V V J... .31.». C35...— :21
inﬁaﬁseﬁwﬁ. 3.3:". . , .
Burr.)
, .515:
1:}.

9%
.

.3
n1.» :
3:

y.

o: 395d . . . “:Pﬁth. . . n . .
, V ‘ Arr“... {I}.
, . . in .JVJ _ A V . . . V . s L. a
r. . v . {um} I: I.
. V , . . .af. 3 1.. (.53
.n. uﬂaﬂaﬁxﬁéw
:. A)
.r:

.
1 . V ,
. V . . .3!) a

. 32;.

I.
a

1!
595..
.3: x

2
.... 2.2.
A." «2:...

$33.53; , .
$4.53.. . in ram:
hyaﬁwss; f

V n

at impair”: 2!. ,
«l... L51.¢¢:ln‘.3h. A ”dry
2 .5? .1 . Va: F12.

2:
V I 1‘.
. .2...
a :2.
in}; 3
5.9.356, .93..
divaﬂurﬂvﬂ .:.ul: hr}!
1! . 1!!

«.n. «M.

1.5.55? 25...... x
5959...}:
V .1.» . ac

n. .1 .v.

.l ’15:».
. ) .12... .9.
Ban 32:325.}?
A {23:33

51 a! .
vir :53.
V

 

 

 

 

 

 

 

 

1112313

.‘ If." '5’ }

 

LIBRARY
Michigan State
University

 

 

 

This is to certify that the

thesis entitled

COMPOSITION OF SPUTTERED TITANIUM NICKEL AND
TITANIUM NICKEL PALLIDUIM FILMS

presented by
THOMAS BRADLEY LAGRANGE

has been accepted towards fulﬁllment
of the requirements for

M.S. d . Material Science &
egree 1n

 

 

Mechanics

 

Major professor

Date M

0-7639 MS U is an Afﬁrmative Action/Equal Opportunity Institution

 

PLACE IN RETURN Box to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE

DATE DUE

DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 cu‘CIRC/DatoDuopes-pts

COMPOSITION OF SPUTTERED TITANIUM NICKEL AND TITANIUM NICKEL
PALLIDUIM THIN FILMS

By

Thomas Bradley LaGrange

A THESIS
Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of
MASTER OF SCIENCE
Department of Material Science and Mechanics

2001

ABSTRACT

COMPOSITION OF SPUTTERED TITANIUM NICKEL AND TITANIUM NICKEL
PALLIDUIM THIN FILMS

By
Thomas Bradley LaGrange

The properties of shape memory alloys (SMA) are strong function composition,
thus tight control of composition is necessary during material manufacturing processes.
The goal of the present work is to observe the effect of sputter deposition process on
composition of TiN i and TiNiPd SMA thin ﬁlms. TiNi and TiNiPd ﬁlms made by dc.
magnetron sputtering have shown composition shiﬁ with target erosion, which does not
occur in do magnetron triode sputtered ﬁlms. In addition, deposition parameters such as
working gas pressure, throw distance, and substrate temperature have been shown to
affect the composition. Due to the complexity of the sputter deposition process, it has
been difﬁcult to determine the particular phenomena that most affects the ﬁlm
composition, however preferential resputtering, thermalization, diffusion, sticking
coefﬁcients, and angular distributions of sputter particles are discussed as possible
contributors. Further experiments will be presented that may lead to a solution for the

composition control problem.

ACKNOWLEDGEMENTS

I would ﬁrst like to thank Dr. Ron Averill and Dr. Thomas Pence for their support.

I would like to thank Reza Lolee for his assistance on the triode deposition experiments.

I would like to thank to the MSM department secretaries, Jo Ann Peterson, Lorna
Couliter, Debbie Conway, and Iris Taylor.

I would to thank J.C. Brenton, whose design and machining skills help to make my
experimental setups possible.

I would especially like to thank Dr. David Grummon who inspired me, and gave me this
unique opportunity to study shape memory thin ﬁlms. Without his guidance and wisdom,
this thesis would not be possible.

I would like to tlmnk my parents who have given their support and love that helped me to
become successful student and adult.

Finally, I would like to give my unconditional love and gratitude to my wife stands by me
through thick and thin and gives me the loving support that I need to achieve my goals.

iii

TABLE OF CONTENTS

List of Tables ................................................................................... viii
List of Figures ................................................................................... ix
Chapter 1 Introduction .......................................................................... 1
Chapter 2 Literature Review .................................................................... 4
2.1 Shape Memory Properties of TIN i ................................................ 4

2.1.1 Martensitic Transformations ........................................... 4

2.1.1.1 Phenomenological Theory .................................. 5

2.1.1.2 Martensite Internal Twin Structures ...................... 6

and Variant Interfaces

2.1.2 The Shape Memory Effect ............................................. 9
2.1.3 Equilibrium Phase Structure and Phase Stability of TiN i ....... 10
2.1.4 Effects of Composition on Transformation Temperatures . . ....11
TN i and TIN iX Alloys
2.1.5 Mechanical and Metallurgical Studies of Sputtered TiN i ...... 14
SMA Thin Films
2.1.5.1 Binary TiNi ................................................. 14

2.1.5.2 Ti-(Ni+Pd) Structure and Mechanical Properties. . ....16

2.1.6 TIN i Thin Film Sputtering Composition Experiments ........... 17

2.2 Sputter Deposition ............................................................... 21
2.2.1 Principles of DC Sputtering Glow Discharge ..................... 22

2.2.2 Sputtering Mechanisms ............................................... 24

2.2.3 Sputtering Yield ....................................................... 26

2.2.4 Sputtering Sources .................................................... 31

2.2.4.1 Magnetron Planar Diode .................................. 31

iv

 

2.2.4.2 Triode Source ............................................... 33

2.2.5 Film Growth and Sticking Coefﬁcients ............................ 34
2.2.6 Sputter Deposition of Alloys ........................................ 37
2.2.6.1 Topographical Evolution of Multi-component. . . . .....38
Targets
2.2.7 Effect of Recoil Angular Distributions on Film .................. 39
Composition
2.2.8 Effects of Thermalization on Thin Film Composition ........... 42
and Microstructure

2.2.9 Effects Preferential Resputtering and Sputter Yield .............. 47

Ampliﬁcation on Composition
Chapter 3 Materials and Methods ........................................................... 52
3.1 Materials ........................................................................... 52
3.1.1 Binary TiNi Targets .................................................. 52
3.1.2 TiNiPd Targets ........................................................ 53

3.1.3 TiNiPd Target Chemical and Microstructural Analysis... .. ....53

3.1.4 P/M Target Performance ............................................. 55
3.1.5 Substrates .............................................................. 56
3.2 Sputtering Deposition Process ................................................. 5 7
3.2.1 Binary TiN i Magnetron Diode Deposition Setup ................ 57
3.2.2 Triode Deposition Setup ............................................. 58
3.2.3 TNiPd Magnetron Sputtered Thin Film Deposition ............ 59
3.3 Qualitative Assessment of Film Appearance and Performance ............ 61
3.4 Resistivity Measurements ........................................................ 65
3.5 Differential Scanning Calorimetry (DSC) ..................................... 65

3.6 Thin Film Composition Measurements ........................................ 66

3.6.1 Binary TiNi Composition Shifts Measurements due to..........66

Cathode Wear
3.6.2 TiNiPd Thin Film Composition Measurements ................... 67
3.6.2.1 TiNiPd Thin Film Composition Standard ............... 67
Preparation
3.7 TEM Microscopy ................................................................. 69
3.7.1 TEM Sample Preparation ............................................ 69
3.7.2 TEM Microscope Setup .............................................. 70
3.8 Thin Film Thickness Measurements ............................................ 70
Chapter 4 Results and Discussion ............................................................ 71
4.1 Sputtered Binary Tm i Composition Experiments ............................ 72
4.2 Thin Films Studies of Films Sputtered from P/M Composite TiNiPd.. . .74
Target
4.2.1 Composition Studies of Sputtered TiNiPd Thin Films ........... 74
4.2.2.1 The Effect of Target Erosion on Composition in. .....74
TiNiPd Thin Films
4.2.2.2 The Effect of Deposition Temperature on .............. 81
Composition
4.2.2.3 The Effect of Working Gas Pressure on ................ 81
Composition
4.2.2 Transformation Temperatures of Sputtered TiNiPd Thin ...... 82
Films
4.3 Mechanisms that Affect Film Composition ................................... 85
4.3.1 Partial Sputtering Yields ............................................. 85
4.3.2 Composition Effects Resulting from Differential ................. 89
Angular Distributions
4.3.3 Thermalization ......................................................... 91

vi

 

4.3.4 Mechanisms that Affect the Composition at the ................... 93

Film Surface

4.3.4.1 Sticking Coefficients ...................................... 93
4.3.4.1 Preferential Resputtering ................................. 93
4.3.5 The Effects of Target Surface Diffusion on Film ................. 96

Composition
Chapter 5 Conclusions ......................................................................... 99
Bibliography ................................................................................... 105
Figures ......................................................................................... 1 10

vii

 

 

LIST OF TABLES

Table 2-1. Twinning modes [Matusumoto, 1987] ......................................... 7
Table 2—2. Sputtering Conditions and Film Composition [Ishida, 1993] .............. 18
Table 2-3. Threshold Energy in (eV) for Argon Sputtering ............................. 27
Table 2-4. Sputtering Yield Data for Ni, Ti, and Pd ..................................... 31
Table 2-5. Mean Free Paths of Ti, Ni, Pd, and Ar ....................................... 44
Table 2-6. Thermalization Fractions of Ti, Ni, Pd, and Ar .............................. 45
Table 2-7. Atomic Mass Parameters (against Argon) .................................... 48
Table 3-1. Nominal TiNiPd P/M Sputtering Target Compositions .................... 53
Table 3-2. Deposition Parameters and Properties ....................................... 62
Table 4-1. TiNiPd Film Deposition Parameters and Properties ......................... 76
Table 4-2. Transformation Temperature Data for TiNiPd ﬁlms from .................. 82
DSC Curves.
Table 4-3. Transformation Temperature Data from Electrical ........................... 82
Resistivity Curves.

viii

 

LIST OF FIGURES

Figure 2-1 Schematic of Shape Memory Eﬂ'ect and Transformation ................. 110
Temperatures Observed in Electrical Resistivity vs.
Temperature Plot [Duerig, 1990].

Figure 2-2 Binary Phase Diagram of TiNi [Binary Alloys Phase ..................... 111
Diagram (ASM), 1987].

Figure 2-3 Equilibrium Phase Diagram Near Equiatomic Composition at ........... 112
Low Temperature of TiN i Binary System Proposed by
Wasilewski et al.

Figure 2-4 Martensite Start Temperatures as a Function of Ni Content ............... 113
[Melton, 1990;Wang, 1965].

Figure 2-5 (a) Effect of Aging Time on Ni- rich TiN i Thin Films ..................... 114
[Miyazaki, 1999].
(b) Effect of Aging Time on Ti-rich TiN i Thin Films
[Miyazaki, 1999].

Figure 2-6 DSC Results of the Effects of Aging Time on Ms .......................... 115
Temperature for Ni and Ti-Rich TiN i Thin Films
[Miyazaki, 1999].

Figure 2-7 The Effect of Solutionizing Temperature on Transformation ............ 116
Temperatures in Ni—rich TiN i Thin Films Observed by
Strain-Temperature Curves at Constant Stress [Ishida, 1995].

Figure 2-8 The Effect of Solutionizing Temperature on Transformation ............ 117
Temperatures in Ti-rich TiN i Thin Films Observed by
Strain-Temperature Curves at Constant Stress [Ishida, 1995].

Figure 2-9 Titanium-Nickel-Oxygen Phase Diagram .................................... 118
[Chattopadhyay, 1983]

Figure 2-10 The Effect of Cu and Pd Content on the Thermal Hysteresis ............ 119
[Miyazaki, 1999].

Figure 2-11 M, Temperature as a Function of Pd and Pt at. % ........................ 120
[Lindquist, 1988].

Figure 2-12 The Effect of Composition on Austenitic Strength ........................ 121
[Hodgson, 2000].

Figure 2-13 Stress-Strain Curves for Various TiN i Thin Films ........................ 121
Compositions [Ishida, 1999].

Figure 2-14 The Effect of Precipitate Diameter on Critical Stress for Slip ........... 122
in Austenite [Ishida, Microstructure 1999].

Figure 2-15 Strain-Temperature Curves at Constant Stress of Nickel- ............... 123
Rich TiN i Thin Films Annealed at Various Temperatures
for 3.6 ks [Ishida,1995].

Figure 2-16 (a) Stress-Strain Curve ofTi-48.2 at.% Ni thin ﬁlm ....................... 124

Annealed at 773 K for 300$[Ishisda, 1999].
(b) Stress-Strain Curve of Ti- 48.2 at.% Ni thin ﬁlm
Annealed at 873 K for 3.6 ks [Ishisda, 1999].

Figure 2-17 Strain-Temperature Curves at Constant Stress of Titanium- ............ 125
Rich TiN i Thin Films Annealed at Various Temperatures
for 3.6 ks [Ishida,1995].

Figure 2-18 Lattice Parameter Data for B19 Martensite as a Function of ............ 126
Pd at. % Reported by Lindquist (I), Sivokha et a]. (O), and
Shugo et 01. (X).

Figure 2-19 Lattice Parameter Data for B2 Austenite as a Function of Pd ........... 126
at. % Reported by Lindquist (I) and Shugo et al. (C).

Figure 2-20 Strain-Temperature Curves for Ti-(Ni+Pd) Alloy ........................ 127
[Miyazaki, 1999].

Figure 2-21 The Relative Change Ti content of IBAD TiN i Films to the ............ 127

expected Film Composition as Function of I/A Ratio in the
Absence of an Assisted Beam [Chang, 1992].

Figure 2-22 The Effect of Target Erosion on Composition [Lee, 1994] .............. 128

Figure 2-23 The Effect of Pressure on the Composition [Lee,l994] .................. 128

Figure 2-24 The Effect of Target-Substrate Distance on Composition ................ 129
[Lee, 1994].

Figure 2-25 The Effect of Power on Composition [Lee, 1994] ........................ 129

Figure 2-26 Schematic of Sputtering Apparatus [Bunshah, 1994] ..................... 130

Figure 2-27 The Voltage-Current Relationship a DC Discharge ...................... 130
[Mahan, 1999].

Figure 2-28 The Luminous and Dark Regions of DC Glow Discharge ............... 131
[Mahan, 1999].

Figure 2-29 A Collision Cascade Resulting in a Sputter Atom or Recoil ............ 131
[Mahan, 1999].

Figure 2-30 Three Energy Regimes of Collision Cascade (a) Single .................. 132
Knock-on Regime, (b) Linear Cascade (c) Spike Regime
[Sigmund ,1981].

Figure 2-31 The Effect of Surface Binding Energy on Yield ........................... 133
[Mahan and Vantomme, 1999].

Figure 2-32 The Projectile Energy Dependence on Yield for Tungsten .............. 134
Target Struck by Various Projectiles [Anderson, 1981].

Figure 2-33 A Contour Plot Showing the Effect of Projectile Mass on ............... 135
Sputtering Yield with Target Atomic Number (a) Yield Data
ﬁom Empirical Formula Developed by Matsunami et al.
[1984] (b) Yield data ﬁ'om a Simpliﬁed Collision Model
[Mahan and Vantomme, 1999].

Figure 2-34 The Effect of Ion Incidence Angle on Sputter Yield ( solid ............. 136
lines are theoretical calculation by Yamamura et a1. [1983]
and circles are experimental data ﬁ'om Oechsner [1973]).

Figure 2-35 The Electron Trapping Effect of Applied Magnetic and ................. 137
Electrical Field [Mahan, 1999].

Figure 2-36 Schematic of a Triode Sputtering Source and Conﬁguration ............ 137
[Bunshah, 1994].

Figure 2-37 Schematic of Film Formation and Growth [Leawer, 1981] .............. 138

Figure 2-38 Schematic of an Unbalanced Magnetron [RossnageL 1982] ............ 139

Figure 2-39 Schematic Diagram of Sputtering Yields ﬁ'om Initial and ............... 139
Altered Layer Target Surface Compositions
[Bunshah, 1994].

Figure 2-40 Schematic Drawing of the Formation of Cones ............................ 140
[Bunshah, 1994].

Figure 2-41 SEM Photomicrograph of Cone Structure on the Surface of ............ 140

a Sputtered Ni-rich TiN i Target [Lee, 1994].

xi

 

Figure 2-42 Angular Distribution of Sputter Atoms [Kay, 1962] ..................... 141

Figure 2-43 Lateral Composition Shifts Studies in TbF e alloy Sputtered ............ 142
Thin Films [Murakami, 1990].

Figure 2-44 The Correlation of Emission Angle with Sputtering Yield .............. 143
[Betz, 1970].

Figure 2-45 The Fraction of Sputter Particles That Reach the Substrate ............. 144

[Turner, 1992].

Figure 2-46 A Plot Showing the Binding Energy and Mass of Range of ............. 145
Elements Overlayed with Contour Lines that Indicate the
Percentage of Atoms Reaching the Substrate, which was
Interpolated ﬁ'om Turner et al. Data on Incident Flux
Reduction at 3.5 mTorr. ['Turner, 1992].

Figure 2-47 Thermalization Distance as Function of Pressure for Various .......... 146
Atomic Masses and Initial Energies [Westwood, 1978]
(— , 5 eV) (mu, 1000 eV).

Figure 2-48 The Average Energies as Function of Working Gas-Distance .......... 147
Product [Somekh, 1984].

Figure 2-49 The Experiment Data of Intrinsic Resputtering as Function ............ 148
Atomic Mass Parameter and the Resputtering Fractions
Calculated for Ti, Ni, and Pd [Hoﬂinam 1990].

Figure 2-50 Static TRIM Simulations of Partial Sputtering Yield of a ............... 149
l-nm Thick Al Layer as Function of Substrate Atomic
Number for Various Incident Ion Types [Berg, 1999].

Figure 3-1 Literature Values for Transformation Temperatures of .................... 150
TiNiPd Alloys.

Figure 3-2 Lo-Wu-Wayman’s 1990 TiNiPd Transformation ........................... 150
Temperature Data.

Figure 33 Optical photomicrograph of Ti 53.5/Pd 25/N i 21.5 at. % .................. 151
Target taken at 50X.

Figure 3-4 Optical photomicrograph of Ti 53.5/Pd 25/Ni 21.5 at. % .................. 151
Target taken at 500X.

Figure 3-5 SEM photomicrograph at 100 X of the Ti 53.5/Pd 25/N i .................. 152

21.5 at. % Target.

xii

Figure 3-6 SEM photomicrograph at 1200 X of the Ti 53.5/Pd 25/N i ................ 152
21.5 at. % Target.

Figure 3-7 X-ray Map for Ti, Ni, and Pd of Target Pdl at 1000X ..................... 153
Figure 3-8 SEM Photomicrograph Taken at 500X Overlayed with 5 50 .............. 154
Point Grid used for Void Volume Fraction Calculation.
Figure 3-9 Kurt J. Lesker Sputter Deposition System ................................... 155
Figure 3-10 Aluminum Substrate Holder ................................................. 156
Figure 3-11 Torus 3M Magnetron Sputtering Gun Assembly .......................... 157
Figure 3-12 Triode Substrate and Gun Position Setup .................................. 158
Figure 3-13 Basic Schematic of Thin Film Deposition Unit Used in ................ 159
Magnetron Sputtered TiN i and TiNiPd Thin Film
Composition Experiments.

Figure 3-14 5-inch Wafer Platen Used for Depositing TiNiPd Thin Films .......... 160
Figure 3-15 DCS Curve Used to Calculate the Crystallization ........................ 161
Temperature for Specimen # 02240001 (weight was 5.5 mg

and scanned at 5°C/min).
Figure 3-16 Resistivity Device Schematic ................................................ 162
Figure 3-17 Schematic of Annealing Sample Setup ..................................... 162
Figure 3-18 (a) TEM Microphotograph of Specimen #03020001- .................... 163

AN0518-500 (100 kX).
(b) TEM Diffraction Pattern of B19 Martensite
(Camera Length 1.2m).

Figure 3-19 (a) TEM Microphotograph of Specimen #03020001- ................... 163
AN0518-500 at 300°C (50 kX).
(b) TEM Diffraction Pattern of B2 Austenite
(Camera Length 1.2m).

Figure 4-1 Ni Content (at. %) Verses Magnetron Diode Sputtering Time ............ 164
ﬁom 51 at% Ni/48 at% Ti Target.

Figure 4-2 Surface Plot of Ni Composition as Function Sputtering Time ............ 165
and Distance from Center of Diode Cathode ﬁ'om 51 at% Ni/48
at% Ti Target.

xiii

Figure 4-3 Rough Schematic of Smoothed surface plot of Ni Composition .......... 166
as Function Sputtering Time and Distance ﬁom Center of
Diode Cathode ﬁom 51 at% Ni/48 at% Ti Target.

Figure 4-4 Ni Content Verses Radial Distance ﬁ'om Diode Cathode Center... ......167
for ﬁrst 12 Hours of Sputtering.

Figure 4-5 Ni Content Verses Radial Distance ﬁ'om Diode Cathode Center. . . . . ....l67
for 15 to 24 Hours of Sputtering.

Figure 4-6 Ni Composition Verses Triode Sputtering Time ﬁom 51 at.% ........... 168
Ni/ 48 at% Ti Target.

Figure 4-7 Ni Content Verses Radial Distance from Triode Cathode Center ........ 169

Figure 4-8 Composition Verse Elapsed Target Wear Time of Ti 53.5%/ ............. 170
Ni 21.5 %/ Pd25% Target.
Figure 4-9 Ti Content of the Ti(Ni+Pd) Films as a Function of Deposition ......... 171
Temperature.

Figure 4-10 Film Composition vs. Deposition Temperature for the Ti ............... 172
53.5%/ Ni 21.5 %/ Pd 25% Target.

Figure 4-11 Film Composition vs. Pressure-Distance Product for Thin .............. 173
Films Sputtered from the Ti 53.5%/ Ni 19.5%/ Pd 27% Target.

Figure 4-12 Composition vs. Working Gas Pressure for Thin Films .................. 174
Sputtered from the Ti 53.5%l Ni 19.5%/ Pd 27% Target.

Figure 4-13 Transformation Temperatures in As-deposited Thin Films vs .......... 175
Palladium Content.

Figure 4-14 Ms Temperature as Function of Palladium Contents: Published ........ 176
Values Compared to As-deposited and Annealed Thin Films
[Lindquist, 1988].

Figure 4-15 Ms Temperature as a Function of Titanium Content for As- ............ 177
Deposited Thin Films.

Figure 4-16(a) M, Temperature Verse Deposition Temperature for As- ............. 178
deposited Thin Films ﬁom 53.5% Ti/ 19.5% Ni/ 27% Pd
Target.

(b) M, Temperature Verse Deposition Temperature for As-
deposited Thin Films from 53.5% Ti/ 21.5% Ni/ 25% Pd
Target.

xiv

 

 

Figure 4-17 Schematic of Surface Diffusion in Magnetron Sputtering that ......... 179
Results in a Non-Steady State Sputter Flux.

Figure 4-18 The Effects of Recoil Mass on Preferential Sputtering of Alloy ........ 179
Calculated ﬁ'om Matsunami et al. Sputter Yield Formula using
Argon Projectiles.

Figure 4-19 Schematic of the Possible Surface Topography Developed in .......... 180
PM Targets at Steady State.

XV

CHAPTER 1 INTRODUCTION

Shape memory alloys (SMAs) have been used for many applications, such as
eyeglass frames and arterial stents. With recent advancements in thin ﬁlm technology and
micromachining techniques, there has been a growing interest in shape memory thin ﬁlms
for use in micro-actuator systems. The shape memory effect (SME) is dependent on
thermoelastic phase transformations at temperatures that are a strong function material
composition. These transformation temperatures deﬁne the therrnomechanical operational
parameters of micro-actuators, thus tight control of composition is necessary to obtain the
desired properties of the SMA and device operation.

Since titanium and nickel have very different vapor pressures, vacuum evaporation
is impractical, and sputtering has been the preferred way to manufacture TiN i shape
memory thin ﬁlms. Sputtering is the most widely used manufacturing process in the thin
ﬁlm industry because it can produce ﬁlms of any desired composition with high substrate
adhesion [Bunshah, 1994]. The physical nature of the sputtering plasma is not well
understood. In recent years, there has been a signiﬁcant effort to understand the
sputtering deposition process for complex alloys [Bunshah, 1994]. The goal of the
present work has been to understand the effect of sputtering deposition parameters on
composition of sputtered TiNi and TiNiPd alloys.

There have been many studies on the metallurgical and mechanical properties of
TiN i thin ﬁlms, but very few research studies have focused on the effect of sputtering
parameters on the production TiN i SME thin ﬁlms. The sputtering process is so

complex, and little is known about the effects of deposition parameters on the composition

of sputtered alloys. Many experimental results are empirical and are concerned with alloys
of particular use in the semiconductor and computer industry. Compositional shifts from
preferential resputtering have been observed in sputtered superconductive YBaCuO and
BiSrCaCuO thin ﬁlms. In this case, small deviations in composition ﬁ'om stoichiometry
can dramatically change the superconductivity characteristics [Grace, 1991; Selinder,
1991]. Likewise, small deviations fi'om stoichiometry in TiN i alloys may shift the
martensite start (M,) temperature signiﬁcantly and also change the mechanical strength
and ductility of the alloy [Wang, 1965].

The above-mentioned studies on superconductive thin ﬁlms used X-ray
ﬂuorescence to observe changes in composition with respect to the target as a function of
sputter gas pressure and deposition temperature [Grace, 1991; Selinder, 1991]. In the
same manner, much of present work has been focused on tracking composition changes as
a ﬁmction of deposition parameters such as target erosion, working gas pressure,
substrate-target distance, and deposition temperature. This thesis will present results and
theoretical models of sputtering yield found in the literature to help develop an
understanding for these composition shifts. In addition, ideas for future experiments will
be presented that will give a further understanding of the sputtering phenomena that affect
composition.

The technical goal of this research work has been to develop a sputter deposition
process that gives reproducible results and yields ﬁlms of a predictable composition. Due
to complexity of sputtering process, no deﬁnitive conclusions have been drawn regarding
the actual mechanisms that determine composition in sputtered TiN i and TiNiPd alloy.

Instead, a list of all the possible phenomena associated with the sputter deposition process

that effect ﬁlm composition will be evaluated in light of the experimental results along
with future experiments, which may lead to better understanding and a solution to this

very important composition control problem.

CHAPTER 2 LITERATURE REVIEW

The purpose of this chapter is to introduce the Shape Memory Effect and the
effects of composition on the thermomechanical properties of shape memory materials.
In addition, background information on the sputtering process will be presented along
with mechanisms that are thought to result in composition shifts during sputter deposition

of alloy systems.

2.] Shape Memory Properties of TiNi

2. I . I Martensitic Transformations

To understand the Shape Memory Effect in TN i, an explanation of the process of
martensitic transformations is useful. The martensitic transformation in shape memory
alloys is a displacive phase transformations that does not require atomic migration to
form the new crystal structure. These phase changes are ﬁrst-order transformations
where heat is liberated when the martensite is formed upon cooling. There is a thermal
hysteresis that deﬁnes the associated transformation temperature and a temperature range
where both martensite and austenite coexist. This type of martensitic transformation is
sometimes described as athermal since the driving force for the phase transformation is
the free energy change of the transformation. Thus, the reaction is not rate dependent,
since thermal activation is required.

The martensite transformation produces austenite/martensite (AIM) phase
boundaries, which are highly glissile and are either partially or fully coherent. At ﬁrst

observation, the lattice appears to undergo a simple shear between the (NM) phases, like

twins, however, this simple shear will not produce the martensite symmetry or lattice
parameters. The actual deformation is a lattice invariant-plane strain that gives the
undistorted habit for deformation and accommodates the new lattice symmetry
martensite. The mechanisms of transformtions are complex, and the phenomenological
crystallographic theory, or Wechsler Lieberman Read (WLR) theory, offers
crystallographic explanation of the behavior of martensitic phase transformation based on

experimental observations. The following paragraphs discuss this theory in more detail.

2. I . 1.1 Phenomenological Theory

Since martensitic transformation happens sometimes at the speed of sound, and
since actual observation of the transformation is impossible, the WLR theory offers
logical steps for the transformation that correlate with experimental conclusions. The
ﬁrst logical step in the martensitic transformation is a lattice deformation, which creates
the new lattice symmetry. This deformation is often termed as a Bain strain or distortion,
since Bain was the ﬁrst to observe this phenomenon. If the accommodation deformation
does not cause any structural change in the newly formed crystal, there must be an
undistorted crystallographic habit plane to produce the coherent phase boundary between
the austenitic and martensitic structure. A simple shear created ﬁ'om twinning will not
produce an undistorted reference plane. Therefore, there must be an invariant plane
deformation that produces an undistorted habit plane and the experimentally observed
martensite phase symmetry and lattice parameters [Wayman, 1983]. The ﬁnal step in the

phenomenological theory is a rigid body rotation of the invariant plane to bring it into

lattice correspondence with parent and product phases, which can be imagined by

superimposing the parent and product lattices.

2.1.1.2 Martensite Internal Twin Structures and Variant Interfaces

The lattice deformation in TiN i is shear—dominated, unlike steels that have both a
shape and volume change [Duerig, 1990]. The shape change must be accommodated by
the surrounding austenite lattice structure, which can be accomplished by either slip or
twinning. The slip deformation is permanent and is common in steel martensitic
transformations, while twinning is the preferred accommodation deformation in shape
memory alloys and creates twin boundaries that are reversible [Duerig, 1990]. The twin
structure is self-accommodating due to the creation of internally twinned martensite
variants (called transformational twins) that result in no net macroscopic shear strain in
the crystal, and there can be as much as 24 martensite variants formed from one austenite
crystal [Duerig, 1990].

There have been many investigations into the crystallography of the internally
twinned structures that are produced in the TiN i alloys. Matusumoto et al. have used the
deformation twinning theory of Bilby-Crocker to calculate the twinning elements of TiN i
single crystals observed by Transmission Electron Microscopy and X-ray Diffraction.
Table 2-2 shows the values for the twinning modes in the martensitic transformation.
Matsumoto et al. used a tensile stage to deform a TiN i single crystal, and the twin plane
orientation of the stress-induced martensite was observed with a Laue camera and
diffractometer. Their theoretical calculations in Table 2-2 conﬁrmed that the two

twinning systems possible during the experiment were the {111} Type I and <110> Type

H twins, however the experimental results concluded that Schimid factor was zero for
both {111} Type I twinning. The habit plane calculated ﬁ‘om this study was (0.8684
0.26878 0.4138), which deviated from the theoretically predicted plane by 3.3 degrees

[Matusumoto, 1987].

Table 2-1. Twinning modes [Matusumoto, 1987]

 

 

 

 

 

TWiIl K1 Tl] K2 n2 5
{l l 1} [.054043 (0.24695
(111) [211] 0.30961
Type I 0.45957 1] 0.50611 1)
{111} [1.51172 (0.66875
(111) [21]] 0.14222
TypeI 0.51172 1] 0.3375 1)
<110> (.72053
[011] (011) [1.57271 1 1] 0.28040
Type H 1 1)
Compound (00]) [100] (100) [001] 0.23848

 

 

 

 

 

 

 

 

Nishida et al. did a similar study of the transformation twin system in TiN i
martensite using high-resolution transmission electron microscopy of room temperature
electropolished martensite thin foils. The ﬁrst observation of this study was that the
(001) compound twin was formed by an instability of the martensite during high
temperature electropolishing procedures and was not observed at lower electropolishing
temperatures. This is in accordance with the theoretical calculations by Matsumoto et al.,
which show that the compound twin mode should not exist in the martensite. Nishida et
al. also observed by electron diffraction and microscopy that the <110> Type II twin was

found in the martensite. However, the two studies conﬂicted on the issue of the {111}

Type I twin system, which Nishida et aI. believed to exist due to the additional mutual
self-accommodation between the martensite variants by <110> Type II twinning. The
transformation TiN i produces pairs of <1 10> Type II transformation twins in separate
variants that are accommodated by {111} Type I twinning. However, Matsumoto et al.
believed tlmt this was an artifact of formation of TEM thin foils and was not an
accommodation mechanism. Since the Matsumoto study focused on a stress-induced
martensite that had a tensile axis orientated such that there was no resolved shear stress
on K plane of the {111} Type I twin system, it was concluded that Type I twins do not
form during stress induced martensite formation. The basic conclusion of the above
studies is that the type of self-accommodating twin structure is governed by both internal
and external stress ﬁelds in the austenite.

Twin boundaries in the martensite have a low energy and are quite mobile
[Durieg, 1990]. If a stress were applied to the self-accommodated twin structures, the
variant and twin boundaries between the transformation twins would easily move to
produce a shape strain that would accommodate the applied stress. This would result in
the conversion of non-stress-preferred variant crystal to the orientation of the stress-
preferred variant. If the martensite is sufﬁciently strained, a single martensite variant can
be produced. This mechanism of internal twin boundary motion is termed “detwinning”
and is an important aspect of shape recovery in shape memory materials as will be

discussed in the following section.

2.1.2 The Shape Memory Effect

The Shape Memory Effect (SME) arises ﬁom the crystallographic reversibility of
the self-accommodated twin structure [Duerig, 1990]. The martensite phase has a lower
symmetry than the austenite. As a consequence, there are many ways in which the
martensite variants can be formed and distributed, but only one possible way to revert
back to austenite structure due to the high chemical ordering of the austenite phase
[Duerig, 1990]. Consequently, biased martensite variant distributions created by
detwinning stress must revert to the original lattice orientation, and as a result, any shape
strains caused by the detwinning deformation are recovered. This recovery of shape
strains incurred by detwinning of martensite is known as the Shape Memory Effect
(SME).

The shape memory effect can be described schematically with reference to the
cooling and heating curves shown in Figure 2-1. There is no shape change when the
austenite is cooled to martensite phase. When the SMA material is deformed below the
martensite ﬁnish (Mf) temperature, it will remain deformed until it is heated above the
austenite start (A,) temperature. The shape recovery begins at As and is completed at Af.
Once the shape is recovered at Ar, a subsequent cooling to Mr does not cause a shape
change. The shape memory can only be reinitiated by a repeated deformation of
martensite and heating to Af. Therefore, the shape memory effect shown in Figure 2-1 is
a one-time occurrence, which can be referred to as the one-way SME. The operation of
SME is deﬁned by the characteristic transformation temperatures of the alloy, Ms, Mf, As,
and Ar. Figure 2-1 illustrates the use of electrical resistivity versus temperature curves as

a means for determining the transformation temperatures of a material. Other methods,

such as differential scanning calorimetry (DSC), are also commonly used to gather
transformation temperature and enthalpy data of SMA materials.

In addition to one-way SME, another manifestation of the thermoelastic
martensitic transformation is the Superelastic Effect (SE). SE occurs by isothermal
reversible martensitic phase transformation at temperatures above Af. When the stress is
applied to austenitic alloy, it nucleates a preferred martensite variant within the austenite
matrix. The shape strains caused by the accommodation of the preferred martensite
variant in the austenite can be recovered by releasing the stress due to the reversible
nature of the AIM interfaces. SE can be thermodymmically described by a modiﬁcation
of the Clausius-Claperon Equation, which gives the transformation temperatures as a

function of stress.

2.1.3 Equilibrium Phase Structure and Phase Stability of T iNi

Observing the TiN i binary phase diagram, stoichiometric TiN i can exist at high
temperature and has a steep solvus boundary on the Ti-rich side and a slight solubility of
nickel (see Figure 2-2) on the Ni-rich side of stoichiometry. Under 1250 K, TizNi will
precipitate out of 32 TiN i matrix when the compound is Ti-rich, and TiNi3 is equilibrium
precipitate phase in Ni-rich alloys'. The equilibrium phase diagram has traditionally
shown an eutectoid decomposition of TiNi—9 TiNig + TizNi at 900 K, however,
Wasilewski et al. proposed that the TiN i may remain a stable phase down to room
temperature with a very narrow stoichiometric range, as shown in Figure 2-3.

Furthermore, Van Loo et al., Nishida et al., and Kim et al. reported the existence of

 

l Itsexistence depends on temperature and amount of excess nickel, since excess Ni is
partially soluble in stoichiometric compound.

10

metastable phases other than Ni3Ti on the Ni-rich side, such as Ni3Ti2 and NiaTi3. The
equilibrium phase diagram for the Ni-rich side at room temperature is not well
established, and the composition and structure of the intermellatic phases have been a
source of debate. Most commercially used TiN i SMA is single phase that has small
amounts of intermellatic precipitate phase, usually NiaTi3, distributed in the B2 matrix.
These precipitate phases are generally isomorphous with TN i and produce coherency
strains that affect the transformation and mechanical characteristics of the alloy, which is

discussed in the following section.

2.1.4 Effects of Composition on Transformation Temperatures T iNi and TiNiX Alloys

The martensitic start temperature (M,) is strongly dependent on the composition,
especially on the Ni-rich side of stoichiometry, as shown in Figure 2-4. The M,
temperature can change up to 100 K per atomic percent enrichment in nickel above
stoichiometry. On the Ti-rich side of stoichiometry, the transformation temperatures are
only slightly affected by the enrichment of titanium (15-25 K/ per at% Ti). The changes
in transformations are related to solubility of nickel and titanium in the B2 matrix and to
the formation of precipitates.

The M, temperature’s stronger dependence on nickel content is related to the high
solubility of Ni in the B2 matrix above 775 K [Melton, 1990]. The excess nickel
dissolved in the B2 matrix causes coherency strains and stresses that suppress the Ms
temperatures [Miyazaki, 1999]. In addition, precipitation of nickel rich phases can
increase the M, temperature by forming a more Ti-rich matrix that has a higher M,

temperature [Ishida, 1995]. However, over-aging of nickel-rich precipitates, such as

11

Ti3Ni4 and TiNig, can create internal stress fields that suppress the M, temperature by
constraining the formation of martensite variants when precipitates become semi-
coherent with the matrix [Xie, 1990].

Miyazaki and Ishida explored the aging effect and distribution of Ti3Ni4 on the
transformation temperatures in Ni-rich TiN i thin ﬁlms. They reported that as aging time
increased, the observed M, temperature (See Figure 2-5 (a) and 2-6) increased as a result
of formation and growth of Ti3Ni4, which depletes the dissolved excess Ni in the matrix
[Miyazaki, 1999]. In addition, Miyazaki and Ishida [1999] investigated the aging time of
Ti-rich TiN i thin ﬁlms on the transformation temperatures and showed that it had little to
no affect on the transformation temperatures, as shown in their results in Figure 2-5 (b)
and 2-6. This is expected since Ti is almost completely insoluble in the B2 matrix. The
excess titanium forms TizNi precipitates, which tend to nucleate at the grain boundaries,
and the aging and growth of these precipitates has no affect on the M, temperature
[Melton, 1990; Miyazaki, 1999]. Miyazaki and Ishida [1999] observed the presence of
the R-phase transformation above the Af and lower than expected M, temperatures in Ti-
43.9% Ni ﬁlms (See Figure 2-5 (b) ). This was attributed to ﬁne internal metastable Ti-
rich precipitatesz, which exist as very thin platelets or GP zones distributed 5-10 nm apart
on {100} B2 planes [Miyazaki, 1999]. They believed that these structures have the same
effect as dislocations and semi-coherent Ni-rich precipitates in the bulk material, which
suppress the M, tenrperature [Miyazaki, 1999].

The annealing temperature greatly affects the transformation temperatures due to

difference in amount of precipitate formation and growth with temperature. As

 

2These precipitates were observed in Ti-rich ﬁlm that were crystallized ﬁ'om amorphous
as—deposited thin ﬁlms.

12

mentioned previously, nickel is slightly soluble in the B2 matrix, and according to the
binary phase diagram in Figure 2-2 and at temperatures around 700 ° C, 2-3 at. % excess
Ni is dissolved in equiatomic TiN i. Therefore, transformation temperatures can decrease
if the high solutionizing temperatures drive excess nickel in the solution. The effect of
solutionizing annealing temperature on Ni-rich TiN i thin ﬁlms is shown in Figure 2-7.
Ishida et al. observed Ti3Ni4 precipitates at the lower annealing temperatures but also
observed the absence of these precipitates at the higher annealing temperature of 973 K,
which is related to dissolution of excess nickel [Ishida, 1995].

On the other side of stoichiometry, Ti is almost completely insoluble B2 matrix at
any temperature. The transformation temperatures of Ti—rich ﬁlms increase with
increasing solutionizing temperature, as seen in Figure 2-8. The increase in annealing
temperature will further promote the growth TizNi precipitates [Ishida, 1995]. However,
TizNi precipitate growth should not affect the transformation temperatures, and Ishida’s
discussion about the effects of TizNi growth resulting in increased transformations
temperatures is ambiguous and contrary to results shown by Wang [1965] (see Figure 2-
4). However, this increase in transformation temperatures can be related to the relief of
the localized strain ﬁelds in austenite grains due to growth of TizNi and their incoherence
with the austenite crystal, and therefore facilitate growth of the martensite at higher
temperatures due to this internal stress relief.

Elemental impurities, such as O, N, C, and H, in TiN i can also affect the
transformations temperatures. Since Ti is very chemically reactive, it easily forms
compounds with the fore mentioned impurities that deplete the matrix of Ti, making it

Ni-rich, which would lower the M, temperature. According to Melton [1990], the

13

presence of oxygen in the TiN i alloy can decrease its stoichiometric range and result in
the coexistence of both Ti and Ni-rich precipitates (See Figure 2-9). Thus, Ni-rich
precipitate phases can be present in Ti-rich NiTi, and TizNi can be present in Ni-rich
alloys. The stable oxide formed with TiN i is TiaNiZO, that is isostructural with TizNi,
which makes it difficult to identify.

The addition of ternary elements that substitute for Ni or Ti in stoichiometric TiN i
can shift the transformation temperatures and decrease the thermal hysteresis. The
substitution of Cu for Ni in TN i has shown to narrow the thermal hysteresis, as seen in
Figure 2-10 [Miyazaki, 1999; Melton, 1990]. This same effect also occurs when Pd is
substituted for Ni [Miyazaki, 1997], while the substitution of Hf for Ti widens the
hysteresis [Johnson, 1995]. Palladium additions over 10 at. % can dramatically increase
the M,temperature, while Pd content between 6-10 at. % can decrease M, temperature

below 0°C, as seen in Figure 2-11 [Lindquist, 1988].

2.1.5 Mechanical and Metallurgical Studies of Sputtered TiNi SMA Thin Films
2.1.5.1 Binary T iNi

The strength of the austenite controls two key properties in TiN i alloys: its
resistance to deformation during manufacturing processes, and the force that the alloy can
exert during shape recovery [Duerig, 1990]. High yield strengths can make drawing and
forming of TN i very difﬁcult, at the same time it is desirable to have large temperature
ranges for superelasticity and high recovery forces in shape memory alloys. The
austenitic strength and total plastic strain of TiN i is a strong function of material

composition (See Figure 2-12 and 2-13), since precipitate structures can increase the

14

strength of the material by impeding dislocation motion, but as consequence, it also
reduces the ductility of the material.

The aging of Ni-rich precipitates, such as Ti3Ni4, can increase the austenitic
strength and critical stress for slip, but over-aging can also decrease the austenitic
strength The size of the Ti3Ni4 precipitates has a profound effect on the critical stress for
slip when it is less than 100 nm in diameter, as seen in Figure 2-14 [Ishida 1999a]. The
high yield stress of Ni-rich ﬁlms can be attributed to ﬁne Ti3Ni4 precipitates and ﬁne
grain structure3 [Ishida, 1996]. In addition, transformational strains are changed by
annealing Ni-rich ﬁlms, which increase with increasing annealing temperature, as seen
Figure 2-13 [Ishida, 1995]. This increase in transformation strain is attributed to the
growth of Ti3Ni4 precipitates, which lose their ability to suppress the growth of
preferentially oriented rmrtensite variants as they increase in size [Miyazaki, 1999].

Ti-rich ﬁlms that are sputtered at room temperature and crystallized at 773 K also
exhibit higher tensile strength than equiatomic TiN i thin ﬁlms, due to precipitate
strengthening from the Ti-rich GP zones, as seen in Figure 2-13 [Ishida, 1999]. Bulk
(melt-solidiﬁed) TiN i materials that have over 1 at. % excess-Ti are very brittle and do
not show ductility anywhere near to the 20 % reported by Ishida et al. in sputtered ﬁhns
of the same composition [Ishida, 1999]. This lack of ductility is related to the
precipitation of TizNi at the grain boundaries, which promotes intergranular ﬁ'acture
[Matsrmaga, 1999]. Ti-rich ﬁlms sputtered by Ishida et al. have a different precipitate
structure than bulk materials, and consist of Ti-rich GP Zones that increase the strength

of austenite but allow the material to remain ductile. Over-aging or high temperature

 

3 Sputtered TiN i ﬁlms have an average grain size 200-1000 nm, which is much smaller
than the bulk materials, where the average grain size is 100-200 um.

15

annealing of Ti-rich ﬁlms leads to a slight increase in tensile strength, but greatly reduces
the overall ductility, as seen in Figure 2-16 (a) and (b). The ﬁlms in Figure 2-16 (b)
showed no apparent plastic deformation, but still exhibited a large recoverable
deformation strain from stress-induced martensite transformation and martensite variant
detwinning [Ishida, 1999]. The effects of annealing temperatures on transformation
strain are shown in Figure 2-15 [Ishida, 1995]. The strain-temperature curves at low
constant stress have a two-stage transformation hysteresis that is associated with R-phase
and martensitic transformation, while high constant stress strain-temperature curves have
a single-stage martensitic transformation. As seen in Figure 2-17, there was very little to
no plastic strain at low constant stress, while high plastic strains exist at high constant
stresses, which suggest that the R-phase transformation can suppress dislocation

formation in martensitic transformation [Ishida, 1995].

2.1.5.2 T i-(Ni +Pd) Structure and Mechanical Properties

As stated earlier, substituting more than 10 at. % Pd for Ni in equiatomic TiN i is
useful for increasing the Ms temperature. The Pd content can also change the crystal
symmetry of martensite, which is B19’ monoclinic for composition less than 15 at. % Pd,
B19 orthorhombic for composition greater than 20 at.% Pd, and mixed monoclinic and
orthorhombic when compositions are between 15 and 20 at. % Pd [Sivokha 1983].
Sivokha et al. also reported the parent phase of all compositions and additions Pd of Ti-
Ni+Pd to be the B2 structure. The effect of Pd composition on the lattice parameters of
Ti-(Ni+Pd) B19 martensite are shown in Figure 2-18, which is from the data reported by

Lindquist, Sivokha et al., and Shugo et al. [Lindquist,1988]. Lindquist reported that 10

16

at. % Pd alloy was B2 austenite at room temperature (300 K) and had a lattice parameter
a=0.309 nm, which was determined using the Cohen method and the difﬁ‘action peaks of
(211) and (222) [Lindquist, 1988]. The lattice parameter as function of Pd content data
collected by Lindquist and Shugo et al. are shown in Figure 2-19.

Limited research work has been done on the mechanical properties of sputtered
TiNiPd ﬁlms. However, Miyazaki et al. have shown that Ti-26.4 at. %Ni-21.8 at. % Pd
ﬁlms have high transformation temperatures, a high resistance to slip deformation, and a
perfect shape recovery at a constant stress of 200 MPa, as seen in Figure 2-20 [Miyazaki,

1999].

2.1.6 T iNi Thin Film Sputtering Composition Experiments

The compositions, crystal structure, morphology, and transformation temperatures
of Ti-(Ni+Cu) and periodic multi-layered Ti-(Ni+Cu) and Ti thin ﬁlms were studied by
Chang et aI. [Chang,199l; Chang 1992]. These ﬁlms were sputtered using a triode
sputter gun, and in the composition results, a Ti depletion in sputtered—deposited ﬁlms
relative to target composition was observed. Chang reported the reason for this depletion
was related to energetic bombardment of neutral species on the growing ﬁlm, resulting in
a preferential resputtering effect of Ti [Chang, 1992]. This contradicts published data on
the sputtering yield of pure Ti and Ni, since Ni has a much higher elemental sputtering
yield than Ti [Waits, 1978]. Thus, a Ni depletion in the growing ﬁlm may have been
produced by a preferential resputtering effect. However, Chang indicated that the
sputtering of TiN i compounds might have different sputtering properties than pure

elements, which caused the Ti-loss. To understand the sputtering behavior of amorphous

17

 

TiN i alloys, Chang designed a series of ion beam assisted deposition (IBAD) experiments
and plotted the Ti-depletion as function ion to atom arrival ration (VA) for various ion
energies, which is shown in Figure 2-21 [Chang, 1992]. Ti depletion increases with
increasing I/A ratio, and dramatic increase in titanium-nickel sputtering ratio (Y=Yn/YNi)
was also observed at low ion beam energies [Chang, 1992]. In low ion energy regimes
(<100 eV), the sputtering yield changed dramatically and was sensitive to the slight
variation in ion energy. At intermediate level ion energies of 200 eV in argon, pure
titanium sputtering yield is 0.2 atoms/ ion, while pure nickel has a sputter yield of 0.7
atom/ion [Vossen, 1971]. This would indicate that titanium-nickel sputtering ratio should
be less than 1. However, Chang concluded ﬁ'om experiments that the sputter yield of Ti
was greater than nickel in the low to intermediate regimes (<500 eV), and the sputter
yield ratio of Ti to Ni changed from 9 at 50 eV to 1.75 at 500 eV. These results are
inconsistent with elemental sputtering data, but offer an insight into the preferential
sputtering Ti in amorphous TiN i alloys. The discrepancy is probably related to the

relative binding energies in the pure elements and in the compound.

Table 2-2. sputtering‘ Conditions and Film Composition [Ishida, 1993].

 

 

 

 

 

 

 

Ar Gas Pressure RF Power Substrate Temperature Fllm Composrtlon (at.
Run %)
(Pa) (W) (K) T, N,
1 13.3 400 523 50.2 49.8
2 6.7 600 573 50.5 49.5
3 6.7 400 523 50.6 49.4
4 6.7 200 523 51.6 48.4
5 0.67 400 523 48.6 51.4

 

 

 

 

 

 

 

4Target Composition was Ti-SO at. % Ni, and target-substrate distance was 82mm.

18

 

Ishida et al. produced TiN i thin ﬁlms using RF magnetron sputtering5 and
reported the composition for different RF powers, working gas pressures, and substrate
temperatures, which is shown in Table 2-3 [Ishida, 1993]. Their studies were not intended
to observe the effects of sputtering parameters on composition, however it is interesting
to observe the tends in composition that resulted. Ishida et al. used an electron probe
micro-analyzer (EPMA) to determine that composition, which has a precision similar to
energy dispersive spectroscopy (EDS) methods (i025 at. %). Two interesting titanium
composition shifts are noted, a Ti- depletion at low working gas pressure in Run 5, and a
Ti enrichment at low RF power in Run 4. This result coincides with the results observed
by Chang, who found a Ti-depletion at low working pressures [Chang, 1992]. Reﬂected
neutrals leave the target with a large fraction of the initial ion energy, and these low ion
energies might not create reﬂected neutrals that have sufficient energy to breach the
sputtering threshold of Ti at the growing ﬁlm Therefore in Run 4, there might not be
preferential resputtering of Ti due to low ion energies, which results in Ti enrichment.
However, other factors may be involved in this particular observation of Ti-enrichment,
such as differential angular distribution of sputter particles and non-steady-state sputter
conditions at the target surface. Therefore, none of these results and comments can be
conclusive, however the observed composition shifts due to various sputtering parameters
is prevalent.

J. C. Lee has also done an insightful investigation on the effects of plasma energy,
working pressure, and working distance on the composition of sputtered TiN i thin ﬁhns.

He believed that during steady-state conditions, where the sputter ﬂux has the same

 

5 .RF magnetron sputtering does not offer any advantage over dc magnetron sputtering,
Since TiN i is conductive material.

19

composition as the target, the preferential resputtering of Ti was the dominating effect for
the overall Ti-depletion observed regardless of the effects of angular distributions and
scattering differences between Ni and Ti [Lee, 1994]. However, when an enrichment of
Ti was observed, it corresponded to formation of target surface morphologies exhibiting
microscopic cones, etch pits, and ridge structures, which lead to the preferential
sputtering of Ti over Ni [Lee, 1994].

Lee also studied the effect of working gas pressure and throw-distance on
composition in binary TiN i, which is shown Figures 2-23 and 2-24 [Lee, 1994]. The
increase in pressure will thermalize or reduce the energy of incident particles to the
growing ﬁlm and cause sputtering ﬂux to become diffuse due to scattering events. The
former effect reduces the energy of energetic neutrals and thus would reduce the
resputtering effect, which makes Lee’s conclusion of increase in resputtering
questionable. However, since Ti is the lighter sputtering species it would scatter more
diffusely, and less Ti would reach the growing ﬁlm, which is a more probable reason for
the observed Ti-depletion. The increase in working distance has the same effect as
increasing pressure, since more collisions with sputter gas particles and scattering are
possible at longer distances. Again, longer distances do not increase resputtering and in
fact should reduce the resputtering effect.

Lee also conducted some experiments to examine the effect of sputtering power
on the composition of TiN i ﬁlms. The ﬁlms sputtered at 150 W had a composition
resulting from steady-state sputtering conditions and a Ti-depletion that Lee attributed to
resputtering. While, the Ti enrichment observed at 300 W sputtering power was

attributed to non-steady state sputtering due to topographical effects, such as the

20

formation of cones. The off-axis depletion of Ti was attributed to an unbalanced
magnetic ﬁeld that results in an additional bombardment of electrons and ions at the outer
edge of the substrate. The enrichment of Ni in sputter ﬂux and ﬁlm composition at 450
W was concluded to be caused by resputtering effects and rarefaction of the sputtering
gas, which is explained by the increase in the plasma voltage as the power increased ﬁom
300 W to 450 W.

The conclusions made by Lee are questionable, but his observations and data
offer insight on composition control problems in sputter deposition of TiN i and the
complexities of the sputtering process, which are discussed in detail in the following

sections.

2. 2 Sputter Deposition

Sputtering can be deﬁned as the ejection of particles ﬁ'om a condensed matter
target material due to the bombardment of energetic projectiles. The greatest advantage
of sputter coating is that virtually any target material can used to form a thin ﬁlm.
However, since sputtering is a complex multi-parameter process, it is difﬁcult to create
predictable results, especially in the case of alloys. Thus much research has been
conducted in order to theoretically understand the physical nature of sputtering and to
develop an empirical formula for processing materials of interest by sputtering.

Sputter deposition is a vacuum coating process that employs a target material
(source material for the thin ﬁlm), a substrate where the thin ﬁlm is grown, a noble or
reactive sputtering gas, and sputter gun that provides for water cooling of the target and

magnetic enhancement of the plasma. Typically, the substrate is mounted opposite of the

21

target material in a vacuum chamber, which is initially evacuated to 1045 to 10'‘0 Torr6.
Then the chamber is backﬁlled with the sputtering gas to pressure between 1 and 100
mTorr, depending on the type of source and thin ﬁlm process. A negative potential7 is
applied to the target (cathode) to ionize the gas and initiate ion bombardment of the
target, while the substrate (anode) is grounded or kept at a negative potential relative to
the sputtering plasma“. A schematic of a sputtering apparatus is shown in Figure 2-26,
which employs a diode type gun [Bunshah, 1994]. The plasma used in the sputtering

process is complex and is discussed in the section below.

2. 2.1 Principles of DC Sputtering Glow Discharge

The energetic ions that bombard the target during sputtering are generated by a
glow-discharge, which is named for the light emitted ﬁ‘om excited atoms. The glow
discharge used in sputtering is known as an ‘abnormal negative glow’ [Coburn, 1980],
which is self-sustaining type of plasma in which there are an equal number of positive
ions and negatively charged particles [Chapman, 1980]. The voltage-current
relationships for a DC discharge are shown schematically in Figure 2-27 [Mahan, 1999].
The primary region of interest on V-C curve is the breakdown regime, where a sudden
drop in voltage occurs that sustains the discharge at a ﬁxed current value. During
breakdown, ions acquire enough kinetic energy that collisions with the cathode generate
secondary electrons, which are emitted from the target surface of the cathode and

accelerated towards the anode and collide with sputter gas atoms producing ions. Thus,

 

6 The ultimate base pressure depends on process and the amount of allowable residual gas
contamination, and reactive sputter species can be greatly affected by residual oxygen
and water vapor.

7 Typical voltages range from 5 keV to 200 eV.

22

the cycle of electron-ion formation is repeated, and the plasma is self-sustaining. After
breakdown, current increases at a constant due the increase in secondary electron
emission at asperities on the target surface that cause a non-uniform current density
across the target surface. The plasma is sustained mainly by impact ionizations in the
cathode sheath, but this is not the practical plasma used in sputtering. After normal glow
regime, the abnormal glow conditions is reached where the entire target surface area used
and current density is uniform. The abnormal glow regime used in DC magnetron
sputtering typically obeys the voltage-current relationship given by Equation 1.

I at (V)" (2 — 1)
where n in is the range of 5-15 [RossnageL 1988].

The self-sustaining discharge has a particular structure, as seen Figure 2-28. The
Crookes Dark Space or Cathode Dark Space, according to Maissel et al. [1970], is the
“region of primary interest its thickness is approximately the mean distance traveled
by an electron ﬁ'om the cathode before it makes an ionizing collision.” The term “dark”
comes for the fact that this region does not emit light or contain excited particles since
too few electrons are present. This region is commonly referred to as the cathode sheath
in processing plasmas but it is several times thicker than the actual cathode sheath. The
actual cathode sheath9 in practical plasmas is a region in which ions are accelerated at
energy equivalent to the cathode voltage drop (‘fall’). This voltage potential between the
plasma and cathode is larger than applied negative potential to the cathode since the

plasma slightly positive [Mahan, 1999].

 

: This is known as biased substrate sputtering.
A sheath is a volume space charge region, where electrical charge density is not equal to
zero and it surrounds the cathode and anode.

23

The negative glow region of the self-sustaining plasma is where primary electrons
expend their energy and impact ionizations occur. It has a visible light emission with
colors related to the excited neutral species. Classically, the positive column extends
ﬁom the negative glow to the anode, which has an electrical ﬁeld sufficient to transport
the discharge current to the anode and produce ionization that makes up for wall losses
[Bunshah, 1994]. However, most modern sputtering setups place the substrate in the
negative glow region, and the positive column does not exist. As a consequence, higher
sputtering gas pressures must be used due to the small inter-electrode spacing, which
causes loss of electrons to the walls or anode before they collide with gas atoms to

produce ions.

2. 2. 2 Sputtering Mechanisms

Sputtering is a statistical process that results from a momentum-exchange
collisional cascade in the near-surface region of the target that is initiated by
bombardment with incident energetic projectiles [Bunshah, 1994]. If an ion strikes the
target at near surface-normal angles, it is obvious that a single collision will not produce a
sputter particle (sometimes referred to as ejected recoil). The momentum of the incident
ion will drive atoms into the lattice structure of the solid ﬁ'om the single binary collision,
and the momentum vector of the target atom clearly cannot be altered by more than the
90 ° that is necessary for it to leave the solid. Figure 2-29 shows a schematic of collision
cascade resulting for an incident ion that creates a sputter particle. Computer simulations
on Cu with 600 eV Ar ions shows that the collision cascade radius is on the order of 10

nm, and those that intersect surface and transfer enough momentum to result in recoil are

24

few [Harrison, 1978]. Statistical analysis shows that sputter ejection rarely occurs ﬁ'om
collision cascades that develop more than 5 atomic monolayers below the surface, and
most of the energy transferred to the lattice is lost to heat [Bunshah, 1994]. Because of
the large amount of heat transferred to the target upon sputtering, most sputtering systems
and processes provide for cooling of the targets to avoid equipment damage and bulk
diffusion of constituents in multi-element targets.

There are three different energy regimes of collision cascades classiﬁed by
Sigmund [1981], which are displayed schematically in Figure 2-30. The low energy
single ‘knock-on regime’ is related to low projectile energies just above the threshold
energy (energy required to produce sputter atoms) that result in very few sputtered atoms.
In this regime, sputtering results from a single binary collision between surface atoms and
internally backseattered ions or neutrals. In the linear cascade regime, the sputter yield is
proportional to the surface binding energy, and the number of recoils is still low. Most of
the models regarding sputter yield assume this type of cascade since it is the dominant
meclmnism in most commercial and experimental sputtering processes. The last regime
is the spike regime that is sometimes referred to as thermal sputtering since atoms are
ejected in ways similar to vaporization, which is a thermal process rather tlmn collisional
[Sigmund, 1981]. This regime is typiﬁed by a high sputtering yield, termed as ‘excess
yield’, and it lms a nonlinear type of cascade, which results ﬁ'orn a combination of high

projectile energies, heavy mass projectiles, and low surface binding energies.

25

2. 2.3 Sputtering Yield

The sputtering process is quantiﬁed in terms of sputtering yield, which is deﬁned
as the number of recoils per incident ion. The yield is dependent on the surface binding
energy of the target, the energy of the bombarding species, the mass of projectile ion,
incidence angle of the bombarding ion species, and the amount of target conditioning
caused by previous sputtering. The sputtering yield is relatively insensitive to the target
temperature in that most sputtering applications that operate in the linear cascade regime,
where evaporation rates are orders of magnitude lower than the sputtering rates
[Sigmund, 1981].

When the projectile type and energy is held constant, the yield increases with
volatility of the target material, which is related to its surface binding energy. Kelly
[1984] observed a periodic trend in sputtering yield versus atomic mass of the target
material, which was related to the heat sublimation of the material that is roughly equal to
the surface binding energy of the material. He determined that the yield was inversely
proportional to the surface binding energy. These trends are shown in Figure 2-31.

In the sputtering, a threshold projectile energy must be produced to initiate
particle ejection that is also related to the surface binding energy. These values have
been measured experimentally for a variety of materials, and the values for Ni and Ti are
shown in Table 2-3. Theoretical values of threshold energy can be calculated using a
simpliﬁed hard sphere collision model that predicts the amount of energy transfer for a
projectile to a target atom and their ﬁnal direction after impact, which can be determined
by conservation of momentum. Mahan [1999] used this model and developed the

following equation for the threshold energy assuming collision cascade of 4 atoms.

26

E”, = 4U” (2 — 2a)

7
4m pm,

my (2 — 2b)

where U,b is the surface binding energy , 7 is the projectile energy transfer ﬁlnction, mp is

 

the mass of the projectile, and m, is the mass of the target atom

Table 2-3. Threshold Energy in (eV) for Argon Sputtering.

 

 

 

Element Up,“ 7 Theoretical Eh Experimetal" Ed,
Ti 5.10 0.992 20.6 32
Ni 4.60 0.964 19.1 24

 

 

 

 

 

 

 

' Thermodynamic data, U,b=A,.,H(298 K)/N,, from Weast [1975]
" Anderson and Bay [1981]

After the threshold energy is reached, there is a sharp increase in yield with the projectile
kinetic energy, followed by constant yield with increased ion energy [Mahan, 1999].
Figure 2-32 shows the yield curves for tungsten as function of the projectile energy. The
general shape of the yield curves are due to the energy transfer mechanisms in the
collisonal cascade and the nuclear stopping power of the projectile, which is the rate of
nuclear energy loss of the projectile per unit distance in the target material [Mahan,
1999]. The downward trend in the Tungsten yield data at high ion energies is due to the
decreased nuclear energy loss probability in the near-surface region and the electron
energy loss, which is a projectile stopping mechanism that does not generate recoils
[Mahan, 1999].

Argon is the typical choice for sputtering gas used in commercial processes

because it is inexpensive and gives moderately good yields. However, heavier noble

27

gases can result in much higher sputtering rates. Figure 2-33 shows the effect of
projectile type on sputter yield as a function of target atomic mass in the form of contour
plots. For heavy atomic mass targets, the yield increases with increasing projectile mass,
while lighter mass targets have a maximum yield at intermediate projectile masses,
usually Krypton. The behavior of the projectile mass on sputter yield is related to
variations in the projected range of projectiles in the target [Mahan and Vantomme,
1999]. If the projectile can transfer most of its energy to atoms that are near the surface,
it will lead to higher sputtering rates. Thus, heavy mass projectiles in light atomic mass
targets will have longer projectile ranges and will transfer most of their energy to atoms
at deeper depths in the target, which results in fewer ejected recoils.

In addition to projectile mass, the angle of incidence of the energetic ion can
affect the sputtering yield. The increase in sputtering at speciﬁc angles is related to the
penetration depth in atomic lattice, where shorter depths lead to higher sputtering rates.
Thus, the yield is inversely proportional to depth. Assuming a given, the projected range
of a projectile at a ﬁxed ion energy and mass, the depth of penetration varies with the
cosine of the angle of incidence measured from the surface normal (depth=Rpcose). This
would indicate that maximum yield would be obtained at glancing angle near 90°.
However due to high probability of reﬂection or escape of the still-energetic projectile at
high angles of incidence, the number of recoils produced is greatly reduced, and thus the
sputter yield is reduced [Mahan, 1999]. Yamamura et al. have performed experimental
studies on angle of incidence and developed empirical relations of yield as a function of

projectile incidence angle. Figure 2—34 shows Yamamura’s theoretical calculations of

28

yield compared with experimental results that show a yield maximum for incidence at
around 70° for several materials.

Lastly, target condition has a strong effect on the sputtered yield. The presence of
oxide layers, evolution of surface morphologies during sputtering, and the accumulation
of implanted ions can cause reproducibility problems in yield results [Mahan, 1999].
Towsend et al. recommended that yield measurements be performed on targets which
have achieved steady state by previous conditioning such that the yield is constant with
further sputtering. The acceptable steady state of the target surface is accomplished for
an ion dose of 1016 ions/cm2 that amorphorizes the surface without blisteringlo. The
effects of the target surface morphology on sputtering will be discussed in more detail
later.

Much of the sputtering yield data found in literature are determined
experimentally, and many of the theoretical models and formulae are empirically related
to experimental data. Matsunami et al. [1984] has compiled a broad range of sputtering
yield data and developed at empirical yield formula that is described below.

The empirical formula for yield by Matsunami is as follows.

* 8.478Z Z “2 2'8
Y(E)=0.42 “ 93"“) - 55m” ' “(Tl (2—3)
U,[1+0.35U,,s,(a) (2,, +2, ) (mp+m,) E

 

where U,b is the surﬁace binding energy, Zp and Zr are the atomic number of projectile
and target respectively, mrand mp are the mass of recoil and projectile, and E is the
energy of the projectile. The three empirical parameters are at“, Q, and E,;., which

physical meaning were not explained in detail. a* and E,;. are calculated using the

 

1° Macroscopic erosion phenomena caused by occluded gas [Scherzer, 1983].

29

following empirical formulas, and Q is tabulated in Appendix B of Chapter VII in

Physical Vapor Deposition of Thin Film by Mahan [1999].

 

 

mp mp

0.4 1.29
a*=0.08+0.1640[m’] +0.0145-[m'] (2—4)

and

-1 1.24
Erh=Usb'[:1.9+3.80[mr] +0.134-[m'] ] (2-5)
mp mp

The reduced energy is given by the following.

 

 

 

0.03255 . m, E (2_6)

= 2/3 273 1/2
2P2,(2p +2, ) m, +m,

e(E)

Lindhard’s elastic (nuclear) reduced stopping cross section is approximated with

Equation 2-7.

_ 3.44172 1n(s + 2.718) -
1 + 6.35572 + a(- 1.708 + 6.8825)

 

s..(£) (2 — 7)

The inelastic (electronic) reduced stopping cross section is approximated in Equation 2-8.

 

(m + mr)3/2 22/132,1/2
s,(s) = 0.079 0 trig/2’"? (2:3: 2,2,3)“ 5‘” (2 — 8)

Table 2-4 displays yield calculations for Ti, Ni, and Pd for various projectile energies
using Matsunami et al. formulas along with experimental data by Vossen. The empirical
calculation of yield using Matsunami are close to the values reported by Vossen but slight
differences exist, and the disparities increase with increasing ion energy and recoil mass.
The disparties between experimental and empirical calculated yield can be explained by a
deviation from the assumed binary collision model that occurs at high ion energies and in
materials with low surface binding energies and high mass [Mahan, 1999]. In general,

Matsurnani et al. formulas are well established and are used by many researchers, hence

3O

their yield data calculated for Ni, Ti, and Pd will be in the remaining discussion. As will

be seen later, the critical question regarding the use of these empirical equations such as

these concerns the correct valve of U,b for a species in the alloy. Though U,b may be

higher for Ti than Ni in monoatomic targets, it is not at all clear that this would be the

case for the same species bound in the intermetallic compound.

Table 2-4. Sputtering Yield Data for Ni, Ti, and Pd.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Element Nickel Titanium Palladium
Surface Binding Energy in
eV (W east, 1975) 4.6 5.1 4.22
Surface Binding Energy in
eV (Gschneider, 1964) 4'47 4'9 3'91
Surface Binding Energy in
eV (Matsunami, 1984) 4.44 4.85 3.89
Usb, best 4.50 4.95 4.01
Q (Matsunami, 1984) 1.06 0.58 1.1
Atomic Mass 58.6934 47.867 106.42
Z 28 22 46
alpha-star 0.295 0.275 0.374
Sputtering Threshold Energy,
Eth 21.2 25.9 15.1
epsilon(E) 0.00382 0.0035 0.00193
Sn(epsilon) 0.154 0.149 0.119
Matsunami Yield at 200 eV
(atoms/ions) 0.45 0.19 0.75
Matsunami Yield at 300 eV
(atoms/ions) 0.66 0.3 1.1
Matsunami Yield at 600 eV
(atoms/ions) 1.1 0.52 1.73
Vossen Experimental Yield at
200 eV (atoms/ions) 0'7 0'2 1'0
Vossen Experimental Yield at
600 eV (atoms/ions) 1'5 0'6 2'4
2. 2. 4 Sputtering Sources

2. 2. 4. I Magnetron Planar Diode

Magnetron Sources were developed in the early 19705 and are capable of higher

deposition rates than simple diode sources. In simple diodes, the secondary electrons

31

 

emitted from the target are typically lost to anode or walls. Thus, secondary electrons in
a simple diode have a low probability to impact ionize the sputtering gas. If the path of an
electron were lengthened then it would have a higher probability to impact an ion.
Magnetron sources use a magnetic ﬁeld, which is parallel to the target surface and
perpendicular to the applied electrical ﬁeld, to trap electrons near the surface of the
target. The trapping effect occurs by the drift of electrons in the —E x B direction creating
a cycloidal motion, as shown Figure 2-35. Since many of the electrons are conﬁned near
the surface of the target, there is a higher probability to ionize the neutral atoms. Thus,
the plasma is intensiﬁed in this region due to increased impact ionization. This leads to a
higher ion current density that results in a higher deposition rate, and sputtering is
possible at much lower pressure with magnetron than with a simple diode.

The motion of the electrons in a planar—ring rmgnetron is shown schematically in
Figure 2-36 (c). The radius of the electron motion is deﬁned in Equation 2-9 and 2-10
[Mahan, 1999].

F = qv x B (2 — 9)

where F is exerted force on the electron, q is the charge of the electron, v is the velocity

of the electron, and B is strength of the magnetic ﬁeld.

r= ‘ (2-10)

 

where r is radius of gyration of the electron and m, is the mass of the electron.
Assuming the electron is initially at rest on the target, it is deﬂected away from
the target when a negative voltage or electrical ﬁeld is applied. The force of the magnetic

ﬁeld deﬂects electrons back towards the target at a radius described in Equation 2-10.

32

The maximum height of the electron’s deﬂection from the target surface is deﬁned in

Equation 2-11 [Green, 1976].

 

ymax =l[2me (V_VT)] (2‘11)
B 4

where ym, is the maximum deﬂection and (V- V1) is the cathode fall.

The continuous path of the electrons is annular in shape due to the E x B ﬁeld that
develops a high concentration of ions above a region, which is called the plasma ring.
Subsequently, the target is eroded much more quickly in this region and creates a
macroscopic erosion proﬁle known as a ‘racetrack’ [Bunshah, 1994]. Thus, the main
disadvantage of a planer ring magnetron is poor target material utilization. Later, we will
discuss whether or not the development of this proﬁle can affect composition in

conjunction with the angular distribution of the sputter ﬂux.

2. 2. 4.2 T riode Source

Triode magnetrons use an additional electrode independent of the target to sustain
the plasma. A typical triode conﬁguration is shown in Figure 2-36 [Bunshah, 1994].
Electrons are emitted by a thermionic ﬁlament and are used to replenish the plasma
electron density in addition to the secondary electrons emitted by ion bombardment. This
relaxes the ionization requirement for sustaining a discharge, and consequently, low
working gas pressures and cathode voltagesll can be used. Conﬁning the electrons with
an axial applied magnetic ﬁeld minimizes the radial plasma losses, however the ﬁeld

produces distortions that cause uneven distribution of ion current over the area of the

 

11 Pressure between 0.5 and 4 mTorr and applied voltage (100—500 V) can have ion
currents of several amperes [Bunshah, 1994].

33

target. Typically, the target is eroded in the shape deﬁned by the parallel path of the
electrons across the target, and due to the axial conﬁnement of the mgnetic ﬁeld, the
outer edges of the target are eroded less efficiently. This causes a U-slhaped proﬁle
known as ‘saddling’. The saddling proﬁle is somewhat less pronounced than the
‘racetrack’ proﬁle of the diode magnetron and does not create steep sloping surfaces.
This may be the reason for more compositional uniformity in sputtered ﬁlms as compared
to magnetron diodes.

Triodes achieve much higher deposition rates than simple planar diodes at lower
gas pressures and voltages. However, triodes are less accepted in industrial plasma
processes due to problems in scalability, complex and expensive power supplies, and

vulnerability of the thermionic ﬁlament to reactive gases [Busch, 1977].

2. 2.5 Film Growth and Sticking Coeﬂicients
It has been observed by electron microscopy that ﬁhns form in a series of stages.
These stages are shown schematically in Figure 2-37 and are as follows:

1. Nucleation- small nuclei are statistically distributed over the surface of the
substrate

2. Growth of the nuclei and formation of large islands

3. Coalescence of the islands and formation of a connected network of empty
channels.

4. Filling of the Channels to form a continuous ﬁlm.
As the sputter particles impinge on the surface, they are adsorbed and are referred to as

adatoms. The energy loss of the particle as it impinges on the surface is characterized by

34

the accommodation or sticking12 coefficient that is deﬁned in Equation 2-12 [Eckertova,
1977].

TC — TV

T. _ (2—12)

 

a:

“"1

where T, is the temperature corresponding to the energy of the incident particle, T, is the
temperature of emitted particle, and T, is the substrate temperature. I

The sticking coefficient varies ﬁ‘om 0 to 1, where 0 refers to elastic reﬂection of
the particle and 1 refers to complete accommodation or sticking on the surface. It has
been found that the sticking coefﬁcient is generally 1 if the incident particles are less than
or equal to 25 times the desorption energy on the substrate, which is around 1-5 eV for
most materials [Eckertova, 1977]. Considering the energy distribution of sputter particles
for a linear cascade deﬁned by the Thompson13 formula (Equation 2-13), most of sputter

particles ejected for the target have energies equal to half the surface binding energy.

dY/dEo ocZIE—ELZIL)? (2—13)
0+ sb

where Y is the yield and E, is the energy of the recoil.

Most materials have surface binding energy 1.5 —8 eV, and the energy of particles
impinging on the substrate is typically much less due to scattering and thermalization
events during transport through the sputter gas. Therefore, most sputtered particles do not
posses the energy to readily desorb ﬁom the surface, even at high substrate temperatures.
The kinetic energy that the sputtered particles do posses aides in nucleation and island

formation by encouraging migration of adatoms thru surface diffusive jumps. The

 

‘2 This sometimes is also referred to as condensation coefﬁcient, and different literature
sources interchange the names of this parameter but all have same general meaning.

35

probability of desorption is estimated by considering the mean diffusion time to
physically or chemically absorb at preferred nucleation sites, and the mean residence time
of adatoms [Eckertova, 1977]. Typically, diffusion times and distances are short due to
high deposition rates in sputtering, and re-evaporation or desorption does not
signiﬁcantly reduce the condensation rate of the growing ﬁlm.

There are many factors that can affect the quality of the growing ﬁlm, like
substrate temperature, chemical nature, crystallographic structure, and surface roughness
[Eckertova, 1977]. In addition, the number density of residual reactive gas particles is an
important factor since they can affect the composition and mechanical properties of the
ﬁlm. As discussed earlier, the ultimate base pressure is below 10'7 Torr, and the use of
pure sputtering gas sources can greatly reduce the partial pressures of these containments.
Sputtering gas incorporation” can also have adverse effects of the mechanical properties
by reducing ductility of the ﬁlm. In addition, energetic neutrals, negative ions and high-
energy electrons that are accelerated across the Crookes Dark Space strike with high
energy, and at high doses they can transfer a signiﬁcant amount of thermal energy to the
growing ﬁlm. This can occur especially in the case of the so-called unbalanced
magnetron that has magnetic ﬂux lines that intersect the substrate and causes electrons to
reside near the surface of the substrate, as shown schematically in Figure 2-38. The
energy of incident sputter particles and the heat generated by particle bombardment on

the substrate plays an important role in the growth and development ﬁlm microstructures.

 

'3 The formula shows a maximum number of sputter particles at U,b/2 and a rapid (1/E,)2
falloff at higher energies.

This is caused by energetic sputtering gas ions and neutrals, which are reﬂected by the
target during sputtering, striking the surface of growing ﬁlm.

36

Thus, sputter parameters, such as working gas pressure, substrate temperature, and

sputtering power, can have key effect development of certain microstructures.

2. 2. 6 Sputter Deposition of A lloys
An important advantage of sputter deposition is that the ﬁlms formed have the
same composition as the target, provided that the following conditions exist [Bunshah,
1994]:
1. The target is maintained sufﬁciently cooled such that bulk diffusion of
constituents does not occur.
2. The target does not decompose.
3. Reactive gas and containments are not present.
4. The sticking coefﬁcients are the same for all the components.
5. The gas-phase transport is the same for all the components, such that composition
shifts of the ﬁlm relative to target do not occur due to differences in
thermalization and angular distributions of sputter species.

6. Preferential resputtering of a particular component does not occur.

The sputtering of alloys is complex and ﬁlms compositions are in reality affected
by deposition parameters given in the above list. The sputtering yields of components in
alloys will not be the same as yields measured and calculated for pure elements, since the
surface binding energies are different in the alloy [Bunshah, 1994]. If the surface binding
energies are similar, the component with lighter mass will sputter more readily, and if the

masses are similar, then the more weakly bonded species will sputter faster [Sigmund,

37

1980; Coburn, 1979; Anderson, 1981]. Preferential sputtering yields in alloys will have

an effect on surface morphology as well ﬁhns composition, which is discussed below.

2. 2. 6. I Topographical Evolution of Multi-component Targets

The details of ion/ surface interactions of multi-component materials are poorly
understood and quite complex. Consider the case of a homogenous target that is
composed of two elements that have different sputtering yields. During the initial stages
of target sputtering, the component with higher yield will sputter more readily from the
surface, which enriches the sputter ﬂux in that component, and the composition of the
target surface will become lean in the high sputtering component. This surface
development continues until the lower sputter yield component surface concentration
becomes high enough that sputtering yields of both the high and low components
equilibrate (see Figure 2—39). This surface condition is known as the steady state altered
layer, which creates a sputter ﬂux equal to the bulk target composition. The transition
period to form this alter layer depends on the method of target manufacture, the alloy,
and deposition process. This transition period to a steady state condition can be long for
hot pressed targets made for coarse powders (50-130 pm), which may require 20 um of
target material to be sputtered to achieve steady state. [Dahlgren, 1972]. The thickness
of this surface altered layer is a measure of the escape depth of recoils, which depends on
the target material and sputtering conditions [Bunshah, 1994; Mahan, 1999]. Typically,
thickness can range from a few nanometers to several um [Eltoukhy, 1980; Gillam, 1959;

Greene, 1978].

38

Sputtering of two-phase alloys in which the phases have signiﬁcantly different
sputter yields can develop an irregular surface topography [Navinsek, 1976; Nobes, 1969;
Carter, 1971; Sigmund, 1973]. The faster sputtering material will erode quickly and
remove material around the slower sputtering phase or impurity. This results in etch pits,
ridges, and a surface topography phenomena known as cones, which are shown
schematically in Figure 2-40. The angles of the cones are related to incident ion angle
that results in the highest sputter yield. Lee [1994] observed the formation of etch pits,
ridges, and cone in the sputtering of Ni-rich TiN i targets containing precipitates, which
are shown in Figure 2-41. These cones will eventually be sputtered away but will expose
new impurities and second phases, which will result in the formation new cones.

The sputter ﬂux will reach steady state and have the same composition as bulk
target after an incubation time required to develop the altered surface layer and cone
structures. This period of time is typically referred to as the target “burn in” [Bunshah,
1994], and most commercial sputtering processes presputter the target before depositing a
thin ﬁlm Since there are many factors that can affect the required time for burn in, little
is known about the sputtering times required to condition the target surface, which could
be a few minutes to several hours. Mahan [1999] suggests that the appropriate target
‘burn-in’ is accomplished by presputtering for 1 hour. Evidence in this study for powder-

metallurgy Ti, Ni, Pd targets indicates that this period may be considerably longer.

2. 2. 7 Effect of Recoil Angular Distributions on Film Composition

Behrisch [1981] stated, “For normal incidence that angular distribution may

mostly be described in a ﬁrst approximation by a cosine distribution.” However, Wehner

39

et al. [1960] has observed differences in angular distribution between elements, and
Olson et a1. [1979] has found differences in the angular distributions between constituents
in alloys that vary with incident ion energy. The cosine law of emission is given by the
following equation.

jn(t9)=Yz” cosB/rt (2—14)
where j g is the angular distributed emission ﬂux, Y is the sputter yield, 2+ is the ion
impingement rate, and 0 is emission angle.

This cosine law is related to isotropic angular distribution of recoil velocities in
the collision cascade over an inﬁnitesimal area on an ideal sputter source, but in general,
the extended source overall can be described with an under-cosine type of distribution
[Mahan, 1999]. Blech et al. developed an expression for describing angular distributions
for real sputtering sources, which is shown in Equation 2-15.

dY = Y cost9

—— A ~ 2—15
of!) WI . (p2 sin2 0+cos2 6) ( )

 

where I’m; is the total integrated yield, p is an empirical parameter, and d!) is the
differential solid angle in the direction of emission.

p <1 gives an under-cosine distribution, and p >1 give an over-cosine distribution.
Information on the mass and Z-dependence of p are scant in the literature, but the data
,which exists, strongly suggests that p is more overcosine for less massive species in an
alloy [Olson, 1979].

The type of angular distribution is related to the preferred angle of ejection. If the
preferred direction of emission is normal to the target, the angular distribution has an

over-cosine shape. Likewise, if the preferred direction of emission is more parallel to the

40

target, the distribution has an under-cosine shape. These distributions are shown
schematically in Figure 2-42. As mentioned above, the lighter species of an alloy will
tend to have over-cosine distribution when the ion energies are greater than 100 eV
[Olson, 1979]. This is related to binary collisions between two species that have different
masses. The lighter mass species is more likely to backseatter at angles that would tend
to favor ejection in normal direction, based on the reﬂective collision model developed
by Olson [1979].

Murakami et al. used Olson’s model to explain the lateral composition shifts in
binary alloy TbxFeH. Using an experimental setup that collects ﬁlms at various angles
with the target, they showed that the lateral composition shift was signiﬁcant and had
opposite compositional effects for two different types of targets”. The setup and results
of their experiment are shown in Figure 2-43. The composition shift in the compound
(alloy) target was explained to have resulted by mechanisms described by the Olson’s
reﬂective model. The observed enrichment of the lighter mass Fe at center positions was
attributed to a preferred emission normal to target (over-cosine), while Tb was concluded
to have an under-cosine distribution causing enrichment at lateral position away from the
target. The ﬁlms that were deposited ﬁom the composite (powder) target enriched in Tb
in the center and Fe at lateral position away ﬁ'om the center of target, which was opposite
of the trend observed in the compound target. The composite target was composed of
large distinct regions of pure elements, and the sputtering mechanism can be assumed to
be similar to pure materials. Thus, the angular distributions were concluded to be similar

to pure elemental results. Since Ni, Mo, Fe, Pt, and Ge were reported to have under-

41

cosine distribution by Wehner et al. [1960], it was assumed that elemental Tb had a more
over-cosine distribution than Fe and thus enriched at the center positions. It is obvious
from work of Murakami et al. that angular distributions can have a strong effect on ﬁlm
composition.

For oblique ion incidence, the angular distribution has an asymmetric distribution,
where the maximum yield is shifted to the opposite side (as compared to the incidence
angle) of the surface normal. These asymmetric distributions were observed in Si at an
oblique angle of 60° by Okutani et al. [1980], and the preferred ejection angle from the
surface normal was found to be 4.9° for 10 keV Ar+. Betz et al. [1970] correlated the
sputtering yield to preferred emission angle and developed a universal curve to describe
the relationship, which is shown in Figure 2-44. At high yields (over 50 atoms/ion), the
emission angle is nominally zero, while emission angle approaches 90 ° for yields under

2 atoms/ ion.

2. 2.8 Eﬂects of Thermalization on Thin Film Composition and Microstructure

Most ballistic trajectory models for sputtering and evaporation assume that the
mean ﬁ'ee path is much longer than the throw distance (the distance from the substrate to
the target). In evaporation and ion beam assisted sputtering“5 that operate at pressures of
10'6 and 10“ Torr, respectively, the mean free paths of both processes would be greater
than 30 cm. Thus, the assumption of longer mean paths than throw distances is valid for

typical ﬁlm deposition setups that use throw distances of 5-10 cm. In the case of

 

‘5 The compound target was a homogenous alloy manufactured by hot pressing powders.
The %omposite target was manufactured by randomly embedding 8mm Tb disk into 125
m e p ate.

‘6 This is a method of sputtering using a separate ion gun to generate the projectiles.

42

magnetron sputtering processes that operate at pressures on the order of 1-5 mTorr, the
mean ﬁ'ee path is on the order of a few centimeters, ﬁlrthermore, non-magnetic assisted
diodes that operate at high pressures have even shorter mean ﬁee paths. Thus, the
sputtering ﬂux in magnetron deposition may be appreciably scattered and diﬁuse
[Mahan, 1999].

The scattering is caused by collisions between the energetic sputter particles or
reﬂected neutrals and sputter gas. This transfers energy for energetic particles to the
sputtering gas that can reduce the characteristic temperature1 7 of sputter particles (several
thousand degrees) to the temperature characteristic of plasma sputter gas (~5 00 K). This
process is termed thermalization. The reduction in energy of these particles can affect
properties of the growing ﬁlms, particularly the microstructure. Thermalization can be
beneﬁcial since it increases ﬁlm thickness uniformity by creating a diffuse distribution of
sputter particles that gives better composition control by reducing the preferential
resputtering of constituents in alloy ﬁlms, which is discussed in the next section.

The fraction of thermalization can be deﬁned by the following equations by
Mahan [1999].

F =1— at“ (2 — 16)
where y is the throw distance, k’ is an empirical parameter (often on the order of 10), and
A. is the mean free path deﬁned by equation 2-17 and is the distance traveled before the
ﬁrst collision occurs.

kT

,2, =
‘5sz

(2—17)

 

 

'7 5 eV is equivalent to 60,000 K, which is a typical value from Thompson’s assumption
of sputtering particles having half the surface binding energy.

43

here, It is the Boltzrnan constant, P is the pressure, d is the particle diameter, and T is the

temperature.

Table 2-6 shows mean ﬁ'ee paths of Ti, Ni, Pd, and Ar at various pressures and

temperatures.

Table 2-5. Mean Free Paths"3 ofTi, Ni, Pd, and Ar.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Mean ﬁee path at 300 K (m)
Pressure (mTorr) 2.5 3 4 4.5
ickel , 45.3] "T ' '38:" m -2337 _" ' T25i3llmeL’.
Titanium 32 26.7 20 17.8
ualladuiumvﬂ 37.3 _ 31.1 2. _ 23.3 20.77 m i "m
A__rgon 19.1 15.9 11.9 10.6
Mean free path at 800 K (m)
Pressure (mTorr) 2.5 3 4 4.5
ickel ' i . I 121.5" 8 ’ 101:3 ’ ”75.9 67.5
itanium 85.3 71 53.3 47.4
alladium 99.5 83 62.2 55.3
[argon 50.9 42.5 31.8 28.3

 

 

 

 

 

It is obvious from Table 2-5 that nickel has the longest mean path of all the

species, since it has the smallest atomic diameter. The mean ﬁee path is a strong ﬁrnction

of temperature, which is related to the reduction of gas density at higher temperatures and

thereby reduces the probability of interactions with sputter particles. Table 2-6 shows the

thermalization fraction for Ti, Ni, and Pd at different working gas pressures and distances

assuming k’=10.

 

‘8 The values chosen for pressure, temperature, and working distance will become
obv1ous in Chapter 4, since these values correspond to experimental setup parameters.

44

 

Table 2-6. Thermalization Fractions of Ti, Ni, Pd, and Ar.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

  

  
 

  

 
   
 

    
   

 

 

 

 

 

Fraction of Fraction of Fraction of Fraction of
Thermalization at Thermalization at Thermalization at Thermalization at
300 K, 2.5 mTorr, 300 K, 3 mTorr, 300 K, 4 mTorr, 300 K, 4.5 mTorr,
and y=131 mm and y=l3l mm and y=l3l mm and y=l31mm
% °/o % %
itanium 33.6 38.8 48.1 52.2
on 49.6 56.1 66.6 70.9
Fraction of' Fraction of Fraction of Fraction of'
Thermalization at Thermalization at Thermalization at Thermalization at
300 K, 2.5 mTorr, 300 K, 3 mTorr, 300 K, 4 mTorr, 300 K, 4.5 mTorr,
and y=75 mm (%) and y=75 mm (%) and y=75 mm (%) and y=75 mm (%)
ickel 15.2 17.9 23.2 25.6
alladi 18.2 21.4 27.5 30.3
Fraction of Fraction of Fraction of Fraction of
Thermalization at Thermalization at Thermalization at Thermalization at
300 K, 2.5 mTorr, 300 K, 3 mTorr, 300 K, 4 mTorr, 300 K, 4.5 mTorr,
and —67 mm % and -67 mm % and —67 mm % and -67 mm %
itanium 18.9 22.2 28.5 31.4
on 29.6 34.4 42.9 46.8
Fraction of Fraction of Fraction of
Thermalization at
800 K, 2.5 mTorr, 800 K, 3 mTorr, 800 K, 4 mTorr,
and y=75 mm (%) and y=75 mm (% and y=75 mm (%
6

 

 

 

45

 

 

The effect of the mean free path on thermalization is obvious, since the species
with the lowest mean free path will have the highest amount thermalization. Large
relative differences in thermalization can affect the composition of the ﬁlm by the
preferential reduction in the ﬂux of one species relative to the other even at steady state
conditions where the sputter ﬂux has the same composition as the target.

Turner et al. [1992] calculated the fraction of the incident ﬂux which would reach
the substrate after suffering scattering events during transport, as function of atomic mass
and pressure, using Monte Carlo simulations at a throw distance of 5 cm in Argon, which
is shown in Figure 2-45. They considered both the mass of the species and its binding
energy, which was introduced to account for the initial energy of the sputtered species in
the sense that higher binding energy would imply higher initial energy. Turner did not
explicitly simulate deposition rates for Ni or Ti, but did produce results for Na, Sn, C, Cd,
Pd, Nb, Pb, Au and W. These results are shown in Figure 2- 46 with a superimposed set
of estimated contour lines giving the percent of atoms that actually reach the substrate, as
a ﬁrnction of binding energy (y-axis) and mass (x-axis). The contour lines indicate that
Ti and Ni would suffer the same amount of scattering and reduction incident ﬂux.

The thermalization distance was quantiﬁed by Westwood [1978], which was
calculated as function of pressure for a variety of mass and energy regimes that are
shown in Figure 2-47. His result shows that thermalization distances are short for most
sputter particles masses at pressures commonly used during sputter desposition.

Somekh [1984] did computer simulations on the thermalization of different
atomic mass sputter species in different types working gas with primary interest in the

effective energy reduction of particles according to Thompson sputter particle energy

46

distribution. He based his results (see Figure 2-48) on pressure- throw distance products
and showed that light mass atoms retain more of their initial energy than heavy mass
atoms above 100 Pa-mm The unambiguous conclusion of these studies is that
thermalizat ion is signiﬁcant under practical sputter conditions, and its effect on

composition and microstructures of sputtered thin ﬁlms cannot be ignored.

2. 2. 9 Eﬂects Preferential Resputtering and Sputter Yield Amplification on Composition
Resputtering is removal of sputter deposited material by bombardment of
energetic neutrals or energetic ions. Resputtering can reduce the net condensation ﬂux as
an inadvertent result of the bombardment of the target by energetic ions, which recoil or
reﬂect as neutrals ﬁ‘om the target surface at energies on the same order of magnitude as
the cathode fall potential [Hoffman, 1990]. Resputtering of the ﬁhn can be as little as
few percent to values that approach 100 %. Applying a negative voltage to the substrate
can increase the resputtering rate, thru increased ion bombardment, which is know as
biased substrate sputtering. This is beneﬁcial for microstructure control since the heat
generated by ion bombardrrrent can increase diffusion processes and the development
preferred rrricrostructures. Hoffman [1990] measured the intrinsic resputtering of ﬁlms
for a variety of projectile types and target materials, and he found that the fraction of
resputtering falls on a universal curve. Figure 2-49 shows the Hoffman results for the
ﬁaction of resputtered material as a function of atomic mass parameter (AMP) and the
resputtering ﬁ'actions of the Ti, Ni, and Pd (the AMP for these elements are shown in

Table 2-7).

47

 

Table 2-7. Atorrric Mass Parameters (against Argon).

 

Nickel Titanium Hafnium Palladium Cgrper

 

 

 

 

 

 

10.19004 0.09018 0.63424 0.45414 8' 0.22801

 

 

Hofﬁnan’s results show that for heavy sputtering gas and light target atoms (AMP
<0), the resputtering is a constant 3.5%, which is probably due to self—sputtering. While
light gas atoms and heavy target atoms (AMP >0) are linearly related with the atomic
mass, which is due to an increase in reﬂected neutrals from the sputter target as the target
mass increases.

Resputtering can have an effect on composition when atomic species have
intrinsic resputtering yield differences, like those that occur by sputter yield
ampliﬁcation. The sputter yield ampliﬁcation (SYA) effect is an enhancement of partial
sputter yields to values larger than those of pure elements [Berg, 1999]. The increase in
partial sputtering yield is related to underlying substrate material and the thickness of the
sputtering layer.

The mechanisms for SYA are related to the collisional cascades within the thin
ﬁlm. As mentioned earlier, typical collision cascades that result in ejected recoils occur
at depths of few monolayers of atoms (1-2 run). If the ﬁlm thickness is on same order as
the depth of the collision cascade and the underlying substrate is composed of heavier
mass atoms (as compared to the ﬁhn), a substantial ﬁ-action of the lighter ﬁlm atoms in
the collision cascade will be reﬂected off the heavier substrate atoms and back towards
the surface. Since reﬂection occurs at small impact parameters, this mechanism is
sensitive to the relative difference of atomic mass and not the atomic density of the

materials. Furthermore, the lighter ﬁlm atoms will lose very little energy due to a small

48

kinematical factor, thus reﬂection through multiple collisions between ﬁlm and substrate
atoms is a reasonable explanation for sputter yield increase [Berg, 1999]. The end result
of increased backseattering of the lighter atoms is an increase in the density of collisional
cascades, which are conﬁned to the surface, and thus enhance the sputtering yield. Berg
et al. performed computer simulations of partial sputtering yield of 1 nm thick Al layer
on different Z (atomic number) substrates and show that as Z increases the partial
sputtering yield increases (see Figure 2-50). This conﬁrms that there is a possible SYA
effect, since at high Z substrates the yield is ampliﬁed by twice the sputtering yield
measured for pure Al.

For thin sputtering layers one or two mono layers thick, the mechanism is different
than the one previously mentioned. The partial sputtering yield is independent of the
atomic number of the substrate but it is dependent on the atorrric density of the substrate.
This is due to small thickness of the sputtering layer; the majority of the collision cascade
density occurs in the substrate. Thus, the sputtering of a thin layer depends the collision
cascade density of the substrate, which is roughly proportional with the atomic density of
the substrate [Berg, 1999].

In addition, the SYA effect is observed in alloy target materials containing
elements of differing mass [Berg, 1999]. The enhancement of the sputtering yield is
related to projected range of the bombarding species, which is related to masses of the
target material and projectile. The increase in energy loss of the projectile is in the near
surface of the target. If a target is doped with a heavy species, the projected range of the
projectile will decrease with increasing concentration of the heavier mass species. Thus,

most of the projectile energy will be transferred to regions near the surface that increase

49

the density of collision cascades, resulting in an increased sputtering yield. Another
result of the heavier species/ion interactions is that the ion will retain much more energy
on reﬂection as compared to the interaction with the lighter mass species, due to
kinematical fraction differences. Thus, the reﬂected ion can transfer more energy to near
surface atoms. The consequence of doping the target with a heavier species is that the
partial sputter yield of the light component can be double the value of the pure material
yield. This conclusion is important when considering preferential sputtering, since the
species in multi-element alloy target, which are preferential sputtered is not obvious ﬁom
the pure element sputter yield data.

Preferential sputtering in target exists at early stages sputtering. However as
mentioned earlier, the sputter ﬂux will eventually be equivalent to the target composition
due to the conservation of matter. This steady state condition does not occur in processes
where preferential sputtering exists thru the whole process, like the resputtering of a
growing ﬁhn. Preferential resputtering will tend to decrease the composition of a species
even though the steady state sputter ﬂux is the same as the target composition Thus, the
ﬁlm composition will have a different composition compared to the target. Preferential
resputtering has been observed in a number of multi-component sputtering processes.

For instance, Grace et al. have observed the selective resputtering of Bismuth in Bi—Sr-
Ca-Cu-O ﬁlms. Their results stated that Bi loss was strongly related to the working gas
pressure, where increased loss was found at low working gas pressures and compositions
that resemble the target stoichio metry at high pressures. Resputtering caused by the
bombardment of growing ﬁlm by energetic neutrals, and since thermalization of neutrals

increases with increasing pressure, the resputtering rate was a strong ﬁlnction of the

50

working gas pressure. Selinder et al. also observed a‘ preferential resputtering of Cu in
YbaCuO ﬁlms that was eliminated at high working gas pressure (>9 Pa). They also
observed that Cu loss was a strong ﬁlnction of substrate temperature and concluded that
relevant differences in sticking coefficients of the components resulted in the diminished
Cu. Neither Selinder nor Grace et al. eluded to mechanisms which drive preferential
resputtering. However, conclusions can be made based on the SYA effect and increase

sputtering yields for the lost species.

51

CHAPTER 3 MATERIALS AND METHODS

The goal of the present work has been to study the effect of sputter deposition
parameters on the composition of TiNi and TiNiPd thin ﬁlms. The binary TiNi thin ﬁlms
were formed by two methods, d.c. magnetron and dc. triode sputtering, and experiments
were conducted to compare the effects of target erosion on composition between these
two methods. Additional experiments were conducted to observe the effects of deposition
parameters on the composition of TiNiPd ﬁom sputtered ﬁom powder-metallurgy targets.
This chapter discusses the procedures used to fabricate the thin ﬁlms and to gather data on

composition, thickness, and transformation temperatures.

3 . 1 Materials
3.1.1 Binary TiNi Targets

The dc magnetron sputtering targets used in the binary composition study had a
composition of NisrTiao, as supplied by Special Metals, Inc. and were electrical-discharge
machined (EDM) ﬁom 76.2 mm wrought bar into 3” diameter 1/4” thick disks. The
targets used in triode sputtering experiment were EDM-cut from the same bar stock as the
diode targets but were machined into 2-1/4” diameter 3/8” thick disks. The surfaces of
these targets were polished before use with 600 grit SiC paper to remove the hydrocarbon
deposits caused by the EDM process, and then ultrasonically cleaned in acetone and

isopropyl alcohol.

52

3.1.2 T iNiPd Targets

The targets used in TiNiPd experiments were produced by Pure Tech, Inc. and
shipped to MSU Material Science Department by Al Knebel of Delpi Energy and Engine
Management Systems. Each TiNiPd target was a 3” diameter, 1/8” thick powder
metallurgy (P/M) disk that had the following composition as reported by Pure Tech, Inc.

Table 3-1. Nominal TiNiPd P/M Sputtering Target Compositions.

 

 

 

 

 

Composition, atomic %
Target #
Ti Ni Pd
Pdl 53.5 21.5 25.0
Pd2 53.5 20.25 26.25
Pd3 53.5 19.5 27.0

 

 

 

 

 

 

The three TiNiPd target composition were chosen to provide SMA sputtered thin ﬁhns
that would have 50 at. % Ti and 27 to 29 at. % Pd that would produce an austenite ﬁnish
temperature between 200—225°C, that was determined from literature results shown in
Figure 3-1 and 3-2. The additional 3.5 at. % Ti-enrichment was chosen to compensate for

the expected Ti loss during sputtering observed in previous results [Grummon, 2000].

3.1.3 T iNiPd Target Chemical and Microstructural Analysis
Energy dispersive spectroscopy (EDS) was performed on the targets to gather

information on composition to conﬁrm the reported values by Pure Tech. However, due

53

to the extreme variability of the measurements, the results were inconclusive. This may
have been a result of inappropriate standards, which were solid thin ﬁlms rather than P/M
materials similar in nature to the target. ZAF corrections were accomplished using
software that accompanied the LINK EDS system.

Optical microscopy and metallography were performed to gain information about
the PM TiNiPd target microstructure. Target Pdl was mohanically polished for
metallographic observation with a sequence of silicon carbide papers and crystalline
diamond polishing compounds to 0.25]; ﬁnish, and then ultrasonically cleaned in acetone
and rinsed with isopropyl alcohol. To reveal the microstructure, a modiﬁed aqua regia
etchantl was used that contained 2ml HCl and 101111 HNOt An optical photomicrograph
was taken at 50X from target Pdl and is shown in Figure 3-3. The nricrostructure is
coarse and inhomogeneous with large features on the order of 200 pm At a
magniﬁcation of 500x, void porosity and reaction zones were visible, as well as zones of
apparent unreacted pure metals, as seen in Figure 3-4. The observation of zones that
contained pure metals coincided with slightly magnetic nature of the targets, since
elemental Ni is magnetic.

Figure 3-5 displays a secondary electron image2 of target Pdl at 100X. The
microstructure consisted of regions of pure Ti, Ni, and Pd, conrplex intermellatic regions,
and numerous voids. When higher magniﬁcations were used, like the SEM

photomicrograph in Figure 3-6 taken at 1200X, the nricrostructure revealed complex

 

' The modiﬁed aqua regia is general etchant for nickel and palladium based superalloys
and does not etch titanium.

2 SEM images were acquire at an accelerating voltage 25KeV, a beam emission current of
140uA, a #3 objective aperture, and working distance of 8-6mm. Pdl target was
prepared using the procedure for metallo graphic observation

54

reaction zones that may have resulted ﬁom partial melting during the manufacturing of the
target. The inhomogeneous distribution was also apparent from X—ray EDS maps, which
is displayed in Figure 3-7 at 1000X that show regions of elemental Ni, Ti, and Pd along
with melted/reaction regions of combination of these elements.

A stereographic analysis was performed to determine the approximate void fraction in
target Pdl using the grid pattern shown in Figure 3-8. The volumetric void ﬁaction was
determined using V=Q/N, where Q is the number of grid points intersecting voids, and N
is the total number of grid points (see Figure 3-8). The void volume ﬁaction calculation
was performed 10 times with different grid sizes, and the results were averaged. The
volume ﬁ'action of void porosity was found to be z7.1%. The target was therefore
concluded not to be hot isostatically pressed (HIPed), but was rather uniaxially hot

pressed or simply sintered.

3.1.4 P/M Target Performance

The three TiNiPd powder-metallurgy (P/M) targets used in this study performed
well in the 3” Torus 3M guns, allowing reliable plasma ignition at pressures below 2.5
mTorr. Early in the work, copper backing plates were soldered to the targets to increase
their thickness to 1/ ” in order to achieve the proper ground—shield gap in an older model

of the 3” Torus-3M gun3. In one case, target heating at 250-watt power levels caused

 

3 The Torus 3M design uses an enhanced magnet structure speciﬁcally designed for
sputtering magnetic materials. Two guns were used in the study: One, built in 1990
(designated as the ‘old’ gun in Table 1) had apparently been damaged by overheating and
was later replaced with a new unit. The old gun had a substantially weaker magnetic ﬁeld
but was still capable of producing high-quality ﬁlms.

55

melting of the solder4 holding the plate. For subsequent tests using a newly purchased
gun, these plates were removed.

High void porosity was found during metallo graphic analysis, as discussed
previously. This did not cause obvious problems, such as large virtual leaks from the
voids, nor did the slight magnetic character of the targets. The latter factor would,
however, need to be taken into account for thicker targets or for use in standard
rmgnetron guns not equipped with enhanced magnet structures since it can reduce the
deposition rate. The guns used in these studies did use an enhanced nugnet structure
speciﬁcally intended for magnetic materials, which gave good plasma performance at the
expense of less eﬂicient target utilization.

As mentioned previously, sorrre question remained at the end of the study with
respect to the actual target compositions, or the possrhility that compositional uniformity
had been compromised by uneven distrlhution of the powder components during
manufacture. Fimlly, it should also be mentioned that no attempt was made here to
determine the oxygen content of the as-supplied targets. Though none of the results
implied that ﬁlm oxidation had occurred, and oxygen contamination of TiN i alloys is

notoriously difficult to identify5 and cannot be completely excluded.

3 . I . 5 Substrates
Binary TiNi thin ﬁlms were deposited on glass slides that were 3” X 1” X 1/8”

thick, which were used in composition measurements as a function of cathode wear and

 

:The solder has a melting point of around 160 °C.
TizNiO oxides are isostructrual with TizNi, therefore making distinctions between the
phases is almost impossible.

56

lateral distance from the cathode. The slides were adhered to the wafer platen with
vacuum grade copper tape after cleaning with 15% HF solution, and rinsing with isopropyl
alcohol, respectively. 2.5” square, 1/4" thick ﬂint glass plates were used as substrates for
the triode experiments, and were cleaned in the same nranner as the glass slides. One
comer of the glass plates was cut to use as a locater for the position of the plate with
respect to the thermionic ﬁlament of the triode gun. 5” wafers of (100) p-type silicon
wafers were used for TiNiPd thin ﬁlm experiments. In addition, 5” silicon wafers with

100 nm oxide layers were used for TNiPd thin ﬁlms deposited and annealed at high
temperatures. The silicon oxide layer acted as a diﬂirsion barrier to stop the interdiffusion
of Si in TiNiPd at high temperatures, which was known to cause the ﬁlms to become

brittle and have no apparent SME quality.

3. 2 Sputtering Deposition Process
3. 2. I Binary T iNi Magnetron Diode Deposition Setup

Films were sputtered in a 28” Kurt J. Lesker (KJL) Supersystem-II apparatus
equipped with CT 8 cyropump capable of ultimate pressures below 1X10'8 Torr (see
Figure 3-9). The chamber was internally heated with quartz lamps to temperatures greater
than 150°C for 24 to 48 hours to outgas residual water vapor and improve the vacuum.
All ﬁlms were deposited with base pressure better than 5x10’7 Torr. A 24.5” diameter
aluminum octagon shaped substrate holder was constructed for this experiment, which had
eight 2.5” square hole substrate positions (see Figure 3-10). Three glass slides were
centered over the 2.5” square hole and adhered to substrate holder with vacuum grade

copper tape. A KJL Torus 3-M magnetron sputter source (see Figure 3-11) was ﬁtted

57

with a 3” diameter, 1/4” thick TinNin target. Eight thin ﬁlms‘5 approxirmtely 12.3p.m
thick were sputtered sequentially onto ﬂint glass microscope slide substrates at ambient
chamber temperature using a sputtering power of 250 watts, an argon working gas
pressure of 4.5 m Torr, and a substrate-target distance of 7 8mm. Each deposition lasted
for 3 hours, and the eight depositions (including a 30 minute cathode bum-in period) was
predicted to have consumd approximately 75% of the total cathode life based on the

estimated target thickness at the end of the deposition series.

3. 2.2 Triode Deposition Setup

The triode experiments were conducted on a Sirnard Type 383 source and used a
binary 2-1/4” diameter, 3/8” thick TiaoNisr target. The 2.5” square glass plates were
placed over nine 2” holes, as seen in Figure 3-12. The cut end of the glass plate
corresponded to the position relative to the thermionic ﬁlament. A system base pressure
of 4x10'8 was obtained before deposition. The thin ﬁlms were deposited at power 440
watts7 with a thermionic ﬁlament current of 30A, an argon working gas pressure of
2.8mTorr, and working distance of 131 mm. This setup achieved a sputtering rate 0.8
nm/s with 2 hour sputtering time that produced nine ﬁlms with an approximate thickness
of6.111m The target showed erosion proﬁle known as “saddling”, which is convex shaped
erosion formed by path of energetic electrons emitted by the thermionic ﬁlament across

the target. This proﬁle can be avoided by rotating the target, which allows for better target

 

6 One thin ﬁhn deposition run deposited a thin ﬁlm on three glass slides, and there were
eight available positions on the substrate holder, which gives a total of 24 glass slides with
deposited thin ﬁlms.
7 440 watts sputtering corresponded to a sputtering voltage of 550Vand a current of 0.8A.

58

utilization. However, the target was not rotated in these experiments to observe the

worse possible effects of the target proﬁle on the compositional shift of the ﬁlms.

3. 2.3 T iNiPd Magnetron Sputtered Thin Film Deposition

Films were sputtered in KJL 28” chamber shown in Figure 3-9 with KJL Torus 3M
magnetron sputtering source in a working gas of ultra-high purity argon. All the thin ﬁlms
were deposited at 250 watts on the constant power regulation setting of the Advanced
Energy MDXI K sputtering power supply, which is capable of delivering 1kW at 1000V.
A schematic of the thin ﬁhn deposition chamber in Figure 3-13 shows the relative
positions of pumps, sputter gun, fumace, and wafer platen. The typical setup for ﬁlm
deposition was a “sputter down” conﬁguration where the gtm was above substrates and
the wafers were heated on the backside. The chamber, wafer furnace, and wafers were
outgassed at 150, 600, and 575 °C, respectively, for 24-48 hours before sputtering to
improve the vacuum A few hours before each deposition runs a titanium sublimation
pump, specially built for these experiments, was used to getter reactive gases, such as 02
and H20, by evaporating Ti8 unto the water-cooled walls of the sublimation pump.
Typical oxygen partial pressures were below the measurement sensitivity of the Ametek
Dycor Quadrupole residual gas analyzer (P02<10'lo Torr), and water vapor partial
pressures were below 1 x 10’8 Torr. All the ﬁlms were deposited with initial base pressure
better 5x10'7 Torr and most often better than 8 x 10’8 Torr.

Wafers were heated using a US Inc. four-inch non-contact inconel substrate heater

that radiatively transfers heat to the backside of the wafer. The wafer temperature was

59

measured with a K-type l/16” stainless steel sheathed grounded-tip thermocouple in
contact with center of the wafer on the backside (see Figure 3-13). A 304 stainless steel
wafer platen was specially constructed for this research and had eight positions that
centered the wafers under the sputter gun (see Figure 3-14). The wafers were held in
place by 1-2 oval shape clips that were bolted to wafer platen.

During deposition, thermal equilibrium with the sputtering process ﬂux energy was
reached approximately 20 minutes into the deposition run. When no 11ch power was
applied, the wafer backside temperature typically equih'brated about 175°C. Working
distances of 75mm and 67mm were used, and the working pressure, which was held at 2.5
or 4.0 mTorr for most deposition runs, was carefully controlled with a micrometer ﬂow
valve and measured with an MKS capacitance manometer. Extreme care was taken to
avoid atmospheric contamination of the working gas lines by careful purging of the lines
before use and maintaining them at a constant 60-psi positive pressure.

The choice of deposition temperature was made on the basis of the measured
crystallization temperature of the amorphous TiNiPd ﬁlm, and deposition temperatures
were explored above and below this crystallization temperature. A TA Systems 2250
DSC apparatus was used to measure the crystallization temperature of a ﬁlm sputter from
target Pdl at room temperature. The unit was calibrated using a 99.9% pure Zn sample a
heating rate of 5°C/min, and TA Systems DSC commercial software. Two DSC samples
from specimen #0224001 were made from ﬁhn delaminated ﬁ'om the Si substrate with

sample weights of 5.5 and 4.4 mg, respectively. These samples were scanned at 5°C/min

 

8 Titanium was evaporated from 85%Ti-15%Mo ﬁlament with 40A/3V sublinmtion power
supply for 2 minutes at 15-minute intervals.

60

rate under a N2 shield gas, and the crystallization temperature was found to be 480 °C, as
shown in Figure 3-15.

A master table of deposition parameters and thin ﬁlm characterization results for
the TiNiPd ﬁlms used in the composition study is given in Table 3-2. The designation old
and new sputter gun used refers to identical KJL Torus 3M gun designs, however, tests on
the older sputtering gun revealed the permanent magnetic assembly had been damaged and

was replaced with new gun that had a stronger9 magnetic ﬁeld.

3. 3 Qualitative Assessment of Film Appearance and Performance

Qualitative assessment of room-temperature ductility, adhesion, and shape memory
were performed on all samples listed in Table 3-2‘0. Ductility was assessed by deforming
a mechanically peeled sample by manual crumpling. If the ﬁlm ﬁ'actured it was considered
to be brittle. Shape recovery was assessed for ductile ﬁlms by observation of deformed
ﬁlms during indirect heating". The adhesion test was based on the ease that the ﬁlm could
be delaminated from the wafer. Released ﬁlms were often observed to curl extensively

inward due to through-thickness stress variations.

 

9 The new KJL Torus 3M gun had magnetic ﬁeld strength of 1650 Gauss 10mm above the
cathode cooling plate, while the old gun had 423 Gauss magnetic ﬁeld strength.

'0 G,F,P symbols in Table 3-2 corresponds to a good, fair, and poor qualitative rating.

1' Films were heated in an aluminum pan by butane lighter.

61

Table 3-2. Deposition Parameters and Properties

Specimen #

100010£0#
1000z0:0#
loooeosor
10006090#
ZroooOI£0#-' _.

1000172201?
IOOOSZZO#
'8 I SO'NV" 10002090111

Substrate

Working Gas Pressure

Base Pressure (1x1

at

Crystallization

Test

 

62

Table 3-2 (cont’d).

Specimen #

rooozzcon
roooszvos
10009190#
'5090'NV'10009-190#
proooszsos
zoooozsoa
seahorses .
‘ harnesses ,

Substrate

Power

Working Gas Pressure

Base Pressure (1x1

at

Crystallization

Resistivity Test

 

63

Table 3-2 (cont’d).

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Performed

 

 

 

 

 

 

 

4: =11: at: :11: a:
a e e e e
or 04 u» at or . j .
:2 e e a e , ,,
c e c
a e 2 § § 3 3 - §
, - g 4 u at u at g u g , N e e
Specimen # U- . U! r U: r = . Us . <1 or a
°> °> °> °> °§ S_—S S
2 2 2 2 c e a
'c 'c 'e 'e :5 u "‘ "‘
a a a at q , ,
e N N N e
S" ‘r‘ 3" 1' '3
Gun Used new new new new new new new new
Target Pd% 27 27 27 27 27 - 25 726.125 26.25
81/ Si/ Si/ Si/ Si/ Si/
Substrate SiOz SiOz SiOz 3102 3102 51/ 5" $102 81/ 5"
5" 5" 5" 5" 5" 5"
SP““°"°3 PW" 250 250 250 250 250 250‘ 250 250
(watts)
Sputtering Volta e (V) 294 295 295 294 294 328 328 331
D°P°s“‘°‘(’,TC°)mP°m“’° 548 518 518 548 548 425 ,. 425 [1425'
w°rkmg Gas Pressure 4.0 4.0 4.0 4.0 4.0 4.0 4.0 4.0
(mTorr)
Working Distance (mm) 67 67 67 67 67 67 67 67
Base Pressure(lx10 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8
‘ Torr) , ‘
i Sputter Time (Min) 45 45 45 45 45 60 i 60"" 60 7
Thickness (pm) 3.53 3.45 3.45 3.53 3.53 4.64 4.34 4.46
Post Anneal . i l
‘ Tern erature (°C) 575 748 655 600 550 - - a“.
Post Anneal Time(Min) 75 60 60 60 60 - - -
Base Pressure at Heat .
Treatment (1x10" Torr 30 50 55 20 0'45 ' ' '
C’yStamza‘m 481 481 481 481 481 479 482 482
Temperature (°C)
‘ DSC Test Performed no no no no no no no no
Resistivity Test
no no no yes yes yes yes yes

 

 

 

64

3. 4 Resistivity Measurements

To gather accurate transformation temperatures in a non-destructive test, an
electrical resistivity device was constructed (see Figure 3-16). The apparatus used a
Jandel Macor four-point probe with a sprirrg-load tungsten pins spaced 1 mm apart. The
two outer probes were used to apply a constant current of 100mA, and the inner probes
measured the voltage drop. The specimen was heated and cooled at constant 5°C/min by
OMEGA PID temperature controller that measured temperature with a ﬁne thermocouple
wire near the contacting probes. The inner probe voltage drop and thermocouple voltage

was output to a chart recorder, which plotted the voltage drop verses temperature.

3.5 Differential Scanning Calorimetry (DSC)

A TA Systems model 2250 DSC system using LN2 chiller was used to gather
information about the transformation temperatures of TiNiPd thin ﬁlms. The DSC was
cahhrated using a 21 mg pure indium sample at a heating rate of 10 °C/min. The DSC
samples were prepared from ﬁlm delaminated from the substrates, which were weighed on
an electronic scale. The delaminated ﬁlm was placed in specially designed TA Systems
aluminum DSC pans, which were then sealed using a commercially supplied pan crirnper.
The DSC was programmed to equilibrate at 75°C, heat to 220°C at a rate 10°C/min, and
cool to 75°C at a rate 10°C/min. The DSC curves were graphed and analyzed using the

Universal Analer program supplied by TA Systems.

65

3. 6 Thin Film Composition Measurements
3. 6. I Binary T iNi Composition Shifts Measurements due to Cathode Wear

Film compositions were measured by energy dispersive X-ray spectroscopy (EDS)
on a Hitachi S-2500C scanning electron microscope (SEM) that was equipped with Link
EDS X-ray spectrometer using the following setup for all measurements: objective
aperture setting #3; detector-specimen angle 45°; specimen working distance of 15mm;
accelerating voltage 25 keV; LaB, ﬁlament current 90uA; acquisition setup requiring a
total spectral integral count of 10,000 Ti-Kor counts under a Ti Kat energy window
between 4.3474668 keV; Ni K01 energy window between 7287-7668 keV; a dead time
about 15% and count rate 1 kcps”; and beam rastered over a scan area of 3.3x4.1 mm
(30x magniﬁcation setting). Since the ﬁlms were thicker than 6pm, it was assured that no
characteristic X-rays were generated ﬁ'om the substrates. The relative horizontal distance
from the center of the cathode on the glass substrate was measured by the SEM stage
micrometers and aligned by fudicial marks on the stage. Sixty X-ray spectra were taken
for each specimen at the measured center-cathode position, while 30 spectra were taken at
distances away from the cathode center for each specimen. The composition was
determined by a ZAF correction of integral count ratio using a standard with a

composition of TiaoNisr.

 

'2 The count rate was constant for each specimen by an adjustment of the condenser lens
setting.

66

3. 6.2 TiNiPd Thin Film Composition Measurements

The composition study was performed on a Hitachi S-2500 C SEM that was
equipped with Link EDS X-ray spectrometer operated a 25 keV with 140 11A emission
current. The collection parameter were as follows: 50x magniﬁcation, 15 mm working
distance, detector-specirnen angle 45°, 2 kcps count rate, 30% deadtirne, and 60 second
preset livetime. Thin Film standards were used along with a Link ZAF correction
software. The standards used were taken from sample #0302000], which was assumed to
have a composition TisoNizoPd3o based on observed transformation temperatures and TEM
microstructure analysis (see the section below). Spectra from the standard were collected
ﬁrst to develop the necessary ZAF correction factors and proﬁles”, and then spectra from
the unknown TiNiPd thin ﬁlms were collected. Thirty spectra were collected for each

unknown, which were tabulated and statistically analyzed.

3. 6.2. I TiNiPd Thin Film Composition Standard Preparation

Transmission electron microscopy (TEM) examination of a ﬁlm that has been heat-
treated to achieve thermodynamic equilibrium can be useful since the amount and type of
precipitates formed gives unambiguous information on deviation from equiatomic
stoichiometry. Typically in bimry TiN i alloys, a long heat treatment of 24 hours at
500°C” will drive excess Ni out of solution and form NiaTi3 precipitates. Long hold times
cause temperature cause the formation of large incoherent precipitate particles. If the

alloy is titanium-rich, TizNi will be commonly observed, which are spherical in shape, and

 

‘3 The ZAF correction factors and proﬁles were standardized every 15-20 minutes during
the unknown spectra collection process.

67

often form at grain boundaries and other high-energy sites. Since palladium substitutes for
nickel in the ternary alloys, the heat treatment processes and trends should be similar to
binary alloys, but little is known on the precipitate structures that form in ternary
compositions. An equilibration heat-treatment of a ﬁee standing TiNiPd thin ﬁlm was
performed on sputtered sample #0302000]. A piece of this specimen was sandwiched
between two quartz plates and vacuum annealed at 505°C for 24 hours to produce the
specimen designated as 03020001-An0518-500 (See Table 3-2).

Figure 3-17 shows a schematic of sample preparation conﬁgtuation. The specimen
was phced on a wafer substrate heater in the sputter chamber and held in place by the
slight tension of a thermocouple wire. The temperature of the ﬁlm was calculated by
averaging the temperature measured on the surface of the substrate heater and backside of
the top quartz plate. The vacuum chamber was pumped down to a base pressure 6x10"3
Torr with a water partial pressure of 2x1 0'9 Torr and oxygen partial pressure of 4x1 0'10
Torr. The temperature was slowly ramped to 505°C for 1 hour then held at that constant
temperature for 24 hours at which time the substrate heater was turned off to let the ﬁlm
cool slowly to room temperature. Subsequent TEM observation showed the existence of
martensite a room temperature with structure shown in Figure 3-18. The foils could be
rmde to transform in- situ to the austenite phase by using TEM heating stage, yielding
microstructures as shown in Figure 3-19. A small number of precipitates with sizes under

20 nanometers were visrble, mainly at austenite grain boundaries. These precipitates were

 

'3. {his annealing temperature is well below the solutionizing temperature for excess
me e

68

too small to adequately index with SAED diﬂ‘raction methods, but were assunwd to be
Ti2(Ni+Pd)‘5 particles.

The transformation temperatures of the post-annealed specimen were
approximately 50°C higher than in the as-deposited condition, which according to Figure
3-1 corresponds to a minimum of 30 at. % Pd“. The composition of the calibration
specimen was therefore estimated to be approxirmtely 50.2 at. % Ti, 30 at. % Pd, and

19.8 at. % Ni.

3. 7 T EM Microscopy
3. 7.] TEM Sample Preparation

The TEM foil preparation was began by adhering a 3mm molybdenum support ring
that had a 1 mm hole to the thin ﬁlms using M-Bond 610 adhesive. This epoxy resin was
post cured at 175° C for 12 hours. After curing the resin, excess thin ﬁlm was cut away
using scissors, which was done very careﬁllly so cutting process would not damaged the
adhered ﬁlm and bend the support ring causing unwanted dislocations and bend ﬁ'inges in
observed TEM photomicrographs. These 3 mm disks were then electropolished on a
Struers Tenupol-3 twin jet electroplisher using 25% concentrated Sulﬁrric acid in
Methanol. The best foils that were observed used a current density of 0.7 A/cmz, 8V, an
electrolyte temperature of—20 °C, and ﬂow rate setting of 3. The foils were rinsed in

water for 1 min and then rinsed in methanol.

 

'5 This assumption was based on their size, shape, and observed location.
'6 The observed martensite start temperature was 223°C, which about 30 at.% Pd in

Figure 3-2.
69

3. 7. 2 T EM Microscope Setup
The TEM samples were observed a using double-tilt holder, and TEM was aligned

using #2 condenser aperture and a spot size of 5 pm. The objective and condenser
astigmatism was corrected as needed during TEM observation. The #5 SAD aperture was
inserted at in the thin area to observe the crystallinity and phase of the material, and
diffractions patterns were taken at different camera lengths and an exposure time of 15
seconds. The #6 objective aperture was inserted and centered on the transmitted beam to
form bright ﬁeld images at various magniﬁcations.

To image in-situ phase transformation and the austenitic structure of TiNiPd, a
double tilt stage equipped with electro-resistive fumace was employed. The temperature
of this furnace and TEM foil was measured by Pt-PtRh thermocouple, whose output was
displayed on an mV meter and was converted to temperature in Celsius using the
manufacturer’s calibration curve. The ﬁlrnace was outgassed for 2-hours at 600°C to
remove contamination and absorbed gases before use. During in-situ transformation
observation, the ﬁrmace was ramped slowly from room temperature to 300°C at rate

10°/rnin. After the furnace temperature reached 300°C, the foil was allowed to eqruhhrate

and fully transform to austenite before taking photomicro graphs.

3.8 Thin Film Thickness Measurements
The thickness of the as-deposited ﬁlms was used to calculate the deposition rate of

the sputtering process. To obtain the thickness proﬁle, a Dektak H proﬁlometer was used

to scan 1 mm wide. 5-10 thickness measurements were rmde for each sample and

averaged.

70

CHAPTER 4 RESULTS AND DISCUSSION

The sputter deposition of alloys is complex and complicated by the large number
of physical and chemical processes involved. Previous discussions on the mechanisms of
sputtering as well some process parameter effects on ﬁlm composition was presented in
Chapter 2.

The ﬁrst part of this chapter will present data gathered from compositional studies
of sputtered TiN i and TiNiPd ﬁlms. This includes a discussion on magnetron sputtered
TiN i ﬁlm composition as compared to the triode results, and in addition experimental
data on composition and the transformation temperatures of TiNiPd ﬁlms will be
presented.

The second part of chapter will discuss the mechanisms possibly responsible for
the ﬁlm composition shifts, which include the following: partial yields, angular
distributions of the sputter particles, thermalization, mechanisms that effect composition
of the growing ﬁlm such as sticking coefficients and resputtering, and surface diffusion in
the target.

Deﬁnitive conclusion will not be made on the actual mechanisms that affect the
ﬁlm compositions, however, experimental results will show that ﬁlm compositions do
vary with the deposition process making composition control problematic. Suggestions
for further experiments will be made in Chapter 5 that may lead to a better understanding

of these mechanisms and a solution to composition control problem.

71

4.1 Sputtered Binary T iNi Composition Experiments

Films were sputtered in a 28” Kurt J. Lesker (KJL) Supersystem-II apparatus
equipped with a KJL Torus 3-M magnetron sputter source, which was ﬁtted with binary
3” diameter, 1/4” thick Ti49Ni51 target. Eight thin ﬁlms approximately 12.3 pm thick
were sputtered sequentially onto ﬂint glass slide substrates. Each ﬁlm was sputtered for
3 hours at ambient chamber temperature using a sputtering power of 250 watts, an initial
base pressure better than 5x10'7 Torr, an argon working gas pressure of 4.5 mTorr, and a
substrate-target distance of 7 8mm. Figure 4-1 shows the ZAF corrected ﬁlm composition
measured at the center of the each ﬁlm as a function of elapsed time on the binary TiN i
cathode at the end of each deposition. The results show that the ﬁlm composition
changes continuously with time, beginning in the Ti-enriched range (relative to the
sputter cathode composition), and ending in a Ni-enrichment of the ﬁlm with an overall
change in Ni composition around 2 %. The ﬁlm composition matches that of the sputter
cathode material only for a brief period after 12 hours of sputtering at 250 watts. A quasi
steady-state condition is apparently approached at 14 hours of sputtering at a composition
that is Ni-rich relative to the target composition.

Figure 4-2 shows a surface plot of both the lateral and the temporal variation in
ﬁlm composition. The latter variation is measured as a function of the radial distance
away ﬁ'om the projection of the diode cathode center. If the scatter in composition data
with lateral distance was assumed to be related to random errors and poor counting
statistics, the surface can be smoothed to give the surface plot shown in Figure 4-3. The
variation in the Ni content with lateral distance seems to diminish as the sputtering time

increases. This is also apparent in the 2-D plots shown in Figures 4-4 and 4-5. The linear

72

fit through all the data points in Figure 4-5 would indicate that the Ni content does not
change signiﬁcantly with lateral distance from the cathode center at long sputter times.
However at initial sputtering of the target, variation in the composition is evident, where
a Ti-enrichment is apparent toward the edge of the specimen and above the weartrack of
the target. This data may be compared with results of a similar experiment using a triode-
type sputter gun, which is discussed below.

Nine 6 pm thick thin ﬁhns were deposited ﬁom Sirnard 383 triode source using a
binary 2-1/4” diameter, 3/8” thick Ti.,,Ni51 target at power of440 watts' with a
thermionic ﬁlament current of 30A, an argon working gas pressure of 2.8mTorr, and
working distance of 131 mm. These ﬁlms were used to observe the composition change
verses sputtering time behavior of the triode source, which is shown in Figure 4—6.
Although even greater overall Ti-depletion is apparent, it can be seen that the output of
the triode source is nominally constant with time, showing approximately 0.25% shift
ﬁ'om beginning to end of the run, which is an order of magnitude better than the
performance of the diode source. Figure 4—7 additionally shows slightly worse lateral in-
plane composition uniformity for the triode output as compared to the diode performance.
This Ni-content of the ﬁlms was found to increase by about 0.75 at % at distances of 20

mm for the cathode center.

 

' 440 watts sputtering corresponded to a sputtering voltage of 550Vand a current of 0.8A.

73

4. 2 Studies of Films Sputtered from P/M Composite TiNiPd Target
4. 2. 1 Composition Studies of Sputtered T iNiPd Thin Films

Table 4-1 gives a master list of the ﬁlms fabricated, together with their
clulracteristics and some of their deposition parameters (their depositions parameters are
shown in detail in Table 3-2). Targets used in this study were made with 53.5 at. % Ti, or
3.5 at. % in excess of equiatorrric stoichiometry. The results show that the titanium
content of as-deposited ﬁlms ranged ﬁ‘om 49.4 at% to 52.9 at%. The Ti depletion during
sputtering was not as high as was anticipated. The mean Ti content of 25 ﬁlms was 51.4
at%, for an average overall Ti-loss during sputtering of 2. 1%. The standard deviation in
the measurements of the Ti-content for the same 25 ﬁlms was 1.2 at%. The scatter2 in
the EDS data was minimal as compared to overall variation in Ti-content, which is
indicative of substantial process parameter and target erosion effects on ﬁhn composition,

as will be discussed further below.

4. 2. 2. I The Effect of Target Erosion on Composition in TiNiPd Thin Films

Figure 4-8 shows conrposition of ﬁlms made from target PDl as a function of the
time for a cathode that has been sputtered at 250 watts. In this case, the titanium ‘output’
is initially depressed by over 4% relative to the nominal 53.5 at% in the as-fabricated
cathode, but rises over the ﬁrst 5 hours of cathode wear, ﬁnally stabilizing at near 50 at%.
The trend toward overall Ti-depletion in the sputtered ﬁlms is therefore similar to that for
both diode and triode sputtering from the binary melt-solidiﬁed target, but the trend

toward increased Ti output over time is the opposite case observed for the binary material

 

2 The standard deviation in the composition measurement of single specimen ranged
between 0.25 and 0.5 at. %.

74

sputtered in a similar diode gun. Palladium content of the ﬁlms is enriched with respect
to the 25 at% cathode composition, and the output of this species also rises over the ﬁrst
8 hours of cathode life. The nickel output is also initially enriched, but over time drops to
nearly the initial as-fabricated cathode composition. This is also different ﬁom the trend
observed in the binary alloy experiments using the diode source. Thus, both Ni and Pd
were initially enriched in the ﬁlms at the expense of Ti. The latter species steadily
increased whereas the Ni level dropped during the ﬁrst 8-10 hours of sputtering. For all
three species, and apparent steady-state output condition was reached, but not before the

cathode was approximately half consumed.

75

an!» at? .225” 35.. 63830: megana and 3.8913

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

o v
Henna . . 385: new 5.238 m m. n. m S
macaw—:2. P. nos—SUE...— HaBu. 5.82:6 9:38 m. M" m. . W 3“ zen up“ a?“ 3...» 3m a?” >9
LA”? gv ADA. 0v 00 Ass—.Oﬁﬂv Ann—av ” m .0... . 3 KO in. in VIC v—An v10 *0 inc!
8 m u
3 2” my:
tonnaoos Mm a5.— Bb Nch am PM aw Pu no m. w .. t .. r t r t t ..
nonnmocg Nu bob NPm NPm 3m Pm am 955. «am who who 0%. 3» GM :0 5e 3m 0.» :q a»
#8925— Nm nob web we mum PM um Mun «mm m. 0 O 3— 3o H3 Sq Ga :5 73 Ga
acmonocg Nu us; 3.0 we muo PM um 93 no m. 0 0 3c 3N 3m Em 3a am am :_

 

 

 

76

a.»

mo; S

we

Her—m 5L €2.35.

8931119an

gm Gm

 

77

Hes—o ar— A8513.

 

 

 

 

 

 

n m
. l. V 0
Hanna. . . 53852. new mecca: a n. n S
meson—=3 w.— neNHeW—We: #2:... W533 9.558 W. W m. W N ManxAW—MXWXMMXMWXMWEWXMW
jun. .5. o no Aide—.5 GEE .m m m. .M. H 1
m
.3 7: ma
#8383 3 avg 5h 3.» mam Na .3 no «am w w w .. .. - r r .. - -
$8383 3 3 :3 3% who Nm .3 Mum <8 and w m. 73 :q :w Ge - - .. ..
ainmoog 3 web Sb 3.» moo PM 3 $55.03. 0 v :28 :w as SN 5m . - - ..
“53253 3 mobgpurq Ame PM 3 —m.3 so 0 v - . . .. r - r - ..

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

78

nos—~352—

ﬂ 7: mi

382
3 3 ﬂ.

qmo

Hus—o A; Ana—1.3.

um sump!“

Gm H

 

79

macaw-3:

035.352.
Ann. o\ov

Hue—a A; A8535.

w
I.
m
a

 

80

4. 2. 2.2 The Eﬂect of Deposition Temperature on Composition

Figure 4-9 shows titanium content for as-deposited (not annealed) ﬁlms as a
function of deposition temperature. Although hot-substrate depositions showed slightly
elevated Ti levels as compared to the few runs conducted at ambient temperature
conditions (solid trendline in Figure 4-9), the trend for elevated-temperature deposition
shows that Ti to decreases slightly with increasing substrate temperature (dashed
trendline). The effects of deposition temperature on the atomic ﬁ'action of all elemental
species from Ti 53.5%/ Ni 19.5 %/ Pd 25% Target is shown in Figure 4-10. Overall
trends in composition (solid trendline) show that Ti and Pd content increase with
temperature, while the Ni content decreases. If the low temperature result is excluded,
the trend is the opposite; Ni content increases with temperature while Ti and Pd decrease
(dashed trendline). The latter trend is probably more valid, since composition

measurement of the low temperature non-crystalline ﬁlms may be in error3.

4. 2. 2.3 The Eﬂect of Working Gas Pressure on Composition

Deposition pressure exerted a more pronounced effect on composition, as shown
in Figure 4-11, which plots the effect of the PD (pressure x working distance) product on
thin ﬁlm composition. Figure 4-12 is a plot of composition verse working pressure alone.
In Figure 4-12, it can be seen that as pressure was increased, the titanium and nickel
levels increased while the palladium concentration dropped by a complementary amount.
This suggests that the primary effect of increasing pressure is to diminish to the

concentration of the heavier species in the deposition ﬂux.

 

3 Slight errors in composition can result was analyzing amphorous materials with
crystalline standards in ZAF correction, since electron scattering and channeling are
different between amorphous and crystalline materials.

81

4. 2.2 Transformation Temperatures of Sputtered T iNiPd Thin Films

Transformation temperatures were determined for selected ﬁlms by differential
scanning calorimetry (DSC) and by electrical resistivity measurements. Results for as-
deposited ﬁlms are shown in Tables 4-2 and 4-3 (For DSC and Resistivity results
respectively), and in Figure 4-13. A number of depositions could not be evaluated by
DSC since the high degree of adhesion achieved and the lack of material that could be

removed from the silicon substrate (about 10 mg was needed for each test).

Table 4-2. Transformation Temperature Data for TiNiPd ﬁlms from DSC Curves.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Sample # Ms] Mn A31 An Msz Mn A32 An Ms3 MB M" A'
(°C) (°C) (°C) (°C) (°C) (°C) (°C) (°C) (°C) (°C CC) (°C)
02250001 165 154 172 190 147 115 145 169 - - 161 184
03010001 173 161 176 192 161 151 145 167 124 112 158 188
03020001 137 125 149 161 123 93 101 148 - - 110 133
03220001 149 125 102 144 125 90 - - - - 106 126
05290001 144 100 118 166 - - - - - - 117 142
Table 4-3. Transformation Temperature Data from Electrical Resistivity Curves.
Ms] Mn As! An Ms: Ma A52 An
Simple # (°C) (°C) (°C) (°C) (°C) CC) (°C) (°C)
02250001 172 155 179 189 155 97 117 168
03010001 171 161 187 197 157 140 143 156
03020001 170 162 178 188 156 132 155 171
03030001 167 153 170 185 153 147 160 170
03100001 174 167 162 189 167 110 123 162
03220001 145 117 113 159 n/a n/a n/a n/a
04250001 1 13 84 102 128 n/a n/a n/a n/a
05290001 147 119 167 175 n/a n/a 137 167
05290002 143 112 155 177 112 108 n/a n/a

 

 

 

 

 

 

 

 

 

The DSC and resistivity curves generally exhibited two and sometimes three

 

 

transformation peaks, which indicates that sequential transformations occurred on both
heating and cooling. Sequential transformations involving orthorhombic martensite can

be eliminated as a cause for this behavior, as well as any Monoclinic or R-phase

82

transformations, which are known to be fully suppressed by Pd substitution [Lindquist,
1988]. One possible cause for the multiple transformation peaks is through-thickness
structure gradients (such as ﬁner grain size near the ﬁlm-substrate interface) allowing
lamellar transformation zones to develop in association with differentiated near-surface
and near-substrate microstructures or stress-states. Another possibility, which is
indicated by composition study data, is that composition varies with thickness, and thus
different phases can form through the thickness, such as precipitates. As discussed
earlier in Clmpter 2, precipitates can have an adverse effect by suppressing the martensite
transformation, and differing densities of precipitates in the ﬁlm would explain the
existence of multiple transformations at different temperatures. However, at present the
reasons for the multiple peaks remain unclear.

Transformation temperature data for as-deposited ﬁlms plotted as a function of
palladium content are shown in Figure 4-13, and for reference, are superimposed on the
literature data by Lindquist in Figure 4-14. The observed as-deposited transformation
temperatures fell within the range predicted by previous work but are generally on the
low side, which probably reﬂects the ﬁne grain size of the ﬁlms in comparison to ingot-
metallurgy alloys used in most of the studies in the literature. The annealed thin ﬁlms
transformation temperature coincided with the literature, which indicates that the
palladium content measured by EDS is close to the true value. However, it is unclear
how off-stoichiometric compositions and precipitate formation affects the transformation
temperature, thus further heat-treatment and TEM studies are necessary.

The M8 temperature was observed to increase by approximately 11 degrees per

at% increase in Pd content, which is close to the 13 degree per at% slope from Wayman’s

83

1990 study (see Fig. 3-2). The [Aer] hysteresis varies widely, ﬁom as high as 65 °C to
as little as 22 °C, with low values of hysteresis tending to occur for high transformation
temperatures. The Ms temperature also was noted to vary sharply with Ti-content in the
ﬁlms as shown in Figure 4-15, with the maximum in M, occurring, as expected, at near-
stoichiometric compositions, but dropping by about 12.5 °C per at% Ti. This result is
similar to composition relationship between Ti and Mstemperature in binary bulk
materials that Wang [1965] observed, which had about 15°C per at. % Ti trend (See
Figure 2-4).

The two outliers observed in Figure 4-15 could be related to the sputter deposition
temperature. At low deposition temperatures, a mixture amorphorous and ﬁne grain
structure can be produced that can depresses the Ms temperature. This could be
attributed to the low observed Ms temperature of the outlier that was deposited at 549 °C.
In the case of the high M, temperature outlier, the deposition temperature was 605 °C and
observed M, temperature was higher than the observed trend. This ﬁlm may have
developed a higher degree of lattice ordering that does not sufﬁciently depress the Ms
temperature.

As shown above, deposition temperature can exert a strong effect on
transformation temperatures, as shown in Figures 4-l6a and 4-l6b. In ﬁlms sputtered
form targets Pdl and Pd2, the transformation temperatures increased with increasing
deposition temperature. This may be related to the ultra ﬁne grain size that is possibly
being formed at low deposition temperatures and is suppressing the M3 temperature.
Larger grains are generally formed at high temperature temperatures that do not depress

the Ms temperature, which may account for the higher transformation temperatures.

84

In summary, the transformation temperatures are strong function composition as
well as ﬁlm microstructure. As expected, the transformation temperature increased with
increasing Pd content. The comparisons with literature data show that in general TiNiPd
ﬁlms were close to stoichiometry. It is not clear how off-stoichiometric composition
effect composition, but excess Ti seems to decrease the M3 temperature. Some comment
is appropriate at this point regarding the sputtering mechanisms that are possibly

affecting ﬁlm composition. This mechanism will be discussed in detail below.

4.3 Mechanisms that Aﬂect Film Composition

There are several aspects of the alloy deposition process that have a signiﬁcant
effect on the ﬁlm composition: partial sputtering yields, angular distributions of sputter
particles, mechanisms that affect composition at the ﬁlm surface such as sticking
coefﬁcients and resputtering, and surface difﬁision in the target. These mechanisms will

be discussed in terms of experimentally observed composition trends in proceeding

paragraphs.

4. 3. 1 Partial Sputtering Yields

Sputtering is mechanism of ballistic ejection of atoms from the cathode in which the
accelerated ion ﬁrst penetrates the target surface, and then undergoes a series of atomic
collisions (known as a collisional cascade) which eventually transfer sufﬁcient
momentum to near-surface atoms to cause ejection. The likelihood of the ejection event

is primarily dependent on the surface binding energy of the cathode species, but also

85

depends on its mass and the mass of the projectile. Hence, the probability of ejection
(sputter yield in atoms per incident ion) is species dependent.

Pure Ti has a lower sputter than Ni because its surface binding energy is higher
and its atomic number density is lower. Furthermore, energetic ions bombarding pure Ti
targets will penetrate and transfer most of their energy at greater depths than in Ni,
therefore making it less probable that recoils will have enough energy to over come the
surface binding energy resulting in a lower yield. If pure material sputter yield data are
assumed, Ni would be expected to be preferentially sputtered ﬁ'om the TEN i target since
Ni has more than twice4 the sputter yield of Ti as calculated using the Matsunami
formulation (see Table 2-4). This would cause the ﬁlm to enrich in Ni at earlier stages of
sputtering, which is contrary to the observed Ti rich composition at 3hrs in Figure 4-1.
Thus, we must consider sputtering in TNi alloy.

In TiN i alloys, the surface binding energies, atomic number densities, and
projectile depths should be the same for Ti and Ni atoms, since their bond structures are
similar. In this case, Ti has as lighter mass that is closer to the projectile mass (Ar) than
Ni, thus, more projectile energy is transferred to Ti atoms than Ni, and Ti recoils are
more likely to be ejected from the surface. It is possible that Ti rather than Ni would
have the higher sputter yield, which may be ampliﬁed and exhibit a higher sputtering
yield than the pure material. This could account for the Ti-rich composition at 3 hrs in
Figure 4-1, but contradicts the observed Ni rich composition in the triode ﬁlms.
However, it should be noted that the ﬁlm composition in triode could be affected by
resputtering which in fact favors the idea of preferentially sputtered Ti. This will be

discussed in more detail later. The key factor in increased partial yields is the mass

 

4 Ni had a sputter yield of 0.66 at 300 eV, and Ti had 0.3 sputtering yield.

86

relationship between the projectile and sputter species, thus partial yields of Ni and Ti
can change with projectile type, which is discussed below.

If the above theory for preferential sputtering is true, Ni would preferential
sputtered when Krypton is used as sputtering gas, which is closer in mass to Ni. This
implies tint amount and type of preferentially sputter species is a function of difference
between recoil and projectile mass. A calculation can be made of relative yield in which
U,b and Z, has been held constant, and by only varying the recoil mass in the terms from
Matsunami’s empirical sputter yield formula (Equation 2-3) that are not associated with
the energy loss cross section. This analysis gives the result shown in Figure 4-18 for
Argon projectiles. The magniﬁed portion of the curve given in the inset shows the
isolated mass effect (as it impacts the values of Ed. and a* in Equation 2—3) and indicates
that, other factors being equal, Ti (mass 47.88) should sputter faster than Ni (nmss 58.7)
by a factor of approximately 1.3%. Heavier species should sputter at even lower rates,
and in case of TiNiPd ternary alloy, Pd should not be enriched since it has highest mass
of the constituents, which is not observed. However, the assumption for equal binding
energies for ternary compositions is less obvious than the in the case for the NiTi binary
compound, and the above model may be less valid. Some comment is appropriate at this
point regarding the performance of composite targets (made up of incompletely reacted
elemental powders by hot-pressing).

During the ‘burn-in’ period, composition can be expected to shift sharply as the
effective cathode surface composition shifts to account for the different sputter-yields of
the components. It is not known at present exactly how long this burn-in period should

be expected to last, but the triode experiment indicates that it is not longer than two

87

hours. Also, to a ﬁrst approximation the alloy and PM composite targets should reach
steady state by similar mechanisms, but may have a sufﬁciently longer burn in time as
apparent in Figure 4-8 in which it apparently takes 5 hours before steady state is reached.
The surface of the target is initially depleted in the more rapidly sputtered species, whose
effective sputtering rate then drops as fewer atoms (or particles) of that type are left
exposed at the surface. Eventually, a surface “alter-layer” develops that equilibrates the
partial sputter yield between species, which schematically shown in Figure 2-40. Barring
continuous resupply of this element to the surface by bulk diffusion, steady state (if
attained) must correspond to a condition where the sputter ﬂux precisely matches the
composition of the cathode. In the case of the PM target, surface composition is
determined by the relative distribution of the powder particle types at the surface, but the
steady state argument applies once a depth of a few particle diameters has been eroded.
In fact, diffusion conditions for the achievement of steady state are someth less
stringent for PM materials, and bulk diffusion is not likely to be a signiﬁcant factor in a
target composed of pure materials, since no continuous concentration gradients exist.
However, the development of local sputter topography, such as etch pits, in the case of
powder targets may be at least, in part, responsible for the changes over the cathode life
since erosion of the rapidly sputtering species will tend to cause it to become occluded by
slower-sputtering particles in a way tlmt encourages redeposition onto other particles in
the aggregate, as indicated schematically in Figure 4—19. This would not affect the
steady-state argument to the ﬁrst approximation but could alter the cosine-distribution of
the ﬂux, which tends to narrow the distribution of the more rapidly sputtering element

relative to the other species in the cathode. From the target wear results it seems this

88

surface morphology takes at least 5 hours to develop. After its development, there seems
to be a steady state ﬁlm composition with target wear, but it is different from target.

As alluded to above, even if the ﬂux at the cathode surface matches the bulk
composition, the angular distribution (under- or over-cosine) of the ejected atoms will
still be capable of altering the ﬂux composition, and thus the ﬁlm composition, at any
distance away from the cathode surface. Little is presently known about factors affecting
these angular distributions, but atomic mass is expected to play a role in determining the
over— or under-cosine nature for a given species in an alloy. At this point, a discussion on

composition effects of differential angular distributions is appropriate.

4. 3. 2 Composition Effects Resulting from Diﬂerential Angular Distributions

Difference in angular distribution of sputter particles can affect composition by
enriching the substrate in one species relative to the other at positions deﬁned by a
preferred angle of ejection. In the case of PM targets, the angular distributions of sputter
particles should be similar to pure elements. Wehner et al. found Ni to exhibit an under-
cosine distributions, however, there are no results in the literature regarding the type of
distribution Ti and Pd exhibit. Thus, it is difﬁcult to explain the possible effects that the
angular distributions might have on the PM ﬁlm composition. However, the above
model for PM target erosion would suggest that Pd (the fast-sputter species) has forward
bias distribution and may account for observed enrichment in the ﬁlm.

In the case melt-solidiﬁed alloys, the angular distributions should be different
than pure nmterials. According to Olson’s reﬂective collisional model of mass effects on

angular distribution, the lighter species (Ti) would have preferential emission normal to

89

the target surface and would enrich at the substrate positions above the cathode center.

As a result, Ni should enrich at lateral distances away for the wear track. In the case of
magnetron sputter targets that develop a wsertrack, Ti should be enriched directly above
the wear track, while Ni would be enriched at location above the cathode center. This is
consistent with the results shown in Figure 4-4 at early hours of sputtering. In the case of
triode composition a result, a slight enrichment (relative to lateral positions) of Ti at
center of the ﬁlm is noted (see Figure 4-7) which is consistent with favored normal
ejection angles for Ti.

If this is true, the above analysis would predict a continual enrichment of titanium
in the central region of the growing ﬁlm, and thus it is of little help in explaining the
observed composition shifts. It may, however, bear on this issue of composition shifts
over the life of the cathode in diode sputtering, where cathode depletion is accompanied
by the formation of a pronounced surface ‘wear-track’ topography. The evolution of the
wear track alters the orientation of the local surface normal with respect to the substrate
plane. If one species is sharply overcosine with respect to another, this change in
orientation may cause composition to change as the wear track develops.
Experimentally, this possibility is strongly supported by data (shown in Figure 4-1 and 4-
6) showing that composition shifts quite markedly in the case of the diode sputter gun,
but shifts hardly at all when a triode sputter source is used, which has a very even, ﬂat
cathode erosion proﬁle. However, this model is very inconclusive, since there is no
literature data or theoretical models that support this phenomenon. In addition, the

diﬁ‘erential angular distribution does not support the observed Ti-loss in triode ﬁlm

90

However, thermalization could offer a possible mechanism which might explain the Ti

depletion in the ﬁlms.

4. 3.3 Thermalization

Sputtered cathode species and reﬂected neutral working gas atoms may suffer
collisions with gas atoms during transport to the substrate. Such collisions reduces
energy of both the sputter ﬂux, which leaves the cathode surface with as much as 10 eV
(equivalent to a temperature of ~1 00,000 Kelvins), and the reﬂected neutral atoms that
has an energy an order of magnitude higher than sputter atoms. Thermalization has two
key implications: First, thermalization of the cathode alloy species disperses the ﬂux in

a way that is mass (and therefore species) dependent, and secondly, thermalization
reduces the energy of the reﬂected neutrals and diminishes any effects due to
resputtering of the growing ﬁlm. The former mechanism may account for titanium loss
in the ﬁlms, as discussed later.

The fraction of sputter particles that reach the substrate is related to thermalization
distance, the mass of sputtered atom, the rmss of the sputtering gas atoms, and the energy
of the ejected the sputter particle. Turner et al. calculated the ﬁaction of sputtered atoms
which reach the substrate after suffering scattering events during transport. He
considered both the mass of the species and its binding energy, which was introduced to
account for the initial energy of the sputtered species in the sense that higher binding
energy would imply higher initial energy. As in the Hoffman study, Turner did not
explicitly simulate deposition rates for Ni or Ti, but did produce results for Na, Sn, C, Cd,

Pd, Nb, Pb, Au and W. These results are shown in Figure 2-47, on which we have

91

superimposed a set of estimated contour lines giving the percent of atoms that actually
reach the substrate, as a function of binding energy (y-axis) and mass (x-axis). It is
apparent from this data that thermalization should not affect the composition in the case
of TiN i ﬁlms since both Ti and Ni fall on the same contour near 82.5%.

The initial energy of most of the ejected recoils is approximately half its surface
binding energy [Behrisch 1981; Thompson, 1981]. In the case of ejected recoils of Ti
and Ni ﬁ'om pure element targets, the recoils would have higher initial energy for Ti since
it has a higher surface binding energy. However, since the mean path and thermalization
distance is lower for Ti (see Table 2-6 and 2-7), it will transfer more of its energy or
thermalize at shorter distances than Ni. In addition, the less massive Ti atoms are
scattered more diffusely than the heavier mass Ni atoms. A combination of higher initial
energies, lower thermalization distance, and higher scattering probability of Ti could
result in its reduction in the incident ﬂux as compared to Ni, as seen in Figure 2- 46
where Ni and Ti lie on the same contour line. The key difference between pure materials
of Ni and Ti and the binary alloy is that surface binding energies and initial energies of
the recoils can be assumed to be the same in the TiN i alloy. If these assumptions were
true, Ti would have a lower incident ﬂux than Ni, because it is less rmssive and has
shorter thermalization distances. This would create a continual ﬂux to the substrate that
is Ni rich. Even though, this theory is probable, it cannot be concluded as the dominating
effect for Ti-depletion since other mechanisms can affect composition at substrate such as

sticking coefﬁcients and preferential resputtering, which are discussed in the below.

92

4. 3.4 Mechanisms that Aﬂect the Composition at the Film Surface
4. 3. 4. 1 Sticking Coeﬂicients

The sticking coefﬁcient is probability that an incident atom is adsorbed by the
substrate. If the sticking coefﬁcients are less than 1, only a ﬁ'action of the incident ﬂux
will condense on the surface. Thus, if there are differences in the sticking coefficients
between two sputtered species, the growing ﬁlm will be lean in the species with lower
sticking coefﬁcient. The sticking coefficients are thought to be close to unity for metals
with high melting points [Mahan, 1999]. However, small differences between the
sticking coefficients of Ti and Ni could account for the observed Ti-loss during
sputtering. If a difference in sticking coefﬁcient does exists, it would most likely be
lower for Ni since Ti is more reactive and thus is easier to be chemically absorbed by the
substrate. Furthermore, the energy of the sputtered particles promotes rapid surface
diffusion, allowing them to ﬁnd traps more quickly and thus be less likely to desorb.
Unfortunately, there is no data on the sticking coefﬁcients for Ni and Ti that can
conclusively eliminate it as the cause for observed Ti depletion.

Additional mechanisms need to be considered with respect to events occurring at
the surface of the growing ﬁlm Resputtering has been mentioned above, and can alter
the composition of the growing ﬁlm via any mechanism in which the probability of

resputtering for one species is greater than another.
4. 3. 4. 2 Preferential Resputtering

Ion projectiles (i.e., Ar) may be elastically backseattered from the cathode and can

then (having been neutralized) travel with high energy in the direction of the substrate.

93

Thus, the growing ﬁlm is also subject to projectile bombardment that causes a portion of
the ﬁlm to be ‘resputtered’. Here, the species dependence of the sputter yield becomes
important in determining ﬁlm composition

Preferential sputtering of alloys can cause a ﬁlm to have different composition
than the deposition ﬂuxes and occurs ﬁ'om differences in surface binding energy and non-
proportional kinetic energy transfer between the constituents due to differences in
kinematical factors. Preferential sputtering of the target only exists at the early stages of
sputtering, and eventually due to conservation of mass, the target and deposition ﬂux
composition will equilibrate. This steady state condition does not occur in the
resputtering of ﬁlms, since material is continually being deposited, and the surface is
continually being depleted in the preferentially sputtered material. Thus, the ﬁlm
composition will be lean in the preferentially resputtered species with respects to
deposition ﬂux. The differences in resputtering rates for Ti and Ni are discussed in the
following paragraphs.

Hoffman’s data on intrinsic resputtering shown in Figure 2-50, predicts that the
resputtering percentages of Ni and Ti are similar and would not account for preferential
loss of Ti. Further, if a difference between resputtering rates were derived ﬁom
Hoffman’s data, Ni would be found to be preferentially lost rather than Ti, since it has a
slightly higher atomic mass parameter. However, Hoffman used data gathered from the
resputtering of pure materials, thus his results may not reﬂect the resputtering rates that
occur for the same species in alloys. As mentioned above, it is possible that there are

sputtering mechanisms in TiN i, which could result in higher Ti sputtering rates than Ni

94

(using Argon). Thus Ti rather than Ni would be preferentially resputtered, which is
plausible explanation for the Ti depletion observed in the Triode ﬁlms.

If the sputtering yields of Ni and Ti in the binary alloy were assumed to be similar
and do not cause the Ti-depletion in the ﬁlms, there could be another mechanism to
explain the preferential loss of Ti. For instance, if Ni diffuses into the bulk and creates a
Ti rich surface composition, more Ti than Ni will be lost due to resputtering, resulting in
a overall ﬁlm composition depleted in Ti. The diffusion of Ni from the surface, rather
than Ti, assumes that Ni is the more rapidly diffusing species, but there must be a driving
force for diffusion. Normal ﬁckian diffusion as a result of concentration gradients is not
an obvious mechanism for the Ni diffusion, since we assume that Ni increases in
concentration away for the surface. However, Ni could diffuse by the Soret effect,
which is diffusion down an activity gradient which arises from the presence of a
temperature gradients. There certainly are very steep temperature gradients in the ﬁrst
few monolayers that could be as high as several hundred degrees Kelvin per micrometer
due to the heat generated by ion bombardment [Bunshah, 1994]. These temperature
gradients could lead to large activity gradients, and Ni, being a rapid diffuser, could
diffuse in the bulk ﬁlm via the Soret Effect and thus protect itself to the same degree
ﬁ'om resputtering.

An experiment was conducted by Ellesworth [2000] to calculate the Soret
coefﬁcients for Ti and Ni. The composition of TiN i alloy was not observed in this
experiment to change in the direction of the temperature gradient, and it was concluded
that the temperature gradients did not create a sufﬁcient activity gradient for Soret eﬂect

diffusion to occur.

95

Despite this negative result, the possibility still exists that the surface ﬁ'ee energy
of Ti is lower than that of Ni, which would lead to the same consequence. For the pure
elements, the surface energy of Ti is indeed lower than that of Ni (m = 1.69 J/m2 whereas
7N, = 1.81 J/mz). If this difference holds for binding in the intermetallic compound, then
Ti might again ﬁnd itself preferentially exposed to resputtering events. In this context it
should be remembered that for this mechanism, as well as for the other phenomena
discussed here, only a small difference in diffusion properties between Ni and Ti is
necessary to account for the observed composition shifts. It is the exquisite sensitivity of
the shape- memory effect to composition that makes the understanding of these subtle
phenomena important from a manufacturing standpoint.

None of the above mechanisms can account for the composition trend seen in
Figure 4-1. Preferential resputtering, thermalization, and differences in sticking
coefficients can only shift this composition trend to higher or lower relative composition
and cannot account for the observed time-evolution. Thus, a different mechanism must be

responsible for the trend such as target surface diffusion, which is discussed below.

4. 3.5 The Eﬂects of Target Surface Diﬁ'usion on Film Composition

In the case of alloys with different sputter yields, a surface altered-layer forms at
steady state that is depleted in the fast sputtering species. The erosion and formation of
new altered-surface layers are self-perpetuating such that after its initial formation, a
steady-state condition exists with continued target erosion. Assuming above arguments
on preferential sputtering, Ti will be the fast sputtering species and thus depleted in the

alter-layer. This surface altered layer creates a large composition gradient in the direction

96

normal to the target surface which will drive a diffusive ﬂux of the rapid-sputtering
species ﬁ'om the bulk to the surface. However, if the cathode is well-cooled, and the
target erosion rate is high enough, bulk diffusion will not have suﬂicient time to
inﬂuence the surface composition signiﬁcantly before it is simply eroded away. A
steady-state condition is theoretically attained in which the sputtered ﬂux must match the
bulk composition of the cathode. In the case of the diode magnetron, however, a second
complication arises. As a result of the inhomogeneous magnetic ﬁeld strength that gives
rise to the racetrack proﬁle, the overall sputtering rate will be at a maximum at the center
of the racetrack. Since the erosion rate is highest at this point, depletion of the rapid-
sputtering species will occur here ﬁrst, leading to the formation of a surface composition
gradient. This gradient will occur in a region of the cathode that is probably quite hot,
that is also under continual ion bombardment, so the surface diffusion rates can be
expected to be quite high. This could lead to a continual supply of the rapidly-sputtering
species ﬁ'om the edges of the wear track to the center of the track. This would, in turn,
tend to deplete the cathode at the edges of the wear track in the rapid-sputtering species,
under circumstances where steady state is not achieved due to the low erosion rate. The
result would thus be twofold. First, the initial enrichment in the rapid-sputtering species
would be larger, and the time required to achieve steady-state conditions would be longer
than would be otherwise expected. Secondly, the surface-diffusion mechanism would
result in a steady rise in output of the slow-sputtered species as the cathode becomes
steadily less and less able to supply the high-rate species from the edge to the center of

the wear track. This surface diffusion process is schematically shown in Figure 4-17.

97

It should be noted that for the case of sputtering from the triode source, where the
racetrack topography does not form, the relatively even cathode erosion proﬁle would not
result in surface composition gradients, and steady state would be achieved vary rapidly
under conditions where the bulk cathode was well-cooled. In addition, non steady state
conditions due to surface diffusion are not likely to be a signiﬁcant factor in a PM target
composed of pure materials, since no continuous concentration gradients exist.

In conclusion, this surface diffuse mechanism is a possible cause for the
composition shift, but it needs to be veriﬁed by quantitative analysis before it can be

attributed to the time-evolved composition trend.

98

CHAPTER 5 CONCLUSIONS

It is apparent from the experimentally observed composition shifts that the
deposition process is problematic for producing desired ﬁlm compositions. Due to the
complexity of the sputtering process, it was difficult to determine the principle
mechanisms and parameters that determine composition in sputtered TiN i and TiNiPd
alloys. Instead a list of possible mechanisms were discussed in terms of the of the
observed composition trends, which are outlined below. Further experiments will be
presented that may clarify composition problem.

The following composition trends were observed during sputtering of TiN i and

TiNiPd:

1. Magnetron sputtered TiN i ﬁlm compositions showed a continually changing
composition with target erosion The ﬁlm composition varied ﬁ'om 2 at. % Ti-rich
at 3 hours of sputtering to 1.5 at.% Ni enrichment at 24 hours of sputtering, with
an overall composition change of 3.5 at. %. There was an enrichment of Ti at
lateral positions and especially above the wear track at earlier hours sputtering,
which diminishes as sputtering time increases.

2. Triode ﬁlm compositions were nominally constant with target wear. The ﬁlm
composition showed a consistent 3.5 at. % Ti-loss relative to the target
composition, and a slight lateral increase Ni content from the center position.

3. In the case of TiNiPd ﬁlm compositions, titanium was initially depressed by over

4 at. % relative to the 53.5 at. % cathode composition, but raised over the ﬁrst 5

99

hours of cathode wear to ﬁnally stabilize at near 50 at. %. Ti content was
observed to decrease with increasing substrate temperature and was observed to
increase with working gas pressure. Pd content was also observed to decrease
with increasing substrate temperature and working gas pressure. Composition and

deposition temperatures showed a strong effect on ﬁlm transformation

temperatures.

The following mechanisms may possibly cause the experimentally observed ﬁlm

compositions shifts:

l. The partial sputter yields of Ti in the alloy may exhibit higher yields than pure
materials.

2. Formation of surface topography that creates a preferential forward ejection of
fast sputtering species.

3. Differences in the angular distributions of the sputter ﬂux between alloy species.

4. Differences in sticking coefficients.

5. Differential reduction in incident sputter ﬂux caused by scattering due to
collisions with gas neutrals.

6. Preferential resputtering of the fast sputtering species or by a surface enrichment
of the slow diffusing species.

7. Surface diffusion of titanium into the wear track that causes a continually

changing sputter ﬂux and ﬁlm composition with target erosion.

100

Triode sources give good compositional uniformity over the life of the cathode,
however, they create a Ti depletion in the ﬁlms. It is difﬁcult to determine the
dominating mechanisms that cause Ti-loss. It is possible that a combination of
preferential resputtering, differences in sticking coefﬁcients, and the reduction in incident
ﬂux due to scattering are responsible for the Ti depletion. The observed enrichment of Ni
at lateral positions in these ﬁlms may be related to differential angular distributions, due
to a mass effect and energy transfer difference between species that prefers the forward
ejection of Ti and enriching it in the ﬁlms at the center and Ni at lateral distances.

Both the melt-solidiﬁed and PIM targets seem to have composition control
problems related to formation of surface topography in magnetron sputtering. In the case
of melt-solidiﬁed targets, it is possible the formation of the wear track, and the new
surface altered-layers promotes a continually changing Ti surface diffusion ﬂux into the
wear track that may be responsible for non-steady state composition shift with time. P/M
targets may have long burn time associated with the formation of a pitted surface
morphology before steady state composition ﬂux is reached. Also, this surface
morphology may create a preferred ejection of Pd normal to the target that results in an
enrichment of Pd in the center of the ﬁlm.

To gain a more deﬁnitive understanding of sputtering process effect on
composition further experiments are necessary. The following list suggests some the

experiments that could clarify the compositional control problem:

101

1.

The surface diffusion Ti into wartrack my be understood by a quantitative
analysis that models mean surface diffusion time as function erosion rate of target
and partial sputter yields of Ni and Ti.

Further investigations must be conducted on the effects of target wear and
racetrack surface evolution on angular distributions of Ti and Ni. One such
investigation is the development of computer simulated models that track angular
distribution and ﬁlm composition change as function of surface angle change
using integrated sputter yields summed form a collection of points on the target
surface. This simulation may also explain the variation of Ni and Ti content with
lateral distances, which are not completely understood.

There should be ftrrther experiments in process techniques that eliminate rather
understand composition effects of angular distribution. For example, an
investigation of the ﬁlm compositions created by a substrate that oscillates over
the target, which will collect sputter atoms at central and lateral positions of target
at all positions of the substrate. In theory, this should result in more uniform ﬁlm
composition.

Experiments at high working gas pressures should be investigated since uniform
ﬁlm compositions can be created due to the signiﬁcant scattering of sputter
particles and the elimination of any preferred angular distribution. In addition, the
resputtering rate should be signiﬁcantly reduced at high working gas pressure due
thermlization of energetic neutrals. If the Ti-loss is not observed at higher
pressures (9 Pa), it may prove that resputtering is a dominating factor in the

change ﬁlm of composition. High pressure also can reduce composition changes

102

scattering by fully thermalizing both species and randomizing the deposition ﬂux
such that there is not a overall difference in incident ﬂux between Ti and Ni. In
general, the high working gas pressure may solve the compositions by it has been
known to create ﬁlms with poor mechanical properties

. If you assume that resputtering causes the Ti depletion in the thin ﬁlms due to
differences in the partial sputter yield between Ti and Ni, then mixtures of Ar and
Kr sputtering gas can be used to enhance the sputtering rate of Ni in the alloy.
Krypton will transfer more energy to Ni atoms rather than Ti, since it is more
similar in mass to Ni. The addition of krypton may increase the Ti content in the
ﬁlms, but it does not solve the problem of the temporal variation in the
composition. It may be possible given a certain set of ﬁxed sputtering parameters
(such as working gas pressure, substrate-target distance, sputter power, initial
target composition) to calibrate the variation of composition with time and the
ratio Ar/Kr, such that composition no longer varies with time. This could be done
thru a set of experiments that measure composition variation using different Ar/Kr
ratios, and then by using a computerized gas controller, the gas ratio is varied
precisely with target erosion. This may be a good method for solving the
composition problem, but it will give no scientiﬁc explanation fro the occurrence
of the composition variation with time.

. Insight could be gained in the preferential sputtering nature of binary alloy by
conducting sputter Auger composition measurements. The Auger surface
compositions can be measured for unsputtered piece of binary TiN i and compared

with surface compositions after a few minutes of sputtering. If differences in

103

surface composition exist, then it is indicative of preferential sputtering. These
experiments may bring clarity to the effects of resputtering and partial yield
ampliﬁcation.

. Finally, composition studies of cylindrical magnetron sputtered ﬁlms should be
investigated. In cylindrical magnetron sputtering, the substrate lies in the center
of a hollow cathode, and substrate receives sputter ﬂux for all angles. This would
smear out any effects of differential angular distributions. Furthermore, the
sputtering occurs in a closed system, therefore the ﬁlm must have the composition
due conservation of mass. This type of deposition may solve the composition

problem entirely.

104

BIBLOGRAPHY

105

BIBLIOGRAPHY

Anderson, H. H. and H. L. Bay, “Sputtering Yield Measurements,” in Sputterr_ng° by
Particle Bombardment 1, ed. Behrish, Springer-Verlag, Berlin, 1981

Anderson, H. H, in “Symposium on the Physics of Ionized Gases 1980,” B. Cobic, ed.,
Boris Kidric Austria Nuclear Sciences, Belgrad, Yugoslavia (1981)

Behrisch, R., “Introduction and Overview,” in Sputtering by Particle Bombardment I. ed.
R. Behrish, Springer-Verlag, Berlin, 1981

Berg, S. and I. V. Katargdjiev, “Preferential Sputtering Effect in Thin Film Processing,”
Journal of Vacuum Science Technology, Al7(4), 1999

Betz, G., R. Dobrozemsky and F. P. Viehbock, Bed. Tijdschr. Voc. Tech, 8, 203 (1970)

Binary Alloy Phase Diagrams, Vol. 2, ed. T. B. Massalski, American Metals Society,
Metals Park, (1987), 1768

Bunshah, R. F.,Llandbookof Deposition Technologies for Films and Coatmg' s, Science
Technology and Applications, Park Ridge, New Jersey, Noyes Publications, 1994

B_u_scp, T. and E. D. McClanahan, Thin Solid Films, 47:29] (1977)

Chapman, 8., “Glow Discharge Processes”, John Wiley & Sons, New York (1980)
Chang, L. and D. S. Grummon, Scripta Metall., 25(1991), 2079

Chang, L., PhD. Dissertation, Michigan State University, (1992)

Chattopadhyay, G. and H. KleyKamp. Z Metallkd., 74 (1983), 1982

Carter, G., J. C. Colligon and M. J. Nobes, J. Mat. Sci., 6:115 (1971)

Coburn, J. W. and M. Chen, J. Appl. Phys, 51, 3134 (1980)

Coburn, J. W., Thin Solid Films, 64:37] (1979)

Duerig, T. W., K. N. Melton, D. Stockel, and C. M. Wayman. Engineering Aspects of
Shape Memory Alloys. Butterworth-Heinemann: London 1990

Eckertova, L., Physics of Thin Films, Plenum Press, New York, 1977

106

Ellesworth, K., Unpublished Data, Bachelor Thesis, Michigan State University,
2000

Eltoukhy, A. H. and J. E. Greene, J. Appl. Phys., 51:4444 (1980)

Gillam, E., J. Phys. Chem. Solids, 11:55 (1959)

Grace, J. M., D. B. McDonald, M. T. Reiten, J. Olson, R. T. Kampwirth, and K. B. Gray,
“Selective Resputtering of Bismuth in Sputtered Bi-Sr-Ca-Cu-O Films”, Journal
of Applied Physics, Vol. 70 (7), 1 Oct. 1991, pp 3867-75

Green, P. A. and B. N. Chapman, J. Vac. Sci. Technol., 13 (1970), 165

Greene, J. E., B. R. Natarian and F. Sequeda-Osorio, J. Appl. Phys., 49:417 (1978)

Gschneider, K. A “Physical Pr0perties and Interrelationships of metallic and
Semirnetallic Elements”, Sol. St. Phys. 16, 275, 1964

Harrison, D. E., P. W. Kelly, B. J. Garrison, and N. Winograd, Sur. Sci., 76:311 (1978)

Hodgson, D. E. and J. W. Brown, “Using Nitinol Alloys” Shape Memory Application,
Inc., (2000) SMST—2000

Hofﬁnan, D. W., “Intrinsic Resputtering-Theory and Experiment”, J. Vac. Sci. Technol.,
A8(5), 3707, 1990.

Ishida, A., M. Sato and S. Miyazaki, “Mechanical properties of Ti-Ni shape memory thin
ﬁlms formed by sputtering”, Mat. Sci. and Eng., A273-275 (1999a), 754-757

Ishida, A. and S. Miyazaki, “Microstructure and Mechanical Properties of Sputter-
Deposited Ti-Ni Thin Films”, Journal of Engineering Materials and Technology,
Jan. 1999, Vol. 121.

Ishida, A., M. sato, A. takei, K Nomura and S. Miyazaki, Metall. Trans. A, 27A (1996),
3753

Ishida, A., M. Sato, A. Takei and S. Miyazaki, “Shape Memory Behavior of Ti-Ni
Annealed at Various Temperatures”, Mat. Re. Symp. Proc., Vol. 360 (1995)

Ishida, A., A. takei and S. Miyazaki, Thin Solid Films, 228 (1993), 210

Jardin, P., “Deposition Parameters for Sputter-Deposited Thin Films TiN i”, Mat. Res.
Soc. Symp. Proc., Vol. 360, 1995, pp. 293-8

Johnson, A.D., V. V. Martynov and R. S. Minners, J. Phys. (France), IV 5 (1995) C8-783

107

Kay, E., Adv. Electronics and Electron Phys, 17 (1962), 245

Kelly, R, “The Mechanisms of Sputtering-Part. I Prompt and Slow Collisional
Sputtering”, Sol. St. Phys. 16, 275, 1984

Kim, J. J., P. Moine and D. A. Stevenson, Scripta Metall., 20 (1986), 243
Leawer, K. D. and B. N. Chapman, “Thin Films”, Wykeham Publication, London, 1981
Nobes, M. J., J. C. Colligon and G. Carter, J. Mat. Sci. , 4:730 (1969)

Oechsner, H., “Sputtering of Polycrystalline Metal Surfaces at Oblique Ion Bombardment
in 1 keV Range”, J. Phys. 261, 37, 1973

Okutani, T., M. Shikata, S. Ichimura and R. Shimizu, “Angular Distributions of Si Atoms
Sputtered by keV Ar+ ions”, J. Appl. Phys, 51(5), 1980

Olson, R. R, M. E. King and G. K. Wehner, “Mass Effects on Angular Distribution of
Sputtered Atoms”, J. Appl. Phys, 50(5), 3677, 1979

Rossnagel, S. M. and R. S. Robinson, J. Vac. Sci., 20(1982), 506

Rossnagel, S. M. and H. F. Kauﬁnan, “Current-Voltage relations in Magnetrons”, J. Vac.
Sci. Technol. A6(2), 233, 1988

Selinder, T. I., G. Larsson and U. Helrnersson, “Resputtering Effects on the
Stoichiometry of YBazCU30x Thin Films”, Journal of Applied Physics, Vol.
69(7), 1 Jan. 1991, pp 390-5.

Scherzer, B. M. U., “Development of Surface Topography Due to Gas Ion
Incorporation”, in R Behrisch, ed., Sputterr__ng' by Particle Bombardment II.
Springer-Verlag, Berlin, 1983

Shugo, Y., Y. Yagi and T. Honma, Bull Res. Inst. of Mineral Dressing and Metallurgy,
Tohoku University, 43 No. 2, 139(1987)

Sigmund, P. and C. Claussen, “Sputtering ﬁ'om Elastic Collision Spikes in Heavy -Ion —
Bombardment Metals”, Phys. Rev. 184(2), 383, 1981

Sigmund, P., J. Vac. Sci. Technol., ”3% (1980)
Sigmund, P., J. Mat. Sci., 8:1545 (1973)

Sivokha, V. P., A. S. Savvinov, V. P. Vomin and V. N. Khachin, Phys. Met. Metal, 56
(112-116), 1983

108

Somekh, R E., “The Thermalization of Energetic Atoms during the Sputtering Process”,
J. Vac. Sci. Technol., A2(3), 1285, 1984

Townsend, P. D., J. C. Kelly and N. E. W. Hartely, Ion Iprplantatiop, Sputtgr_rng' , and
their Applications, Academic Press, London, 1976

Turner, G. M., I. S. Falconer, B. W. James and D. R Mckenzie, “Monte Carlo
Calculation of the Properties of Sputtered Atoms at a Substrate Surface in a
Magnetron Discharge”, J. Vac. Sci. Technol., A 10(3), 455, 1992

Van Loo, F. J. J., G. F. Bastin and J. H. Leenen, Journal of Less Common Metallurgy, 57
(1978), 11

Vossen, J. L., J. Vac. Sci. Technol., 8:812 (1971)

Waits, P. K, in “Thin Film Processes”, eds. J. L. Vossen and W. Kern, Academic Press,
New York, (1978), 131

Wang, F. E., W. J. Buehler, and S. J. Pickart, Journal of Applied Physics, Vol. 36, 1965,
pp.3232

Wasilewski, R. J., S. R. Butler, J. E. Hanlon and D. Worden, Metall. Trans, 2(1971), 229

Wayman, C. M., “Phase Transformation, Nondiffusive”, Physical Metallurgy, 3rdl ed.
Elsevier Science Publishers (1983), pp. 1032-78

Weast, a (2., ed., HandbooleClﬂristry and Physics. 56‘”, CRC Press, Cleveland OH,
1975

 

Wehner, G. K and D. Rosenberg, “Angular Distribution of Sputtered Material”, J. Appl.
Phys, 31(1), 1960

Westwood, W. D., “Calculation of Deposition Rates in Diode Sputtering Systems”, J.
Vac. Sci. Technol., 15(1), 1, 1978

Xie, C. Y., L. C. Zhao and T. C. Lei, Scripta Metall., 24(1990), 1753
Yamamura, Y., Y. Itikawa and N. Itoh, ﬁgplar Deppndence of Spuﬂeﬁnglields of

Monpatomic Solids, Research Information Center, Institute of Plasma Physics,
Nagoya University, Japan, 1983

109

R

  
  
 

 

 

 

 

 

 

 

 

   

’ //

888% z

35 /§ / %

at £8 /‘ p

‘5 /32 /

%8/. r

a V ///i &
Temperature

above A} b01333“: below Mg below Mg .d'l_'°"“'d
srPecmeuJ J “ r l —.

“.0

 

 

Figure 2-1 Schematic of Shape Memory Effect and Transformation Temperatures
Observed in Electrical Resistivity vs. Temperature Plot [Duerig, 1990].

110

Temperature °C

 

12mg @383 2323.

 

 

 

 

 

 

 

 

 

... ....... “reason. a s... e e .2
d eeeee — <<<<<<<<< 1 1.e:1d1d14441‘111“111441<—11‘.44J1id4-<ll«44<e-u<.44: 4 d .......
h
v
I
H. .
=3...
W...O T
eon _.._..., h
95 n
n
-..l .
as:
in. .
N .
i r
r. '- I
w
v
v
n
n
.
I'II'I---' '0--'I§°hllll-"l H
nnnnn .t<-11<.‘4u<‘<-1144¢44:...sqeqten. tier-d<<<441«.~4‘<..:
50 no 8 .3 co co .8
>338 page: Enra— z.

mam—=6 N-» 9an 35mm Use—mama om iv: 3ng >=o<m 3.80 UmmmBB Gamay Bmd.

111

 

 

 

 

 

 

 

1400
I a" p K. x \
, tozs 220° ,/
IOOO' ‘ ﬂ
ﬂ+TlNl,
.‘i
g 800'-
3
g 625 3 20°
0 ’ . .
2‘ “of Ti, imp 1’ t «44513,;
c ' ' .
'- 400- i 15133.15,I ___...t
200‘. N ' °t ‘QuﬂiaT'z
T' i+mor arm 0 ‘
o—“--—-1—‘-%—qb-—---——I—-—.-
45 50 55

Ni , atomic percent

Figure 2-3 Equilibrium Phase Diagram Near Equiatomic Composition at Low
Temperature of TiN i Binary System Proposed by Wasilewski et al.

112

 

 

 

200 j l I l l
o
150L 0 -
e
100 — ‘ ~
{3 a . A ‘
O
9 50 . N ‘9 -
‘3 I I {3‘ I
3 A...
E if
0 __ _.
.8 . u
U o
_ 0 Wang et al. A 4
—50 .
AlHarnson et al.
olHanlon et al. ‘4
400 __ ulPurdy et al.. __
O
-150 1 1 L 1 1

 

 

47 48 49 50 51 52 53
Nickel atomic (%)

Figure 2-4 Martensite Start Temperatures as a Function of Ni Content [Melton, 1990;
Wang, 1965].

113

 

 

 

 

M. W
i - n X-SB
3 Mm
f #—
Goolng
é m ,,,,,,,,,,,,,,,
. 3 -h.° ------- \J\”":
----- . - m. ’4'

X-gg, ......... \V

we ............ x

sun—O". \’

Hauling A'
lLLLllJJIlIIlJLIIJIJ]
150 zoo 250 30° 350

TWoo

Figure 2-5 (a)Effect of Aging Time on Ni- rich TiN i Thin Films [Miyazaki, 1999].

“-43.le thin ﬂm
973K 3.610 V1.0. - 77% X It! W.G.

 

Hoot Flow 0.1(mWIg) I....l

 

 

mating
lLlllllJlllllllJl
'200 260 300 350

Temperature (K)
Figure 2-5 (b)Effcct of Aging Time on Ti- rich TiN i Thin Films [Miyazaki, 1999].

 

114

 

30° . Ti-51.Qat°/oNI thin film

i 973K 3.6ks w.o. - 773K 3.6ks w.o.

250‘ g #674:

 

 

 

 

 

3" 200 - 9

U) r
E 1

i o Ti-51.Qat%Ni (DSC)

1250- A Ti-51.Qat°/oNi (MEC

A “-43.939/0Nl (DSC)

o 11-43.9at%Ni (MEC

100 If........ - .-.-..-1 L L-u--.l - --.l-.

o 1 10 100 1000
X (ks)

Figure 2-6 DSC Results of the Effects of Aging Time on Ms Temperature for Ni and Ti-
Rich TiN i Thin Films [Miyazaki, 1999].

115

 

o M (ZOOMPa)
oA (ZOOMPa)
I M (360MPa)
DA (360MPa) '
A M (520MPa)
AA (520MPa)

 

§

 
  
  

3

Transtormation temmrarure (K)

 

 

l —

 

5%

773 395" " 973
Annealing temperature (K)

Figure 2-7 The Eﬁ‘ect of Solutionizing Temperature on Transformation Temperatures in
Ni-rich TiN i Thin Films Observed by Strain-Temperature Curves at Constant

Stress [Ishida, 1995].

116

 

 

§

 

§

 

L¢/‘;/i '
hit/7333;:3

I M (ZOOMPa)
u A (ZOOMPa)
A M (scours) u
A A (360MPa) ’
o M (520MPa)
o A (520MPa)

Transformation temperature (K)
N
8 §

 

Tia 873 973
Annealing temperature (K)

 

I l

 

§

Figure 2-8 The Effect of Solutionizing Temperature on Transformation Temperatures in
Ti-rich TiN i Thin Films Observed by Strain-Temperature Curves at Constant
Stress [Ishida, 1995].

117

  
           
 

no2
Ti 0 M

1200 K

4 7 Ti o -—

3 5 ’

Tizoa/

TiOlYl

Y" ___':; t
5/ \I

\‘ V

(U

\Q\

‘ l

1 OTizNiiC”
tI-Ti oTiNi; Tian

OTINI;
a- Ti 0 TizNilOI

I ‘uu’\ oNiiTi

B-Ti le NI TiNi TiNi; NiiTll

Ti Ni

1.!

Figure 2-9 Titanium-Nickel-Oxygen Phase Diagram [Chattopadhyay, 1983].

118

 

 

 

 

 

OA‘O‘
°o 5 * 13””1‘5“ ’20
Cu-oontent(at%)
4° Tl-Ni-Pd thin films
'— ' 973K-3.6ks W.Q.

 

Hysteresis (K)

 

 

 

 

0o 5 10 15 20 25 30
Pd(8t°/o)

Figure 2-10 The Effect of Cu and Pd Content on the Thermal Hysteresis[Miyazaki, 1999].

119

 

1100 a -
'= MNI(50-X)de) ;

1000 -= “50(Nl(so-x)Pix)
o= Donkersloot and Van Vucht 1970 .0.
goo
aoo
7oo
o
0 ~ 600
‘2
3 500
2
o 400
g .
a, 300 l
I'-
g 200
100
0 I
-100 ‘

 

0 10 20 30 40 so
Composition, x

Figure 2-11 Ms Temperature as a Function of Pd and Pt at. % [Lindquist, 1988].

120

 

 

 

 

ta ‘ 0.0318 PIC‘FL "Imam

120 can
100-
U
0
o‘ ”"
g .600
d
i 60H
40-
8
I}
g 20- .400
i
0..
l1

 

-20—1

 

 

 

i i i i
4915:. 50/50 51149
nickel Titania Ate-to ’5

Figure 2-12 The Effect of Composition on Austenitic Strength [Hodgson, 2000].

 

 

 

 

1600
51.5%!"
1200 _ 48.3%!“
g 50.0%”
.5.
g m h
400 -
o L L j 1
O 10 20 so 40
Strain (%)

Figure 2-13 Stress-Strain Curves for Various TiN i Thin Films Compositions
[Ishida, 1999].

121

 

 

 

 

600
' 0 773K
«a °' “gist
‘ o
O' A A 623K
2 400 _ o
D.
g \
on:
E ------- El- - - o ___________ _ .
g 200 p o
'o
00 200 400
Preclpitate diameter.dlnm

Figure 2-14 The Effect of Precipitate Diameter on the Critical Stress for Slip in Austenite
[Ishida, Microstructure 1999].

122

Strain (%)

 

—

 

 

NNOX mwux muwx
angst. t onoivn r unogtn
P NOOKVI ‘ 803V. Nooivo
1...”... 1.1.1.1 33v- tllllllllllli so u
_ mapgﬁu . _ _ a p _

 

 

goo

woo

08 .80 ..oo woo «8 3o
333328 :3

.oo woo woo Aoo

Enid Ni; mnmmaéoabonmﬁﬁn 0:28 3 00383 mqnmm omzmowomaor .27: HE:
255 >38on 2 5595 #333358 men we rm 533.38%

123

 

1w

 

 

 

 

 

0 5 10 15 20 25
Strain (96)

0

Figure 2-16 (It) Stress-Strain Curve of Ti—48.2 at.% Ni Thin Films Annealed at 773 K for
3005 [Ishisda, 1999].

 

1500

1 73K 287K

 

 

 

 

o 5 1b 15 20 25
Strain (‘16)

Figure 2-16 (h) Stress-Strain Curve of Ti- 48.2 at.% Ni Thin Films Annealed at
873 K for 3.6 ks [Ishisda, 1999].

124

Strain (%)

 

   

     

   

 

 

 

 

ﬂux Sax $25.: max
9.3:."-
- 30:3 - t
o ”.333
302:: 802.3 . t

a r 8535..
82s? ”oozi-

‘ .l .I .l

N t i r 32:"...

F u _ . rim _ _ .
‘00 woo 000 500 ‘00 woo 000 #00 .00 woo woo boo

aoaooqmacqo :3

Ems—d V: mﬁBEéoBuaBﬁﬁo 0:28 A: 025:5» maaomm om meECB-Eo: 122m HE:
355 gnu—om 2 $508 HoBUQﬁEom we we 5 EmrEP_oc&.

125

 

Lettlce Parameter. nm

 

 

 

 

 

 

 

 

‘:
E
E
0280 E
E
E
0275 E-
E
t
om D AAAAAAAA 1 AAAAAAAAA LAW ..... l ........ LL; AAAAAAA
10 20 30 4O 50
Atom Percent. X

Figure 2-18 Lattice Parameter Data for B19 Martensite as a Function of Pd at. %
Reported by Lindquist (I), Sivokha et al. (C), and Shugo et al. (X).

 

0.310

I V ' V V U I V

0305

Lattice Parameter. nm

 

 

 

 

Figure 2-19 Lattice Parameter Data for B2 Austenite as a Function of Pd at. %
Reported by Lindquist (I) and Shugo et al. (C).

126

Ti-26.4Ni-21.8Pd(at%)
973K-3.6ks W.Q.

Heating

1 °/o

Strain

 

 

 

 

 

 

 

 

 

 

 

 

 

300 350 400 450
Temperature(l()
Figure 2-20 Strain-Temperature Curves for Ti-(Ni+Pd) Alloy [Miyazaki, 1999].
age 0 (500 0V) 19
22.2“ (5° "L: 242; o 5100 M 9

0 ‘ i l 7,_"—1—'—"_'—2
P. E E E O 50 eV 1
‘0-02 7' """ Z' ““““““ i """""" t """" I 100 .v ----- A
: .5 . § 5 . 500 eV :
'0-04 C' """" : """" ‘I‘i """"""" ‘: """""" ’: """""" ‘E‘ """""" 1
g s = I i a a -
as -0...) """""" ° """""" a """" t " g “1
3 I E E i i i I
i: -0-08 r """"" :‘ """"" t """""" ‘: """"""" '5 """"""" 1' """""" 1
< - s i 5 s s .
'0-1 2' """"""" """"" i """"""" """""" ‘; """""" "' """""" -:
~0-12 :- """""" 2‘ ----------- i --------- é ----------- -§ ----------- -§- --------- -:

-o.14 - i i '1 ' i

0 0.1 0.2 0.3 0.4 0.5 0.6

Ion/Atom Arrival Ratio
Figure 2-21 Comparison of the Ti Content of IBAD TiN i Films with the Expected Film

Composition as Function of HA Ratio in the Absence of an Assisted Beam
[Chang, 1992].

127

 

 

 

 

 

 

 

 

 

 

 

 

 

 

52.01- \
51.5 T
g \¥ ..
2 51.0" __
.1
§ ........ .......
E 50.5 s- """" “' """" "“
50.0--
49.5 E I ‘1
lsthour 2ndhour 3rdhour 4thhour
TargetEmsionTimc

Figure 2-22 The Effect of Target Erosion on Composition [Lee, 1994].

  
   

 

 

 

53.0 E Center
52.5 i -""°——25mm off axis

g: 52.0 -

3 51.51

8

35 51.0 -
50.5 *:::::::::::::: 2::::::::::::::::":::::::::

E T \ Sputter Target
50.0 » Compomuon
49.5 , l J t l 1 L J :

0.44 Pa 0.64Pa 1.01 Pa

Pressme

Figure 2-23 The Effect of Pressure on the Composition [Lee,l994].

128

51.8 -

‘

 

 

51.6 r-
? 51.4 --
2’ 51.2 4'
1:
"‘ 51.0 " Sputter Target
E Composition
50.8 " /
50.6 iE22222212222222IZZZIIZIZZZZZZZZZZIZZZZ--2222:
50.4 L i
51 um 76 m

Figure 2-24 The Effect of Target-Substrate Distance on Composition [Lee, 1994].

52.5

52.0 r

51.5
51.0
50.5
50.0
49.5

Composition (Ni at%)

“‘I

W

W

  
 

*:::::::::- ::::::-- -- - “man's“
' ' Composition

 

V

V

 

49.0

150 Watts 3m WM 450 Watts

Figure 2-25

Plasma Power

The Effect of Power on Composition [Lee, 1994].

129

— WORK ING
GAS

    
  
  

    

 

f“
l
-’ ' .' f. .1: ‘.'.' '

.-:-:-;.;I;' , . ’4 ...... . .
~-: ;~:-:-.-: ..-:-;;t.'-:i'-:r .1 _ ,4 ._. :1-:- ,. .
, . . . . . .‘hv.;.;.: 0.3:" ::.'::;;;._ '1- :;;;.:::::‘:'::;:::' ukvfrk'
. _,'....._.',; t_._. c ‘.'.. '3 c - '
'-.'-‘.----, .5:

        

...........
...........

PWER
SUPPLY

     

...M... A...

 

 

 

 

VACUUM
VACUUM CHAMBER

PUMPS

 

 

 

Figure 2-26 Schematic of Sputtering Apparatus [Bunshah, 1994].

 

 

 

 

A
Jramic Conduction
log(CIrrent)
Figure 2-27 The Voltage-Current Relationship in a DC Discharge [Mahan, 1999].

 

130

Negative Anode
Cathode Glow Glow
Glow Pos'nlve Column

1

CATHODE

 

Figure 2-28 The Luminous and Dark Regions of DC Glow Discharge [Mahan, 1999].

 

Sheath (1) (3)

 

 

 

Figure 2-29 A Collision Cascade Resulting in a Sputter Atom or Recoil [Mahan, 1999].

131

‘0‘» e e e e I e e \- """ e ' e
e .:§;..w. e e e .9. ”I“ e e
I”: e e e e a -—-" ' v ”e e e .-,..‘/
e e e e e e e e e e e\‘\
e e e e e e a e e
(a) (b)

//

Figure 2-30 The Three Energy Regimes of Collision Cascade (a) Single Knock-on
Regime, (b) Linear Cascade (0) Spike Regime [Sigmund ,1981].

132

EXPERIMENTAL TARGET-DEPENDENCE OF
YIELD FOR 1-IteV ARGON PROJECTILES

 

 

 

 

 

 

 

9 ‘ 7 fr ' 1 Ordﬁ I V r iﬁ
8 ~ ..
o 7 P J
E 5 r 1
2' 5 * P” .
4 - ..
3 i i
)- .1
go ' C'Fcu Ge Au 1
2 :39 Flu Rial FlaIr Pt Th ‘
I " 816.0% M ZrMoNb RhSn Ht Oa '9

L l A L A LTaw l . I

103 L20 30 40 50 60 70 80 9O

TARGET ATOMIC HUBER

(a)

0 RECIPROCAL SURFACE BINDING ENERGY

0.9
0.8 J
0.7 - 1
0.8
o 5 » P“ i
r t

0.4 L 1
0.3 “NW3 :3 Sn Au .
Pt U.

0.2 EC Fm HI '. ‘

0.1 NO RU Ta wlso Th

 

v
A 1

Cd
0 .

' r

etc
I

 

tIU-u (0V4)

 

 

 

010 2030405060M7080 90

TARGET ATOMIC MJMBER
(b)

Figure 2-31 The Effect of Surface Binding Energy on Yield [Mahan and Vantomme,
1999}

133

 

IIIIIII T 1 IIIIIII 1 1711111; r IirTunr-i'l
’

I I TTTIII

N
T

 

I—o —

b .I‘

l- .1

F- u

l— -i

51— ‘

P- w

h- d

2- "
4

IO— —

__ .4

p d

,_ u

,- -l

91-- "

Y b a

2i- "‘

— _
’- ‘4
l- d
P -1
.- Oi

0t- 5 “
.. 0 ‘ -
P .4

ﬁg

 

 

 

 

 

*— 1
I n
i- 4
5" -
( d
2:- .1
4 / .l .‘ 11114051511
”1 11111 1 ’1 1144111 1 1 LllLLlj 1 1 1 111111
5 0,] 2 9 IO 3 5 IO 2 5 m
(a) ENERGY mvt

Figure 2-32 The Projectile Energy Dependence on Yield for Tungsten Target Struck by
Various Projectiles [Anderson, 1981].

134

 

90 r l r l ' I r I ' I ' 1 ' I ' I ' l
70 r a
‘ Xe

«Kr

PROJECTILE ATOMIC NUMBER

‘ Ar
* Ne
- He

too

 

 

 

 

 

1R"

Xe

Kr

"Ar

PROJECTILE ATOMIC NUMBER

 

 

 

o 10 20 so 40 so so 70 so so 100
(b) TARGET ATOMIC NUMBER

 

Figure 2-33 A Contour Plot Showing the Effect of Projectile Mass on Sputter Yield with
Target Atomic Number (a) Yield Data ﬁ-om Empirical Formula Developed
by Matsunami et al. [1984] (b) Yield data from a Simpliﬁed Collision
Model [Mahan and Vantomme, 1999].

135

 

5.0 I T I l I T f1 I

H. 00011806 (1973)

Vl‘vrj
l

1 L 1

4.0

O:1.05keVAr’-Ii

Vial/Viol
YMNM

 

 

 

 

 

5.0

4.0, L H. calm-titers) f.
o: t.05kaVAr‘~Ta

3.0

2.0

lLJillllJll

IIIIIIT]

  

 

1.0

LlLl

 

 

 

o o 1 so so so
Angleot incidence (degree) Nubotinddenco (m1
(8) (b)
5.0
4.0 L H. Oedtener (1973) J
. o : 1.05 keV Ar ~ Ag ~
.. 3.0 - -
g . .
a
>-

 

 

 

 

Angle 01 itcidence (degree)
(6)

Figure 2-34 The Effect of Ion Incidence Angle on Sputter Yield ( solid lines are
Theoretical calculation by Yamamura et al. [1983] and circles are
experimental data from Oechsner [1973]).

136

[3:lll

ie-ii

, (a)
a
Target (cathode) \ u» /

,0 Target
Ar' .' /' (C)

\\\\\:§\\\\\\\\\\\\\;\\‘.;\\\\\\\\\\\\\\\\\‘
“‘73 s I Annular Permanem

,NJ
(b)

 
         

 
 

 

Mamie Pole Pcece
Figure 2-35 The Electron Trapping Effect of Applied Magnetic and Electrical Field
[Mahan, 1999].
NE CAT I VE
HIGH VOLTAGE
THERMIONIC I moor

    
  
 

 

SUBSTRATE
HOLDER

Figure 2-36 Schematic of a Triode Sputtering Source and Conﬁguration [Bunshah, 1994].

137

(a) Single Atom Arrives

 

 

  

Islands Growing

 

 

 

 

 

 

Substrate
(b) Migration Re-cvaporation (t) IslandShape
. . Cross-Section
0*O\ l/ ms
' ti.
0 .
(c) Collision & Combination of
Sin le Atoms

 

 

.\m/O

 

 

 

 

 

(d) Nucleation

"Island" of Atoms

 

 

 

 

 

 

Figure 2-37 Schematic of Film Formation and Growth [Leawer, 1981].

138

 

  
 

MAGNETICS

 

 

 

Figure 2-38 Schematic of an Unbalanced Magnetron [Rossnagel, 1982].

 

men
WELD ‘0'" me» me:
YIELD YIELD YIELD
MATERML MATERIAL museum unemu
00.0.0.0

00000000000,

ex; .' 4, . ’: ‘1" ,. ;' 4572""
7‘39” ‘ TARGET

BEFORE SURFACE EWILIBRATION AFTER SURFACE EWILIWTION

Figure 2-39 Schematic Diagram of Sputtering Yields ﬁom Initial and Altered Layer
Target Surface Compositions [Bunshah, 1994].

INCIII’NT ION

MATERIAL WITH
LON SPUTTIRING Vltl0

 

Figure 2-40 Schematic Drawing of the Formation of Cones [Bunshah, 1994].

 

Figure 2-41 SEM Photomicrograph of Cone Structure on the Surface of a Sputtered Ni-
Rich TiNi Target [Lee, 1994].

140

Over Cosine Surface Normal

Under Cosine
Cosine

 

 

 

 

Figure 2-42 Angular Distribution of Sputter Atoms [Kay, 1962].

141

 

RF 200W
5
.5 A 3.
8. o\°
g 5i: Composite (Tb+F e)
e :3, /
‘ 0.9.0
Compound “12517675

 

 

 

 

 

 

 

 

 

\ Erosion (Wear Track)
'I
\
\ Target
\
I ‘ "-2—— J y—LL
l l

 

Figure 2-43 Lateral Composition Shifts Studies in TbFe alloy Sputtered Thin Films
[Murakami, 1990].

142

Spurrenmc new. (grams PER mm
0.

 

‘0
r

 

A!

'0‘ Ir 20° 30°
ANGLE OF PREFERRED EMSSION

 

 

Figure 2-44 The Correlation of Emission Angle with Sputtering Yield [Betz, 1970].

143

, ,100—

80—

‘m

 

 

 

 

-s\\\\‘0
&\

 

 

 

 

Fraction (%)

 

 

 

Light Medium Heavy
Mass

Figure 2-45 The Fraction of Sputter Particles That Reach the Substrate ['Tumer, 1992].

 

 
 
 

0

Binding Energy /eV 9.

(‘3‘

 

 

 

Atomic Muss lamu

Figure 2-46 A Plot Showing the Binding Energy and Mass of Range of Elements
Overlayed with Contour Lines that Indicate the Percentage of Atoms
Reaching the Substrate, which was Interpolated from Turner et al. Data on
Incident Flux Reduction at 3.5 mTorr. [Tumer, 1992].

145

12 EN)

 

 

 

 

 

1200400 400 _ g
100_ " Q. I I IIIIIII r rIrlu=
E \ i
10.__
. t:
1::
‘E _.
3 F-
D -
"=—
E
'1 t t Lllllll i Irriitll l Lllll _
.01 ,1 1 1o

pIPa)

Figure 2-47 Thermalization Distance as Function of Pressure for Various Atomic Masses
and Initial Energies [Westwood, 1978] (— , 5 eV) (nu , 1000 eV).

146

 

 

105

to3

 

L A l __.l. A

o ‘ too

  

l

‘200 ’ 300

Pressure Distance. Pa-mm.

Figure 2-48 The Average Energies as Function of Working Gas-Distance Product

[Somekh, 1984].

147

 
   

 

   
 

l - L L

1.00

 

 

A LLA‘

 

I
TARGET I “W yield)

Al (LS)
Cr IL?)
Cu (3.03
MoILZI
A0 (4.4)
Sn (2.2)

£— lntrinsic Resputtering, %

 

20—

Pd Pd

 

VI (0.0)
PI (2.0)
Au (3.6)

000000-60

IO“-

i---‘-‘ -------- I-

   

 

 

 

Ti f

'0—3
x

 

 

 

O
i-
o t.._--- -.......

 

Mtorget ' Mgos )

+M

ATOMIC MASS PARAMETER ( M
target gas

Figure 2-49 The Experiment Data of Intrinsic Resputtering as Function Atomic Mass
Parameter and the Resputtering Fractions Calculated for Ti, Ni, and Pd
[Hoffmam 1990].

148

 

‘ U I ' U V U V I V V V V f v y

" j Melons jl
08 r ah; .._...‘

 

 

 

 

 

0.6

 

 

 

 

0.4

 

0.2

 

 

 

 

Al partlal sputterlng yleld [atoms/Ion]

 

Atomic number of substrate

Figure 2-50 Static TRIM Simulations of Partial Sputtering Yield of a l-nm Thick Al
Layer as Function of Substrate Atomic Number for Various Incident Ion
Types [Berg, 1999].

149

500

Trans Temp, C

_. to 00 ->

o 8 o o o
O O O

Kleinherenbrink 91, '-°- wu‘w'ymfp-

Johnson. 95, Bulk MI
Zhu SMM94 Ms 0
PU. SPIE 94 Ap O
MOORE a VI. 92 AI. Mt
LO & wu. 01 At, Mf
QUANDT. 95. AI, MI 0
Lotkov 82. Ms

 

0

At% Pd

 

  

50

Figure 3-1 Literature Values for Transformation Temperatures of TiNiPd Alloys

700

600

500

400

Temperature, C

(N i+Pd)
Wu&Wayman, Ap, Mp
Scr Met 24 1574 1990

 

10 20 30

Atomic % Pd

Ap

Mp
Mf

p-subscript
above indicates
location of the
DSC peak. Start
and ﬁnish
temperatures
are estimated.

40 50 60

Figure 3-2 Lo-Wu-Wayman’s 1990 TiNiPd Transformation Temperature Data

150

 

Figure 3-3 Optical photomicrograph of Ti 53.5/Pd 25/Ni 21.5 at. % Target taken at 50X

- . t . .
~ .-.. ' >‘-.(‘{‘. A ,p _ .
t*w"g>_ a;

\-_ ’ Q i .
I ‘ g- ”f? . ~ .
{gem .5 region
A.""3¢.ﬁ ( ‘ _, o ‘

Voids

 

Figure 3-4 Optical photomicrograph of Ti 53.5/Pd 25/Ni 21.5 at. % Target taken at 500x

151

Ni/Pd

Ni/Pd
structure

Voids
Ti-Pd/N i
structure
Ti

region

 

Figure 3-6 SEM photomicrograph at 1200 X of the Ti 53.5/Pd 25/Ni 21.5 at. % Target

152

 

 

 

3-7 X-ray Map for Ti, Ni, and Pd of Target Pdl at 1000X

Figure

153

 

Figure 3-8 SEM Photomicrograph taken at 500x Overlayed with 550 Point Grid used for
Void Volume Fraction Calculation.

154

 

Figure 3-9 Kurt J. Lesker Sputter Deposition System

155

 

 

Glass
Slides

 

 

 

 

 

Copper
Tape-

 

 

 

 

   

 

Three Glass slides were centered over
the 2.5” square hole and adhered to
substrate holder with vacuum grade
,_..__,g copper tape. The glass slide substrates
I I were sandwiched between the substrate . \
. i I holder and the existing wafer platen. I j
I

 

 

 

 

 

Edges Spring
Clipped to
Existing
Chamber Wafer
Platen

 

 

 

 

 

 

 

Figure 3-10 Aluminum Substrate Holder

156

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

[ J
=
l l
l J
thne.
| l
r ' 1
F l
ﬁatﬁ
E :2

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 3-11 Torus 3M Magnetron Sputtering Gun Assembly

157

DARKSPACBW

FIXTURESCREW

PRESET GAP
MOUNTING HOLE
HOLD mWN SCREW

undersea) DOWN
TARGET

000nm PLA‘IBSCRBW

'O‘RING
WWLY

BASE I'LA‘IE
rowanmn new

'O’RINO

'O'RINO

CAM BODY

mm

FLAT WASHER

M'IOFLANOBW

MOUNTING POST

 

 

The substrate
positions were
rotated clockwise
with computer

controlled stepper
motor. The shutter
position was used
during target burn
it).

 

 

 

 

 

 

 
  
  

 

 

 

' ‘ The cut
Cut _. . ‘ comer of the
Comer ‘ __ ‘ substrate
‘ located the its
position
/" relative to the
Substrate mermmmc
ﬁlament.
Target / .

 

 

 

Anode

 

 

 

Figure 3-12 Triode Substrate and Gun Position Setup.

158

Ti

Sublimation
Pump

 

 

I”

85% Ti-Mo
Filaments

 
 
  
 
 
   
  
 

§\\\\\\\\\\\\\\\\\\\‘z

Ill

 

 

 

Sublimation
Power Supply

I

 

 

 

Variable voltage
supply with PID
temperature
controller that
measure temperature
at backside wafer

 

 
  
 

 

  

    

 

Hi-Vacuum
Chamber

Valve

Roughing
Valve

  
   
   
  
 
 
  

Rough
Pump

 

Argon Gas
Cylinder

 

 

Schematic Illustrations of
Temperature Measurement

and Heating Setup

 

 

Figure 3-13 Basic Schematic of the Thin Film Deposition Unit Used in Magnetron
Sputtered TN i and TiNiPd Thin Film Composition Experiments

159

Q4 5%
I,A:IOIVTI ”:3
Steel F‘Iote

1/4" ﬂeet LII’IC'E'

/ 1U

I (.1

 

I’IQI E’

,w-FTQIH; I. LLrI

c: L‘ ‘ .\ FL; I 0 \
_ , '\
.7:- iﬁ’“ o s O 0’ \
J" \
[7 \i‘
a; I \
II I \.
\x //’l x
\x ,v

 

/

 

 

I,‘ . r .
‘ I /.' Q 97; \\ ‘3“ . v, ‘
‘\ x/ \. \p_ ' ‘ a \ ‘\
/ (I II \ ‘* - ” ‘ \\ \\
/ II - \
,/ r I III/I \ \
// \\\\\: /// \\ \ ,,)(’—E'8 5
,1 \L / x \ g/ \
: l @0000 f/f
‘ 35.00 II 1" m 1 J \ p.750
1 ,, I / q 7 r- . ’ 0;: put . /./
p j I L how \I
- 71.3 LSUIJ ’
.4 [Exp burp
‘\ WIIL 5’9

 

k“: ““1 If‘“ I " I” ”r 9 I ‘3‘ Taped holes were used
:1 r; U m [I (I P (”I I In

' Cir” U 5mm LIL: gr, I: to mount wafer to

Wafer were held in place rotation feed through

by 1-2 oval shape clips that 90%

were bolted to wafer platen

 

 

 

 

 

 

 

Figure 3-14 5-inch Wafer Platen Used for Depositing TiNiPd Thin Films

160

Heat Flow (W/ g)

oba

obs

Pow

obo

.98

.9?»

 

30.300 Ongﬁwnoe

9890330 08w

.\

Dianne: wow—G

 

 

 

seamen
Simon

33sec
Meade

 

 

 

_ . —

3o moo So

4 . 4

sue to go

 

HoBRBHEo A09

3mg.» uLm Own 956 C80 8 030508 :5 OQwSENm—mo: HoBmeHEd m2. £00930: a Subsec—
Aiommr» 200 Mm 3m «50 8853 m" MOOSE;

161

 

 

 

Chart Recorder

Assembly used t -

 

 

 

 

 

 

 

 

 

 

 

 

   

Convecting
Box Fan

 

 

 

 

Quartz
Plates

Thin Film
Specimen

 

Figure 3-17 Schematic of Annealing Sample Setup

162

(222) m9

 

Figure 3-18 (a)TEM Microphotograph of Specimen (b)TEM Diffraction Pattern of B19
#03020001-AN0518-500 (100 kX) Martensite (Camera Length 1.2m)

(IIHHI‘C

 

Figure 3-19 (a)TEM Microphotograph of Specimen (b)TEM Diffraction Pattern of B2
#03020001-AN0518-500 at 300°C(50 kX) Austenite (Camera Length 1.2m)

163

Approximate
Target
Composition

 

Figure 4-1 Ni Content (at. %) Verses Magnetron Diode Sputtering Time ﬁ‘om 51 at% Ni/
48 at% Ti Target.

NIWﬂm‘. “)
400 m4 :1: no

‘5

 

 

Figure 4-2 Surface Plot of Ni Composition as Function Sputtering Time and Distance
from Center of Diode Cathode from 51 at% Ni/ 48 at% Ti Target.

165

 

 

 

 

\* Viv \tga \\<

Sputtering Time and Distance from Center of Diode Cathode from 51 at%

Figure 4-3 Rough Schematic of Smoothed surface plot of Ni Composition as Function
Ni/48 at% Ti Target.

166

 

Figure 44 Ni Content Verses Radial Distance from Diode Cathode Center for ﬁrst 12
Hours of Sputtering.

Amortimste
Cm of the
War Truck

Linear Fit

ThruAllData

 

Figure 4-5 Ni Content Verses Radial Distance ﬁom Diode Cathode Center for 15 to 24
Hours of Sputtering.

167

Approximate
Target
Composition

 

Figure 4—6 Ni Composition Verses Triode Sputtering Time ﬁ'om 51 at.% Ni/ 48 at% Ti
Target.

168

 

 

 

 

 

 

 

 

 

 

 

 

Figure 4-7 Ni Content Verses Radial Distance ﬁ'om Triode Cathode Center.

169

 

 

I
l
Axerage Center Linear it
Composition For Thru Data
Triode Films

+2 hrs
+4 hrs
——6 hrs
*8 hrs
+10 hrs
+12 hrs
—B—-14 hrs
+16 hrs
+18 hrs

 

 

Figure 4-8 Composition Verse Elapsed Target Wear Time of Ti 53.5%l Ni 21.5 %/ Pd
25% Target.

170

 

Figure 4-9 Ti Content of the Ti(Ni+Pd) Films as a Function of Deposition Temperature.

171

Linear Fit thru
all data points

 

Figure 4-10 Film Composition vs.Deposition Temperature for the Ti 53.5%l Ni 21.5 %/
Pd 25% Target.

172

 

Figure 4-11 Film Composition vs. Pressure-Distance Product for Thin Films Sputtered
ﬁ'om the Ti 53.5%l Ni 19.5%/ Pd 27% Target.

173

 

Figure 4-12 Composition vs. Working Gas Pressure for Thin Films Sputtered from the
Ti 53.5%/ Ni 19.5%/ Pd 27% Target.

174

 

 

 

 

 

 

/

 

y=11x- 167.5
I

 

 

 

 

 

 

 

 

Figure 4-13 Transformation Temperatures in As-deposited Thin Films vs. Palladium
Content.

175

1100

1000

900

800

700

9. .0.
2

2 500
2

0 400

g. I

0 300
I—

g 200

100

O

-100

Figure 4-14 M. Temperature as Function of Palladium Contents: Published Values
Compared to As-deposited and Annealed Thin Films[Lindquist, 1988].

I= MNKSO-x) de)

0= MNI(5o.x)Plx)
0= Donkerstoot and Van Vucht 1970

0 Ms tor at measured os- ,0
deposited thin film Pd at % .'

0 Ms tor tllm post annealed ,0
at 650 °C tor I hr ,o'
0 Ms tor ﬁlm post annealed c
at 600 °C tor 1 hr o'
I Ms tor ﬁlm post annealed 0
at 550 °C tor 1 hr '

°C'L'II'L'I.’ 42.";

8

10 20
Composition, x

176

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 4-15 M. Temperature as a Function of Titanium Content for As-Deposited Thin
Films.

177

 

Figure 446(0) M. Temperature Verse Deposition Temperature for As-deposited Thin
Films from 53.5% Ti/ 19.5% Nil 27% Pd Target.

 

1w

 

 

175

 

 

 

 

 

 

 

Martensite Shit Temperature (degrees C)
a;

 

 

 

 

 

 

 

 

500 525 550 575 600 625 650
Depodtlon Temperature (degrees C)

 

 

 

Figure 4-l6(b M. Temperature Verse Deposition Temperature for As-deposited Thin
Films from 53.5% Ti/ 21.5% Ni/ 25% Pd Target.

178

Ti Surface Diffuses

Ti—fich to the Weartrack:
sputter 4 edge composition
ﬂux e—becomes Ti-lean I

 

 

 

.7
‘/Ti diffusion I

wear track ﬂux

 

 

 

Figure 4-17 Schematic of Surface Diffusion in Magnetron Sputtering that Results in a

 

 

 

Non-Steady State Sputter Flux.
2 _ T 31.79 _ _ _ I _
1.9
'.<.
1.8 2
n.

 

1.7 \

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5 :
2. 1.6 \\ 1.6 i
P- 50
a 15 \ 7 1 RecollMass -
o - ‘*
5 1.4 \
:r . \

1.3 \

1.2

1.1

1

 

 

 

 

0 20 4O 60 80 100 120 140 160 180 200
Recoil Mass

 

 

 

Figure 4-18 The Effects of Recoil Mass on Preferential Sputtering of Alloy Calculated
from Matsunami et al. Sputter Yield Formula using Argon Projectiles.

179

tnltid Condltion: Surface Composition = Bulk Composition

 

EmdedCondlllm: SurfaoeWIn O spades.

TMWMWMWIUMMW.

 

Figure 4-19 Schematic of the Possible Surface Topography Developed in PM Targets at

Steady State.

180

 

 

 

 

 

 

IIIIIIII

 

02