 

THESIS

:1 003

This is to certify that the
dissertation entitled
PROCESSING AND CHARACTERIZATION OF HIGH-TEMPERATURE

NICKEL-TITANIUM-HAFNIUM SHAPE MEMORY THIN FILMS

presented by

I Jinping Zhang

has been accepted towards fulfillment
of the requirements for

 

Ph.D. degree in Materials Science

 

 

 

Major [ﬁessor

Date i! '2sz

MS U is an Afﬁrmative Action/Equal Opportunity Institution 0- 12771

 

LIBRAFW
Michigan State
University

 

 

 

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 cJCIRC/DateDuo.p65—p.15

 

PROCESSING AND CHARACTERIZATION OF HIGH-TEMPERATURE
NICKEL-TITANIUM-HAFNIUM SHAPE MEMORY THIN FILMS

By

Jinping Zhang

A DISSERTATION

Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Chemical Engineering and Materials Science

2002

ABSTRACT

PROCESSING AND CHARACTERIZATION OF HIGH-TEMPERATURE
NICKEL-TITANIUM-HAFNIUM SHAPE MEMORY THIN FILMS

By

Jinping Zhang

Near-equatomic TiNi thin ﬁlms are of interest as robust actuator materials in
microelectromechanical systems (MEMS) due to the high mechanical energy density
associated with reversible martensitic transformations. High-temperature (Ti+Hf)Ni
shape memory thin ﬁlms have been of particular interest because their high
transformation temperatures allows use at higher temperatures at which binary TiNi ﬁlms
become inoperative, and also because they can improve the operation frequency for
actuation. In this thesis, (Ti+Ht)Ni ﬁlms having different hafnium contents were
prepared by magnetron sputtering. Crystalline ﬁlms exhibiting martensitic
transformation were obtained either by depositing the ﬁlms at elevated temperatures or by
conducting a post-deposition anneal of as-sputtered amorphous ﬁlms. Phase
transformations, microstructure, mechanical properties and shape memory properties of
ﬁlms obtained by different procedures were investigated in order to ﬁnd the best way to
produce high-temperature shape memory ﬁlms.

In ﬁlms obtained by post-deposition annealing, the heating rates during annealing
were found to signiﬁcantly affect mechanical properties. Conventional vacuum
annealing with heating rates ranging from 5°C/min to 40°C/min resulted in brittle ﬁlms
even though the as-deposited amorphous ﬁlms were ductile, but rapid thermal annealing
(RTA) with a heating rate of 6000°C/min resulted in ductile ﬁlms having improved

martensitic transformation characteristics. Embrittlement of the ﬁlms annealed by

conventional annealing occurred at temperatures well below the crystallization
temperature. The RTA treated Ni49JTimel,7 ﬁlms had room-temperature ductility of
5.8%, fully recoverable strain of 1.7%, maximum recoverable strain of 1.9%, and
maximum stress for full strain recovery 200 MPa.

In-situ deposited crystalline Ni489Ti36beI45 ﬁlms demonstrated substantially
better properties than the RTA treated samples. The transformation temperatures were
higher, transformation hysteresis was lower, and most importantly, the mechanical
properties and shape-memory properties were signiﬁcantly improved. In-situ deposited
crystalline Ni48O9Ti366Hfl45 ﬁlms had room-temperature ductility of 8.3%, fully
recoverable strain of 2.8%, maximum recoverable strain of 3.6%, and maximum stress for
full strain recovery of 250 MPa. In addition, the transformation temperatures were
effectively increased as more Hf was substituted for Ti. These properties of in-situ
deposited crystalline Ni-Ti-Hf ﬁlms put them in a good position for industrial
applications where binary TiNi ﬁlms are not applicable because of their lower

transformation temperatures.

ACKNOWLEDGMENTS

I would like to express my profound gratitude to my major advisor, Professor
David Grummon, for his continuous guidance and support during this period of my
education. His careful reading of the preliminary drafts and critical comments helped in
many ways. Also I would like to thank other professors in my Ph.D. guidance
committee, Professor Mercouri Kanatzidis, Professor Jun Nogami, and Professor Donnie
Reinhard, for their invaluable suggestion and comments.

I am thankful to my employer, Delphi Corporation, for allowing me use the
materials obtained at work. Encouragement and support from my supervisors Gregory
Rasmussen and Rick Kerr are appreciated. Assistance from Mr. Brian Jones of Delphi
Delco Electronics in conducting the RTA experiments is acknowledged.

Special thanks are due to my family for their support and patience, which have

been a driving force for me to ﬁnish this work.

iv

TABLE OF CONTENTS

LIST OF TABLES .................................................................................. vii
LIST OF FIGURES ................................................................................ viii
KEY TO ABBREVIATIONS .................................................................... xiv
CHAPTER 1
INTRODUCTION .............................................................................................................. 1
CHAPTER 2
LITERATURE REVIEW ................................................................................................... 4
2.1. SHAPE MEMORY EFFECTS AND SHAPE MEMORY ALLOYS ........................................ 4
2.1.1. SME and Thermoelastic Martensitic Transformation ....................................... 5
2.1.2. Transformational Superelasticity ...................................................................... 8
2.1.3. Two-way Shape Memory Effect ....................................................................... 9
2.2. NEAR-EQUIATOMIC NI-TI SHAPE-MEMORY ALLOYS .............................................. 10
2.2.1. SME and Displacive Transformation in Near-equiatomic Ni-Ti .................... 11
2.2.2. Aging Effect .................................................................................................... 13
2.3. SHAPE-MEMORY ALLOYS WITH HIGH TRANSFORMATION TEMPERATURES ............ 17
2.3.1. Ti-Ni-Pd, Ti-Ni-Pt and Ti-Ni-Au alloys ......................................................... 19
2.3.2. Ni-Ti-Zr and Ni-Ti-Hf Alloys ......................................................................... 26
2.3.3. General Comparison of Different HTSMAS ................................................... 32
2.4. SHAPE-MEMORY ALLOY THIN FILMS ...................................................................... 33
2.4.1. Deposition of Thin Films by Sputtering .......................................................... 33
2.4.2. Sputtered Shape Memory Thin Film ............................................................... 39
2.4.3. High-Temperature Shape Memory Alloy Thin Films ..................................... 53
CHAPTER 3
EXPERIMENTAL PROCEDURE ................................................................................... 59
3.1. THIN FILM FABRICATION ........................................................................................ 59
3.1.1. Thin Film Deposition ...................................................................................... 60
3.1.2. Hot Pressed Powder Target ............................................................................. 61
3.1.3. Post-Deposition Annealing ............................................................................. 62
3.2. THIN FILM CHARACTERIZATION ............................................................................. 64
3.2.1. Transformation Temperatures ......................................................................... 64
3.2.2. Mechanical Property Measurement ................................................................. 65
3.2.3. Shape-Memory Property Evaluation ............................................................... 67
3.3. COMPOSITIONAL, STRUCTURAL AND MICROSTRUCTURAL ANALYSES .................... 71
3.3.1. Composition Analysis ..................................................................................... 71
3.3.2. Crystallographic Structure Analysis ................................................................ 72

3.3.3. Microstructure Observation ............................................................................. 72

CHAPTER 4
RESULTS AND DISCUSSION ....................................................................................... 74
4.1. FILMS CRYSTALLIZED BY POST-DEPOSITION ANNEALING ...................................... 75
4.1 . l. Crystallization by Conventional Isochronal Heating ...................................... 77
'4. l .2. Isothermal Annealing at Low Temperatures ................................................... 85
4.1.3. Isothermal Annealing at High Temperature via Different Ramp Rates .......... 87
4.1.4. Discussion on Ni-Ti-Hf Films Crystallized by Post-Deposition Annealing 93
4.2. IN-SITU CRYSTALLIZED NI-TI-HF FILMS ................................................................ 98
4.2.1. Phase Transformations .................................................................................... 99
4.2.4. Crystallographic and Microstructural Analysis ............................................. 101
4.2.2. Mechanical Properties ................................................................................... 103
4.2.3. Shape-Memory Properties ............................................................................. 110
4.2.4. Discussion on In-situ Crystallized Ni-Ti-Hf Films ....................................... 116

CHAPTER 5
CONCLUSIONS ............................................................................................................. 120
BIBLIOGRAPHY ........................................................................................................... 1 26

vi

LIST OF TABLES

Chapter 2.
Table 2.1. Selected properties of Ni-Ti SMA ........................................... 12

Chapter 3.
Table 3.1. Composition and porosity of the targets used to deposit

Ni-Ti-Hf SMA thin ﬁlms in this study .................................................... 62

Chapter 4.
Table 4.1. Target compositions vs. ﬁlm compositions .................................. 75

Table 4.2. Peak temperatures and transformation enthalpies of
the peaks on the DSC curves during isochronal annealing of
as-sputtered amorphous Ni49JTi362Hf14g ﬁlms at different heating rates ............ 78

Table 4.3. Transformation temperatures and enthalpies of
as-crystallized Ni49,1Ti36,2Hf14.7 ﬁlms after isochronal heating
at different rates .............................................................................. 81

Table 4.4. Comparison of room-temperature ductility of
as-deposited amorphous Ni49,1Ti36,2Hf14_7 samples before and
after isochronal heating from room temperature to 575°C ............................. 84

Table 4.5. Comparison of room-temperature ductility of the
as-deposited amorphous Ni49.1Ti36,2Hf14,-; ﬁlms before and
after 1-hour annealing at 300°C and 350°C ............................................. 86

Table 4.6. Martensitic transformation temperatures and

transformation enthalpies for Ni49_lTi36,2Hf14,-; samples

crystallized by isothermal annealing with different heating

rates to annealing temperatures ............................................................ 89

Table 4.7. Comparison of room-temperature ductility of as-deposited
amorphous Ni49_1Ti36,2Hf14_7 ﬁlms crystallized by isothermal
annealing via different ramp rates to the annealing temperatures ...................... 91

Table 4.8. Martensitic transformation temperatures vs. sample
composition of in-situ crystallized Ni-Ti-Hf SMA ﬁlms ................................ 100

Table 4.9. Room-temperature tensile testing results of In-situ
crystallized Ni-Ti-Hf ﬁlms with different hafnium content ............................ 105

vii

LIST OF FIGURES

Chapter 1.
Figure 1.1. Work per volume versus cycling frequency for various

microactuators [Krulevitch et a1. 1996] ............................................

Chapter 2.
Figure 2.1. Schematic showing physical property changes and

endothermic/exothermic effects during reversible martensitic

phase transformation ...................................................................

Figure 2.2. Theoretical recoverable strain of single-crystal Ni-Ti
SMA as functions of crystal direction developed by

(a) Saburi et a1. [1981], (b) Inoue et a1. [1996] .....................................

Figure 2.3. Equilibrium phase diagram of Ti-Ni alloy [Massalski 1987]......

Figure 2.4. Schematic showing the dependence of austenite yield stress
and reverse transformation peak temperature of Ni-Ti alloy on the

composition of the material [Hodgson and Brown 2000] ........................

Figure 2.5. Isothermal time-temperature-transformation (I 1'] )

diagram for Ti-52at.%Ni alloy [Nishida et al. 1986] ..............................

Figure 2.6. Effect of Pd and Hf composition on reverse martensitic
transformation temperature of NiTi-base ternary alloys summarized

from the literature .......................................................................

Figure 2.7. Effect of Hf composition on transformation temperature

of Ni-Ti-Hf alloys summarized from the literature ...............................

Figure 2.8. Effect of Ni composition on the transformation

temperature of Ni-Ti-Hf shape memory alloy .....................................

Figure 2.9. Deformation Behavior of Ni-Ti-Hf alloys at different

test temperatures [Olier et a1. 1995] .................................................

Figure 2.10. A microactuator structure proposed by Johnson et a1. [1994] . ..

Figure 2.11. A schematic showing realization of vertical cyclic

displacement from a horizontal component under an applied force ............

viii

. 138

139

140

141

. 142

.. 143

. 144

..145

. 146

. 147

148

..149

Figure 2.12. Shape-memory properties of sputtered Ni-Ti-Pd SMA
thin ﬁlm during thermal cycles under different applies stresses

[Miyazaki et al. 1995] ................................................................ 150
Chapter 3.

Figure 3.1. Schematic of the sputtering system used to deposit

Ni-Ti-Hf SMA thin ﬁlms in this study ............................................. 151

Figure 3 .2. Schematic of the apparatus for conducting heating
treatment for Ni-Ti-Hf thin ﬁlm .................................................... 152

Figure 3.3. Typical temperature proﬁle of sample in a rapid thermal
anneal treatment for crystallization of Ni-Ti-Hf thin ﬁlms ..................... 153

Figure 3.4. Schematic of the apparatus for measuring resistance of
the thin ﬁlm sample to determine phase transformation temperature

in this study ............................................................................ 154
Figure 3.5. Schematic of the experimental setup of semi-quantitatively
determining the ductility of Ni-Ti-Hf thin ﬁlms using bending test .......... 155
Figure 3.6. Schematic of the method measuring recoverability of a

pre-strain put to Ni-Ti-Hf martensite ..... A ......................................... 156
Figure 3.7. Schematic of the apparatus measuring shape memory

properties of Ni-Ti-Hf thin ﬁlm in constrained thermal cycles .................. 157
Figure 3.8. Typical hysteresis in a strain-temperature curve for

measuring shape memory properties ofNi-Ti-Hf SMA thin ﬁlms .............. 158
Chapter 4.

Figure 4.1. Composition uniformity of the thin ﬁlms deposited from

an 8”-diameter Ni43,5Ti3-;,5Hf14 target. ............................................. 159

Figure 4.2. Composition variation of ﬁlms at central area with
continuous wearing of the Ni43,5Ti35,5Hf16 target. ................................ 160

Figure 4.3. DSC curves of as-deposited amorphous Ni49_;Ti36_2Hf14,7
ﬁlms during isochronal heating from room temperature to 575°C at
different heating rates ................................................................ 161

Figure 4.4. Dependence of exothermic peak temperatures and
enthalpies on heating rate during heating as—deposited amorphous
Ni49_1Ti36,2Hf14_-, ﬁlms in DSC instrument ........................................... 162

Figure 4.5. Estimation of activation energy for crystallization
transformation and the transformation associated with the
low-temperature exothermic peak using Kissinger’s method ..................... 163

Figure 4.6. DSC curves Of N149,1Tl35.2Hf14.7 ﬁlms crystallized by
isochronal anneal at different heating rates showing reversible martensitic
transformation during thermal cycling between 50°C and 250°C ................ 164

Figure 4.7 . Martensitic transformation temperature of annealed
Ni49,1Ti36_2Hf14,7 ﬁlms as functions of heating rates during isochronal
heating for crystallization ............................................................. 165

Figure 4.8. Room-temperature XRD spectra of as-deposited
amorphous Ni49,.Ti36,2Hf14_7 ﬁlms and the samples crystallized
by isochronal heating to 575°C at different rates .................................. 166

Figure 4.9. TEM micrograph and SAED pattern of as-deposited
amorphous Ni49,1Ti36.2Hf14,7 thin ﬁlms aﬁer crystallized by
isochronal heating to 575°C at different rates ...................................... 167

Figure 4.10. Room-temperature XRD of as-deposited amorphous
Ni49,1Ti36,2Hf14,-; thin ﬁlms and those for ﬁhns after annealed at
300°C and 350°C for 1 hour, respectively .......................................... 169

Figure 4.11. TEM micrograph and SAED pattern of as-deposited
amorphous Ni49,1Ti36,2Hf14_7 ﬁlms and those for ﬁlms aﬁer annealed
at 300°C and 350°C for 1 hour, respectively ........................................ 170

Figure 4.12. DSC curves during thermal cycling between 50°C
and 250°C of as-deposited amorphous Ni49_1Ti36_2Hf14,7 ﬁlms after
crystallized by isothermal annealing at different ramp rates ...................... 171

Figure 4.13. Transformation temperatures and transformation
enthalpy of Ni49_1Ti36,2Hf14,7 ﬁlms isothermally annealed by
different procedures (conventional vacuum anneal and RTA) ................... 172

Figure 4.14. Room-temperature XRD spectra of as-deposited

amorphous Ni49,1Ti36,2Hf14,7 thin ﬁlms after crystallized by

isothermal annealing via different ramp rates ﬁ'om room

temperature to the annealing temperature ........................................... 173

Figure 4.15. TEM micrograph and SAED patterns of as-deposited

amorphous Ni49,.Ti35,2Hf14_-; ﬁhns after crystallized by isothermal

annealing via different ramp rates from room temperature to the

annealing temperatures ................................................................ 174

Figure 4.16. Room-temperature stress-strain curves of as-deposited
amorphous Ni49,1Ti36,2Hf.4,7 ﬁlms after rapid thermal annealing at
650°C for 20 seconds with heating rate of 6000°C/min ............................. 178

Figure 4.17. Schematic strain-temperature curve of RTA crystallized
Ni49,1Ti36_2Hf14,7 ﬁlms during constrained thermal cycling starting
from room temperature (fully martensitic condition) ............................. 179

Figure 4.18. Strain-temperature curves of as-deposited amorphous
Ni49,1Ti36_2Hf14_7 thin ﬁlms crystallized by RTA anneal at 650°C for
20 seconds during thermal cycling under different loads ......................... 180

Figure 4.19. Dependence of strain induced by cooling from austenite

to martensite and that recovered by heating from martensite to

austenite of RTA treated as-deposited amorphous Ni49,1Ti36_2Hf14,7

SMA thin ﬁlms on applied stress during thermal cycling ........................ 171

Figure 4.20. DSC curves of in-situ crystallized Ni-Ti-Hf thin ﬁlms

with different hafnium content during thermal cycling between
50°C and 300°C ....................................................................... 182

Figure 4.21. Dependence of transformation temperatures on haﬁrium
content of in-situ crystallized Ni-Ti-Hf ﬁlms ..................................... 184

Figure 4.22. Room-temperature XRD of in-situ crystallized
N143,9Tl36,6Hf14.5 SMA ﬁlm ............................................................ 185

Figure 4.23. TEM micrograph and SAED pattern of in-situ
crystallized Ni43,9Ti36,6Hf14.5 SMA ﬁlms taken at ambient temperature ........ 186

Figure 4.24. Room-temperature stress-strain curves of in-situ
crystallized T1505Nl495 SMA thin ﬁlms ............................................. 189

Figure 4.25. Stress-strain curves of in-situ crystallized binary
Ti50,5Ni49,5 SMA ﬁlms at different temperatures ................................. 190

Figure 4.26. Room-temperature stress-strain curves of in-situ
crystallized Ni-Ti-Hf thin ﬁlms with different hafnitun content ............... 191

xi

Figure 4.27. Dependence of room-temperature yield stress,
fracture stress and ultimate elongation on hafnium content
in in-situ crystallized Ni-Ti-Hf ﬁlms ................................................ 192

Figure 4.28. Stress-strain curves of in-situ crystallized
Ni43,9Ti36,6Hf14,5 SMA thin ﬁlms at different test temperatures
using cooling-down routine. ......................................................... 193

Figure 4.29. Dependence of yield stress of in-situ crystallized
Ni43,9Ti36_6Hf14.5 SMA on tensile test temperatures between 200°C
and 140°C using cool-down procedure .............................................. 194

Figure 4.30. Dependence of ultimate elongation and yield
stress on test temperature of in-situ crystallized Ni43_9Ti36,6Hf14,5
SMA thin ﬁlms ......................................................................... 195

Figure 4.31. Stress-strain curves of in-situ crystallized Ni43,9Ti36_6Hf14,5
thin ﬁlms elongated at room-temperature upon loading, unloading and
subsequent heating ...................................................................... 196

Figure 4.32. Recoverable strain and recovery rate of in-situ
crystallized Ni43_9Ti36_6Hf14,5 ﬁlms with different amount of
pre-strain introduced at room temperature ......................................... 197

Figure 4.33. Stress-strain curves of in-situ crystallized Ni43,9Ti36,6Hf14,5
thin ﬁlms elongated at different temperatures upon loading, unloading
and subsequent heating. The strain upon unloading was ﬁxed at about
3%, and the direct-heating routine was used ........................................ 198

Figure 4.34. Dependence of recoverable strain and recovery rate on
deformation temperature Of in-situ crystallized Ni43,9Ti36,6Hf14,5
martensite ﬁhns at a ﬁxed pre-strain of about 3% ................................ 199

Figure 4.35. Stress-strain curves of in-situ crystallized Ni4g,9Ti36,6Hf14,5
elongated at different temperatures (in the initially austenitic condition)

during loading and unloading. Before test, samples were heated to 250°C

and then cooled down to various test temperatures ............................... 200

Figure 4.36. Dependence of isothermal strain recovery on deformation
temperature for in-situ crystallized Ni43,9Ti36_6Hf14_5 SMA thin ﬁlm
at a ﬁxed total tensile strain of 4% ................................................... 201

Figure 4.37. Stress-strain curves of in-situ crystallized Ni43,9Ti36,6Hf14_5

thin ﬁlms loaded and unloaded at 190°C for different amounts of
total strain .............................................................................. 202

xii

Figure 4.38. Dependence of recoverable strain and recovery rate on
initial strain of in-situ crystallized Ni43,9Ti36,6Hf14_5 SMA thin ﬁlms
when elongation was done at 190°C .................................................. 203

Figure 4.39. Strain-temperature curves of in-situ crystallized
Ni4g,9Ti36,6Hf14_5 thin ﬁlm during thermal cycling under various
constant stresses ........................................................................ 204

Figure 4.40. Dependence of strain induced by cooling form

austenite to martensite and that recovered by heating from

martensite to austenite of in-situ crystallized Ni43_9Ti36,6Hf14,5

SMA thin ﬁlms on biasing stress during thermal cycling ......................... 205

Figure 4.41. Dependence of transformation temperatures on applied

stress during thermal cycling of in-situ crystallized Ni43,9Ti36,6I-If14,5
SMA thin ﬁlms ......................................................................... 206

xiii

KEY TO ABBREVIATIONS

A ....................................... Austenite phase
Af ....................................... Austenite ﬁnish temperature
Ap ...................................... Austenite transformation peak temperature
As ...................................... Austenite start temperature
CTE ................................... Coefficient of thermal expansion
DSC ................................... Differential scanning calorimetry
EDS ................................... Energy-dispersive spectrum
HTSMA .............................. High-temperature shape memory alloy(s)
LVDT ................................. Linear voltage-displacement transducer
M ...................................... Martensite phase
' Mf ..................................... Martensite ﬁnish temperature
Mp ..................................... Martensite transformation peak temperature
Ms .................................... Martensite start temperature
RTA .................................. Rapid thermal anneal
SAED ................................ Selected-area electron diffraction
SE .................................... Superelasticity (Pseudoelasticity)
SEM ................................. Scanning electron microscope (microscopy)
SIM ................................... Stress-induced martensite (martensitic transformation)
SMA ............................................ Shape memory alloy(s)
SME ............................................ Shape memory effect(s)
TEM ................................. Transmission electron microscope (microscopy)
TTR .................................. Transformation temperature(s)
TIT diagram ....................... Time-temperature-transformation diagram
TWSME ..................................... Two-way shape memory effect
XRD ................................. X-ray diffraction

xiv

Chapter 1
INTRODUCTION

Shape Memory Alloys (SMAS) undergo a reversible displacive phase
transformation between a low-temperature martensite (M) phase and a high-temperature
austenite (A) phase [Wayman and Duerig 1990]. Associated with this reversible phase
transformation are the shape memory effect (SME) and transformational superelasticity,
the two most important properties of SMAS. SME represents the ability to recover a
large amount of strain imparted to the low-temperature phase when the alloys are heated
to the hi gh-temperature phase. Transformational superelasticity represents isothermal
recovery of strain caused by stress-induced A —-) M transformation when the applied
stress is removed. Alloys such as In-Tl, Cu-Zn-Al, Cu-Al-Ni, Au—Cd, Ni-Ti and some
ferrous materials have been found to be shape memory alloys.

Near-equiatomic Ni-Ti alloys are the most economically successful shape
memory alloys developed so far due to their large recoverable strain (up to 8% in uniaxial
mode), large recovery stress (up to 700 MPa), excellent corrosion resistance, and
excellent fatigue resistance [Wasilewski 1972, Beuhler and Wang 1968, Mercier and
Melton 1979, Miyazaki, Sugaya and Otsuka 1989, Ford and White 1996]. Therefore, a
vast majority of current applications related to the shape memory effect and
transformational superelasticity makes use of Ni-Ti alloys. Also important is the
biocompatibility of Ni-Ti alloys with human body. In fact, Ni-Ti alloys have found as
many medical applications as industrial ones [Russell, Hodgson and Basin 1997].

Although they have superior mechanical and shape-memory properties, near-
equiatomic Ni-Ti alloys have phase transformation temperature (Af temperature) below

1 10°C, which is too low for many automotive underhood and aerospace related

applications. Therefore, development of alloys having high transformation temperatures
has been actively pursued [Humbeeck 1999]. The most promising are Ni-Ti based
psuedO-binary alloy systems, which are formed by substituting Au, Pd or Pt for Ni, or
substituting Zr or Hf for Ti in the CsCl-type structure of B-NiTi. It has been found that
the transformation temperatures of the psuedO-binary alloys could reach as high as 600°C
depending on the atomic percentage of the ternary elements, but the mechanical and
Shape-memory properties were often degraded by the third element. Among those alloy
systems, Ni-Ti-Hf alloys showed better shape-memory properties than Ni-Ti-Zr alloys
and were more cost competitive than Ni-Ti-Pt or Ni-Ti-Pd alloys [Humbeeck 1999, Wu
and Ma 2000].

Shape-memory thin ﬁlms have been attracting widespread interest due to their
potential usefulness as nricroactuator materials in microelectromechanical systems in
which high energy-density output is desired. In fact, Ni-Ti binary shape-memory thin
ﬁlms have an energy density at least one order of magnitude larger than any other
microactuator materials currently available (see Figure 1.1), and it is the only shape-
memory alloy thin ﬁlm that has been successfully developed. The mechanical and shape-
memory properties of Ni-Ti alloy thin ﬁlms are comparable or superior to those of their
bulk alloy counterparts [Miyazaki and Ishida 1999]. However, the application of Ni-Ti
shape-memory thin ﬁlms has been limited to relatively low temperatures, and the
response time is too long for many applications. Development of high-temperature
Shape-memory thin ﬁlms has been considered as a measure to overcome these drawbacks.

Although few and far between, attempts to fabricate shape-memory thin ﬁlms
with high transformation temperatures have been made, and Ni-Ti-Pd or Ni-Ti-Hf thin
ﬁlms have shown some promise [Johnson, Martynov and Minners 1995, Johnson,
Martynov and Shahoian 1995, Miyazaki, Hashinoga, Yumikura, Hirikawa, Ueki and
Ishida 1995, Quandt, Halene, Holleck, Feit, Kohl and SchloBmacher 1995]. From the

limited data on Ni-Ti-Hf thin ﬁlms, it has been demonstrated that high transformation

temperature can be obtained through composition adjustment, but evaluation of
mechanical and shape-memory properties of Ni-Ti-Hf SMA thin ﬁhn is uncharted
territory. No literature exists on optimization of thin ﬁlm properties by appropriate
processing strategies.

In order to acquire Shape-memory properties, binary Ni-Ti thin ﬁlms must be
deposited at high substrate temperature [Ikuta, Fujita, Ikeda and Yamashita 1990, Gisser,
Busch, Johnson and Ellis 1992, Hou and Grummon 1995], or the thin ﬁhns must be
annealed at high temperature [Miyazaki and Ishida 1994], because alloys sputtered onto
substrate at temperatures below 350°C are metallic glasses. Thin ﬁlm properties may be
different according to whether in-situ deposition or post-deposition annealing is used.
For the post-deposition annealing treatment, annealing time and annealing temperature
affect the properties profoundly. This thesis work presents results on fabrication of Ni-
Ti-Hf thin ﬁlms having the shape-memory effect, and shows how process parameters,
especially heating rate during annealing, affect their mechanical and Shape-memory
properties. The result will be useful as a reference for the fabrication of high-temperature
shape memory thin ﬁlms.

The organization of this thesis is as follows: A review on shape-memory alloys,
especially high-temperature shape memory alloys, and thin ﬁlms is presented in the
Literature Survey chapter; experimental techniques and procedures are detailed in the
Experimental Methods chapter; the experimental results on mechanical property
evaluation and microstructural investigation of both annealed and in-situ deposited ﬁlms
are presented and discussed in the Results and Discussion chapter; the major results of

this research work are summarized in the Conclusions chapter.

Chapter 2
LITERATURE REVIEW

In this chapter, a literature survey is presented on the shape memory effect and
shape-memory alloys, particularly near equiatomic Ni-Ti alloys. Ni-Ti-base ternary
shape-memory alloys having high transformation temperatures are emphasized. Also, the
fabrication and evaluation of Ni-Ti shape memory thin ﬁlms and Ni-Ti-X (X = Hf, Zr, Pt,
Pd, Au) high-temperature thin ﬁlms are reviewed in detail. Applications of Ni-Ti shape

memory thin ﬁlms are discussed.

2.1. Shape Memory Effects and Shape Memory Alloys

The shape memory effect (SME) represents the ability of a certain group of
materials to recover, upon heating, the plastic deformation sustained at low temperature.
It was ﬁrst found in 1951 by Chang and Read in Au-Cd alloys [Chang and Read 1951].
Since then, a quite number of materials, including metals [Arbuzova and Khandros 1964,
Smialek and Hehemann 1973, Schroeder and Wayman 1979, Delaey, Humbeeck,
Chandrasederan, Janssen, Andrade and Mwamba, 1981, Miyazaki and Otsuka 1989],
ceramics [Muddler and Trail] 1994], and even polymers [Hayashi 1993], have been found
to be shape memory materials. Metallic shape memory materials, which are also called
Shape Memory Alloys (SMAS) display the greatest SME, and the most important SMAS
are near-equiatomic Ni-Ti and Cu-base alloys including Cu-Zn-Al and Cu-Al-Ni [Miura,
Meada and Nakanishi 1974, Sakarnoto, Otsuka and Shimizu 1977]. Between these two
alloy systems, Ni-Ti has signiﬁcantly better mechanical and shape-memory properties
and greater thermal stability, while Cu-base alloys hold some economic advantage for

heavy sections.

2.1.1. SME and Therrnoelastic Martensitic Transformation

The shape memory effect originates from the martensitic transformation, a
displacive transformation like that in carbon steel [Reed-Hill and Abbraschian 1992].
But the martensitic transformation in SMAS differentiates itself from the one in steel by
its reversibility and the highly mobile interfaces [Wayman 1992]. As a result, the
transformation in SMAS takes place thennoelastically, meaning that it is thermally
reversible'. The high-temperature phase is called the ‘parent’ or austenite phase, and the
low-temperature phase is called the ‘daughter’ or martensite phase. Because of the
displacive nature of the phase transformation, atoms do not need to undergo long-range
migration (in contrast to diffusive transformations), and the transformation is independent
of time. Rather, the fraction transformed depends on the temperature of the material.2
Displacive transformations are thus called ‘athermal’ to indicate that they do not depend

on thermal activation.

2.1.1.1. Transformation Temperatures. On cooling, the austenite phase starts to
transform to martensite at M, (the martensitic start temperature), and it becomes fully
martensitic at Mf (the martensite ﬁnish temperature). On heating, the fully martensitic
material starts to transform to austenite As (the austenite start temperature) and becomes
fully austenitic at Af (the austenite ﬁnish temperature). Transformational thermal

hysteresis exists between the forward transformation (A -) M) and reverse transformation

 

' Although being widely used, the term “thermoelastic” is not very descriptive since both stress and
temperature may induce the transformation. Another term “thermotractive” offered by Grummon [Chang
and Grummon, 1997] may be a better choice for the description of SMA alloys.

2 Since the transformation is not time dependent, one may think that it should take place as fast as wanted.
In reality, however, heating transfer between the sample and environment may A -) M limit the
transformation Speed. Transformational latent heat may further slow down the temperature change of the
sample.

(M —-) A), which means that the A, temperature is always higher than the Mf temperature,
and the M8 temperature is always lower than the Af temperature.

The martensitic transformation is a ﬁrst-order phase transformation. It is
accompanied by a discontinuous non-linear change in physical properties such as electric
resistivity and heat capacity. Figure 2.1 schematically illustrates these changes with
respect to the transformation temperatures. The martensitic transformation is also
accompanied by endothermic and exothermic effects, and Differential Scanning
Calorimetry (DSC) is the most common means to determine the transformation
temperatures and enthalpies. The transformation temperatures are determined by the
tangent-intercept method. In the DSC curve, the exothermic peak during cooling
corresponds to martensitic transformation and the peak temperature is deﬁned as Mp; the
endothermic peak during heating corresponds to reverse martensitic transformation and

the peak temperature is deﬁned as Av.

2.1.1.2. Thermoelastic Martensite. On cooling, the austenite phase of SMAS
transforms into martensite with very little volume change (<0.5%) but with large local
shear strains. The martensite may be composed of a number of variants that have same
crystal structure but different orientations, which is possible because the martensite
structure has lower symmetry that the austenite. Although the structures of the austenite
phase and martensite are different, martensitic transformation does not result in
signiﬁcant macroscopic shape change if no stress is applied. This is because the
martensite variants are accommodated with each other so that lattice distortions are
canceled out on macroscopic scale. The interfaces between martensite variants are
usually twin boundaries, and the movement of those boundaries results in rearrangement
of martensite variants.

The shape memory effect results from the unique plastic deformation mechanism

available to the martensite, i. e. , the reorientation of martensitic variants effected by

applied stresses, involving preferential growth of some variants at expense of the others
[Otsuka and Shinrizu 1986]. Ultimately, only one variant may survive, and the surviving
variant is the one that gives rise to maximum transformational shear strain with respect to
the orientation of the applied stress. The strain appears to be plastic because there is no
driving force for the material to return back the original conﬁguration of martensite
variants. However, if the deformed material is heated to the Af temperature, the
martensite will transform into austenite. Due to the displacive nature of the
transformation and restriction of the crystal orientation relationship between austenite and
martensite imposed by the strongly ordered structure, when the reverse transformation is
completed on heating, atoms return to their original positions, thereby the original Shape

is recovered.

2.1.1.3. Recoverable Strain and Recovery Stress. For martensite in shape-memory
alloys, there exists a limit of plastic strain that can be recovered upon heating to austenite,
which is called maximum recoverable strain. Deformation beyond the maximum
recoverable strain may induce permanent strain by dislocation slip, as occurs in ordinary
metals. However, a smaller deformation does not guarantee a full recovery upon heating
to austenite. This is because slip may occur at the same time or before martensite
variants reorient. In order to induce recoverable strain and avoid non-recoverable slip
deformation, the stress applied must be higher than the critical stress for inducing
rearrangement of martensite variants, but lower than the critical stress for slip in the
martensite.

In another consideration, a great deal of stress can be generated against a physical
constraint when a constrained SMA recovers from a deformation imparted to the
martensite. This stress is called recovery stress. Apparently, this stress is also associated
with the critical stress for slip. How to increase the critical stress for slip and decrease

that for martensite reorientation is an interesting issue. In this respect, control of the

precipitation of non-equiatomic intennetallic phases (such TizNi or Ti3Ni, in NiTi SMAS)

is of importance and will be discussed further later.

2.1.2. Transformational Superelasticity

2.1.2.1. Stress Induced Martensitic Transformation. Martensitic transfonnation can
be induced thermally or induced by stress. This is, even without decreasing the
temperature of an austenite down to Ms temperature, martensitic transformation may still
occur with the help of an applied stress. This is because the applied stress can increase
the martensitic transformation temperature. The effect can be described by a modiﬁed
version of the well-known Clapeyron equation [Wollants, DeBonte and Roos 1979,
Wollants, Roos and Delaey 1993]:

ZZ-__A_H_
dT T-V-Ac

(2.1)

where 0', T, AH, V and Ac are the stress, equilibrium transformation temperature,
transformation enthalpy, molar volume of the material, and resolved transformation strain
in the direction of applied stress, respectively. Because of the multiplicity of shear
variants, and the sign of each parameter, the ratio is positive for common SMAS.
Therefore, Ms, M,, A, and A, temperatures rise with the absolute value of the applied
stress. Martensite is stress induced in an otherwise stable austenite when the stress raises

the Ms to the ambient temperature.

2.1.2.2. Transformational Superelasticity. While the thermally induced martensite
is self-accommodating and no signiﬁcant macroscopic deformation occurs, the martensite
formed by stress tends to favor particular martensite variants that give the maximum

transformational strain in the stress axis. On loading, a macroscopic transformational

strain is thus produced.

The unloading behavior of the SMA material is determined by the deformation
temperature. If the deformation temperature is below the A, temperature, i. e. , an
austenite specimen is held at a temperature below the A, but above M,, the plastic strain
remains after unloading, and it can be recovered only when the material is heated to
above A, temperature. However, if the deformation temperature is above A, temperature,
the reverse transformation is automatic upon unloading (with the same hysteresis), and
isothermal recovery of the strain occurs. The usefulness of transformational
superelasticity is not only because the recoverable strain is large, but also because there is
very little strain hardening.

It is known that slip deformation mechanisms produce permanent strain and
should be avoided in SMA applications. In order to ensure that the strain is caused by
stress induced transformation instead of slip, the stress for inducing martensitic
transformation has to be lower than the critical stress for slip in the austenite. While the
stress for inducing martensitic transformation increases with deformation temperature, as
described in Clapeyron’s equation, the critical stress for slip decreases with an increase of
deformation temperature. For a certain SMA, there exists a characteristic temperature at
which those two characteristic stresses are equal, and this temperature, called Md, deﬁnes

the maximum temperature for stress inducing a transformational strain.

2.1.3. Two-way Shape Memory Effect

Normally, shape-memory materials have the ability to restore its high-temperature
Shape upon heating, but can not automatically restore to the low-temperature
conﬁguration upon cooling. In many applications, cyclic displacement is of interest, and

it can be easily realized by using biasing forces with shape memory alloys.

2.1.3.1. Cyclic Displacement by Biasing Stress. External biasing stress such as that
implemented by a biasing spring can be used to achieve cyclic displacement in a SMA-
spring assembly [Ditrnan and Bergman 1994]. The temperature change must cover both
forward and reverse transformation temperatures. Most critically, the biasing stress
should be higher than the critical stress for inducing martensitic transformation, but lower
than the critical stress for slip in either martensite or austenite. By doing so, the biasing
stress can induce a relatively large deformation in the martensite, but can cause only a
negligible elastic deformation of the austenite. As temperature changes, cyclic

displacement is generated between martensite and austenite.

2.1.3.2. Intrinsic T wo-way SME. Even without external biasing stress, cyclic
displacement is sometimes still achievable. This is called two-way shape-memory effect,
and it is actually caused by the internal biasing stress from defects such as dislocations
and precipitate interfaces [Schroeder and Wayman 1977, Perkins and Sponholz 1984,
Nishida and Honma 1984, Liu and McCormick 1990]. Under such conditions, martensite
formed upon cooling is no longer fully self-accommodating, but adopts a preferred
conﬁguration with regard to variant distribution. As a consequence, cyclic shape change

results on cooling (to form martensite) as well as heating (to form austenite).

2.2. Near-equiatomic Ni-Ti Shape-Memory Alloys

Near equiatomic NiTi alloy was found to have SME by Beuler and his coworkers
in 1961 [Beuhler and Wiley 1961, Buehler and Wang 1968]. Since then, there has been a
lot Of success in research and development of this material. Ni-Ti has become the most
important of the shape-memory alloys. It has superior overall shape-memory properties,
the raw material is not prohibitively expensive and SOphisticated fabrication techniques
have been developed. Presently the overwhehning majority of SMA applications use Ni-

Ti materials.

10

Table 2.1 lists selected properties of polycrystalline NiTi alloy and it shows that
the alloy has excellent mechanical and Shape memory properties [Hodgson and Brown
2000]. For instance, one-time recoverable strain is up to 8%, stress for fully strain
recovery is up to 700 MPa, and room-temperature ultimate elongation is up to 50%. The
martensitic transformation temperatures, which are sensitive to composition, can be
varied from —200°C to +110°C (for A,). In addition, its excellent corrosion resistance
makes it viable for working in some harsh environments, and its good biocompatibility

allows many clinical applications.

2.2.1. SME and Displacive Transformation in Near-equiatomic Ni-Ti

The martensitic transformation in Ni-Ti involves the transformation between the
hi gh-temperature austenite phase and the low-temperature martensite phase.3 The former
has the CsCl-type structure (ordered b.c.c., ,6 - phase or B2 phase) with lattice parameter
a = 0.3005 ~ 0.3015 nm; the latter has a monoclinic structure (B19’ phase) with lattice
parameters a = 0.2889 nm, b = 0.4120 nm, e = 0.4622 nm and 7= 98.6° [Matsumoto,
Miyazaki, Otuka and Tamira 1987]. According to the lattice correspondence of
martensitic transformation, there are 12 different martensite orientations that could
possibly form from a single austenite grain, i. e., there are possible 12 martensite lattice-
correspondence variants [Otsuka, Sawarnura and Shimizu 1971].

Crystallographic information on the martensitic transformation in Ni-Ti makes it
possible to calculate the maximum recoverable strain theoretically. The single-variant
model proposed by Saburi and Nenno [1981] predicts that the maximum recoverable

strain is 11% when the uniaxial stress is applied along the [111]B2 direction and the least

 

3 In addition, another displacive transformation, the R-phase transformation, may occur prior to martensitic
transformation in Ni-Ti SMA. Shape memory effect associated with R-phase transformation has also been
observed but the transfonnational strain is generally limited within 1%. However, the R-phase
transformation has small hysteresis, which is of interest in certain applications such as fast actuation
[Miyazaki et a1. 1986]. Since the R—phase transformation does not occur in Ni-Ti-Hf alloys with Hf over
10at.%, it is not discussed further here.

11

Table 2.1. Selected Properties of NiTi SMA
[after Hodgson and Brown 2000]

Transformation Properties

 

Transformation Temperatures ........................................................................... -200 to +110 °C
Latent Heat of Transformation .................................................................................... 5.78 cal/g
Transformation Strain for Polycrystalline Material
for 1 cycle ............... - - - ................................................... max 8%
for 100 cycles ................................................................................................................................ 6%
for 100,000 cycles ......................................................................................................................... 4%
Hysteresis .................................................................................................................. 30 to 50 °C
Physical Properties
Melting Point .................................................................................................................. 1300 °C
Density ...................................................................................................................... 6.45 g/cm3
Thermal Conductivity
austenite ....................................................................................................................... 0.18 W/cm-°C
martensite ................................................................................................... 0.08618 W/cm-°C
Coefﬁcient of Thermal Expansion
austenite .................................................................................................................... 11.0 x 10'°/°C
martensite .................................................................................................................. 6.6 x lO‘PC
Speciﬁc Heat ........................................................................................................ 0.20 cal/g - °C
Corrosion Resistance .................................................................................................... excellent

Electrical and Magnetic Properties

 

Resistivity

austenite ........................................................................................................................ ~ 100 tin-cm

martensite ........................................................................................................................ ~ 80 uQ-cm
Magnetic Permeability ...................................................................................................... <1.002
Magnetic Susceptibility ..................................................................................... 3.0 x 10° emu/g

Mechanical Properties

Young’s Modulus (highly nonlinear with temperature)

austenite .................................................................................................................................. 83 GPa

martensite ....................................................................................................................... 28 ~ 41 GPa
Yield Strength

austenite ....................................................................................................... 195 to 695 MPa

martensite ...................................................................................................... 70 to 140 MPa
Ultimate Tensile Strength

fully annealed .................................................................................. 895 MPa

work hardened ................................................................................................................... 1900 MPa
Poisson’s Ratio ..................................................................................................................... 0.33
Elongation at Failure

fully annealed ..................................................................................................................... 25 to 50%

work hardened ..................................................................................................................... 5 to 10%
Hot Workability .................................................................................................................. good
Cold Workability .............................................................. difﬁcult due to rapid work hardening
Machinability ........................................................... difﬁcult, abrasive techniques are preferred

12

recoverable strain was 2% when the stress axis was [100]B2 (Figure 2.2a). The
recoverable strain for a randomly oriented polycrystalline NiTi can be estimated at 8.4%.
The multi-variant model proposed by Inoue et al. [Inoue, Miwa and Inakazu 1996]
yielded similar results for single crystal (Figure 2.2b), but the merit of Inoue’s model is
its capability to take the texture information, i. e. , orientation distribution ﬁmction, into
consideration.

Excellent superelasticity associated with stress induced martensitic transformation
was also demonstrated in near-equiatomic Ni-Ti [Miyazaki, Imai, Otsuka and Suzuki
1981]. Ltiders-like deformation behaviors were observed when the material was
deformed at temperatures above A,, indicating strain hardening was almost negligible
once the martensite started to form under inﬂuence of the applied stress. The relationship
between the tensile-test temperature and the critical stress for inducing martensitic
transformation obeys Clapeyron’s equation. For polycrystalline NiTi, the value of do/dT
is about 7 MPa/K, meaning the M, temperature increases about 14 to 15°C for every 100

MPa increment in applied uniaxial stress [Hodgson and Brown 2000].

2.2.2. Aging Effect

Thermal treatments are usually adopted to modify the transformation
temperatures, mechanical properties and shape memory properties of near-equiatomic Ni-
Ti SMAS. While alloys with Off-stoichiometric composition are prone to precipitation
reactions during aging, defect structures formed by cold working may also undergo
reconﬁguration. Such microstructural evolution has a profound effect on the properties

of the material.

2.2.2.1. Second Phases in Ti-rich Ni-Ti. The second phase in Ti-rich Ni-Ti alloys is
relatively simple. The phase diagram by Massalski indicates that the solubility of excess

Ti into ,6 - NiTi is less than 0.5at.% (reaching a maximum at 984 °C, see Figure 2.3).

13

Therefore, when Ni-Ti alloys have Ti-rich composition, TizNi precipitates are almost
inevitable.‘ TizNi phases have an f.c.c. structure with lattice parameters a = 1.127 nm ~
1.132 nm [Rostoker 1952, Yurko, Barton and Parr 1959, Mueller and Knott 1963]. They
have particulate morphology, and preferentially appear near grain boundaries.

Transformation temperatures of Ti-rich melt-solidiﬁed Ni-Ti alloy are relatively
consistent because the matrix composition is relatively consistent. In addition, the Ti-rich
material has higher transformation temperature than Ni-rich one. However, due to second
phases around grain boundaries, the mechanical properties of Ti-rich alloys solidiﬁed
from the melt usually are unsatisfactory, especially for alloys with large TizNi particles.
Figure 2.4 shows the relationship between the alloy composition and austenite yield stress
of Ni-Ti alloys. It indicates that the more Ti in the composition is, the less the austenite
yield stress will be. This will, in turn, limit the maximum stress during constrained

recovery.

2.2.2.2. Diﬁ‘usive Transformations in Ni-rich Ni-T i. Second-phase formation is
more complicated in Ni-rich Ni-Ti alloys than in Ti-rich ones. According to the
equilibrium phase diagram, solubility of excess Ni in ,B-NiTi is strongly temperature
dependent. Ni solubility decreases from about 7at.% at 1118°C to zero at 63 0°C.
Therefore, it is easy to obtain over-saturated nickel in the matrix by quenching the
material since excess Ni is readily solutionized. Although Massalski suggested that the
only possible second phase was Ni3Ti in Ni-rich Ni-Ti alloys, more detailed study by
Nishida et al. [Nishida, Wayman and Honma 1986] using TIT (Temperature - Time -
Transformation) diagram (Figure 2.5) revealed that there may be other metastable phases

such as Ni3Ti2 and or Ni4Ti3.

 

‘ Oxide Ti4NiZO, is easy to form in Ti-Ni alloys when oxygen exists in the system. It has a similar effect
on material properties as TizNi precipitates. What is different for the two precipitates is that while TizNi
usually exist only in Ti-rich Ti-Ni alloys, '1‘i4Ni20x may be able to form in equiatomic composition or even
in Ni-rich composition. However, Because Ti4NiZOx and TizNi are isomorphous phases (with very similar
lattice parameters), it is generally difficult to distinguish between them using structure analysis.

14

Nishida et al. [Nishida, Wayman and Honma 1986, 1987] also determined the

crystal structures of those precipitates and the annealing condition for each precipitate to

appear:

13, —> ,6, +TinNil, —> ,6, +TI,NI, —> ,6, +TiNi3; (T, s 680iIO°C)
p, —> ,3; + TI,NI, —> p; + TiNi3; (6801 10°C 5 T, 5 7501mm)
3, —> an TiNi3; (T, _>_ 750°Ci10°C)

Where ,60 is the original supersaturated Ti-52Ni alloy, ,8] is the composition of the
matrix in equilibrium with Ti,,Nil4 precipitates. The same role applies to ,6), ,63, ,6] i ,62'
and [31

Transformation temperatures, mechanical properties and shape memory properties
of Ni-rich Ni-Ti alloys are affected by the formation of second phases. Both matrix
composition and interfacial stress conditions of the precipitates change as the annealing
proceeds, and they both affect the transformation temperatures. For the effect of matrix
composition, 1at.% excess nickel in the matrix will decrease M, temperature by about
100°C [Hodgson and Brown 2000]. The more complete the precipitation reaction is, the
closer the matrix composition will be to equiatomic stoichiometry. "Therefore,
transformation temperatures tend to increase as annealing proceeds. Stress ﬁeld from ﬁne
and coherent particles tends to suppress martensitic transformation, while incoherent
phase boundaries may be preferential sites for martensite nucleation. From this aspect,
extended annealing should also raise transformation temperatures. The mechanical
properties of Ni-Ti alloys are also affected by the formation of second phases during
annealing. Fine and coherent precipitates are much desired for improving the critical
stress for slip, and therefore improve Shape memory properties of the alloy.

Due to lack of interfacial coherency with the matrix, Ni3Ti2 and Ni3Ti precipitates

generally do not have signiﬁcant inﬂuence on the Shape memory properties of Ni-rich Ni-

15

Ti alloys [Nishida and Wayman 1987]. On the contrary, Ni,,Ti3 precipitates, which may
have perfectly coherent, semi-coherent, or incoherent interfaces with the B2 matrix at
different growth stages, are very useful to modify the SME properties by appropriate
annealing procedure [Nishida and Honma 1984, Nishida, Wayman, Kainuma and Honma
1986, Xie, Zhao and Lei 1990]. When Ni-Ti alloys with over-saturated Ni composition
are annealed, Ni4Ti3 precipitates start to appear at annealing temperatures as low as 300°C
~ 350°C [Ishida, Sato, Takei, Nomura and Miyazaki 1996]. At the beginning, Ni4Ti3
precipitates are small and densely distributed in the matrix, and they have perfect
interfacial coherency with the B2 matrix. As a result, martensitic transformation
temperature is greatly depressed. At the same time, the critical stress for Slip is the
highest but it also needs more stress to activate the movement of variant boundaries.
Continued annealing allows the precipitates grow, resulting in larger precipitates, less
dense spatial distribution, and partial loss of interfacial coherency with the BZ matrix.
Therefore, the depression effect on the M, temperature becomes weaker, the
transformation temperatures move up, and the mechanical strength of the material
decreases accordingly. When the precipitates grow sufficiently large so that interfacial
coherency is totally lost, the precipitates promote martensitic transformation by providing
site for heterogeneous martensite nucleation.

Xie et al. [1990] found that very ﬁne precipitates encourage the growth of certain
martensite variants, and Li and Chen [1997] observed similar effect on the R-phase
transformation. This result implies that, with presence of frne Ni4Ti3 precipitates, the
conﬁguration of martensite variants formed by cooling is no longer perfectly self-
accommodating, but has a preferred set, so the material has a certain macroscopic
conﬁguration every time it is cooled. The material can remember both high-temperature

and low-temperature shapes without any external aids.s

 

’ Due to the low ﬂow stress of martensite, little mechanical work can be extracted during the A —+ M
transformation on cooling.

16

2.3. Shape-Memory Alloys with High Transformation Temperatures

High-temperature shape memory alloys are deﬁned as those with A, temperature
above 120°C at the stress-free condition [Humbeeck 1999]. There are quite a few alloys
that have been identiﬁed as potential high-temperature SMAS including F e-Mn-Si, Cu-
Al-Ni, Ni-Al, Ni-Mn, Cu-Zr, Ni-Ti-Hf, and Ni-Ti—Pd [Wu and Ma 2000]. But the
realization of those possibilities has been hampered by various problems such as room-
temperature brittleness, phase decomposition, martensite stabilization, small shape-
memory strain at high temperature, and long—term instability of transformation
temperature. For example, Cu-Al-Ni (or Cu—Al-Ni-Mn-Ti) [Saburi and Nenno 1974,
Delaey et al. 1981] has potential to reach over 100 °C of M, temperature, and the raw
material is inexpensive. However, the material has inferior shape memory, and
martensite stabilization tends to occur during aging (even at room temperature), leading
to gradual loss of SME. The transformation temperatures also shiﬁ substantially during
thermal cycling. Yet, the biggest problem seems to be the room-temperature brittleness
due to large grain size, large elastic anisotropy and the segregation of impurities towards
grain boundaries. It was proposed that Single-crystal samples may overcome this
problem and a prototype actuator using single crystal Cu-Al-Ni has been developed
[Yawny, Sade and Lovey 1997, 1999].

Another example is Ni-Al alloys, which are also capable of undergoing
martensitic transfonnation at high temperature. When fabricated and treated
appropriately, Ni-Al alloys may have M, temperature from room temperature up to 500°C
with a small transformational hysteresis [Potapov, Poliakova and Udovenko 1996].
Theoretical calculation as well as experimental data show that Single-crystal Ni-Al alloys
have the fully recoverable strain nearly 5% if tensile axis is along <110>, and as much as
13% if the tensile axis is <001> [Enami, Martynov, Tomie, Khandros and Nenno 1981].

Polycrystalline Ni-Al, however, only shows about 0.2% of recoverable strain and it is

17

very brittle due to the intergranuar fracture mechanisms [Kim and Wayman 1992].
Although the high-temperature ductility may be improved by a third element such as Fe
or Mn [Chen 1994], problems remain with poor room-temperature ductility and the
transformation from martensite (3R or 7R structure) to Ni5A13 phase, which hinders the
reverse transformation of the martensite to austenite [Yang and Wayman 1993]. So far, it
has not been possible to develop polycrystalline Ni-Al SMAS with acceptable SME
properties.

There are more potential HTSMAS. Binary Ni-Mn has martensitic transformation
temperature between 650°C and 750°C [Adachi and Wayman 1985]. But it is very brittle
and prone to tempering at 400°C [Mulder 1995]. The Zr-based intermetallic alloy Cu-Zr
has M, temperature of 170°C and A, temperature of 25 0°C, and nickel addition to Cu—Zr
may be used to modify the transformation temperature [Koval, F irstov and Kotko. 1992].
F e-Mn-Si alloys Show recoverable strain up to 2% and recovery stress up to 350 MPa
[Kikuchi, Kajiwara and Tomota 1995, Wang, Cai, Gao, Huang and Zhao 1997]. There
are even ultra-high temperature SMAS such as near-equiatomic RuNb and RuTa, which
have transformation temperatures from 800°C to over 1000°C [Fonda and Vandermeer
1997]. All those alloys have poor ductility, especially at room temperature.

The most pronrising and most widely studied HTSMAS are Ni-Ti base ternary
alloys, which are obtained by substituting Ni with same atomic percent of Pd, Pt or Au, or
substituting Ti with same atomic percent of Zr or Hf. Since the replacement takes place
within the same family of the elements, the ternary alloys have phase structures much like
Ni-Ti binary alloy, so they are also called pseudo-binary (N i-Ti) alloys. Because of the
importance of these alloy systems and their relevancy to this study, the literature on these

alloys is reviewed in detail.

18

2.3.1. Ti-Ni-Pd, Ti-Ni-Pt and Ti-Ni-Au alloys

Substituting Ni in near-equiatomic Ni-Ti SMA with Pd, Pt or Au can greatly
increase martensite transformation temperature. At the same time, the shape memory
effect and the mechanical properties are modiﬁed due to this ternary element substitution.
In the following, the effect of ternary composition on phase, phase transfonnation, shape-
memory and mechanical properties will be presented, and because it is most widely

studied among these three alloys, Ni-Ti-Pd alloy will be considered in detail.

2.3.1.1. Phase Transformation and Transformation Temperatures. Martensitic
transformation in Ti-Ni-Pd shape memory alloys has been studied by a number of groups
[Donkersloot and van Vucht 1970, Sivokha, Savvinov, Voronin and Khachin 1983,
Lapin, Egorushkin, Shabalovshaya and Ivanova 1990, Lo and Wayman 1990, Yi, Moore
and Petric 1992, Golberg, Xu, Murakami, Morito, Otsuka, Ueki and Horikawa 1994]. It
has been found that the transformation temperatures depend on the content of additional
palladium as well as that of titanium. The Ti : (N i+Pd) ratio has the same critical effect
on transformation temperature as found in binary Ni-Ti. When titanium is more than
50at.%, transformation temperatures are nearly the highest and change only slightly with
titanium content. But when Ti was less than 50at.%, transformation temperatures
decreased dramatically with decreasing Ti content. Therefore, to ensure a high-
temperature SMA, titanium had to be at least 50at.%. Palladium affected transformation
temperatures in a different way. Figure 2.6 shows the M, temperature of Ni-Ti-Pd alloys
with different Pd compositions using data obtained in varies studies. With palladium
content less than about 8 to 10at.%, the transformation temperature is slightly decreased
with increasing Pd content. With more palladium than 10at.% added, the transformation
temperature markedly increases with increasing Pd content. High A, temperature up to

450°C may be Obtained when the Pd content is increased to 40at.%. The transformation

19

hysteresis of Ni-Ti-Pd was lowered to less than 10°C, substantially smaller than binary
Ni-Ti. The transformation enthalpy of Ti-Ni-Pd was also increased up to over 1.5 lemol
from 1.0 kJ/mol ~ 1.3 kJ/mol for binary Ni-Ti [Lo et al. 1990]. The structure of the
austenite phase of Ti-Ni-Pd has the same B2 structure (CsCl structure) as binary Ni-Ti,
but the lattice parameter was slightly larger. Martensite lattice parameters, as well as
structure were also determined by the amount of ternary palladium. When the palladium
content is small, martensite has a monoclinic structure (Bl9' structure) as that in binary
Ni-Ti. AS palladium content is increased (7 to 15at.%), the martensitic transformation
becomes two-stage: a B19 orthorhombic martensite forms ﬁrst, and as the temperature is
continued down, a B19' monoclinic martensite. When the palladium content is over
about 15at.%, only B2 to orthorhombic martensite transformation has been observed in
normal temperature range [Lotkov, Grishkov and F adin 1982].

When all the nickel is replaced by palladium in the Ti-Ni binary SMA, an
isomorphous Ti-Pd binary intermetallic compound results. Donkersloot and Van Vucht
ﬁrst found the martensitic transformation in this alloy in 1970 [Donkersloot and Van
Vucht 1970]. Later on Khachin et al. systematically studied the pseudo-binary TiPd-TiNi
alloy system, and determined the transformation temperatures and the structures of
martensite as a function of composition [Khachin, Matveeva, Sivokha and Chemov
1981]. According to them, the alloys with nearly Ti,,,Pd50 composition transform from B2
type austenite to the B19 (orthorhombic) martensite. Binary near-equiatomic Ti-Pd alloy
had the highest transformation temperature in the TiNiPd system. Otsuka et al. [Otsuka,
Oda, Ueno and Piao 1993] also studied the phase transformation of the near equiatomic
Ti-Pd alloy. It was determined that, after solution treatment followed by water
quenching, the material had an M, of 817.7 K, M, of 812.6 K, A, of 847.0 K, and A, of
863.8 K.

Aging also affected the transformation temperatures of Ni-Ti-Pd alloys, especially

for samples that have non-stoichiometric composition and are quenched after solution

20

treatment. Shimuzu et al. [Shimizu, Xu, Okunishi, Tanaka, Otsuka and Mitose 1998]
reported that, for Ti-lean compositions, aging at temperatures below 700 K did not result
in appreciable change in transformation temperature, but aging at higher temperature
resulted in a decrease in M, and an increase in M, temperature. For Ti-rich compositions,
aging the sample between 650 K to 800 K resulted in the lowest transformation
temperatures.

Lindquist studied the phase transformation and Shape memory properties of Ti-
Ni-Pt alloys [Lindquist 1978, Lindquist and Wayman 1990]. He found that platinum
substitution could increase the martensitic transformation temperatures in a way more
effective than palladium additions, even though less than 10at.% of platinum also slightly
decreased the transformation temperatures. With 20at.% of platinum, the M, temperature
was about 300°C, and when the platinum reached 30at.%, the M, temperature increased to
about 620°C. Wu and Wayman [1987] reported martensitic transformation in Ni-Ti-Au
ternary alloys with Au substituting Ni. The M, temperatures of TisoNimAu,0 and Ti,,,Au50

were 440°C and 610°C, respectively.

2.3.1.2. Shape-Memory Properties. Shape-memory properties of Ti-Ni-Pd high
temperature SMA have been characterized by different research groups, usually in
uniaxial tension or bending modes. It was generally found that as palladium content went
up, the shape memory properties were degraded. At the same time, the stress for
martensite reorientation was increased. Therefore, shape-memory strain is more difﬁcult
to obtain. However, depending on palladium composition, the material could still have
recoverable strain up to 4%. Khachin et al. [1981] also made a high temperature torsion
test for a Ti,,,Pd_.,7Nil3 alloy, and reported that about 4% strain was attained by stress-
induced transformation under a uniaxial applied stress of 200 MPa, and it was recovered
completely upon reverse transformation. Wu et al. [Wu, Liu, March and Tseng 1994]

tested Ti-Ni-Pd alloys with different palladium contents in bending and also found the

21

fully recoverable strain was decreased as Pd content was increased. Partial recovery was
Observed when the strain exceeded the fully recoverable strain, but the percentage of
recovery fell dramatically with excess deformation.

As introduced before, binary TisoPd,o undergoes martensitic transformation at high
temperature. Otsuka et al. [1993] also characterized the shape memory properties by
deforming the material in uniaxial mode at different temperatures. At room temperature,
when the sample was elongated to 3.5% plastic strain, about 60% of the strain was
recovered upon heating. As the tensile test temperature was increased to 573 K, while the
stress for the same amount of strain decreased to 400 MPa, the recoverable strain became
less. When the testing temperature increased to 713 K or above, no recoverable strain
was observed. The reason for this phenomenon, as indicated by Otsuka et al., is that the
stress for martensite variant reorientation is lower than the critical stress for slip (in both
martensite and austenite), especially at elevated temperatures. As a result, slip (which
results in permanent strain) occurred before (or at the same time as) reorientation of
martensite variants took place. Although this explanation was intended for the shape-
memory behavior of Ti,,,Pdso alloy, it could also be true for that of all Ti-Ni-Pd alloys.

Perfect transformational superelasticity had not been Observed in Ti-Ni-Pd or Ti-
Pd alloys. The reason may, as indicated by Otsuka et al. [1993], be also related to low
critical stress for slip at high temperature. In fact, the critical stress for slip in the
austenite in Ti-Ni-Pd alloys was lower than that in the martensite, which was just
opposite to the situation in Ti-Ni binary alloys [Golberg et al. 1994]. Considering that
transformational superelastic deformation has to be made at a temperature above A,, the
stress for inducing martensitic transformation is high. As a result, slip occurs before the
applied stress can induce martensitic transformation. Under this circumstance, permanent
strain was generated.

The two-way shape memory effect in HTSMAS is more difficult to implement,

especially by training processes. This is because at elevated temperature, dislocations are

22

prone to recover. However, Sun et al. reported limited success in developing two-way
SME in the Ni-Ti-Pd and Ti-Ni-Pd-B shape memory alloys by mechanical training [Sun
and Wu 1994]. The training process that was used included: 1), deforming the material to
a given constant strain at room temperature; 2) unloading and heating to A, + 30°C, and
ﬁnally, 3) quenching the sample into water. The effect of training strain and training
number on TWSME was studied. It showed that applying larger training strain and more
training numbers increase the strain for TWSME. However, most TWSME strain had
been developed in the ﬁrst ﬁve cycles. Sun et al. also found that as the palladium content
increases, the TWSME was decreased for a ﬁxed training numbers at a ﬁxed training
strain. Boron additions did not affect the 2-way SME developed by training.
Transformation temperatures were also measured for samples after different training

numbers. They seemed to be very stable over thermal cycling.

2.3.1.3. Improvement of Mechanical Properties of T i-Ni-Pd SMAS. Substituting Ni
with Pd in Ni-Ti SMAS caused a decrease in the critical stress for slip at high temperature
and decreasing ductility in the alloys. While low critical stress is the reason for inferior
shape memory prOperties, brittleness made the material less durable and harder to
process. Efforts have been made to improve the ductility of the material and increase the
critical stress for slip. AS reviewed below, the strategies that have been developed
included: 1) increase the critical stress for slip by adding a 4th element; 2), strengthen the
austenite by age hardening, 3) strengthen the austenite by a thermomechanical treatment.
Yang and Mikkola, among other research groups, attempted to improve the room-
temperature ductility of Ni-Ti-Pd alloys (Tuominen and Biermann 1988, Yang and
Mikkola 1993) by adding a small amount of boron. According to them, the room-
temperature tensile elongation of as-cast sample that has no boron addition was only 2%.
With about 2000 ppm of boron, tensile elongation reached 8%. More boron additions

tend to decrease the ductility. The reason for the ductilization by boron addition was

23

because the grains were reﬁned. Micrographic investigation revealed that the as-cast
sample with boron has the grain Size 4 times smaller than that without boron. Also,
boron additions clearly reduce the rate of grain growth that occurred by subsequent
annealing of the as-cast samples. As a result, boron is also effective in increasing the
materials ductility at high temperature [Suzuki, Xu, Morito, Otsuka and Mitose 1998].
For example, with the same starting grain size with or without boron in hot rolled
TisoPdmNizo, after annealing at 900°C for 1 hour, the sample with boron had a grain size
of less than 50 microns, whereas the sample without boron had a grain size of larger than
120 microns.

Boron is also beneﬁcial to high-temperature ductility and strength of Ti,,,Pd,,,Ni20
alloy. The ultimate tensile strength and elongation were about 780 MPa and 16% for
TisoPd30Ni,9_,Bo,2, respectively, compared to 460 MPa and 8% for Ti,,,Pd3,,Ni20 alloy.
Boron existed as TiB2 precipitates with an average size larger than 1 micron, and these
precipitates distributed within grains, rather than on grain boundaries. More importantly,
additional boron had no noticeable negative inﬂuence on high-temperature shape memory
characteristics. Suzuki et a1. [1998] reported that in Ti,,,Pd3,,Ni20 alloy after
homogenization at 1000°C for 5 hours followed by hot rolling at 923K and annealing 1
hour at 1173 K, with or without boron, the recoverable strain was over 3%. However, the
stress that allows a full strain recovery was only about 100 MPa. It was suggested that if
small and densely distributed TiB2 borides could be achieved through using proper
thermal treatments, it was possible to improve the high-temperature SME by precipitation
hardening.

Precipitates, such as Ni4Ti3 in Ni-rich Ni-Ti binary alloys were useful to improve
the shape memory properties because the critical stress for slip was increased. Shimizu et
al. [1998] investigated an approach for inducing precipitates in non-equiatomic

Ti50,,Pd3oNi,9, alloys. They found that the shape memory characteristics were improved

24

by aging the alloy at 773K for 1 hour. The recovery rate6 was about 90% for a total of
6% strain for deformation at 473 K, which was about 10% higher than that for the
equiatomic Ti,,,Pd3,,Ni20 alloy. The reason for the improvement was found to be the
hardening effect by homogeneously distributed ﬁne precipitates Ti2(Ni+Pd), which had
TizNi type structure and were produced by aging at 773 K of the Ti-rich alloy. The
appearance of ﬁne Ti2(Ni+Pd) may suggest that the solubility of excess Ti is greater in
ternary alloy than that in binary Ni-Ti.

The third approach to improve the shape-memory property of Ti-Ni-Pd is
thermomechanical treatment, as suggested by the success of Ti-Ni binary alloys. Golberg
et al. reported their work on a Ti,,,Pd,,,Ni20 composition in this regard [Golberg et al.
1994, 1997]. The solution treated sheet was ﬁrstly cold rolled to 15 to 25% thickness
reduction, then annealed at 673 K for a recovery treatment 7, and then deformed to a
tensile strain of 5.3% at 443 K (20 to 40°C below M,). This deformation was fully
recovered upon heating to A, temperature. In comparison, if the cold rolled material was
solution treated at 1173 K, the perfectly recoverable strain was only about 2%, even
though the solution treated sample was easier to deform than the recovery treated sample.
The reasons for the improvement were because that the critical stress for slip in the
martensite was increased to 400 MPa aﬁer recovery treatment compared to 200 MP3 for
the sample having only solution treatment. Superelasticity resulting from stress induced
martensitic transformation was also observed in the recovery treated sample, but it was
absent in the solution treated sample.

It should be pointed out that, similar to the results for Ti-Ni binary alloys, proper
thermal treatment was crucial to a successﬁrl implementation. This is because during

high-temperature annealing, recrystallization could occur, which would result in

 

° The recovery rate is deﬁned as the percentage of the plastic strain imparted to martensite that can be
recovered when the sample is heated to austenite.

7 “Recovery” is used here in the classical metallurgical sense, i.e., to represent a process of dislocation
annihilation.

25

substantial grain growth and degraded SME properties. Xu et al. systematically studied
the recovery and recrystallization behavior of Ti,,,Pd3,,Ni20 alloys after cold work [Xu,
Shimizu, Suzuki, Otsuka, Ueki and Mitose 1997]. Hardness drop was linked to the
reverse martensitic transformation, and occurrence of recovery and recrystallization.
TEM observation directly revealed that recrystallization took place during annealing at
temperature higher than 600°C. This result was useful for the optimization of

thermomechanical treatment procedure.

2.3.2. Ni-Ti-Zr and Ni-Ti-Hf Alloys

Although Ti-Ni-Pd (Pt, Au) alloys are very promising high-temperature shape
memory alloys, they contain considerable amount of expensive noble metals. The other
group of Ni-Ti —base alloys, Ni-Ti-Hf and Ni-Ti-Zr, which are obtained by substituting Ti
with Zr or Hf, are more economical HTSMA candidates. Also, the tertiary additions are
more efﬁcient in raising transformation temperatures in Ni-Ti-Hf and Ni-Ti-Zr alloys.
Below, the phases, phase transformations, mechanical properties and shape memory

characteristics of Ni-Ti-Hf and Ni-Ti-Zr alloys are reviewed.

2. 3. 2. 1. Transformation Temperatures. Similar to Pd (Au, Pt) substituting for Ni in
binary NiTi, Zr and Hf substituting for Ti can also effectively increase the martensitic
transformation temperatures. Pu et al. systematically studied the effect of the amount of
Zr [Pu, Tseng and Wu 1995] and Hf [Pu, Tseng and Wu 1994] additions on the
martensitic transformation temperatures. For Ni-Ti-Zr alloys with ﬁxed Ni composition
at 50at.% and varying Zr composition, the transformation temperatures decreased only
slightly with increasing Zr content until Zr reached 10at.%, whereafter the transformation
temperatures increased dramatically. With 20at.% of Zr, the A, temperature reached

about 300 °C and MF, temperature of reached 260 °C. R-phase transformation was also

26

observed in the alloy with small amount of Zr additions.

Substituting Hf for Ti raises the transformation temperatures even more efﬁciently
than with Zr. AS also reported by Pu et al. on solution-treated Ni-Ti-Hf alloys with ﬁxed
Ni content at 50at.% and varying Hf content, the transformation temperatures were also
decreased slightly for the ﬁrst few percents of Hf additions, then increased rapidly with
further increase of Hf content. Angst et al. [Angst, Thoma and Kao 1995] also conducted
a study of the effect of composition on the transformation temperatures in as-cast Ni-Ti-
Hf samples. According to them, with 50at.% or less of Ni content, the apparent minimum
peak transformation temperatures were observed at Hf composition between 0 to 3at.%.
Transformation temperatures slightly increased before the Hf reached 8at.%, then
increased with a large slope as hafnium continuously increased. When the hafnium
reached 30at.%, the A, and M, temperatures reached 600°C and 535°C, respectively. The
transformation energy measured by DSC was in the range of 20 to 25 J/g over the Hf
composition range between 0 to 15at.%. Figure 2.7 summarizes the effect of hafnium
content on M, temperature obtained by different research groups.

Angst et al. also studied the transformation temperatures of the as-cast Ni-Ti-Hf
alloys with off-stoichiometric compositions. As shown in Figure 2.8, the transformation
temperatures did not change with the Ni content changed from 40 to 50at.%. However,
once the Ni content exceeded 50at.%, the transformation temperature dropped
dramatically. With only 0.5at.% more Ni than stoichiometn'c, both A, and M, decreased
more than 100°C.

Annealing may change the transformation temperatures of Ni-Ti-Hf alloys too,
especially for as-melted-then-quenched samples with off-stoichiometric composition, and
cold worked samples. Besseghini et al. [Besseghini, Villa and Tuissi 1999] studied the
Ni,,,Ti,,,I-IfIo alloy and found that annealing at 400°C after solution treatment decreased the
transformation temperature, but annealing at 600°C raised the transformation

temperatures slightly. Besseghini attributed the variation of transformation temperature

27

to the growth of large amount of second phases (about 50% in volume fraction) that are
present in the alloy. Augst et al. [1995] found cold work considerably decreased the
transformation temperature of Ni-Ti-Hf alloy, but the transformation temperatures were
recovered upon annealing at high temperatures.

The stability of transformation temperatures against thermal cycling was
evaluated by Besseghini et al. [1999] in solution treated Ni,,,Ti,,,Hflo alloys. Up to 30°C
decrease was observed for the ﬁrst 20 cycles. But the transformation temperature tended
to stabilize after the ﬁrst 20 cycles. Experiments by Pu et a1. [1994, 1995] revealed
Similar phenomenon for Ni-Ti-Zr alloys. But Pu et al. also found the thermal stability
was improved as more Zr was added. Pu et al. also found that Ni-Ti-Hf had better
thermal stability against thermal cycling than Ni-Ti-Zr alloys, but neither of them was as
good as Ti-Ni-Pd in this respect.

The crystal structure of martensite for Ni,,_,Ti,,,,Hf,, alloys was determined by
Han et a1. [Han, Zou, Jin, Zhang and Yang 1995, Han, Zou, Wang, Zhang, Yang and Wu
1995]. It had the same B19’ structure as martensite in Ni-Ti binary alloys with the lattice
parameters only changed slightly. However, unlike martensite in Ni-Ti binary alloys, the
martensite self-accommodating substructure of Ni,,_,Ti3,,_,Hfls alloy was dominated by
(100) compound twins, and Han et al. attributed this substructure characteristic to the
small recovery strain of the ternary Ni-Ti-Hf alloys.

Second phases may exist in Ni-Ti-Hf alloys, especially for alloys with off-
stochiometric composition. Precipitation and growth of second phases during annealing
were mainly responsible for variation in transformation temperatures, and were also the
reason for alteration of mechanical and shape-memory properties. (Ti+Hf)2Ni
precipitates were very common in nickel-lean Ni-Ti-Hf alloys. (Ti+Hf),NiZO, precipitates
were also present due to inevitable introduction of oxygen. Thoma and Boehm [1999]
observed a linear relationship between the nickel composition and volume fraction of the

precipitates. With nominal 50at.% of nickel there was 0.3% (in volume) precipitates.

28

When the nickel was decreased to about 49at.%, the volume fraction of precipitates was

increased to 6%.

2. 3. 2. 2. Shape-memory Properties of Ni-T i—Zr and Ni-Ti-Hf Alloys. Shape-memory
was observed in both Ni-Ti-Zr and Ni-Ti-Hf alloys with up to 40at.% of Zr or Hf
replacing Ti. However, the Shape memory properties were evidently inferior to binary
Ni-Ti because of the tertiary element replacement. This SME property degradation was
primarily manifested in the reduction of fully recoverable strain, reduction of the stress
allowing full strain recovery, increase of critical stress for martensite reorientation and
decrease of critical stress for slip in austenite at high temperature.

Pu et al. [1995] evaluated the one-way SME properties of Ni-Ti-Zr and Ni-Ti-Hf
alloys, which was done by pre-straining the material at room temperature then heating the
material to 50°C above the A, temperature. It was found that with increase of Zr content,
the strain recovered by heating was reduced. For NisoTi35Zr” alloys, 1.5% pre-strain that
was obtained at room temperature was fully recovered upon heating, and 2.2% of pre-
strain could be recovered by 90%. For alloys with 5 to 10at.% of Zr composition, the
fully recoverable strain was nearly 2.5%. Fully recoverable strain in the Ni-Ti-Hf alloys
was also decreased as more hafnium was added. With 5at.%, 15at.% and 20at.% of
hafnium in the alloy, the fully recoverable strain was 3.0%, 2.5% and 1.8%, respectively.

However, if the martensite was pre-defonned at elevated temperature, the
recovery rate8 could be improved. Olier et al. [1995] studied the dependence of recovery
rate on uniaxial strain in a Ni,,,Ti3,Hf,2 at room temperature. It was found that the
recovery was poorer for the lower deformation (recovery rate was 60% for prestrain =
1%), and it was increased for the higher deformation (recovery rate was 80% for prestrain

= 2.5%). Therefore, there seemed to be a competition between the plastic deformation of

 

° In uniaxial deformation mode, the recovery rate is deﬁned as the percentage of strain recovered on
heating in the total plastic prestrain (elongation). In bending deformation mode, the recovery rate is
measured as the precentage of angle recovered on heating from the initial angle.

29

the austenite and the reorientation of martensitic variants. Defonning the martensite at
higher temperature the recovery was better. For example, a 1% strain deformed at 150°C
(35°C below A, temperature) recovered 93% upon heating.

The best SME was found in Ni,,,Ti,,Hf,2 by deforming the austenite at 150°C, 9 a
temperature just above M,. The alloy fully recovered as much as 4% strain on heating.
Based on this experiment, Olier suggested that the movement Of twin boundaries in
Ni,0Ti3,Hfl2 appeared more difﬁcult than that in binary Ni-Ti. The accommodation
process could be hindered by interface pinning or plastic relaxation, which was in
correlation with the increase of the hysteresis of Ni-Ti-Hf alloys.

Wang et al. [Wang, Cai, Zhu and Zhao 1997] conducted similar experiments on
NisoTi35Hf” alloy. They also found if the tensile test was conducted at 353K, up to 3%
strain could be fully recovered. 4% strain could be recovered by 95%, and 6% strain by
83%, upon heating to above A, temperature.

Transformational superelasticity has not been reported in Ni-Ti-Hf or Ni-Ti-Zr
high-temperature Shape memory alloys. As was the case for Ni-Ti-Pd(Pt, Au) alloys, the
low critical stress for Slip in the high-temperature austenite was thought to be the cause.
Although stress induced martensitic transformation was Observed at temperature just
above M, temperature, the superelasticity occurred only when the deformation
temperature was above A, temperature, which was at least 50°C higher than M, due to the
wide transformational hysteresis of Ni-Ti-Zr and Ni-Ti-Hf alloys.

Two-way SME acquired by training is generally associated with the establishment
of dislocation structures and with local internal stress ﬁelds in austenite developed during
training, which act to favor the formation of certain variants of martensite during cooling.
For HTSMAS, because of its high austenite transformation temperature, the dislocation

arrangements formed during thermomechanical cycling are easy to recover, making it

 

9 Because of the transfonnation temperatures, Ni,,,Ti,,Hfl2 was martensitic if it was heated from a low
temperature (below M, temperature), and it was austenitic if it was cooled down from a high temperature
(above A, temperature)

30

more difﬁcult to train the sample to demonstrate two-way SME. However, Pu et al.
[1994] demonstrated that two-way SME could be implemented into Ni-Ti-Hf HTSMA by
training in a bending mode. Even so, they found the magnitude of TWSME strain was
not comparable to binary Ni-Ti, and the TWSME became saturated by only 3 or 4 cycles
of training [Pu et al. 1994].

2.3.2.3. Mechanical Properties. In general, Ni-Ti-Zr and Ni-Ti-Hf alloys had
considerably poorer ductility and workability than Ni-Ti binary alloys, especially at room
temperature. This was supported by the higher hardness of Ni-Ti-Zr and Ni-Ti-Hf alloys.
However, Thoma et al. [1999] found it possible to achieve a reduction in area of 27%
without cracking when they rolled Ni,,,Ti,,,Hflo alloys at room temperature. Continued
reductions in area above 27% produced signiﬁcant edge cracking of the strip. Wang et al.
[1997] systematically studied the tensile behavior of NisoTinf,0 alloys at different
temperatures. Unlike the stress-strain curve of binary Ni-Ti alloys, a plateau (i. e. , perfect
plasticity) was generally absent. Ultimate elongation, as well as characteristic yield stress
on, was a function of test temperature. After solution treatment at 900°C for 1 hour, the
material had an ultimate elongation of ~10% at room temperature. With the increase of
tensile testing temperature, the ultimate elongation increased substantially, and it reached
over 30% at temperature just above A, temperature. The dependence of 0'0,2 stress with
temperature was just the Opposite with that of ultimate elongation with testing
temperature. At room temperature, 0'0,2 was 550 MPa. As the temperature decreased, it
gradually decreased to the nrinimum (430 MPa) at 510K (just above A, temperature).
Then it climbed to 650 MPa at 590 K.

For Ni-Ti-Hf and Ni-Ti-Zr high-temperature shape memory alloys, the critical
stress for martensite reorientation is high so that permanent strain is easily introduced
when the material is deformed in the martensitic condition. Olier et al. [1995] studied the

yield stress of Ni,,,Ti3,Hf,2 SMA alloy (Figure 2.9) and found that the yield stress was

31

highly temperature dependent. The lowest yielding stress was obtained when the material
was deformed just above M, in austenite (deformation temperature = 150°C, M, ~ 140°C).
The stress-strain curve exhibited a respective elastic contribution of both austenite and
martensite phases, separated by the plateau associated with stress induced martensitic
transformation. The SME properties were also the best because loading at this condition
led to optimal variant formation. Therefore, the need for reorientation of martensite
variant was avoided. This result suggests that Ni-Ti-Hf (Zr) alloys may have optimal
performance in some working conditions, such as applying continuous biasing stress

while cycling the material between low temperature and high temperature.lo

2.3.3. General Comparison of Different HTSMAS

In comparing Ni-Ti-Zr and Ni-Ti-Hf as high-temperature shape memory alloys,
Ni-Ti-Zr has advantages over Ni-Ti-Hf in (1) cost; (2) slightly smaller hysteresis for
alloys with same transformation temperature; (3) potentially better SME properties.
However, Ni-Ti-Hf alloys hold advantages over Ni-Ti-Zr in the following aspects: (1) it
has higher transformation temperatures for the same 3'‘1 element content; (2) Hf has lower
afﬁnity for oxygen as compared to Zr; (3) it has a larger temperature range for hot
working due to the slightly higher peritectic and melting points; and (4) it shows better
stability against thermal cycling. In comparison between Ni-Ti-Pd group and Ni-Ti-Hf
group, Ni-Ti-Pd has advantages over Ni-Ti-Hf in some areas: (1) it has smaller
transformation hysteresis which leads to quick response; and (2) the transformation
temperatures are more stable during thermal cycling. Ni—Ti-Hf has advantages over NiTi-

Pd on some other areas: (1) it is more cost effective; (2) it needs less third element to be

 

'° Binary Ni-Ti SMAS also exhibit better shape memory properties in such conditions. For example,
substantial transformational strain may be obtained at small biasing stress during cooling down, while the
same level of stress may not be enough to produce any shape-memory strain if loaded to the martensite.

32

added for the same transformation temperature and (3) it has potentially larger

recoverable strain.

2.4. Shape-memory Alloy Thin Films
Thin ﬁlm science and technology plays an important role in modern industry.

Maj or applications of thin ﬁlms have been found for coatings and microelectronics.
Interest in fabrication of shape-memory alloy thin ﬁlms has arisen due to their potential
useﬁrlness in microelectromechanical systems (MEMS). As an actuator material, shape-
memory alloys may offer large energy-density output upon reverse martensitic
transformation, which may be conveniently activated by Joule heating. In addition, by
fabricating the shape-memory alloy in the form of a thin ﬁlm, the disadvantage of slow
response for SMA actuator may be greatly offset. In this section, a brief summary of thin
ﬁlm fabrication by sputtering, and the structure and stress characteristics of sputtered thin

ﬁlms is given ﬁrst, then shape memory in thin ﬁlms will be reviewed in detail.

2.4.1. Deposition of Thin Films by Sputtering

There are numerous methods of making thin ﬁlms, but there basically are two
categories: Chemical Vapor Deposition (CVD) and Physical Vapor Deposition (PVD),
according to whether chemical reactions are involved during the deposition. Sputtering,
usually a PVD method unless reactive working gas is used, has high deposition rate and
high ﬂexibility of target material selection, thus is favored in many circumstances. It

remains the only method successfully applied to fabrication of SMA ﬁlms.

2.4.1.1. Sputtering Processes. Sputtering a thin ﬁlm usually takes place in a

vacuum chamber in order to avoid contamination from atmosphere and to allow atom

transport to the substrate. A voltage, either direct current (d.c.) or radio frequency (r.f.),

33

is applied between target (cathode) and substrate (anode), and it is the energy source to
ionize an inert working gas and accelerate the positive ions to the cathode. Energetic ions
strike the target surface, causing the target atoms to eject from the surface and to form
continuous thin ﬁlm on the substrate. Secondary electrons from the target play an
important role in maintaining the discharge.

The quality of the ﬁhns is determined by the sputtering parameters such as power
to the target, deposition rate, working gas species, working gas pressure, distance from
cathode to substrate, and substrate temperature. Among those parameters, working gas
pressure and substrate temperature affect thin ﬁlm quality most dramatically. In general,
lower working gas pressure is more desirable because of three reasons: ﬁrst, less working
gas will be incorporated into the deposit so the deposit will be denser. Second, the
deposited atoms arrive onto the substrate with more kinetic energy so they are more likely
to further diffuse. Third, more energetic reﬂected working gas neutrals arrive onto the
substrate, providing more energy for ﬁlms to further densify. However, ordinary
sputtering has to be operated at high working gas pressure (>10 mTorr) due to reduced
ionization efﬁciency at low pressure. Modiﬁcation of the Simple diode sputtering source
needs to be done in order that the working gas pressure can be reduced. Conventionally
there are two methods to do so. The ﬁrst approach to force the secondary electrons travel
spirally by applying magnetic ﬁeld in front of target (magnetron Sputtering). The other is
to bring additional supply of electrons by introducing a thermoionic source (triode
sputtering). Both magnetron sputtering and triode sputtering can be Operated at working
gas pressure of 1 mTorr or less.

Another important parameter is substrate temperature. In general higher substrate
temperature provides more thermal energy allowing the thin ﬁlm to densify via diffusion.
At very high temperature and slow deposition rate, single-crystal ﬁlms may even be
obtained. However, high substrate temperature is not favorable in industry due to

increased complexity of the equipment. Magnetron sputtering, which can produce thin

34

ﬁlm with reasonably good quality at low substrate temperature, is very popular among all

sputtering methods.

2.4.1.2. Composition, Structure and Microstructure of Sputtered Films. Sputtering
from a Single-element target obviously produces single-element ﬁlm in a PVD process,
even though the thickness distribution of thin ﬁhn may vary across the substrate.
Sputtering from multi-element target, however, may produce thin ﬁlm with composition
deviating from that of the target. Basically, the composition shift may arise from several
reasons. First, yield rates for different elemental species are different. Therefore, the ﬂux
leaving the substrate may have different composition than the target. This is particularly
true during the initial sputtering period of a new target. However, the composition
discrepancy resulting from yielding rate difference will diminish after a proper wear-in or
bum-in period, because the composition change on the surface of target will offset the
yield rate difference.

Second, scattering during transport from target to substrate may also cause a
composition discrepancy between the target and the ﬁlm. This is because the atoms with
different size and mass have different mean free paths, so some atoms may be more likely
to be scattered away than others due to more probable collisions with the background gas
atoms. Third, resputtering after the atoms arrive on the substrate may alter the
composition of thin ﬁlm. Incoming atoms may have enough kinetic energy to dislodge
some adatoms on the ﬁlm. Depending on their atomic mass, bonding strength with
others, and ability to move to low-energy trapping sites, some atom species may more
likely be knocked away from substrate. Despite these uncertainties, it is still possible to
get the desired thin ﬁlm composition by controlling the target composition. This is
because the ﬁrst cause for discrepancy may be eliminated by proper wear-in procedure,

and the second and third causes may be corrected through establishing an empirical

35

relationship between the target composition and ﬁlm composition, and making suitable
compensation in the target composition accordingly.

The structure of thin ﬁlms may be amorphous, partially crystalline, or totally
crystalline, depending on the sputtering parameters that are used. The microstructure of
thin ﬁlm is also determined by sputtering conditions. Thornton [Thornton 1977,
Thornton and Hofﬁnan 1985] developed a zone structure model illustrating the structural
and microstructural characteristics a function of substrate temperature for sputtered ﬁlms.
The model illustrated that the microstructure of sputtered ﬁlms are divided into basically
three ‘zones’ depending on the homologous temperature Tﬂm (where T s is substrate
temperature during deposition, Tm is the melting temperature of the ﬁlm):

Zone I (T W m <0.3): Characteristic microstructure of the ﬁlms is either
amorphous or dentritic. In each case the structure is extended in the grth
direction. The ﬁlms are rather porous, and have very poor mechanical
properties.

Zone 11 (0.3 < T ,JT m <0.5): The most characteristic feature of the rrricrostructure
is a largely uniform columnar morphology with high compactness. The grain
diameter, d, extends throughout the thickness of the ﬁlm, t, but the grain
aspect ratio d/t is usually less than 1.

Zone 111 (T ,JT m > 0.5): Metallic ﬁlms have a columnar structure with a grain
diameter greater than the thickness of the ﬁlm. Films have very strong
texture. At extreme conditions, thin ﬁlm grth in sputtering behaves like
epitaxial growth.

The zone structure can be largely explained in terms of self-shadowing, surface
diffusion and bulk diffusion. These three mechanisms, along with adsorption of adatoms,
constitute four basic processes of ﬁlm formation. In the ﬁrst regime, both adatoms and
sub-surface atoms have very poor mobility. Due to the self-shadowing effect, porous

structures are usually developed. Intermediate substrate temperature allows sufﬁcient

36

surface diffusion of the adatoms, but bulk diffusion of sub-surface atoms is still sluggish.
A more compact ﬁlm is formed, but numerous grain boundaries exist. At high substrate
temperature both surface diffusion and bulk diffusion are sufficient. Therefore, some
grain growth takes place during the sputtering process .

Evidently other sputtering parameters can also affect the zone structure. The most
pronounced one is the working gas species and its pressure. The momentum exchange
between the incoming species and the adatoms provides more energy for adatoms to
diffuse. The kinetic energy of the incoming species diminishes as more collisions with
working gas occur. Low working gas pressure decreases the probability of such
collisions, and the kinetic energy of the incoming species will be mostly preserved. As a
result, lowering working gas is similar in a sense to increase the deposition temperature,

making the zone structure move toward lower temperature regime.

2.4.1.3. Extrinsic Stress and Intrinsic Stress. Stresses are often introduced from
thin ﬁlm fabrication processes and post-fabrication treatment. Stresses in thin ﬁlms have
important effects on many phenomena such as decohesion, defect generation and surface
morphology evolution, and they can substantially modify physical properties of the thin
ﬁlms.

Stresses in thin ﬁhns can be divided into categories: extrinsic stress and intrinsic
stress. Extrinsic stress is thermal stress, and it results form the differential CTES
(Coefficients of Thermal Expansion) between thin ﬁlm and substrate. Assuming that the
thermal stress does not exceed the elastic deformation limit of ﬁlm, and that the substrate

is rigid, thermal stress can be quantitatively predicted as:

TIE

athermal = l 1
T, ‘V f

 

-(a, — a, )dT (2.2)

37

Where T, and T, are starting temperature and ﬁnish temperature, respectively. E, is the
elastic modulus of thin ﬁlm, v, is the Poisson’s ratio of the ﬁlm, and a, and a, are CTES
of ﬁlm ad substrate, respectively.

While the origin of extrinsic stress is quite clear, that of intrinsic stress is much
more complicated. In fact, no single model can explain how intrinsic stress is created.
Nevertheless, it has been found that a number of mechanisms, such as nonequilibrium
growth, defect evolution, interfacial constraint, in-situ phase transformation and in-situ
stress relaxation are related to the Sign (tensile or compressive) and the magnitude of the
intrinsic stress. However, the dominant mechanism for intrinsic stress in sputtered ﬁlm is
unique compared to ﬁlms made by other techniques such as evaporation. In sputtered
ﬁlm, compressive stress was usually observed, while ﬁlms Obtained by evaporation were
usually under tensile stress. D’Heurle [D’Heurle, Berenbaum and Rosenburg 1968,
d’Heurle 1989] ﬁrst pointed out that, in evaporation, the deposited atoms are nearly in
thermal equilibrium with energy about few kT. In a sputtering process, a ﬂux of
depositing atoms and the neutral gas atoms reﬂected from the cathode arrive onto the
substrate with energy up to several e V. Adatoms on the ﬁlm are thus subjected to
bombardment which, at the atomic scale, is similar to the ‘shot peening’ effect, an
approach widely used to put material surfaces into compression in metallurgical
fabrication. It does not matter so much that the energy is small since the damage is
immediately buried under newly deposited layers, but rather that the damage is
accumulated. The effect is increased under conditions of negative bias sputtering when
the substrate is deliberately bombarded with energetic working gas ions. It is particularly
noted that this atomic peening mechanism is predominant only when the working gas
pressure is low, since too much background gas tends to thermalize the kinetic energy of

the working gas atoms reﬂected from sputtering target.

38

2.4.2. Sputtered Shape Memory Thin Film

Since 1990, shape-memory alloy thin ﬁlms as a potential material for
microactuator in MEMS has been investigated by a number of research groups [Busch,
Johnson, lee and Stevenson 1990, Ikata, Fujita, Ikeda an Yamashita 1990, Kuribayashi,
Yosaki and Ogawa 1990, Walker, Gabriel and Mehregany 1990, Chang and Grummon
1992, Miyazaki and Ishida 1994, Hua, Su and Wuttig 1995]. However, only Ni-Ti-base
alloys were successfully made to be Shape-memory thin ﬁlms. In this part, fabrication
and characterization of Ni-Ti shape memory thin ﬁlms are reviewed, and their potential

applications are discussed.

2. 4. 2. 1. Fabrication of Shape Memory Thin Film. Targets made by different
methods may be used to sputter Ni-Ti SMA thin ﬁlms. However, because oxygen has
very strong afﬁnity for titanium and oxygen is known to have grave effects on SME
properties of Ni-Ti alloys [Wu, Khachturyan and Wayman, 1988, Johnson 1991, Lin, Wu
and Lin 1994, Barborini, Pieseri, Mutti, Milani, Biasioli, Iannotta and Gialanella 1998],
oxygen contamination is a great concern for the target fabrication. The most widely used
targets have been pre-alloyed targets. Oxygen level in sputter targets may be well
controlled. Besides pre-alloyed targets, composite targets that are composed of pure -
element powders [Quandt, Halene, Holleck, Feit, Kohl and SchloBmacher 1995], or
sectored pure nickel and titanium material have been used [Bendahan, Canet, Seguin and
Carchano 1995, Tsuchiya and Davies 1998].

The composition of deposited Ni-Ti ﬁlms is of great concern because it is very
important for SME and transformation temperature control. It has been repeatedly
reported that the composition of sputtered Ni-Ti ﬁlms deviate from that of the sputter
target. For example, Johnson et al. reported that sputtering from an equiatomic Ni-Ti

target produced thin ﬁlm containing 51-54at.% nickel [Johnson, Martynov and Minners,

39

1995]. The major reason for composition discrepancy may be because Ti is a
preferentially resputtered species compared to Ni atoms. Sputtering parameters also
affect the composition of thin ﬁlms. Lee [1994] attempted to establish the relationship
between ﬁlm composition and the process parameters in a dc. magnetron sputtering
process. He found that, for a given sputter condition, the ﬁlm composition varied with
the wear history of the target, but steady-state composition of thin ﬁlm was obtained after
the target was wom-in for a period of time. Also, higher deposition power intensiﬁes the
preferential loss of Ti, making the ﬁlm more Ni-rich. However, the composition of ﬁlm
at off-axis position was less prone to change with the deposition power than that at on-
axis positions. Bendahan et al. [1996] demonstrated that in an r.f. magnetron process, Ni-
Ti ﬁlm composition was sensitive to the product of working gas pressure and working
distance (PA, x D). The lower the value of (P, X D) was, the less Ti the sputtered ﬁlm
contained. These results generally point to a resputtering effect dominating the
composition shift, but the situation remains uncertain.

Different strategies have been proposed to realize composition control of Ni-Ti
ﬁlm. Bendahan et al. [1996] showed that the composition of a plasma could be measured
in-situ by optical emission spectroscopy, and a small composition change could be
realized by changing the working gas pressure. It was also shown that Ni-Ti SMA
sputtering may be accomplished by using separate Ni and Ti targets. Almost arbitrary
composition of the thin ﬁlm can be obtained by changing the relative ratio of deposition
powers.

Although amorphous Ni-Ti ﬁlms may be obtained by different methods, they do
not possess shape memory properties, and a separate crystallization treatment is needed to
confer SME. A crystalline Ni-Ti ﬁlm may also be Obtained at higher substrate
temperature (in-situ crystallization). Apparently in-situ crystallization has certain
engineering advantages over the post-deposition annealing, such as shorter processing

time. But there are some other issues. Krulevitch et al. [Krulevitch, Lee, Ramsey,

40

Trevino, Hamilton and Northrup 1996] pointed out that the surface smoothness and stress
at room temperature needed to be considered. Generally, ﬁlms sputtered at high
temperature have rougher surfaces due to thermal etching and columnar grain growth.
Because of the differential CTES between substrate and the deposit, tensile stress will be
developed during cooling down from deposition temperature. This tensile stress may

result in delamination of the ﬁlm from substrate, especially when undercuts exist.

2. 4. 2. 2. Crystallization of Amorphous NiTi Films. The amorphous structure is
metastable in NiTi at all temperatures below the melting point. When amorphous ﬁhns
are annealed at sufficiently high temperature, crystallization occurs. Crystallization is a
ﬁrst-order exothermic polymorphic phase transformation. Kim et al. [Kim, Moire and
Steveson 1986] studied the crystallization behaviors of magnetron sputtered Ni,,,Ti,4
amorphous thin ﬁlm by DSC measurement and TEM investigation, and reported that the

crystallization temperature was 791 K and the crystallization sequence was:

NiTi (amorphous) 9Ni4Ti 3 '9 ﬂ1(NiT i) + Ni4T i 3 7" ,62(NiTi) + Ni 3Ti 2

Where the ,6, and ,6; phases were the crystalline B2 matrix when the precipitates were
present.

Ishida et al. [Ishida, Sato, Takei, Nomura and Miyazaki 1996] observed the
microstructure of initially amorphous Ti,.,,Ti4,_2 ﬁlms after crystallization annealing. The
crystallization sequence was that Guinier-Preston (G.P.) zones formed ﬁrst, then a ﬁne
Ti-rich metastable b.c.t. phase appeared in the matrix, then small spherical TizNi
uniformly formed in the B2 matrix and at last TizNi grows, resulting larger precipitates
and more sporadic distribution. While small TizNi existed inside the matrix of grains,

large TizNi preferentially appeared at grain boundaries.

41

The crystallization temperature of amorphous near-equiatomic NiTi alloys
reported by different research groups ranged from 400°C to over 550°C. There are a
number of reasons for this inconsistency. First, the crystallization temperature was
composition dependent. Ti-rich ﬁlm had lower crystallization temperature than Ni-rich
ﬁlms. Secondly, the crystallization temperature was related to the heating rate for
isochronal heating, such as during DSC measurements. The higher the heating rate, the
higher the measured temperature. Thirdly, when the material was held isothermally, the
crystallization temperature could be substantially lower if sufficient annealing time was
allowed. Zhang [1997] observed the occurrence of crystallization Ni-rich Ni-Ti at 400°C
annealing temperature using a wafer curvature measurement. The activation energy for
crystallization was 477 kJ/mol.

There have been efforts to decrease crystallization temperature since substrate
material may not survive high temperature exposure. Madsen and J ardine [1993]
reported that by heavily cold working of the amorphous binary Ni-Ti, the crystallization
temperature may be dramatically dropped to as low as 380°C. That was attributed to the

large number of defects and stored strain energy introduced by cold working.

2. 4. 2. 3. Annealing Induced Embrittlement of Amorphous Materials. Since post-
deposition heat treatment is indispensable for as-deposited amorphous Ni-Ti ﬁhns to
become shape memory thin ﬁlms, the behavior of the amorphous ﬁlm upon annealing is
important. Ishida et al. [Ishida, Takei and Miyazaki 1993] found that that volumetric
defects in amorphous Ni-Ti ﬁlms were most likely to survive the subsequent
crystallization annealing, therefore, poor-quality amorphous ﬁlms turned out to be brittle
ﬁlms after crystallized. It was also found that deposition parameters, mainly working gas
pressure and deposition power, affected the porosity of amorphous ﬁhns. At higher
deposition power and/or lower Ar pressure, the resulted ﬁlms appeared featureless (i. e. ,

ﬁrlly dense) under SEM, suggesting a low-porosity good-quality ﬁlms. On the contrary,

42

at lower cathode power and/or high working gas pressure, the resulted ﬁlms appeared
columnar, suggesting porous low-quality ﬁlms. The results were related to in-s'itu
densiﬁcation which was more complete with the help of energetic bombardment of
reﬂected neutrals.

There have been widespread studies showing that amorphous materials (or
metallic glasses) may become brittle during annealing prior to the occurrence of
crystallization.” The embrittlement was attributed to various structural relaxation
phenomena such as phase separation [Haasen 1983, Gruene, Oehring, Wagner and
Haasen 1985, Yamasaki and Ogino 1986], element separation [Qu and Xie 1991],
formation of microinhomogeneities [Glezer, Molotilov and Utevskaya,1984], evolution
of short-range order of atoms [Wu and Spaepen 1990, Niikura, Tsai, Inoue and
Masumoto 1993] and stress localization [Li, Quan and Hu 1997].

In their study on rapid quenched amorphous TimBe40Zrl0 and Ni,,,Nb39 ribbons,
Gruene et al. [1985] observed spatially extended compositional waves in aS-quenched
Ti,,,Be,,,,Zr,0 alloys. The amplitudes and wavelengths increased upon aging below the
crystallization temperatures. At the same, such compositional variation was not observed
in Ni,,,Nb39 alloys. An embrittlement induced by annealing in Ti,,,Be,0Zrlo alloys was
linked to the compositional waves that were aggravated at higher temperatures.

Redistribution of elements along the through-thickness direction was correlated to
embrittlement in Ni,,,P20 metal-metalloid amorphous ribbons. Qu and Xie [1991]
detected, by Auger Electron Spectrometry, that P was enriched on the free surface and
depleted in the subsurface layers. The amplitude of P enrichment and the thickness of the
depletion zone increased with annealing temperature. Planar defects were observed by
high-resolution TEM, and these defects were believed to be caused by element
redistribution and had a negative impact on the material’s ductility.

Glezer et al. [1984] observed another form of defects, nricroinhomogeneities, in a

 

“ A review on this topic is presented here because it is related to the experimental results.

43

series of F e-B-Ni amorphous samples by SEM and TEM. The microinhomogeneities,
presumed to have different compositions than the matrix, formed during annealing as a
result of structural relaxation. The degree of embrittlement was correlated to the spatial
distribution of these defects. Additional elements such as Ce and Sb were introduced to
the alloys so the formation of these defects was inhibited, thereby the embrittlement can
be alleviated.

Short-range ordering is another possible cause for embrittlement of amorphous
materials induced by annealing. There are two kinds of short-range ordering, topological
Short-range ordering and chemical Short-range ordering. Topological short-range
ordering is a densiﬁcation process in which the excess free volume is annealed out
[Sonius, Thijsse and Beukel 1983]. It is an irreversible process and can be kinetically
described by a single activation energy [Muraleedharan 1991]. Wu and Spaepen [1990]
measured the ductile-to-brittle transition temperature of a Fe-B-Si-C glassy material and
correlated the transition temperature with the change in enthalpy due to structure
relaxation. They developed a phenomenological model to describe the embrittlement
phenomenon based on the change in free volume during relaxation and a minimum free
volume for ductility.

In contrast to topological short-range ordering, chemical short-range ordering
involves a rearrangement of constituent atoms and may be reversible on temperature
changes if atomic mobility is high enough. Kinetic description of chemical short-range
ordering requires an activation energy spectrum. Both topological Short-range ordering
and chemical short-range ordering contribute to the enthalpy change. In general the
contribution from chemical Short-ordering is overshadowed by that from topological
short-range ordering. However, the sensitivity of the enthalpy change on composition
may suggest that the occurrence of chemical short-range ordering. For example, Niikura
et al. [1993] measured the enthalpy change for structural relaxation in two alloy series,

Mg9,,_,,Cu,,Y10 and Mg90,,Cu,oY,, of amorphous ribbons, and the enthalpy change was

44

correlated to chemical short-range ordering due to a change in bonding state (Mg-Cu and
Mg-Y).

In summary, structural relaxation of amorphous materials during annealing can
cause a marked loss of ductility and there are numerous mechanisms attributing to the
embrittlement phenomenon. Although the atomistic basis of metallic glass embrittlement
remains the subject of much debate, empirically, the factors favoring annealing induced
embrittlement are well known: a major metallic constituent that has the b.c.c. structure in
the crystalline state, a high metalloid content, testing at low temperature or high strain
rate, and most importantly, annealing below the crystallization temperature [Wu and

Spaepen 1985].

2.4.2.4. Transformations in Crystalline Ni T i Films. As indicated above, the
microstructure of as-crystallized Ni-Ti ﬁlm is typically a B2 matrix with possible
precipitates. Displacive transformations from B2 -) (R phase) 9 B19’ martensite may
take place upon cooling, and reverse transformations take place upon heating, the same
way as in bulk Ni-Ti material. The structures of austenite, R-phase and martensite are the
same as in bulk material, and the transformation temperatures are also comparable to
those of bulk materials.

While the displacive transformations are frmdamental for shape memory
properties, the second phases have profound effect on the SME properties. However, the
possible second phases were distinctively different for Ni-rich and Ti-rich ﬁlms. For Ni-
rich Ni-Ti ﬁlms, the most important second phase is Ni4Ti3, which forms at intermediate
annealing temperature between 350°C to 600°C. At temperatures higher than 650°C, Ni-
rich alloys become solid solutions [N ishida et al. 1986]. Rapid quenching from the
solution treatment temperature inhibited the formation of Ni4Ti3 and the resulting
microstructure was the B2 phase (at temperatures higher than M,) with supersaturated Ni.

Due to this fact, M, was far below room temperature and the R-phase transformation was

45

absent. Aging at intermediate temperature after solution treatment resulted in the
precipitation of Ni,Ti,. Ishida et al. observed progressive grth of the precipitates at
different aging temperatures and times in Ni,,,3Ti,,,, thin ﬁlms, then correlated the size of
the precipitate with shape memory behaviors. The dependence of transformation
temperatures on precipitate size was demonstrated. In general, larger precipitates resulted
in higher transformation temperatures, and the effect of precipitate size on martensitic
transformation was more evident than on R-phase transformation. In addition, precipitate
density in the matrix resulting from different annealing temperatures had an impact on
transformation temperatures. This may partly result from composition differences in the
matrix, as the over-saturated Ni decreased the M, temperature dramatically. Also, the
transformation hysteresis (A, - M,) has a monotonic relationship with the precipitate
diameter, i. e., the hysteresis decreases rapidly with the increase of precipitate diameter
when the precipitate diameter is smaller than 100 nm, and it is quite stable when the
precipitate diameter is larger than 100 nm. This is an indication that the coherency of the
precipitates with the matrix is a strong barrier to martensitic transfonnation.

Bulk (melt-solidiﬁed) Ti-rich TiNi binary SMAS are of less industrial interest than
the Ni-rich counterpart because of inferior mechanical prOperties and poorer workability
that result from TiZNi second-phase particles that usually appear on grain boundaries.
However, Ti-rich thin ﬁlms behave differently than their bulk counterparts when the
former are crystallized from the amorphous condition. Ishida et al. studied the effect of
heat treatment at annealing temperatures ranging from 500°C to 700°C for different times
on an initially amorphous Ti,,,7Ni,,,,3 ﬁlm [Ishida, Sato, Takei and Miyazaki 1995]. After
crystallization treatment, TizNi second phases were formed. Usually the TizNi phases
appeared to be evenly distributed inside the matrix grains, which was distinctively
different in the situation for melting-solidiﬁed bulk material. Essentially this is because

Ti can be super-saturated in amorphous TiNi and caused to precipitate out at low

46

temperature. Only at extreme annealing conditions (such as high temperature at 700°C,
long time at 1000 hrs), do the precipitates grow preferentially at grain boundaries.

Unlike Ti-rich bulk material, crystalline Ti-rich ﬁlms obtained by annealing of the
amorphous phase display an R-phase transformation as well as martensitic transformation
during cooling from austenite. Ishida et al. [1995] reported that martensitic
transformation temperatures increased with the increasing annealing temperature and time
with the annealing temperature ranging from 500°C to 700°C. Kawamura et al.
[Kawamura, Gyobu, Horikawa and Saburi 1996] studied extensively both martensitic
transformation temperature and R-phase transformation temperature. They found that R-
phase transformation temperature is evidently more stable than the martensitic
transformation temperature. In their unpublished work, Zhang and Grummon [1998]
studied the effect of heat treatment on the transformation temperatures of originally
amorphous Ti-rich ﬁlms. It was found that the transformation temperatures were
increased as the annealing temperatures increased from 500°C to 700°C. In the mean
time, the transformation hysteresis was decreased.

Recently, a non-equilibrium phase was found in the Ti-rich thin ﬁlms when the
materials were heat-treated at temperatures a little above the crystallization temperature,
which had been never found in the bulk materials. Under transmission electron
microscope, the new phase, which was identiﬁed as a b.c.t. phase, appeared to be disk-
shaped platelets Of Ti-rich composition with sub-nanometer thickness and 5 to 10 nm
diameter. They strongly suppressed martensitic transformation. High-resolution TEM
revealed that they were formed at the interface between crystallized domains during the
crystallization process. It was also found that these non-equilibrium phases were formed
at low annealing temperature in an initially amorphous ﬁlm for a fairly long annealing
time (for example, 745 K for 1 hour). At relatively high annealing temperature, the very
early stage (for example, 773 K for 10 minutes), a platelet second phase can be formed.

With increase in annealing time, spherical precipitates started to appear. Those spherical

47

precipitates were identiﬁed as TizNi, and the nuclei formed evenly in the sample. AS the
annealing time increased, they grew preferentially along the grain boundaries.

It is well-known that the crystalline Ni-Ti ﬁlms are obtainable at higher substrate
temperature [Ikuta et al. 1990, Gisser et al. 1992, Hou and Grummon 1995, Krulevitch et
al. 1996]. As-deposited ﬁlms possessed SME without necessity Of any additional thermal
treatment. Hou [1998] has studied systematically the phase transformation and shape
memory behaviors of Ni-rich NiTi binary thin ﬁlm deposited at elevated substrate
temperatures. He developed a model showing the structure/microstructure dependence on
the deposition temperature, which was believed to the dominating parameter. According
to Hou, by using a Ni-rich target, magnetron sputtered Ni-rich ﬁlm was amorphous when
the substrate temperature was below 300°C. With increased substrate temperature, very
ﬁne grain started to form inside the amorphous matrix. Grains appeared to be columnar,
and the diameter of the grains increased as the thickness of thin ﬁlm continuously grew.
Therefore, although the diameter of columnar grains may be relatively large on the
growing surface, it was very small at areas adjacent to the substrate. There was a
tendency that the thin layer of deposit next to substrate surface was still amorphous even
the substrate temperature was high enough to produce crystalline structure elsewhere.
Further increasing the substrate temperature resulted in larger columnar grain diameter
(on the growing surface) and Ni,Ti3 precipitates started to form inside the grain, and the
size of precipitates was larger at higher substrate temperature.

One of the important results in Hou's work was that crystalline ﬁlm could be
obtained at substrate temperature considerably lower than the generally reported
crystallization temperature. Hou attributed the surface diffusion during deposition to this
temperature difference compared to bulk diffusion in the other method. As we know, the
activation energy for diffusion is different for different diffusion mechanisms. Because
there is less lattice resistance, surface diffusion is easier than bulk diffusion, therefore, the

activation energy is smaller, and it lead to a crystalline ﬁlm growth at lower temperature.

48

Other mechanisms for thin ﬁhn growth become operational at higher temperature. Large
grains and large precipitates at high substrate temperature were mainly associated with

bulk diffusion occurring during the sputtering process.

2. 4.2.5. Mechanical and SME Properties. The properties of Ni-Ti shape memory
thin ﬁlms are comparable to those of bulk Ni-Ti. The ultimate elongation can reach more
than 10% in Ni-rich ﬁlms and can be as high as 40% for nominally equiatomic
composition. The tensile fracture stress is usually over 700 MPa for nominal equiatomic
composition and can reach 1.7 GPa for Ni-rich compositions. The recoverable strain
associated with martensitic transformation nearly 6%, and the maximum stress for full
strain recovery over 500 MPa can be achieved.

Like bulk materials, Ni-rich ﬁlms are more extensively studied than Ti-rich ﬁlms.
This is because the precipitate in Ni-rich ﬁlms, Ni4Ti3, helps to optimize the shape
memory properties. Ishida et al. studied effect of Ni4Ti3 precipitates, particularly the size
of the precipitates, on shape-memory characteristics in originally amorphous NimTiw,7
ﬁlms [Ishida et al. 1996]. Note that the grain size has not been changed appreciably aﬁer
different aging procedure so this effect was excluded. It was found that when the
precipitates were small, the critical stress for slip could be over 500 MPa, but it decreased
rapidly to below 300 MPa when the precipitates grew to 100 nm in diameter. Further
growth of the precipitates resulted in even lower critical stress, but in a less effective
manner. Maximum recoverable strain was also higher for the samples with smaller
precipitates, but it also depended on annealing temperature. It was suggested that the
dependence of aging temperature is associated with the size and spatial distribution of
precipitates.

Mechanical and shape-memory properties of Ti-rich thin ﬁlms that crystallized
from originally amorphous samples were also studied. Surprisingly Ti-rich ﬁlms were

not as brittle as their melt-solidiﬁed bulk counterparts. Ductile SMA ﬁhns were obtained

49

with titanium as high as 56.1at.%. The reason is because the distribution of second phase
TizNi was different in bulk material and thin ﬁlm samples. In thin ﬁlms, precipitates
were evenly distributed inside the grains, while they formed preferentially on grain
boundaries in the melt solidiﬁed bulk samples, which was the cause for embrittlement.
Using thermal cycling at constant load, Ishida et al. [1996] demonstrated the annealed
Ni,,_3Ti,,_7 ﬁlms had recoverable strain over 4% and the maximum stress for full strain
recovery up to 400 MPa. Due to the good ductility of Ti-rich ﬁlms, they are favored over
Ni-rich ﬁlms under some circumstances because both relatively high transformation
temperatures and good shape memory properties can be obtained, and Ti-rich ﬁlms are
more tolerant of small errors in composition.

Ogawa et al. observed the precipitation sequence of Ti-rich Ni-Ti ﬁlms and
correlated the precipitate characteristics with the shape memory properties [Ogawa, Sato
and Miyazaki 1997]. As described in section 242.3, second phases having morphology of
disk-shaped platelets were found in Ti-rich ﬁlms crystallized at lower temperature, and
they evolved into TizNi precipitates at higher annealing temperature. Disk-shaped
platelets formed inside the matrix during annealing amorphous ﬁlms for 1 hour at 400°C.
The recoverable strain was near 4%, but critical stress for slip was the highest (at 250
MPa). Increasing the annealing temperature to 500°C caused very ﬁne spherical
precipitates of TizNi to form inside the matrix. Correspondingly, the recoverable strain
increased to the maximum at near 6% but the critical stress for slip decreased to about
100 MPa. When the annealing temperature was increased to 600°C, large TizNi
precipitates formed along the grain boundaries. At that time, the recoverable strain
decreased back to 4%. Therefore, annealing temperature higher than 550°C was not
recommended because maximum recoverable strain and critical stress for slip were
decreased.

Ni-Ti ﬁhns deposited at elevated substrate temperature showed SME without

necessity of crystallization annealing. Hou conducted tensile tests of the Ni-rich ﬁlms at

50

different temperatures. He found that at low temperature when the material was
martensitic, there was a plateau-like segment on the stress-strain curve and the strain
associated with the plateau-like segment was up to 4.3%. At high temperature when the

materials were austenitic, superelastic properties were demonstrated.

2. 4. 2. 6. MT i SMA Films in MEMS and Related Research. SMA thin ﬁlms are
regarded as potential materials for nricroactuators in microelectromechanical systems due
to their robustness and large recoverable strain. Busch and Johnson proposed a method to
fabricate silicon-base NiTi SMA microactuator/microvalve using a Si etch process
[Johnson and Ray 1994]. In their method, thin ﬁlms of amorphous Ni-Ti were sputtered
onto Si wafer and crystallized at high temperature. Photolithography was used to pattern
both sides of the wafer and then SMA etching and Si etching were performed. Finally a
device having a rectangular Si frame, a narrow NiTi SMA strip running through middle
of the frame, and a small Si island at the center of the frame was made. In operating the
device, dc. current is applied to the SMA thin ﬁlm to heat the material, and on-and-off
switching can change the material structure between austenite and martensite. When the
SMA is stressed such as by virtue of a small biasing spring located in the middle, the
center island can move up and down by supplying on—and-Off electrical power (square
wave), therefore it can be used as a microactuator or microvalve. The structure of a
prototype of NiTi microvalve is Shown in Figure 2.10.

The displacement of the center island depends on the length of SMA strip as well
as recoverable strain. Assuming that initially the SMA strip does not sag nor under any
tension, cyclic displacement due to reversible transformation can be estimated as (Figure

2.11):

 

y=§(\/(l+a‘o +5)2 —1—,/(1+g,)2 —1 (2.3)

51

where y is the vertical displacement, L is the length of'the SMA strip, so is the
conventional elastic strain resulting from applied force F ,‘2 and a is the recovered strain
resulting from reversible martensitic transformation under biasing load F. It may be
noted that, although the force F is constant, the stresses are different at high-temperature
and low-temperature positions due to different deﬂection angle. The applied force to the
sample, which may be converted to uniaxial stress (a) if the dimensions of the SMA strip

are known, can be calculated using similar triangular principle:

l+co+a, F
= X
2-\/(l+so+a,)2—l 1"”

 

a (2.4)

 

where F is the force applied vertically to the SMA strip, t and w are the ﬁlm thickness
and ﬁlm width, respectively.

Estimation using equation (2.3) tells us that considerable displacement can be
realized even when the lateral dimension of the actuator is rather small. For example,
assuming the effective length of the thin ﬁlm strip is 2mm, elastic strain (80) due to
applied force F is 0.5% and the recoverable strain a is 2%, the displacement at the center
of the strip is about 124 microns.

Ion-beam sputtering is another way to form Ni-Ti SMA ﬁlms, and crystalline
ﬁlms can be obtained at relatively low substrate temperature. In a study reported by
Tsuchiya et al. [1998], near—equiatomic Ni-Ti ﬁlms were deposited using Kaufman-type
gridded ion beam source to bombard a composite sectored Ti/N i target at a 45° angle

through an aperture. During deposition, no external heating was applied to the substrate.

 

'2 The elastic modulus of Ni-Ti SMA is a function of temperature/phase. Therefore, temperature change
also results in different elastic strains if the applied force is constant, and this change also causes cyclic
displacement of central island in the microactuator assembly. However, the change in elastic strain is
typically one order of magnitude less than the transformation strain.

52

However, due to the energy generated by ion beam bombardment, the substrate
temperature increased to 170°C. The aS-deposited thin ﬁlms were crystalline, and they
showed R-phase and martensitic transformations as well as shape memory properties.
This result, which was in agreement with an earlier experiment conducted by Grummon’s
group [Walles, Chang and Grummon 1992], suggested that by using ion beam sputter
deposition, the as-deposited NiTi thin ﬁlms possessed SME without additional heating of
substrate. Thus, polyimide substrates can be used. However, one of the major
restrictions was the low deposition rate comparing to d.c. or r.f. magnetron sputter
deposition. At the ion beam current density about 1 mA/cmz, the deposition rate was
only about 1 micron in 1 hour of deposition.

Grummon et al. suggested a novel way to bias the martensite formation to create
cyclic displacement in a SMA thin ﬁlm [Grummon and Gotthardt 2000]. By heavily
implanting N+ ions into a pre-strained martensitic Ni-Ti thin ﬁlm, the top layer of the
material was arnorphized. This amorphous layer did not have SME, but could provide
biasing force to the martensite formation. Cyclic out-of-plane shape change upon heating
and cooling was demonstrated in the ion implanted Ni-Ti SMA thin ﬁlm bimorph

structure.

2.4.3. High-Temperature Shape Memory Alloy Thin Films

High transition temperature SMA thin ﬁlms have been of great interest due to the
limitation of transition temperature of binary NiTi SMAS. However, there have been
only a few successful attempts that have been reported. Johnson et al. [1995] studied
alloy systems that potentially have A, temperatures over 100°C. In their study, three
kinds of alloy thin ﬁlms, Cu-Al-Ni, Ni-Al and Ni-Ti-Hf, were sputter deposited at
without substrate heating. No martensitic phase transformation was detected in Cu-Al-Ni

thin ﬁlms after traditional annealing treatment. Martensitic phase transformation was

53

detected in Ni-Al thin ﬁlms but shape memory was not observed due to brittleness. Ni-
Ti-Hf thin ﬁlms showed some promise because the annealed ﬁhns were relatively ductile
and had high transformation temperatures. Quandt et al. [1995] and Miyazaki et al.
[1995] reported their work on Ti-Ni-Pd SMA thin ﬁlms. Both had limited success. SO
far only Ni-Ti-Hf and Ti-Ni-Pd alloys are regarded as promising candidates for high-
temperature SMA thin ﬁlms.

2. 4. 3. 1. Ni-Ti-Pd SMA Thin Films. There are only a few reports in the literature
concerning Ni-Ti-Pd HTSMA thin ﬁlms. Quandt et al. [1995] Obtained the Ni-Ti-Pd thin
ﬁlms by post-deposition annealing of the as-sputtered amorphous ﬁlms, and the thin
ﬁlms showed two-way shape memory effect. But no data were presented to Show phase
transformation and detailed Shape memory properties of the thin ﬁhns. Miyazaki et al.
[Miyazaki, Hashinaga, Yumikura, Horikawa, Ueki and Ishida 1995] also investigated Ni-
Ti-Pd SMA ﬁlms. The transformation temperatures and shape memory properties were
studied more extensively.

Miyazaki et al. [1995] reported several results that were worthy of mention. First,
they found that the composition of the ternary thin ﬁlms was slightly Ti rich, i. e. , the
atomic percent of titanium should be between 50% to 51%. By doing this, the
transformation temperatures of thin ﬁlms could be effectively increased by the ternary
element while the mechanical properties remained satisfactory. Secondly, transformation
temperatures (particularly M, temperature) of the 700°C annealed ﬁlms depended on the
atomic percentage of palladium. With increasing Pd content from 0 to 6at.% the M,
decreased Slightly, but with continued increasing Pd content, the M, temperature
increased rapidly. With 21 .8at.% of palladium, the M, temperature reached 117°C. Also,
the transformation hysteresis was decreased due to additional palladium. With more than
6at.% of palladium, the hysteresis was generally lower than 20°C, and sometimes was

below 10°C. These results were generally in agreement with those reported for bulk Ti-

54

Ni-Pd research [Pu et al. 1994]. Thirdly, good shape memory properties were found in
those high-temperature thin ﬁlms treated by secondary annealing. By thermal cycling the
thin ﬁlms under applied stresses that were stepwise increased for different cycles, the
shape memory effect was demonstrated at stresses as low as 30 MPa. About 2% strain
was produced when the sample was cooled under 150 MPa of applied stress, and
maximum recoverable strain was about 2.5%. Due to the solution hardening by
palladium, the critical stress for slip was close to 300 MPa (Figure 2.12).

Pre-martensitic transformation and martensite structure in Ni-Ti-Pd thin ﬁlms
were also related to palladium content, as reported by Miyazaki et al. [Miyazaki and
Ishida 1999]. When the palladium content was less than 6at.% or so, R-phase
transformation occurred prior to the martensitic transformation. The martensite was
monoclinic in ﬁlms having 6a.t% Pd or less. When the palladium content was higher
than 6at.%, R-phase transformation was absent, instead, on cooling, austenite transformed
to orthorhombic martensite, which transformed to monoclinic martensite upon further
cooling. As Pd content increased above 6at.%, the start temperature for austenite to
orthorhombic martensite increased. At the same time, the start temperature for
orthorhombic martensite to monoclinic martensite transformation decreased rapidly. At
palladium contents more than 15%, the start temperature of monoclinic martensite was
lower than -120°C, so normally this transformation was not considered.

Both Miyazaki and Quandt showed that Ti-Ni-Pd had smaller hysteresis than Ti-
Ni binary alloys. Winzek and Quandt [1999] also studied the hysteresis of Ti-Ni-Pd ﬁlms
under different constraint conditions", which were realized by depositing the SMA onto
substrates (thermal silicon oxide, 50-micron thick molybdenum and Fe-,2Cr,,Nilo alloy)

with different coefficients of thermal expansion. SMA ﬁlms along with their substrates

 

'3 Transformational hysteresis is generally deﬁned as A, - M,. However, in the stress-temperature
hysteresis loop or strain-temperature hysteresis loop, the transformation hysteresis is typically measured as
the difference in temperatures at which the same stress or strain level is obtained during forward
transformation and reverse transformation, respectively (Figure 2.1).

55

were annealed 1 hour at 700°C for crystallization. Curvature of the composite was
measured during thermal cycles to get hysteresis loops. It was found that the hysteresis
was smaller under substrate constraint, no matter whether tensile or compressive, than in
free-standing form. The smallest hysteresis (nearly none to 35°C) was found in a 10at.%
Pd ﬁlm and under compressive stress. However, substrate constraint did not change the
temperature difference between A, and M, temperatures.

Fatigue of Ti-Ni-Pd SMA thin ﬁlms was also evaluated by Winzek and Quandt
[1999] using curvature measurement during thermal cycling. The stress change resulting
from transformation continuously decreased due to thermal cycling. It was reduced to
half of the initial value aﬂer the frrst 100 cycles, although it became relatively stable upon
further cycling. It was noted that the fatigue behavior of Ti-Ni-Cu thin ﬁlms was much
better than that of Ti-Ni-Pd, the former showing no reduction of stress change from

transformation for the ﬁrst 2000 cycles.

2. 4.3.2. Ni-Ti-HfSMA Thin Films. Material cost has been a consideration for high
temperature alloys containing noble metals such as Au, Pt and Pd. Therefore, alternative
ternary alloys that substitute Zr or Hf for Ti have been attractive. Due to the success of
making NiTin high-temperature shape memory alloy in bulk form, fabrication of Ni-Ti-
Hf thin ﬁlms was studied by a number of research groups. Reportedly, Johnson was the
ﬁrst to fabricate Ni-Ti-Hf high-temperature SMA thin ﬁlms [J ohnson, Martynov and
Shahoian 1995]. He sputter deposited amorphous Ni-Ti-Hf ﬁlms and then acquired SME
by a secondary annealing at 520°C and high transformation temperature was obtained.
With 19.5at.% hafnium in the thin ﬁlm, an M, temperature of about 100°C was
measured. The resulting ﬁlms appeared ductile, although no quantitative data was
presented for comparison. Independently, You et al. [You, Chung, Han, and Gu 1997]

also successfully Obtained moderately high transformation Ni-Ti-Hf SMA thin ﬁlms by

56

Sputtering amorphous ﬁlms and then annealing for crystallization. The peak temperature
for martensitic transformation was 52°C for 12.5at.% hafnium ﬁlms and it was increased
to 92°C in a ﬁlm with 20at.% hafnium. Microstructural observation indicated that the
grain size was only about 0.1 micron. Mechanical properties and shape memory
properties were not evaluated.

Besides sputter deposition, laser ablation was also investigated as a method to
fabricate Ni-Ti-Hf high-temperature SMA thin ﬁlms. In an experiment reported by Gu et
al., 6-micron thick Ni-Ti-Hf ternary thin ﬁlms having Hf content up to 13.3at.% were
produced using a KrF excimer laser ablation system with different composite targets in
high vacuum. Laser pulses had 10Hz frequency, 248nm wavelength, and 34ns pulse
duration. The laser was directed to the surface of polished composite target at an
incidence angle of 45°. The circular composite target consisted of three fan-like plates:
one TisoNiso, one pure Ni, and one pure Hf. The growth rate was lum/hr. The as-
deposited ﬁlm was amorphous and crystallization temperature was 513°C. Annealed ﬁlm
showed R-phase and martensitic transformations on cooling, but the transformation
temperatures were quite low: the forward martensitic transformation temperature was
only 31°C. Fine grain size and Ni-rich composition were believed to be the reasons for
low transformation temperatures.

In Spite of the cost advantage, Ni-Ti-Hf ternary SMA ﬁlms had some drawbacks
compared to Ti-Ni-Pd thin ﬁlms. As Shown in Johnson's reports, the hysteresis of Ni-Ti-
Hf SMA ﬁlms was about 60 to 100°C, which was considerably larger than that of Ti-Ni-
Pd ﬁlms. Wide hysteresis results in longer response time for cyclic actuation, and is
therefore not favorable for applications in microelectromechanical systems.

Although encouraging, the work by Johnson et al., You et al. and Gu et al. on
development of Ni-Ti-Hf high-temperature SMA thin ﬁlm was only rudimentary.
Johnson demonstrated that high transformation temperatures could be Obtained in Ni-Ti-

Hf thin ﬁlms while Gu showed martensitic transformation with low transition

57

temperatures. Neither group has characterized the microstructure or was able to evaluate
the mechanical and shape memory properties of the thin ﬁhns. In this thesis study, both
post-deposition annealing and in-situ deposition are used to produce crystalline Ni-Ti-Hf
thin ﬁlms that possess shape memory at high temperatures. Different annealing
procedures are investigated to optimize the mechanical properties. Microstructure
characterization and mechanical evaluation are conducted to correlate the performance

and microstructural features.

58

Chapter 3
EXPERIMENTAL PROCEDURE

This study concerns preparation of high-temperature shape-memory alloy thin
ﬁlms, characterization of their properties, and correlation of their properties with
structural and microstructural features. In SMA thin ﬁlm preparation, magnetron
sputtering has been used, and depending on the nricrostructure of the as-deposited ﬁlms, a
secondary annealing was canied out to crystallize the samples. Thermal analyses to
determine phase transformation, mechanical property tests and shape-memory property
evaluation were carried out. Structural and microstructural analyses utilizing X-ray
diffraction techniques, electron diffraction and transmission electron microscopy have
been conducted. This chapter describes in detail various experimental methods and

procedures used in this study.

3.1. Thin Film Fabrication

Experimental procedures for shape-memory alloy thin ﬁlm fabrication consist in
two parts: thin ﬁlm deposition, and post-deposition thermal treatment to give the ﬁlm
SME properties if the as-deposited ﬁlms do not have SME. For ﬁlm deposition using
magnetron sputtering, target preparation is necessary, and the quality of deposit is directly
related to that of the target. In this section, procedures of target preparation and thin ﬁlm

deposition are outlined, and thermal treatment is detailed.

59

3.1.1. Thin Film Deposition

Ternary Ni-Ti-Hf thin ﬁlms were made by planar d.c. magnetron sputtering
method in a Perkin-Elmer Ultek sputtering system. A schematic of the sputter system is
illustrated in Figure 3.1. As shown, the system was composed of a 29"-diameter vacuum
chamber, a Ni-Ti-Hf target (cathode) with magnets behind, substrate (anode) on a
substrate holder, a substrate heater, and a dc. power supply that was connected to target
and the substrate. The vacuum chamber, evacuated by an Edwards A6117DC36A
mechanical pump and an Ultek-12 cryO-pump, could be pumped to 10'7 torr. The target
was 8 inches in diameter and 0.25 inch thick. The substrate holder was 3.0mm thick
160mm—diameter circular molybdenum plate which provided sufﬁcient thermal
conductivity across the entire plate. The substrate heater was composed of three parallel
500W quartz lamps that were supplied with variable power to achieve different substrate
temperatures, which was measured by an ungrounded exposed K type thermocouple
located near the center of the substrate. Although the quartz-lamp heater may create non-
uniform heating, the massive molybdenum plate helped achieve a temperature uniformity
better than i=5°C across 5 inch diameter area. The DC. power supply had maximum

capacity of 5 kW.

Thin ﬁlm deposition was made after about 10*5 torr base vacuum was achieved.
99.999%-purity Kr was introduced during sputtering as working gas, and the pressure of
working gas was controlled by MKS 1259B mass ﬂow controller and measured by an
MKS 127 Barometer. The composition of gas in vacuum was analyzed by an ion pump
backed UTI 100C reduced-pressure RGA (Residual Gas Analyzer) before and after the

introduction Of the working gas.

Due to the differential sputtering rates of different atomic species, the composition
of the deposit is not expected to be at steady-state in the early stage of the sputtering [Lee
1994, George 1992]. Therefore, a fresh target was usually burned in for 20 hours at IkW

60

Of deposition power to achieve compositional stability. In addition, before each

deposition run, preheating was applied to outgas the substrate and vacuum chamber.

The thin ﬁlm fabrication process is determined by a number of process parameters
including base vacuum, working gas species, working gas pressure, deposition power,
substrate-to-target distance, and substrate temperature. During this study, base pressure
was below 10" torr, typically 5x10‘7 torr. The pressure of the working gas krypton 1.0
mTorr. The sputter power, which was deﬁned as the product of the voltage between
target (cathode) surface and substrate (anode) and the current, and was set constant at 1
kW during each sputter run." The working distance, the distance ﬁom the cathode to the
substrate, was set at 81mm. The substrates were 5-inch diameter thermal oxide
passivated (100) single-crystal silicon wafers. The substrate temperature was varied in
order to get desired crystallinity of the as-deposited ﬁlms. Without external substrate
heating, the plasma can gradually raise the substrate temperature up to 200°C. At the

deposition parameters described above, the sputtering rate was about 15 A/sec.

3.1.2. Hot Pressed Powder Target

The targets for this study were made by hot pressing a mixture of pure metallic
powders in the appropriate ratio by the vendor (Pure Tech Co.). There are three steps

involved:
1). Mixing the pure metallic powders and placing the mixture in a mold;

2). In protective environment of Ar, heating the mixture to about 1000°C and then

applying pressure of about 100 MPa to 200 MPa to the mixture;

 

'4 As the target is continuously worn in, the voltage decreased slightly and the current increased at constant
deposition power.

61

3). Machining the consolidated target material to appropriate dimension and

soldering it to a copper backing plate.
4). Grinding the surface of the target, and sealing it in Ar until ready to use.

The density of ﬁnal target piece was measured. The porosity of the target was
calculated by comparing the measured density with theoretical value. Table 3.1 is the
lists of those data for the targets with different compositions used in this study and shows
that the porosity in the targets was between 11.4% and 15.5% of the full density.

Table 3.1. Composition and Porosity of the Targets Used to Deposit Ni-Ti-Hf
SMA Thin Films in This Study.

 

 

Target Composition Ni,,,,Ti,,,,Hf,, N i,., ,Time" Ni,._,Ti,,,,Hf,, Ni,,,,Ti,,_,Hf,,
Theoretical Density (g/cc) 8.83 8.27 8.08 7.83
Measured Density (g/cc) 7.46 7.24 7.16 6.85
Porosity (%) 15.5 12.5 11.4 12.5

 

3.1.3. Post-Deposition Annealing

Although the ﬁlms deposited at elevated substrate temperature Showed SME at
the as-deposited state, those deposited without sufﬁcient substrate heating were
amorphous, and crystallization annealing was necessary to render a material with SME
properties. In this study, conventional vacuum annealing, as well as rapid thermal anneal
(RTA), was used to achieve different heating rates in the thermal treatment for

crystallization.

Conventional heat treatment was conducted in a custom built vacuum furnace that
was equipped with an Edwards ES/ED 200 rotary mechanical pump and a Varian VHS 6"
diffusion pump having a pumping capacity 2400 liters/sec. A schematic of the fumace is
illustrated in Figure 3.2. As shown, the furnace consisted in two 10 by 12 inch 5 kW

62

heating plates, which were placed about 2 inches apart. The Ni-Ti-Hf samples, which are
in free-standing form, were housed inside 5.5 x 5.5 x 0.5” pure titanium box, which was
situated between the two heating plates during the armealing treatment. A K-type
thermocouple was placed on each side of the Ti box. Temperature control was
accomplished by time proportional cycle, which was realized by an Omega CN3 000
temperature dual input/output controller. The vacuum of the chamber was maintained
below 5x10" torr during the annealing process. After annealing was completed, the
sample remained as shiny as before annealing, suggesting that the vacuum and protective

titanium box were effective to protect the sample from oxidation.

Due to radiative heat loss, the maximum heating rate was limited and was
temperature dependent with slower heating rate at higher temperature. For the
conventional vacuum annealing system described above, the maximum achievable
heating rate was about 40°C/min at 400°C. The RTA technique (radiant type thermal
annealing), which provided much higher heating rate, was also used to crystallize the
amorphous Ni-Ti-Hf ﬁlms. During the RTA treatment, the sample was attached to the
substrate. A Heatpulse 410 system was used and the procedure included the following
steps:

1). Program the temperature proﬁle for the annealing treatment;

2). Load the wafer sample into the system;

3). Purge the system using 99.99% Ar at a ﬂow about 200 cc/min for about 60

seconds;
4). Apply power to the quartz lamps and heat the sample according to the program;

5). Remove sample from the oven when the reading of sample temperature is below

1 00°C.

63

Depending on sample load, the heating rate can be as high as 300°C/min.
However, the heating rate for this experiment was chosen at 100°C/sec in order to
minimize temperature overshoot. After isothermal annealing, the samples were cooled at
50°C/sec. Only minor discoloration of the sample was observed after the RTA treatment.

Figure 3.3 shows a typical temperature proﬁle for an RTA treatment.

3.2. Thin Film Characterization
The properties of the deposited ﬁlms or annealed ﬁlms were evaluated for their
phase transformation temperature, mechanical properties and shape-memory properties.

In the following, the detailed experimental procedure for each category is described.

3.2.1. Transformation Temperatures

Martensitic transformation is a ﬁrst-order phase transformation, and is
accompanied by enthalpy, entropy, and other physical property changes. Therefore,
Differential Scanning Calorimetry (DSC) has been used to measure to
endothermic/exothermic effects, and the transformation temperatures can be determined
from the endothermic/exothermic peaks. DSC measurements were conducted on a Perkin
Ehner 7 Thermal Analysis System. To ensure good resolution, the weight of the sample
was at least 10mg. The heating/cooling rate was set to 20°C/min, and the temperature
ranges were from ambient temperature up to 350°C. A typical DSC curve for an SMA
thin ﬁlm is shown in Figure 2.1(b). As indicated in the ﬁgure, the transformation
temperatures were deﬁned by the tangent-intercept method, and they may vary slightly
according to the segment that was chosen to extrapolate. However, the endothermic and

exothermic peak temperatures (M, and A,) were well deﬁned.

64

Electrical resistivity measurement was also used to determine the transformation
temperatures. The experimental set-up is schematically shown in Figure 3.4. The
samples, preferably in the form of strips, were connected to a DC power supply (in
Constant Voltage mode) with a 3k!) resistor in series, and the potential drop across the
sample was recorded by a Data Acquisition System (DASZOOO). Because the sample
resistance (<2 Q for typical sample Size) is much smaller than 3kg, the electrical current
passing through the sample was virtually constant regardless varying electrical resistivity
of the sample resulting from varying temperature". The potential drop across the sample
is proportional to the sample resistivity, so variation of the voltage reﬂects the resistivity
change with the temperature. During the test, the sample was put into a 12-inch long
circular ﬁrmace. The temperature of the sample was measured by a K-type thermocouple
placed nearby and it was also recorded by a Data Acquisition System. A CN4800 Omega
controller was used to control the temperature of the sample. For a typical set-up, the
voltage of the DC power supply was about 10V, heating rate was 15°C/min, and
temperature ranged from room temperature to 300°C. During cooling, the furnace was
turned Off so the sample cooled down naturally. Figure 2.1(a) is the typical resistance-
temperature curve. The tangent intercept method was used to determine the

transformation temperature.

3.2.2. Mechanical Property Measurement

Mechanical properties of the SMA thin ﬁlms were mostly measured by tensile
testing on a 5564 Instron system that is capable of conducting tensile test at temperatures
up to 350°C. Various temperature can be realized by programming an Eurotherrn 2408

temperature controller. The load cell has a maximum load capacity of 50 Newtons with

 

'5 In this setup, the current ﬂowing through the sample is about 3.3 mA and the power is about 22 uW.
This level of power is 3 to 4 orders of magnitude less than can cause appreciable Joule heating to the
sample. Therefore, there is no sample heating due to the current ﬂow in the resistivity measurement.

65

an accuracy of 0.025 Newton. The displacement of the cross-head had an accuracy of
measurement at about 25 microns. Test samples were free-standing ﬁlms, obtained by
manually peeling from the substrate. The sample was cut into a 70x3 mm strip and
loaded onto the tensile testing machine with gauge length set at 50 mm. In order to
prevent Slipping, the ends of the sample were attached to folded copper foil and then
clamped between stainless steel blocks having ridged faces. During the test, the sample
was elongated at 1 mm/min, or at strain rate about 0.02 mm/mm/min. Strain
measurement was usually taken by the displacement of upper grip, but a video
extensometer was also used. In the latter, a gauge length of 20 mm was marked on the
sample with black ink. The contrast signal was picked up from video so the distance

between black marks can be measured. The resolution of video extensometer was 5 pm.

For tensile tests that were conducted at elevated temperature, the ramp rate was
usually set at 10°C/min. In order to ensure temperature uniformity, the sample was held
for 5 minutes at the targeted temperature before tensile test started. Most samples were
elongated to failure to get the ultimate elongation at different temperatures, and at least 5
samples of the same kind were tested to get the average value. The stress-strain data was

collected using Merlin V4.42 software.

In some circumstances where samples were too brittle to be clamped to the
clamping ﬁxture, a bending test method was used to semi-quantitatively evaluate the
ductility of the sample. A schematic of the test method is shown in Figure 3.5. Basically
the minimum curvature radius was measured and the strain (tensile or compressive,

depending on which side) at the sample surface was calculated as:

gsurface =t/(D—t)

where t is the thickness of the sample and D is the curvature diameter of the deformed
sample. As shown in Figure 3.5, the distance of the two blocks was adjusted to be

slightly wider that the diameter of the metal wire that was used to force the strip into the

66

gap between the two blocks. The radius of the metal wire was considered to be the
bending radius. Samples were deformed at larger bending radius ﬁrst and then the
bending radius was decreased step by step until the sample failed. Because the sample
was very thin, it can be deformed to very small curvature radius even when the material

has very poor ductility.

3.2.3. Shape-Memory Property Evaluation

Shape-memory properties were evaluated by measuring the recoverability of a
prestrain, recoverable strain and maximum stress for full recovery during constrained
thermal cycling between martensite and austenite. The recoverability of a prestrain can

be measured in the following procedure and schematically showed in Figure 3.6:

1). load the sample in the 5564 Instron material test system which has capability of

varying test temperature;

2). bring the sample to the test temperature and allow sufﬁcient time for thermal
equilibrium and zero the strain measurement system (making sure that the

temperature does not exceed the austenite start temperature);
3). apply load to elongate the sample to a predetermined strain (8,0101);

4). unload the sample completely, load the sample again very slightly in tension (1

MPa), and then measure the strain (8p);

5). slacken the sample sufﬁciently by lowering the upper grip, and then heat the
sample to 50°C above the A, temperature, making sure the sample is NOT under

stress;

6). cool the sample to temperature at which the sample was elongated, allow

sufﬁcient time for thermal equilibration;

67

7). bring the sample slightly under tension (1 MPa) and measure the strain of the

sample at this time (8",) (This is the non-recoverable strain).

The strain obtained at Step 4 is the plastic strain in martensite, and the difference
between 8mm] and 8,, is the elastic strain of the martensite. Of the plastic strain in
martensite 8p, recoverable strain is 8,. = 8,, - 5nr~ Therefore, the recovery rate of the

plastic strain of martensite is:'6

g 6‘ —£
RR=_L=———P "’ x10(P/o (3.1)
8P 8P

Measuring 8p and am- at the same sample temperature eliminates the error from
thermal expansion. However, if the strain is measured from the displacement of the
upper grip, it is important to make sure that the shaﬁs are at the same temperature when
measurement is taken. Otherwise, thermal expansion of shafts of the tensile machine
could cause substantial error. To solve this problem, two methods can be used. The ﬁrst
one is to wait enough long time to let the shaﬁs to come to thermal equilibrium. This
includes both before the initial elongation (step 3) and before the measurement of residual
strain (step 7). In order to establish the waiting time that is necessary for shaﬁs to
become thermal equilibrium, a non-SMA dummy sample (such as a strip of copper foil)
was loaded on the Instron machine, and was put through every step except step 2 and
unloading. If enough elapsed time was allowed, 8p and 8”,- should be equal, and the
waiting time should be used as reference for SMA ﬁlm measurement. In the other

method, video extensometer can be used to measure the length change of the sample. A

 

'° Two-way memory effect may be introduced by prestraining the SMA material severely and repeatedly.
In this case, when the sample is cooled down from austenite at high temperature, a small amount of

elongation may occur, so the measured recovery rate may be lower than the actual one. The error may be
minimized by taking two measures: 1) keep pre-strain at small level; and 2) use fresh sample every time.

68

20 mm gauge length was marked before test started. Gauge length was recorded in step
3, 4, and 7. In doing so, there is no need to wait until the shafts to reach thermal
equilibrium if video extensometer is used to measure sample elongation, although it is
still desirable to wait sufﬁcient time to allow the sample to reach the deformation

temperature.

Measurement of recoverable strain at different applied stresses was determined by
thermal cycling at a constant load, which was conducted in a custom built experimental
setup that is schematically shown in Figure 3.7. As in the tensile test, sample strips,
which usually have dimensions of 40mm x 3mm x 4.5 pm, were used in the constant-load
test. The sample strip was loaded into the circular fumace by clamping between two
pairs of copper grips. Samples were always placed in the middle of the circular furnace
to achieve maximum temperature uniformity. Low-CTE molybdenum strips were
attached to the copper grips to extend the ends out of furnace. While the upper end of
molybdenum strip was ﬁxed, the lower end was attached to the location where the load
was applied to the sample. LVDT (Linear Voltage-Displacement Transducer) cores were
also attached to the lower end. During thermal cycling, the length of sample changed, the
position of the LVDT core was changed accordingly. The LVDT device converted the
length change into electrical signal, which was picked up and recorded in a DAS-2000
data acquisition system. Sample temperature was measured by a K-type thermocouple
that situated near the middle of the sample, and reading was also recorded. A CN3 000
Omega temperature controller is used to realize various iterations of the sample

temperature proﬁles.

The length change of sample reﬂected the strain of the sample. However, the
LVDT could only detect the overall length change resulting from sample length change
and ﬁxture length change. If ignored, thermal expansion of the ﬁxture (mainly

molybdenum extension strips) during thermal cycle could give a great deal of error. This

69

is particularly true for short samples, which were usually used to ensure good temperature
uniformity. For example, considering a situation that 10-inch molybdenum extension
Strip was heated up by 250°C uniformly, the elongation resulting ﬁ'om temperature

increase is:
AI = 5.9x10'°(mm/mm/°C) x 10 x 25.4(mm) x 250°C = 0.375 mm.

this elongation, if mistakenly considered as the elongation of a 50mm sample, would
cause an absolute error of 0.75% which would be subtracted from the overall shape
memory strain. Although thermal expansion of ﬁxture might be compensated just by
calculation, it is very complicated to do that, because it is necessary to know the
temperature distribution over the entire length of the molybdenum ﬁxture. The following
calibration procedure was used to eliminate the error from the thermal expansion Of

ﬁxture. It involves three steps:

1. For a certain temperature proﬁle running, record the displacement of the sensing
bar in LVDT assembly for a dummy sample (a molybdenum strip with the same

length of the SMA sample to be measured) during a thermal cycle;

2. Using the same temperature proﬁle, record the displacement of the LVDT sensing

element for a real SMA sample during a thermal cycle;

3. Subtract the data obtained in step 1 from the data obtained in step 2, then add the

thermal expansion compensation of the dummy sample.

For certain temperature proﬁle, the displacement of LVDT sensing bar for a
dummy sample was measured several times to ensure repeatability. It should be pointed
out that, although the displacement for the dummy sample during thermal cycles repeated
very well for the same temperature proﬁle, when the temperature proﬁle is changed, the
calibration procedure has to be repeated. This is because with different heating/cooling

rate, the temperature distribution of the molybdenum ﬁxture (with time) also changes. A

70

typical strain-temperature curve during a thermal cycle under a constant stress is shown
in Figure 3.8, where sample thermal expansion, recoverable strain and residual strain are

shown.

3.3. Compositional, Structural and Microstructural Analyses

In order to explore the underlying reason for the mechanical behavior of the
material, it is necessary to conduct phase analyses, which involve compositional,
structural and microstructural analyses. The following describes each technique used to

carry out those experiments.

3.3.1 . Composition Analysis

The composition of thin ﬁlms was analyzed by a LINK AN -1000 X-ray EDS
(Energy Dispersive Spectrum) system mounted on a JEOL S-2500 SEM (scanning
Electron Microscope) operating at 20 kV. Gain calibration was made on Co-KOI radiation
and bulk calibration was done using LINK ZAP-4 code. The analyzed samples were
usually 5 pm thick ﬁhns and were attached to Silicon substrates or glass slides. Area
scanning mode at 100x magniﬁcation was used to illuminate the sample in order to get
average compositional data over a relatively large area. It should be pointed that,
although composition analysis by EDS is expeditious, considerable error may exist.
While the accuracy of the data is usually within :20 (O is the standard deviation, usually
<0.25%), The precision is more difﬁcult to tell unless a standard alloy specimen (N i-Ti-
Hf) is measured. However, a careful calibration using pure specimens (Ni, Ti, and Hi)

may minimize the errors.

71

3.3.2. Crystallographic Structure Analysis

Crystallographic structure analysis was conducted by X-ray diffraction using a
Scintag-2000 diffractometer with accelerating voltage of 35 kV, tube current of 25 mA,
Cu-K, radiation and normal scanning mode. The scan step was 0.03°/step, the scan rate
was 0.25 to 2°/min and the scan range (20) was ﬁom 20° to 70°. All the X-ray spectra
were taken in free-standing ﬁlms, and the ﬁlms was attached to a glass Slide (amorphous)
during X-ray scanning in order to avoid extraneous peaks. Nevertheless, it was found

that with ﬁlm thickness greater than 3 pm, few X-rays from the substrate were detected.

In indexing of the X-ray diffraction Spectra, the interplanar distance (d)
corresponding to each X-ray peak was calculated by the well-known Bragg equation.
Then the measured dis compared to interplanar distance of possible phases with known
structures to ﬁnd the best match. The tolerance was usually set at 0.01A. In case that
unknown structures existed, the d value for each peak was calculated by Debye equation,
then the values of d2 were listed in sequence from small to large, and normalized by
dividing all the value by the smallest n Xaﬂ(n = 1, 2, 3, ...). If an integral sequence was
obtained, then a cubic phase was detected. In cases that more complicated phase
structure existed, the measured d values were listed in Excel spreadsheet, and a structure
similar to known structure based on literature was assumed. A list of d spacings were
calculated then compared with the measured values. The lattice parameters were

modiﬁed until a match between calculated values and measured values was obtained.

3.3.3. Microstructure Observation

The in-plane microstructure of the Ni-Ti-Hf thin ﬁlms was observed on a Hitachi
H-800 TEM operated at 200 kV. Free-standing ﬁlms were Obtained by manually peeling
the ﬁlms from substrate. Samples were cut into 3 x 3 mm dimension and then glued with

acid-resistant lacquer on 3mm-diameter nickel grids with 2 mm-diameter holes. Electron

72

transparency was achieved by twin-jet electropolishing method using a fresh electrolyte
that contained 20% nitric acid and 80% methanol in volume percent. The
electropolishing was done on a Tenupol-3TM (Struers A/S. Denmark) twin jet polisher
operated at 8V (which gave a current density of 12mA/mm2). The temperature of
electrolyte was kept between 233 K and 243 K by continuously introducing liquid

nitrogen into the electrolyte.

Selected-Area Electron Diffraction (SAED) was also used to determine phase
structure. SAED was done on a Hitachi H-800 Transmission Electron Microscope
(TEM) The TEM was operated at 200 kV acceleration voltage and the sample was
installed in a double-tilting specimen holder. Indexing of electron diffraction spot/ring
was done by comparing the d value measured on the pattern with those calculated from
known structures. Camera length was calibrated from the diffraction pattern of phase

determined by X-ray diffraction.

73

Chapter 4
RESULTS AND DISCUSSION

The main purpose of this research is to ﬁnd a method to produce Ni-Ti-Hf thin
ﬁlms having high transformation temperatures, good shape memory properties and good
mechanical properties that are useful for industrial applications. Additional objectives are
to correlate the properties of the shape memory thin ﬁlms with details of the fabrication
process, to characterize the microstructure of SMA thin ﬁlms with different properties,
and to improve the material properties through process selection based on the
characterization results. Our ﬁrst goal was to achieve suitable crystallinity of the thin
ﬁlm, and it was accomplished by either sputtering in-situ crystallized material, or
applying post-deposition thermal treatment to crystallize the aS-deposited amorphous
material. The in-situ crystallized thin ﬁlms were obtained by choosing proper deposition
parameters, particularly by using sufﬁciently high substrate temperature. Because
crystalline phase formed during deposition, this procedure was also called in-situ
crystallization. The post-deposition thermal treatment for crystallization was a high
temperature anneal, either isothermal or isochronal. In this thesis, two classes of
deposition technique, one characterized by relatively low substrate temperature and the
other by relatively high substrate temperature, were used to control the crystallinity of the
as-deposited samples. The as-deposited material was either ﬁrlly amorphous or ﬁrlly
crystalline '7. For those having fully amorphous structure, separate annealing was carried

for crystallization. In the following, crystalline Ni-Ti-Hf ﬁlms Obtained by post-

 

” Partially crystalline samples can be obtained using intermediate substrate heating. In this study,
however, the partially crystalline samples were generally brittle, and this prevented us from doing
extensive evaluation on their mechanical and shape-memory properties.

74

deposition annealing and those by in-situ crystallization will be discussed separately.
Results for each process class will be presented before proceeding to a discussion of their

implications.

4.1. Films Crystallized by Post-Deposition Annealing

A number of Ni-Ti-Hf ﬁlms with different compositions were sputter deposited
using different target compositions. With deposition parameters described in Section
3.1.1 (working gas Kr pressure 1.0 mTorr, working distance 81 mm, deposition power 1
kW, and substrate temperature 200 a: 15°C), the composition of the thin ﬁlms was slightly
different than their targets. Table 4.1 shows the composition of ﬁlms at the central area
deposited from different targets. Measurement was taken after the target were worn
about (equivalently) 20 hours at 1 kW of deposition power (about 1/5 to 1/6 lifetime of a
target). In general, there was more nickel in the ﬁlms in the central area than in the target
and less titanium. Hafnium content was mostly higher in the ﬁlms than in the target.
This composition shift may be related to combined effects of preferential loss of some
species of atoms due to collisions with background gases during atom transport from the
target to the substrate, and preferential resputtering of the growing ﬁlm [Fursenko,
Galjukov, Makarov, Lutovinov and Rarnm 1995].

Table 4.1. Target Compositions vs. Film Compositions

 

Targetc°mpositi°n Ni.7..Time2. Ni,,,n,,,m., Ni,,,Ti,,_,Hf,, Ni,,,n,,,m,,
Film C0mp05iti°n Ni48.7Ti29.2Hf22.1 Ni4a.aT133.4Hf17.s Ni497Ti339Hf164 Ni491Ti36 2Hfr47

 

 

ANi +1.2 +1.3 +1.2 +0.6
ATi -3.3 -l.l -l.6 -l.3
AHf +2.1 -0.2 +0.6 +0.7

 

75

The composition uniformity of the thin ﬁlms (deposited from the Ni,,,,Ti37,,HfM
target) across 5”-diameter wafer was evaluated and the results were shown in Figure 4.1.
It can be seem that within 3” diameter, the compositions were very uniform (deviation
<0.5%). More deviation was measured at locations outside of 3”-diameter area.
Therefore, in order to guarantee the maximtun composition uniformity, only ﬁlms within
the 3”-diameter limit were used to conduct further study. The non-unifonnity of
composition is generally associated with angular distribution of sputtered atoms [Olson,
King and Wehner 1979, Fursenko et al. 1995].

The composition of ﬁlms was also varied as the target was continuously
consumed. Figure 4.2 shows the composition of ﬁlms at central area for different time-
on-power for the Ni,,,,Ti3,_,Hf,, target. It can be seen that as the target was continuously
wom-in, Ni and Hf contents in the ﬁlms decreased and Ti content increased. In addition,
the composition varied most during the ﬁrst 20 hours of wear at 1 kW, then a relatively
stable period (20 to 80 hr-kW) followed. More variation was observed as the lifetime of
the target was approached. This change trend of composition is generally associated with
the geometrical change of the target surface since the target was not evenly worn out in a
magnetron sputtering process.

Films with Ni,,9,,Ti3,_2Hf,,7 composition were annealed and studied in detail.
These ﬁlms were deposited from a Ni,,_,Ti37_,Hf,, target at substrate temperature around
200°C, and crystallized by conventional vacuum annealing. Phase transformations,
microstructural characteristics and mechanical properties were analyzed. One of the
major ﬁndings was that after annealing, the samples became brittle, but using faster
heating rate alleviated this problem. In order to clarify the causes of embrittlement,
different annealing procedures were conducted. The mechanical properties of as-
annealed samples were evaluated and their microstructures were examined. Rapid
thermal annealing was also utilized in order to conﬁrm the heating rate effect and to

assess the beneﬁt of using high heating rates to crystallize Ni-Ti-Hf ﬁlms.

76

4.1.1. Crystallization by Conventional Isochronal Heating

Like binary Ni-Ti alloy thin ﬁlms, amorphous Ni-Ti-Hf thin ﬁlms must be
crystallized in order to become shape memory materials. Crystallization may take place
either by isothermal annealing or isochronal heating, and the annealing parameters must
be established experimentally according to the characteristics of crystallization. The most
important parameter for crystallization is the crystallization temperature, which is
heating-rate dependent, but provides an important guideline for crystallization annealing
treatment. In this experiment, the crystallization temperature was determined by DSC

analysis.

4.1.1.1. Crystallization Temperature and Activation Energy. Figure 4.3 shows the
DSC curves of the as-deposited amorphous Ni,.,,,Ti,,,,2Hf,,,7 thin ﬁlms during isochronal
heating excursion to 575°C at rates of 5°C/min, 10°C/min, 20°C/min and 40°C/min,
respectively. Apparently there are two exothermic peaks in the course of isochronal
heating. The peak temperatures as well as transformation enthalpies of exothermic peaks
of different DSC curves are listed on Table 4.2 (and also in the insets in Figure 4.3) . In
each curve, the second exothermic peak, which was sharp and located at higher
temperature, was associated with amorphous-to-crystalline transformation. Comparing
the curves that were obtained using different heating rates, one can see that the peak
temperature and transformation enthalpy were heating-rate dependent (Figure 4.4a).
With the heating rate changed from 5°C/min to 40°C/min, the crystallization peak
temperature monotonically increased from about 513°C to 545°C. At the same time, the
transformation enthalpy increased from 12.6 to 17.8 J/g. The peak temperature of the
crystallization transformation of Ni,,,,,Ti3,2Hf,,7 amorphous thin ﬁlms was higher than
has been observed for amorphous near-equiatomic Ni-Ti ﬁlms [Kim et al. 1986, Gong et

al. 1994, Zhang 1997].

77

Table 4.2. Peak Temperatures and Transformation Enthalpies of the Peaks
on the DSC Curves During Isochronal Annealing of As-Sputtered
Amorphous Ni,,,,Ti,,,Hf,,,., Films at Different Heating Rates.

 

Heating Rate (°C/min) 5 10 20 40

 

1" Peak Temperature (°C) 342 360 372 382

1’t Peak Enthalpy (J/g) 6.2 5.5 3.6 2.5
2“ Peak Temperature (°C) 514 525 533 545
2Ml Peak Enthalpy (J/g) 12.6 14.9 16.9 17.8

 

AH,+AH,(I/g) 18.8 20.4 20.5 20.3

 

The shift of crystallization temperature during isochronal annealing resulting from
different heating rate is well understood, and was observed in other amorphous SMA
materials [Chang and Grummon 1997]. The reason is because crystallization is a
diffusive process that depends on thermal activation and is controlled by both temperature
and time. Lower heating rate may allow more time for difﬁrsion, and the crystallization
temperature is lower. The activation energy for crystallization may be calculated based
on the shift of crystallization temperature with heating rate using Kissinger’s method

[1957]:

= _R dan<¢ / T: ))
d(1/T,,)

 

(4.1)

where R is the molar gas constant (R = 8.314 J/mol), ¢ is the heating rate, and Tm is the
temperature at which the reaction rate reaches a maximum. It would be a good estimation
to consider Tm as the peak temperature in the DSC measurement (for a certain heating
rate). Figure 4.5(a) shows the linear ﬁtting result according to Kissinger’s equation. The
activation energy was estimated at 347 kJ/mol. That value was slightly smaller than the

data (385 kJ/mol) from Chang and Grummon for Ti-rich Ti,,,0Ni,,,Cu,, multi-layered

78

ﬁlms [1997] and much lower than that (477 kJ/mol) from Zhang [1997] in Ni-rich Ti-Ni
ﬁlms.

The ﬁrst exothermic peak locating at temperature between 300°C and 400°C was
very shallow compared to the peak at higher temperature. The peak temperature as well
as the corresponding heat varied with the heating rate (Figure 4.4b). The peak
temperature increased with increasing heating rate, whereas the heat associated with these
peaks decreased. That suggests the reaction was suppressed at high heating rate. In
addition, although the enthalpies for both the high-temperature peak and low-temperature
peak varied evidently with heating rate, their sum was relatively constant, which may
suggest that the reaction at the lower temperature have produced a product that would
contribute less heat to the crystallization transformation. Unlike the high-temperature
sharp peaks that are associated with crystallization transformation, the low-temperature
peaks are relatively diffuse. However, estimation of transformation activation energy for
those peaks using Kissinger’s method was also attempted (Figure 4.5b), and it gave a
value of 162 kJ/mol.

4.1.1.2. Martensitic Transformation Temperatures and Enthalpies. Samples were
thermally cycled between room temperature and 250°C in the DSC system directly after
the crystallization annealing"3 to determine the displacive transformation temperatures.
Figures 4.6 Show the DSC curves for samples crystallized by isochronal heating with
different heating rates. In each of those ﬁgures, an exothermic peak associated with A —>
M transformation occurs during cooling at temperatures ranging from about 70°C to
about 100°C, and an endothermic peak associated with M -> A transformation appears

during heating at temperatures ranging from about 120°C to 175°C.

 

'3 The cooling rate from 575°C (end temperature for isochronal heating) to 250°C was 40°C/min, and the
scan rate was 20°C/min during thermal cycling between 50°C and 250°C.

79

Start and ﬁnish temperatures of martensitic transformation were Obtained from
Figures 4.6 using tangent-intercept method, and the results are listed on Table 4.3. The
correlation between those temperatures and the heating rate during isochronal heating iS
illustrated in Figure 4.7. It is indicated that the transformation temperatures, especially
the peak temperatures for calorimetric peaks in the DSC curves, did not vary much with
the heating rate during isochronal annealing for crystallization. The transformational
hysteresis did not seem to change substantially with crystallization heating rate either.
However, the transformational enthalpy evidently varied with the heating rate. For the
sample annealed at the lowest heating rate (5°C/min), the average value of enthalpies for
forward and reverse transformations is 7.2 J/g (0.52 kJ/mol). As the heating rate was
increased, the average enthalpy was increased consistently. For sample heated at
40°C/min, the transformation enthalpy increased to 10.2 J/g (0.74 kJ/mol). The variation
of transformation enthalpy with heating rate for crystallization anneal may suggest more
martensite formed as the material cooled down to the M, temperature. In addition, the
transformation enthalpies were generally lower even than that for binary Ni-Ti ﬁlms
(which normally had AH of 1.0 kJ/mol to 1.3 kJ/mol [Lo et al. 1990]), just opposite to the
prediction that Hf would increase the transformation enthalpy [Cesari et al. 1999].
Therefore, it is suggested that there was a considerable fraction of material did not
participate martensitic transformation.

Because the martensitic transformation temperatures and transformation
hysteresis did not varied evidently with the heating rate for crystallization, it is likely that
the composition of the transforming phase was similar in ﬁlms crystallized at different
rates. Also, the microstructural characteristics are not expected to vary substantially with

the heating rate. Evidences to this effect will be sought in next section.

80

Table 4.3. Transformation Temperatures and Enthalpies of As-Crystallized
Ni,,_,Ti,,,Hf,,,., Films After Isochronal Heating at Different Rates.

 

 

 

Heating Rate (°C/min) 5 10 20 40
M, (°C) 100 100 100 97
M, (°C) 70 72 70 78
A,(°C) 120 110 110 134
A, (°C) 172 170 173 166
M, (°C) 85 86 85 87
A, (°C) 155 153 158 152
Hysteresis (°C) 70 67 73 65
Average Enthalpy (J/g) 7.2 8.3 9.3 10.2
Average Enthalpy (kJ/mol) 0.52 0.60 0.67 0.74

 

4.2.1.3. Structure and Microstructure Analyses. X-ray difﬁaction and TEM
observation were conducted to investigate the crystallography and microstructure of the
annealed samples. Spectrum (a) in Figure 4.8 is the X—ray diffraction spectrum of the as-
deposited sample taken prior to isochronal annealing for crystallization. Spectra (b) ~ ((1)
in Figure 4.8 are for samples after isochronal heating from room temperature to 575°C at
rates of 5°C/min, 10°C/min, 20°C/min and 40°C/min. The sole wide hump at 2 6 angle
ranging from about 34° to 46° in curve (a) in Figure 4.8 conﬁrms the as-deposited sample
was totally amorphous. After isochronal heating, the wide hump disappeared and sharp
diffraction peaks emerged, verifying that the amorphous structure had transformed to
crystalline phase during the heating. In the spectrum of sample annealed at 5°C/min
heating rate, the highest peak appears at 26angles of 41 .6° (d = 2.17A), and two minor
peaks appear at 2 Hangles of 291° (d = 3.06A) and 604° (d = 1.53A), respectively. In
addition, there are two weak peaks at 26 angles of 384° (d = 2.34A) and 44.4° (d =

2.04A). As the heating rate for isochronal annealing was increased, the intensity of the

81

peak at 41 .7°C decreased, while the peaks at 384° (d = 2.34A) and 44.4° (d = 2.04A)
became stronger.

Analysis of the crystalline peaks indicated that three strong peaks (at 41 .6°, 29.1°
and 604°) belong to a cubic structure with lattice parameter of a = 3.06 A. Because of
the existence of the (001) reﬂection, the unit cell of the cubic structure is either a simple
cubic or an ordered b.c.c. structure (B2 structure). Since the intensity of (001) diffraction
peak is very weak compared to that of (110) diffraction, it is likely a superlattice
reﬂection, and the structure is thus ordered. In addition, the measured lattice parameter
matched the predicted value of an ordered b.c.c. structure)9 In fact, the crystal structure
is the same as that of the austenite phase in binary Ni-Ti alloy except that the lattice
parameter is slightly larger. The weak peaks emerging at higher heating rate belonged to
a monoclinic structure with lattice parameters of a = 2.63 A, b = 4.01 A, c = 4.77 A and ,6
= 99.1°. This monoclinic structure is similar to B19’ martensite in binary Ni-Ti alloy (a =
2.889 A, b = 4.120 A, c = 4.622 A and ,6 = 98.6°) except that the lattice parameters are
slightly different. The lattice parameters are close to those of the martensite in
Ni49Ti36Hf15 alloy obtained by Han et al. from SAED (a = 2.63A, b = 4.11A, c = 4.73A
and ,B = 100.1°). Changes in the relative intensity of the XRD peaks with different
heating rate indicate that the phase ﬁ'action was strongly affected by the heating rate for
crystallization. At higher heating rate during crystallization heating, more cubic
structured phase formed and there was less monoclinic martensite phase in the ﬁlms.

TEM micrographs of as-sputtered amorphous samples after isochronal heating to
575°C at different rates are shown in Figure 4.9 (a) and (b). In the micrograph of the
sample heated at 5°C/min rate (Fig.4.9(a)), the grain size was 0.2 microns or less, which

varied substantially from one grain to another. Some grains have twin-like substructure

 

'9 Using the known lattice parameters of 3.01 A for Ni-Ti B2 structure, the covalent radii of Ni, Ti and Hf
atoms were modiﬁed to 1.214, 1.393 and 1.520 A, respectively. Assuming Hf taking Ti sites randomly, the
effective radius for (Ti, Ht) would be 1.429 A. Therefore, the expected lattice parameters in Ni(Ti,Hf) 82

structure is a z 2(R,,, + M,)/«l3 = 3.05 A.

82

while others do not. The twin-like substructure suggests that those grains were
martensite at room temperature. Grains without twin-like substructure are presumably
the cubic phase that was detected from the X-ray diﬁraction spectrum. It was noted that
the shapes of the grains looked irregular, and the grain growth seemed to be hindered at
some points. In the TEM micrograph of the sample heated at 40°C/min (Fig.4.9(b)), there
are also both apparent martensite grains and cubic phase grains. There did not seem to be
major differences between the TEM micrograph of the sample heated at 5°C/min and that
heated at 40°C/min. It is noted that the grains size in annealed Ni-Ti-Hf ﬁlms was much
smaller than that in annealed binary Ni-Ti ﬁlms. The latter typically was on the order of
micron or larger [Ishida et al. 1996, Meng and Catalan 1999].

4.1.1.4. Mechanical and Shape Memory Properties. Preliminary examination of the
as-annealed samples after isochronal crystallization treatment revealed that the samples
were very brittle although the amorphous ﬁlms before crystallization annealing were
fairly ductile. Quantitative measurement by tensile test turned out to be very difﬁcult
because the as-annealed samples were so brittle that it was almost impossible to clamp
the samples without failure at the clamping area. In this experiment, the ductility of the
annealed samples was semi-quantitatively evaluated by bending the strip sample to
different radii until the smallest bending curvature radius the samples could survive with
was found.

The results of room-temperature ductility evaluation for different samples are
listed on Table 4.4. The original amorphous samples had ultimate uniaxial tensile
elongation of about 2.5%. However, after the samples were crystallized by isochronal
annealing with 40°C/min heating rate, the samples became so brittle that tensile test was
no longer applicable. In bending tests, most samples survived at 1mm bending diameter
(corresponding bending strain is 0.5%) but failed at 0.5mm bending diameter

(corresponding bending strain is 1.0%). For the samples crystallized by isochronal

83

annealing at heating rate of 5°C/min, most failed at 1mm bending diameter
(corresponding bending strain is 0.5%) although they survived at 2mm bending diameter
(corresponding to 0.25% bending strain). Therefore, it was qualitatively demonstrated
that embrittlement had occurred in the isochronal annealing, and but higher heating rate

during isochronal annealing may alleviate the embrittlement to some degree.

Table 4.4. Comparison of Room-Temperature Ductility of As-Deposited
Amorphous Ni,,,,Ti3,,,Hf,,_, Samples Before and After Isochronal Heating
from Room Temperature to 575°C

 

 

Sample Heating Rate Tensile Elongation Bending Diameter RT Ductility

(4 kinds) (°C/min) (%) (mm) (%)
Amorphous N/A 2.5 N/A 2.5
Annealed #1 40 N/A 0.5 ~ 1 mm 0.5 ~ 1
Annealed #2 5 N/A 1 ~ 2 mm 0.25 ~ 0.5

 

Reversible Shape change of the as-annealed samples was observed upon heating
and cooling. At room temperature, the sample strips were usually curled (radii 5 ~ 10
mm), but when the samples were heated on a hot plate, they usually suddenly became
straight at temperatures around 150°C. Reverse shape change occurred when the samples
were cooled down to around 90°C. Although thermal expansion may also result in shape
change (curling and uncurling), it was not the case because curling/uncurling would
happen gradually. In addition, the temperature for shape change seemed to be consistent
with the transformation temperatures of the as-annealed samples. Therefore, the
curling/uncurling shape change was most likely related to the shape memory effect
resulting from reversible martensitic transformation. 2° However, quantitative
measurement of SME either by free-recovery method or constrained thermal cycling was

impossible to conduct due to the brittleness of the samples.

 

2° The Shape memory effect described here was two-way SME, which was possible for samples having
stress gradient.

84

4.1.2. Isothermal Annealing at Low Temperatures

The DSC curves of amorphous Ni,,,_,Ti3,,2Hf,,_7 ﬁlms during isochronal heating
indicated that there was also shallow exothermic peak in each of those curves (Figure
4.3). The enthalpies associated with those shallow peaks were affected by the isochronal
heating rates. Larger enthalpy was measured in the samples that were heated at lower
rates. This Observation seemed to be linked with the embrittlement which was more
severe in the sample heated at the lower rate. Therefore, it was interesting to clarify what
caused the Shallow exothermic peak and if there was any connection between those
shallow peaks and embrittlement of the annealed samples.

To do so, we isothermally annealed the aS-deposited amorphous Ni,,,,,Ti3,_2Hf,,_7
thin ﬁlms for 1 hour at 300°C and 350°C in vacuum, respectively. These temperatures
covered the range of the shallow peaks in the DSC curves of Ni,.,_,Ti3,,,2Hf,,_7 ﬁlms during
isochronal heating (Figure 4.3), but substantially lower the corresponding crystallization
temperatures (510°C ~ 550°C). In order to minimize the inﬂuence from heating rate, fast
heating (40°C/min) was used to bring the sample the target temperature. Annealed
samples were evaluated for their room-temperature ductility, and crystallographic and

microstructural characteristics.

4.1.2.1. Room-Temperature Ductility. The ductility of the samples before and after
low-temperature isothermal annealing at different temperatures were semi-quantitatively
evaluated by tensile or bending tests. The results are shown on Table 4.5. For the
original amorphous samples, the room-temperature ductility was measured using tensile
tests which showed fair ductility with ultimate elongation at 2.5%. After the sample was
annealed for 1 hour at 300°C, the sample remained fairly ductile at room temperature and

a tensile test showed ultimate elongation at a similar level as the original amorphous

85

sample. If tested in bending mode, these two kinds of samples could be bent back
without failure (bending curvature radius 2 ﬁlm thickness). Samples annealed at higher
temperature showed obvious embrittlement. After annealing for 1 hour at 350°C, the
sample was too brittle to be loaded on to tensile test machine. Therefore, semi-
quantitative measurement using bending test was conducted. It showed that the minimum
curvature diameter for the annealed samples to be bent was between 1 mm to 2 mm,
which corresponds an elongation (or compression) at the sample surface at about 0.25% ~

0.5%.

Table 4.5. Comparison of Room-Temperature Ductility of the As-Deposited
Amorphous Ni,,,,Ti3,,,Hf,,,., Films Before and After l-hour Annealing

 

 

 

at 300°C and 350°C.
Sample Annealing Tensile Elongation Minimum RT Ductility
(4 kinds) Temperature (°C) (%) Bending Diameter (%)
As-deposited N/A 2.5 ~ 10 microns 2.5
Annealed #1 300 2.7 ~ 10 microns 2.7
Annealed #2 350 N/A > 1mm < 0.5
4.1.2.2. Crystallographic and Microstructural Characterization. X-ray diffraction

analyses and transmission electron microscopy were conducted for the samples before
and after isothermal annealing. The X-ray diffraction spectra for different samples at
room temperature are shown in Figure 4.10. It can be seen that the spectrum of the
sample annealed at 300°C and that annealed at 350°C have features similar to those of the
original amorphous sample, which showed only a wide hump, indicating that the
amorphous phase remained aﬁer 1 hour annealing at 300°C, and even at 350°C.
Therefore, the embrittlement must have occurred while the sample was still amorphous.

However, careful examination of the spectra for sample without annealing and those for

86

samples with 300°C and 350°C annealing indicated some subtle changes. The XRD
spectra of as-deposited sample and 300°C annealed sample look very Similar, but the
XRD spectrum of 350°C annealed sample looks Slightly different: the peak is at a higher
angle and has broaden, perhaps, indicating the presence Of a second peak at 422°.
Therefore, even though still amorphous, some structural change has taken place when the
annealing temperature was increased ﬁ'om 300°C to 350°C.

Microstructure observation was also conducted for the samples annealed at
different temperatures to verify the results from X-ray diffraction analyses. The TEM
micrographs, as well as SAEDS, are shown in Figure 4.11. Both the as-deposited sample
and the one annealed at 300°C Show featureless microstructures, and the electron
diffraction patterns of the samples showed only diffuse rings, consistent with the TEM
and XRD results. However, for the sample was annealed at 350°C for 1 hour, a
modulated contrast is apparent although the electron diffraction pattern still shows only a
diffuse ring. Therefore, after the sample was annealed at 350°C, there was some
structural and/or chemistry change within the amorphous structure, but no detectable
crystalline structure had formed. The nature of this change will be discussed in later

sections.

4.1.3. Isothermal Annealing at High Temperature via Different Ramp Rates

Above, it was shown that the properties of the Ni,,,,,,Ti3,_2Hf,,,7 thin ﬁlms depended
on the heating rate during the isochronal crystallization anneal. Higher heating rate for
crystallization treatment apparently improved the properties of the materials. In this
section, amorphous Ni,,,_,Ti3,_2Hf,,7 thin ﬁlms were crystallized by isothermal annealing at
high temperatures, but the heating rates from room temperature to isothermal annealing
temperatures were different. Three annealing procedures were chosen. The ﬁrst two

were conventional vacuum annealing at 550°C for 1 hour with ramp rates of 5°C/min and

87

40°C/min, respectively. The third one was a rapid thermal annealing at 650°C for 20
seconds with ramp rate of 100°C/sec (6000°C/rrrin).21 The annealed samples were
evaluated and characterized for their transformation temperatures, microstructure,

mechanical and shape memory properties.

4.1.3.1. Transformation Temperatures. The martensitic phase transformation
temperatures of the three annealed samples were measured by DSC during thermal
cycling between 50°C to 250°C. The DSC curves are shown in Figure 4.12. The
transformation temperature data were taken from those DSC curves and are listed on
Table 4.6. The difference in transformation temperatures in different annealing process
are illustrated in Figure 4.13. In general, reversible martensitic transformations were
observed in all three samples, but the transformation temperatures and transformational
enthalpy varied, more or less, with different annealing procedures. Between the two
vacuum annealed samples, the transformation temperatures did not seem to be strongly
affected by the heating rate, but the transformation enthalpy was considerably higher for
the sample heated at higher rate. The RTA treated sample (heating rate 6000°C/min)
seemed to have slightly higher martensitic transformation temperature (M, temperature =
103°C) and slightly smaller transformational hysteresis (= 63°C) than the two vacuum
annealed samples. The transformation enthalpy was higher as the heating rate increased.
The RTA treated sample had an enthalpy for martensitic transformation of 14.7 J/g,
which was considerably higher than that for the sample heated at 40°C/min (AH = 10.5

J/ g) and that for the sample heated at 5°C/min (AH = 7.8 J/g). The difference in AH
strongly suggests a larger volume of material underwent martensitic transformation in the

RTA treated sample. Therefore, using very high heating rate in an RTA process has

 

2' The reason that the annealing temperature for RTA treatment is 650°C in stead of 550°C is because
550°C may not be high enough to crystallize amorphous Ni,,,,Ti,,2Hf.,_, thin ﬁlms due to the very high
heating rate and short isothermal duration in an RTA process. The crystallization temperature is estimated
at 631°C using Kissinger’s method if taking the heating rate as 100°C/sec and activation energy as 347
kJ/mol.

88

clearly improved the transformation characteristics of the Ni-Ti-Hf shape memory thin

ﬁlms crystallized ﬁom amorphous samples.

Table 4.6. Martensitic Transformation Temperatures and Transformation
Enthalpies for Ni,,,,,Ti,,2Hf,,,7 Samples Crystallized by Isothermal Annealing
with Different Heating Rates to Annealing Temperatures.

 

 

 

 

 

Samples Annealed #1 Annealed #2 Annealed #3
Process Vacuum Anneal Vacuum Anneal RTA
Heating Rate 5°C/min 40°C/min 6000°C/min
Annealing Temperature 550°C 550°C 650°C
Annealing Time 1 hour 1 hour 20 seconds
Martensite Start (°C) 98 97 103
Martensite Finish (°C) 68 78 87
Martensite Peak (°C) 85 87 95
Austenite Start (°C) 132 134 148
Austenite Finish (°C) 170 166 163
Austenite Peak (°C) 152 152 158
Hysteresis (A, — M,) 67 65 63
Enthalpy (J/g) 7.8 10.5 14.7
Enthalpy (kJ/mol) 0.565 0.760 1.064

 

4.1.3.2. Crystallographic and Microstructural Analyses. X-ray diffraction
experiments were conducted and TEM observation was carried out for the samples
isothermally annealed in vacuum and for those crystallized by the rapid thermal anneal.
Figure 4.14 shows the X-ray spectra for different samples using 20diffraction condition
and CuK, irradiation. Indexing of the diffraction peaks is also shown in the ﬁgure. For

samples using ramp rates of 5°C/min and 40°C/min, the result was similar to what has

89

been found in samples crystallized by isochronal anneal at similar heating rates. The
annealed samples were composed of monoclinic martensite and cubic phase. The lattice
parameters of the monoclinic phase were a = 2.63 A, b = 4.01 A, c = 4.77 A and ,6 =
99.1°, and the lattice parameter of the cubic phase was a = 3.06 A. The change in relative
peak intensity indicated that the phase fraction of martensite was higher for the sample
heated at higher rate. The RTA treated sample showed the highest martensite content, but
still contained some of the cubic phase.

TEM micrographs and SAED pattern of the isothermally annealed samples are
shown in Figure 4.15(a) and Figure 4.15(b). For samples treated by conventional vacuum
annealing with ramp rates of 5°C/min and 40°C/min, the nricrostructure looked very
similar to those of isochronal annealed Ni-Ti-Hf ﬁlms at similar heating rates. That is,
the microstructure consisted in both martensite and cubic phases with varied grain sizes
of 0.2 pm or smaller. The shapes of grains were irregular and the grain boundaries
seemed to have been pinned locally during grain growth. The microstructure of the RTA
treated samples (Figure 4.15(c)) was very different. It was highly uniform with very ﬁne
martensite clusters (characterized by twin-like substructure) dominating the
microstructure. There were also ﬁne grains not having twin-like substructure which were
probably the cubic phases revealed by XRD experiments. The size of the martensite
grains was generally about 50nm, and the width of twin plates was 10nm to 20nm, as
indicated in Figure 4.15(d). The micrograph also indicated that large grains or large

second phases were generally absent in the RTA treated samples.

4.1.3.3. Mechanical and Shape Memory Properties. Mechanical properties of the
isothermally annealed samples were evaluated at room-temperature by bending
deformation or uniaxial tensile tests. The results are shown in Table 4.7. Note that the
ultimate elongation for conventional vacuum annealed samples was measured by bending

test and that for RTA treated samples was measured by uniaxial tensile test. Again, this

90

is because the samples annealed in vacuum furnace were not ductile enough for clamping
which is necessary in uniaxial tensile test, and the RTA sample was so ductile that it can
be bent back (bending curvature radius z ﬁhn thickness) without failure. The table shows
that, the conventionally annealed samples were fairly brittle. The maximum achievable
strain at the sample surface was less then 0.5% at room temperature. The RTA treated
sample, however, is much more ductile. It has a uniaxial elongation of 5.8% and fracture
stress of 1370 MPa (Figure 4.16). Between the two vacuum annealed samples, the one
heated by higher heating rate seemed to be more ductile than the one heated at lower rate.
The effect of heating rate on ductility observed in this series of samples was in agreement

with the results of samples crystallized by isochronal annealing (see Table 4.4).

Table 4.7. Comparison of Room-Temperature Ductility of As-deposited Amorphous
Ni,,,,,Ti,,,2Hf,,,7 Films Crystallized by Isothermal Annealing via Different Ramp
Rates to the Annealing Temperatures

 

 

Sample Annealed #1 Annealed #2 Annealed #3
Process Conventional Conventional RTA
Heating Ramp Rate 5°C/min 40°C/min 6000°C/min
Annealing Temperature 550°C 550°C 650°C
Annealing Time 1 hour 1 hour 20 seconds
Min. Bending Diameter 2 1 mm 0. 5 ~ 1 mm ~ 10 um
Tensile Elongation (%) N/A N/A 5.8
Ductility (%) S 0.5 0.5 ~ 1 5.8

 

Similar to that in samples crystallized by isochronal heating, the two-way shape
memory effect in all the samples was observed, with sudden cyclic shape change upon
both heating and cooling. However, no quantitative data was Obtained for samples
annealed in vacuum furnace due to their brittleness. The RTA treated samples, however,

were good enough for quantitative measurement. Thermal cycling under various constant

91

loads 22 was conducted between 50°C and 250°C, and the applied stresses were up to 350
MPa.

In conducting thermal cycling under different applied stress, the strain recovery
path (heating path) on the ﬁrst half-cycle was markedly different than on subsequent
cycles. Figure 4.17 schematically shows such differences. The sample was loaded at
room temperature and the load was kept constant throughout subsequent thermal cycles.
Under a constant load, the sample was ﬁrstly heated to about 250°C (above A,
temperature) , and the corresponding strain-temperature curve was A-B-C. Then the
sample was cooled down to 50°C and the strain-temperature curve was C-D-E. Then, the
sample was heated up again to 250°C and the corresponding strain-temperature was E-F-
G. The two heating curves A-B-C and E-F-G are markedly different. This is because
little strain was produced by the initial loading at room temperature (or a temperature
substantially lower than the M, temperature), but a large recoverable strain was produced
during cooling from high temperature (curve D-E-F-G). This phenomena suggested that
for RTA crystallized Ni,,,,Ti3,,2Hf,,_7 ﬁlms, transformation strain can be induced at
temperature just above M,, whereas the same level of stress was not sufﬁcient to cause
substantial martensite variant reorientation when applied to a thermally induced
martensite. Therefore, in order to determine the recoverability of shape memory strain, it
is important to apply the load to the material in the austenite condition, i. e. , thermal
cycling should start from high temperature.

Figure 4.18 Shows the strain-temperature plots for RTA treated samples during

thermal cycling at different ﬁxed loads.23 It can be seen that an appreciable amount of

 

2’ Although the applied load was constant during thermal cycling, the true applied stress changed slightly
due to cross-area change resulting from elongation.

2’ For the same strip, the load started from 100 MPa for the ﬁrst thermal cycle, and then was gradually
increased to 150 MPa , 175 MPa, 200 MPa, and 225 MPa for subsequent cycles (Fig.4.l8 and Fig.4.19).
Thermal cycling under 250 MPa, 275 MPa, 300 MPa and 350 MPa stresses was conducted using fresh
samples because a large number of dislocations could be generated during thermal cycling at those stresses
and may have effect on the shape memory characteristics of the materials. Applying stresses greater than
350 MPa usually resulted in ﬁlm breakage, which often occurred during cooling when transfonnational
strain began to form.

92

strain (cc ) was obtained by cooling at applied stresses as low as 100 MPa during thermal
cycling, and the material could be thermally cycled at biasing stresses as high as 350 MPa
without immediate failure. At lower biasing stress (5 200 MPa), closed strain-
temperature hysteresis loops were obtained, indicating the strain generated on cooling
was fully recovered on heating. At higher biasing stress, the hysteresis strain-temperature
loops were open, indicating the strain generated on cooling was not fully recovered.
Figure 4.19 summarizes the shape memory characteristics from this series of strain-
temperature plots. It indicates that the RTA crystallized Ni,9_,Ti3,,2Hf,,,7 samples had fully
recoverable strain of 1.8% and the maximum stress for full stain recovery of about 200
MPa. Recovery strain increased slightly at stresses beyond 200 MPa and reached a
maximum at 1.9% before decreasing with increasing biasing stress. Residual strain
started to develop at biasing stresses higher than 200 MPa, and the higher the biasing
stress during thermal cycling, the more residual strain was developed for a complete
thermal cycle. The maximum strain induced by cooling may be as high as 3% but almost

half of it was not recoverable.

4.1.4. Discussion on Ni-Ti-Hf Films Crystallized by Post-Deposition Annealing

4.1.4.1. Transformation Temperatures. Producing Ni-Ti-Hf SMA ﬁlms using post-
deposition crystallization annealing has been attempted by a number of groups [Johnson
et al. 1995, Gu et al. 1995, You et a1. 1997]. Although there have been reports that
indicated the transformation temperatures Obtained were higher than those in binary Ni-Ti
ﬁlms, the effectiveness of haﬁrium in the ﬁlms in increasing transformation temperatures
was disappointing compared to that in melt-solidiﬁed materials. Moreover, no literature
disclosed the mechanical characteristics of the annealed samples.

In this thesis, Ni,,,,_,Ti3,,2Hf,,,7 SMA ﬁlms produced by annealing an amorphous

precursor exhibited M, temperatures between 80 and 90° and A, temperatures between

93

150°C and 160°C. Although evidently higher than those in binary Ni-Ti ﬁlms, those
temperatures were generally about 80 to 90°C lower than those of melt-solidiﬁed bulk Ni-
Ti-Hf alloy having similar hafnium content [Han et al. 1995], which were about 170°C
and 235°C for M, and A, temperatures, respectively. The results were in general

agreement with those reported by others.

4. 2. 4. 2. Annealing Induced Embrittlement of Amorphous Ni-Ti-Hf Films. The
mechanical behavior of the annealed Ni-Ti-Hf ﬁlms was one of the major concerns in this
study. It was found that common annealing procedures suitable for crystallizing binary
Ni-Ti ﬁlms typically resulted in brittle Ni-Ti-Hf ﬁlms after annealing, yielding basically
unusable product. Further study indicated that embrittlement occurred at temperatures
between 340°C ~ 390°C, far below the typical crystallization temperatures (510°C ~
550°C), when the samples were still amorphous. X-ray diffraction analyses and
transmission electron microscopic Observation suggested that during this period, the
materials underwent structural change even though no major phase transformation had
taken place.

The structure change was most likely to be structural relaxation of amorphous
material. In fact, structural relaxation has been often Observed in amorphous materials
and linked to embrittlement reactions [Chen 1976, Haasen 1983, Makino et al. 1990, Wu
and Spaepen 1990, Qu and Xie 1991, Niikura et al. 1993, Barth et a1. 1997]. The most
widely observed phenomenon is the short—range ordering which can be divided into
topological Short-range ordering and chemical short-range ordering [van den Buekel and
Huizer 1985], during which the atoms diffuse or rearrange even though the structure of
the material remain amorphous. In the cases of amorphous Ni-Ti-Hf thin ﬁhns in this
study, embrittlement occurred prior to formation of any detectable crystalline phase,
leading to a suspicion that the embrittlement was caused by structural relaxation of

amorphous ﬁlms.

94

In fact, topological short-range ordering or chemical short range ordering was
very likely in sputtered Ni-Ti-Hf ﬁlms during a post-annealing process. This is because
that structural relaxation resulting from such processes has been observed in melt-spun
amorphous Ni-Hf ribbons [Buschow and Bekmans 1979] and amorphous Ni,,,Ti,o
produced by ion implantation [Bodin, Moine and Eymery 1987], although the effect on
mechanical properties was not discussed. It is reasonable to make comparison between
sputtered Ni-Ti-Hf amorphous ﬁlms and melt-Spun ribbons, because the sputtered
amorphous ﬁlms, like melt-Spun amorphous ribbons, contained a fraction of free volume
due to self-shadowing effect of the sputtered ﬁlms, and it was the free volume that
determined the brittleness of the amorphous material [Wu and Spaepen 1990]. It seemed
that the free volume in as-sputtered amorphous ﬁlms was well distributed since the
material was fairly ductile. However, post-deposition annealing may change the fraction
and distribution of the free volume which would result in embrittlement of the amorphous
material.

The occurrence of Short-range ordering was supported by experimental results
from X-ray diffraction (Figure 4.10) and TEM investigation (Figure 4.11). In the X-ray
diffraction spectra, the shift of amorphous peak towards to higher 20 angle indicated the
average interatomic distance in the amorphous structure decreased, suggesting that
densiﬁcation took place. In the TEM micrograph, the wavelike contrast suggested there
may be regions having compositional enrichment or material enrichment (densiﬁcation),
also an indication of occurrence structural relaxation involving short-range ordering
[Glezer et al. 1984, Loirat et al. 2000]. The TEM micrograph indicated that chemical
short-range ordering may be a signiﬁcant factor for embrittlement.

The apparent activation energy of the process associated with the low-temperature
exothermic peak also supported the occurrence of short-range ordering. The activation
energy was 162 kJ/mol (1.68 eV/atom). It was comparable with the typical activation

energy for structural relaxation in amorphous material (For example, 168 ~ 179 kJ/mol

95

for Alngeer, [Wang et al. 2002], 154 kJ/mol for Fe7,B-,Sils [Takahara 1997], 159
kJ/mol for Fe40Ni38M04Bm [Keupers et al. 1985]). In addition, the value of the activation
energy was close to that for Hf diffusion in amorphous Ni,,,Zr,, alloy which was 171
kJ/mol (1.78 eV/atom) [Loirat et al. 2000]. It is known that Ni is a faster diffuser in a Ni-
Ti (Zr, Hf) system (activation energy for Ni diffusion was 1.33 eV/atom in Zr-Ni
amorphous [Gradj ean and Limoge 1997]), the fact that the activation energy for the
embrittlement reaction was determined by Hf diffusion may suggest that chemical short-
range ordering was more signiﬁcant than topological short-range ordering in the
structural relaxation process.

There are some other possible mechanisms for embrittlement such as segregation
of impurities and oxidation. While no evidence for impurity segregation was found in the
TEM observation, oxidation induced embrittlement can also be ruled out. Although our
observation indicated that Signiﬁcant oxidation of Ni-Ti-Hf ﬁlms on the surface took
place at temperature as low as 250°C in extended exposure to atmosphere, the reduction
of ductility was fairly moderate. In addition, pure Ti was used as oxygen getter during
the annealing seemed to have successfully inhibited oxidation on the surface (no
observable discoloration was noted on the surface) even though the vacuum for the
annealing in this experiment (~5x10‘ torr) was not considered to be a very hard vacuum.
Therefore, oxidation does not seem to be severe enough to cause signiﬁcant
embrittlement of the material. Nevertheless, there may still be oxygen in the ﬁlms
through absorption or incorporation followed by diffusion to the interior via grain

boundaries and they may play a role in precipitation which is discussed next.

4.1.4.3. Have Precipitates Formed At Low Temperatures? Although have not been
directly observed, crystalline phases such as (Ti+Hf)2Ni or (Ti+Hf),NiZO are likely to
form prior to the formation of (Ti+Hf)Ni B2 phase. This is because f.c.c. TizNi
crystallized at lower temperature (3 50°C ~ 400°C) than B2 NiTi phase [Buchwitz,

96

Adlwarth-Dieball and Ryder 1993] and local ﬂuctuation of composition was very
common in amorphous ﬁlms. Furthermore, the possible presence of oxygen in the
interior of ﬁlms promoted the likelihood of formation of theses phases, since oxygen
atoms stabilize the (Ti+Hf)2Ni phase and nucleation of (Ti+Hf)2Ni phase could occur
without necessity of local Ti enrichment. Although XRD and TEM studies were not able
to detect (Ti+Hf)2Ni or (Ti+Hf),NiZO phases, it was possible that they may still exist in a
very ﬁne form and very small volume percentage, since the composition of the ﬁlms was
very close to stoichiometry. It was noted that formation of these phases would lower the
Ti content in the matrix. It is likely the matrix composition became Ni rich, thereby the
transformation temperatures would be depressed.

The existence of ﬁne (Ti+Hf)2Ni or (Ti+Hf),NiZO heterostructures in the
amorphous matrix was suggested from the shape of grain boundaries after the sample was
fully crystallized. Under normal conditions, crystallization formed spherical B2 grains,
which evolved into regularly equiaxial grains as the boundaries met with each other.
Spherical crystalline grains in the amorphous matrix were Observed in melt-spun
Ti,,,Ni2,Cu25 alloys [ROsner et al. 1999]. That was expected because the surface energy
would be minimized in this way. However, the grains in the conventionally annealed Ni-
Ti-Hf samples looked irregular, and it seemed that the grain boundaries had been pinned
as the grains grew. It was suspected that the irregular grain boundaries may contribute to
the brittleness of the crystallized ﬁlms, since crack initiation and propagation can be
easier at these locations.

Formation of (Ti+Hf)2Ni or (Ti+Hf),NizO phases was further supported by the
relatively low transformation temperature of the crystallized ﬁlms and residual B2 phases
when the samples were cooled to room temperature. As observed by Han et al. [1995],
the (Ti+Hf)2Ni or (Ti+Hf),NizO phases have higher Hf : Ti ratio than in the matrix.
Therefore, Ti and Hf in the matrix somehow was depleted due to the formation of

(Ti+Hf)2Ni or (Ti+Hf),Ni20 phases, and the transformation temperatures were thus

97

decreased. Ni-enrichment in the residual B2 phase may also result from depletion of
(Ti+Hf) in the matrix, which drove the martensitic temperatures to a very low level (see

Figure 2.8).

4.1.4.4. Suppression of the Embrittlement via Fast Heating. Although brittle Ni—Ti-
Hf ﬁlms resulted if crystallized by conventional vacuum annealing, using high heating
rate such as in an RTA process to get fully crystalline material seemed to minimize or
eliminate the embrittlement. This was probably because at high heating rate, the
time/temperature window for embrittlement reaction was shortened, allowing the
crystalline (Ti+Hf)Ni phase to form before embrittlement could occur due to the
difference in incubation time.

Besides making the material more ductile, using high heating rate to crystallize
amorphous Ni-Ti-Hf ﬁlms had another beneﬁt. As shown in Figure 4.8 and Figure 4.14,
the as-annealed samples were composed of B19’ martensite and B2 phase at room
temperature. There was more room-temperature B l 9’ martensite in the samples annealed
via higher heating rate. Therefore, the transformation characteristics were improved by

using high heating rate for crystallization.

4.2. In-Situ Crystallized N i-Ti-Hf Films

In the ﬁrst part of this chapter, Ni,,,,,Ti3,_2Hf,,,7 SMA ﬁlms were discussed that had
been obtained by post-deposition annealing the originally amorphous ﬁlms. In this part,
Ni-Ti-Hf SMA ﬁlms are discussed that were produced by sputtering at elevated substrate
temperature so that the as-deposited ﬁlms were crystalline and showed shape memory
properties without necessity for further heat treatment. As shown below, it was found
that these in-situ crystallized Ni-Ti-Hf SMA ﬁlms showed greatly improved mechanical
and shape-memory properties. Samples with different Hf compositions were

investigated.

98

4.2.1. Phase Transformations

As-deposited crystalline Ni-Ti-Hf thin ﬁlms having different compositions were
Obtained by sputter deposition from targets with various compositions using the same
deposition parameters for depositing amorphous ﬁlms (i. e., working gas Kr pressure 1.0
mTorr, working distance 81 mm, deposition power 1 kW) except that the substrate
temperatures were between 400 to 450°C 2‘. DSC curves for these samples during
thermal cycling between 50° and 300°C are shown in Figure 4.20. In each of these
curves, an endothermic peak occurred during heating, and an exothermic peak occurred
during cooling, corresponding to a transformation from austenite to martensite, and
martensite to austenite, respectively.

Transformation temperatures including M,, M,, M,, A,, A, and A, were
determined from those DSC curves. Table 4.8 shows those temperatures as well as the
transformational hysteresis (A, - M,). For comparison, the transformation data for the
RTA treated Ni,,,,,Ti3,,,Hf,,_7 ﬁlms are listed in the brackets in the last column. It can be
seen that substantially higher transformation temperatures than binary Ni-Ti thin ﬁlms
(typically M, < 60°C, A, < 110°C [Hunbeeck 1999, Hodgson and Brown 2000]) were
Obtained in the Ni-Ti-Hf thin ﬁlms. An M, up to 200°C was achieved when the hafnium
content was about 21 .4at.%. As expected, the transformation temperatures depended on
the hafnium content, since all ﬁlms were (Ti+Hf) rich. Figure 4.21 illustrates the
transformation temperatures as a ﬁmction of hafnium content. AS the hafnium content

increased, transformation temperatures increased monotonically. The peak temperature

 

2" Because Hf addition tends to increase the crystallization temperature for the same annealing conditions,
one may expect that the minimum substrate temperature required for obtaining fully crystalline Ni-Ti-Hf
SMA thin ﬁlm is also composition dependent. The higher hafnium content, the higher the substrate
temperature required for obtaining full crystalline structure. In choosing the substrate temperature, it is
important to keep the substrate temperature just slightly above the minimum temperature required for fully
crystalline structure, because if the substrate temperature is too high, not only will the microstructure
change dramatically, but also the surface roughness will become excessive. That could affect the
mechanical properties of the thin ﬁhns substantially.

99

of martensitic transformation (M,) was 135°C for the sample with 14.5at.% Hf, and it
increased to 182°C for the sample having 21 .4at.% hafnium. The same effect was found
for other transformation temperatures. Also, transformational hysteresis changed
considerably with increasing hafnium content. For example, the transformational
hysteresis of the sample containing 14.5at.% hafnium was 45°C, and it increased to 63°C
when the hafnium content rises to 21 .4at.%. The average of enthalpy change for both
forward and reverse transformation was about 17 to 18.3 J/g (or 1.231 ~ 1.501 kJ/mol),
with higher transformation heat was measured in samples having higher Hf content.
These values are generally higher than the transformation enthalpy measured in Ni-Ti

binary alloys (1.0 ~ 1.3 kJ/mol, [LO et al. 1990]).

Table 4.8. Martensitic Transformation Temperatures vs. Sample Composition
of In-Situ Crystallized Ni-Ti-Hf SMA Films

 

Target C0mp03iti°n NinsTisstfzo Ni47.5Ti34.5Hfls Ni48.5Ti35.5Hf16 NiusTinstM

 

Film COII‘POSIIIO“25 Ni48,2Ti30.4Hf21.4 Ni48.OTi33.9Hf18.1 Ni49 3Ti34.5Hfl6.2 NiassTlssstus

 

 

 

 

M, (°C) 200 157 150 140 (103)I
M, (°C) 160 143 130 127 (87)
A, (°C) 213 202 181 172 (148)
A, (°C) 260 217 202 184 (163)
M, (°C) 182 150 140 135 (95)
A, (°C) 245 210 190 180 (158)
Hysteresis (°C) 63 60 50 45 (63)
Enthalpy (J/g) 18.3 17.6 17.4 17.0 (14.7)
Enthalpy (kJ/mol) 1.501 1.345 1.300 1.231 (1.064)

 

IValues at far right in parentheses are for RTA post-deposition annealed Ni,,,,,Ti,,,2Hf,,7 ﬁlms.

 

25It is noted that the compositions on this table are slightly different than those on Table 4.1 even though
the same targets were used. This slight disparity mostly resulted from the difference in deposition
temperature. The typical trend is, the Ni and Hf contents in ﬁlms decreased at increasing deposition
temperature while the Ti content increased.

100

Comparing the transformation data of in-situ crystallized Ni,,,9Ti3,,Hf,,_, and that
of the RTA treated Ni,,,_,Ti,,,2Hf,,7 ﬁlms, the in-situ crystallized ﬁlms had considerably
higher transformation temperatures and smaller transformational hysteresis. In addition,
the martensitic transformational enthalpy was higher for in-situ crystallized ﬁlms. These
results indicated that the in-situ crystallized ﬁlms exhibited improved transformation

characteristics.

4.2.4. Crystallographic and Microstructural Analysis

In order to compare the crystallographic and microstructural characteristics of in-
situ sputtered Ni-Ti-Hf and post-deposition annealed Ni-Ti-Hf ﬁlms, X-ray diffraction
and TEM micrographs of Ni,,_9Ti,,,_,Hf,,_5 ﬁlms were studied and the results are presented

below.

4.2.4.1. X-ray Diﬁ'raction. Figure 4.22 shows X-ray diffraction spectra of in-situ
crystallized Ni,8_9Ti3,,Hf,,, thin ﬁlms. These spectra were taken at room temperature
using CuKOt irradiation and 29mode (in-plane position). Three major peaks were found
at 26angles 452°, 41 .5° and 382°. Also there are several minor peaks at other 20
positions. All the peaks can be indexed a phase having monoclinic structure with lattice
parameters of a = 2.63 A, b = 4.01 A, c = 4.77 A and ,6 = 991°. The highest peaks are
indexed to (020), (1 1-1) and (200) planes, respectively, and the two minor peaks at 20
angle of 44.42° and 60.29° may be indexed as (012) and (022) planes, respectively. The
monoclinic phase was the same as that identiﬁed in the RTA treated Ni,,,9Ti3,,Hf,,,
samples, but the X-ray spectrum looked different. For RTA treated sample the strongest
X-ray diffraction peaks for martensite were (002)M and (012)M, but for as-sputtered

crystalline sample the strongest peak was (O20)M. This may suggest that the texture of

101

as-sputtered crystalline sample was different with that of crystallized sample by RTA
treatment. Also, unlike the RTA treated sample in which there were cubic phases besides
the martensite, the as-sputtered crystalline sample did not explicitly show any phases
other than martensite. However, because the ﬁlms are slightly (Ti+Hf)-rich, it is
suspected that (Ti+Hf)2Ni or (Ti+Hf),NizO, precipitates may exist in the 319’ matrix. X-
ray diffraction spectra cannot rule out the existence of those phases because their peaks
could be concealed into the high peaks of the martensite because the volume fraction of
the second phases could be very small. Therefore, TEM observation is necessary to

conﬁrm whether there are second phases or not.

4. 2. 4. 2. Transmission Electron Microscopy and Electron Diﬁfaction. Figure 4.23
(a) ~ (c) Show the TEM micrographs of in-plane sample of an in-situ crystallized
Ni,,,,Ti,,,_,Hf,,_5 ﬁlm as well as corresponding electron diffraction patterns. They
indicated that the microstructure was composed of grains that had twinned substructure, a
typical feature of the martensite of Ni-Ti based Shape memory alloy [Otsuka and
Wayman 1986]. The grain size was on the order of 100 nm, which was comparable with
that of in-situ deposited Ni-rich TiN i ﬁlms in the interior layer, but evidently larger than
the ﬁne grains in the layers close to the substrate surface [Hou 1998]. The grain size was
also much smaller than typical grain Size for melt-solidiﬁed Ni-Ti-Hf alloys (>1 um)
[Han et al. 1995, Olier et al. 1995]. In the TEM micrographs (Figure 4.23 (b), (c))
having higher magniﬁcation, clear twinned substructure was revealed. The thickness of
twin plate was on the order of 10 nm or smaller. Consistent with the X-ray diffraction
results, the micrographs did not reveal any second phases.

A selected-area electron diffraction pattern was also taken and it is shown in
Figure 4.23. The aperture size was 50 microns so there may be hundreds of grains. The

diffraction pattern was composed of a series of rings. The sharpness of those rings is

102

generally not good because some of the interplanar distances are very close to each other

in this low-symmetry structure.

4.2.2. Mechanical Properties

In order to assess the effect of Hf addition on the mechanical behavior of Ni-Ti
SMA ﬁlms, Ni,,,,Ti,,,_5 ﬁlms were prepared using the in-situ crystallization technique and
were studied as a benchmark. The binary ﬁlms were deposited from a Ni,,Ti,3 target
using the same deposition parameters as for ternary deposition (i. e. , working gas Kr
pressure 1.0 mTorr, working distance 81 mm, deposition power 1 kW) except that the
substrate temperature was about 380°C ~ 400°C. The transformation temperatures of the
binary ﬁlms are M, = 55°C, M, = 37°C, A, = 78°C, and A, = 90°C. Figure 4.24 showed a
room-temperature stress-strain curve which is typical of martensitic NiTi SMA and is
composed of a number of deformation stages. The ﬁrst segment A-B represents a linear
elastic deformation of the martensite. The following segment B-C is characterized by
low ﬂow stress with negligible strain hardening, and it is associated with martensite
variant reorientation. The third segment C-D is a transitional segment, which is followed
by another linear elastic one (D-E). Finally, the segment E-F is a plastic deformation as a
result of dislocation slip, which is a common mechanism of plastic deformation occurring
in ordinary metallic materials. The second elastic segment (D-E) had a smaller Young’s
modulus than the frrst elastic segment (A-B). This may result from either a combination
of plastic/elastic deformation or elastic anisotropy.

It is well known that the mechanical behavior of SMAS is temperature dependent.
Therefore, tensile tests of the in-situ crystallized Ni49,,Ti,o,, ﬁlms were conducted at
various temperatures. The results are Shown in Figure 4.25. Note that all of the curves
except curve (f) were obtained for specimens that were heated ﬁom room temperature.

Therefore, curves (a) (at 25°C), (b) (at 50°C) and (c) (at 75°C) were all obtained when the

103

samples were fully martensitic. Curve (d) (at 100°C) and curve (1) (at 125°C) were
obtained when the samples were fully austenitic. The test temperature for curve (f) was
75°C but the sample was cooled down to this temperature from 150°C, so the sample was
fully austenitic before the test started. Comparing curves (a), (b) and (c), one can see that
the ductility increased when the tensile test temperature was increased. At the same time,
the yield stress, which corresponds to the onset stress for martensite reorientation,
decreased as the test temperature approached A, temperature. For curve ((1) which was
obtained at temperature higher than A,, only a small strain associated with plateau-type
deformation was Observed.26 At higher temperature (125°C) when T >> M,, plateau
deformation did not occur and the stress-strain curve looks Similar to that of common
metallic materials. The test temperature of curve (f) was closer to M,, and the plateau
segment, which was associated with stress-induced martensitic transformation, gave
almost 8% elongation. Plasticity by stress-induced martensite formation will be
discussed further below.

The stress-strain curves showed that the strain hardening rates during martensite
reorientation or stress-induced martensitic transfonnation (SIM) were very low. In
curves (a), (b) and (c) where the ﬁlms were elongated in the martensite, the typical strain
hardening rates dO/dc were about 0.75, 0.5 and 0.4 GPa respectively. Therefore, when
the temperature approached A,, the strain hardening rate for deformation associated with
martensite reorientation decreased. In curve (f) where the plateau segment was associated
with stress induced martensitic transformation, the corresponding strain hardening rate
was even lower (do/dc = 0.32 MPa). Therefore, it seemed that transformation strain can
be more easily to be created by inducing martensitic transformation than by martensite

reorientation.

 

2" The low stress for the deformation indicated that some martensite was still present at temperature above
the nominal A,.

104

4.2.2.1. Tensile Tests of Ni- Ti-Hf Films at Room Temperature. Room-temperature
mechanical properties are of interest because it is convenient to manipulate the materials
at room temperature. Figure 4.26 shows the stress-strain curves of in-situ crystallized Ni-
Ti-Hf thin ﬁlms with different Hf compositions. The strain rate was 0.04 mm/mm/min
(or 2 mm/min of cross-head Speed for 50mm long sample) and the samples were
elongated until failure. Yield stress (O'y), ultimate elongation (smax) and fracture stresses
(cf) of those samples were obtained by averaging the results of ﬁve samples and the
results are listed in Table 4.9. Typically, samples with different Hf compositions can be
elongated to ultimate strains from 5 to 9% at room temperature, and the fracture stresses
ranged from 1480 MPa to 1820 MPa. The hafnium content affected the yield stress,
ultimate elongation and fracture stresses of the ﬁlms, which is illustrated in Figure 4.27.
As hafnium content increased, the ultimate elongation decreased and the fracture stress
increased. Most importantly, hafnium content had a strong effect on room-temperature
yield stress of the ﬁlms. As the hafnium content was increased from 14.5% to 21 .4at.%,
the yield stress at room temperature was increased from 480 MPa to 770 MPa. This

result suggests that solution hardening resulted from substituting Hf for Ti.

Table 4.9. Room-Temperature Tensile Testing Results of In-situ Crystallized
Ni-Ti-Hf Films with Different Hafnium Content.

 

 

Film Composition Ni,,_2Ti,o,Hf2,, NimTimem Ni,.,,Ti,.,,,Hf,,2 Ni,,,,,Ti,,,,Hf,,S
Ultimate Elongation (%) 5.1 6.8 7.4 8.2
Fracture Stress (MPa) 1480 1610 1780 1820
Yield Stress (MPa) 770 640 570 480

 

Comparing the room-temperature tensile curves of binary and ternary ﬁlms, one

may see that the tensile behavior of the martensite in Ni-Ti-Hf thin ﬁlms, especially the

105

ones having higher haﬁrium content, differs from that of martensite in binary Ni-Ti SMA
ﬁlms. The plateau region in tensile curves for ternary ﬁhns showed much more working
hardening than occurred in the binary ﬁlms. However, considering the large difference in
transformation temperatures between binary ﬁlms and ternary ﬁlms (AA, = 95°C ~
160°C), difference in stress-strain curves for binary and ternary ﬁlms was not unexpected,
since the mechanical behavior of SMAS depends on the temperature with respect to their
transformation temperatures [Miyazaki et al. 1986]. Therefore, in order to compare the
mechanical behavior of binary and ternary ﬁlms, it is necessary to conduct tensile tests at

elevated temperatures.

4.2.2.2. Tensile Tests at High Temperatures. In order to ﬁnd the temperature effect
on the mechanical behavior of Ni-Ti-Hf SMA ﬁhns, in-situ crystallized Ni-Ti-Hf ﬁlms
were tested at various temperatures. Films having composition Ni,,,9Ti3,,,Hf,,_, were
selected to conduct such tests because the post-deposition annealed ﬁlms discussed
previously had a similar composition and thus a direct comparison can be made.27 The
samples were heated to 250°C (to make them fully austenitic) before cooling down to the
test temperature." In doing that, the material can be austenitic at lower temperature so
stress induced martensite formation can be more extensively studied.

Figure 4.28 shows a series of stress-strain curves of Ni,,_9Ti,,_,Hf,,,, ﬁlms
elongated at different test temperatures. Basically the temperature affected the tensile
behavior of the Ni4,,9Ti3,_,Hf,,, ﬁlms in a way similar to that for melt-solidiﬁed Ni-Ti-Hf

 

’7 Another reason of choosing this ﬁlm was its transformation temperatures. It was intended to test the
material at both martensite and austenite. On one hand, with material having M, above 100°C, test at
martensite can be facilitated in heating fumace since cooling down to sub-ambient temperature is not
necessary. On the other hand, very high A, temperature is troublesome because oxidation of the material
can occur rather rapidly. With the material having A, temperature below 200°C, testing at austenite can be
done in non-vacuum furnace without signiﬁcant oxidation problems.

2" This test procedure is referred as to “cool-down routine” in contrast to a more frequently used procedure

“direct-heating routine” in which the sample is heated to the test temperature from room-temperature at
which Ni,,,Ti,,_,Hf,,, ﬁhns are fully martensitic.

106

materials [Wang et al. 1997]. This is, at temperature much higher than A,, the ﬁlm
behaved like ordinary metallic material. Plateau-like deformation associated with SIM
took place as the test temperatures approached M,. The onset stress for SIM reached a
minimum as the temperature decreased to about M,. As the test temperature was
decreased to below M,, martensite reorientation replaces SIM as a mechanism for plastic
deformation, the yield stress started to increase, and the plateau-like behavior started to
diminish. It was also noted that the effect of temperature on ternary ﬁlms was similar to
that on binary ﬁlms although major differences exist [see Figure 4.25 and Figure 4.28].
The most marked differences are: (1), the strain associated with SIM or martensite
reorientation in the ternary ﬁlms was smaller than that in the binary ﬁlms (2 ~ 4% for
ternary vs. 6 ~ 8% for binary) which indicating that Hf substitutions reduced the
transformational strain dramatically; and (2) the strain hardening rate for SIM or
martensite reorientation related deformation was higher in ternary ﬁlms than in the binary
ﬁlms.

In attempting to quantitatively evaluate the plastic deformation associated with
SIM or martensite reorientation, we have estimated the corresponding strain hardening
rates, and the results are shown in Figure 4.28. The magnitude of the rates ranged from
3.7 to 14.2 GPa, with the maximum at 25°C (the lowest temperature, corresponding to
martensite reorientation) and the minimum at 160°C (corresponding to SIM). This
behavior indicated SIM related deformation in the Ni,,,9Ti3,_,Hf,,, ﬁlms was easier than
the deformation related to martensite reorientation. However, the strain hardening rate
for Ni,,,,9Ti3,,,,Hf,,,5 ternary ﬁlms was typically an order of magnitude higher than that for
binary Ti50,,Ni,9,, ﬁlms. That means that transformation strain (whether by SIM or by
martensite reorientation) in Ni4,_9Ti3,,Hf,,, ternary ﬁlms is much more difﬁcult to achieve
than that in binary ﬁlms, and the resistance increases as more strain is generated. It is
suggested that increasing stress will be likely to activate dislocation generation and

movement, resulting in poor strain recovery.

107

The relationship between the yield stress and tensile temperature was shown in
Figure 4.29. The yield stress increases almost linearly with the tensile test temperature,
which is consistent with the prediction from Clapeyron’s equation. Linear curve ﬁtting
gives the slope (do/d7) at about 6.23 MPa/°C, and this is close to the ratio (6.9 MPa/°C)
for stress induced martensitic transformation found for binary Ni-Ti alloy [Hodgson and
Brown 2000].

Theoretical transformation strain or optimal transformation strain may be

estimated using a number of measured data and Clapyron’s equation:

51.9.___AH_
dT T-V-Aa

(4.2)
With dor’dT= 6.23 MPa/K, AH = -1.23 kJ/mol, To, the equilibrium temperature,
simpliﬁed as (A, + M,)/2 = 435 K, and the molar volume V= N A x a3 /2 = 8.63 x 106 m3
/mol, (N A, the Avgadro’s constant, = 6.022 x 1023/mol, and a, the lattice parameter of
CsCl-type structure ofNi,,,9Ti3,,,Hf,,, austenite, z 3.06 A), then As was estimated at
5.2%. This value approaches that for in-situ crystallized binary Ni-Ti ﬁlms [Hou 1998].
However, the strain associated with the plateau-like deformation was about 2 ~ 4%,
which was considerably lower than the expected value from Clapeyron’s equation. The
disparity suggests that not all the martensite variants formed from stress induced
martensitic transformation were optimal for generating transformation strain. Grain
boundaries in fme-grain microstructure may force the formation of variants less favorable
for strain generation due to continuity restriction [Inoue et al. 1996].

In Figure 4.29, it is interesting to note that the yield stress at M, (140 °C for
Ni,,,9Ti,,,Hf,,,) was 325 MPa, because, theoretically, zero yield stress at M, temperature
is expected by linear regression using Clapeyron’s equation, as Observed by Hou in in-

situ crystallized Ni-Ti ﬁlms. It seemed that there existed a threshold stress to prevent

108

SIM from happening, and this threshold stress may or may not change with the tensile
temperature. It is also noted that the measured yield stress at the M, temperature was 375
,MPa instead of zero, but this is similar to the behavior of binary ﬁlms, because Hou
measured the yield stress at M, to be 100 MPa. The reason may be related to the high
local lattice energy resulting from a phase transformation in a very small volume ﬁaction.
Imaginably SIM can not keep going at zero stress because the local lattice distortion
works against it.

The tensile temperature also had inﬂuence on the ultimate strain. Figure 4.30
shows the relationship of the ultimate elongation and yield stress with temperature from
25°C to 250°C. The ultimate elongation ranged from about 6% to 11% and the yield
stress ranged from 375 MPa to 690 MPa. The dependence of yield stress and ultimate
elongation on test temperature may be divided into three regions. Region I (from 25°C to
150°C), Region II (from 150°C to 200°C) and Region 111 (from 200°C to 250°C). In
Region I, the yield stress decreased with increased test temperature. In the meantime, the
ultimate elongation was increased slightly. Considering that in this region the material
was at least partially martensitic, decreasing of yield stresses suggested martensite
reorientation became easier at higher temperature. Correspondingly, martensite
reorientation may be more complete and gave better ductility. In Region II, the ductility
remained near the maximum, and SIM stress increased linearly with temperature, which
was discussed previously. In Region III, the yield stress decreased again with increased
test temperature, and the ultimate elongation also decreased. This was related to the fact
that materials strength was decreased at higher temperature. Thus, the amount of
martensite formed by SIM transformation dramatically decreased, resulting a reduction of

overall elongation.

109

4.2.3. Shape-Memory Properties

Shape-memory properties of in-situ crystallized Ni,,,9Ti3,_,,Hf,,, ﬁlms were
evaluated by both the constrained thermal cycling method and the free recovery method.
Because most SMA applications are analogous to one of the two test methods, evaluation
by both methods may provide information useful for application design of Ni-Ti-Hf SMA

ﬁlms.

4.2.3.1. Free Recovery of Deformation in Martensite By Heating. Figure 4.31
shows a series of room-temperature stress-strain curves for load-unloading of
Ni,,_,Ti3,,,,Hf,,_5 ﬁlms at overall pre-strain of about 2%, 3%, 4% and 5%. The remaining
strain after unloading was measured at 0.9%, 1.8%, 2.7% and 3.5%, respectively, and
they are plastic strains (8p). Recovery took place upon heating the sample to 250°C
(65°C above A,), and residual strain was also measured. Figure 4.32 is a plot showing the
recoverable strain as well as recovery rate as a function of the plastic pre-strain. The x-
axis is the amount of pre-strain left aﬁer unloading (an), the left-side y-axis is the amount
of strain that is recovered (arc), and the right-side y-axis is the recovery rate (RR). It is
indicated that the pre-strain at 25°C (147°C lower than the A,, 100°C below M,) was
partially recoverable when the material was heated to austenite. The magnitude of
recoverable strain increased as the amount of pre-strain increased. Up to 2% of strain
was recoverable when the sample was strained at 5% (which gave rise to a 3.5% plastic
strain after unloading). The recovery rates ranged from about 55% to over 80%. Among
the four different pre-strains of martensite at 25°C, the smaller recovery percentage was
found at the lowest pre-strain and largest pre-strain. The maximum recovery rate was
about 83% when the overall pre-strain was 3% (i. e. , plastic pre-strain was 1.8%). The
difference of recovery rate at different pre-strain level suggested that slip occurred even at
the very initial stage of plastic deformation. After an initial hardening, most strain

resulted from martensite variant reorientation. When the (plastic) deformation went

110

beyond 3% or so, slip - mediated strain became more dominant.

The recovery rate measured in ﬁlms presented above was generally less than those
measured in melt-solidiﬁed bulk Ni-Ti-Hf SMAS [Pu et al. 1995; Wang et al. 1997], for
which up to 3% fully recoverable strain was reported. However, that does not certify the
properties of Ni-Ti-Hf ﬁlms was much inferior to that of melt-solidiﬁed NI-Ti-Hf SMAS,
because the recovery rate measured in melt-solidiﬁed Ni-Ti-Hf alloys involved bending
strain, while the strain for Ni-Ti-Hf ﬁlms was uniaxial. SMAS usually recover far better
from bending strain than from uniaxial strain.

Pre-straining of the martensite at higher temperatures and subsequently unloading
and heating the material to austenite was carried out to determine the recoverability of the
strain. Figure 4.33 shows the stress-strain curves of this series of experiments. In those
experiments, samples were deformed at 25°C (147°C below the A, temperature), 50°C
(122°C below the A, temperature), 100°C (72°C below the A, temperature), and 150°C
(22°C below the A, temperature), respectively. At each of those temperatures the
martensitic ﬁlm was strained to about 4.3% ~ 4.5% elongation and the plastic strain after
unloading was about 3%. The sample was then heated to 250°C (66°C above A,
temperature) so that the reverse martensitic transformation temperature was completed.
Residual strain was measured after the sample was cooled down the test pre-straining
temperature. Figure 4.34 shows the amount of strain recovered upon reverse
transformation and the recovery rate as functions of deformation temperature of
martensite. It is indicated that at ﬁxed pre-strain, the strain recoverable by heating
monotonically increased as the deformation temperature was approached A,. The
recovery rate also increased at the same time. Among the four tests at temperatures,
deformation at 25°C (147°C below the A, temperature) had the lowest recovery rate at
63%, and the deformation at 150°C (22°C below the A, temperature) had the highest

recovery rate at 83%.

111

4. 2. 3. 2. Transformational Superelasticity. As presented previously, SIM related
plateau-type stress-strain curve occurred when an initially austenitic sample was
deformed. Figure 4.35 shows the stress-strain curves of the sample deformed at various
temperatures (210°C, 200°C, 190°C and 180°C, which are 70°C, 60°C, 50°C and 40°C
above the M, temperature, respectively), to which the samples were cooled down from
250°C. The overall elongation was 4% at each of those temperatures, and both loading
and unloading paths at the test temperature were also recorded. Although incomplete,
isothermal recovery upon unloading was seen at different temperatures.” Figure 4.36
illustrates the recovery strain upon unloading as a function of the deformation
temperature. It Shows that best recovery occurs at 190°C (50°C above the M,) and 200°C
(60°C above the M,), and the recovery rates were between 80% to 90%. At higher
temperature (210°C) and lower temperature (180°C) the recovery rate fell to only 40% to
50%. It was noted that the recovery rates in superelastic deformation was typically
higher than SME recovery of the strain imparted to martensite.

The isothermal recovery of tensile strain seen in this experiment demonstrates that
the in-situ crystallized Ni,,,,Ti3,,Hf,,5 ﬁlms had superelastic properties. The dependence
of superelastic recovery on test temperature was related to the transformation
temperatures, and also to the dependence of material strength on temperature. At 200°C
and 190°C (60°C and 50°C above the M, temperature, respectively), a portion of the strain
resulted from stress induced martensitic transformation, and reverse transformation upon
unloading took place isothermally, giving rise to a recovery of the strain. However, the
low strain recovery rate at high temperature and at low temperature occurred for different
reasons. Lower recovery seen at higher temperature (210°C, 70°C above M,) may be
because at this temperature, more strain resulted from slip related deformation. This is

supported by the fact that the material yielded at lower stress at 210°C than at 200°C.

 

2" For simplicity, in calculating the recovery rate in this set of experiments, the conventional elastic strain
was included.

112

Lower recovery seen at lower temperature (180°C, 5°C below A,) was because the
temperature was so low so that the reverse transformation was not completed. This is
supported by the fact that the A, temperature of this sample was above the tensile test
temperature, and more recovery was Observed when the sample was heated to higher
temperature.

The effect of overall strain on isothermal recovery rate was also investigated.
Figure 4.37 shows stress-strain curves of Ni,,,,Ti3,,,Hf,,_5 SMA thin ﬁlms deformed at
190°C (5 0°C above the M, temperature and 6°C above the A, temperature) to total initial
strain of 2%, 3%, 4%, and 5% respectively. Isothermal recovery upon unloading was
also shown in the ﬁgure. Figure 4.38 shows the recoverable strain upon isothermal
unloading and recovery rate as functions of the initial strain. It is indicated that larger
recovery strain was increased as the initial strain was increased. Up to about 4% strain
was isothermally recovered upon unloading when the initial strain was 5%. In the
meantime, the recovery rates range from about 80% to about 95%, but the strain recovery
rate decreased monotonically as the initial strain was increased. This phenomenon may
suggest that the transformational strain contributed less and less to the overall strain as

the deformation continued.

4.2.3.3. Thermal Cycling Under Constant Load. Thermal cycling under various
constant loads between 50°C and 300°C was conducted for in-situ crystallized
Ni,,_9Ti3,,Hf,,,, ﬁlms. It was found that if the constrained thermal cycling started at
temperature well below M, temperature (such as room temperature which was about
100°C below M,), the ﬁrst heating excursion Showed little strain recovery while the
second heating excursion had substantial strain recovery, the same phenomenon that was
Observed in the RTA crystallized Ni,9,,Ti3,_2Hf,,,7 samples. This observation suggested
that for in-situ crystallized Ni,,,,Ti3,_,Hf,,_5 ﬁlms, transformation strain can be most easily

induced at temperature just above the M, temperature via optimal variant formation.

113

Figure 4.39 shows a series of strain-temperature curves during thermal cycling
starting from 300°C (115°C above A,) under different preset constant loads. It indicates
that transformational strain was obtained when material was cooled down from high
temperature and the strain may be completely or partially recovered upon re-heating the
sample to high temperature, forming a hysteresis loop.

From each strain-temperature hysteresis loop at different biasing stresses, a
number of data were taken including transformation temperatures, the strain resulting
from cooling to martensite from austenite ( cc ), and stress recovered on heating from
martensite to austenite (ch ). The residual strain er was Obtained by subtracting the
recovered strain ch (on heating) from the cooling induced strain cc. Figure 4.40 Shows
the cooling induced strain and recovered strain (on heating) as functions of applied stress
during thermal cycling. It is indicated that a small amount of transformational strain may
be induced at biasing stress as little as 100 MPa. With increased applied stress during
thermal cycling, the strain induced by cooling from high temperature (austenite)
increased continuously. Under applied stress less than 250 MPa, the strain recovered by
heating from low temperature equaled the strain generated on cooling from high
temperature, indicating a full recovery, corresponding to a closed loop in strain-
temperature curves. Under higher biasing stress during thermal cycling, the strain
recovered on heating was less than that generated on cooling, resulting residual stress
from each constrained thermal cycle. Therefore, 250 MPa was consider the maximum
stress for perfect recovery during constrained thermal cycling, and the corresponding
strain generated on cooling (which equaled the one recovered on heating), which was
measured at 2.8%, was the maximum fully recoverable strain for this sample. Figure 4.40
also shows that, as the biasing stress during thermal cycling was increased beyond the
maximum stress for complete recovery, more strain was obtained on cooling, and it
reached nearly 5% as the biasing stress was increased to 500 MPa. At the same time, the

recovered strain upon heating did not seem to be saturated at 3.6% (which was regarded

114

as maximum recoverable strain). Therefore, residual strain for a complete HT —> LT —)
HT cycle was increased as the biasing stress as further increased beyond the maximum
stress for full strain recovery.

The dependence of transformation temperatures (M,, M,, A, and A, temperatures)
on biasing stress during thermal cycling is illustrated in Figure 4.41. As is shown, the
transformation temperatures increased roughly linearly with increasing biasing stress
during thermal cycling. Linear ﬁtting 3° gave a number of dT/do slopes: dMS/do = 0.152
°C/MPa, dMde = 0.098 °C/MPa, dAs/do = 0.157 °C/MPa and (lAfldO' = 0.243 °C/MPa.
Noticeably the slopes of dMS/do and dAS/do were very close to each other, which the
other two slopes varied substantially.

In stress-induced martensitic transformation, Clapeyron’s equation predicted a
linear relationship between the transformation temperature and the applied stress. That
was consistent with our experiments because the strain produced during cooling occurred
at temperature above the M, temperature measured when the material was stress-free. In
addition, the values of dMS/do and dAs/do are in good agreement with the value of
dO,/dT measured in Figure 4. 28 in the stress-strain curve at different temperature (= 6.23
MPa/°C which gives an inverse of 0.161 °C/MPa). Using the measured data of dMS/do
and dAS/do and Clapeyron’s equation (4.1), the theoretical transformation strain was
estimated at from 4.9% to 5.1%. The difference between the measured value and
predicted value may suggest the formation of martensite variant under stress was still not
optimized for strain generation.

The linear ﬁtting for the transformation ﬁnish temperatures with biasing stress did
not go as well as that for transformation start temperatures with biasing stress.
Furthermore, the values of dedo and dAf/dO' looked rather aberrant from the data from
the values of dMS/do and dAs/do. Ironically, using the measured de/do value and

 

3° The last few data that were associated with high biasing stress were excluded in the linear ﬁtting because
at high biasing stress, deformation of the material relating to slip became more signiﬁcant.

115

Clapeyron’s equation, a 3.2% theoretical transformation strain was Obtained which was
fairly close to the measured value (3.6%). Nevertheless, the aberrance of de/dO' and
dA/do may be understandable if there exist components (such as ﬁne precipitates) that
can hinder the transformation signiﬁcantly. In that case, one may see strain still creeping
in at very low temperature at which a vast majority of austenite became martensite
(“actual” M, temperature was higher), and prolonged strain recovery at high temperature
at which only very little austenite still exited (“actual” austenite A, temperature was

lower).

4.2.4. Discussion on In-situ Crystallized Ni-Ti-Hf Films

In this section, as-deposited crystalline Ni-Ti-Hf SMA ﬁlms that were produced
by in-situ crystallization are discussed. High transformation temperatures as well as good
mechanical properties were obtained. For ﬁlms having 14.5% hafnium, the
transformation temperatures (M, = 135°C and A, = 180°C) were comparable with those
for melt-solidiﬁed alloys with similar hafnirun composition (Han et al., M, = 170°C, A, =
23 5°C for Ni49Ti36Hf,5 alloy). The samples had good ductility (up to 10%) and high
mechanical strength (Fracture stress over 1800 MPa at room temperature) that promise
industrial applications.

For the ﬁrst time, the mechanical properties of in-situ crystallized Ni-Ti-Hf ﬁlms
were studied at different temperatures. A side-by-Side comparison was made between Ni-
Ti-Hf ﬁlms and binary Ni-Ti ﬁlms. In general, the mechanical properties of the ternary
ﬁlms was affected by temperature in a way similar that for binary ﬁlms. That is, the yield
stress of martensite increased as the temperature decreased. The difference between
binary and ternary ﬁlms was that the ternary binary ﬁlms typically had higher yield
stress. In addition, strain-hardening rate for deformation associated with martensite

reorientation in the ternary ﬁlms was typically one order of magnitude higher that that in

116

the binary ﬁlms. The strain-hardening rate for strain related to stress-induced martensitic
transformation was even lower (0.32 GPa at 75°C). The ternary Ni,,_,,Ti3,,,Hf1 4,, ﬁlms
had much higher strain-hardening rate (3.7 ~ 14.2 GPa) in the deformation segment that
was supposed to be associated with martensite reorientation or stress induced martensitic
transformation. This result suggests that martensite reorientation encountered increasing
resistance as deforrrration proceeded, especially when the deformation was done at
temperatures substantially lower than the M, temperature. When the stress was high
enough, permanent strain resulting from slip was unavoidable. Similar to the results for
binary Ni-Ti ﬁlms, the strain-hardening rate associated with SIM in ternary
Ni,,_9Ti3,,,Hf,,_, ﬁlms seemed to be lower than that related to martensite reorientation,
which suggested that SIM was a better way to obtain transformational strain. This
observation was generally in agreement with the result from bulk Ni-Ti-Hf materials
[Olier et al. 1995].

Shape memory properties were another major topic that was investigated for the
in-situ crystallized Ni-Ti-Hf SMA ﬁhns. The evaluation was done by two methods, free
recovery method and thermal cycling. In the results obtained by the free-recovery tests, it
was found that recovery rate of the strain imparted to martensite Ni4,_9Ti3,_,Hf,,_, ﬁlms was
higher when the deformation was done at higher temperature (closer to A,). That means
more strain was generated from martensite reorientation. However, the strain recovery
was never 100%, suggesting slip accompanied martensite reorientation all the time.

The results of SME evaluation using constant-load thermal cycling demonstrated
a much better SME than that obtained from free-recovery tests. Nearly 3% of fully
recoverable strain was observed at an applied stress of 250 MPa. The discrepancy may
arise from the fact that the mechanisms of getting shape-memory strain are different in
the two testing methods. Unlike the situation in the free-recovery test where the shape-
memory strain resulted from reorientation and/or detwinning of existing martensite

variants, the shape memory strain generated by cooling the austenite under stress was due

117

to the formation of preferred martensite variants that gave maximum strain in the stress
direction. Apparently reorientation/detwinning of existing martensite variants may
experience considerable lattice resistance so it required high applied stress, which may
unavoidably activate other deformation mechanisms such as slip. Therefore, permanent
strain may occcur at the same time. Transformational strain generated from stressed
thermal cycling required much smaller stress. It seemed that 100 MPa applied stress was
adequate to affect the formation of martensite variants by SIM, and it was far below the
critical stress for martensite variant reorientation. The maximum stress for a full recovery
was as high as 250 MPa, indicating thermal cycling under such stress did not activate any
deformation mechanisms that produced non-recoverable strain. Comparison between the
results of the two SME testing methods suggests that stressed thermal cycling should be
the optimal working condition of the Ni-Ti-Hf SMA thin ﬁlms. However, if the material
has to be deformed isothermally, it should take place at temperatures that are best for
inducing the formation of preferred martensite variants, which corresponds to the
temperature that is just above M, temperature with the material still being austenitic.
Another interesting phenomenon worthy attention was that illustrated in Figure
4.17. That is, when Ni-Ti-Hf SMA ﬁlms were stressed at room temperature (martensite
state) and started cycling from there, there was little strain change in the ﬁrst heating
excursion. However, considerable amount of strain, depending on biasing stress, can be
generated on cooling. This phenomenon was also related to the fact a strain by martensite
reorientation was much harder to achieve than that by SIM. In design of an application,
this characteristic of Ni-Ti-Hf SMA ﬁlms needs to be considered. For example, in an
actuator fabricated in a Si-etch process, thin ﬁlm strip freed from the substrate is most
likely under slightly tension at temperatures below M, temperature. If one tries to deform
the ﬁlm in order to close a valve, it will need a great deal of stress and a permanent strain
will occur. However, instead of stressing the material at room temperature, if one heats

the SMA ﬁlm to austenite then applies a biasing stress, much lower stress will be needed

118

to get the necessary strain, and because this strain is related to SIM, it will recover when
the ﬁlm is heated back to austenite.

In comparison between Ni-Ti SMA and Ni-Ti-Hf high-temperature SMA, one of
the undesirable properties of the ternary material is that its austenite strength is low at
high temperature. That accounts for absence of superelasticity in Ni-Ti-Hf high
temperature SMA [Olier et al. 1995]. However, superelasticity was observed in Ni-Ti-Hf
SMA thin ﬁlms for the ﬁrst time, even though perfect recovery has not been realized.
This may result from high austenite strength of the Ni-Ti-Hf SMA thin ﬁlm at
temperature slightly above A, temperature. The strengthening of the material may
beneﬁt from the ﬁne-grain microstructure and also ﬁ'om the fact that the materials were

made in the form of thin ﬁlms.

119

Chapter 5
CONCLUSIONS

In this study, Ni-Ti-Hf high-temperature SMA ﬁhns were produced by both post-
annealing method and in-situ crystallization. For ﬁlms crystallized by post-annealing,
embrittlement occurred prior to the amorphous-to-crystalline phase transformation, even
though the crystallized ﬁlms still underwent martensitic transformation when the
materials reached proper temperature range. By using a high heating rate for
crystallization, the embrittlement was evidently alleviated. Rapid thermal annealing,
which had ramp heating rate as high as 6000°C/min, produced crystalline ﬁlms having
better shape-memory and mechanical properties than conventional vacuum annealing
with much lower ramp heating rate.

In-situ crystallized ﬁlms exhibited much better properties than those produced by
RTA treatment for the materials with the similar Hf composition. That was manifested
by several facts: First, the transformation temperatures were about 20°C to 50°C higher
in in-situ crystallized ﬁlms and the transformation hysteresis was also lower (~45°C for
in-situ crystallized ﬁlms versus over 65°C for post-deposition annealed samples). Higher
transformation temperatures and lower hysteresis are desired for actuators in MEMS
systems. Secondly, the in-situ crystallized Ni-Ti-Hf ﬁlms demonstrated better ductility
than the RTA treated samples (~8.2% for in-situ crystallized ﬁlms versus ~5.7% for post-
deposition annealed samples at room temperature). Thirdly, the in-situ crystallized Ni-
Ti-Hf ﬁlms had larger fully recoverable strain (~2.8% for in-Situ deposited crystalline
ﬁlms versus ~1.7%°C for post-deposition annealed samples) and higher maximum stress
for perfect strain recovery thermal cycling (~250 MPa for in-situ crystallized ﬁlms versus

~200 MPa for post-deposition annealed samples).

120

The difference of properties between in-situ crystallized samples and RTA treated
samples may result from different microstructural characteristics. Both X-ray diffraction
Spectra and TEM micrograph indicated that the microstructure of RTA treated samples
consisted in both martensite in very ﬁne clusters of variants (~50 nm, see Figure 4.15(d))
and B2 phase at room temperature. In contrast, the microstructure of in-situ crystallized
ﬁlms was exclusively martensite with ﬁne grain size at ~0.2 micron. While the
martensite variant reorientation or the martensite induced by applied stress during a
constrained cooling was responsible for the shape memory properties, the untransformed
BZ phase in the RTA treated samples was expected to have a negative effect on the
recoverable strain of the material. This was because the 82 phase, which did not reorient
or transform under applied stress, did not contribute to the formation of a transformation
strain. On the contrary, they may resist martensite reorientation or formation of
martensite under inﬂuence of an applied stress. Therefore, the maximum transformation
strain was decreased and higher biasing stress was needed to produce the same amount of
transformation strain compared to the situation where fully martensitic microstructure
may be obtained.

The transformation temperatures of RTA treated sample were apparently lower
than the in-situ crystallized ﬁlms. The reason may be ﬁne grain structure and existence
of the untransformed BZ phase in the RTA treated ﬁlms. Very ﬁne structure in
martensites typically decreases the martensitic transformation temperature [Su et al.
1994, You et al. 1997]. The B2 phase may also suppress the transformation due to
interfacial constraint, in the meantime, it is possible the composition of the matrix was
altered because of the cubic phases.

In comparison, from the standing point of material properties, in-situ
crystallization is the process of choice for making Ni-Ti-Hf high temperature SMA ﬁlms.
In-situ crystallization may offer advantages in other areas. For example, it is a one-step

process, and contamination can be minimized. Also, the stress in the ﬁlms may be

121

smaller because intrinsic stress (more likely compressive) and thermal stress (tensile

stress for Ni-Ti-Hf on Si or on glass) are cancelled on cooling. However, RTA treatment

is a fast process. In addition, the substrate is relatively cool compared to the ﬁlm due to

the unique means of heating in the RTA process. Therefore, thermal damage to the

substrate could be minimized if appropriate parameters are selected, even though the ﬁlm

is heated to high temperature during the process.

1.

The major conclusions of this research are highlighted as follows:

Two exothermic peaks appeared during isochronal heating of amorphous ﬁlms
prepared by sputtering. The ﬁrst peak, which occurred at temperatures from 300°C to
450°C at heating rates ﬁ'om 5 °C/min to 40°C/min and had an activation energy Of 162
kJ/mol, led to embrittlement of the amorphous ﬁlms; the second peak, which occurred
at temperature between 510°C to 560°C at heating rates from 5°C/min and had an
activation energy of 347 kJ/mol, resulted from the amorphous-to-crystall'me

transformation.

Amorphous Ni,9_,Ti3,,2Hf,,7 ﬁlms crystallized by conventional vacuum annealing
Showed martensitic transformation temperatures when cooled down to room
temperature. The martensitic transformation temperatures were increased by
substituting Hf for Ti but they were considerably lower than those in melt-solidiﬁed
Ni-Ti-Hf alloys. Heating rate for post-deposition crystallization did not affect the
transformation temperatures. However, the enthalpy for martensitic transformation in
post-deposition crystallized Ni,,,,Ti,,,_2Hf,,_7 ﬁhns, which in general was smaller than

that for crystallized binary Ni-Ti ﬁlms, increased as the heating rate was increased.

122

3. The ﬁlms crystallized by vacuum annealing contained Bl9’ martensite and B2 phases
at room temperature even though the M, temperature was typically over 80°C. The
volume fraction of B2 phase decreased as the heating rate increased. The grain

boundaries were irregular compared to classic metallographic microstructure.

4. Films crystallized by rapid thermal anneal at 650°C with a heating rate of 6000°C/min
Showed improved martensitic transformation characteristics compared to a
conventional vacuum anneal that had heating rates from 5°C/min to 40°C/min. The
RTA treated Ni,9,,Ti3,_2Hf,,,7 ﬁlms had room-temperature ductility of 5.8%, fully
recoverable strain of 1.7%, recoverable strain of 1.95% and maximum stress for full
strain recovery of 200 MPa. The microstructure of RTA treated samples displayed

ﬁne grains with a typical size less than 0.05 pm.

5. In-situ crystallized Ni4,_9Ti3,,,Hf,,, ﬁlms exhibited better mechanical and shape-
memory properties than the RTA treated ﬁlms. The room-temperature ductility was
8.2% and fracture stress was 1820 MPa. The martensitic transformation temperatures
were higher those in the RTA treated samples but still lower than those in melt-
solidiﬁed alloys having sarrre amount of hafnium addition. In-situ crystallized
Ni,,,9Ti3,,Hf,,_, ﬁlms had fully recoverable strain of 2.8%, maximum recoverable

strain of 3.6% and maximum stress for full strain recovery of 250 MPa.

6. The free-recovery method was used to evaluate shape-memory properties of the in-
situ crystallized Ni,,,,Ti,,,_,Hf,,,5 ﬁlms. The recovery rate was both pre-strain and
temperature dependent, but 100% recovery was never Observed. Transformational
superelasticity was observed at temperature above A, although the isothermal

recovery was not perfect.

123

7.

10.

The martensitic transformation temperatures as well as mechanical properties in in-
situ crystallized Ni-Ti-Hf ﬁhns were affected by the hafnium composition. As
hafnium composition in the ﬁlms increased from 14.5at.% to 21 .4at.%, the
transformation temperature increased from 180°C to 245°C for A,, and transformation
hysteresis increased from 45°C to 63°C. At the same time, room-temperature ductility

of the ﬁlms was reduced from 8.2% to 5.1%.

The mechanical properties of aS-sputtered crystalline thin ﬁlm was highly temperature
dependent and thermal history dependent. Higher ultimate elongation was obtained
around reverse (M —> A) transformation temperatures if the material was heated from
martensite, and better ductility was obtained around forward (A —) M) transformation
temperatures if the material was cooled down from austenite. The yield stress of

martensite decreased at higher temperatures.

The room-temperature microstructure of in-situ crystallized Ni,,,9Ti3,,,Hf,,5 ﬁlms was
fully martensitic with typical grain size on the order of 100 nm.. The martensite had a
monoclinic structure with lattice parameters of a = 2.63 A, b = 4.01 A, c = 4.77 A and
,3 = 991°. The microstructure of the RTA treated Ni49,,Ti3,_2Hf,,_, ﬁlms consisted in
martensite having the same crystallographic structure as that in in-situ crystallized

Ni,,.9Ti3,_,Hf,,, ﬁlms and a cubic phase having a lattice parameter a = 3.06 A.
Considering mechanical properties, transformation temperature and shape memory

properties, sputtering at elevated substrate temperature is the method of choice to

make high-temperature Ni-Ti-Hf SMA thin ﬁlms.

124

BIBLIOGRAPHY

125

BIBLIOGRAPHY

AbuJudom, D.N., Thoma, P.E., Kao, M.-Y. and Angst, D.R., “High Transformation
Temperature Shape Memory Alloy”, US Patent # 5,114,504, May 19, 1992.

Adachi, K. and Wayman, C.M., Metall. Trans, 16A, 1567, 1985.

Adachi, K., Perkins, J. and Wayman, C.M., Acta Metall., 36, 1343, 1988.
Angst, D.R., Thoma, PE. and Kao, M., J. de Physique IV, 5, pC8-747, 1995.
Arbuzova, LA. and Khandros, L.G., Phys. Met. Metalloved, 17, 390, 1964.

Barborini, E., Pieseri, P., Mutti, S., Milani, P., Biasioli, F ., Iannotta, S. and Gialanella, S.,
Nanostructured Materials, 10, 1023, 1998.

Barth, E.P., Spaepen, F., Bye, R. and Das, S.K., Acta Mater., 45, 423, 1997.

Bendahan, M., Canet, M., Seguin, J .-L. and Carchano, H., Mater. Sci. & Eng, B34, 112,
1995.

Bendahan, M., Seguin, J.-L., Canet, P. and Carchano, H., Thin Solid Films, 283, 61,
1996.

Besseghini, S., Villa, E. and Tuissi, A., Mater. Sci. & Eng, A273-275, 390, 1999.

Bodin, D., Moine P. and Eymery, J .P., Nuclear Instruments and Methods in Physics
Research, Section B: Beam Interactions with Materials, B23, 337, 1987.

Buchwitzz, M., Adlwarth-Dieball, R. and Ryder, P.L., Acta Metall. et Mater., 41, 1885,
1993.

Buehler, W.J. and Gilﬁich, J .V. and Wiley, R.C., J. Appl. Physics, 34, 1475, 1963
Buehler, W.J. and Wang, F.E., Ocean Eng, 1, 105, 1968.

Buehler, W.J. and Wiley, R.C., Naval Ordnance Laboratory Report, p61-75, 1961.
Busch JD. and Johnson, A.D., IEEE ’ 90, 40, 1990.

Busch, 1., Johnson, A.D., Lee, CH, and Stevenson, D.A., J. Appl. Physics, 68, p6224,
1990.

126

Busch, J .D., Johnson, A.D., Hodgson, D.E., Lee, CH. and Stevenson, D.A., Materials
Science Forum, 56-58, 729, 1990; J. Appl. Physics, 68, 6224, 1990.

Buschow, K.H.J. and Bekmans, N.M., J. Appl. Physics, 50, 6348, 1979.
Buschow, K.H.J., J. Appl. Phys, 56, 304, 1984.

Cesari, E., Ochin, P., Portier, R., Konomytsev, V., Koval Yu., Pasko, A. and Soolshenko,
V., Mater. Sci. & Eng, A273-273, 738, 1999.

Chang, F .C. and Read, T.A., Trans. AIME, 49, 191, 1951.

Chang, L. and Grummon, D.S., Materials Research Society Symposium Proceedings,
246, p141, Pittsburgh, PA, 1992.

Chang, L. and Grummon, D.S., Phil. Mag, A76, 163, 190, 1997.

Chang, L., Ph.D. Dissertation, Michigan State University, East Lansing, MI, 1994.
Chang, LC. and Read, T.A., Trans. AIME, 49, 191, 1951.

Chen, H.S., Mater. Sci. & Eng, 25, 59, 1976.

Chen, T., Scripta Metall. et Mater., 31, 1187, 1994.

Cross, W., Kariotis, A. and Stirnler, F., “Nitinol Characterization Study NASA CR-
1433, 1969.

d’Heurle, F.M., Metall. Trans, 1, 725, 1970; Inter. Mat. Rev., A34, 53, 1989.
d’Heurle, F .M., Berenbaum, L. and Rosenburg, R., Trans. AIME, 242, 502, 1968.
Davis, L.A., Metallic Glasses, American Society of Metals, Metals Park, Ohio, 1978.

Delaey, L., van Humbeeck, J ., Chandrasederan, M., Janssen, J., Andrade, A. and
Mwamba, N., Metals Forum, 4, 164, 1981.

Ditrnan, IE. and Bergman, L.A., in Proc. of 2nd inter Conference on Intelligent
Materials, ed. Rogers CA. and Wallece, G.G., p450, Virginia, Technical Publishing CO.
Inc., 1994.

Donkersloot, HO and van Vucht, J .H.N., J. of the Less-Common Metals, 20, 83, 1970.

Enami, K., Martynov, V.V., Tomie, T., Khandros, L.G. and Nenno, 8., Trans. Japn.
Inst. Metals, 22, 357, 1981.

127

Filip, P. and Mazanex, K., Scripta Mater., 35, 349, 1996.
Fonda, R.W. and Vandermeer R.A., Phil. Mag. A., 76, 119, 1997.
Ford, D.S. and White, SR, Acta Mater., 44, 2295, 1996.

Fursenko, A.A., Galjukov, A.O., Makarov, Yu.N., Lutovinov, D.S. and Rarnm, M.S., J.
Crystal Growth, 148, 155, 1995.

George, J ., Preparation of Thin Films, Marcel Dekker Inc., New York, 1992.

Gisser, K.R.C., Busch, J .D., Johnson, AD. and Ellis, A.B., Appl. Physics Lett, 61, 1632,
1992.

Glezer, A.M., Molotilov, B.V. and Utevskaya, O.L., Physics of Metals and
Metallography, 58, 147, 1984.

Golberg, D.V., Otsuka, K., Horikawa, H. and Mitose, K, United States Patent #6,64l,364,
Jun. 24, 1997.

Golberg, D., Xu, Y., Murakami, Y., Morito, S., Otsuka, K., Ueki, K. and Horikawa, H.,
Scripta Metall. et Mater., 30, 1349, 1994.

Gong, F ., Shen, H., Shen, G., Yang, Y., Yang, Z. and Wang, Y., Shape Memory
Materials ’94, Proceedings of the International Symposium on Shape Memory Materials,
p272, International Academic Publishers, Beijing, 1994.

Gruene, R., Oehring, M., Wagner, R. and Haasen, P., Proceedings of the Fifth
International Conference on Rapid Solidiﬁcation, p761-765, North-Holland, 1985.

Grummon D.S. and Gotthardt, R., Acta Mater., 48, 635, 2000.

Grummon, D.S., Hou, L., Zhao, Z. and Pence, T.J., J. de Physique IV, Colloque C8, 665,
1995.

Grummon, D.S. and Zhang, J., Phys. Stat. Sol., 186, 17, 2001.
Grummon, D.S., Zhang, J. and Pence, T.J., Mater. Sci. & Eng. A, A273-275, 722, 1999.

Gu, H.D., You, L., Leung, K.M., Chuang, C.Y., Chan, KS, and Lai, J.K.L., Appl.
Surface Science, 127-129, 579, 1996.

Gyobu, A., Kawamura, Y., Horikawa, H. and Suburi, T., Mater. Sci. & Eng, A273-275,
749, 1999.

128

Haasen, P., J. Non-Crystalline Solids, 56, 191, 1983.

Han, S., Zou, W., Jin, S., Zhang, Z. and Yang, D., Scripta Metall. et Mater., 32, 1441,
1995.

Han, X., Zou, W., Wang, R., Zhang, Z. and Yang, D.Z., Acta Mater., 44, 3711, 1996.

Han, X.D., Zou, W.H., Wang, R., Zhang, Z., Yang, D.Z. and Wu, K.H., J. De Physique
IV, 5, C8-753, 1995.

Hayashi, S., Int. Progress in Urethanes, 6, 90, 1993.

Hodgson, DE. and Brown, J .W., “Using Nitinol Alloys”, Shape Memory Applications,
Inc., San Jose, CA, 2000 (SMST—2000).

Horikoshi, H. and Tamura, N., Jpn J. Appl. Phys, 2, 328, 1963.

Hou, L. and Grummon, D.S., Scripta Mater., 33, 989, 1995.

Hou, L., Pence, T.J. and Grummon, D.S., Mat. Res. Soc. Proc., 360, 369, 1995
Hou, L., Ph. D. Dissertation, Michigan State University, East Lansing, MI, 1998.

Hua, 8., Su, C.M., and Wuttig, M., MRS Symposium Proceedings, 360, p375, MRS
Society, Pittsburgh, 1995.

Humbeeck, J.V., Trans. of ASME, 121, 98, 1999.

Ikuta, K., Fujita, H., Ikeda, M. and Yamashita, S., Micro Electo Mechanical Systems,
IEEE Proceedings No. 90CH2832-4, p38, IEEE, Piscataway, CA, 1990.

Inoue, H., Miwa, N. and Inakazu, N., Acta Mater., 44, 4825, 1996.

Ishida, A., Sato, M., Takei, A., Kase, Y. and Miyazaki, S., J. de Physique IV, 5, C8-70,
1995.

Ishida, A., Sato, M., Takei, A. and Miyazaki, S., Mater. Trans, JIM, 36, 1349, 1995.

Ishida, A., Sato, M., Takei, A., Nomura K. and Miyazaki, S., Metall. and Mater. Trans,
27A, 3753, 1996.

Ishida, A., Takei, A. and Miyazaki, 8., Thin Solid Films, 228, 210, 1993.

Jardine, A.P., Materials Research Society Symposium Proceedings, 271, p31, MRS,
Pittsburgh, PA, 1990.

129

Johnson, A.D., J. MicroMech. Eng, 1, 39, 1991.
Johnson, A.D., Martynov, V.V. and Minners, RS, .1. de Physique IV, 5, C8-783, 1995.

Johnson, A.D., Martynov, V.V. and Shahoian, E.J., Proceedings of SPIE, 2441, 165,
1995.

Johnson, AD. and Ray, C.A., “Shape memory alloy ﬁlm actuated microvalve”, US
Patent #5,325,880, July 5, 1994.

Kawamura, Y., Gyobu, A, Horikawa, H. and Saburi K., J. de Physique IV, 5, C8-683,
1995; J. Jap. Inst. of Metals, 60, 921, 1996.

Khachin, V.N., Matveeva, N.A., Sivokha, VP. and Chemov, D.V., Doklod. Nauk SSSR,
257, 167, 1981.

Kikuchi, T., Kajiwara, S. and Tomota, Y., J. Physique IV, 5, C8-451, 1995.
Kim, J .J ., Moire, P. and Steveson, D.A., Scripta Metall., 20, 243, 1986.
Kim, T., Ph. D. Dissertation, University of Maryland, 1994.

Kim, Y., Katsurai, M. and Fujita, H., Sensor and Actuator, 20, 33, 1989.
Kim, Y.D. and Wayman, C.M., Metall. Trans, 23A, 2981, 1992.

Kinuma, R., Matsumoto, M. and Honma, T., Spring Meeting of Japan Institute of Metals,
Sedai, p221, 1986.

Kissinger, H.E., Anal. Chem, 29, 1702, 1957.
Knowles, K.M. and Smith, D.A., Acta Metall., 29, 101, 1981.
Koval, Yu.N., F irstov, GS. and Kotko, A.V., Scripta Met. et Mat, 27, 1611, 1992.

Krulevitch, R, Lee, A. P., Ramsey, P. B., Trevino, J. C., Hamilton, J. and Northrup, M.
A., J. Microelectromechaincal Systems, 5, 270, 1996.

Krulevitch, P., Ramsey, P.B., Makowiecki, Lee, A.P., Northrup, MA. and Johnson, G.C.,
Thin Solid Films, 274, 101, 1996.

Kuribayashi K., Yoshitake, M. and Ogawa, 8., IEEE, p217, 1990.

Lapin, V.B., Egorushkin, V.B., Shabalovshaya, SA. and Ivanova, O.P., Solid State
Communications, 73, 471, 1990.

130

Lee, A., Dino, R.C., Krulecitch, A.P., Lehew, S., Trwvino, J. and Northrup, M.A., 8th
Int. Conf on Solid-State Sensors and Actuators and Eurosensors IX, Stockholm, p368,
1995.

Lee, J ., MS. Thesis, Michigan State University, 1994.

Li, D.Y. and Chen, L., Acta Mater, 45, 471, 1997.

Li, J.M., Quan, M.X. and Hu, Z.Q., J. Mater. Sci. Lett, 16, 921, 1997.

Lin, H.C., Wu, SK. and Chang, Y.C., in Shape Memory Materials ’94, ed. Chu Y. and
Tu, H., International Academic Publishers, Beijing, 1994.

Lin, H.C., Wu, SK. and Lin, J .C., Materials Chemistry and Physics, 37, 184, 1994.
Lindquist, P., Ph.D. Dissertation, University of Illinois, 1978.

Lindquist, P. and Wayman, C.M., in “Engineering Aspects of Shape Memory Alloys”,
eds. Duerig T.W., Melton K.N., Stochel D. and Wayman C.M., Bettherworth-Heinemann
Ltd., Boston, 1990.

Ling, CH. and Kaplow, R., Metall Trans, A1 1, 77, 1981.

Liu, Y. and McCormick, P.O., Acta Metall. et Mater., 38, 1321, 1990; Materials
Transactions, JIM, 37, 691, 1996.

Lo, Y.C. and Wayman, C.M., Scripta Metall. et Mater., 27, 1097, 1992.

LO, Y.C., Wu, SK. and Wayman, C.M., Scripta Metall. et Mater, 24, 1571’, 1990.
Loirat, Y., Bocquet, J .L. and Limoge, Y., J. Non-Crystalline Solids, 265, 252, 2000.
LotkOv, A.I., Grishkov, V.N. and Fadin, V.V., Phys Stat Sol, a(70), 513, 1982.

Lu, S., in Engineering Aspects of Shape Memory Alloys, eds. Duerig, T.W., Melton, K.N.,
Stockel, D., and Wayman, C.M., Buttherworth-Heinemann Ltd., Boston, 1990.

Madsen, J .S. and Jardine, A.P., MRS Proceedings, 308, 45, 1993.

Massalski, T.B., (ed), Binary Alloy Phase Diagrams, 2, p1768, American Metals Society,
Metal Park, Ohio, 1987.

Mathews, S.A., Wuttig, M. and Su., Q., Metall. and Mater. Trans, 27A, 2859, 1996.

Matsumoto, 0., Miyazaki, S., Otuka, K. and Tamira, H., Acta Metall., 35, 2137, 1987.

131

Melton, K.N., Proc. Inter. Conference on Martensitic Transformation, p1054, 1986.
Mercado, PG. and Jardine A.P., Proceedings of SPIE, 2441, 185, 1995.

Mercier, O. and Melton, K.N., Metall. Trans, A10, 387, 1979.

Miura, S., Meada, S. and Nakanishi, N., Phil. Mag. 30, 565, 1974.

Miyazaki, S. and Ishida, A., Mater. Trans, JIM, 35, 14, 1994; Mater. Sci. & Eng, A273-
275, 754, 1999.

Miyazaki, S. and Otuska, K., Metall. Trans, A17, 53, 1986; ISIJ Int, 29, 353, 1989.

Miyazaki, S., Hashinoga, T., Yumikura, K., Hirikawa, H., Ueki, T. and Ishida, A.,
Proceedings of SPIE, 2441, 156, 1995.

Miyazaki, S., Imai, T., Otsuka, K. and Suzuki, Y., Scripta Metall., 15, 853, 1981.
Miyazaki, S., Nomura, K. and Ishida, A., J. de Physique IV, 5, C8-677, 1995.

Miyazaki, S., Sugaya, Y. and Otsuka, K., MRS International Meeting on Advanced
Materials, 9, 251, 1989.

Moberly, W.J. and Melton, K.N., in Engineering Aspects of Shape Memory Alloys, eds.
Duerig, T.W., Melton, K.N., Stockel, D., and Wayman, C.M., Buttherworth-Heinemann
Ltd., Boston, p46, 1990.

Mrualeedharan, R.V., Phy. Stat. Sol. (A) Appl. Research, 125, 83, 1991.

Muddle, BC. and Traill, R.M., in Shape Memory Materials ’94, ed. Chu Y. and Tu, H.,
p689, International Academic Publishers, Beijing, 1994.

Mueller, M.H. and Knott, H.W., Trans. of the Metall. Society of AIME, 227, 674, 1963.

Mulder, J .H., “Investigation of High Temperature Shape Memory Alloys from the Ni-Ti-
Zr and Ni-Ti-Hf systems”, Ph.D. thesis, Univ. Twente, Netherlands, 1995.

Nakata, Y., Tadaki, T., Sakarnoto, H., Tanaka, A. and Shimizu, K., J. de Physique IV, 5,
C8-671, 1995.

Niikura, A., Tsai, A.P., Inoue, A. and Masumoto, T., J. Non-Crystalline Solids, 159, 229,
1993.

Nishida, M. and Honma, T., Scripta Metall., 18, 1293, 1984.

Nishida, M. and Wayman, C. M., Mater. Sci. & Eng, 93, 191, 1987.

132

Nishida, M., Wayman, GM. and Honma, T., Metall. Trans, A17, 1505, 1986; Metall.
Trans, A18, 785, 1987.

Nishida, M., Wayman, C.M., Kainuma, R. and Honma, T., Scripta Metall., 20, 86, 899,
1986.

Ogawa, K., Sato, K. and Miyazaki, S., Metall. and Mater. Trans. A, 28A, 1985, 1997.

Olier, P., Brachet, J.C., Bechade, J .L., Foucher, C. and Guenin, G., Journal de Physique
IV, 5, C8-741, 1995.

Olson, R.R., King, ME. and Wehner, G.K., J. Appl. Physics, 50, 3677, 1979.

Otsuka, K. and Shimizu, K., Scripta Metall., 11, 757, 1972; Int. Met. Rev., 31, 93, 1986.
Otsuka, K., Oda, K., Ueno, Y. and Piao, M., Scripta Metall. et Mater., 29, 1355, 1993.
Otsuka, K., Sawamura, T. and Shimizu, K., Phys. Stat. Sol, A5, 457, 1971.

Otuska, K., in Engineering Aspects of Shape Memory Alloys, eds. Duerig, T.W., Melton,
K.N., Stockel, D., and Wayman, C.M., Buttherworth-Heinemann Ltd., Boston, 1990

Perkins, J. and Sponholz, R.O., Metall. Trans, 15A, 313, 1984.
Potapov, P.L., Poliakova, NA. and Udovenko, V.A., Scripta Mater., 35, 423, 1996.

Potapov, P.L., Shelyakov A.V., Gulyaev, A.A., Svistunova, E.L., Matveeva, NM. and
Hodgson, D., Mater. Lett., 32, 247, 1997.

Pu, Z., Tseng, H.K., and Wu, K.H., Proceedings of SPIE, Smart Materials, 2189, 289,
1994; Proceeding of SPIE, 2441, 171, 1995.

Qu, Z. and Xie, T., Mater. Sci. & Eng, A134, 955, 1991.

Quandt, E., Halene, C., Holleck, H., F eit, K., Kohl, M. and SchloBmacher, P.,
Transducer '95 Eurosensors, 202, 1995.

Quandt, E., Halene, C., Holleck, H., Feit, K., Kohl, M. and SchloBmacher, P., the 8th
Inter Conf on Solid-State Sensors and Actuators and Eurosensors IX, Stockholm, 282-
PA1 1, 1995.

Rai, AK. and Bhaattacharya, R.S., Mater. Sci. & Eng, 85, 137, 1987.

Ray, C.A., Sloan, C.L., Johnson, A.D., Busch, J .D. and Petty, B.R., MRS proceedings,
276, 161, 1992.

133

Reed-Hill, RE, and Abbraschian, R., Physical Metallurgy Principles, 3rd edition, PWS-
KENT Publishing Company, Boston, 1992.

ROsner, H., Shelyakov, A.V., Glezer, A.M., Feit, K. and SchloBmacher, P., Mater. Sci. &
Eng, A273-275, 733, 1999.

Rostoker, W., Trans. AIME, 194, 209, 1952.

Russell, S.M., Hodgson , DE. and Basin F ., in Proceedings of2"d International
Conference on Shape Memory and Superelastic Technologies, eds. Pelton, A., Hodgson,
D., Russell S. and Duerig T., Paciﬁc Grove, CA, p429, 1997.

Saburi, T. and Nenno, S., Scripta Metall., 8, 1362, 1974; in Proceedings of the
International Conference on Solid-State Phase Transformations, eds. Aaronson, H.I.,
Langhlin, D.E., Sekera, RF. and Wayman, C.M., pl455, The Metall. Soc. AIME, New
York, New York, 1981.

Saburi, T., Nenno, S. and Fukuda, T., J. the Less-Common Metals, 125, 157, 1986.
Sakamoto, H., Otsuka, K. and Shinrizu, K., Scripta Metall., 11, 607, 1977.
SchloBmacher, B, Materials Letters, 31, 119, 1997.

Schroeder, TA, and Wayman, C.M., Scripta Metall., 11, 225, 1977; Acta Metall., 27,
405, 1979.

Shimizu, S., Xu, Y., Okunishi, E., Tanaka, S., Otsuka, K. and Mitose, K., Mater. Lett.,
34, 23, 1998.

Sivokha, V.P., Savvinov, A.S., Voronin, VP. and Khachin, V.N., Phys. Met. Metall., 56,
112, 1983.

Smialek, J .L. and Hehemann, R.F., Metall. Trans, 4A, 1571, 1973.

Sonius, S.E., Thijsse, B.J. and van den Beukel, A., Scripta Metall., 17, 545, 1983.
Su, Q., Hua, S.Z. and Wuttig, M., Proceedings of SPIE, 2189, 409, 1994.

Su, Q., Kim, T., Zheng, Y. and Wuttig, M., Proceedings of SPIE, 2441, 179, 1995.
Sun, L. and Wu, K.H., Proceedings of SPIE, 2189, 298, 1994.

Suzuki, Y., Xu, Y., Morito, S., Otsuka, K. and Mitose, K., Mater. Lett., 36, 85, 1998.

Takahara, Y., Mater. Sci. & Eng. A, A231, 128, 1997.

134

Thoma, P., Blok, A. and Kao, M., in Fundamental Aspects and Applications of SMA, ed.
Liu, C.T., Kunsmann, H., Otsuka, K. and Wuttig, M., MRS proceedings, 321, 1991.

Thoma, PE. and Boehm, J. J., Mater. Sci. & Eng, A273-275, 385, 1999.
Thoma, P.E., AbuJudom, D.N. and Fariabi, S., US Patent #4,881,981, 1989.
Thornton, J .A. and Hoffman, D.W., J. Vac. Sci. T echnol., A3, 576, 1985.
Thornton, J .A., Ann. Rev. of Mater. Sci, 7, 239, 1977.

Tsuchiya, K. and Davies, S.T., Nanotechology, 9, 67, 1998.

Tuominen SM. and Biermann R.J., J. Metals, February Issue, 32, 1998.

Ullakko, K., Huang, J .K., Kokorin, V.V. and O’Handley, R.C., Scripta Mater, 36, 1133,
1997.

van den Beukel, A. and Huizer, E., Scripta Metall., 19, 1327, 1985.

van den Beukel, A., Rapidly Solidiﬁed Materials, Proceedings of an International
Conference, San Diego, CA, USA, p193, ASM, Metals Park, Ohio, 1985.

van Vucht, J .H.N., J. the Less-Common Metals, 11, 308, 1966.

Walker, J .A., Gabriel, K.J. and Mehaegany, M., 5th Int. Conf on Solid State Sensors &
Actuators, Montreanx, Swiss, p123, 1989; Sensors & Actuators, A21-A23, p243, 1990.

Wang, L., Ma, L., Chen, M., Kimura, H. and Inoue, A., Mater. Sci. & Eng, A325, 182,
2002.

Wang, Y.Q., Cai, W., Zhu, W.H. and Zhao, L.C., Proceedings of the 2"“ Int. Conf on
Shape Memory and Superelastic Technologies, eds. Pelton, A., Hodgson, D., Russel, S.
and Duerig, T., Paciﬁc Grove, CA, USA, p77, 1997.

Wang, Z., Cai, W., Gao, Z.Y., Huang, L.Q. and Zhao, L.C., Proceedings of the 2"“ Int.
C onf on Shape Memory and Superelastic Technologies, eds. Pelton, A., Hodgson, D.,
Russel, S. and Duerig, T., Paciﬁc Grove, CA, USA, p113, 1997.

Wasilewski, R., Scripta Metall., 4, 127, 1971; NASA Technical Report, 1972.

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

135

Wayman, CM. and Duerig, T.W., in “Engineering Aspects of Shape Memory Alloys”,
eds. Duerig T.W., Melton K.N., Stochel D. and Wayman C.M., Bettherworth-Heinemann
Ltd., Boston, 1990.

Wayman, C.M., The Int. Symp. on Shape Memory Alloys, ed. Chu, Y.Y., Guilin, China,
p59, 1986; Progress in Materials Sciences, 36, 203, 1992.

Winzek, B. and Quandt, E., Z. Metallkd, 90, 10, 1999.

Wollants, P., DeBonte M. and R003, J.R., Z. Metallkd., 70, 113, 1979.

Wollants, P., Roos, J .R. and Delaey L., Progress in Materials Science, 37, 227, 1993.
Wu, K.H. and Ma, J .L., SMST Proceedings, 2000.

Wu, K.H., Liu, Y.Q., March, M. and Tseng, H.K., Proceedings of SPIE, 2198, 306, 1994.

Wu, SK. and Wayman, C.M., Metallography, 20, 359, 1987; Acta Metall., 37, 2085,
1989.

Wu, S.K., Khachaturyan, A.G. and Wayman, C.M., Acta Metall., 36, 2065, 1988.

Wu, S.K., Lin, HQ and Chou, T.S., Acta Metall. et Mater., 38, 95, 1990.

Wu, T.W. and Spaepen, F ., Acta Metall., 33, 2185, 1985; Phil. Mag, B61, 739, 1990.
Xie, C., Zhao, L. and Lei, T., Scripta Metall., 24, 1753, 1990.

Xie, C., Zhao, L., Duerig, T.W. and Wayman, C.M., Scripta Metall., 24, 1807, 1990.

Xu, Y., Shimizu, S., Suzuki, Y., Otsuka, K., Ueki, T. and Mitose K., Acta Mater., 45,
1503, 1997.

Yamasaki, T. and Ogino, Y., Journal of Jap. Inst. of Metals, 50, 595, 1986.

Yang J .H. and Wayman C.M., Materials Lett., 16, 254, 1993.

Yang, W.S. and Mikkola, D.E., Scripta Metall. et Mater., 28, 161, 1993.

Yawny, A.A., Sade, M. and Lovey, M., Materials Research Society Symposium
Proceeding, 459, p369, MRS, Pittsburgh, PA, 1997; Mater. Sci. & Eng, A273-275, 789,
1999.

Yi, H.C., Moore, J .J . and Petric, A., Metall. Trans. A, 23A, 59, 1992.

136

You, L., Chung, C.Y., Han, X.D. and Gu, H.D., Proceedings of the 2"”, International
Conference on Shape Memory and Superelastic Technologies, Asilomar Conference
Center, eds. Pelton, A., Hodgson, D., Russel, S. and Duerig, T., Paciﬁc Grove, CA, USA,
189, 1997

Yurko, G.A., Barton, J .W. and Parr, J .G., Acta Cryst., 12, 909, 1959.

Zhang, C., Zhao, L, Duerig, T.W. and Wayman, C.M., Scripta Metall., 24,1807, 1990.
Zhang, J ., MS. Thesis, Michigan State University, East Lansing, MI, 1997.

Zhang, J. and Grummon, D.S., Mat. Res. Soc. Symp. Proc., 459, 281, 1997.

Zhang, J. and Grummon, D.S., unpublished work, Michigan State University, 1998.

137

Work/Volume (J/m3)

108

 

 

 

 

0 1a SMA
7 .
1012 SL 9 1b SMA
106 .3TP
<> 4TB <>5a EM(a)
5a ES(a)0
105 . 97a PE
ZT
.Sb EM (P )
.SM
104
.6b ES
3 95c EM
10 ' 06c ES
09MB ”PE
102 1 I 1 1 1 1(Zn0)+
10 102 103 104 105 106 107

Cycling Frequency (Hz)

Figure 1.1. Work per volume versus cycling frequency for various microactuator.
1a: Ni-Ti SMA with the maximum one-time output (o= 500 MPa, 8 = 5%);
1b: Ni-Ti SMA with thousand of cycles (0'= 300 MPa, s= 2%),

° 999:195092459

Solid- liquid phase change (SL);
Thermo-pneumatic (TP);

Thermal expansion (TE) (nickel on silicon);
Electra-magnetic (EM);

Electrostatic (ES);

Piezoelectric (PE) (PZT);

Muscle (M);
Microbubble (MB).

: ideal value; 0: measured value [Krulevitch et al. 1996].

138

 

 

  

hysteresis

 

 

 

 

 

 

 

Resistivity, speciﬁc volume, etc

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

4;
Temperature
Mp

g exothermic

a:

3 44:

__, j endothermic
Mf Ms A; .Af

 

 

 

Temperature

Figure 2.1. A schematic showing physical property changes (a) and
endothermic/exothermic effects (b) during reversible martensitic phase
transformation. Transformation temperatures are determined by

the tangent - intercept method.

Ms: martensite start temperature; Mf: martensite ﬁnish temperature;
As: austenite start temperature; Af: austenite ﬁnish temperature;

Mp: martensite peak temperature; A,,: austenite peak temperature.

139

111
9.7

 

 

 

 

10.4
11
(a) .
s
7
5 6
2.1 A ' ‘7.6
001 011
“91.
11 ﬂ
7
4
3
2
1'00 01618

Figure 2.2. Theoretical recoverable strain of single-crystal Ni-Ti SMA as functions of
crystal direction developed by (a) Saburi et al. [1981], (b) Inoue et al. [1996].

140

Atomic Percent Nickel

 

    

 

 

 

 

 

 

 

 

 

 

0 10 20 30 40 50 60 70 80 90 100
1 8m 1 J J l l l l l l
1670 °c r
O 1600-"; '-
2, ‘ L
a . 1455 °c .
a 1400.: 1300 °c -
E 3 ‘
E. 3 ‘ 1304 °c
[2 1200.; -
g 3 1118 °c (Ni)
"3 (011)
E 1000— 942 0C P-
09. 3
g ; 882 °c ,.
a i 2'
[— 8001 M 765 °C 1: -
V E 630 °C 1
<— 0!“ {5‘
6m I T I V I ET I I r' I ﬁiﬁﬁ ' '1? I I """"" I ' ' '
0 10 20 30 40 50 60 70 80 90 100
Ti Weight Percent Nickel N‘

Figure 2.3. Equilibrium phase diagram of Ti-Ni alloy [Massalski 1987].

141

 

 

 

 

 

 

 

120 800
G 100— ._700 g
a. 80* a
E ._600 g.
H _J (D
a: 60 500 :.<.
._ (D
g 40— E
E“ 9
g 20~ --—400 a
m A
g 0-1 ,_ z
1: 300 3'3

-20-
l l l l 200
49/51 50/50 51/49

Figure 2.4. Schematic showing the dependence of austenite yield stress and
reverse transformation peak temperature of Ni-Ti alloy on the composition of the
material [Hodgson and Brown [2000].

142

Aging Temperature (°C)

 

oTiNi OTiNi+TiuNil4

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

o
ATiNi+Til,Nil4+Ti2Ni3ATiNi+TizNi3
90 1:1 TiNi+ TizNi3+TiNi3 ITiNi+TiNi3
800 “—‘0—
o
‘ ‘1’
IA A
700 ;—+———Al
o 5*
TN
0
§
600110 0 I. 0*
o o T
500W ' T “
0.1 1 10 100 1000 10000
Aging Time (hrs)

Figure 2.5. Isothermal time-temperature-transformation (I'I'I ) diagram
for Ti-52at.%Ni alloy [Nishida et al. 1986].

143

 

 

 

 

 

 

 

 

 

 

 

 

0 10 20 3O 40 50 60
Hfor Pd content (at.%)

[A F 1100 I. Hf'AbuJudom'
l : x . Hszan

18 600 . ; x o 9 . HfDlier

_‘ F

1 E 400 L . . 0 0 szLo

“3 _

l 8 200 l 3 o x Pdﬂolberg

t a. » 00.1 i . .

o ﬁe" o szLmdqmst 1
. H _ x o
| a. O o *-- 4‘
1“ ;

2 -200 I 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 ‘
l
l

 

Figure 2.6. Effect of Pd and Hf composition on reverse martensitic transformation
temperature of NiTi-base ternary alloys summarized from the literature.

144

 

 

 

 

 

 

 

 

 

 

 

600 . o AbuJudom

i a W
,1 500 1-. A-“ u
g) E .Potapov
g 400 ’5 1 . ————-———J x Mulder
S 300 :~__",____w . . XHan
é- ' o Olier
3 200 1__._ O . _,_
an ; >2
2 100 ~ 9.0 ‘

e ‘ I

o 0| x D
of‘w .........
0 10 20 30 40
chontent(at%)

 

 

 

Figure 2.7. Effect of Hf composition on transformation temperature of Ni-Ti-Hf
alloys summarized from the literature.

145

 

 

 

 

 

 

 

l
‘40 111111111i
{ 404142434445464748495051

Atomic % Ni

 

 

 

Figure 2.8. Effect of Ni composition on the transformation temperatures of Ni-Ti-Hf
shape memory alloy [Augst et al. 1995].

146

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

  
   

 

 

 

 

 

 

 

 

1 A ’ + -‘ 1—
1 800 800
1| NiTi 1411in1,
1
1 7; 400 3. 400 .
1 a 200 -- A = as 200 _ 0R(M)=535MPa
'1 o 1 1 1 1 o 1 1 1 1
1 0 1 2 3 4 5 0 1 2 3 4
1 Strain (%) ' Strain (%)
1’ “ 1 '1‘
, 800

, 80° Nme11 ‘1 11111111.,
; ,3 600 _T=150°C(Martensite) ,8 500 1T=150°c (Austenite)
1 g; 400 1 7; 400 ~

*7 200 _ = m 200
: VJ O’R(M) 390MP3 051M(M)=335MP3
1 0 1 1 1 1 0 1
1 0 1 2 3 4 5 0 1 2 3 4
1
1 Strain (%) Strain (%)
11 L _ 4 L

 

 

 

Figure 2.9. Deformation Behavior of Ni-Ti-Hf alloys at different test
temperatures [Olier et al. 1995].

147

Bias Sprin .

 
 
 
      
  
 

Pressure Vent
Actuator Die

TiNi SMA Fil

' , 1.1.-Ea

 
  
 

SST Spacer

Pneumatic Oriﬁce
4- Control Pressure

Fluid out 4— _. .— 4— Fluid in
Fluid Oriﬁce

Polymer Membran :

Bias Sprin .

Housin iPoppet P V t
ressune en
TiNiSMA Fil Actuator Die
' +

 
 
  

 

 
      
 
 
 
    

      

SST Spacer i a—‘
7*".- Pneumatic Oriﬁce
Polymer Membran {i 4— Control Pressure
13;;
Fluid out 4-7-n— 4- Fluid in

Fluid Oriﬁce

Figure 2.10. Structure of Si-based microvalve actuated using SMA Thin ﬁlm.
Upper: close position when the SMA ﬁlm is powered off;

Lower: open position when the SMA ﬁlm is powered on.
[Johnson et al. 1994].

148

 

 

 

 

 

high-temperature sta

low-temperature state

 

L/2

 

F/2, ((L(1+eo+e,))2-L2)°'5/2

 

 

Figure 2.11. Schematic showing realization of vertical cyclic displacement from a
horizontal SMA component under an applied force. The stress may also be calculated
using similar triangular principle.

149

 

 

 

 

1 Ti-26.4Ni-21.8Pd(at%)
2.5 i 973K-3.6ks W.Q. 1:1
2 L 1:: 9 ‘
g A
1.5 I Q
1 1 III
III [3 Strain on cooling
0.5 W ‘ Strain on heating
“ m
0~111LL11111111L11111111m11111

 

 

 

0 50 100 150 200 250 300

Figure 2.12. Shape-memory properties of sputtered Ni-Ti-Pd SMA thin ﬁlms
during thermal cycles under different applies stresses [Miyazaki et al. 1995].

150

Sewage ea <5 E192 .é :5 wetness as do censoﬁm 3. 2:5

mega

..llllllozobaoo 32m Iw1mq|l1|me 2.83

0]

 

 

 

 

  

 

    

 

   

 

 

 

 

 

 

 

 

 

 

 

c>_e>mmv o>ﬁ> 23851 g
lacuna: 5.8% \.B.. 61d
.m. .520: Seaman—m go? .8830
. cow weed wanmo \3 Hou—Em
_ Hilllllllonamno 832$
_ :m 833
Hou—95“.... H......w...: <

Swans“... 08. :. 2.38%

151

Th 1 B
Thermocouple A ermocoup e

 

Heating Plate A

 

 

 

 

 

 

 

Ti box for holding samples

Figure 3.2. Schematic of apparatus for conducting heating treatment for Ni-Ti-Hf thin ﬁlm.

152

 

 

 

700 é-

600 i” ~*
$500 - if \
‘ E400E / \

100 / —«
l l 1 l 1 1 1 i 1 '1 l 1 1 1 1 1 J l l 1

Time (secmd)

 

 

 

 

 

 

 

 

Figure 3.3. Typical temperature proﬁle of sample in a rapid thermal anneal treatment
for crystallization of Ni-Ti-Hf thin ﬁlms.

153

55% £8 E 2382::8 coumgommaeb omega egos—c 8
2958 SE 55 05 mo oogmmmoa mam—=32: a8 manage 05 mo oﬂeEonom 4mm 2:? m

aim a. 231';on a muin.

 

 

 

conned 3316:: U

 

 

 

c

 

   

850m ow§o> «51%on

 

S Emmi 39:0
A.

2 5808.5

 

seem co. m

 

 

 

 

 

 

 

154

Force

Metal wire having diameter d

SMA ﬁlm stripe

 

Figure 3.5. Schematic of the experimental setup of send-quantitatively determining
the ductility of Ni-Ti-Hf thin ﬁlms using bending test.

155

Stress

 

 

E
X I}: "X5 311a] ' ’
8'" Aﬁer heating 9 o Stram

 

Figure 3.6. Schematic of the method measuring recoverability of a pre-strain put to
Ni-Ti-Hf martensite.
A-B-C: loading curve of the sample;

C-D-E: unloading curve of the sample;
F: position that after sample was heated to >Af then cooled down to test temperature.

156

/ / / /
Mo extension strip

 

Cu gp'ppers

  
   

SMA thin ﬁlm
Thermocoup e

 

J.
:. 10V input
Cylindrigl ﬁrmace

Output signal

I__

 

 

 

Wei t and en rt
LVDT core

   
 

 

 

 

 

 

LVTD corecase
/ /

 

 

/

Figure 3.7. Schematic of the apparatus measuring shape memory properties of Ni-Ti-Hf
thin ﬁlm in constrained thermal cycles.

157

 

 

“1
J

Strain
c1

 

 

 

 

 

 

Temperature

Figure 3.8. Typical hysteresis in a strain-temperature curve for measuring shape
memory properties of Ni-Ti-Hf SMA thin ﬁlms.
A-B-C: First cooling curve, slope A-B results from thermal contraction
of the sample;
C-D-E: First heating up curve, slope C-D results from thermal expansion.
8c: strain induced by cooling the austenite to martensite;
eh: strain recovered by heating the martensite to austenite;
8,: residual strain during a complete constrained thermal cycle.

158

Composition (at% Hf)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

55 _ fs/ _ 40
: Hf _ :
54 :__ (M —>:14 39
53 : :13 38
I :
r: 52 T 12 “ 37
i : Tl W W 1 9
3.: *5- 1 ' ——> « a
3 51 t 111 p 36 "8
g 1 :1 1 3-
3.3 50 1 10 ‘ 35 8'
E 49 —— ‘9 i 34 §
8 :l : 2:!
48 :3 : 33
47 I :7 32
46 _ L Area for investigation I 31
45*1115111111‘51111L111111330
-3 -2 -l 0 l 2 3

Distance from Wafer Center (inch)

Figure 4.1. Composition uniformity of the thin ﬁlms deposited from an 8”-
diameter Ni485Ti375Hf14 target.

Deposition parameters: working gas Kr 1.0 mTorr, deposition power 1 kW,
working distance 81mm, substrate temperature 200 :l: 15°C, target life 20 kW- Hr.

159

Composition (at% Hf)

 

 

 

 

Composition (at% Ni)
(u. we) Honssodwoo

 

 

 

 

 

 

O 20 40 60 80 100

Target Wear Time (Hr- kW)

Figure 4.2. Composition variation of ﬁlms at central area with continuous
wearing of the Ni485'l‘i355Hf16 target. Deposition parameter: working gas Kr 1.0

mTorr, deposition power 1 kW, working distance 81mm, substrate temperature
200 i 15°C.

160

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

I I 4 I I I
4 Heating rate (°C/min) 5 10 20 40 l
1"peaktemperature (°C) 342 360 372 382
1St peak enthalpy (J/g) 6.2 5.5 3.6 2.5
2"d Peak Temperature (°C) 514 525 533 545
2"d peak enthalpy (J/g) 12.6 14.9 16.9 17.8
AI
\ 4) = 40°C/min
N 4%,; l
3
O
E
*5
Q)
:E.‘
\ ¢ = 20°C/min
/ K
4*. ¢ =10°C/min # Jk~
T ¢=5°C/min A /J \"~
50 100 150 200 250 300 350 400 450 500 550 600
Temperature (°C)

Figure 4.3. DSC curves of as-deposited amorphous Ni49JTi362Hfl4J ﬁlms during
isochronal heating from room temperature to 575°C at different heating rates.

161

Peak Temperature (°C)

Peak Temperature (°C)

Figure 4.4. Dependence of exothermic peak temperatures and enthalpies on heating
rate during heating as-deposited amorphous Ni49.,Ti36_2Hf14_7 ﬁlms in DSC instrument.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

550 _ J 25
540 Ed /0 (a) 20
I; / ’73 - m
530 : M A " 15 a
t : .5"
C A//:/ \_—’ . 4%
520 5 I 10 ~<
: <4 JI Q
r r 00
510 : - ; 5 V
500 : c 4 I 14 L I 145* o
o 10 20 3o 40 50
Heating Rate (°C/min)
400 : 4 1o
(b) 1
380 5‘—\ 1 3
I
360 g ; 6 g
340 j 4 B
I Q
320 ” 5 2 ‘3
300 I I . 3 +3 T I 0

 

 

 

Heating Rate (°C/min)

(a) high-temperature peak (2nd peak); (b). Low-temperature peak (lst peak).

162

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

-95 _ I I l I ﬁﬁ I I F f I I— F T I r I I _
.— H
: L°\ A: crystallization j
-10 H“ ' I
: I \ 1
A . I N ..
N " ...k __ ;_ 2 k '1
E-1O.5 " = ‘—‘
t : ) \ E 347 kJ/mol :-
.e. .. .1
E.“ " Slope = - 41.708 \ I -
_. -11 :— _ ‘ , C\ -
Ti 4 \o ,
i I
-12 I I L L I I I I I I I I I J I I
1.2 1.22 1.24 1.26 1.28 1.3
l/T, K" x 1000
‘9 _ I j I I T I f I I I I I I I I I I I Y I I ﬁI I I I I I q
I o .
-95 ' \\ B: low-temperature peak I
- '————- ______1.
A 10 : l R E: 162 kJ/mol 1
H ‘10.5 -_—’ = -
; : Slope 19.395 0 \ j
V : I \ 1
E -II _= I \o -
-115 Eh? I
-12 I. I I I I I I I I I I I I I I I I _l_ I I l I I I I I J I
1.5 1.55 1.6 1.65

1/T, K"x1000

Figure 4.5. Estimation of activation energy for crystallization transformation (2nd
peak, or the peak at higher temperature) and the transformation associated with the low-
temperature exothermic peak (lst peak on DSC curve) using Kissinger’s method.

163

 

87°C
\A ‘/

AH = 10.2 J/g

 

Rate = 40°C/min

\H#\/

/

 

 

 

 

 

f 85°C 152 c

N

\__—
AH = 9.3 J/g .
5 N Rate = 20°C/mm

158°C
5 86°C ,
8“”

Rate = 10°C/min

 

\g/

 

 

 

 

 

 

 

 

85°C 153C
‘__A +—-————— 1
AH = 7.2 J/g Rate = 5°C/min
k
ﬂ V
155°C
llllIllllIlllLIllllIlllL
0 50 100 150 200 250
Temperature (°C)

Figure 4.6. DSC curves of Ni49.lTi36_2Hf,4_-, ﬁlms crystallized by isochronal anneal
at different heating rates showing reversible martensitic transformation during

thermal cycling between 50°C and 250°C.

164

200

 

 

 

 

 

 

 

 

 

 

 

O E
g 150 Hr+ ______________ a “Afr"
a E
8. :
g 100 ................................................
P" ” 18+ + 4.; Mp
.3 :
g 50 ————————————————————————————————————————————————
ﬁ
0 1 I l l 44
0 10 20 30 4O 50
Heating Rate (°C/min)
12
9 .
En :
B 6 5
>4 _
D.
3
3,3 3 : ________________________________________________
O : l 1 l L l L
O 10 20 30 40 50

Heating Rate (°C/min)

Figure 4.7. Martensitic transformation temperatures (Ap and Mp) of annealed
Ni49_lTi36.2Hfl4.7 ﬁlms as functions of heating rates during isochronal heating
for crystallization. ‘

165

X-ray intensity (relative cps)

 

 

A

0
2:3 2
£8 “9'. 8‘
$8 9 8
"" v

(e)

 

 

(d)

 

 

(C)

 

 

 

{110}Cubic

 

km:

200

 

 

 

 

 

 

 

 

 

 

20 3o 40 50 60 7o
29 angle (°)

Figure 4.8. Room-temperature XRD spectra of as-deposited amorphous
Ni49JTi362HfM7 ﬁlms and the samples crystallized by isochronal heating to
575°C at different rates.

(a) as-deposited amorphous; (b), d) = 5°C/min; (c) o = 10°C/min;

(d) o = 20°C/min; (e) d) = 40°C/min.

166

 

Figure 4.9(a). TEM micrograph and SAED pattern of as-deposited amorphous
Ni49_lTi36_2HfM_7 thin ﬁlms after crystallized by isochronal heating to 575°C at
5°C/min heating rate.

167

 

Figure 4.9(b). TEM micrograph and SAED pattern of as-deposited amorphous
Nidg'lTilrﬁlefM.7 thin ﬁlms after crystallized by isochronal heating to 575°C at
40°C/min heating rate.

168

 

 

X-ray Intensity (relative unit)

__-__.._____..__.___._.__‘_____.___.._..__..._--____._~._._._--—_—_.___—_-

30 35 4O 45

2 theta angle (°)

Figure 4.10. Room-temperature XRD of as-deposited amorphous NinTimszm7

thin ﬁlms and those for ﬁlms after annealed at 300°C and 350°C for 1 hour,
respectively.

169

 

50

 

Figure 4.1 1. TEM micrograph and SAED pattern of as-deposited amorphous
Ni49JTi362HfM7 ﬁlms and those for ﬁlms after annealed for 1 hour at low temperatures.
(a) and (b): as-deposited amorphous; (c) and (d): after 1 hr anneal at 300°C;

(e) and (f): after 1 hr anneal at 350°C.

170

 

95°C /
Heating rate = 6000°C/min

AH =14.7J/g //

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

A \
\/
158°C
87°C
A /
Heating rate = 40°C/min
3 AH = 10.5 J/g
0
g _____——->\/
152°C
85°C
+—_— ’
AH = 7.8 J/g Heating rate = 5°C/min
ff
152°C
0 50 100 150 200 250
Temperature (°C)

Figure 4.12. DSC curves during thermal cycling between 50°C and 250°C of as-deposited
amorphous Ni49_,Ti36.2Hf14.7 ﬁlms aﬁer crystallized by isothermal annealing at different
heating rates.

171

— N
UI O
O O

perature (°C)

—
O
O

50

Transformation Tem

 

 

 

 

 

. D Vacuum Anneal at 556°C with I
5°C/min I
I ll Vacuum Anneal at 550°C with I

; 40°C/min
T ‘ I RTA at 650°C with 6000°C/min k”

 

 

 

— H
o u. 8

(S/f) Kdleqrug uoneuuogsu'er 1,

U"

Figure 4.13. Transformation temperatures and transformation enthalpy of Ni49.lTi36.2Hfl 4.7
ﬁlms isothermally annealed by different procedures (conventional vacuum anneal and RTA).

172

 

1M
[M

<2)M

(m 11..
(01‘
(002)M

 

(0(12)M

 

(100),,l
(-17 1..

(c): RTA
6000°C/min

—

W

(b): 40°C/min ‘

 

 

 

 

 

 

X-ray intensity (relative cps)

 

 

 

.2

'3

U
I“
E’.

U W :3
P'W L2
C
e B
O—

...» E3.

 

 

 

 

 

  

(a) 5°C/min

- _ l l l | -
20 3O 4O 50 60 70
29 angle (°)

   

 

 

 

 

 

Figure 4.14. Room-temperature XRD spectra of as-deposited amorphous
Ni49.,Ti36_2Hf14.7 thin ﬁlms aﬁer crystallized by isothermal annealing via different
ramp rates from room temperature to the annealing temperature.

(a) vacuum annealing at 550°C for 1 hour with a ramp rate of 5°C/min;

(b) vacuum annealing at 550°C for 1 hour with a ramp rate of 40°C/min;

(0) rapid thermal annealing at 650°C for 20 seconds with a ramp rate of 6000°C/min.

173

 

Figure 4.15 (a). TEM micrograph and SAED patterns of as-deposited amorphous
Ni49_lTi36.2Hf,4_7 ﬁlms aﬁer crystallized by isothermal annealing at 550°C for 1
hour with a ramp rate of 5°C/min from room temperature to 550°C.

174

 

Figure 4.15 (b). TEM micrograph and SAED patterns of as-deposited
amorphous Ni‘w‘lTimszM'7 ﬁlms aﬁer crystallized by isothermal annealing at
550°C for 1 hour with a ramp rate of 40°C/min from room temperature to 550°C.

175

 

Figure 4.15 (c). TEM micrograph and SAED patterns of as-deposited amorphous
Ni49JTi362HfMJ ﬁlms aﬁer crystallized by rapid thermal annealing at 650°C for
20 secs with a ramp rate of 6000°C/min.

176

 

Figure 4.15 ((1), TEM micrograph (higher magniﬁcation) of as-deposited
amorphous Ni49.,Ti36.2Hfl4_7 ﬁlms aﬁer crystallized by rapid thermal annealing at
650°C for 20 secs with a ramp rate of 6000°C/min.

177

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

I I I
I I I I
I I
1200 _ ( DI erﬂﬁﬂ I
I I
g 900 z / I
a /{ I
U? I
E 600 __ I
U)
I I
300
I
O 1 I 1 1 I 1 1 1 1 1 1 1 1 m 1 1 . 1 1 1 1 1 1 1 1 1
0 1 2 3 4 5 6 7I
Stra'n (%)

 

 

Figure 4.16. Room-temperature stress-strain curve of as-deposited amorphous
Ni49_1Ti36_2Hfl4.7 ﬁlms after rapid thermal annealing at 650°C for 20 seconds with
heating rate of 100°C/sec.

178

 

0.5%

1111

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

”a I F
\
° _. E =2 0...
§ — a
35
SI : I
r
’— O— W
~ D
I l I L l l l I 1 I I l f I l l l I l | I
O 50 100 150 200 250
Temperature(°C)

Figure 4.17. Schematic strain-temperature curve of RTA crystallized Ni49_1Ti36.2Hfl4_-,
ﬁlms during constrained thermal cycling starting from room temperature (fully
martensitic microstructure).

1st heating excursion: A —) B —) C;

1st cooling excursion: C —> D —> E;

2nd heating excursion: E —-) F —) G.

179

 

 

 

 

 

I’ i

L T
1%
v

A

 

350MPa ,

 

 

 

ﬁrﬁ

\ 3OOMPa

1
EL

\ \ 2501141):
I

 

 

 

I/ I

 

 

 

II

 

Strain

 

 

 

 

 

 

T’i—X

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

"" fZOOMPa
; lSOMPa
2: IOOMPa
l l l l l l I I I I l I I I 4 l I I I I
O 50 100 150 200 250
Temperature (°C)

Figure 4.18. Strain-temperature curves of as-deposited amorphous
Ni49_lTi36‘2Hf,4.7 thin ﬁlms crystallized by RTA anneal at 650°C for 20 seconds
during thermal cycling under different loads.

180

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

3.5 -r I I I l I I I I I I I I I I I I I I I I I I I I I V I I l l d

3 L -O- sc: strain on cooling I :

E £1— 8,,: strain on heating 5

2.5 :—— -A- 8,: residual strain :

83‘ 5 3

E 1 5 : )1 \1: I

3 1

w - d

E /z/ 3

0.5 _ (a q
50 100 150 200 250 300 350 400

Applied stress (MPa)

Figure 4.19. Dependence of strain induced by cooling form austenite to martensite and
that recovered by heating from martensite to austenite of RTA treated as-deposited
amorphous Ni49.1Ti36.2Hf,4‘7 SMA thin ﬁlms on applied stress during thermal cycling.

181

 

 

 

 

 

 

 

 

 

 

 

 

   

 

 

 

 

Ji 182°C 3 II
B h-
0
c:
*5 AH = 18.3 J/g
o I— 1 . m.
m I A1=21V=260°C
245°C
N148.2Tl30.4Hf21.4
l I I I l
O 50 100 150 200 250 300
Temperature (°C)
LII
_ Mp = 150°C
3 I—
O
a
8 _
:1: Mf= 143°C M3 = 157°C
’7 AH = 17.6 J/g _ L
J 1 """ Af = 217°C
Ap = 202
N148.0T133.9Hf18.1 I 1 AP = 210°C
0 50 100 150 200 250 300
Temperature (°C)

Figure 4.20. DSC curves of in-situ crystallized Ni-Ti-Hf thin ﬁlms with different
hafnium content during thermal cycling between 50°C and 300°C.
(a) Ni482Ti304Hf214§ (b) Ni48.oTi33.9Hf18.1; (C) Ni493Ti345Hf162§ (d) Ni48.9Ti36.6Hfl4.5

182

Figure 4.20 (continued)

 

    

 

 

 

 

 

 

 

 

 

 

 

 

- Mp = 140°C
E T
r: ..............
£3 — AH = 17.4 J/g

_ / A1 :11“ As =202°C

.. . . A =190°C

N1493T1345Hf162 p
41 I l I I
O 50 100 150 200 250 300
Temperature (°C)
1119:1351: El
3 I—-
a Mf= 127 M5 = 140°C
8 A: \
£ ' AH = 17.0 J/g
I ‘ 1 —g a
— A8 = 172°C A, = 184°C
Ni48.9Ti36.6Hfl4.5 Av = 1°°°C
I I I I I
O 50 100 150 200 250 300
Temperature (°C)

183

 

300 1.8

 

 

 

 

 

 

 

 

 

g 250 #
° ” 5
a g 5'”
8 200 B
9° ~ 2’.
E 5-
o 1 :1
H 5 ['11
g 150 a
7‘3 _ E
E — u
o _ °<
E 100 E
a : < 1' gt; g:
50 - - ...................... a 0.3 V
I Ap'Mp I
O 1 1 . 1 1 J 1 1 1 1 4 0
10 15 20 25
Hafnium Content (at.%)

Figure 4.21. Dependence of transformation temperatures (Mp, AP),
transformational hysteresis (Ap - Mp) and transformational enthalpy (AH) on
hafnium content in in-situ crystallized Ni-Ti-Hf ﬁlms.

184

 

 

 

 

 

 

 

 

 

 

 

 

 

 

200
,E
8
8
150
2
2
3100 a:
.33: ,E 8
a 8
8 o
E v
50 I ,
1:?
T 33
O 30 40 50 6O 7O
2 theta angle (degree)

Figure 4.22. Room-temperature XRD of the in-situ crystallized Ni48.9Ti36_6Hf14.5
SMA ﬁlm. The bars with arrows show the peak locations for TizNi or Ti4Ni20x
phases from powder diffraction data. '

185

 

Figure 4.23(a). TEM micrograph and SAED pattern of in-situ
crystallized NiwlgTiwﬁHf”.5 SMA ﬁlms taken at ambient temperature.

186

 

Figure 4.23(b). TEM micrograph (higher magniﬁcation) of in-situ
crystallized N14,,.§,Ti36.6Hf14.5 SMA ﬁlms taken at ambient temperature.

187

 

Figure 4.23(c) TEM micrograph (high magniﬁcation) of in-situ
crystallized Ni48_9Ti36_6Hf,4.5 SMA ﬁlms taken at ambient temperature.

188

 

 

 

 

 

 

 

800
600—
,8 _
Q‘ —
:2; ..
a 400— B
2 _
c7) _
— D
200L
— C
1 1 1 1 1 1 1 1 1 1 1
5 10
Strain (%)

Figure 4.24. Room-temperature stress-strain curve of an in-situ crystallized

A-B: Conventional elastic region;

B-C: Martensite reorientation and detwinning region;

C-D: Transition region;

D-E: 2nd elastic region accompanied by partial plastic deformation;
E-F: Conventional yield region.

189

15

 

1200 MPa

 

/I

(1) 125°C, T >> AAA)
|

 

7/

X

 

4—

(e) 100°C, T > A,(A)

 

/I/

((1) 150°C -)75°C ‘

X

 

 

 

 

 

 

 

 

 

T > M101)
/ (c) 75°C, T < As
,3 do/de = 0 32 GPa 4 I
a r x
E? / (b) 50°c,T<A.(M)
é // / I
do/de = 0.40 Gpam / l
/

 

 

/1

 

 

 

 

 

 

 

 

 

 

 

 

c, = M..- / I (a) 25°C,T<Mf(M)
/ x
1 l l l l l 1 L
10 15 20
Strain (%)

Figure 4.25. Stress-strain curves of an in-situ crystallized binary Ti505Ni495 SMA
(Mf = 55°C, M5 = 37°C, As = 78°C, Ar = 90°C) at different temperatures.

(a) T = 25°C; (b) T = 50°C; (c) T = 75°C, (d) T = 75°C but this temperature was
cooled from 150°C (53°C above Af); (e) T = 100°C; and (f) T = 125°C.

190

Stress (MPa)

 

500 MPa

Ni48.2Ti30.4Hf21.4

 

 

 

Ni48.0Ti33.9Hf18.1

 

 

X

2"} i

NI49.3T1345Hf 16.2 ﬂ

 

\

 

/.'.
/
/

 

 

 

 

 

 

 

 

 

 

 

 

 

 

// Ni48.9Ti36.6Hf 14.5
/ / X
/
T150 V
2.0 4.0 6.0 8.0 10.0
Strain (percent)

Figure 4.26. Room-temperature stress-strain curves of in-situ crystallized Ni-Ti-Hf
thin ﬁlms with different hafnium content.

191

2000 10

 

 

 

 

 

 

 

 

Cg :
3 8
C!
8 1500 j E.
I
g 16 a
o 1 ['11
E 1000 1 a
O I 0%
1E 44 8.
\ I I O
a _ I E:
g 500 -I 2 x
U) r n v
2 . I
.2 1 1
>4 1 I
0 L A m I J 1 l 1 1 1 1 0
10 15 20 25

Hafnium (at.%)

Figure 4.27. Dependence of room-temperature yield stress, ﬁ'acture stress and ultimate
elongation on haﬁiium content in in-situ crystallized Ni-Ti-Hf ﬁlms.

192

Cool-down mode
MS =140 °C;Mf=127 °C;
AS =172 °C;Af=184 °C.

(id/‘18 = 4.7 GPa

do/de = 3.7 GPa

do/de = 3.3

dO/de = 6.3 GPa

do/ds “ 8'3

dole!6 ‘ H-

d6 oldﬁ’ ‘2

dome;

 

O 3.0 6.0 9.0 12.0

Figure 4. 28. Stress-strain curves of in-situ crystallized Ni48 9Ti36 6Hf14 5 SMA thin
ﬁlms at different test temperatures using cooling-down routine.

193

 

800

__-__-_—-————————<

  
 

600 Z _______________________________
F 6.23 MPa/°C

 

Yield Stress (MPa)
A
o
o

200 Z- _ _ _ - ____________________ Cool-down mode _
- Ms =14o °C;Mf=127 °C; l
As=172 °C;Af=184 °C.

 

 

 

 

1 L l 1

 

L 1 1 l

180 190 200 210

 

140 150 160 170
Tensile Temperature (°C)

Figure 4.29. Dependence of yield stress of in-situ crystallized Ni48.9Ti36.6Hf145
SMA on tensile test temperatures between 200°C and 140°C using cool-down

procedure.

194

12 1000

 

 

 

A 10 800

ex:

.3 8 d g.
*5 600 "‘
a“ . 3‘
a“: 6 a?
*3 400 A
.33 4 5 §
5 Ms =14O °C;Mf=127 °C; 1 v

2 : AS=172°C;Af=184°C. ; 200
_____________________ r-_______-_-_---_-_:
h ‘ Region I II III , «

 

 

 

 

 

 

 

 

 

0 50 100 150 200 250 300

Tensile Test Temperature (°C)

Figure 4.30. Dependence of ultimate elongation and yield stress on test
temperature of in-situ crystallized Ni4wTi3MHfl4.5 SMA thin ﬁlms using cool-

down mode.

195

Qazoahﬁjx

W—

 

.wzumon Bologna EB mﬁumoiz $532 com:
038358-88“ “a 33983 "WEE £5 nzmmvomﬁqaﬁz BNESmbo 5.8-5 mo mots Ewﬁémobm ._m.v onE

g seam

anus... a:

gsﬁﬁnu28< @Egﬁnhzjw

 

\

a

 

\‘

\

\ \\

 

\ LY

 

 

 

 

 

 

 

 

 

 

o\oo._

 

 

 

 

 

 

 

 

dos: "Liven: n.<
6.51.262:ng

 

I.

 

 

 

_ _ _ lﬁ

 

Em

C
O
\0

(Raw) ssans

O
O
O\

cog

ooﬂ

196

2.5

’3
o\
v
'U
23
0
>
O
U
Q)
s...
.C:
a
3:!
CD

 

0.5 :

 

 

,_____-___.._-____

M8 = 140 °C;

 

Mf= 127°C;
As =172 °C;Af=184 °C.

 

 

ll 1 141 _1__Ll

 

l

 

 

O 1 2

Plastic Pre-strain 8p (%)

3

Figure 4.32. Recoverable strain and recovery rate of in-situ crystallized
Ni48.9Ti36.6Hfl4.5 ﬁlms with different amount of pre-strain introduced at room

temperature.

197

(m1) 912)] MQAOOQH

 

ﬂow: 83 28:8 masses

.88% 08 EB ..XF. 53m 3 Bin 33 wage—S. com: 58% 2:. $528: 626038 28 wﬁwmois $532 can
358388 Engage E Bongo—o mEE E5 35%.»: H3: Z team—aha 52.5 mo 8.55 gmémobm .mmé 0.5me

 

 

 

 

 

 

 

 

 

 

 

 

 

3 seam
\33 5:. b1 \ \waea .32 \wsewwmk mﬁmﬁmﬁ \_ o
\ \ \ m
25 a
N.\ \\\ 8.. m
“woo: n a. 0&2 u a $8 1. e e 0.2.. n H
I 82

. 0.3: 1:103: 16

$3 . .. .
Alv 0.5 u 2 .92: 1.2
d _ _ _ 4 82

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

198

 

 

 

 

 

 

 

 

 

 

 

 

e: 3 - ‘ 0.8
'53: a
o O
:3; L 0.6 §
§ 2 ~21
a . : 0.4 g
E, t : 8
m 1 ; --------- , ------- Ms=l40°C;Mf=127°C; 3 0.2

. As=l72°CgAf=l84°C. j

0 50 100 150 200

Pre-straining Temperature (°C)

Figure 4.34. Dependence of recoverable strain and recovery rate on
deformation temperature of in-situ crystallized NiMLQTimﬁHfht5 martensite
ﬁlms at a ﬁxed pre-strain of about 3%. The samples were heated from room
temperature to the test temperature before elongation.

199

.mobdﬁoqaow 68 305$ 8
@208 55 28 0.63 8 Bowen Bo? $388 .63 880mm .wﬁumois can wEEE maize 33558 03563 .2335
05 EV 358388 acouohmu um uoamwago nzmmﬁmﬁﬁdz cons—mambo 83-2.. mo moPSo mmobméﬁbm .mmé 8:me

 

 

 

 

 

 

 

 

 

 

 

 

g :55
mama... \ \ o
x \ \ \\ \ \ com
S
N \ \ k \ \ Oco my
\\ \\ R 8a m
9% u H 0&2 u H uooom a sac: n H
82
$3 0.3: u..< mmom: "J.
LIV 6.51s: 623%: 82

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

200

 

 

 

 

 

 

 

 

 

 

g 5

a”

g 4

E; 3 f

’3 <

§ : §

2 1 ______________________ Ms=140°C;Mf= 127°C; j 02

'g As=172 °C;Af=184 °C. 1

m 0p!1H1111414#¥¥111111111140
170 180 190 200 210 220

Tensile Temperature (°C)

Figure 4.36. Dependence of isothermal strain recovery on deformation temperature
for in-situ crystallized Ni48.9Ti36_6Hf14.5 SMA thin ﬁlm at ﬁxed total tensile strain of
4%. The samples were cooled from 250°C to the test temperature before
elongation.

201

00oz a 382: Ea evens Em as “sumesaﬁz 853%? 5.2.5 do 835 amaéﬁm .52. 2:5

g £25

38% :39 mo 355:8 Bogota no.“

 

\ \cm

 

\

 

\ \

cc

 

\

\ _\

ca

 

TN

 

 

 

 

 

 

 

o\oo.~

 

 

 

 

 

 

 

 

 

 

.93: "£60.65 "J
612L262: ",2

_ _ h

 

 

 

com _

 

of

 

(em) ssans

202

Isothermal Recovery Strain (%)

O)

 

03-501

 

N

 

9123 Human

 

: Ms=140°C;Mf=127°c; A 02
: ----------------------- AS=172°C;Af=184°C. 3

0 1 2 3 4 5 6

A

 

 

 

 

 

 

W—rT
..

Deformation Strain (%)

Figure 4.38. Dependence of recoverable strain and recovery rate on initial strain
of in-situ crystallized Ni48'9Ti366HfM.5 SMA thin films when elongation was done
at 190°C.

203

Strain (%)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

\ K
W \
OOMPa
\ \300Mpa
’3——”'\
\\ \ ZSOMPa
\
§ 200MP3_
A “=F—"'\
\k k ISOMPa
J #—
lOOMPa
llllllllllllllllll lllllle-l-AT—ﬁll lllllllll 111111111
0 50 100 150 200 250
Temperature (°C)

Figure 4.39. Strain-temperature curves of in-situ crystallized
Ni48.9,Ti36.6Hf14_5 thin ﬁlm during thermal cycling under various constant

Stl'CSSCS.

204

300

 

 

Cl ac: strain induced by cooling
5 E - A 8h: strain recovered on heating -----------------
: 0 8,: residual strain in a C/H cycle

    

Strain (%)

  

 

 

l

 

l l l J

O 1 00 200 300 400 500 600

Biasing Stress During Thermal Cycle (MPa)

Figure 4.40. Dependence of strain induced by cooling form austenite to
martensite and that recovered by heating from martensite to austenite of in-
situ crystallized Ni435,Ti36.6Hf14.5 SMA thin ﬁlms on biasing stress during

thermal cycling.

205

 

 

 

 

 

300 300
£3
0 250 H i 250
E
8.
a
0) i- r
E; 200 200
.2
‘3
E
‘H
g 150 150
E-*
100 4 1 i lift L 1 1 J 1 a, 100
o 100 200 300 400 500 600
Biasing Stress (MPa)

Figure 4.41. Dependence of transformation temperatures on applied stress during

thermal cycling of in-situ crystallized Ni489Ti366Hf145 SMA thin ﬁhns.

206