Growth and surface characteristics of oxide thin films for chemical sensor applications

MISSING IMAGE

Material Information

Title:
Growth and surface characteristics of oxide thin films for chemical sensor applications
Physical Description:
xviii, 222 leaves : ill. ; 29 cm.
Language:
English
Creator:
Burgess, Darren R
Publication Date:

Subjects

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

Notes

Thesis:
Thesis (Ph.D.)--University of Florida, 1999.
Bibliography:
Includes bibliographical references (leaves 184-195).
Statement of Responsibility:
by Darren R. Burgess.
General Note:
Typescript.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 030471188
oclc - 42450688
Classification:
lcc - LD1780 1999 .B955
System ID:
AA00013548:00001


This item is only available as the following downloads:


Full Text








GROWTH AND SURFACE CHARACTERISTICS OF OXIDE
THIN FILMS FOR CHEMICAL SENSOR APPLICATIONS















By

DARREN R. BURGESS


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


1999






























Copyright 1999

by

Darren R. Burgess





























To My Wife,

Dana











ACKNOWLEDGMENTS


I would like to express my gratitude to Professor Tim Anderson, my thesis advisor,

whose forward thinking in the area of graduate education made this highly valuable

industrial Ph.D. thesis research program possible. His numerous trips to Wilmington, DE,

during the progress of this research are greatly appreciated.

I am extremely grateful to Dr. Pat Morris Hotsenpiller for serving as my mentor and

thesis co-advisor. Her patience and guidance during my time at the Experimental Station

are greatly appreciated. Her steadfast teaching of the value of pursuing scientific excellence

by example in her own research for The DuPont Company has been invaluable in my

training. Also, her end goal of having me pursue whatever career I believed would be

fulfilling is an invaluable trait which I greatly admire.

The DuPont Company especially Drs. Jan Lerou, Kathy Saturday and Jim

Trainham are thanked for providing me with the research funding and opportunity to gain

industrial experience during my graduate education.

Mr. George Wilson will always have my appreciation for his constant willingness

to teach me about the construction, operation and maintenance of all types of laboratory

equipment. Without his broad knowledge in these areas, many experiments would have

been far more difficult to perform.

Drs. Jim Hohman, Stephens The and Sandy Witman, along with other former

members of the reaction engineering research group, are thanked for their efforts in helping

to start my research at DuPont and always being available to help in any way thereafter.






Drs. Kirstine Myers and Jim McCambridge are thanked for their efforts in teaching

me to use the x-ray diffraction equipment which is under the supervision of the DuPont

superconductivity group.

Mr. Scott McLean is thanked for his unselfish willingness to perform a large

amount of analysis by atomic force microscopy which proved to be very valuable in the

explanation of various results.

Dr. Victor Lusvardi is thanked for sharing his expertise in UHV material analyses

and his willingness to do x-ray photoelectron spectroscopy analysis. His camaraderie and

encouragement in the last year of my research are also appreciated.

The staff and my colleagues at the University of Florida Department of Chemical

Engineering especially Shirley Kelly, Nancy Krell, Bill Epling, Todd Dann and Michelle

Griglione are thanked for helping with the many tasks I could not perform while I was

away from campus.

My family and friends are thanked for their endless support and encouragement

during my graduate education. Dr. Howard Gerlach is especially appreciated for being a

mentor in recent years.

Very special thanks and gratitude are reserved for my wife, Dana, who has shown

unlimited patience and understanding throughout my graduate education. Ultimately, I am

thankful to the Lord for His grace and mercy.











TABLE OF CONTENTS

Page

A CKN O W LED G M ENTS .................................... .......................................... iv

LIST OF TABLES ..................................................................................................... xi

LIST OF FIG URES ................................................................................................. xii

ABSTRA CT .......................................................... ...............................................xvii

CHAPTERS

1. INTRO D UCTIO N........................................................................................... 1

1.1 M otivation......... 1
1.2 O objective ................................................................................................ 2
1.3 Approach.............................................................................................. 2

2. LITERATURE REVIEW .................................................................. 4

2.1 Rutile Thin Film Deposition...................................................................4..
2.1.1 Physical Vapor Deposition ..................................................... 5
2.1.2 Chemical Vapor Deposition ................................................... 5
2.1.2.1 M etalorganic chem ical vapor deposition.....................7
2.1.2.2 CVD related techniques..............................................7
2.1.3 Heteroepitaxy ........................................................................ 8
2.2 Rutile G as Sensors.................................. ..................................... 11
2.2.1 O oxygen Sensors...................................................................... 11
2.2.2 Humidity Sensors 12
2.2.2 Hum idity Sensors................................................................... 12
2.3 Rutile Properties ................................................................................. 13
2.3.1 Crystal and Band Structure ................................................. 13
2.3.2 Conductivity .......................................................................... 16
2.3.2.1 Defects and m echanism .............................................. 16
2.3.2.2 Activation energy and m obility ................................. 19
2.3.2.3 Intrinsic band-gap states............................................ 21
2.3.3 Doped rutile .......................................................................... 21


vi







2.3.3.1 Donor and acceptor defect chemistries...................... 21
2.3.3.2 Structure and conductivity effects............................. 22
2.3.4 UHV Surface Characterization..... .............................. 24
2.3.4.1 (110)-oriented rutile................................................... 25
2.3.4.2 (100)-oriented rutile.................................................. 25
2.3.4.3 (001)-oriented rutile ................................................... 25
2.4 Other Metal Oxides .............................................................................26
2.4.1 Gallium Oxide ............................................. 26
2.4.1.1 Deposition .................................................................26
2.4.1.2 Gas sensitivity ........................................................27
2.4.2 Aluminum oxide ............................... 27

3. EXPERIMENTAL METHODS ..............29

3.1 Thin Film Deposition ...................................... .................................... 29
3.1.1 Reactors............................................................................ 29
3.1.2 Precursors ........................................................................... 33
3.1.3 Substrates 34
3.1.4 Ion-Beam Sputter Deposition................................................. 35
3.2 Physical Characterization..........................................................................36
3.2.1 X-ray Diffraction................................ ..............................36
3.2.2 Rutherford Backscattering Spectrometry............................... 38
3.2.3 Atomic Force Microscopy ..................................................... 41
3.2.4 X-ray Photoelectron Spectroscopy ....................................... 45
3.2.5 Secondary Ion Mass Spectrometry and Auger Electron
Spectroscopy ................................................................... 46
3.3 Electrical Characterization........................................................................ 48
3.3.1 Sample Preparation.............................................................. 48
3.3.2 Environmental Chamber......................................................... 50
3.3.2.1 Chamber design........................................................ 50
3.3.2.2 Testing procedure...................................................... 52
3.3.2.3 Sample heaters .......................................................... 53
3.3.3 Impedance Measurements ............................................... ....... 53
3.3.3.1 Theory of measurement............................................. 54
3.3.3.2 Calculations from impedance measurements..............55
3.3.3.3 Data approximations .................................................. 57
3.4 Equipment Safety .................................................................................. 59


vii






4. RUTILE FILM DEPOSITION.................................................................. 60

4.1 Undoped Rutile Films............................................................................. 60
4 .1.1 D eposition................................................................. .. ....... 60
4.1.1.1 Precursor deposition efficiency ................................. 60
4.1.1.2 Precursor .............................. 61
4.1.1.3 Activation energy ................................................... 63
4.1.2 Crystalline Quality ................................................................ 65
4.1.2.1 Bulk crystalline quality ............................................65
4.1.2.2 Heteroepitaxy .......................................................... 68
4.1.3 Morphology ......................................................................... 70
4.1.3.1 Orientational effects ........ ............................... 70
4.1.3.2 Substrate effects ........................................................79
4.2 Doped Rutile Films................................................................................ 81
4 .2.1 D position ..................................................................... ...... 82
4.2.1.1 Bulk dopant concentration ..........................................82
4.2.1.2 Surface dopant concentration.............................. ....... 83
4.2.2 Nb-doping Structural Effects.......................................... ........ 87
4.2.2.1 Bulk crystal structure................................................. 87
4.2.2.2 Surface morphology ................................................91
4.2.3 Ga-doping Structural Effects................................................ .. 94
4.2.3.1 Bulk crystal structure .............................................. 94
4.2.3.2 Surface morphology ..................................................98
4.3 Sum m ary ................................................................................................. 10 1
4.3.1 Undoped Rutile Film Deposition ............................................ 101
4.3.2 Doped Rutile Film Deposition.......................................... ....... 102

5. RUTILE GAS SENSORS.................................................................................... 104

5.1 Film Orientation Results Summary ........................................................ 104
5.2 Film Orientation Discussion .................. ....................................... 106
5.2.1 Measurements............................................................. 106
5.2.2 M echanism s............................................................................ 108
5.2.2.1 Film conductivity....................................................... 108
5.2.2.2 Activation energy ..................................................... 109
5.2.2.3 Oxygen sensitivity................................................ 111



viii






5.2.2.4 H humidity sensitivity ................................................... 112
5.2.3 (101) and (001) Orientations Comparison.............................. 113
5.2.3.1 Surface Morphology ................................................ 113
5.2.3.2 Surface Chemistry ................................................... 121
5.2.3.3 Electron M obility .....................................................1... 23
5.2.4 (100) Orientation Discussion ...................................................125
5.3 Film Composition Results Summary....................................................... 128
5.4 N b-doped film s .........................................................................................130
5.4.1 Nb-doped Film Results.......................................................... 130
5.4.2 Nb-doped Films Discussion....................................................132
5.4.2.1 O xygen sensitivity .................................................... 132
5.4.2.2 Humidity sensitivity............................................ ....... 133
5.5 Other Compositions.............................................................................. 134
5.5.1 Undoped and Ga-doped Films Results........................... ....... 134
5.5.2 Undoped Films Discussion ....................................................135
5.5.3 Ga-doped Films Discussion .................................................... 139
5.5.3.1 Carbon monoxide sensitivity ..................................... 140
5.5.3.2 Sensitivity with other dopants.. .............................. 141
5.5.3.3 P-type conducting films.................................... ........ 142
5.5.3.4 Adsorption enhancement by a thin surface layer........ 145
5.5.3.5 Adsorption enhancement by film reduction................ 148
5.5.4 Mobility and Composition...................................................... 155
5.6 Engineering Sensitivity Assessment......................................................... 156
5.7 Sum m ary ............................................................................................... 159

6. METAL OXIDE FILM DEPOSITION...................................................... 161

6.1 Gallium Oxide 161
6.2 Aluminum Oxide ............171
6.3 Sum m ary ................................................................................................. 174

7. CONCLUSIONS AND RECOMMENDATIONS FOR FUTURE WORK .....176

7.1 C conclusions ....... ............................................ 176
7.1.1 UndopedRutile Film Deposition ............................................ 176
7.1.2 Doped Rutile Film Deposition ............................................ .... 177
7.1.3 R utile G as Sensors ........................................................... 178






7.1.4 Deposition of Other Metal Oxides......................................... 180
7.2 Recommendations for Future Work.............................................. 180
7.2.1 Undercoordinated Surface Cation Reactivity........................... 180
7.2.1.1 Z inc O xide ........................... ................................ 180
7.2.1.2 Sensitivity to other donor- or acceptor-type
m olecules................ ............................ 181
7.2.1.3 U H V analyses ........................................................... 181
7.2.2 T ailored M materials ...................................................................182
7.2.3 P-type Conducting Sensors .................................................... 182
7.2.4 Gallium Oxide Films................................................... 183

R E F E R E N C E S ........................................................................................................184

APPENDIX

SAFETY DOCUMENTATION FOR GAS SENSITIVITY
E X PE R IM E N T S ............................................................................... 196

BIOGRAPHICAL SKETCH ...................................... 222











LIST OF TABLES


Table Page

2.1 Cationic and anionic lattice mismatch for heteroepitaxial rutile
deposited on sapphire substrates.............................................. ........ 10

2.2 Oxygen sensor examples..............................................................12

2.3 Humidity sensor examples................................................................ 13

2.4 Dominant impurity trends in TiO2.............................................. ........ 19

2.5 Representative activation energy and mobility data for TiO2.............. 20

3.1 Growth variable ranges for quartz reactor growth study .....................30

3.2 Sublimation temperatures of metal precursors determined by TGA.... 34

4.1 Reported rutile growth rate dependence on Ti precursor ....................62

4.2 Reported MOCVD of TiO2 growth rate activation energies.................65

5.1 A measure of photoreactivity as a function of thin film orientation..... 123

6.1 Summary of Ga203 growth from Ga(TMHD)3 precursor.................. 164

6.2 Dangling bond densities of surface oxygen atoms........................... 166

6.3 Summary of A1203 growth from AI(TMHD)3 precursor.................... 173











LIST OF FIGURES


Figure Page

77
2.1 Rutile unit cell. From A. von Hippel et al.77, Fig. 2...................14

3.1 Solid source MOCVD reactor schematic................................ 30

3.2 Precursor schematic 33

3.3 Thin film x-ray diffraction geometry ................................................ 38

3.4 Schematic of essential Rutherford backscattering equipment ............. 39

3.5 RBS spectra of Ti02 on sapphire with RUMP model.........................41

3.6 AFM equipment schematic (Courtesy of R. S. McLean).................... 43

3.7 TM AFM images of a Ga203 thin film (a) normal and
(b ) p hase shift.................................................................................. 44

3.8 Schematic of alumina sample holder with attached film...................... 49

3.9 Environmental chamber feedthrough schematic.................................. 51

3.10 Environmental chamber gas flow architecture schematic..................... 51

3.11 Impedance bridge simplified measurement circuit schematic...............55

3.12 Complex impedance data as a function of sample temperature
for an undoped (101)-oriented rutile thin film in N2 ....................... 57

3.13 Complex Z plot for le9 Q resistor in parallel with 1 pF capacitor ......58

3.14 Complex Z plot for 5e9 QL resistor in parallel with 1 pF capacitor ......58

4.1 Rutile film growth rate as a function of Ti or 02 partial pressure........61

4.2 Logarithm of growth rate as a function of reciprocal temperature....... 64

4.3 Rc fwhm of undoped rutile XRD peaks as a function of substrate
temperature and deposition reactor.................................................. 66

4.4 (100) rutile XRD peak rc fwhm as function of PO2 .......................... 68

4.5 (001)-oriented 70 A rutile film AFM image ....................................... 71






4.6 (001)-oriented 400 A rutile film AFM image..................................... 71

4.7 (001)-oriented 3350 A rutile film AFM image.................................... 72

4.8 (101)-oriented 70 A rutile film AFM image ....................................... 72

4.9 (101)-oriented 400 A rutile film AFM image.......................................73

4.10 (101)-oriented 3350 A rutile film AFM image.................................... 73

4.11 (100)-oriented 70 A rutile film AFM image ...................................... 74

4.12 (100)-oriented 400 A rutile film AFM image...................................... 74

4.13 (100)-oriented 3350 A rutile film AFM image .................................. 75

4.14 RMS roughness as a function of thickness taken from AFM
images of undoped oriented rutile films and their respective
sapphire substrates........................................................................ 75

4.15 Characteristic planar dimension of surface features as a function
of film thickness for undoped, oriented rutile films......................... 77

4.16 Ratio of characteristic height to planar dimension of surface
features as a function of film thickness for undoped, oriented
rutile film s............................ .............................................. 79

4.17 (0001)-oriented sapphire.................................................................. 80

4.18 (1120)-oriented sapphire with unremovable surface features............. 80

4.19 (101)-oriented 70 A rutile film on "featured" (1120) sapphire
substrate .......................................................... 81

4.20 (101)-oriented 3350 A rutile film on "featured" (1120) sapphire
substrate .......................................................... 81

4.21 XPS spectra of Ga 2p3/2 core electron binding energy for Ga
atom in (001)-oriented rutile thin film........................................... 84

4.22 Ga 2P3/2 core electron binding energy as a function of Ga
oxidation state .................................. 85

4.23 Nb 3d5/2 core electron binding energy as a function of Nb
oxidation state .................................. 86

4.24 XPS spectra of Nb 3d5/2 and 3d3/2 core electron binding
energies for Nb atom in (001)-oriented rutile thin film.....................87

4.25 (002) rutile plane d-spacing as a function of Nb doping .................... 89

4.26 (101) rutile plane d-spacing as a function of Nb doping .................... 89


Xiii






4.27 (001) and (101) rutile peak 0 rc fwhm as a function of
N b doping................................................................................. .... 90

4.28 (001) and (101) rutile 4 scan peak fwhm as a function of
N b doping ..................................................................................... 9 1

4.29 Nb-doped (001)-oriented thin film AFM images (a) 70 A
(b) 3350 A ....................................... 92

4.30 Nb-doped (101)-oriented thin film AFM images (a) 70 A
(b) 3350 A ................................... ......................................... 92

4.31 Nb-doped (100)-oriented thin film AFM images (a) 70 A
(b) 3350 A............................ 93

4.32 (002) rutile plane d-spacing as a function of Ga doping.....................94

4.33 (101) rutile plane d-spacing as a function of Ga doping......................95

4.34 (001) and (101) rutile peak 0 rc fwhm as a function of
G a doping ................................................................................... 97

4.35 (001) and (101) rutile scan peak fwhm as a function of
G a doping ................................................................................... 98

4.36 Ga-doped (901)-oriented thin film AFM images (a) 70 A
(b) 3350 A ............................ ............................................... 99

4.37 Ga-doped (101)-oriented thin film AFM images (a) 70 A
(b) 3350 A ............................. ............................................... 99

4.38 Ga-doped (100)-oriented thin film AFM images (a) 70 A
(b) 3350 A ............................................................................... 100

5.1 Conducting electrons removed as a function of film orientation
calculated from resistance values measured in N2 and 02
atmospheres at 240 C .............................................................. .... 105

5.2 Conducting electrons added as a function of film orientation
calculated from resistance values measured in N2 and
humidified N2 atmospheres at 240 C...................................... ....... 106

5.3 Activation energy of conductance as a function of film orientation
in different atm ospheres................................................................ 110

5.4 Surface proton conduction on undoped thin film near room
temperature and electronic conductivity at higher temperatures
measured over several temperature cycles........................................ 113

5.5 Orientation stability diagram for rutile at 1273K................................. 115






5.6 AFM images of (001)-oriented rutile thin films (a) deposited by
MOCVD (b) deposited by IBSD..................................................... 116

5.7 Schematic of facets on (001)-oriented rutile surface........................... 117

5.8 AFM images of (101)-oriented rutile films (a) deposited by
MOCVD on "smooth" substrate (b) deposited by MOCVD on
"featured" substrate (c) deposited by IBSD....................................117

5.9 Schematic of facets on (101)-oriented rutile IBSD and MOCVD
deposited surfaces....................................................................... 119

5.10 Electrons donated to an undoped (101)-oriented film as a function
of m obility ................................................................................... 125

5.11 AFM images of (100)-oriented rutile films (a) deposited by
MOCVD (b) deposited by IBSD..................................................... 126

5.12 Conducting electrons removed as a function of film composition
calculated from resistance values measured in N2 and 02
atmospheres at 240 C.................................................................... 129

5.13 Conducting electrons added as a function of film composition
calculated from resistance values measured in N2 and humidified
N2 atmospheres at 240 C............................................................. 130

5.14 [V02*] calculation as a function of PO2 at 240 C by Marucco
m odel ......................................... ............................................... 137

5.15 Calculated electrons resulting from [Vo2*] as a function of [V2"]
enthalpy of form ation...................................................................... 138

5.16 Comparison of conducting electrons added by humidity to an
Fe-doped film compared to other compositions...................... ........ 142

5.17 Estimated time required for 02 diffusion through a 3300 A thick
rutile film as a function of temperature............................................. 144

5.18 Comparison of conducting electrons added by humidity to a film
with a Ga-doped surface layer (Ga*) to other compositions............ 147

5.19 Comparison of conducting electrons removed by oxygen from
a film with a Ga-doped surface layer (Ga*) to other
com positions ................................. ............................... .. ....... 148

5.20 N2 atmosphere resistance values measured at 2400C and
calculated activation energies for (001)-oriented, reduced and
unreduced sam ples....................................................................... 150

5.21 Impedance spectra as a function of atmosphere at 240 OC for
(001)-oriented rutile film reduced 12 hrs in H2 at 500 OC.................151






5.22 Impedance spectra as a function of atmosphere at 240 C for
(001)-oriented rutile film reduced 6 hrs in H2 at 500 C.................152

5.23 Comparison of conducting electrons removed by oxygen from
reduced film s to other compositions ................................................ 153

5.24 Comparison of conducting electrons added by humidity to
reduced films to other compositions .................................. ............. 154

5.25 Sensitivity of various films to the introduction of 1.332e-3
atm 02 (R) to an atmosphere of less than le-lO atm 02 (Ro)..........157

5.26 Sensitivity of various films to the introduction of humid N2
(R) to an atmosphere of dry N2 (Ro).............................................. 158

6.1 Schematic of B-Ga203 unit cell (a) Oblique (b) Normal to b-axis....... 162

6.2 AFM images of B-Ga203 deposited on (0001) sapphire
(a) 600 C deposition, normal image (b) 600 C deposition,
phase contrast image (c) 700 C deposition, normal image
(d) 700 C deposition, phase contrast image................................. 168

6.3 AFM images of B-Ga203 deposited on (111) Si
(a) 600 C deposition, normal image (b) 600 "C deposition,
phase contrast image (c) 700 "C deposition, normal image
(d) 700 "C deposition, phase contrast image............................. ....... 170

6.4 AFM images of B-Ga203 deposited on (111) at 700"C and
annealed at 800"C in air for one hour (a) normal image
(b) phase contrast im age................................................................ 171











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


GROWTH AND SURFACE CHARACTERISTICS OF OXIDE
THIN FILMS FOR CHEMICAL SENSOR APPLICATIONS

By

Darren R. Burgess

August 1999


Chairman: Timothy J. Anderson
Major Department: Chemical Engineering


The effects of surface chemistry, orientation and morphology on the gas sensing

properties of rutile thin films were investigated. The motivation for this work was to

improve the limited understanding of gas sensing mechanisms in semiconducting materials.

The metalorganic chemical vapor deposition of rutile phase TiO2 thin films on sapphire

substrates was studied and thoroughly developed to prepare materials with well-defined

physical characteristics for the gas sensor investigation.

Heteroepitaxial (001)-, (101)- and (100)-oriented rutile films were deposited on

(1010)-, (1120)- and (0001)-oriented sapphire substrates, respectively, using solid source

metalorganic precursors. X-ray diffraction (XRD) analysis determined the films'

crystalline characteristics. Atomic force microscopy identified the facet planes on (001)-

and (101)-oriented film sufaces.

The films were doped with Ga, an acceptor-type dopant, or Nb, a donor-type

dopant. XRD proved that heteroepitaxy was maintained to 6.5 at%. The lattices of (001)-

and (101)-oriented films were observed to expand linearly with increasing dopant


xvii






concentration. Changes in surface morphologies due to doping were related to increased

film stress and defects caused by lattice expansion.

The gas sensor properties of each undoped film orientation and (001)-oriented films

with various compositions were investigated. Impedance spectroscopy measurements were

used to arrive at overall changes in the number of conducting electrons as a function of

changing environment. The number of conducting electrons added by dissociatively

adsorbed water or removed by oxygen was substantially greater for a (101)-oriented film

than an (001)-oriented film. This behavior was related to the density of Ti atoms with low

oxygen coordination present on facet planes and their intersections. A previously unknown

relationship was discovered between surface cation oxygen coordination, photochemical

reactivity and enhanced adsorption of acceptor- or donor-type molecules.

Undercoordinated surface cations were also created by Ga-doping and hydrogen reduction

resulting in enhanced adsorption. The electronic compensation of Nb dopant atoms also

resulted in enhanced oxygen adsorption.

Oriented thin films of Ga203 and A1203 were also deposited from solid

metalorganic precursors. The amount and relative orientation of crystalline material was

found to increase with increasing deposition temperature.


xviii











CHAPTER 1
INTRODUCTION


1.1 Motivation


Increasing needs for pollution monitoring and control have given rise to the need

for gas sensors which demonstrate higher selectivity and sensitivity.'2 Among the more

familiar examples is the oxygen sensor contained in an automobile's exhaust system. The

function of this sensor is to control the air-to-fuel ratio to minimize the concentration of

uncombusted hydrocarbons or nitrous oxides released into the atmosphere.3'4 This sensor,

which is commonly made from zirconium oxide, suffers from inadequate operation at low

temperatures. This characteristic renders it unable to control the engine combustion process

until a built-in heater brings the material to an adequate temperature. Therefore, a large

amount of pollution is released just after a car is started due to a lack of sensitivity over the

desired temperature range of operation.

Many efforts have been made to improve sensitivity and selectivity of gas sensors.

Approaches have included modifying a material's bulk or surface characteristics through

doping or annealing. Also, the observed electrical response has been changed through the

use of different contact materials and configurations. These efforts have met with varying

degrees of success. A lack of understanding of gas sensing mechanisms has limited the

innovation in gas sensor technology.

To engineer an improved gas sensor, a better understanding of the fundamental

processes occurring during sensing is necessary. The mechanism of gas sensing can be

separated into the adsorption of a gaseous species and its observable effect on the electrical

conductivity. Each must be understood individually as a function of the material and the






environment. This knowledge can then be correlated to create a complete picture of the

sensing mechanism and variables of interest. The primary motivation for this work is to

improve the understanding of gas sensing mechanisms by investigating the relationship

between a material's bulk and surface properties and its sensitivity.


1.2 Objective


The first objective of this work is to further the understanding of complex metal

oxide thin film deposition by metalorganic chemical vapor deposition (MOCVD). This thin

film deposition method is characterized by high throughput which makes it attractive for

mass production. The primary goal for using MOCVD in production is to not sacrifice

crystalline film quality at the faster growth rates. The investigation of films' physical

characteristics as a function of growth conditions and doping furthers the understanding of

MOCVD as applied to more complex material systems.

The second objective of this work is to create changes in a gas sensitive material by

doping and relate those changes to the observed gas sensing behavior. To accomplish this

objective, a gas sensitive material must be obtained in a configuration where both the bulk

and surface properties can be characterized. The ability must also exist to identify the

changes caused by the addition of dopant atoms using these same characterization methods.

The actual effect of various dopants on the gas sensitivity of a material can then be more

accurately assessed as an aid to future gas sensor design.


1.3 Approach


To fulfill the objective of creating a material whose bulk and surface properties can

be identified with available methods, rutile phase titanium dioxide (TiO2) was chosen.

TiO2 has been widely studied as a model for the behavior of transition-metal oxides






because of its stability over a wide range of temperatures and environments. In addition,

TiO2 exhibits semiconductive behavior at high temperatures as a function of oxygen partial

pressure. A wealth of literature exists on the expected physical, electrical and adsorptive

characteristics of TiO2.

Among the available forms of TiO2 (e.g., powder, single crystal), non-porous,

crystalline thin films lend themselves to the simultaneous study of bulk and surface

properties. The first objective of this work is met by depositing rutile TiO2 thin films by

MOCVD. The (001), (101) and (100) crystal orientations are grown heteroepitaxially on

the (1010), (1120) and (0001) orientations of sapphire, respectively. The films are doped

with donor- and acceptor-type atoms to assess their effects on deposition and later,

sensitivity. The films are primarily characterized by x-ray diffraction for their crystalline

quality and atomic force microscopy for their surface morphologies.

Having created undoped and doped rutile thin films with well-defined bulk and

surface properties, the second objective is met by investigating these films as gas sensors.

An environmentally-controlled chamber is used to subject the films to reducing and

oxidizing atmospheres and humidity. The observed changes in the electrical behavior are

analyzed with respect to the rutile electrical properties literature and the ultra-high vacuum

surface characterization literature to create a mechanistic model of oxygen and humidity

sensing.













CHAPTER 2
LITERATURE REVIEW



2.1 Rutile Thin Film Deposition



The basic processes involved in thin film deposition are adsorption, nucleation and

growth.5-7 First, a sufficient number of atoms from the vapor phase must adsorb onto the

growth surface to form a permanent nucleus of material. Depending on the binding energy

of the depositing material, relative to itself and the growth surface, and the energy supplied

through heat by the growth surface, material will deposit by one of three mechanisms. One

mechanism is two-dimensional or layer-by-layer growth. In two-dimensional growth,

material addition to a stable nuclei is predominantly in the plane as opposed to building

upward. The growth takes place in a sheet-like manner and is more typical of slow

deposition rates and low strain in the growing material. The opposite of two-dimensional

growth is island or Volmer-Weber growth which is characterized by growth in three

dimensions. Island growth can occur for a couple of reasons. One is when the adsorbing

atoms are more strongly bound to themselves than the growth surface. This condition

makes adsorption on top of previously deposited material more favorable. Another cause is

deposition occurring at a higher rate than the atoms can transport across the growth surface

to extend a stable nucleus in two dimensions, so the atoms simply pile up. The Stranski-

Krastonov growth mode is a mixture of the layer and island mechanisms. Atoms initially

deposit in layers but due to a strain induced defect or other instability, island formation

begins to occur.

TiO2 thin film deposition first received interest for optical applications such as anti-

reflection coatings and for electrical applications because of the material's high dielectric







constant.8 In the ensuing period, the deposition of TiO2 has been attempted by a variety of

physical and chemical vapor deposition (CVD) methods. Rutile is the high-temperature

phase of TiO2 which is desirable for most applications. The deposition rate and

temperature dependence for obtaining this phase has been documented for different

deposition techniques.9-14 Rutile phase films have been deposited by CVD at a substrate

temperature of 450 C and higher. Physical vapor deposition (PVD) methods have used a

substrate temperature as low as 550 OC. The fraction of rutile material has been observed to

decrease using either deposition method at high deposition rates in combination with low

deposition temperatures resulting in a portion of the material being amorphous or anatase

phase.


2.1.1 Physical Vapor Deposition


The objective of PVD is to transfer atoms from a source to a growth surface where

the film is deposited atomistically. Traditional PVD processes are characterized by high

vacuum and the absence of chemical reactions.5 Among the PVD methods with higher
energy requirements, rutile TiO2 has been deposited by rf and de plasma sputtering,15-20

ion-assisted evaporation21-24 and ion beam sputter deposition.14 Increased film density,

reduced film stress, and increased control of doping at very low levels are some of the

advantages of these methods.


2.1.2 Chemical Vapor Deposition


CVD is the process of reacting a volatile compound, often with other gases, to

produce a solid which deposits on a surface. The volatile compounds and gases are known

as precursors, and the deposition surface is known as a substrate. The substrate is

normally heated to facilitate a reaction which forms a solid film. The desired product may






form in the gas phase and deposit molecularly, or one or more of the precursors may

adsorb on the substrate and react to create the film. Which of these processes occurs

depends on the energy required to separate the desired atom(s) from the remainder of the

precursor molecules and to motivate a reaction. The advantages of CVD are its ability to

controllably create films with widely varying stoichiometry, and its adaptability to semi-

continuous or batch processing.5

The foremost way in which the CVD process has been enhanced is the use of a

plasma.25-29 A plasma aids in decomposing the precursor molecules so that their reaction

is not solely dependent upon energy provided by a heated substrate. Plasma-enhanced

CVD is especially helpful when the energy to create a reactive form of a precursor is

beyond the capabilities of a resistively heated substrate. A good example is the deposition

of nitride compounds such as WN or TiN. The plasma forms nitrogen radicals which are

reactive with metal atoms from normally inert N2 gas.3034

When one of the reactants is a compound containing a metal atom bound in an

otherwise organic molecule, the compound is labeled a metalorganic and the deposition

process is known as MOCVD. Metalorganic precursors first received prominent attention

for the deposition of room temperature semiconductors. Compounds such as

trimethylgallium and triethylarsine have been used along with many others.35

The CVD of Ti02 has been widely studied. TiCI4 is the non-metalorganic

precursor which has received the most attention.9-11,26,36 Deposition with this precursor

has been motivated by two factors. First, TiCl4 is low cost in comparison to other Ti

precursors.9'36 Second, TiCl4 is used in the manufacture of TiO2 powder for white

pigment applications.37 The biggest disadvantage associated with this precursor is the

chlorinated byproducts resulting from the TiO2 formation reaction. Depending on the

precursor used with TiCl4, HCI or Cl2 can be formed. These compounds can degrade the

reactor and are considered harmful pollutants in the environment. The low cost advantage

of this precursor is therefore reduced by the cost of properly disposing of its byproducts.








2.1.2.1 Metalorganic chemical vapor deposition. Outside of the use of TiC14 as a

precursor, a large portion of CVD of TiO2 has been carried out with metalorganic

precursors. One of the advantages of metalorganic precursors is that film contamination is

minimized because the reaction byproducts are hydrocarbons in low concentrations.

Another advantage is that films can be deposited with relatively low-energy input to

separate the relatively weakly bound metal atom from the organic ligands. Several

metalorganic compounds have been used in research, with titanium isopropoxide
(Ti(OC3H7)4 chief among them.2728'38-52 This liquid precursor is normally transported

by an inert carrier gas or 02 while being maintained at a temperature of -100 OC. Other

approaches to delivery have been to heat the liquid to 270 OC and allow the reactor's

pressure gradient to force transport toward the substrate53 and to enhance the liquid's

dispersion in the reactor by pulsed injection using an ultrasonic nozzle.54 Titanium
ethoxide (Ti(OC2H5)4 is another liquid metalorganic precursor which has been employed

for the deposition of TiO2.5556 A summary of these CVD studies is that the film density

and crystalline phase of is most affected by substrate temperature and partial pressures of

the precursors.


2.1.2.2 CVD related techniques. Value exists in noting several non-traditional
CVD techniques which have been applied to TiO2 deposition. Atomic layer deposition

(ALD) consists of alternating the introduction of precursors such that nominally one
monolayer of TiO2 is deposited for each cycle. This technique has been performed with

both Ti(OC3H7)4 and TiCl4 precursors.10,13,55,57 The advantages of ALD are obviously a

high degree of control and the uniformity of film thickness. As a result, this technique can

be applied to substrates with very large surface areas. Some researchers substitute the term

epitaxy for deposition in ALD, although this can only accurately be applied when the film is

epitaxial as discussed later.







Sol-gel processing requires dramatically lower energy input. The deposition

consists of dipping a substrate into a titanium precursor and stabilizer mixture. The liquid

film is allowed to dry and is then heat treated at temperatures in the 100 C to 300 C

range.58 The advantages of sol-gel processing are the ability to use heat sensitive

substrates and low cost. The films are, however, amorphous and porous. The deposition

of TiO2 thin films on organic self-assembled monolayers is another technique which

requires low energy input.59,60 Sulfonate surface functional groups on Si wafers act as

nucleation sites for TiO2 from Ti precursors in solution. Uniform, polycrystalline anatase

films were grown from a TiCI4 in aqueous HCI solution and Ti(OC3H7)4 in non-aqueous

ethanol solution over a 80 to 100 C temperature range.


2.1.3 Heteroepitaxy


The word epitaxy is derived from the Greek words 'epi' which means 'resting

upon' and 'taxis' which means 'arrangement'.5 In the field of thin film deposition, the

term refers to the growth of a 'single' crystal film on a crystalline substrate. Homoepitaxy

is the growth of a film on a substrate of the same material. An example is the growth of Si

on Si. The advantages are that a film can be deposited with fewer defects and a higher

purity than the substrate. Also, there is a higher control of doping.6

Heteroepitaxy is the growth of a crystalline film on a crystalline substrate of a

different material.5'6 A well known application of heteroepitaxy is the manufacture of

compound semiconductor devices. In these devices, it is desirable to have multiple layers

of high quality crystalline materials which exhibit different electrical properties. The

pertinent issues are the minimization of interlayer diffusion and defect densities which

degrade the electronic properties. Interdiffusion is activated by temperature. Therefore, the

deposition of all layers must take place below a critical temperature to avoid interdiffusion.







Defects fall into five principle categories.7'35 An example is the propagation of a

substrate defect such as a screw dislocation into a depositing film. The atoms preferentially

deposit at the lower energy ledge sites of the dislocation continuing its pattern. The

continuation of a screw dislocation would be heavily dependent on the growth taking place

by a layer or two-dimensional mechanism. Of the five defect categories, there are two main

types of dislocations pertinent to the three-dimensional or island-type growth to be

presented in this thesis. They are the formation of low-angle grain boundaries and twins

and misfit dislocations. Low-angle grain boundaries and twins occur when islands of

different orientation with respect to the plane of the film coalesce. In the case of twins, the

islands coalesce such that their orientation is the mirror image of one another across the

boundary perpendicular to the plane. Misfit dislocations occur when the elastic strain

energy in a film caused by it being in compression or tension to fit the lattice of the

substrate becomes greater than the energy of the film with the inclusion of a defect. For

example, a film whose lattice in a direction is smaller than the lattice of the substrate will be

in tension in that direction. With layer upon layer of the film being in tension there is an

energy release at a critical thickness which relaxes the tension. As a simple illustration, a

relaxed layer could grow such that five unit lengths of the lattice match to four unit lengths

of the lattice in tension. This allows the film to grow with less strain on top of this misfit

layer. Further explanations, examples, and discussion of thin film defects, especially with

respect to the deposition of room temperature semiconductors, are available.7'35
Heteroepitaxial TiO2 thin films are desirable for their improved optical and electrical

properties. The use of an electrically insulating substrate allows one to probe and exploit
only the characteristics of the TiO2 without interference. The heteroepitaxial deposition of

TiO2 has been accomplished by several deposition methods. Representative of the

literature are growth by reactive ion cluster beam (RICB) deposition,22 ion-beam sputter
deposition (IBSD),14 CVD using a TiC!4 precursor9,26,36 and MOCVD using a

Ti(OC3H7)4 precursor.38-42,44 Of these studies, the crystalline quality of the films grown







by IBSD14 are representative or superior as judged by the x-ray diffraction (XRD) theta
rocking-curve full-width-at-half-maximum (rc fwhm). The 0 rc fwhm is a measure of the
crystalline alignment perpendicular to the plane of the film. Table 2.1 is a summary of the
three rutile orientations important to this work along with the parallel directions in the plane

and the calculated cationic mismatch.
Given the large mismatch values in Table 2.1, it is the opinion of many researchers

in the field that matching between the oxide sublattices as opposed to the cation lattices is
the driving force for the observed TiO2 growth orientation. Excellent schematic

illustrations can be found for (101) and (100) oriented rutile films on the respective

sapphire substrates,42 and values have been calculated for the (001) orientation.14 The
mismatch is dramatically reduced by this analysis as shown in Table 2.1.


Table 2.1 Cationic and anionic lattice mismatch for heteroepitaxial rutile
deposited on sapphire substrates


Rutile Sapphire In-plane crystal Cationic Anionic
lattice relationship Mismatch Mismatch


(001) (1010) [100] //[1120] 11.5% 3.6%
[010] //[0001] 6.0% 5.7%

(101) (1120) [101]//[1100] 14.8% 0.9%
[010]//[0001] 6.0% 5.8%

(100) (0001) [010]//[1210] 3.6% 3.8%
[001] //[1010] 3.0%* 7.3%

* Mismatch of 3 lattice units







2.2 Rutile Gas Sensors


2.2.1 Oxygen Sensors


TiO2 has received considerable attention as an oxygen sensor because of the

automotive industries search for a suitable oxygen sensor to be placed in an automobile's
exhaust system.4,61,62 The semiconducting characteristics of TiO2 at higher temperature

have been utilized for this purpose. Semiconduction occurs when Ti4+ and Ti3+ ions

coexist or synonymously when oxygen vacancies are present. In this regime the resistance
(R) varies with the 02 partial pressure P02 and the absolute temperature (T) according to

the following equation


n E/kT
R= Ro POe e (2.1)



where E is the activation energy for electronic conduction and the remaining variables are

constants.3,4,62 Therefore, any change in the oxygen partial pressure may be measured
because it directly affects TiO2's resistance.

Table 2.2 shows some representative oxygen sensors from the literature.61-64 The

second column of the table shows the form of the rutile sensor. Thick films are normally

more porous than thin films and are deposited as a printed paste and annealed to increase

uniformity.65 The term ceramic as used here is representative of materials that are pressed

from powders and sintered. The second column is the logarithm of the resistance change in
response to the change in 02 partial pressure shown in the third column. Note that

resistance increases with increasing PO2 as explained above. The polycrystalline ceramic

and thick film materials tend to be more sensitive due to higher available surface area. This

increased sensitivity suggests that a surface phenomena is important in the sensing

mechanism. The majority of oxygen sensor studies ignore specific surface characterization







of the materials as a means of understanding more about the interaction of the surface with

oxygen.



Table 2.2 Oxygen sensor examples



Response* -Log P02 (atm)
Material Form Log 10x (ohms) response' range T(C) Reference


Ti02 Thick Film 3-7 35-5 350 61

0.2-3.5 16-0 800 61

TiO2/9at% Nb Thin Film 2-4 3-0 1000 63

Ti02 Thin Film 2.2-2.4 4.3-1.7 1000 64

TiO2 Ceramic 3.5-8 30-10 400 62


2.2.2 Humidity Sensors


Deposited thin films and ceramic forms of TiO2 are commonly used as humidity

sensors for industrial processes and human comfort.66,67 The majority of TiO2 humidity

research has studied the ceramic form at or near room temperature. Several sources explain

the mechanism by which the humidity sensing proceeds at room temperature.66-68 The

first layer of water chemisorbs on a ceramic surface. Subsequent layers are physisorbed

and dissociate into hydronium and hydroxyl ions due to the high electrostatic fields in the

chemisorbed layer when an electric field is applied. Ionic conduction occurs when a

hydronium ion releases a proton to another physisorbed water molecule and this becomes a

chain reaction. The behavior of these sensors is governed by the number of chemisorption

sites and the porosity of the ceramic (i.e., surface area). Researchers have doped with ions

such as Li+, K+, and Cr3+ with some success in increasing the number of chemisorption








sites.67,69 Mixed results have been obtained when others have tried to improve sensor
response through engineering the porosity by mixing TiO2 with other oxides such as V205

and Nb205.68,70,71 The characterization of a humidity sensor at high temperatures where

the behavior and effects of chemisorbed water can be studied cannot be found in the

literature.

Table 2.3 displays data from representative research on humidity sensors.65,72-74

Column three shows the logarithm of the decrease in resistance of each sensor with the

corresponding increase in percent relative humidity (RH). Some researchers have chosen

to measure the DC resistance while others measure the AC resistance at a single frequency.

Using either of these approaches it is impossible to determine whether or not a change in

the material-contact barrier resistance or the actual material is being probed.



Table 2.3. Humidity sensor examples


Response* %RH range Response/
Material Form Log 10x (ohms) of response* T(C) Freq (Hz) Reference


TiO2/
0.5%Nb205 Ceramic 8-4 10-90 RT AC/400 65

Ti02 Ceramic 7-3 15 -95 25 AC200 72

Ba.5Sr.5TiO3 Thick Film 8-5 10-90 RT AC/400 65

TiO2 reactively
sputtered Thin Film 11-5 0-90 RT DC 73

85%Ti0215%V Thin Film 7-5 0-100 25 AC50 74







2.3 Rutile Properties


2.3.1 Crystal and Band Structure


TiO2 can exist in three crystalline structures: rutile, anatase, and brookite. The

phase transition from anatase to rutile occurs around 700 OC, and the reverse transition is

kinetically limited. The unit cell of rutile is tetragonal with a=4.595 A and c=2.959 A. A

picture of the unit cell is shown in Figure 2.1 with bond angles and lengths shown. Filled

circles represent Ti atoms in the figure. For rutile, the space group is D4 and symmetry is

P42/mnm.75


Figure 2.1 Rutile unit cell. From A. von Hippel et al.77, Fig. 2


1







A number of rutile's properties are of interest for gas sensor applications. First,

rutile has a very high refractive index and dielectric constant. The following equations give

the refractive indices as a function of wavelength, X (A):


n2 = 7.197 + 3.322x07 / (X 0.843x107) extraordinary (2.2)

n2 = 5.913 + 2.441x107 / (2 0.803x 107) ordinary.76 (2.3)


At room temperature, the low frequency dielectric constants are 170 and 86, and the high

frequency dielectric constants are 8.427 and 6.843, parallel to the c- and a-axes
respectively.75 Also, TiO2 is a wide band gap semiconductor. Controversy exists among

different studies as to the fundamental optical absorption edge in rutile. Values ranging

from 3.033 to 3.062 eV have been reported for the direct band gap perpendicular to the c-

axis. The transition is indirect parallel to c-axis, and values from 3.049 to 3.101 have been

reported. Both sets of values are at 1.6 K and 280 K, respectively.75


Finally, the intrinsic electrical conductivity is an important piece of information

which can be described by the equation


In a = A B/T (2.4)


where a is the conductivity in (ohm-cm) and T is the absolute temperature75

Perpendicular to the c-axis, A equals 7.92 and 11.10, and B equals 17,600 and 21,200,

below and above 1123 K respectively. Parallel to the c-axis, A equals 8.43 and 11.30, and

B equals 17,600 and 21,200, below and above 1223 K respectively.







2.3.2 Conductivity


2.3.2.1 Defects and mechanism. As described above, conductivity in pure TiO2 is

a function of temperature and oxygen partial pressure. The combination of these two

variables creates oxygen vacancies or titanium interstitials. These defects are believed to be

the source of conduction electrons. The electron density dependence on the oxygen partial

pressure can be determined from an analysis of defect equilibrium which is well-known in

the literature. For example, using Kroger-Vink notation oxygen vacancies are formed

according to the defect equilibrium equation78-83


Oox 1/2 O,(g) + iVk + ke' (2.5)



where k is the charge of the oxygen vacancy relative to the lattice and e' is a quasi-free

electron. At high temperatures k may have the value 1 or 2 representing a partially or

completely ionized oxygen vacancy, respectively. There is also the equation


null -> e' + h (2.6)


where h' represents a hole. Holes are, however, considered to be a minor defect in TiO2,

except at extremely high temperatures and oxygen partial pressures or high concentrations

of acceptor impurities.78'80 Outside of that realm, the corresponding equilibrium constants

for reaction 2.5 are


Ki = [V0] n P021/2 (2.7)

K2 = [Vo"] n2 P021/2 (2.8)









where k = 1 or 2, respectively. Brackets denote the concentration of the species. The

electroneutrality condition is then


n = [Vo[] + 2[Vo"] (2.9)



Solving for the electron concentration using equations 2.7, 2.8 and 2.9 gives


n3 = K1 n P02-1/2 + 2K2 P02-1/2. (2.10)

There are two limiting cases to equation 2.10: [Vo'] > [Vo"] and [Vo] >> [Vo*] which

represent singly ionized oxygen vacancies or doubly ionized oxygen vacancies being
predominant, respectively. If singly ionized oxygen vacancies are predominant then K1

K2, and the electroneutrality equation simplifies to



n = K11/2 PO2-1/4. (2.11)



If doubly ionized oxygen vacancies are predominant then K2 >> K1, and the

electroneutrality equation simplifies to


n = (2K2)1/3 P02-1/6. (2.12)



Given that the conductivity, o, is a function of n according to the equation



cr=ne (2.13)


one can see that the conductivity should exhibit a -1/4 or -1/6 dependence on PO2

depending on which defect is predominant.


- ---------







By the same analysis of the titanium interstitial defect formation equation78-83


TiTix + 2Oox Tiij + 02(g) + VTiJ' (2.14)


at equilibrium, the dependence of conductivity on P02 when partially ionized titanium
interstitials, Tii3, or fully ionized titanium interstitials, Tii4*, are the predominant defects

can be found. When [Tii3] > [Tii4'], oy a P02-1/4, and a PO2-1/5 in the reverse limit.

Obviously, the PO2 dependence of conductivity is the same in the limit of [Tii3] or [Vo]

being the predominant defect. These defects are expected only to be observed in very
nearly stoichiometric TiO2. Their observance is open to a great deal of speculation since

conductivity dominated by the presence of acceptor impurities exhibits the same P02

dependence.82,84
A review of representative literature reveals several important trends concerning the
dependence of TiO2 conductivity on temperature and partial pressure. Over an oxygen

partial pressure range from 100 to 10-20 atmospheres where the transition between the
terms "low P02" and "high PO2" is around 10-7 atm, these trends are displayed for single

crystal and ceramic samples in Table 2.4. The temperature ranges are guidelines given that

there must be some transition region between each case. Obviously, other potential defects
such as oxygen interstitials and titanium vacancies can be associated with TiO2. This table

is shown as a summary of the available literature and is not intended to imply that all

possible experiments will conform to these trends.






Table 2.4. Dominant impurity trends in TiO2


Sample Temperature Dominant Defect
Treatment Range (OC) low P02 high P2O


Unreduced80'84'85 1000-1200 Impurity, Tii4* Vo"

1200 1500 decreasing Tii4 Tii4*
impurity I I

Reduced79'83 800 1000 V o Vo"

1000-1200 Tii4* Vo"

1200 1500 Tii 4 Ti4

Highly reduced82 60- 225 Vo" V"

250 550 Ti,4" Tii4*


A summary of these trends found in the literature is that with increasing degree of
reduction, which is a function of higher temperatures and lower oxygen partial pressures,
titanium interstitials appear to become the predominant defect over oxygen vacancies.


2.3.2.2 Activation energy and mobility. In addition to studying the relationship
between the conductivity and defect chemistry of TiO2, there have been assessments of the

activation energy and electronic mobility of conduction. Table 2.5 is a summary of typical
research and the results.







Table 2.5. Representative activation energy and mobility data for TiO2


Sample Temperature Activation Electronic Mobility Comments
Treatment Range (OC) Energy (eV) (cm2/Vsec)


Unreduced86


Unreduced84


Reduced87


Reduced82


Reduced83


Reduced88


Reduced89


Reduced90

Reduced91


350 850
850-1400

1000 1500


1.53
1.83


0.1 10


1.9 (air)
1.5 (1 atm 02)


0.04- 0.08


25-350
350 600


0.2 (Ic)
1.0 (l|c)


0.19
0.50


<400
high T


25
60-300


25-325


a T-3/2
- constant


0.028


0.1 -1.0

1e6(T/K)-2-5


None on temp. dep. of p


None on p


>150K Ea(L) = 0.10 eV
Measurements in vacuum

Ea at all PO2
None on p

p ind. of defect type or con-
centration at const T >> p is
not a hopping process

p. fxn of acoustic phonon scat.
. weak fxn of T by optical
phonon scattering

Measurements in vacuum


p fxn of optical phonon scat.

p fxn of phonon scattering


The data show that there is a great deal of sample to sample variation even when tested

under the same conditions. The trends that can be observed are the differences between

unreduced and reduced samples. There is a dramatic change in the activation energy from

greater than 1.5 eV for unreduced samples to 0.5 eV and much less depending on the

degree of reduction. The activation energy of a highly reduced sample is attributed to the

conduction of free electrons affected by phonon scattering. This is the consensus of all

studies on reduced samples in the analysis of activation energy or mobility data.82'83'87-90







Conduction was not found to be a strongly activated or "hopping" process which would be

indicative of a polaronic influence. In the case of the unreduced samples, the high

activation energies are indicative of conduction by electrons which are trapped at defects

such as oxygen vacancies.84'86 After being activated into the conduction band, they may

conduct long distances without being retrapped.


2.3.2.3 Intrinsic band-gap states. To separate the activation energy of conduction

from the mobility term, others have performed investigations into the electronic nature of

defects as to their place in the band structure. Several modeling efforts have identified
electronic states in the band gap ofTiO2 resulting from oxygen vacancies. Without respect

to crystal orientation, levels were calculated to be 0.7 eV below the conduction band

edge.92-94 Oxygen vacancy band gap states were calculated at 0.95 eV below the

conduction band for the (001) surface and 1.78 eV for the (110) surface.95 Investigations

on single crystals by ultraviolet photoelectron spectroscopy found oxygen vacancy band

gap states around 1 eV below the conduction band edge for (001)- and (100)-oriented

single crystals.96'97 Angle resolved photoemission spectroscopy also identified an oxygen

vacancy state at approximately 1 eV for an (001)-oriented single crystal.98,99 Finally, an

analysis of thermally and photonically induced conductivity data revealed an electronic state

0.62 eV below the conduction band edge due to an oxygen vacancy.0


2.3.3 Doped rutile


2.3.3.1 Donor and acceptor defect chemistries. In addition to the oxygen vacancy
and titanium interstitial defects created in pure TiO2 as a function of temperature and

oxygen partial pressure, defects may be created by dopants. Dopants that are of interest for
their effects on TiO2's electrical properties include both donors and acceptors. Along with

the defect equations 2.5, 2.6, and 2.14 for the creation of oxygen vacancies and titanium








interstitials in pure TiO2, the addition of dopants can create these same defects and others.

Ga in its most stable oxidation state, 3+, is a representative acceptor atom with respect to
the TiO2 crystal because the Ti cation is in a 4+ oxidation state. The inclusion of Ga in the

crystal substitutionally on a Ti site may be compensated by oxygen vacancies or titanium

interstitials according to the following defect formation equations:231'232


Ga203 +- 2GaTi' + 3Oox + Vo" (2.15)


and TiO2 + 2Ga203 -> 4GaTi' + 80ox + Tii4*. (2.16)



Nb in its most stable oxidation state, 5+, is a representative donor atom with respect to the
TiO2 crystal. The inclusion of Nb in the crystal substitutionally on a Ti site may be

compensated electronically or by titanium vacancies according to the following defect

formation equations:116,232-234


Nb205 -> 2NbTi' + 40ox + 2e' + 1/2 02(g) (2.17)


and 2Nb205 <- 4NbTi + VTi4' + 1000x. (2.18)


2.3.3.2 Structure and conductivity effects. The doping of TiOz with Ga has rarely

been investigated, however, doping with a variety other trivalent atoms has been. A

notable study investigated which valence state of a number of atoms would be most stable

in the band gap using a two body shell model.100 The 3+ valence state of V, Fe, Cr and

Ga were found to reside in the band gap. It is significant that in the case of Ga and Fe, 3+

is also a stable, naturally occurring valence state. It was also calculated that the most

energetically favorable lattice site for a Ga atom is in substitution for a titanium atom as

opposed to being interstitial. Another modeling study calculated that the 2+ valence state of







Cr and Mn and the 3+ valence state of V and Mn to be in the band gap approximately 1 eV

below the conduction band edge. Fe2+ was calculated to be in the conduction band.101 Of

these four atoms, only Mn3+ residing approximately 1.8 eV below the conduction band and

Fe2+ in the conduction band could be verified by experiment.
Aluminum is a common impurity in TiO2 single crystals, so it has received the most

attention in analyses of electrical conductivity data for trivalent atoms. Representative

studies have been performed at temperatures above 1000 OC where titanium interstitials are

known to be the dominant defect as was shown in Table 2.5.64,80,102 The data revealed
that Al doping decreased the n-type conductivity of TiO2 at low PO2 and resulted in a

transition to p-type conductivity at high PO2. The decrease in conductivity resulted from Al

acting as an electron acceptor deep in the band gap with no discussion of the titanium

interstitial effect. By explaining the data in terms of titanium interstitials, these studies have

assumed Al to be substitutionally incorporated via equation 2.16. There is some

disagreement, however, from x-ray diffraction and thermogravimetric studies which

concluded that Al is incorporated interstitially.103,104 Other trivalent dopants like Cr3+
Fe3+, and Mn3+ in TiO2 also exhibit a decrease in conductivity and transition to p-type

behavior at high temperatures and oxygen partial pressures.105-109

A great deal of work has been done on the electrical and structural effects of doping
TiO2 with Nb. The modeling work mentioned earlier further calculated that the Nb5+

valence state is stable in the band gap, and it is energetically favorable for Nb5+ to

incorporate substitutionally on a Ti site in the lattice.100 Other studies confirmed the
substitutional replacement of Ti with Nb by density measurements as a function of PO21 10

and thermogravimetric experiments.111 The rutile structure was observed to accommodate

seven to eight atomic percent Nb before creating a secondary phase. 11-116 Since Nb5+ is

a larger ion than Ti4+, the lattice volume of substitutionally doped crystals would be

expected to expand. This was the finding when investigated.110'111,114







The defect equations 2.17 and 2.18 show that Nb can be compensated electronically

or by titanium vacancies. A dramatic increase in conductivity was attributed to electronic

compensation.116-118 In one case, an activation energy of 0.08 eV was measured around

room temperature. Mobility values of approximately 0.8 and 0.5 cm2/V*sec were

measured at 25 and 225 OC, respectively. Mobility was only found to be activated below

60 K.87 The donor ionization energy measured at and below room temperature for another

sample was 0.03 to 0.12 eV. In this instance the energy was thought to be characteristic of

electron ionization from a Nb donor level. Mobility was shown to be inversely related to

temperature due to scattering mechanisms. 118



2.3.4 UHV Surface Characterization



Research in the field of heterogeneous catalysis has motivated the surface

characterization of metal oxides. The goal is to understand the relationship between

specific surface characteristics and the bonding and reactivity of adsorbed species.119-123

Among other applications, this type of research has permitted substantial advances in the

photo-oxidation of carbon monoxide and methyl chloride for the reduction of

environmental pollutants.124127

Two key adsorbents of interest are oxygen and water. The common practice in this
field is to characterize single crystal TiO2 surfaces reduced by heating, heating with

exposure to H2, or noble gas ion bombardment in comparison to unreduced surfaces. The

type of defects present and their adsorption characteristics are investigated by a very large

number of techniques.123 The literature indicates that Ar ion sputtering is the most reliable

means of adequately reducing the surface which results in a high enough defect density to

be detected.119







2.3.4.1 (11 0)-oriented rutile. The (110) surface of TiO2 is the most stable because

Ti atoms in the surface plane are five coordinated.95'128 Therefore, this orientation has

received a greater amount of investigation than others. Representative work on (110)

surfaces shows that oxygen was chemisorbed at oxygen vacancies on a reduced
surface.91'96'129'130 Both molecular, 02-, and dissociated, O2-, oxygen species were

observed on the surface.91'96 Water was also observed to adsorb dissociatively on (110)

with both a large and small amount of defects, suggesting that no combination of

neighboring defect sites was required to induce the dissociation.130-134 Using a more

highly sensitive technique, synchrotron radiation photoemission, an apparent charge

transfer from the adsorbed species to the material was observed.131


2.3.4.2 (100)-oriented rutile. The (100) orientation of TiO2 has also been the

subject of a few studies. Both oxygen and water were found to physically adsorb in a

disordered manner on the stoichiometric surface. On the reduced surface, however,
oxygen was observed chemisorbed as 02-, behaving as an electron acceptor.135 Water

was also observed to dissociatively chemisorb on the reduced surface.135'136

Nondestructive SIMS detected a number of different ways in which the water was bound to
the surface including OH-, O2H-, TiOH+, and TiOOH+. A thermal programmed

desorption investigation of this surface found evidence for both chemisorbed and

physisorbed water on a partially reduced (100) surface.137


2.3.4.3 (001)-oriented rutile. The (001) surface of TiO2 is not a naturally occurring

face because of its relative instability compared to the (110), (101), and (100) faces. On an

unfaceted (001) surface the Ti atoms are only four-coordinated as compared to five or six

coordination for (110) surface Ti's for example.128 Due to this relative instability, the

(001) surface has been observed to reconstruct to form { 101) facet planes when annealed

below 1200K and { 114} facet planes when annealed at higher temperatures.97'128'138







These planes consist of five coordinated Ti's which is the driving force for their formation.

The defects on this interesting surface structure have been thoroughly investigated as a

function of different reducing and annealing techniques.119,123 Investigation of oxygen

adsorption on this surface as to the molecular or atomic structure of the adsorbate has,

however, been generally ignored. Water, on the other hand, has been observed to adsorb

dissociatively on the reduced and {101 facet plane reconstructed surface. Both

spectroscopic and thermally induced desorption techniques were used to reach this

conclusion. The adsorbed species was detected as OH-. There was further evidence that

the excess H atom bonded to a surface O for a net formation of two hydroxyl radicals per

adsorbed water molecule. 139-141



2.4 Other Metal Oxides



2.4.1 Gallium Oxide



2.4.1.1 Deposition. Several low-temperature, metastable phases of gallium oxide

have been reported. The hexagonal a-phase is the only one to have been consistently
reported.142'143 The only stable form of Ga.3 is the P-phase which has a monoclinic unit

cell.142-145 The dimensions are a = 12.230.02, b = 3.040.01, and c = 5.800.1 A

with an angle of 103.70.30 between the a and c axes. Care must be taken in reviewing

this literature due to some authors reversing the designation of the a and c axes. The unit
cell consists of four Ga203 molecules. There are two crystalographically non-equivalent

Ga atoms in the unit cell. One is surrounded by a distorted tetrahedron of O atoms, and the

other is surrounded by a severely distorted octahedron of O atoms. The coordination of the

Ga atoms results in three non-equivalent O atoms in the unit cell.
Ga,03 has been deposited by many techniques including MBE,146 thermal147 and

e--beam148 evaporation, spray pyrolysis,149 ALD,150 IBSD,151 and rf magnetron








sputtering.152 No deposition of Ga203 by CVD has been reported. Only ALD and IBSD

resulted in the deposition of any crystalline material. The films were found to be mixtures

of both amorphous and polycrystalline material when deposited at 430 to 520 C and 500

C substrate temperatures, respectively. A portion of the polycrystalline material in the

IBSD film was identified as P-phase. Annealing in air at 800 C and above converts all

material to the P-phase.149,151


2.4.1.2 Gas sensitivity. The most common application of Ga203 is gas sensing.

Ga2O3 in thin and thick film and ceramic forms has been investigated as a sensor of

oxidizing gases,152-154 reducing gases155 and various hydrocarbons.156-159 Because the

proposed environment is a combustion gas products stream, the sensor should be stable

with respect to high temperatures. Therefore, research for this application has focused on
obtaining P-phase material on substrates which do not interdiffuse with the Ga203.

Among the substrates tested, BeO was found to be superior by exhibiting no detectable
interdiffusion with the Ga203 below 1200 C.151 The deposition of thin layers of Ga203

are also of interest as an As free layer on the surface of GaAs.150,160 The presence of As

in the oxide surface layer has been shown to degrade the electrical properties in metal-

oxide-semiconductor field-effect transistors.


2.4.2 Aluminum oxide


Aluminum oxide is well-known as an electrically insulating material. This property
is utilized in thin layers of A1203 in a multi-layered electronic device.161"163 Al203 has

been deposited by a number of techniques including MBE,164 CVD,165

MOCVD,162'163.166-169 e--beam evaporationl0 and ALD.171 The majority of these

efforts have resulted in amorphous films, and the electrical properties of amorphous films

are satisfactory for most applications. In one investigation, however, a heteroepitaxial film




28

of y-phase A1O03 was grown on (111) Si at 750 to 900 C substrate temperature by MBE

to serve as the substrate for a semiconducting layer.164 In another, a polycrystalline A1203

film was deposited on (100) Si over a 700 to 900 C temperature range by CVD using
AICI3, CO2 and H2.165













CHAPTER 3
EXPERIMENTAL METHODS


3.1 Thin Film Deposition


3.1.1 Reactors


Two reactors were employed to grow the thin films. A schematic diagram

representing the essential components of both is shown in Figure 3.1. The majority of the

undoped work investigating growth rate as a function of temperature and precursor partial

pressures was done in a prototype reactor. This reactor was constructed from quartz glass

to test the applicability of an inverted pedestal, hot-walled reactor to solid-source MOCVD

for the growth of metal oxides. The idea for using solid precursors in metal oxide film

deposition originated at Hewlett-Packard.172 The simple design of the quartz reactor

necessitated that it be opened to the atmosphere to introduce or remove precursors and

substrates between runs. A mechanical pump was used to reach a base pressure of less

than 10 mTorr prior to beginning the deposition process. The standard growth conditions

in this reactor were 730 OC substrate temperature, 2.5 Torr 02 partial pressure, 3.9 mTorr

Ti precursor partial pressure, 5.5 Torr total pressure, and a total flow rate of 300 seem.

Table 3.1 shows the value range each variable was varied over in that study. When 02

partial pressure was decreased in this study, He flow rate was increased to maintain a total

pressure of 5.5 Torr. This reactor was also used to deposit gallium oxide and aluminum

oxide thin films. Growth variables within the ranges shown in Table 3.1 were used in the

deposition of both these types of films on a variety of substrates.







Table 3.1 Growth variable ranges for quartz reactor growth study


Substrate temperature

02 partial pressure

Ti precursor partial pressure

Total pressure


To traps &
mechanical
pump













02 feed


400 775 C

0.0 2.5 Torr

0.39 20 mTorr

5.5 Torr


Controllable
vertical
translation
rate



Cooling
water
jacket


Figure 3.1 Solid source MOCVD reactor schematic



After successful exhibition of metal oxide thin film deposition in this type of

reactor, a second reactor was constructed of stainless steel. This reactor used both a

mechanical and a turbomolecular pump to maintain a base pressure of 10 pTorr and will

heretofore be referred to as the HV reactor. Substrates were introduced through a load-lock

system. A precursor was placed in an apparatus that could be removed from the reactor

and reattached using a secondary mechanical pump without breaking system vacuum. The







reactor was equipped with three separate precursor introduction arms, each with this

removable piece, which allowed for deposition with multiple metals or other non-gas phase

precursors. The load-lock and precursor introduction modifications allowed the reactor

body to remain isolated from the atmosphere except in the case of maintenance and

cleaning. This minimizes the introduction of excess humidity and other gaseous and solid

contaminants.

In reference to Figure 3.1, both reactors can be described in three basic sections:

precursor introduction, growth chamber and exhaust system. Solid precursors, which will

be described in detail later, were used as sources of the metal atoms. Precursors were

delivered to the reactor by passing the solid material through an optically heated zone, at

which point the material was flash sublimed in a He carrier gas. A weighed amount of the

precursor was first tightly packed into a hollow glass rod with an internal diameter of

approximately 2 mm. This step was performed in a dry box with an Ar atmosphere to

avoid any potential reactions between a precursor an air or moisture prior to placing the

precursor in the reactor. The packed precursor's height in the rod varied 3 % suggesting

that uniform packing density was a reasonable assumption. Using this height, a

precursor's molecular weight, the mass of material in the rod, and the desired molar

addition rate, a simple calculation determined the necessary translation rate of a packed

glass rod.

Precursor introduction for deposition began by lowering a rod at a set rate from the

water cooled portion of a precursor introduction arm (Figure 3.1) into the focused light of a

Xe filament bulb. Sublimed material then escaped through a lengthwise slit in the glass

rod. Carried by He gas, the now gas-phase material flowed through the remainder of a

fully heated arm to the growth chamber. The sublimation process was allowed at least 15

minutes to reach steady-state while 02 was fed to the top of the reactor's growth chamber.

Both were exhausted at the bottom of the chamber. Steady-state sublimation could also be

confirmed by viewing a distinct point in a lowering rod below which all precursor material







was gone. This visual confirmation was done through welder's glass to avoid eye damage

from a lamp's bright light.

Upon reaching a steady state precursor delivery, the 02 introduction and exhaust

were switched to the bottom and top of the growth chamber, respectively. The 02 and

precursor could then meet at the bottom of the reactor and proceed toward the substrate via

the pressure gradient established across the growth chamber. All portions of the reactor

prior to the cold trap on the exhaust line were nominally heated to 2750C to avoid precursor

condensation prior to the growth surface. This temperature was assumed to be low enough

to avoid an appreciable amount of wall deposition as evidenced by a lack of change in the

general appearance of the chamber walls viewed during maintenance and cleaning. All

walls in both reactors were heated by wrapping with resistively heated strips of insulating

material ("heat tape"). The temperatures were controlled by Northrup programmable

controllers for the HV reactor and directly by setting the current flow to the heat tape on the

quartz reactor. Temperatures were monitored by placing J-type thermocouples at a number

of spots on a reactor's body.

Substrates were attached to a permanent nickel base using Ag paste (Ted Pella), so

that the entire substrate heating element was removed for each run in the quartz reactor.

For the HV reactor, substrates were attached to a nickel plate which was introduced via a

load-lock system as mentioned previously. This plate was brought into physical contact

with a permanent nickel plate which the heater was behind. After a great deal of trial-and-

error heater design, a Ta wire coil vacuum sealed in quartz glass was found to have a long

lifetime. This type of heater was constructed at the DuPont Experimental Station glass

shop. The heaters were controlled by Fuji fuzzy-logic type controllers. The temperature

was monitored by an S type thermocouple, and the actual substrate temperature was

calibrated using an optical pyrometer. A nominal loss of 150 C was found between the

temperature measured by the thermocouple which was buried in the permanent Ni plate and

the real substrate temperature.







3.1.2 Precursors


The precursors used are of a family of compounds known by their common ligand
2,2,6,6-tetramethyl-3,5-heptane dionato. A schematic of the Ti precursor of this family is
shown in Figure 3.2. Note that the metal atom is attached to oxygen atoms. These
precursors were purchased from STREM chemicals in powder form. They have a minimal
room temperature vapor pressure. TiO2 films were grown using the Ti precursor and
doped with Ga or Nb by directly mixing the precursor powders at the desired dopant level.
Ga203 and A1203 films were also grown from their respective metal precursors.








HC HH, CH, HC H, 3CH

HCH
H3C 3

HC CH


H43c-, -M HC--''-

Figure 3.2 Precursor schematic


The notable characteristic of these precursors is their well-defined sublimation
temperature. A thermal gravimetric analysis (TGA) was performed on several of the
precursors. A TGA is a simple experiment which consists of monitoring the mass of a
precursor on a combined scale / hot plate as the temperature is ramped upward. For each







precursor tested, approximately 90 % of a 7 to 10 mg mass was sublimed within a 10 C

window. Table 3.2 shows the chemical formula, molecular weight and sublimation

temperature for the precursors used in the growth studies of this thesis. TGA's of the Fe
and Cr precursors were not performed. When Fe(TMHD)3 or Cr(TMHD)3 was mixed

with Ti(TMHD)3 at one atomic percent, clean sublimation behavior from the precursor rod

was observed in the same way as for the Ga and Nb precursors. This suggests that their

sublimation temperatures fall in the range of 250 to 350 C in agreement with the precursors

for Ga and Al which are also stable in the 3+ oxidation state.


Table 3.2 Sublimation temperatures of metal precursors determined by TGA


Chemical Molecular Weight Sublimation
Formula (g/mol) Temperature (oC)

Ti(TMHD)3 597.7 350

Ga(TMHD)3 619.5 300

Nb(TMHD)4 826.0 375

AI(TMHD)3 576.8 260

Fe(TMHD)3 605.7

Cr(TMHD)3 601.8


3.1.3 Substrates


A variety of substrates were used in the growth studies. TiO2, Ga203 and A1203

films were deposited on one or more of the following: Si, quartz and sapphire substrates.

For the majority of the growth data presented in this thesis, the substrate preparation was

limited to a degreasing which consisted of consecutively wiping a substrate surface with






cotton swabs soaked in acetone, methanol and ethanol. Later in this work, however,

atomic force microscopy (AFM) became routinely available. AFM images showed that

although the historical degreasing procedure was capable of removing most surface

contaminants, some could remain. In addition, salts were found to potentially precipitate

on a surface as a function of the cleaning procedure and reagent purity. The adequacy of a

number of substrate preparation techniques were characterized by AFM. The following

recipe was found to be capable of removing all organic surface contaminants and deterring

the precipitation of salts. First, a substrate was doused with high purity methanol.

Second, a piece of low-lint lens paper was used to scrub the surface but not dry it before

redousing with methanol. Next, the substrate was doused with deionized water and

immediately blown off with pressurized air which had been filtered for particles greater

than 0.2 microns. A good sign that a surface was clean was beading of the deionized

water. This procedure was used to clean the substrates for all the film morphology

investigations presented in this thesis.


3.1.4 Ion-Beam Sputter Deposition


To better analyze some of the data presented in this thesis, comparisons are made to

TiO2 films deposited by ion-beam sputter deposition (IBSD) by Morris-Hotsenpiller in the

same laboratories. The IBSD reactor had a base pressure of 0.1 tTorr. Like the HV

MOCVD reactor, substrates were introduced via a load-lock system to maintain system

vacuum. Substrates were Ag pasted on a thin Ta plate which was heated from behind by

halogen lamps. The temperature was monitored by a thermocouple after being calibrated

with an optical pyrometer. Deposition occurred when a high-purity Ti target was sputtered

with a Xe ion beam from a 3 cm Kaufmann-type source in the presence of 02. The ion

beam energy was 1 keV, and the current was 20 mA. The total pressure during deposition







was 0.2 mTorr due to equal partial pressures of Xe and 02. The data presented in this

thesis are from films grown at 725 OC and approximately 3 A/min for 12 hours.


3.2 Physical Characterization



3.2.1 X-ray Diffraction



When atoms are arranged periodically on a lattice as they are in a crystalline solid,

the x-rays scattered by them have a defined phase relation. Most incident beams are

scattered in random directions and destructively interfere with one another. A diffracted

beam is formed when constructive interference occurs in a scattering direction. Given that

the incident and diffracted beams and the normal to the plane of reflection are coplanar, a

diffracted beam is formed for all scattered beams which satisfy Bragg's law



=2d sin0 (3.1)



where X = x-ray wavelength

d = spacing of atomic planes

0 = angle between incident or diffracted

beam and the atomic plane.

The larger the number of scattered beams which satisfy this equation, the higher the

intensity of the diffracted beam. The relationship between the planar spacing d and the unit

cell dimensions of a crystal is dependent on the unit cell geometry. Rutile has a tetragonal

unit cell and the plane-spacing equation is


l/d2 = (h2 +k2)/a2 + 12/c2


(3.2)







where h, k & 1 = crystal plane indices

a = 4.595 A, dimension of rutile unit cell

c = 2.959 A, dimension of rutile unit cell.

Combining equations 3.1 and 3.2 gives


sin20 = X2/4 [(h2 +k2)/a2 + 12/c2] (3.3)


which relates 0 to the indices of the planes) from which the beam is diffracting. If the

values of h, k and 1 result in sin20 = 1 then diffraction from the plane with those indices is

impossible. XRD tables are typically given in (hkl), 20 and d.

The typical diffractometer is capable of one geometry where the plane of incident

and diffracted x-rays is perpendicular to the surface plane of the film and the x-ray source

moves with respect to the film. This set-up was used in a Rigaku diffractometer for this

thesis to check the bulk orientation and phase of all deposited films. The consistency of

plane-to-plane spacing can also be assessed with this set-up by holding 20 constant and

varying 0. This is known as the 0 rocking curve (rc). The full-width-at-half-maximum

(fwhm) of the 0 rc peak gives a relative measure of the consistency of plane spacing. Other

diffractometers are capable of moving the film with respect to the x-ray source. An

important capability of this type of machine is the assessment of heteroepitaxial film growth

and the in-plane crystal direction relationships between a heteroepitaxial film and its

substrate. As shown in Figure 3.3, when a film is moved so that the surface normal is no

longer in the plane of the x-ray beam, the angle through which the film is tilted is known as

X. In a simple cubic lattice for example, the (101) plane would have a X value of 450 with

respect to the (001) plane. If in this position, a film is then rotated about the normal to the

(101) plane, the angle of rotation is known as (. If all planes of a film have the same in-

plane orientation, the film is said to be heteroepitaxial. In the simple case of a cubic lattice,

the diffraction pattern of a bulk-oriented film moved to X = 450 and rotated through the 3600







range of < would have four distinct peaks separated by 900 if the film is heteroepitaxial.

This scheme was used to investigate the heteroepitaxial characteristics of rutile films on

sapphire substrates and the matching in-plane directions between the two crystals. The

fwhm of peaks were also measured as a relative indication of the in-plane orientational

consistency for some films. This work was performed on a Philips diffractometer.

Thorough treatments of these topics and other XRD methods and applications are available

in the literature.173,174



normal to X
0 x-ray plane*"
Incident \r
Incidxbea- x-ray plane in plane
x-ray beam of film


diffracted beam

Figure 3.3 Thin film x-ray diffraction geometry


3.2.2 Rutherford Backscattering Spectrometry


Rutherford backscattering spectrometry (RBS) is a simple experiment in theory

which requires some very large and high voltage equipment. A beam of monoenergetic and

collimated alpha particles is impinged perpendicularly on a target. The particles which are

scattered backward by angles of more than 900 from the incident direction can be detected.

Figure 3.4 shows the basic components of an RBS experimental system.







Ion source Quadrupole
I- focusing magnet Magnetic
I -I analyzer

Slits
Accelerator
-- -Slits
Analog and []
digital electronics
Sample
PreamplifierSam
and detector"


Vacuum
pump


Figure 3.4 Schematic of essential Rutherford backscattering equipment


Charged particles are generated in the ion source. Their energy is raised to several

MeV by an accelerator. The beam then travels through the quadrupole focusing magnet,
magnetic analyzer and slits which together serve to collimate or focus the beam and filter it
for the desired type of particle and energy. Next, the beam enters the scattering chamber
and impinges on the sample. Some of the backscattered particles hit the detector. An

electrical signal is generated by these particles which is amplified and processed by the

analog and digital electronics. The resulting data is in the form of a spectrum which is why
the technique is known as spectrometry. To gain information from the spectrum, it must be
modeled.
RBS for this thesis was performed at the Laboratory for Research on the Structure
of Matter (LRSM) which is cooperatively managed by several science and engineering
departments at the University of Pennsylvania. The equipment is maintained by LRSM and

for a fee users do the most basic processes of taking data and leave with a spectra file for
each sample. The ions in the LRSM equipment are 4He+ and typically have an energy of 2







MeV. This information along with the defining angles of the source-sample-detector

geometry and the spectra are contained in each data file. Figure 3.5 shows a typical

spectrum for a TiO2 film deposited on sapphire (A1203) with an applied model (smooth

line). The spectra is in the form of peak intensity as a function of energy. The greater the

mass of an element, the higher on the energy axis it appears. From left to right, the first

knee is the oxygen peak, the second is the viewable end of the aluminum peak and finally

the titanium peak can be viewed entirely.

A modeling program known as RUMP was developed at Cornell University for

analyzing RBS spectra. The high energy or front edge of a peak for any atom has a well-

defined energy as a function of the beam energy and equipment geometry. The spectra

must be calibrated to fit the model and therefore the Ti and 02 peaks were used for

calibration of TiO2 films deposited on sapphire. The substrate is modeled as 50,000 A or

infinitely thick with respect to the film. By inputting the atomic density of the crystalline

film, the thickness of the film corresponds to the width of the Ti peak. The model is

believed to be accurate to within 50 A. As the film becomes thicker, the peaks increase in

width toward lower energy. At approximately 5000 A the Ti peak will intersect the Al

peak and modeling becomes more difficult and less accurate. The relative amounts of

atoms correspond to their peak height. For example, the model of a heavily oxidized TiO2

film would show the model peak as being higher than the actual Ti peak because Ti and O

are not at a 1:2 ratio as input to the model. Using this same analysis any dopant which is

distributed throughout the film will have a peak of the same width as the Ti peak. The

height of the dopant peak defines its concentration relative to Ti. Dopant concentrations can

be accurately modeled within 0.1% down to approximately 0.1% where the peak becomes

indifferentiable from the signal noise. The model in Figure 3.5 shows that the film is not

significantly oxidized or reduced and is 2400 A thick. A complete treatment of RBS and its

other abilities and applications is available in the literature.175









25


20


Energy (MeV)
1.0


0.5


5 -


0 -
50


1.5


Channel


Figure 3.5 RBS spectra of TiO2 on sapphire with RUMP model


3.2.3 Atomic Force Microscopy


Atomic force microscopy (AFM) was used to measure the size and shape of surface
features of the thin films. AFM can be used in a number of different modes to investigate
surface morphology with the same equipment. A schematic of the equipment set-up used
to take images for this thesis is shown in Figure 3.6. The essential parts of the equipment
are the physical measurement apparatus consisting of the laser and tip, the conversion of
the signal caused by changes in the reflected laser beam to information about the tip
movement, and the feedback control loop which seeks to keep some measure of a physical







interaction between the tip and the surface constant. The mode of investigation defines

which physical interaction is held constant.

With respect to the equipment shown in Figure 3.6 contact AFM (the tip is

theoretically in constant contact with the surface) can be used in constant force or constant

height mode. In constant force mode the vertical position of the tip would be continually

adjusted to avoid deflection of the tip by the surface. The height adjustment information

would be used to create the surface image. In constant height mode, the vertical position of

the tip is not changed as it is moved across the surface. Deflections caused by the tip

coming into contact with surface features are the information recorded and a control loop is

not actually used. This technique is only useful for extremely flat surfaces because large

features can cause very large deflections in the tip causing it to lose contact with the

surface. Also, the tip has a limited vertical range when not being purposefully moved, so

the tip could be damaged by large features in this measurement mode. The tip deflection

information would be used to create the surface image. In a non-contact mode, the

attractive van der Waals force which is present when the tip is close to the surface is used.

The cantilever which is connected to the tip is vibrated at a slightly higher than its resonance

frequency. As the tip approaches a surface feature out of the plane of the surface, for

example, a change in the intensity of the attractive force changes the amplitude of the

vibrational frequency. The height of the tip would be adjusted to maintain constant

frequency, and as in constant force mode contact AFM, the height change information

would be used to create a surface image.

Tapping-mode (TM) AFM is a hybrid between the modes discussed thus far.

Tapping-mode consists of an oscillation of the tip normal to the plane of the surface. This

intermittent contact technique has been discussed in detail in the literature.176-179 TM AFM

allows one to circumvent the adhesion and capillary forces which lead to uncontrollable

high tip-sample contact forces in conventional contact-mode scanning AFM.176 These

forces can lead to surface damage in some materials. In tapping-mode, height data are








complemented with simultaneously measured phase-shift data. Topography is measured as

the tip height is adjusted to retain a constant amplitude of the oscillating tip. "Phase" is

measured as the phase shift between the drive signal from the feedback control loop and the

actual tip response signal. Phase shift images are particularly useful for imaging the hard

and soft segments of copolymers and the faceted surfaces of crystalline thin films. Figure

3.7 shows a normal image (a) and a phase shift image (b) of a Ga203 film. The different

crystal faces of the film's faceted structure are more clearly exhibited in the phase shift

image.


Figure 3.6 AFM equipment schematic (Courtesy of R. S. McLean)







To obtain the images shown in this thesis, TM AFM was performed by S. McLean

at the DuPont Experimental Station to obtain height and phase imaging data simultaneously.

The piece of equipment used was a Nanoscope IlIa AFM from Digital Instruments.

Microfabricated cantilevers or silicon probes (Nanoprobes, Digital Instruments) with 125

gm long were used at their fundamental resonance frequencies. These frequencies varied

from 270 to 350 kHz depending on the cantilever. Cantilevers had a very small tip radius

of 50 to 100 A. The AFM was operated in ambient conditions with a double vibration

isolation system. Extender electronics were used to obtain the height and phase

information simultaneously. The lateral scan frequency was about 1.2 Hz. The images

presented in this thesis are not filtered.



















Figure 3.7 TM AFM images of a Ga203 thin film
(a) normal and (b) phase shift


The primary source of error associated with AFM imaging of hard, crystalline

surfaces is the size and shape of the tip. Surface features which are indented with respect

to the surface plane and are smaller than the tip size cannot be fully probed. Also, small

features which are out of the surface plane and are smaller than the tip will appear to have

the same shape as the tip. These errors can lead to the groups of small features appearing







as one larger feature. In addition to size limitations, the sharpness of the tip can cause

features such as facets which intersect at well defined to be rounded if the angle of

intersection is greater than the angle of the tip.


3.2.4 X-ray Photoelectron Spectroscopy


X-ray photoelectron spectroscopy (XPS), also known as electron spectroscopy for

chemical analysis (ESCA) is a widely used method for assessing the chemical composition

of surfaces. Surface analysis by XPS involves irradiations of a solid sample with

monoenergetic x-rays and energy analyzing the emitted electrons. Since the mean free path

of electrons with 100 to 1000 eV energies is small, only the electrons which are emitted

from the top few layers are detected.180 XPS is an excellent tool for determining the

composition of a multicomponent system because every element has a unique spectrum.

The exact position of spectral peaks can be used to determine the chemical state of the

element within the solid, which makes XPS able to discern the chemical state of elements in

a multicomponent solid.181-183

Monoenergetic x-rays (Al Ka, 1486.6 eV) interact with the surface atoms by the

photoelectric effect, causing the emission of electrons. Applying the conservation of

energy, one can relate the kinetic energy (KE) of the emitted electron to the binding energy

(BE) of the electron in the atomic orbital where it originated by the following equation


KE = hv BE s (3.4)


where hv = photon energy

Os = work function.183
The electron energy analyzer collects a spectrum which represents electron intensity as a

function of the electron kinetic energy. The electron energy analyzer acts as a filter, only







allowing electrons with a given kinetic energy to pass through to the detector (electron

multiplier). The analyzer allows electrons within a range of energies (pass energy) about a

given kinetic energy to pass, therefore the pass energy is very important for determining the

breadth of spectral features. Increasing the pass energy results in an increase in electron

detection and peak width, therefore pass energy must be optimized to produce a spectrum

which still contains adequate intensity without sacrificing spectral resolution. To maintain

adequate resolution of a surface component with a small concentration, multiple scans of a

given region are often performed and the spectra are averaged.

XPS was performed on a few samples for this thesis by V. Lusvardi at the DuPont

Experimental Station. The samples were doped and undoped rutile thin films. The tests

served a few purposes. First, near surface concentration of dopants was measured to

investigate their diffusion during the deposition process. Second, the oxidation state of the

dopant atoms was investigated in light of discrepancies in the literature on this matter,

especially for Nb. Third, the oxidation state of atoms at the surface was investigated to

determine if reduced cations could be identified in as-grown thin films within the resolution

of this technique. Finally, the work served Lusvardi by helping to define the behavior of a

new piece of UHV equipment with a well-characterized material.


3.2.5 Secondary Ion Mass Spectrometry and Auger Electron Spectroscopy


Secondary Ion Mass Spectrometry (SIMS) and Auger Electron Spectroscopy (AES)

were used to investigate a few samples by depth profile. SIMS depth profiling was

performed by Charles Evans East, an analytical company. SIMS uses an energetic beam of

primary ions to collide with the surface and dislodge them as secondary ions. Upon

collision with the surface, the primary ion is implanted in the solid and its kinetic energy is

transferred to atoms already in the solid. The transferred energy causes desorption of a

solid species at the surface producing a gas phase ion. This process is equivalently known






as sputtering. The mass spectra of these secondary ions are then detected. Both positive

and negative ions can be detected.

The escape depth of the secondary ions ranges from the immediate surface to more

than 20 A. The escape depth is a function of the material and the energy of the primary ion.

The predominant secondary ion species observed in SIMS are singly charged atomic and

molecular ions. Isotopic composition aids in identification of fragments which can be used

as more information about the chemical composition of atomic layers at or near the surface.

By using higher primary ion current densities, material is etched away and a profile of the

film composition as a function of depth can be investigated. The primary source of error in

this procedure is that the chemical composition of the material in the deeper layers can be

changed during the bombardment of the surface. In addition to chemical reactions, more

volatile components can be "boiled" off. Finally, surface material can be decomposed by

the higher energy ion source making it appear that there are species present on the surface

which were not initially present. SIMS depth profiles were performed on several Ga-

doped rutile thin films for this thesis to investigate diffusion of the dopant atoms to the

surface during growth as compared to annealing. In-depth discussions of materials

analysis by SIMS and the associated sources of error are available in the literature. 184185

In AES, the incident electron beam ionizes an atomic core level electron. The

energy released when a higher energy electron falls back to fill the core level is transferred

to a valence electron, which is ejected into the gas phase. This gas phase electron is

detected as an Auger electron with an electron spectrometer. The kinetic energy of the

Auger electron is dependent on the difference between the Auger electron's energy level in

the atom and the energy released when the originally excited electron returns to the core

level. Since this relationship is atom specific, the energy of the Auger electron can be used

to identify the atom it was released from. The Auger electron escape depth is only 4 to 25

A depending on the material being analyzed, so it is a very surface sensitive technique.

AES can be used in cooperation with sputtering to investigate the composition of a film as a






function of depth. This technique was performed at the University of Pennsylvania LRSM

on a series of undoped rutile films grown at varying oxygen partial pressure to investigate

carbon contamination in the films. In depth discussions of materials analysis by AES and

the associated sources of error are available in the literature.186'187



3.3 Electrical Characterization


3.3.1 Sample Preparation


After a film was deposited, it was removed from the nickel plate described in

3.1.1 using a razor blade edge and hand pressure. Overzealous scribing of a substrate

back to denote in-plane crystal direction often resulted in fracturing during removal from

the nickel plate in early experiments. This error resulted in changing to a very light single

mark scribing of substrate backs and more careful notation of how substrates were oriented

on the nickel plate with respect to in-plane crystal orientation. Upon detaching a film from

the nickel plate, dried Ag paste residue remained on the substrate back. For very high

resistance films the residue could potentially offer a path of lower resistance across the back

of the substrate. For this reason the residue was scraped away with a previously unused

razor blade to a level below eye detection.

To probe the electrical properties of a thin film on an insulating substrate such as

sapphire, surface electrical contacts were used. Depositing a contact on a substrate surface

prior to deposition and then a second contact on the film surface was considered. This

contact arrangement, however, introduces two potential problems. First, as will be

demonstrated in Chapter 4, surface impurities can have dramatic effects on film growth and

surface morphology. The contact layer on the substrate would most likely be a deterrent to

heteroepitaxial film formation and would definitely effect the surface morphology in an

uncontrollable manner. Since a primary goal of this thesis is to investigate well-defined

thin film materials, either of these effects would be unacceptable. Second, because the






melting point of potential contact materials like Ag and Au is approximately 1000 C or

less, diffusion of the contact material through the film during deposition would most likely

occur. Because of its adaptability to evaporation and availability in paste form, Au was

chosen as the contact material.

A mask was created from Kapton tape which is known for its ability to withstand

higher temperature than more common tapes. For the portion of the mask that was in

contact with a film, pieces of desired shape were made by putting the adhesive side of two

pieces of tape toward each other and joining these pieces with tape on the top side of the

mask. The final mask consisted of two open spaces 0.5 mm wide and 8.0 mm long

separated by 2.2 mm. The mask was capable of fully covering a film.

A film was placed on a glass slide and the mask secured over it. This object was

placed in a Au evaporation chamber (Balzers Union). A turbo pump evacuated the chamber

to at least 20 pTorr. Nominally 0.25 g of high purity (99.999%) Au wire (Alfa Aesar) was

attached to a W filament. High current flowing through the filament created enough heat to

evaporate the Au which condensed on a film and mask surface. After removing the mask,

0.05 mm diameter Au wire was attached to the contact pads using Au paste (Ted Pella).

These wires were put into physical contact with mechanically stiff Au wires (<=0.76 mm)

which were a permanent part of the sample holder. A schematic of the holder with attached

film is shown in Figure 3.8.


Mechanically
stiff Au wire Alu
Alumina
sample
holder


Evaporated
Au contact pads YT hin Au wire
pasted to contact


Figure 3.8 Schematic of alumina sample holder with attached film







3.3.2 Environmental Chamber


Small, coaxial cable (==0.99mm) was used to bring current from an 8-pin electrical

feedthrough to the permanent wires of the sample holder. The cable was grounded to an

excess thermocouple feedthrough whose atmosphere end was grounded to the building

framework. Normal coaxial cable (0=5.0mm) connected the high current and voltage ports

of the impedance bridge to a pin of the electrical feedthrough and likewise for the low

current and voltage ports. These cables were also grounded to the building framework.


3.3.2.1 Chamber design. These feedthroughs existed on the face of an

environmental chamber designed and constructed for the electrical characterization work of

this thesis by the investigator. Figures 3.9 and 3.10 are schematics of a detail of the

feedthrough arrangement and gas flow architecture, respectively. The dimensions of the

cubic chamber were 10 in. on the outer edge with a 6 in. ID opening on each face. The

feedthroughs detailed on Figure 3.9 are 1.5 or 2.5 in. ID. All flanges were Conflat type

for UHV applications.

As shown in Figure 3.9, the chamber was equipped for impedance measurements,

pressure and temperature measurements, current throughput for an internal heater, gas flow

and evacuation. Figure 3.10 shows the gas flow architecture was capable of introducing

dry and humidified N2, moderate levels of 02 (1 to 760 Torr) and CO. Scientific grade

gases were used (MG Industries, 99.9995%).










wiring to sample
for AC impedance
measurements ,

TC for sample
temperature
observation

cold cathode -
pressure gauge


quartz glass
viewport


gas flow


turbo pump


cryo pump


Figure 3.9 Environmental chamber feedthrough schematic


bypass


Water
bubbler
N2 line may be removed from
mixing intersection M and
connected to bubbler. Bubbler
outlet line connects to valve 1
In place of N2 inlet line from valve 6.


KEY
intersection

Stwo-way valve

Sone-way valve

| |mass flow
controller

N2 MFC 1 may also be
directly connected to water
bubbler. Bubbler output is
directly connected to
chamber inlet.


Valve 11 controls house
air flow to MFC solenoids
(not shown all on MFC's)


Figure 3.10 Environmental chamber gas flow architecture schematic







3.3.2.2 Testing procedure. The general procedure used in testing a film was as

follows. After placing a sample in the chamber and resecuring the removed flange, N2 was

flowed while the sample was heated at approximately 2 C per minute to around 180 'C.

Gas flow was then shut off, and the chamber was evacuated for approximately 14 hours.

During the evacuation the chamber was wrapped in heat tape followed by Al foil and heated

to 90 C. This procedure served to remove the original moisture content from the chamber.

After removing the foil and heat tapes, the chamber was filled with N2 to atmospheric

pressure, and the sample was heated at 2 OC/min to approximately 225 C. The conditions

were held constant for 24 hours while all chemically adsorbed water was driven off a

sample. The bake-out procedure was carried out a second time to remove all humidity from

the chamber. During this second bake-out, nominally 10 minutes were required for the

chamber pressure to reach 2 mTorr. At this point the sample was cooled to 180 OC. After

the bake-out, the chamber was filled with N2 to atmospheric pressure and 150 seem flow

was established. The sample was reheated to 225 OC. From this point impedance data

were taken as a function of sample temperature. 02 was then introduced and impedance

data were taken for changing temperature at one or more partial pressures. The chamber

was then evacuated for approximately 20 hours and pure N2 flow was reestablished. Flow

was then switched to humidified N2 for the collection of impedance data. The cited

evacuation times resulted in a pressure not greater than 5 LTorr. The long equilibration

times and bake-out procedure resulted in a minimum time of one week to collect reliable

data in the three atmospheres described.

Through a great deal of trial and error, minimal partial pressures of humidity were

discovered to be the source of inconsistent impedance data after reestablishing a pure N2

atmosphere. Chemically adsorbed water is known to desorb from a rutile TiO2 surface

above 200 C. Only through the two bake-out procedure could all humidity be removed

from the chamber and a film surface so that a consistent impedance value was measured in

a pure N2 atmosphere independent of previous atmospheres in the chamber. Because of







the large mass and internal volume of the chamber, approximately 1.5 hours were required

to establish equilibrium conditions after a change in sample temperature.


3.3.2.3 Sample heaters. Finally, sample heating was a source of some difficulty in

this thesis. As was shown in Figure 3.8, the final sample holder design iteration allowed a

sample to sit directly on a heater for heated surface. The original heater set-up consisted of

a flat coil-type heater inside a glass finger which protruded into the chamber. The sample

sat on a flat portion of the glass finger. The coiled heater was capable of producing a

temperature near 600 C, but over 200 OC was lost across the air gap and the glass to an

actual temperature measured at a film surface. The ability to take a film to higher

temperatures was desired, so wiring and the glass protrusion were redesigned to

accommodate a ceramic, disk-type heater. This heater worked very well in non-oxidizing

atmospheres. Its most significant design flaw was the use of carbon contacts. Pt was

sputtered over the C in an effort to separate the contacts from the atmosphere and extend the

life of the heater. Unfortunately, over time the contacts degraded and the time between

breakages became so short combined with the repair time that little progress was being

made in film testing. Therefore, the original heater set-up was reinstalled. The majority of

the data presented in this thesis was taken at temperatures below the maximum sample

temperature of this coil-type heater, 310 C.


3.3.3 Impedance Measurements


Two bridges were employed to characterize a film's impedance behavior in the

various atmospheres as a function of temperature. One was the 1689M RLC Digibridge

manufactured by GenRad. This device was computer controlled and averaged the

capacitance and loss tangent values from three measurements at each frequency taken at the

slowest available measurement speed (~1 sec). The frequency range was from 13 Hz to







100 kHz. The Hewlett-Packard 4275A Multi-Frequency LCR Meter was also employed to

extend the measurement frequency range to 10 MHz. The two devices' frequency ranges

overlapped from 10 to 100 kHz giving confirmation to a portion of the data.



3.3.3.1 Theory of measurement. The Digibridge RLC tester uses a patented

measurement technique which is described in detail in the equipment manual. The

advantage of the particular measurement technique and circuitry is that the only calibration

adjustment in the Digibridge is the factory setting of the test-voltage-level reference. The

only precision components in the instrument are four standard resistors and a quartz crystal

stabilized oscillator. The capacitance and loss tangent values are calculated by the

microprocessor from a set of eight voltage measurements, the frequency, and the calibrated

resistance and charge of the applicable standard resistor. Figure 3.11 shows a simplified

diagram of the measurement circuit in the bridge. A sine wave generator drives a current Ix

through a film (or any device being tested) represented by Zx and a standard resistor Rs

which are in series. Two differential amplifiers with the same gain K produce voltages el

and e2 where



ei = K Z I, (3.4)


and e2 = K Rs Ix. (3.5)



Combining these equations, one gets an equation for a film's impedance


Zx = Rs e / e2 (3.6)



which is a complex ratio. The capacitance and loss tangent are automatically calculated by

the Digibridge's microprocessor from the impedance, frequency and other information.






















Figure 3.11 Impedance bridge simplified measurement circuit schematic


3.3.3.2 Calculations from impedance measurements. Having obtained the

capacitance and loss tangent as a function of frequency at a selected atmosphere and

temperature, the following equations were used to calculate the real (Z') and complex (Z")

portions of the total impedance (Z) to create a complex impedance plot where


Z = Z' + jZ". (3.7)


The computer program returned a table of values which had frequency (f) in kHz and the

capacitance (C) in pF and loss tangent (D) at each f. From these parameters, Z' and Z" can

be calculated in two ways. One is via the time constant, T.


Z'= R / (1 + 02 T2) (3.8)


and Z" -w R / (1 + o2 t2) (3.9)


t=RC


where


(3.10)







and the resistance R is calculated by



R= 1 /(tC D) (3.11)


where = 2 7t f. (3.12)



An equivalent calculation uses the loss tangent directly:



Z'=D2 R/(1 +D2) (3.13)


Z" = -1 / [oC (1 + D2)]. (3.14)



C and D values were put into a spreadsheet containing both sets of equations to

calculate Z' and Z" at each f. These values were then transferred to a graphics package to

create a complex impedance plot. A C value around 1 pF is typical of all data presented in

this thesis with the observed changes in the impedance values resulting primarily from

changes in D. Figure 3.12 shows the resulting complex impedance plots for an undoped,

(101)-oriented rutile thin film at a series of sample temperatures in N2. Note the symmetric

shape of the plots. This behavior is also representative of all data presented in this thesis.

The intercept of the complex Z plot with the Z' axis is the DC resistance of a film. This DC

resistance is the value used in this thesis to calculate all data and comparisons based on

resistance values.










8 10


6107

N
4107


2107


n


233 C
....- ........ 247 oC
................ .................. .............. .................. ................. 0 247 C
258 oC
o 270 C
. .............. .................. ................... ............................ .................. ...............
*

.

.. ..A A A
: O:

7 7 7 7 8 8 8
~0


0 2107 4107 6107 8107 110 1.21081.410'

Z'

Figure 3.12 Complex impedance data as a function of sample temperature
for an undoped (101)-oriented rutile thin film in N2



3.3.3.3 Data approximations. Some of the data presented in this thesis are derived

from resistances with values greater than le9 Q. At resistances of this magnitude, a

complex impedance plot is no longer a complete semicircle. The portion that remains can

be very difficult and error prone to fitting statistics to determine the Z' intercept. When the

low frequency data no longer intercepted the real impedance (Z') axis, the Z' value at 40 Hz

was used as an approximation to the Z' intercept value and therefore the DC resistance. As

a test of this approximation and the high resistance measurement capabilities of the bridge,

resistors with nominal values of le9 Q and 5e9 2 were obtained. Each was soldered in

parallel with a 1 pF capacitor, connected to the sample holder and placed inside the chamber

in the same was as a film being tested. Figures 3.13 and 3.14 show the complex Z plot

obtained for both circuits with the R value at 40 Hz inset. The figures show that this

approximation is valid.








7108 --az .... ... .I.. .I

6 108 -.....1----..... ....... 1.04e9 .

6 10 8 .................... ........................ ... ..................... ...... ......

3108
5 1 0 .................... .................... ................................................. ....................... ......... .


S: o ..
3 1 0 ....................................................................... ....................... ............ Q ................
O
2108 O

1108
S 1 0 -- .. ............. I ... ...... I ............ .. ...... ... ......................I
1 1 0 ......................................... ........................ ........................ ....................... ........


0 2108 4108 6108 8108 1109
Z' (n)


Figure 3.13 Complex Z plot for le9 2 resistor in parallel with 1 pF capacitor




3 10 9 ...................... ...................... R a t 4 0 H z ...................... .......................
4.7e9 0Q
2 .5 1 0 9 ............................................ .. .....................
2.5 109 -O



2 1 0 9 .......... ...................................................................................... ........................
!o
2~............. ............... .......................- -- -- -- ------- ------- --- -- --
S 1.5109 ............. I....


-0 .

S1 0 ................. ............ .......................... .... ................ ..............
0
0 1 109 2109 3109 4109 5109


Figure 3.14 Complex Z plot for 5e9 Q2 resistor in parallel with 1 pF capacitor







A value of 4.7e9 1 obtained for the 5e9 0 resistor is within error range for a

resistor of this value according to the electrical professional by whom it was provided.

This approximation method was used for values obtained at 40 Hz up to 2e10 a. Beyond

this resistance, the C and D values at even the lowest frequencies became spurious as

judged by the inconsistency of the three values averaged at each f. Any value in this thesis

noted as an extrapolation was obtained with a linear extrapolation of at least three R values

below 2e10 Q. Extrapolation was only necessary for undoped and Ga-doped (001)-

oriented rutile thin films in oxidizing atmospheres.


3.4 Equipment Safety


Safety is paramount to any research conducted at a DuPont Company site. To

operate both the MOCVD reactor and the environmental chamber, a complete safety audit of

the equipment and proposed experiments had to be performed. This audit consisted of an

assessment, description and solution for the avoidance of all potential hazards associated

with pressure, temperature, electrical shock and others. In addition, a complete operating

procedure for all experiments must be written in detail noting the use of any necessary

personal protective equipment and hazards associated with each step of the procedure. The

safety audit completed for the environmental chamber is include in Appendix A.












CHAPTER 4
RUTILE FILM DEPOSITION


4.1 Undoped Rutile Films



A series of experiments investigated the effects on growth rate (R) of changing

various deposition parameters. As described in detail in 3.1.1, a prototype reactor

constructed from quartz and a high-vacuum (HV) reactor constructed from stainless steel

were used to deposit all films. The majority of the growth rate studies were carried out in

the quartz reactor while the morphology and gas sensitivity studies characterized films

grown in the HV reactor.


4.1.1 Deposition


4.1.1.1 Precursor deposition efficiency. A calculation was performed to assess the

deposition efficiency of the precursor. The unit cell volume of rutile TiO2 is 6.38e-23 cm3.

For a representative film thickness of 3300 A and dimensions of 1 x 0.5 cm, the volume is

1.65e-3 cm3 of rutile. Knowing that there are two TiO2 molecules per rutile unit cell, a

simple calculation revealed 8.6e-7 mol TiO2 in a 3300 A film. Eight such films could be

grown simultaneously under conditions of adding le-5 mol per minute of Ti precursor for

90 minutes for a total of 9e-4 mol precursor. Calculating a growth efficiency by the

following equation


EG = mol Ti in films (4
mol Ti precursor added to reactor






yields 0.76% of the Ti added to the reactor became a part of the films. Even at this small

efficiency, growth rate was found to be a function of Ti precursor partial pressure as will

be shown later.


4.1.1.2 Precursor. As detailed in 3.1.2, a solid Ti precursor of the 2,2,6,6-

tetramethyl-3,5-heptanedionato (TMHD) family was used as a precursor along with 02 gas

for the TiO2 depositions described here. A plot of the growth rate as a function of

Ti(TMHD)3 or 02 is shown in Figure 4.1. The Ti(TMHD)3 partial pressure was varied

from 0.39 to 20 mTorr in the presence of excess 02 (P02=2.5 Torr, quartz reactor) at a

total pressure of 5.5 Torr and 731 OC substrate temperature. The observed slope of the
Ti(TMHD)3 data yielded the following growth rate expression


-R = k p. (4.2)


I i 51111151 I I 1 1 4i11 i i 1111411 i 1 $i$|mell> 1 111111
WI
1 V





Cr
~ 1 -
2 / Oxygen

2 U Titanium
0.0001 0.001 0.01 0.1 1 10

Partial Pressure (Torr)

Figure 4.1 Rutile film growth rate as a function of Ti or 02 partial pressure







The non-integer dependence suggests a complex growth mechanism possibly involving

more than one molecule during the stages of transformation to TiO2.

Table 4.1 shows growth rate dependencies on Ti(OC3H7)4 and TiCI4 precursors

for a variety of substrates and growth conditions. Only TiCl4 at substrate temperatures

below 827 OC displayed a simple dependence of 1.0. Like Ti(TMHD)3, Ti(OC3H7)4 must

undergo a more complex series of reactions to deposit TiO2 as evidenced by the non-unity

precursor dependencies. One explanation for the increased order of reaction with a TiC14

precursor above 827 C is decomposition of the molecules becomes partially driven by heat

independent of the 02.



Table 4.1 Reported rutile growth rate dependence on Ti precursor



Precursor Substrate Temp (OC) PTi (Torr) Ptot (Torr) Order Ref.


Ti(OC3H7)4 Si(111) 360 < 6.8E-4 11 1.35 43
> 6.8E-4 0.14
TiCl4 Si(111), 612-827 0.03-0.9 760 1.0 9
(100),(110) >827 >1.0
Ti(OC3H7)4 Cu 220 0.04-0.3* Equal to the 2.0 50
250 0.04-0.4' Ti(OC3H7)4
280 0.04-0.8* partial
300 0.04-1.0' pressure
220-300 >P*, resp. 0


Experiments showed that TiO2 was deposited from the Ti(TMHD)3 precursor

without the addition of 02, which suggests that two of the O atoms in the precursor

molecule remained bonded to Ti when depositing TiO2. The motivation for conducting the

series of experiments at higher PO, was to determine if there was a change in the reaction

mechanism in the presence of O2. Given the large amount of H and C contained in the

Ti(TMHD)3 molecule, 02 might be expected to have influenced the decomposition






mechanism. The data suggest that excess 02 did not significantly increase the deposition

rate at this temperature and therefore the growth was limited by decomposition of the Ti

precursor or a fragment of the precursor.


4.1.1.3 Activation energy. Figure 4.2 is a plot of the logarithm of growth rate

versus reciprocal temperature for (100) rutile deposited on (0001) sapphire. Growth rate

was found to be substrate independent at all growth conditions studied, thus the same

behavior was displayed for (001)- and (101)-oriented films. The data exhibit regions of

increasing, decreasing, and flat growth rate dependence on the substrate temperature.

First, the high temperature portion of the data measured using the quartz reactor show a

positive slope. The decreasing growth rate is believed to have resulted from homogeneous

decomposition of the precursor to yield non-depositing species. The decomposition was

promoted by radiation or conduction from the hot nickel plate. The growth rate of films in

the stainless steel reactor, however, continued to increase in the same temperature range.

This disagreement resulted from differences in the materials of construction and operating

pressures of the two reactors. Compared with the wall in the stainless steel system which

was maintained at 275 C for all experiments, the quartz reactor wall acted as an insulator,

and the wall temperature was observed to increase above 275 C when the substrate

temperature exceeded 650 C. For example, the quartz wall temperature was 380 C for a

731 C substrate temperature. The combination of the hotter reactor wall and higher

operating pressure in the quartz reactor produced higher gas phase temperatures and longer

reactant residence times, thus increasing the extent of homogeneous decomposition. The

nearly horizontal segment of the low temperature quartz reactor data shown in Figure 4.2

suggests mass transfer limited growth under these conditions.

The data from the stainless steel reactor follow an Arrhenius relation consistent with

heterogeneous reaction limited growth. The growth activation energy calculated from the

temperature dependence is 37 4 kJ/mol and is independent of the sapphire orientation







used as a substrate. The primary barrier to growth is believed to have been the

transformation of the precursor or an intermediate species to TiO2 after adsorption.

Because Ti(III) is an unstable valence state, a first possible step in deposition would be to

decompose one of the TMHD ligands completely, which would result in the more stable

Ti(II) state. In addition, the weakest points in the six member ring are the C-O bonds.

Upon breaking the ring open, many possibilities exist for further decomposition into stable

species while leaving the Ti bound to an O from the ligand.


7270C 636C 560C 496C 441C
0.0I I I

A ;
-0.5 -0.61 A/s

Cc To
c -1.0 N
2 1a
-1.5 N 0.22 A/s

A HV reactor
-2.0 0 Quartz reactor


-2.5 I I I I I I I I I I I I I I I 0.082A/s
0.0009 0.001 0.0011 0.0012 0.0013 0.0014 0.0015

1/T (K')


Figure 4.2 Logarithm of growth rate as a function of reciprocal temperature


Table 4.2 shows a number of TiO2 growth studies and their respective measured

activation energies. The discrepancy in the data of the first and second entries in the table

clearly show that activation energy was a function of more than the precursor. Each study

measured the growth rate over similar temperature ranges and the resulting activation

energies are considerably different. The Ti(TMHD)3 growth rate dependence data and that







of Table 4.1 suggest that some amount of homogeneous decomposition occurred for these

larger precursor molecules. In this case, reactor pressure and geometry, among other

factors, is suggested to affect the extent of homogeneous decomposition.


Table 4.2 Reported MOCVD of TiO2 growth rate activation energies


Temp Ptot Act. Energy
Precursor Substrate Range(oC) (Torr) (kJ/mol) Ref.


Ti(OC3H7)4 Si(111) 325-400 11 10114

Ti(OC3H7)4 (100) Ti02 227- 377 na 578 52

Ti(OC3H7)4 Crown glass 400 500 na 20 (anatase)

Ti(OC3H7)4 Si(100) 300-350 0.75 87 49

Ti(OC3H7)4 Cu 230- 300 0.04-0.15 35
2.0 150

TiCI4 Si(111), <850 760 74.84.2
(100),(110)

TiCI4 Fused SiO2 400 900 760 17
Si(111)
Polyxtalin Si


4.1.2 Crystalline Quality


4.1.2.1 Bulk crystalline quality. As explained in 3.2.1 x-ray diffraction (XRD) is

a tool which can be used in a variety of ways to assess the crystalline qualities of thin films.

The 0 rocking curve (rc) is obtained when the 20 angle is held constant while varying 0.

The full-width-at-half-maximum (fwhm) of a re peak is an indication of the consistency of

plane-to-plane spacing of the planes parallel to a film's surface. Figure 4.3 shows the

range of fwhm values for (001)-, (101)- and (100)-oriented rutile films grown by MOCVD






and IBSD. The numbers to the right of the range endpoints indicate the substrate

temperature range over which the respective films were deposited.


2


1.5


0.5



0




Figure 4.3


S0450 MOCVD
0 IBSD



0 400
............. 5 0 0 ...... -. .................. f .................. .....................................................
.500

0 450 400
S730
............... ................... .............. ........................ ......... ................ .

0725 *775 725
9 775 0450

(001) (101) (100)

Orientation

Rc fwhm of undoped rutile XRD peaks as a function of
substrate temperature and deposition reactor


The data demonstrate that crystalline quality improved as a function of increasing

growth temperature as expected. A fwhm value of 0.200 is indicated for all temperatures

for (100) IBSD films. This value is known to be the resolution of the diffractometer used.

The crystalline quality of these IBSD films is most likely to have improved with increasing

temperature as well, but even the largest fwhm of this orientation was below the machine's

resolution. The lower fwhm values for IBSD as compared to MOCVD films is directly

related to growth rate differences. At the maximum deposition temperatures of the

indicated ranges, the MOCVD growth rate was near 40 A/min whereas the IBSD growth

rate was approximately an order of magnitude slower. In addition, because the Ti source in

IBSD was atoms sputtered from a pure target, no reaction or combustion of a precursor

molecule had to occur on or near the substrate surface prior to the Ti oxidation reaction.






The simple precursor and slow growth rate combined to provide longer time periods for the

diffusion of adsorbed species on the growth surface to build structures with increased long-

range order. The fwhm values shown in Figure 4.3 are comparable to those presented in

the literature for (101)- and (100)-oriented growth.22,39,42,188

The 0-20 scans investigating film orientation normal to the plane of the film found

some secondary characteristics that should be noted. First, small amounts, on the order of

1 to 5% of the (100) rutile peak, of (112) anatase were found in (100)-oriented films. No

anatase was observed in (100) films grown by IBSD which suggests that its presence was

a response to film stresses introduced by the much higher growth rate of MOCVD.

Second, for film thicknesses above approximately 2500 A, a small amount of (100)-

oriented material was detected in (101)-oriented films. This was found to occur in films

deposited by MOCVD and IBSD and was also observed in the literature.40 The appearance

of this second orientation in both MOCVD and IBSD films suggests that it resulted from

stresses present at large film thicknesses. Cracks in IBSD films were actually observed at

thicknesses greater than 4000 A due to such stresses.
Another set of experiments investigated the effect of 02 partial pressure on the rc

fwhm. The corresponding growth rate data were shown in Figure 4.1. Although no

significant growth rate effect was observed, Figure 4.4 demonstrates that the crystalline

quality did decrease as indicated by an increasing rc fwhm. The ability to measure a rc

fwhm indicates that these films continued to be crystalline and oriented. The observed

changes must then have been the result of increased stress in the lattice. Depth profiles

were performed on this set of films using Auger electron spectroscopy. The depth profiles

revealed that less than 2 at% C (technique resolution) was present in a film grown with a
large excess of 02. As the partial pressure of 02 was decreased, however, the C content

increased to approximately 15 at%. This finding suggests that without excess 02 present to

aid in carrying away the TMHD organic decomposition products, a portion was trapped in







the lattice as the film grew. These interstitial organic species expanded the lattice causing

greater stress and larger number of defects as indicated by the fwhm values.


0.50

0.45

0.40

0.35

0.30

0.25

0.20

0.15


0.02 0.04 0.06 0.08

PO2 (Torr)


0.1 0.12 0.14 0.16


Figure 4.4 (100) rutile XRD peak rc fwhm as function of PO2



4.1.2.2 Heteroepitaxy. A diffractometer with three-dimensional sample movement

capabilities was used to investigate heteroepitaxial deposition of the three rutile orientations

on their respective sapphire substrates. (001)-oriented rutile was identified on (1010)

sapphire by a 0-20 scan. To investigate the in-plane crystal directions of the substrate, a (

scan was performed on the (1126) plane (20=57.50, X=54.35) (3.2.1). Four peaks were

observed separated by 490 or 131. A projection of the [1126] direction onto the (1010)

plane shows that [1126] is separated from [0001] by 24.50 (490/2) or 65.50 (1310/2). Next

a 0 scan of the (101) TiO2 plane (20=36.080, X=32.790) was taken. Four peaks were

observed separated by 90. A projection of [101] onto (001) lies along [010]. The (101)

peaks were found to have the same separation from (1126) peaks as (1126) peaks were

calculated to have from (0001). Therefore, [010]//[0001] and along the perpendicular


, ,, I I ,I I 1 ,I I I 1 ,I I I 1 ,,

-
-~15at%C C
by depth profile



0


0 <2at%C
0 I

,,,l*,,I,,,I*,* I,,*I,*, I,,,*,*







direction [100]//[1120]. The heteroepitaxial growth of (001) TiO2 on (1010) sapphire has

not been reported outside of this thesis and the corresponding IBSD research.14
The same type of procedure was carried out for a (100) TiO2 film on (0001) A1203
substrate. A 4 scan of (1126) A1203 (20=57.50, X=42.30) revealed six peaks separated by
60 corresponding to the 3-fold symmetry of this A1203 orientation. A ( scan of (110)
TiO2 (20=27.450, X=450) also showed six peaks at the same angles as the (1126) ) scan
peaks. Only two peaks were expected given the symmetry of (100) TiO2, so there must
have been three equivalent in-plane orientations of TiO2 deposited in this case. The [1126]

direction lies parallel to [1120] when projected into the (0001) plane, therefore one way of
stating the in-plane crystal direction relationships is [010]//[1120] and [001]//[1 i00]. This
relationship has been characterized and reported by others in the field.41'42
Finally, this 4 scan procedure was performed on a (101) rutile film identified on
(1120) A1203 by 0-20 scan. Again, the oblique plane investigated for the substrate was

(1126) (X=47.70), and the scan found two peaks separated by 1800. A 4 scan of the (110)
TiO2 plane (%=67.5) revealed four peaks separated by 100 or 800 where only two peaks

were expected from symmetry. The two extra peaks result from a twin plane parallel to the
surface of the film. The twin plane created two "layers" of material which were rotated
180 in the plane of the film with respect to each other. This plane has been observed by
cross-sectional TEM.44 The [110] projection onto the (101) plane is separated from [010]
by 40, and the (101) peaks were found to have the same relationship to the (1126) peaks.
The in-plane crystal direction relationships are therefore [010]//[0001] and [i01]//[i1100].
This relationship has also been characterized and reported by others in the field.39'44
The fwhm of the 4 scan peak is a value by which relative comparisons of the
degree of in-plane alignment in heteroepitaxial films can be made. The values measured for
(001)-, (101)- and (100)-oriented films were approximately 1.1*, 1.5 and 100 respectively.
The data suggest the (001)- and (101)-oriented films were of comparable in-plane
alignment but the (100)-orientation was far worse. Recalling the data presented earlier for






(100)-oriented films having three equivalent in-plane orientations with respect an (0001)

sapphire substrate, this finding came as no surprise.


4.1.3 Morphology


4.1.3.1 Orientational effects. The morphologies of (001)-, (101)- and (100)-

oriented, undoped rutile films were characterized by atomic force microscopy (AFM).

Tapping mode AFM was used. Details of this analytical technique were given in 3.2.3.

Figures 4.5, 4.6 and 4.7 are normal, as opposed to phase shift, AFM images of thin,

medium and thick (001)-oriented films, respectively. For the remainder of this chapter

these three thickness designations refer nominally to 7010 A, 40025 A and 335075 A.

The gray scales of Figures 4.5 to 4.7 are 100 A, 200 A and 800 A, respectively. Figures

4.8, 4.9 and 4.10 are images of thin, medium and thick (101)-oriented rutile films. The

respective gray scales are 100 A, 200 A and 600 A. Figures 4.11, 4.12 and 4.13 are

images of thin, medium and thick (100)-oriented rutile films, respectively. The

corresponding gray scales are 100 A, 200 A and 800 A. All nine images are of

representative 1 gm by 1 glm areas of a film's surface. Figure 4.14 is a plot of the root

mean square (rms) roughness taken from each of the varying thickness images and a

respective substrate as a function of orientation. These values in combination with the gray

scales allow meaningful comparisons of the images.























Figure 4.5 (001)-oriented 70 A rutile film AFM image


Figure 4.6 (001)-oriented 400 A rutile film AFM image


























Figure 4.7 (001)-oriented 3350 A rutile film AFM image


Figure 4.8 (101)-oriented 70 A rutile film AFM image
























Figure 4.9 (101)-oriented 400 A rutile film AFM image


Figure 4.10 (101)-oriented 3350 A rutile film AFM image

























Figure 4.11 (100)-oriented 70 A rutile film AFM image


Figure 4.12 (100)-oriented 400 A rutile film AFM image





























Figure 4.13 (100)-oriented 3350 A rutile film AFM image


1000


100


10



1


0.1


Substrate 0


1000 2000 3000 4000


Thickness (A)

Figure 4.14 RMS roughness as a function of thickness taken from AFM images of
undoped oriented rutile films and their respective sapphire substrates


0 (001)
0 (101)
A (10 0 ) ... ....... ........................ ................
O

: i i
........ ........... ............... ....... ......................... ...............................................


A.
0 -- -- ....-.. .............. ...............
=


1000



100


10



1



0.1







Characteristic dimensions of the features shown in each oriented thickness image

were measured using the analysis software which accompanied the atomic force

microscope. This allowed substantially more accurate measurement of the planar

dimensions as compared to using ratios with mechanical measurements. Of equal

importance is that the software enabled measurement of a feature dimension normal to the

plane of a film which is impossible to do otherwise.

Figure 4.15 shows the characteristic planar dimension of the features observed for

each orientation plotted as a function of film thickness. The data suggest several important

pieces of information about the films. First, the plot shows two sets of data points for

(101)-oriented films. Obviously, the dimension represented by the open circles is

approximately twice as large as the dimension represented by filled circles at all

thicknesses. A non-unity ratio of these dimensions suggests elongation of the features in

some direction. This elongation can be seen in the features of Figures 4.8 to 4.10. Figure

4.8 confirms this characteristic was present in the earliest stages of deposition. The

heteroepitaxial investigations described previously found that the direction of elongation is

parallel to [101]. Because heteroepitaxial growth was dependent on lattice matching

between a film and its substrate, the degree of matching may have been a driving force for

elongation. As was discussed in 2.1.3, matching between the oxide sublattices of rutile

and sapphire is believed to be the driving force for heteroepitaxial growth of these films.

Table 2.1 showed the anionic mismatch for (101)-oriented films to be 5.8% for

[010]//[0001] but only 0.9% for [101]//[il00]. The small mismatch along the [i01]

direction allowed growth to proceed in that direction with a smaller amount of stress on the

material, and, therefore, fewer defects occurred than in other planar directions of the film.

The feature shape in the (001) and (100) orientations can also be analyzed in terms

of stress due to lattice mismatch. Figures 4.5 to 4.7 of the (001)-oriented films have

features which do not exhibit elongation in any direction. Table 2.1 showed the anionic

mismatch in the [100]//[1120] and [010]//[0001] to be 3.6% and 5.7%, respectively.




77


Although the mismatch is larger in one direction than another, the smaller mismatch was

still too large to induce observable preferential growth. Likewise, Table 2.1 showed the

lattice mismatch between the three equivalent orientations of (100) rutile and (0001)

sapphire to be 3.8% and 7.3%. The considerably smaller mismatch in one direction was

too large to motivate elongation in that direction as demonstrated by the features shown in

Figures 4.11 to 4.13.

Thin film stress is known to increase with film thickness.7'35 This characteristic is

illustrated in these rutile films by the non-linearity of the data in Figure 4.15. The planar

dimensions of the features increased in size with increasing film thickness. A visual

interpolation of the data suggest that the feature size did not, however, substantially

increase for film thicknesses beyond the range of 1500 to 2000 A. Stresses due to many

layers of material being in compression or tension created defects as a relief mechanism. A

limit on feature size in the film plane was the result.


1200 1200


o 1000
r-
0)
E 800
Ca
S 600
0.
400

c 200

0 n


1000

800

600

400

200

n


0 500 1000 1500 2000 2500 3000 3500

Film thickness (A)

Figure 4.15 Characteristic planar dimension of surface features as a function
of film thickness for undoped, oriented rutile films


o (001)
0-... o (10 1) ................. ................ ................ .......... -
o (101)
... A (100) .. ,...... _



0-
.... 9. .............. ................. ...............................-
0o
.........".... ................... ... .......... ................. ................ ................. ..............
.......0 ......... ........ .........i......... ........ .......

" iJLi i J iJ i i~


J


IkJ






Figure 4.16 is a plot of the ratio of the characteristic height to the planar dimension

of the features observed for each orientation as a function of film thickness. Important

insights into film growth mechanism can be gathered from this figure in conjunction with

the AFM images. The thin film images of each orientation suggest an island-type growth

mechanism from the earliest stages of deposition. This mechanism occurred when material

was deposited at a higher rate than previously adsorbed material could transport across the

growth surface to continually extend the growth surface in two dimensions. Although the

thin film features for each orientation are obviously island-like, the data of Figure 4.16

suggest a degree of two-dimensional growth mechanism as well. The height to planar

dimension ratio of these features varies from 5% to 13% for the three orientations.

Correspondingly, the islands were growing at a faster rate in the film plane than normal to

it.

The data of Figure 4.16 further suggest that this mixture of two- and three-

dimensional growth mechanisms was not substantially altered for films that were

approximately 400 A thick. The features in the thin and medium thickness images for each

orientation have similar appearances other than their size difference. The data for a (100)-

oriented thick film indicated that the height approached the planar dimension of the features.

The rms roughness value which is considerably larger than the other two orientations

corresponds with this ratio. A comparison of the data for the two dimensions of a (101)-

oriented film give evidence that the elongated dimension and height feature continued to

increase at similar rates. The smaller dimension increased at a much slower rate. Finally,

the ratio remained low for a (001)-oriented film at all thicknesses. This data corresponds

with this orientation being the smoothest as indicated by its rms roughness value at each

thickness in comparison to the other two orientations.

The cross-sectional views of the AFM images which were used to obtain the feature

height dimension also allow an analysis of the feature shapes. The shapes and the angles

measured from them indicate that (001)- and (101)-oriented film surfaces are faceted, but







the (100) surface was more difficult to resolve. An in-depth analysis of the facet planes

and their relevance to the observed behavior of these films as gas sensors is taken up in the

next chapter.


S0.8

S0.7

E = 0.6
-C-
. 0.5

C" c 0.4

S0.3
.q-0 0.2
O5.
oa
0.1
-a
0


0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0


0 500 1000 1500 2000 2500 3000 3500

Film thickness (A)

Figure 4.16 Ratio of characteristic height to planar dimension of surface features
as a function of film thickness for undoped, oriented rutile films



4.1.3.2 Substrate effects. In the process of carrying out these morphology

investigations, some interesting information was discovered concerning the surface

characteristics of various substrates. An AFM image of a virtually smooth (1010)-oriented

sapphire substrate representative of those used in the morphology investigations is shown

in Figure 4.17. Similar images were obtained for (1120)- and (0001)-oriented sapphire

substrates. The rms roughness data shown for the substrates in Figure 4.14 suggests that

the surfaces were of like smoothness. Such clean surfaces were obtained using the

methanol, deionized water, filtered air procedure described in 3.1.3. In spite of this

procedure, AFM revealed that random substrates had an appreciable density of larger

features. The composition of these features is unknown. They were, however, hard with


1 1 1 I 2 1 I l s # '1 4 1 1 1 1 1 1 .
oE (001) .,. ...... .
..... a (0 0 1) ...................... ................. ................. ..............
S(101)
S (10 1) ..................... ...............................................
S A (100) o



8
- ............. ................. .................. .......... .................. ....... .......... ..............-
S, ,I 0
...................... ................. ................. ................. .............







respect to the ability of phase contrast AFM imaging to differentiate between hard and soft

materials.178 In addition, they were found to be unremovable by any amount of surface

cleaning. These characteristics suggest that the features were material which was

inadequately removed after the polishing step in manufacturing the substrates. Figure 4.18

is a normal AFM 0.5 pm by 0.5 pm image of a (1120) sapphire substrate showing features

typical of what was observed on each sapphire orientation.


















Figure 4.17 (0001)-oriented sapphire Figure 4.18 (1120)-oriented sapphire with
unremovable surface features


Figures 4.19 and 4.20 are AFM images of thin and thick (101)-oriented films

grown on (1120) "featured" substrates. Figure 4.19 demonstrates that the features were

areas of some degree of preferential growth. The characteristic dimension of the (1120)

substrate feature is 140 A with a 15 A height. The thin film imaged in Figure 4.19 has

large features which are approximately 380 A by 200 A and 65 A high, whereas the

remaining material has a 100 A dimension with a 10 A height. This preferential growth

creating larger islands of rutile material in the initial stages of deposition is believed to be

responsible for the larger dimensions of many of the features observed in Figure 4.20 as

compared to those seen in Figure 4.10. Cross-sectional views of each thick film image







showed that the features in both images can be modeled as trapezoidal prisms, but the

relative dimensions on average are larger in Figure 4.20.


Figure 4.19 (101)-oriented 70 A rutile film Figure 4.20 (101)-oriented 3350 A rutile
on "featured" (1120) sapphire substrate film on "featured" (1120) sapphire
substrate


4.2 Doped Rutile Films


Doped rutile films with (001), (101) or (100) orientation were deposited by

MOCVD using solid precursors by the same procedure as the undoped films described in

the previous section. Films were doped with either Ga or Nb. These two dopants were

chosen for a couple of reasons. One, with respect to Ti which has a 4+ oxidation state, Ga

is an acceptor-type dopant as a result of its 3+ oxidation state. Similarly, Nb is a donor-

type dopant because of its 5+ oxidation state. Two, the defect chemistry which results

from acceptor or donor dopants was only of interest for the purposes of this thesis if the

dopant atoms incorporated substitutionally at a Ti lattice site. Ga3+ and Nb5+ have atomic

radii of 0.62 A and 0.70 A, respectively.189 They are similar in size to Ti4+ which has an
atomic radius of 0.68 A. The dopant atoms therefore had little motivation to segregate and







form secondary phases in a film due to misfit strain. The growth of doped rutile films

served two purposes. First, to advance the understanding of metal oxide thin film

deposition with the incorporation of more than one metal. The findings are important to the

deposition of materials such as ferroelectrics where compounds like lead zirconate titanate

are being studied. Second, the known changes in the bulk and at the surface due to the

defect reactions reviewed in 2.3.3.1 were studied for their effects on gas sensitivity.



4.2.1 Deposition


4.2.1.1 Bulk dopant concentration. Rutherford backscattering spectrometry (RBS)

was used to characterize the average dopant concentration in the films. A representative

RBS spectra was shown in Figure 3.5 with an explanation of the analytical technique

(3.2.2). The 4 gpC of total charge used to characterize the thin films was not sensitive to

any segregation of dopants to the surface or film-substrate boundary. Due to noise in the

spectra, the technique was accurate to 0.1 at%.

RBS analysis of the doped films confirmed a nominal one-to-one ratio between

dopant concentration in the dopant/Ti solid precursor mixture and the dopant concentration

measured in a thin film. The relationship was observed for both Ga and Nb concentrations

from 0.5 at% to 6.5 at% on a per Ti basis. This characteristic goes against what has

typically been observed for multiple-metal oxide films grown by the MOCVD

technique.190,191 Several factors in combination are suggested as the cause of this

behavior. Primary, is the unique procedure of mixing the powder precursor materials for a

simultaneous sublimation of each species at the desired ratio. This was accomplished by

setting the UV lamp to provide the heat necessary to sublime the precursor with a higher

sublimation temperature (3.1.2). Using separate rods for each precursor and a ratio of the

sublimation rates was unable to accomplish this same feat at these doping levels. Second,

is the deposition rate. Rutile films were deposited by MOCVD at around 37 A/min as




Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID ECJ6XSX85_OIL8W7 INGEST_TIME 2013-02-14T17:31:32Z PACKAGE AA00013548_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES