL)


U I
0 1000 2000 3000 4000 5000 600C

Beam Irradiation Time (sec)

Figure 21. Typical AES signal intensity measured from the maximum peak to the
minimum peak in the d(E*N)/dE mode for Si (KLL), C(KLL), and O(KLL) as a function
of beam irradiation time. At T=0, the excess carbon is residual contamination prior to
insertion into vacuum. After sputtering at T=1000 sec, the C and Si signals remain
stoichiometric and 0 signals rise from noise to near saturation in less than 2000 seconds
(-30 min).


>N

z
-; *
4 1
iTi <1


)







61

concentrations observed under an electron beam versus irradiation time. The Si (KLL), C

(KLL) and 0 (KLL) peak intensities are shown. The oxygen signal increases with

irradiation after sputtering. The Si and C signals decrease as the oxygen overlayer grows.

Figure 22 shows representative AES signals for the Si (KLL) and C (KLL) transitions a)

before sputtering, b) immediately after sputtering prior to oxidation, and c) after

1 eV Shift


4 2
3 1
iz 2 0 -


2- i 0 -
-a)-3a
o -2 b 0 -, b)
O -3 c) -5 c)
-6t
-4 -6 1

1589 1609 1629 1649 242 262 282 302
Energy (eV) Energy (eV)

Figure 22. Typical AES signals for the Si (KLL) and C (KLL) transitions, a) Before
sputtering; b) Immediately after sputtering before oxidation; c) After saturation of the
oxygen signal.

saturation of the oxygen signal. Before sputtering, the C signal does not indicate carbide

from the matrix, which is consistent with surface layers of adventitiouss'

carbon-containing contaminants.107 After sputtering, the carbon signal becomes carbidic

consistent with the SiC matrix.107 The signal is reduced in intensity but unchanged in

shape with growth of oxide. The Si (KLL) signal shows an equivalent reduction as the

oxygen signal grows. An approximate 1.0 eV shift to lower energy is observed for the Si

(KLL) transition. For pure Si, the KLL transition occurs at 1616 eV. When Si reacts







62

with oxygen to form SiO2, this signal shifts lower to 1608 eV.108 The observed leV shift

indicates that complete reaction to form SiO2 has not taken place and the surface is

characteristic of SiO, which may be a precursor to oxide formation.

Variables affecting electron stimulated oxidation were studied. The effect of

background gas pressure was investigated with a constant electron beam energy (5 keV)

and current (300 nA). The current density, calculated from current and the beam

diameter, was 3.75 mA/cm2. Residual gas analysis confirmed that H20, CO, and CO2

were the prominent gas species in vacuum after H2. Relative concentrations of H20, CO

and CO2 for 3.8 x 10 10 < Presidual < 1.6 x 10-8 Torr

Table 9 Analysis of residual gases in the vacuum chamber during ESO and the associated
relative oxidation rates represented by dIo/dt reported for each pressure condition. H20,
CO, and CO2 are the primary constituents.
Total Chamber H20 CO CO2 dlo/dt
Pressure (Torr) (Arb. Units) (Arb. Units) (Arb. Units) (Arb Units/sec)
1.6 x 108 72.8 35.8 38.6 0.0103
7.1 x 10' 30.4 26 16.7 0.0053
5.5 x 10-9 26 30.5 22 0.0026
2.7 x 10-9 9.5 23 16 0.0013
6.8 x 10-0 4 5 3 0.0005
3.8 x 10-'0 5.3 3.6 2.7 0.0002

are shown in Table 9. The values reported are relative readings for peak intensity from

residual gas analysis and may be compared along the column, but not between columns

since the data are not corrected for the spectrometer sensitivity change with mass. The

values for H2 are not shown because in each chamber condition, the spectrometer reading

was saturated at m/e = 2 AMU. Figure 23 shows the oxygen Auger intensity versus

electron beam exposure time for each pressure studied. The slope of these curves, dlo/dt,

at time = 0 represents the instantaneous change in concentration of oxygen under the










30 min


k/V^
[Af^
,/^
/w
AV
/ 7.1 x
/" AA^


60 min


1.6x 108 Torr


10-9 Torr


5.5 x 10-9 Torr


30 min 60 min
0 1000 2000 3000 4000 5000
Beam Irradiation Time (seconds)

Figure 23. Auger oxygen (KLL) signal intensity (Io) for various chamber pressures as a
function of electron beam irradiation time. The electron beam energy, Ep, was 5 keV and
the current was 300 nA (current density = 3.75 mA/cm2) for each experiment.


0.012

0.01

0.008

0.006 |

0.004 2

0.002


0 -7------
1.0 x 10 5.0 x 10-9


/X dlo/dt = 730000 Pressure

R2 = 0.98





1.Ox 108 1.5x10.8 2.0 x 10-8


Pressure (Torr)

Figure 24. Change in oxygen AES intensity (dIo/dt) at time = 0 as a function of total
residual chamber pressure (Torr).


10

M 8-
z2
* 6-
-o
'1 4-


72
0
>-o
0"
Uas)
i2) Cn



e 0

>1
~I


C)






64

electron beam and is proportional to the initial oxidation rate. These rates are plotted

versus total system pressure in Figure 24 and a positive linear dependence is observed. A

linear dependence is not, however observed when these rates are compared to any one of

the particular residual gas species as shown in Figure 25.

The effects of primary voltage, Ep, on ESO were studied. At a chamber pressure

of 1.6 x 108 Torr and current of 300 nA (3.75 mA/cm2), the quantity dIo/dt (again

assumed to be proportional to oxidation rate) exhibited a moderately linear decrease as Ep

was increased from 3 to 6 keV (see Figure 26).


Z2
o 0.01
00 /7
S0.0081
*d 0.006 /

0
U i 0.004 -- H20
-u-CO
=t 0.002 i cO2

0
0 20 40 60
RGA Concentration (Arbitrary Units)
Figure 25. Change in oxygen AES intensity (dIo/dt) at time = 0 as a function of
concentration of H20, CO, and CO2 (RGA Arbitrary Units). Notice the lack of a linear
trend between oxidation rate and specific residual gas components.

Finally, oxidation rate's dependence on primary beam current, Ip, was studied.

Data were collected at different pressures (2.2 x 107 Torr and 7.1 x 10-9 Torr) for a range

of currents from 25 to 500 nA (current density = 0.31 to 6.25 mA/cm2). A linear

dependence of oxidation rate was previously reported for pressure so the data was









-7
0
0




0 0



U0.

U


dlo/dt = -0.0008 Ep + 0.017


R2 = 0.62





f


Electron Beam Energy, Ep (eV)
Figure 26. Change in oxygen AES intensity (dlo/dt) at time = 0 as a function of Ep
(keV). Ip was 300 nA for each experiment. Notice the decrease of instantaneous
oxidation rate as electron energy is increased.


0.01
0.009
0.008
0.007
0.006
0.005
0.004
0.003
0.002
0.001
0


* 2.2x 10-7 Torr
07.1 x 109 Torr


dlo/dt = 0.0017 ln(Ip) 0.0043

R2 = 0.82


-1 -- -1


0 100 200 300 400 500 600

Electron Beam Current, Ip (nA)


Figure 27. Change in oxygen AES intensity (dlo/dt) at time = 0 as a function of Ip, (nA).
Data collected at 2.2 x 107 Torr or at 7.1 x 10.9 Torr. Ep was 5 keV for each experiment.


0.015


0.014


0.013


0.012






66

normalized to a common pressure (7.1 x 10-9 Torr) and graphed vs. Ip (see Figure 24).

dIo/dt has a negative exponential relationship with increasing current showing reduced

responses above 200 nA regardless of pressure (Figure 27).

Effect of oxidation on secondary electron yield was investigated by totaling the

counts from AES N(E) spectra of SiC and oxidized SiC compared with a Mo surface for

control. Secondary electron yield was measured as a function of primary beam voltage

between 600 and 5,000 eV with a constant current of 10 nA (Figure 28). Data for


1.2 M
Modybdefium
S1. OM
^ ^419SC Oxidized

S0.8 M
W o.6M V _
'Unw
LL 0.6 M 'S'C 0^

S0.4 M :A^-
co0.2 M
0.0 M
500 1500 2500 3500 4500 5500
Electron Beam Energy, Ep (eV)
Figure 28. Secondary electron current yield as a function of Ep (eV) for molybdenum,
SiC and oxidized SiC. The ordinate is total counts collected in leV bin widths for the
secondary signal from 18 to 68 eV with a -20V bias on the sample, Ip, = 10 nA and
counts summed over three 5 second sweeps.
SiCwere collected from a sample sputter cleaned in-situ, while data for oxidized SiC

were collected from samples which had been dipped in 65% HNO3 for 30 min and rinsed

with DI water. Auger analysis confirmed a SiO2 layer on this SiC surface. Data in

Figure 28 show that the secondary yield decreased in all cases with increasing Ep. For

the conditions investigated, Mo had the highest electron yield, followed by SiC, and then

oxidized SiC.










4.1.2 Discussion

Irradiating solid surfaces with moderate to high-energy electron beams results in

many processes. Some of these include electronic excitation of surface species;

adsorption, dissociation and/or desorption of gas species; bond breaking; diffusion; and

charge build-up.56109 With the exception of CH4 and H2, oxygen containing gases (H20,

CO, and CO2) make up the major fraction of the gas phase in our experimental vacuum

system as indicated by RGA analysis. The electron beam may excite or dissociate

gaseous or absorbed species causing them to interact more strongly or rapidly with the

irradiated surface. Adsorbed or gas-phase molecules which contain oxygen can be

dissociated directly by the primary beam and/or by secondary electrons reaching the

surface from within the bulk material. Both primary and secondary electrons have been

shown to lead to molecular dissociation.56 Atomic oxygen liberated from these

dissociative processes can then rapidly and easily oxidize the substrate.

Electron-beam stimulated oxidation depends on several factors. These include:

concentration of oxygen-containing species under the electron beam, probability of

dissociation into a reactive species, diffusion over the surface or to the oxide/substrate

interface, and the reaction rate constant between oxygen and silicon to form SiO2.

Oxidation is enhanced by any mechanism that aids dissociation of absorbed species on

the substrate surface.56,51 In previous experiments on Si substrates, the

electron-stimulated oxidation rate was controlled by maintaining a known pressure of

oxygen-containing gas in the analysis chamber.55'56'5751 In the present case on SiC,

ambient residual gases contribute to oxidation. No reagent gases were leaked into the

system.







68

Because the Gibbs Free Energy of SiO2 (-203.33 kcal/mol) is considerably lower

than that of SiC (-14.4 kcal/mol), oxidation of the Si in the matrix is favored. Oxidation

rate, however, is expected to be a diffusion-limited process, as dissociated atomic oxygen

must diffuse through the growing oxide layer to the SiC/oxide interface.1 0 Rate limiting

was observed in Figure 23 by the decline in reaction rate (saturation of the oxygen signal)

for extended electron beam exposure times. The AES (KLL) signal monitored for Si

should shift in energy from pure Si (1616 eV) to oxidized Si (1608 eV) if complete

reaction to form SiO2 was observed.108 The minor energy shift (- 1.0 eV) of the AES

signal in Figure 22 indicates that a mixture of Si oxides or sub-oxides may be forming

rather than pure SiO2. Because the Si signal is attenuated similarly to the C signal, from

the oxygen overlayer, the conclusion that reaction has not preceded to completion and

carbon is not removed from the matrix to the gas phase as CO or CO2 is valid.

4.1.2.1 Dependence on chamber pressure

The oxygen Auger intensity (Io) versus time is shown in Figure 23 for several

total chamber pressures. These curves are well described by an exponential relationship

of the form:

Io = c (l-exp(-13t)) (12)

where (x is the saturation level of the oxygen signal at t=oo, and P3 is related to the rate of

adsorption, dissociation, and/or reaction, whichever is rate limiting. Tougaard et al.55

reported a similar exponential dependence of oxide growth on Si as a function of

adsorption kinetics. In his model, P3 was in the form of SoVP/No where So is the sticking

coefficient of oxygen, V = 3.48 x 1020 molecules/cm2s Torr is the impingement rate of

oxygen per unit Torr, No is the number of sites for a molecule on the substrate surface,







69

and P is the partial pressure of oxygen. Since the background gases (other than H2) in our

system are predominantly oxygenated species and we did not leak reagent gases into the

chamber, oxidation rate should be directly related to the total chamber pressure. The

derivative of Equation 12, cap, represents the instantaneous change in surface oxygen

concentration or growth rate at time zero. Values for c, 3, and their respective errors

were calculated by curve fitting the data in Figure 23 with Equation 12 and the expected

linear trend was observed as shown in Figure 24.

RGA spectra confirmed the relative concentrations of H20, CO, and CO2 for each

experiment, as shown in Table 9. The 28 AMU peak is assumed to be predominately CO

rather than N2 because there was no appreciable peak at 14 AMU for N. While the linear

trend for oxidation rate was observed for total chamber pressure, it does not follow for

any single residual gas species. These data suggest more than one residual gas species is

contributing to the oxidation of the surface. Tougaard, Kirby, and Coad report that both

CO and H20 physiadsorbed on the surface are dissociated by interaction with primary

and/or secondary electrons to form reactive atomic oxygen for oxidation.55'56'57'58

Regardless of the parent species, the rate of oxidation is expected to depend on (i)

the concentration of atomic oxygen on the surface and (ii) the diffusion probability for

oxygen to the oxide/SiC interface for reaction with Si to form SiO,. It is commonly

assumed that reaction of atomic oxygen with Si is very rapid. Coverage of the substrate

surface by oxygen containing compounds would affect step (i). The linear correlation

between overall gas pressure and oxidation rate supports the conclusion that the coverage

of oxygen containing species was the limiting factor in these experiments.









4.1.2.2 Dependence on beam energy

A negative correlation between Ep and oxidation rate is shown in Figure 26.

Kirby et al.56 observed an increasing oxidation rate for Si substrates upon increasing Ep

from 500 to 2500 eV. They suggest higher Ep resulted in greater secondary electron yield

and thus higher dissociation rates for oxidation. The secondary electron yield as a

function of Ep for SiC, shown in Figure 28 clearly indicates a decreasing yield for Ep >

600 eV. This is consistent with secondary electron yields demonstrated for pure elements

that typically increase with respect to Ep at very low energies and then decrease at

energies above I keV.6 Secondary electron yield is material sensitive and thus also

affected by oxidation. Since all oxidation experiments were conducted with 3000 eV

< Ep < 6000 eV, according to Figure 28, secondary electron yield should be constant or

decrease with increasing Ep. Since these electrons have been shown to contribute to the

dissociation of adsorbed species providing atomic oxygen for oxidation, a reduction in

secondary electron yield would correspond to a reduction in oxidation rate. This is

consistent with the data shown in Figure 26. In addition, the cross section for direct

impact dissociation of adsorbed or gas phase gas species should decrease for primary

electrons as the electron energy is increased.

Finally, Kirby, et a156 suggest beams with Ep < 1000 eV will be significantly

attenuated by even one monolayer of oxide due to inelastic scattering by electrons near

the surface. The resulting, near surface, secondary electrons would have a higher

probability of escape to the vacuum without collision. Their data show that the rate of

oxidation increased with energy. In the present study, however, with higher energy

beams (>3 keV) the electrons have a greater probability of penetrating deep below the







71

surface layer. Attenuation of the primary beam is not expected to be a critical factor.

With secondary electron generation well below the surface, recombination before

reaching the surface is likely. The probability of secondary electrons aiding in surface

dissociation is reduced. Figure 28 supports this mechanism. Based on a reduced

cross-section for primary electrons at high energies and secondary electron generation

well below the surface, a decrease in oxidation rate above 3 keV is not surprising.

4.1.2.3 Dependence on beam current

The effects of pressure on oxidation rate are contained in the term 3 from

Equation 12. Holloway et al.' 1 have modeled the data for electron-stimulated oxidation

of ZnS by a number of different gases with a model similar to Equation 12. In their

model, 13 was dominated by current density. Figure 27 shows the instantaneous oxidation

rate at t=0 for various primary beam currents. There is little dependence of oxide growth

on Ip > 300nA. Coad et al.58 report similar results on Si with background pressures of 5 x

10' Torr, Ep = 2keV, and beam currents of 650 nA to 1600 nA. The apparent lack of

dependence upon Ip above 300nA suggests that the current density (3.75 mA/cm2) was

sufficiently large so that oxidation rate is limited by arrival of oxidizing species under the

beam. At a pressure of 2.2 x 10-7 Torr, oxidation rate did increase with increasing current

for Ip < 200nA, suggesting current limited dissociation. Above 200nA, however,

oxidation rate appeared to become independent of current. This observation is consistent

with both the model of Tougaard, et a155 and Holloway, et al1' showing current limited

oxidation with low Ip and impingement rate limited oxidation with moderate to high Ip.







72

4.2 Re Contacts

The conditions necessary for contaminant free in-situ deposition analysis with an

electron beam have been presented in Section 4.1. In this section, results for the

development of a high-temperature Re contact on 6H-SiC are reported. Re was deposited

on fourteen (14) samples with variable SiC surface chemistries (carbon-rich, silicon-rich,

and stoichiometric). The contacts were characterized as-deposited and annealed to study

contact formation as a function of experimentally controlled interface chemistry.

4.2.1 Results

As described in Section 2.2, surface preparation is a key component to SiC

processing and metallization. The surface preparation decision tree shown in Figure 2

shows how to modify the SiC surface for a desired chemistry. To fully understand a

metallization on SiC, it is important to analyze metal reaction to the components of the

SiC matrix. A silicon-rich surface may be desired for metals that form stable silicides at

elevated temperatures such as Ni, Co, and Pd.4'14'92 A carbon-rich surface may be desired

for carbide forming reactions. Finally, a stoichiometric surface may promote a mixture

Table 10. Prepared surface chemistry and conditions analyzed for each sample.
Sample Surface As-Deposited Annealed
1 Carbon X
2 Carbon X
3 Carbon X
4 Carbon X
5 Carbon X
10 Carbon X
14 Carbon X X
6 Stoichiometric X X
7 Stoichiometric X X
8 Stoichiometric X X
9 Stoichiometric X X
11 Silicon X X
12 Silicon X X
13 Silicon X X







73

of species from reaction with both constituents as shown with Ti, Cr, Mo, Fe, Nb, Pt, and

W.4'14'92 Techniques described in Section 3.3.2 have been used to create all three of these

desired configurations to systematically study Re reaction. Table 10 shows the surface

chemistry and conditions analyzed for all 14 samples.

Carbon-rich surfaces were prepared by resistively heating samples to 1300 C in

order to volatilize surface Si prior to Re deposition. Anneal duration and temperature

profiles as determined by a two color optical pyrometer are shown in Figure 29 for

samples exposed to this treatment. The contact scheme after deposition is also depicted


Sample 1 Sample 2 Sample 3 Sample4

1300 C 1300 C 1300 C 1280 C 1305 C 1191
10 min 10 min 10 min 10 min

Image after treatment Image after treatment Image after treatment



Sample 5 Samnle 10 Samle 14

1300 C 1180C 1214C 1310C 1210C 1300 C
20 min 10 min 5 min

Image after treatment



II I Re Contacts :,". SiC Substrate
AB C Pad
Figure 29. Anneal temperature profile (C) and duration (minutes) for each sample
during the graphitization process. The contact pattern that would result after deposition is
shown in black for reference orientation.

for reference orientation. Because resistive heating required sample contact with both

electrical leads to prepare the carbon-rich surface prior to deposition, these samples did







74

not have a sacrificial fragment for pre-anneal AES characterization as depicted in Figure

16. As a result, approximately twice as many samples were prepared so data could be

collected for as-deposited and annealed conditions.

Table 11. Conditions for Re deposition on samples with stoichiometric surfaces.
Sample Pressure E-beam I Time Thickness Notes
(#) (Torr) (mA) (min) (A)


14 C-rich 1.5 x 10-9



6 Stoich. 1.2 x 10-9

7 Stoich.6.4 x 10-10
8 Stoich.6.5 x 10-1

9 Stoich.3.9 x 10-10


11 Si-rich 5.3 x 10 -10
12 Si-rich 4.5 x 10'10

13 Si-rich 7.0 x 10-10


280 to 300
250

230 to 260


190 to 210

170 to 230

120 to 140


120 to 170



150 to 200

170 to 210
180 to 220

165 to 170


145 to 160
125 to 140

125 to 145


1005
999

960


712

1074

1000


1010



997

1029
1009

932


1013
997

994


1 C-rich
2 C-rich

3 C-rich


Silicon-rich Samples 11, 12, and 13 required the deposition silicon onto the

surface of the SiC prior to Re deposition. 50 angstroms of Si were deposited on the

surface of Sample 11 while 100 angstroms were deposited on the surface of Sample 13.


2.7 x 10-9
2.5 x 10-9

1.3 x 10-9


4 C-rich

5 C-rich

10 C-rich


NO ANNEAL
ANNEAL 120 min/1000 C.
Capped with 198 A Au.
NO ANNEAL. 2 phases to
deposition-- Realigned beam on
Re target. Capped with 195 A
of Au.
NO ANNEAL. E-beam went
through Re target to C crucible.
NO ANNEAL. Capped with
188 A of Au.
NO ANNEAL. Contacts
delaminated before
characterization.
NO ANNEAL in-situ. After
characterization Returned to
chamber ANNEAL 120
min/1000 C
Up to air to replace crystal in
oscillator. Baked twice.

Main piece appears cleaner than
fragment
Re "spit"/overheated (261 A/10
min)and a pinhole in Re target
formed (445 A /20 min).
50 A of Si
100 A of Si over pattern region
only on SiC
100 A of Si


2.9 x 10-9

2.4 x 10-9

8.8 x 10-10






75

Sample 12 also had 100 angstroms of Si deposited, however, the shadow mask for Re

patterning was covering the surface during this process so that Si was deposited only

where Re contacts would be formed.

All samples, regardless of surface chemistry were flash annealed, as previously

described in Section 3.3.2.1, immediately prior to Re deposition. Table 11 shows details

for the deposition conditions of each sample.

Carbon-rich surfaces
Sample 2 Sample 14

1000 C 1000 C

Stoichiometric surfaces
Sample 6 Sample 7 Sample 8 Sample 9
II '1 II I11 II 1I ,.--
1000 C 1000 C 943 C 1000 C 819 C 1009 C

Silicon-rich surfaces
Sample 11 Sample 12 Sample 13
I I i'm I I It
975C 1000 C 1030C 1000 C 997C 1007C




I I I Re Contacts '.".,-:. SiC Substrate
AB C Pad
Figure 30. Heating temperature profiles (C) during 120 min anneals for each SiC
surface.

Samples were characterized as-deposited and then resistively annealed in

evacuated chambers with a pressure of not more than 1.0 x 10-6 Torr as described in

Section 3.3.2.1. An optical pyrometer was used to monitor the temperature profile of the










samples during annealing. The temperature profile for each sample is depicted in Figure

30. All samples were fairly uniformly heated at approximately 1000 C except Sample 9

(stoichiometric surface). For this sample, a large temperature gradient (AT = 200C) was

observed from the leads to the middle of the sample.

Physical differences in appearance between samples were noted for both as-deposited and

annealed conditions. Table 12 documents these observations. During deposition of

Sample 4 (carbon-rich), RGA analysis indicated a high presence of C and CO in-situ.

The electron beam was grazing the carbon crucible during depositions on Samples 1 to 4

(carbon-rich) adding additional unintentional carbon compounds to the Re flux.

Crucibles were subsequently coated with Re foil and the C flux reduced to noise as

confirmed by RGA analysis for Samples 5 to 14.

Table 12. Physical appearance of surfaces for each sample as-deposited and annealed.
Sample As-Deposited Annealed
1 C-rich Specular Degraded (uniformly dull/cloudy gray) < Tube furnace
Iday in atmosphere Contacts disappeared
2 C-rich Specular in-situ Specular
No atmosphere exposure
3 C-rich Flaky delaminated regions
4 C-rich Uniformly dull gray/rough
5 C-rich Specular on Pad
Dull gray on A, B, and C
A, B, and C disintegrated during characterization
10 C-rich Flaky delaminated regions
Contacts disintegrate with minimal agitation
14 C-rich Partially Specular in atmosphere < 3hrs Specular
6 Stoich. Specular Degraded (dull gray patches < 3 hours in Specular
atmosphere)
7 Stoich. Same as 6 Specular
8 Stoich. Same as 6 Specular
9 Stoich. Same as 6 Specular
11 Si-rich Specular Degraded (dull gray large patches -24hr) Specular
12 Si-rich Specular- Degraded (dull gray small patches ~24hr) Specular
13 Si-rich Specular Specular







Carbon-rich samples

12
3 AMM 14
5 UN Fragment

Stoichiometric samples

S 6 -mi, r 7
9
"- Fragment r: Fragment

Silicon-rich samples

j ii U~17INI 12
W Fragment .7, Fragment

SReonSi .Fragm13
l Reon Si r Fragment


Figure 31. Visual appearance of annealed contacts after two years of exposure to
atmosphere. The "holes" in each PAD region are from AES depth profiling immediately
after the annealing phase. Films are increasingly specular as a function of interfacial
silicon concentration before the annealing phase. Note carbon-rich Samples 1, 2, 3, and 5
have been polished to remove films and show relative surface graphite concentration.
As-deposited films were more specular and appeared to adhere with more stability
to surfaces that were rich in Si concentration. Films on surfaces rich in carbon were
degraded and usually delaminated or disintegrated with minimal agitation. Sample 5







78

(graphitized for 20 minutes) showed this variance with respect to carbon concentration

the most. The Re deposition appeared specular on the Pad region but was darker gray

and diffuse on contacts A, B, and C. Figure 29 indicates contacts A, B, and C were 120

degrees hotter during graphitization than the Pad. As a result, a thicker graphite layer

should have formed for this region of the specimen. During characterization, contacts A,

B, and C began to disintegrate, however the Pad remained intact. In addition to samples

with intentionally prepared carbon-rich surfaces, the fragments used with stoichiometric

Samples 6, 7, 8, 9 and silicon-rich Samples 11, 12, and 13 exhibited less smooth and

more diffuse surfaces than their counterpart main samples for as-deposited contacts.

These samples were not flash annealed immediately prior to Re deposition and

presumably had physiadsorbed carbon species on the surface.

Contacts that remained intact prior to annealing became specular with heat

treatment. They remained mirror-like throughout characterization after being annealed.

Figure 31 shows these samples as they appear today, 2 years after annealing. The oval

"holes" visible on the PAD regions of the samples are residual from AES depth profiling

during annealed characterization. Carbon-rich Samples 1, 2, 3, and 5 were polished after

characterization and only show graphite concentration across the substrate surface.

Carbon-rich Sample 14 contacts appear diffuse with strong color variation

suggesting reflective interference from a rough surface. Stoichiometric surfaces are

improved, but are only moderately specular across the sample surface. Films on

silicon-rich surfaces, while showing clear signs of surface oxidation (haze), remain

specular with silver finish suggesting smoother surfaces.









Contact thickness, as determined by Dektak, is summarized in Table 13 for each

contact and graphically depicted in Figure 32

Table 13. Contact thickness reported in angstroms (A) by Detak stylus profilometry for
as-deposited and annealed samples. *Sample 2 (carbon-rich) values reflect a 200 A Au
cap after Re deposition.
Sample A B C Pad A B C Pad
(#) Anneal Anneal Anneal Anneal
1 Carbon-rich 946 1014 1115 1104
2*Carbon-rich 1291 1165 1066 1308
10 Carbon-rich 1174 816 1163
14 Carbon-rich 1089 1099 921 2202 837 886 717 1262
6Stoich. 1778 1775 1808 1589 1212 1253 1095 1193
7Stoich. 1142 1147 1116 845 1027 1064 912 1847
8Stoich. 1521 1761 1782 1509 1056 1053 861 963
9Stoich. 944 734 924 1197 904 874 939 1019
11 Silicon-rich 1620 1515 1407 1356 972 1009 1152 996
12 Silicon-rich 1096 1101 1061 1672 1027 1013 941 1335
13 Silicon-rich 1134 1165 1173 1403 974 891 1177 978

as-deposited and after annealing. The values for carbon-rich Sample 2 reflect a 200 A Au

cap on top of the 1000 A Re contacts. Silicon-rich Samples 11, 12, and 13 as-deposited

SAs-Deposited D Annealed


1 8 0 0 ....... ................ ...... ..................... ....................
Annealed
1600- A B C Pad

1400 -

1200

1000

800
'1 -I-H


-1100
A


-600
A


Sample (Chemistry- #)
Figure 32. Thickness of each contact in angstroms and total variance in angstroms for
as-deposited and annealed samples.










values include additional 50 A and 100 A Si layers prior to 1000 A Re depositions.

Annealing reduced contact thickness, sometimes by as many as 500 A or more, and

yielded smoother film surfaces for all samples.

XRD data for as-deposited and annealed Re films are shown vs. surface chemistry

in Figure 33 and 34. There is minimal variance between individual samples for

stoichiometric and silicon-rich surfaces so the data depicted are averages of the individual

samples. As-deposited Re films on carbon-rich surfaces, however, display a wide

variance of XRD spectra (see Figure 35). Only carbon-rich Sample 14 was characterized

as-deposited as well as annealed so only its data are depicted in Figure 33 and 34. All

samples are compared to a Re standard as reported in JCPDS files (JADE Software

package).

As-deposited films exhibit XRD peaks that align with the Re standard for

stoichiometric and silicon-rich surfaces. The (101) peak is nearly 15 times greater

thanany other peaks (except (100) which is 1/3 the size) for silicon-rich as-deposited

surfaces, suggesting a highly textured as-deposited film. With stoichiometric surfaces,

the (103) peak is the largest followed by the (101) peak, which is 1/3 its size. All other

peaks are roughly 1/3 to 1/5 the size of the (103) peak suggesting surface grains with

varied Re texture. Annealed Re films on silicon-rich surfaces maintained (101) textured

structure.

Annealing increased the texturing along the (103) plane for films on

stoichiometric surfaces. All other peaks are reduced to no more than 1/20 of the value of

the (103) peak. The full width at half maximum (FWHM) of the Re peaks became

narrower after annealing by approximately 35% for stoichiometric and 20% for










Carbon-rich (Sample 14) SiC
Stoichiometric
Silicon-rich
-- Re Standard




SiC )


300

250

200

S150

100

50


11.0

2500 -


Re
(002)


Re
(100)
A


- 4000

t 3000

4 2000

i 0
I


71.0 71.5 72.0
2-Theta (Degrees)


Re
(101)


36.0 37.0 38.0 39.0 40.0 41.0
2-Theta (Degrees)
ann


S Re Re
1 (102) (110)


j I, .
t "







L=


700
600
Arnn


L 400
S300
200
100
0


42.0 43.0 44.0 45.0


50.0 55.0 60.0 65.0
2-Theta (Degrees)


70.0 73.0


78.0 83.0
2-Theta (Degrees)


Figure 33. Average XRD data for as-deposited Re films versus a standard.


13.0 15.0 17.0 19.0
2-Theta (Degrees)


2000


S 1500

1000

500

0


120

100

80

60

40

20

0


88.0


7"








82

- Carbon-rich (Sample 14)
SStoichiometric
Silicon-rich
-- Re Standard


13 15
2-Theta (Degr





Re
(002

Re
(100)


36 37

120 1 Re
100 J (102)


80

60 1

40

20


38 39 40
2-Theta
Re
(110)


r~.~.A~A


41
S(Degrees)


600 j
700

600 -
500 1
400 4
- 300
200
100 i


-1- 0
19 71 71.5 72
2-Theta (Degrees)


Re
(101)


42 43 44 45

Re Re Re
(103) (112) (201)


Re
(004)


Re
(200)





jIll


I I U I-r
50 55 60 65 70 73 78 83 88
2-Theta (Degrees) 2-Theta (Degrees)
Figure 34. Average XRD data for annealed Re films versus Re standard.

silicon-rich surfaces and shifted approximately 0.4 degree (with the exception of the

(002) peak) toward smaller 2-theta or in the direction of larger lattice spacing. No 2-theta


S4000

- 3000

2000

t 1000


2500

2000

1500

1000

500

0







83

shifts or FWHM changes are observed for SiC peaks on annealed stoichiometric surfaces.

Silicon-rich surfaces, however, exhibited FWHM widening by 50% for the 71.6 peak and

by 4% for the 17.5 peak.

As-deposited carbon-rich surfaces exhibited widely varied XRD spectra. Samples

1 and 3 show a broad diffuse region from 36 to 45 2-Theta. The (100), (002), and (101)

Re peaks are indistinguishable with no other Re peaks present. These data suggests

extremely small Re grains with little or no texturing in the film. Sample 4 shows weak

signals for all of the Re orientations, however the peaks are low in intensity and have

FWHM values roughly 10% broader than observed on stoichiometric samples. Sample

14 has texturing along the (002) plane with (101) and (103) signals roughly 25% and

10% respectively as intense.


200 Carbon 1 Carbon 3 Carbon 4 Re Standard
200 -
180 Re Re Re
160 (100) (002) (101)
140
, 120 -
C 100 .
80
60 N
40 -000
20 --------

0 --
36.0 37.0 38.0 39.0 40.0 41.0 42.0 43.0 44.0 45.0
2-Theta (Degrees)
Figure 35. (100), (002), and (101) XRD peaks for as-deposited carbon-rich Samples 1, 3,
and 4 vs. a Re standard.

After annealing, carbon-rich Sample 14 shows enhanced (002) texture with the

(103) peak being only 25% of this value and all other peaks less than 3.5%. Carbon-rich







84

Sample 2 shows polycrystalline texture with the (101) peak being largest, the (002) peak

40% smaller, and all other orientations roughly 20% as intense. All Re orientations

except (002) are shifted to larger d-spacing or smaller 2-Theta by approximately 0.4

degree with annealing similar to observations on silicon-rich and stoichiometric samples.

Re peaks are also narrower by approximately 40% than observed as-deposited. SiC

peaks, with the exception of the 71.6 peak, remain unchanged with respect to FWHM and

2-Theta spacing as compared to as-deposited films. The 71.6 peak is reduced to near

noise.


Silicon-Rich
Stoichiometric
5.5 1 Carbon-Rich
1.5

-0.5
.5
^ -2.5 1V /
-4.5 ReNVV C KLL 0 KLL
6 (176eV) (272 eV) (510eV) ___
-6.5
100 200 300 400 500 600
Electron Energy (eV)
5.5


-1 .5
F. Z_0.5


-4.5 Re MNN
(1799 eV)
-6.5 -, ,_,_,
1300 1400 1500 1600 1700 1800 1900 2000
Electron Energy (eV)
Figure 36. Average AES surveys for as-deposited films on silicon-rich, carbon-rich, and
stoichiometric surfaces.







85

AES surface surveys and depth profiles were collected for each sample. Average

survey data are shown in Figure 36 and Figure 37 comparing as-deposited and annealed

samples for all three surface conditions. As-deposited surfaces are oxygen rich and

contain graphitic carbon contamination. All AES transitions are roughly equal in shape

regardless of sample type. The larger carbon concentration on the surface of carbon-rich

samples masks (reduces) the signals from other transitions.

After annealing, the average oxygen signal appears to remain relatively

unchanged for silicon or carbon-rich surfaces but is reduced by roughly 60% for

stoichiometric surfaces. When comparing the non-averaged data per sample, the initial

surface oxygen signal, 0Io, is reduced in all cases with annealing (Table 14). In all cases,

the carbon signal grows by nearly 300% and shows pure graphite character.

Re(NVV-176 eV) transitions have a smaller inelastic mean free path and are reduced to

Table 14. The oxygen AES signal intensity, Io, on the surface for each sample. Alo
represents the relative change in oxide concentration after annealing. Stoichiometric
Sample 9 shows the least Alo consistent with an anneal <1000 C.
Sample Io Io Alo
(# Type) (As-Deposited) (Annealed)
1 Carbon-rich 5.22
2 Carbon-rich 1.00
3 Carbon-rich 8.72
4 Carbon-rich 0.77
5 Carbon-rich 2.04
14 Carbon-rich 8.80 8.18 0.62
6 Stoich. No Data 3.20
7 Stoich. 5.54 2.78 2.80
8 Stoich. 27.00 6.15 21.00
9 Stoich. 9.36 8.75 0.60
11 Silicon-rich 7.34 1.26 6.09
12 Silicon-rich 19.10 12.90 6.12
13 Silicon-rich 11.10 2.40 8.73









Silicon-Rich
Stoichiometric
4.5 ---Carbon-Rich
2.5

J 0.5
_1.5


-5.5 Re NVV C KLL 0 KLL
-(176 eV) 72eV) (510eV)

100 200 300 400 500 600
Electron Energy (eV)
4.5
2.5 -
0.5
-1Z~.5

6 -3.51
-5.5 Re MNN
i-_(1799 eV)______
-7.5 ---
1300 1400 1500 1600 1700 1800 1900 2000
Electron Energy (eV)
Figure 37. Average AES surveys for annealed films on silicon-rich, carbon-rich, and
stoichiometric surfaces.

near noise from attenuation through the graphite overlayer while the higher energy

Re(MNN-1799 eV) transitions remain relatively unchanged. No detectable AES peak

energy shifts greater than the step size of the collection (1.0 eV), are observed.

Depth profiles for as-deposited and annealed Re films are shown in Figure 38 for

stoichiometric surfaces and Figure 39 for silicon-rich surfaces. On stoichiometric

surfaces, the as-deposited interface is distinct and remains relatively unchanged with

annealing. The only detectable change with annealing is the sizable increase in surface

concentration of carbon (-100-150 angstroms) previously noted with Figure 37 vs. Figure

36. For silicon-rich surfaces, a Si enriched interface is observed as-deposited. The Si










140 -
120 o<
100-
0
80 -
60 %
A f\t


0
0 0000000000000 000


0
0
0


0 0ooo4,feo

0 ^0 ao*0ooo0 ooQ^00000000

0 500 1000 1500
Profile Depth (Angstroms)







1000000000
'IL ~ lffOOC)00000



0 500 1000 1500
Profile Denth (Anptrnms'


0 0 0
0 0 0
0 0 g
0
150 0 00o,


100


5 0 0 0

0~~~~~~~~ 0tarc^Qagtteit'__ 0 00 0


o C6 Anneal
o Si6 Anneal
o Re6 Anneal


2000


--C7
-.-Si7
- Re7
o C7 Anneal
o Si7 Anneal
o Re7 Anneal


2000


--C8
---Si8
- Re8
o C8 Anneal
o Si8 Anneal
o Re8 Anneal


500 1000 1500
Profile Depth (Angstroms)


120
100
80
20




0


-*-C9
--Si9
-Re9
o C9 Anneal
o Si9 Anneal
o Re9 Anneal


0 500 1000 1500 2000
Profile Depth (Angstroms)
Figure 38. Depth profiles of as-deposited and annealed Re films on stoichiometric SiC
surfaces for Samples 6, 7, 8, and 9.


4U
20
0


60
50
40
30
20
10
0
-10


~U ,

- *


2000











+CI1
-.-C11
--Sill
---Rell
o C1 Anneal
o Sil 1 Anneal
o Rel Anneal


2- *

<1




-3 *



f 1


<11
-4 W_
zv


2000


n .te^?08888
|ifltliektlAKf^ __ ___
0 0000


0 500 1000 1500 2000


'_ g MM o o U

0 0 0

1000 1500 2000
Profile Depth (Angstroms)


--C12
--Sil2
--Rel2
o C 12 Anneal
o Sil2 Anneal
o Rel2 Anneal


--C13
--Sil3
--Rel3
o C13 Anneal
o Sil3 Anneal
o Re 13 Anneal


Figure 39. Depth profiles of as-deposited and annealed Re films on silicon-rich surfaces
for Samples 11, 12, and 13.

layer is 47, 88, and 102 angstroms respectively for Samples 11, 12, and 13 reflecting

silicon-rich sample preparation. After annealing, Si and C are stoichiometric at the

interface for Samples 11 and 12 and the Si layer is reduced to 55 angstroms for Sample


0 500 1000 1500
Profile Depth (Angstroms)


- *


80

60

40

20

0$







89

13. There is little change in the sharpness of the Re/SiC interface for Sample 11. Sample

12 and 13, however, show diffused interfaces. Re migrated into the SiC for an additional

250 angstroms vs. as-deposited films. Si and C signals require an additional 300

angstroms in Sample 12 and roughly 500 angstroms in Sample 13 to become

representative of bulk stoichiometric SiC. With the exception of thin surface layers (-60

A) oxygen signals were near zero through the entire contact and are thus not displayed.

Depth profiles for carbon-rich surfaces revealed unique responses to annealing.

Samples 1 to 4 contained as-deposited films that were uniformly contaminated by carbon

from the electron beam grazing the crucible. Figure 40 shows the depth profile of

Sample 1, which is representative of this deposition contamination. The electron beam

current drifted downward slowly over the course of this profile causing the general

reduction in AES signal intensity for all constituents. The annealed film on carbon-rich

Sample 2 (Figure 41) shows a non-uniform distribution of carbon in the film. Two highly

concentrated regions of carbon, "Clump 1" and "Clump 2", resulted at the surface prior to


50
40

4i 30
E 20

10

0 --.
0 100 200 300 400 500 600 700 800 900 1000 1100
Profile Depth (Angstroms)
Figure 40. AES depth profile data for an as-deposited Re film on a carbon-rich surface
representative of Samples 1 to 4 prior to covering the carbon crucible with Re foil to
prevent contamination. These data are from Sample 1.







90

Au capping and inside the Re film. The interface remains distinct similar to as-deposited

conditions.

Clumps 1 and 2

j / ^ SAMPLE
25
20
Z 15
ll10 j- is
5
5 .-'."* "i
0
0 250 500 750 1000 1250 1500 1750 2000 2250
Profile Depth (Angstroms)
Figure 41. AES depth profile data for an annealed Re film on a carbon-rich surface
representative of Samples 1 to 4 prior to covering the carbon crucible with Re foil to
prevent contamination. These data are from Sample 2.

Carbon-rich Samples 5 and 14 were prepared after covering the carbon crucible

with pure Re foil to prevent contamination during deposition. Figure 42 shows the

as-deposited profile for Sample 5 and the annealed profile for Sample 14. The Si (KLL -

1630 eV) peak was tracked for Sample 5 to avoid interference with the Au cap (NVV -

74 eV), while the more surface sensitive Si (LMM 92 eV) signal was tracked for

Sample 14 since Au was not present. The interfaces for both Sample 5 and 14 appear to

be distinct with near stoichiometric or slightly silicon-rich chemistry. It is important to

note that the AES spectra of higher energy transitions are routinely noisier than lower

energy transitions from the same spectrum. The apparent Si rich interface and bulk in

Sample 5 is believed to be an artifact of this increased noise compared to the lower

energy carbon (KLL) signal.







91

The AES signals for surfaces, interfaces, and SiC regions during depth profiling

are nearly identical as-deposited and annealed from sample to sample regardless of

prepared surface chemistry. Figure 43 shows the shape of each signal while depth

profiling through the Re contact on carbon-rich Sample 14 as a reference. The only

chemical change detected was the shift from graphitic carbon at the surface to carbidic


SAMPLE 5


140
120
.^ 100
I -o
Z 80
13"60
, <1 40
20
0


600 800 1000 1200 1400 1600 1800 2000 2200

Profile Depth (Angstroms)

SAMPLE 14


-z
9b3


0 200 400 600 800 1000 1200 1400 1600

Profile Depth (Angstroms)

Figure 42. AES depth profiles for the as-deposited film on carbon-rich Sample 5 and the
annealed film on carbon-rich Sample 14. These films were deposited after covering the
crucible with Re foil.







92
-Surface-Interface SiC

> 5 0:.5 -- ---- 2 -
0.5 -
0 -1.5
Sz -3.5 0
,Q-5 -5.5 -2
SC -7.5 Si Re
-10 ----9.5 ,-r- -4 --.... r.
240 260 280 300 75 80 95 105 1760 1780 1800 1820
Electron Energy (eV) Electron Energy (eV) Electron Energy (eV)
Figure 43. Representative AES transition signals for C (KLL), Si(LMM), and Re (MNN)
while depth profiling through the film. These data are from carbon-rich Sample 14 after
annealing.

carbon after passing through the interface. Carbon-rich Sample 2 was the only exception

to this standard. Figure 44 shows the chemistry of the C(KLL), Si(LMM) and Re(NVV)

transitions while profiling through Clumps 1 and 2, the Re film, and the interface.

Clumps 1 and 2 are graphitic and do not indicate any reaction with Re in the film.

Clumps I and 2 -Surface -Interface SiC
Clumps 1 and 2
S1'Between 1 and 2

I 0 1 0.5


i-2 -2 S -05 Re
-3 -3 -1.5 ,
240 260 280 300 75 80 95 105 145 165 185
Electron Energy (eV) Electron Energy (eV) Electron Energy (eV)
Figure 44. AES C(KLL), Si(LMM), and Re(NVV) transitions while profiling carbon-
rich Sample 2.

I-V data were collected using the TLM pattern and a four-point probe technique

for as-deposited and annealed stoichiometric (Figure 45), silicon-rich (Figure 46), and

carbon-rich (Figure 47) surfaces. The I-V traces for as-deposited films are less linear