-
Combinatorial development and discovery of
ternary and quaternary shape memory alloys
Dissertation
zur
Erlangung des Grades
Doktor-Ingenieur
der
Fakultt Maschinenbau
der Ruhr-Universitt Bochum
von
Robert Zarnetta
aus Karl-Marx-Stadt
(jetzt Chemnitz)
Bochum 2010
-
Dissertation eingereicht am: 23.06.2010
Tag der mndlichen Prfung: 22.07.2010
Erster Referent: Prof. Dr.-Ing. Alfred Ludwig
Zweiter Referent: Prof. Dr.-Ing. Gunther Eggeler
-
Fr meine Frau Christiane
und unsere Tochter Mariella Jocelin
-
The mere accumulation of facts, even an extremely extensive
collection...does not constitute
scientific method; it provides neither a direction for further
discoveries nor does it even deserve the
name of science in the higher sense of that word. The cathedral
of science requires not only equip-
ment, but (needs) to indicate the pathway, by which the most
fruitful new material might be
generated.
Dmitri Mendeleyev 19141
1 J. J. Tennenbaum,
http://american_almanac.tripod.com/mendel94.htm
-
Executive Summary / Kurzfassung In der vorliegenden Arbeit
wurden Methoden der kombinatorischen Materialforschung entwickelt
und angewendet, um neue und/oder verbesserte ternre und quaternre
Formgedchtnislegierungen zu erforschen. Fr die Herstellung
sogenannter Dnnschicht-Materialbibliotheken wurde das
Katodenzerstuben genutzt. Unter Verwendung von keilfrmigen
Multilagenschichten konnten dabei komplette ternre
Legierungssysteme oder groe Zusammensetzungsbereiche quaternrer
Legierungssysteme abgeschieden werden. Zur
Hochdurchsatz-Charakterisierung der Materialbibli-otheken wurden
automatisierte Messverfahren fr die Bestimmung der Zusammensetzung
(energiedispersive Rntgenanalyse), der Struktur (Rntgenbeugung) und
des Phasenumwand-lungsverhaltens (temperaturabhngige
Widerstandsmessung) eingesetzt. Fr die Charakterisierung des
Formgedchtniseffektes dnner Schichten fand die sogenannte
Biegebalken-Methode (tempe-raturabhngige nderung der
Schichtspannung) Anwendung und wurde unter Nutzung
mikrostruk-turierter Si-Biegebalken in Matrixanordnung zu einem
parallelen Hochdurchsatz-Messverfahren weiterentwickelt. Mit den
beschriebenen Methoden konnten in den Legierungssystemen Ti-Ni-Cu,
Ti-Ni-Pd und Ti-Ni-Ag neue Formgedchtnislegierungen entdeckt und
erstmals die vollstndigen Zusammensetzungsbereiche identifiziert
werden, die eine reversible Phasenumwandlung zeigen. Das Ti-Ni-Ag
System zeichnet sich dabei durch die Unlslichkeit von Silber in der
Ti-Ni Phase aus, was zur Bildung von nanoskaligen
Silberausscheidungen und zu herausragenden
Formge-dchtniseigenschaften fhrt. Fr alle Legierungssysteme wurden
die Zusammensetzungs-Struktur-Eigenschafts-Beziehungen bestimmt,
die den Einfluss der Zusammensetzung auf die Phasenbil-dung und den
Einfluss der verschiedenen Phasen auf die
Phasenumwandlungseigenschaften, den Formgedchtniseffekt und die
mechanischen Eigenschaften der Dnnschichten beschreiben. Diese
ermglichten die Entwicklung verbesserter Formgedchtnis-Dnnschichten
fr die Anwendung als Mikroaktoren.
Eine, der in den Dnnschichten neu entdeckten
Legierungszusammensetzungen (Ti39Ni46Cu16) wurde als Massivmaterial
hergestellt, um zu untersuchen, inwieweit die bertrag-barkeit der
Ergebnisse von den Dnnschichten auf Massivmaterial gewhrleistet
ist. Es wurden Hochskalierungseffekte entdeckt und der
voraussagende Charakter der kombinatorischen Dnn-schichtexperimente
fr die Entwicklung von Massivmaterialen besttigt. Basierend auf den
Ergebnissen der ternren Legierungssysteme wurden im quaternren
Legierungssystem Ti-Ni-Cu-Pd neue Legierungen gesucht, fr deren
reversible Phasenumwand-lung eine verschwindend kleine thermische
Hysterese theoretisch vorhergesagt worden war. Durch die gezielte
Variation der Zusammensetzung und damit der Gitterparameter der
Legierungen in den Dnnschicht-Materialbibliotheken konnten
entsprechende Legierungen entdeckt und die Ergebnis-se anschlieend
auf Massivmaterialien bertragen werden. Die quaternren Legierungen
mit ver-schwindend kleiner thermischer Hysterese zeigten eine
beispiellose funktionelle Stabilitt, die die der binren Ti-Ni und
die der ternren Ti-Ni-Cu und Ti-Ni-Pd Legierungen deutlich
bertreffen.
-
Contents 1 Introduction
...............................................................................................................................
1
1.1 Aims of the work
..................................................................................................................
3
2 Fundamentals
............................................................................................................................
4
2.1 TiNi-based SMAs
.................................................................................................................
4 2.1.1 Phase transformation of TiNi-based SMAs
..................................................................
5 2.1.2 Thermal hysteresis of SMAs
.........................................................................................
8 2.1.3 Effect of third elements on transformation temperatures
............................................ 10
2.2 TiNi-based SMA thin films
................................................................................................
12 2.2.1 SMA thin film deposition
............................................................................................
12 2.2.2 SMA thin film stresses
................................................................................................
15 2.2.3 Ageing effects of Ni-rich and Ti-rich Ti-Ni thin films
............................................... 17 2.2.4 Thickness
effects of Ti-Ni thin films
..........................................................................
18 2.2.5 Functional fatigue of SMA thin films.
........................................................................
19 2.2.6 Current and future developments
................................................................................
20
2.3 Combinatorial materials science
........................................................................................
21 2.3.1 Materials libraries
........................................................................................................
21 2.3.2 High-throughput characterization
...............................................................................
22 2.3.3 Data analysis and visualization
...................................................................................
23 2.3.4 Current and future developments
................................................................................
24
3 Methods
....................................................................................................................................
25
3.1 Combinatorial materials science methodology
............................................................... 25
3.1.1 Sputter deposition of SMA thin-film materials libraries
............................................. 25 3.1.2
Compositional analysis energy dispersive X-ray analysis (EDX)
........................... 30 3.1.3 Structural analysis X-ray
diffraction methods (XRD) and temperature-dependent
X-ray diffraction (XRD(T))
........................................................................................
30 3.1.4 Phase transformation properties of thin films
temperature-dependent resistance
measurements (R(T))
..................................................................................................
31 3.1.5 Shape memory effect of SMA thin films cantilever
deflection method .................. 33 3.1.6 Mechanical properties
of substrate-attached thin films nanoindentation .................
35
3.2 Preparation of bulk SMAs
..................................................................................................
36 3.2.1 Phase transformation properties of bulk SMAs differential
scanning calorimeter
(DSC) and alternating current potential drop (ACPD) methods
................................. 36 3.2.2 Shape memory properties
of bulk SMAs
....................................................................
37
3.3 Microstructural analysis Transmission electron microscopy
(TEM) .............................. 37
4 Results and Discussion
............................................................................................................
38
-
4.1 Phase transformation properties of Ti-Ni-Cu SMAs
.......................................................... 38 4.1.1
The Ti-Ni-Cu system
...................................................................................................
38 4.1.2 Phase transformation characteristics of Ti-Ni-Cu thin
films ...................................... 43 4.1.3 Structural
properties of Ti-Ni-Cu thin films
............................................................... 47
4.1.4 Conclusions for Ti-Ni-Cu thin films fabricated using a
multilayer approach ............ 53 4.1.5 Shape memory properties
of Ti-Ni-Cu thin films - identification of optimized
compositions for high-frequency thin film microactuator
applications ...................... 54 4.1.6 High-throughput
measurements of mechanical properties for Ti-Ni-Cu thin films ...
59
4.2 Ti39Ni45Cu16 shape memory thin film and bulk alloys
....................................................... 67 4.2.1
Comparison of the phase transformation of thin film and bulk
Ti39Ni45Cu16 ............. 67 4.2.2 Comparison of shape memory
effects in thin film and bulk
....................................... 70 4.2.3 Microstructural
investigation of bulk Ti39Ni45Cu16
..................................................... 72 4.2.4
Transferability of thin film SMA compositions to bulk material
............................... 76
4.3 Phase transformation properties of Ti-Ni-Pd SMAs
.......................................................... 77 4.3.1
Phase transformation characteristics for Ti-Ni-Pd thin films
..................................... 77 4.3.2 Structural
properties of Ti-Ni-Pd thin films
................................................................ 81
4.3.3 Influence of precipitates on phase transformation
temperatures and thermal hysteresis
of Ti-Ni-Pd thin films
..................................................................................................
82 4.3.4 Ti-Ni-Pd shape memory thin films in view of microactuator
applications ................. 87
4.4 Identification of quaternary Ti-Ni-Cu-Pd SMAs with near zero
thermal hysteresis and unprecedented functional stability
.......................................................................................
88
4.4.1 Geometric non-linear theory of martensite (GNLTM)
............................................... 89 4.4.2 Quaternary
Ti-Ni-Cu-Pd thin film SMAs
...................................................................
89 4.4.3 Functional fatigue of quaternary Ti-Ni-Cu-Pd bulk SMAs
........................................ 94
4.5 Ti-Ni-Ag shape memory thin films
....................................................................................
98 4.5.1 Phase transformation characteristics for Ti-Ni-Ag thin
films ..................................... 98 4.5.2 Structural
characterization of Ti-Ni-Ag thin films
.................................................... 101 4.5.3
Shape memory properties of Ti-Ni-Ag thin films
..................................................... 105 4.5.4
Effect of individual layer thicknesses of the multilayer Ti/Ni/Ag
precursor thin films
on the phase transformation properties
.....................................................................
107 4.5.5 Internal stresses in Ti-Ni-Ag thin films proposed origin
of negative hysteresis .... 110 4.5.6 Conclusions for Ti-Ni-Ag thin
films fabricated using multilayer approach ............. 112
5 Conclusions and Outlook
.....................................................................................................
113
6 References
..............................................................................................................................
115
Publications
...................................................................................................................................
131
Acknowledgements
.......................................................................................................................
133
Curriculum Vitae
..........................................................................................................................
134
-
List of abbreviations & symbols
ACPD [-] alternating current potential drop Af [C] austenite
finish temperature As [C] austenite start temperature at.% [-]
atomic percent B2 [-] cubic austenite phase B19 [-] orthorhombic
martensite phase B19 [-] monoclinic martensite phase CAW [-]
cantilever array wafer CCS [-] continuous composition spread CTE
[-] coefficient of thermal expansion DSC [-] differential scanning
calorimetry [%] strain FIB [-] focused ion beam HAADF [-]
high-angle annular dark field LN2 [-] liquid nitrogen Mf [C]
monoclinic martensite finish temperature (B19) Ms [C] monoclinic
martensite start temperature (B19) Of [C] orthorhombic martensite
finish temperature (B19) Os [C] orthorhombic martensite start
temperature (B19) SEM [-] scanning electron microscope SMA [-]
shape memory alloy SME [-] shape memory effect STEM [-] scanning
transmission electron microscope Rf [C] R-phase finish temperature
Rs [C] R-phase start temperature RT [-] room temperature R(T) [-]
temperature-dependent resistance [MPa] stress (T) [MPa]
temperature-dependent thin film stress SAD [-] selected area
diffraction TEM [-] transmission electron microscopy T [K] thermal
hysteresis T [C] temperature TR [C] reference stress free
temperature XRD [-] X-ray diffraction XRD(T) [-]
temperature-dependent X-ray diffraction
-
1
1 Introduction
TiNi-Shape memory alloys (SMAs) are materials that after being
deformed mechanically can re-
turn to their original shape upon heating. Thus, they seem to
remember their original shape. By
doing so, these materials can induce large transformation
stresses and strains, have the ability of
pseudoelasticity and show high damping capacity, good chemical
resistance and biocompatibility.
Accordingly, shape memory alloys have attracted much attention
as functional materials and for
medical applications and more recently for
microelectromechanical system (MEMS).
Among different SMAs, Ti-Ni alloys exhibit the highest work
output per unit volume,
which if employed as thin film microactuator exceeds that of
other microactuation materials.2
Based on the large displacements and actuation forces possible,
the low operation voltage and high
power density and in combination with the ease of creating
friction free and non-vibrating move-
ment a number of thin film SMA applications have been developed
and demonstrated. Fig. 1.1 and
Fig. 1.2a-c show micropumps47, -valves811, -grippers12,13,
-wrappers14, -cages1, -mirrors15 and
Fig. 1.1 SMA thin film microactuators. (a) SEM picture of a
microcage capturing a micro-polymer ball1; (b) sche-matic drawing
of a SMA-actuated microgripper from Krulevitch et al.2; (c)
stress-temperature behavior of Ti-Ni-Hf and Ti-Ni-Cu thin films
sputtered on a (Mo+Nb) carrier. The nested hysteresis of the two
SMAs allows switching the curved Ti-Ni-Hf/(Mo+Nb)/Ti-Ni-Cu
composite upwards by the heat pulse T1T3T1 (1) and downwards by the
heat pulse T1T4T1 (2); (d) and (e) pictures of the bi-stable
actuator in downward and upward position, respectively.3
-
2
bi-stable actuators for tactile displays3. The compatibility
with micro-fabrication techniques, i.e.
patterning using photolithography and selective chemical
etching16 makes Ti-Ni-based thin films
attractive for MEMS applications. However, the limitations of
SMA actuators, i.e. the low actua-
tion frequency still prevent a widespread use of SMA
microactuators, despite the in this respect
positive scaling effects for thin films.
Additionally, the relatively high crystallization temperatures
of Ti-Ni-based thin films limit the
number of suitable substrates. Only recently Ti-Ni-Cu SMA thin
films were developed, that were
crystalline when deposited on heated polyimide substrates at 270
C, opening up a new field of
application.17
Next to micro-actuator applications, SMA thin films are of
special interest in medical applica-
tions due to large obtainable strains, the constant stress level
during pseudoelastic loading and their
good biocompatibility. In this respect, Ti-Ni microtubes are
attractive materials for biomedical de-
vices, such as micro-catheters and micro-stents19, whereas
superelastic TiNi-polymer-composites
could be used for novel applications in orthodontics and medical
instrumentation.20,21 Fig. 1.2d-e
show SEM images of Ti-Ni microtubes fabricated by rotating-wire
method, which allows to pro-
duce microtubes with uniform wall thickness and
composition.19
Despite the considerable progress in the development of SMA and
SMA thin films, no ground-
breaking actuation application has been commercialized, mainly
due to unsolved material issues
that limit the temperature range of application, the actuation
frequency and the repeatability (func-
tional stability) over a large number of cycles. Thus, a further
improvement of SMA properties
based on an advanced understanding of the underlying
composition-structure-property relation-
ships is needed to improve the material properties in order to
facilitate the wide-spread use of
shape memory alloys for actuator applications.
Fig. 1.2 SMA thin film microactuators and microtubes for
potential medical applications. (a) SEM image of a SMA actuated
micromirror (top view)15, (b) microspring and (c) free-standing
TiNiCu microtweezer structure18, (d) cross-section and (e) close-up
SEM images of the fracture surface of Ti-Ni microtubes fabricated
by rotating-wire method at 30 rpm. A potential starting material
for thin film stents.19
-
3
1.1 Aims of the work
Within this work the aforementioned limitations of Ti-Ni SMAs
will be addressed by investigating
ternary and quaternary SMA thin films using combinatorial
materials science methods for the acce-
lerated development and discovery of improved SMAs and the
underlying composition-structure-
property relationships. The following specific questions were
aimed to be answered:
(I) Can ternary and quaternary materials libraries with
well-defined composition gradients
suitable for investigating SMA properties be fabricated by
sputter-deposition methods?
(II) Can temperature-dependent 4-point probe resistance and
curvature measurements, as
well as nanoindentation be applied as high-throughput
characterization tools for the charac-
terization of the phase transformation properties, the actuator
response and mechanical
properties of SMA thin films, respectively?
(III) Do new, so far unknown SMA thin film compositions or
composition-structure-
property relationships exist for ternary Ti-Ni-based alloys?
(IV) Do quaternary Ti-Ni-Cu-Pd compositions exist that show zero
hysteresis, as post-
ulated by theory?
(V) Can composition-structure-property relationships determined
for thin films be used to
predict bulk materials behavior?
Accordingly, the thesis is comprised of three main parts (I)
introducing the fundamentals of SMA,
SMA thin films and combinatorial materials science, (II) the
specific methods developed and used
in this work, followed by (III) the results and discussion. The
results and discussion part is divided
into several sections, each related to a different ternary or
quaternary alloy system, thin film or bulk
material, respectively. An overview and introduction to each
alloy system will be given at the be-
ginning of the according section.
-
4
2 Fundamentals
2.1 TiNi-based SMAs
The shape memory effect (SME) is the ability of a material to
recover its original shape after a de-
formation that extends well beyond the elastic regime. It
appears in alloys showing reversible
thermoelastic martensitic transformations. The martensitic
transformation is a diffusionless struc-
tural change associated with a large shear-like deformation. The
martensite (low temperature
phase) is deformable and the deformation can be recovered
perfectly by heating to the austenite
(high-temperature phase). The SME was first observed in the
Au-Cd alloy system by Chang and
Read in 195122 and in 1963 the SME in Ti-Ni was discovered by
Buehler et al..23
However, it was only at the start of the 1980s that the shape
memory mechanism in the
Ti-Ni system was extensively investigated, due to difficulties
with the fabrication of single crystals
and the complexity of the martensitic transformation.24 One of
the difficulties being, the compara-
tively small equilibrium composition region for single phase
TiNi at low temperatures (< 600 C:
~49.5 at.% Ni to ~50.5 at.% Ni), as shown in Fig. 2.1, which
leads to the necessity of very precise
composition control. The comprehensive review on the physical
metallurgy of Ti-Ni-based shape
memory alloys by Otsuka and Ren published in 2005 highlights the
significant progress made in
understanding the martensitic transformations in this system in
recent years.26
Fig. 2.1 Equilibrium phase diagram of the Ti-Ni system.25
-
5
2.1.1 Phase transformation of TiNi-based SMAs
The martensitic transformation is a first order diffusionless
phase transformation. Thus, no change
in chemical composition occurs and a distinct phase boundary
exists between the austenite phase
and the transformation product (martensite)28. The
transformation starts upon cooling from the
high-temperature phase A (austenite) at the
martensite-start-temperature (Ms). This temperature is
lower than the thermodynamic equilibrium temperature T0, since
an additional driving force is
needed (GA,M) in order to compensate the nucleation energy of
the martensite and the elastic
stress fields due to the formation of martensite in the
austenite27 which results in an undercooling
(u), as illustrated in Fig. 2.2. The nucleation energy can be
small or even negligible, if the auste-
nite and martensite phase are crystallographically compatible
and thus the elastic stress field ener-
gy is minimized (see also 2.1.2).29
The mechanisms of the shape memory effect (temperature induced
transformation) and superelas-
ticity (stress-induced transformation) are depicted in Fig. 2.3
schematically, using a two-
dimensional crystal lattice as a model. Fig. 2.3 shows that both
superelasticity and the shape mem-
ory effect can occur in the same sample, and which occurs,
depends upon the test temperature. The
austenite crystal structure (red) is completely transformed to
the martensite (blue) upon cooling
below Mf (martensite finish temperature) and the martensite is
characterized by the formation of
different variants (labeled A and B in Fig. 2.3), that have the
same crystal structure but differ-
ent orientations (twinning related). Due to the
self-accommodation mechanism of the martensite
variants the macroscopic shape of the material will not
change.30 The variants in the thermally in-
duced martensite phase are twinning related and upon the
application of an external load the twin-
ning planes are easily moved. The variants will reorient to
accommodate the stress and the coales-
cence of various variants into a single variant will produce a
transformation strain .31 The marten-
site phase at this temperature is stable, so the reoriented
martensite phase is maintained. If the ma-
terial is heated above Af (austenite finish temperature), the
reverse martensitic transformation will
be completed and the transformation strain will be recovered.
This behavior is commonly termed
as the shape memory effect (SME).
Fig. 2.2 Schematic of the temperature dependence of the free
enthalpy G. The enthalpy curves of the low-temperature phase A
(austenite) and the high tempera-ture phase M (martensite) are
shown. The phase transformation starts after an under-cooling (u)
at the martensite start tem-perature Ms (after Hornbogen27).
-
6
For temperatures above Af, the austenite phase is stable,
however on the application of an external
stress the martensite phase can form due to stress-induced
martensitic transformation. The marten-
site phase in this case is unstable since the temperature is
above Af, so upon the release of the ex-
ternal load the material will return to the austenite phase.
During the deformation process the
stress-induced martensite can accommodate a transformation
strain , and therefore it behaves as a
superelastic material. For the case of Ti-Ni the recovery strain
can be as high as 10 %, which is
more than 10 times the elastic limit of normal alloys and
metals.28
Three different transformation paths for the reversible
thermoelastic martensitic transforma-
tion from the austenite (B2) to the monoclinic martensite (B19),
to trigonal martensite (R-phase)
or orthorhombic martensite (B19) are known for Ti-Ni-based
alloys, as shown in Fig. 2.4. For the
latter two, an additional successive transformation step to
monoclinic martensite can be observed,
thus resulting in a two-step transformation.
Unique lattice correspondences between the austenite and
martensite phases exist, such that the
reverse transformation path is restricted. Thus, the
reversibility of the thermoelastic transformation
Fig. 2.4 Three martensitic transformation paths in Ti-Ni-based
alloys. (Adapted from Otsuka and Ren26). Conti-nuous lines indicate
a one-step transformation, the dashed line the possibility of a
successive second transformation step, resulting in a two-step
transformation.
Fig. 2.3 Mechanism of shape memory behavior. Shape Memory Effect
(SME) left hand side, Superelasticity (SE) right hand side. The
temperature scale is indicated at the top, Af is the finish
temperature for the formation of austenite upon heating, Mf is the
martensitic transformation finish temperature, at which the
martensite phase has completely transformed upon cooling; is the
transformation strain (see text for details). Arrows indicate a
transformation or reo-rientation of martensite, due to heating
(red), cooling (blue) or stress/load (black).
-
7
is guaranteed. The lattice correspondence of the B2/B19, B2/B19
and B2/R-phase are summarized
in Fig. 2.5. It is evident that the transformation strain is
dependent on the crystal orientation. This is
seen most clearly for the B2R-phase transformation, where the
R-phase forms by elongation
along any one of the [111] directions of the B2 structure, and
thus the transformation strain along
the [111] direction is largest, whereas the transformation
strain along [001] is nearly equal zero.32
In polycrystalline materials the orientation dependence is
averaged, if no specific texture is present.
Due to the shear-like adeformation of the lattice during the
reversible phase transformation, the
transformation temperatures of SMAs are influenced by stresses,
and vice versa the plateau stress
during superelastic loading by temperature, according to the
Clausius-Clapeyron equation:28,33,34
, (1) where is the uniaxial applied stress, S and H* are the
transformation entropy and enthalpy per
unit volume, strain and T temperature26. The stress rate /T for
B2B19 transformation in
Ti-Ni falls typically in a range between 4 20 MPa/K, whereas the
R-phase shows a characteristi-
cally higher stress rates in the range of 30 70 MPa/K.35 In
other words, the Rs temperature is less
sensitive to stress as compared to the Ms temperature, which in
turn however can result in a change
of the transformation path from a two-step to a one-step
transformation with increasing stress.36
The insensitivity of the Rs results from the significantly
smaller lattice deformation of the R-phase
Fig. 2.5 Lattice correspondence of the martensites in
Ti-Ni-based SMAs. (a) B2 austenite phase with a FCT cell
delineated; (b) orthorhombic martensite B19, formed by
shear/shuffle of the basal plane 110 B2 along 1 0 direc-tion; (c)
monoclinic martensite B19, which is viewed as a B19 structure
sheared by a non-basal shear 001 1 0 B2; (d) trigonal martensite
R-phase formed by stretching the cubic austenite lattice along 111
diagonal direction. The axis a, b and c represent the principal
axes in that deformation (from Otsuka and Ren26).
-
8
transformation, as compared to the B2B19 transformation. Fig.
2.6 shows typical stress-
temperature relations for free-standing thin films and bulk SMAs
and depicts the increase of the
transformation temperatures with increasing stress.
However, for substrate-attached thin films, the
Clausius-Clapeyron relationship is found to
be not applicable for describing the stress-dependence of the
transformation temperatures during
reversible phase transformation or thermal hysteresis39,40, as
discussed in more detail in 2.2.2.
2.1.2 Thermal hysteresis of SMAs
The reversibility of the structural phase transformation has a
profound technological implication
for the application of SMAs and their fatigue life41 and thus
understanding of the underlying me-
chanism govern reversibility, i.e. the thermal hysteresis of
SMAs is essential. Currently, the most
widely known and commonly accepted explanations of hysteresis
are the pinning of interfaces by
defects and thermal activation27,42. However, a close
examination of the limited experimental data,
i.e. for alloys with different dislocation densities as a result
of different quenching methods43 and
theoretical calculations of the energy barriers associated with
nucleation (thermal activation) do not
seem to unambiguously support these ideas.29
More recently, a new theory on the origin of the reversibility
of phase transformations has
emerged29,4447, suggesting that the growth of fully developed
austenite/martensite interfaces is
responsible for determining the size of the hysteresis. Fig. 2.7
shows an optical micrograph and a
schematic of the microstructure of a twinned
austenite/martensite interface, including an elastic
transition layer, which forms due to the crystallographic
incompatibility of the two phases. An
energy barrier is associated with the developing
austenite/martensite interface, due to stored elastic
Fig. 2.6 Stress-temperature relations for free-standing SMA thin
films and bulk. (a) Comparison between the stress-transformation
temperature relation for a solution treated, free-standing Ni-rich
thin film and a Ti49.5Ni50.5 bulk sample37; (b)
stresstransformation temperature relation of a free-standing
Ti50.0Ni48.2Cu1.8 SMA thin film38.
-
9
energy and interfacial energy of the martensite twins, that
needs to be overcome during the forward
and reverse transformation, thus giving rise to transformation
hysteresis.29,41 Next to explaining the
fundamental cause of the transformation hysteresis the geometric
non-linear theory of martensite
(GNLTM) predicts that the hysteresis can be drastically
minimized by improving the crystallo-
graphic compatibility of the martensite and austenite
phase.48,49
The theory specifies several conditions in order for a SMA to
show near zero hysteresis (a
rigorous mathematical derivation can be found in the papers of
Ball and James44,45,49). The first
condition, detU = 1, where U is the transformation stretch
matrix, represents the condition of no
volume change. The second condition: 2 = 1, where 1 2 3 are the
ordered eigenvalues (lat-
tice distortions) of the transformation stretch matrix U,
represents the presence of an invariant plane between austenite and
martensite, i.e. a perfectly coherent interface between both
phases. At
such an interface, the energy contributions to the bulk energy
through the usual elastic transition
layer or interfacial energy of fine arrays of twin bands is
eliminated, thus leading to a decrease of
T. For the cubic (B2) to orthorhombic (B19) martensite
transformation, the following six trans-
formation stretch matrices map the austenite lattice vectors to
the martensite variant lattice vec-
tors45:
0 00 0
, 0 00 0
, 0 0 0
0 ,
0 0 0
0 ,
0 00
0 ,
0 00
0 , (2)
where / , /2 , /2 and a0 is the lattice parameter of the cubic
austenite and a, b, c are the lattice parameters of orthorhombic
martensite. U can be taken to be any one of these six matrices,
since they all have the same eigenvalues. By measuring the lattice
parameters of
the austenite and martensite phases the eigenvalues can be
calculated. A clear correlation between
Fig. 2.7 Interface between austenite and martensite. (a) optical
micrograph of an austenite/martensite interface in Cu69Al27.5Ni3.5
(from James and Zhang49; (b) schematic of the auste-nite/martensite
interface. The different martensite variants are labeled A and
B.
-
10
the middle eigenvalue 2 and was found for binary Ti-Ni50 and
several ternary alloy systems,
i.e. Ti-Ni-X, X = Pt43, Pd41,43, Hf43, Au43, Cu41,43 in both
thin film and bulk alloys as summarized
by Zhang et al.29, whereas the correlation between detU and was
found to be weak.41 Addition-
ally, alloy compositions satisfying 2 = 1 were predicted in the
quaternary alloy system
Ti-Ni-Cu-Pd and postulated to show close to zero .49
2.1.3 Effect of third elements on transformation
temperatures
For the application of SMAs the transformation temperatures are
next to the thermal hysteresis of
significant importance, since they determine the temperature
range of application. The binary Ti-Ni
system shows a distinct composition dependence of the
transformation temperatures for Ti-rich and
Ni-rich compositions. For the former, the transformation
temperatures are constant, while for the
latter a strong decrease is observed with increasing Ni content,
as shown in Fig. 2.8. Much effort
has been made to modify Ti-Ni shape memory alloys by adding
various alloying elements to the
binary system, in order to reduce the strong composition
dependence of the transformation temper-
atures and/or to increase them.
Fig. 2.9 summarizes the effect of third elements on
transformation temperatures reported in
literature.26 Most alloying elements are found to lower the
transformation temperatures (Fig. 2.9b).
Fig. 2.8 Influence of the Ni-concentration on the MS and T0
temperatures for binary Ti-Ni. In region I, MS and T0 (calculated
equilibrium temperature; T0 = (Ms + Af) / 2) are constant. The
concentration range of region I corresponds to the two-phase
region, where Ti2Ni and TiNi are in equilibrium. Region II includes
all chemical compositions which (after quenching) yield single
phase alloys. In region II, T0 / MS decrease from 365/340 K at 50.0
at.% Ni to 227/211 K at 51.0 at.% Ni. Initially, this decrease is
linear, with a slope of -83 C (at.% Ni)-1, followed by a stronger,
non-linear decrease at higher Ni concentrations. From Frenzel et
al..50
-
11
However, the addition of Pd, Pt, Au, Zr or Hf increases them
(Fig. 2.9a) and thus Ti-Ni-X, X = Pd,
Pt, Au, Zr and Hf are considered as candidates for
high-temperature SMAs.51
For future alloy design and the understanding of
composition-structure-property relation-
ships the knowledge of the site preference of the alloying
additions is of fundamental importance,
since the phase formation (e.g. precipitate phases) as well as
the phase transformation characteris-
tics strongly depend on the substitution behavior. Experimental
effort to determine the site occu-
pancy of alloying elements in Ti-Ni has been made by Nakata et
al.52. They employed atom loca-
tion by channeling enhanced microanalysis (ALCHEMI) method to
determine the site preference
of Cr, Mn, Fe, Co, Cu, Pd and found that Fe, Co and Pd
preferentially substitute for Ni, while Mn,
Cr, and Cu seem to substitute for both Ti and Ni with similar
preference. Nakata et al. summarized
Fig. 2.9 The effect of alloying ele-ments on the martensitic
transforma-tion temperatures of Ti-Ni. (a) Pd, Pt and Hf in a wide
alloying range; the circle indicates composition region covered in
b; (b) Fe, Co, V, Mn, Au, Zr, Al, Pd, Pt, Hf and Cr in a narrow
alloying range (low alloying level). From Otsuka and Ren.26
References for the individual sets of data can be found
therein.
-
12
the following: (I) Fe, Co and Pd have strong preference for
entering Ni-sites; (II) Sc has strong
preference for entering Ti-sites; (III) V, Cr, Mn, Cu and Au
seem to have less preference for a par-
ticular site; their occupancy fractions are strongly affected by
the way the alloying element is add-
ed. For example, if adding Cu in a way to give preference for
the Ni-site, i.e. Ti50Ni50-xXx, it is con-
cluded that Cu occupies only Ni-sites; but if adding Cu giving
preference for Ti-sites, i.e.
Ti50-xNi50Xx, there is only about 34% Cu entering into Ni-sites.
This is quite different from the
well-accepted postulation that Cu only goes to Ni-sites.
Similar results are calculated by Sheng et al.53 and more
recently for a wide range of ternary
additions by Bozzolo et al.5456 that strongly motivate the
search for new and/or optimized shape
memory alloys in extended composition regions in ternary and
quaternary alloy systems by combi-
natorial methods. Especially in the quaternary composition
space, where alloy development using
conventional bulk preparation methods is costly and
time-consuming and hence only a very limited
number of investigations are reported5763, the combinatorial
thin film-based methods will be ad-
vantageous.
2.2 TiNi-based SMA thin films
Ti-Ni based thin films are the most frequently used thin film
SMA materials, since they exhibit the
highest work output per unit volume, which if employed as thin
film microactuator exceeds that of
other microactuation materials2, as mentioned above.
Additionally, the positive scaling effect of
surface to volume enables fast cooling, thus significantly
higher working frequencies can be rea-
lized, as compared to bulk SMA actuators. Several thin film
deposition methods were used for the
fabrication of SMA thin films, including laser ablation64, ion
beam deposition65, evaporation (mo-
lecular beam epitaxy)66 and sputter deposition67. The latter is
most frequently applied and will be
discussed in more detail in the following section.
2.2.1 SMA thin film deposition
Sputter deposition of thin film SMAs was pioneered by Kim et al.
in 198667 and in the 1990s sev-
eral groups demonstrated the feasibility of producing thin films
exhibiting reversible phase trans-
formations using alloy targets.6872 In order to obtain
transforming Ti-Ni films a precise control of
the composition is necessary. Fig. 2.10 shows a schematic of a
magnetron sputtering setup and in-
dicates the processes involved. In order to avoid the
characteristic loss of Ti, as the sputtering yield
of Ni is higher than for Ti, due to the different sputtering
rates and angular distributions, the design
and composition of the sputtering target is essential.73
-
13
In previous work of Gyobu et al.74 and Quandt et al.72, the Ti
deficiency was compensated by plac-
ing additional Ti pieces on top of the alloyed target or by
using Ti-rich (54 at.% Ti) targets, respec-
tively. However, since in the case of magnetron sputtering from
alloy targets the deposition rate is
additionally influenced by the depth of the sputter trench and
thus changes with sputter duration2,
other sputter-deposition methods, namely co-deposition75 and
multilayer deposition techniques7680
using elemental targets were developed, in order to control the
thin film composition more precise-
ly. Moreover, the multilayer approach widely eliminates the
influence of sputter power, sputtering
pressure and target-to-substrate distance usually found to
strongly influence the resulting thin film
microstructure. Common to the described thin film deposition
methods is the necessity of a subse-
quent annealing step. Since the Ti-Ni based alloys show a strong
tendency to become amorphous
by sputter deposition at RT26, thin films sputtered from alloy
targets or produced by co-deposition
need to be crystallized at elevated temperatures, as do
multilayer thin films. Amorphization of the
alloys by sputtering is thereby an advantage for the
applications of thin films, since the process of
amorphization and subsequent crystallization leads to small
grain sizes, which are beneficial for the
mechanical properties.81 However, as-sputtered amorphous Ti-Ni
thin films might contain excess
Ti or Ni atoms for Ti-rich of Ni-rich compositions,
respectively, non-equilibrium phases may form
during the crystallization process, as will be discussed
below.
Using heated substrates during the deposition of Ti-Ni based
thin films the subsequent an-
nealing step can be avoided, since crystalline films can be
obtained when deposition temperatures
above 250 C are used.17,65,8285
Fig. 2.10 Schematic of magnetron sput-tering. A magnetic field
confines the orbits of the electrons to maintain an intense plasma
and to increase the colli-sion rate with the Ar gas. The number of
Ar ions created and impinging on the target, knocking out
individual atoms, which are then deposited on the substrate, is
enhanced. For sputter deposition a ultra-high vacuum system and
high-grade Ar gas are beneficial, in order to prevent contamination
of the thin films.
-
14
Fig. 2.11 shows a comparison between the crystallization
behavior of an amorphous and a
multilayer Ti-Ni thin film (Fig. 2.11a) and depicts the alloying
process of the latter
(Fig. 2.11b,e,g).80 The DSC curve during heating of the
multilayer Ti-Ni thin film (Fig. 2.11a)
shows three exothermic peaks79, which correspond to (I) the
formation of an amorphous phase at
each interface (interdiffusion of Ni into the Ti layer86, i.e.
growth of the reaction layer (Fig. 2.11e),
(II) the formation of B2 phase (crystallization) and (III) the
formation of precipitates by inter-
diffusion of Ti atoms into the Ti-Ni B2 from the residual Ti
layers (Fig. 2.11g). The Ti2Ni precipi-
tates remain near the interface of the Ti-Ni B2 phase and no
Ni4Ti3 precipitates were observed. In
contrast, for the as-sputtered amorphous Ti-Ni thin film only a
single exothermic, crystallization
peak80 is observed (Fig. 2.11a) and the microstructure of
Ti48.7Ni51.3 is known to consist of compa-
ratively larger grains with Ni4Ti3 precipitates.87
TEM images show the multilayer structure in the as-sputtered
thin film (Fig. 2.11b), and the
Ti- (bright), Ni- (dark) and a thin reaction layer can be
clearly seen. The corresponding selected
area diffraction patterns show the diffraction spots of Ti and
Ni polycrystals (Fig. 2.11c,f,h) and
diffuse diffraction rings of the amorphous reaction layer in
between (Fig. 2.11d).
Fig. 2.11 Alloying process of sputter-deposited Ti/Ni multilayer
thin films. (a) DSC curves of the as-sputtered multilayer thin film
and as-sputtered amorphous thin film79; (b), (e), (g) bright field
TEM images and (c), (f), (h) the corresponding selected area
diffraction pattern for as-sputtered Ti49.0Ni51.0, heated up to 640
K and 750 K multilayer thin films, respectively80; (d) nano-beam
diffraction pattern of the interface layer between Ti and Ni
layers.80
-
15
Additional insight into the alloy formation, i.e. amorphization
reaction and recrystallization
behavior is provided by studies on Ti/Ni multilayers used for
neutron optics components, such as
highly reflecting mirrors, polarizers, monochromators. Therein
the thermal stability88, structural
and magnetic properties89,90 and the kinetics of alloy formation
at the interface91 were investigated.
2.2.2 SMA thin film stresses
Stresses in SMA thin films can be intrinsic and extrinsic. The
intrinsic stresses depend mainly on
the growth conditions (such as temperature, pressure, growth
rates94,95) and thus play only a minor
role for SMA thin films in general and more specifically for the
multilayer-deposition approach,
since a subsequent high-temperature annealing step is required
for the crystallization of the films
leading to a relaxation of the intrinsic stresses.
Fig. 2.12 shows a schematic stress-temperature () curve of a
constrained SMA thin film92
and an experimental () curve of a Ti-Ni-Cu thin film deposited
from an alloy target onto an oxi-
dized Si substrate93, as determined using the cantilever
deflection method (see 3.1.5). Upon anneal-
ing the compressive stresses of the as-deposited film initially
increase and subsequently relax com-
pletely upon crystallization at high temperatures (Fig.
2.12b).82,93 Upon cooling, the extrinsic
stress, i.e. tensile stress rises linearly due to the different
thermal expansion coefficients of SMA
film and substrate until, through the formation of martensite
starting at Ms, the stress can be partial-
ly accommodated.
The slope of the linear part of the stress-temperature curve
above Ms can be calculated us-
ing an estimate of the coefficient of thermal expansion (CTE)
for TiNi (TiNi = 11.0 x 10-6/K) 96 and
of the Si substrate (Si = 2.5 x 10-6/K)97 and the following
equation98 (the SiO2 diffusion barrier is
Fig. 2.12 SMA thin film stresses. (a) Schematic
stress-temperature curve for transforming constrained SMA film. The
evolution of the film microstructure and misfit is shown (austenite
light grey; martensite black; substrate dark grey; reference
stress-free state at TR) from Roytburd et al.92, (b) experimental
stress-temperature curve of a 1.7 m Ti51.7Ni39.7Cu8.6 film on 180 m
thick, thermally-oxidized Si (100) substrate after deposition.
Heating and cooling rate was 5 K/min from Winzek and Quandt93.
-
16
neglected in this calculation):
/ / 1 , (3) where ETiNi is the Youngs modulus of TiNi austenite
(~80 GPa)40; TiNi is the Poisson ratio of
Ti-Ni film (0.33)40. A slope of -1.03 MPa/K is calculated, which
corresponds well to the experi-
mental data (see 4.1.6). According to (3), the absolute stress
of a SMA thin film at the martensite
start temperature (Ms, Os, Rs) depends on the temperature
interval between Ms (Os, Rs) and TR (ref-
erence stress free state, see Fig. 2.12a), on the thermal
expansion coefficient as well as on the
Youngs modulus of the thin film. These values are likely to vary
with composition in binary, ter-
nary or quaternary SMA thin films.
The effect of thermal stresses on the phase transformation
properties of substrate-attached
(constrained) thin films was discussed in
literature40,82,83,92,93,99,100, but up to now no consistent
pic-
ture has emerged. Roytburd et al. calculated that the
temperature interval of the two-phase equili-
brium is widened for constrained, single-crystal SMA thin films,
which is supported by the expe-
rimental work of Liu and Huang38, but opposed by results
reported by Winzek and Quandt93. The
latter find that the temperature interval of transformation
remains constant, if constrained and free-
standing thin films are compared. Similarly, while the
transformation temperatures are calculated
to shift to higher temperatures92 both, experimental
verification93 and falsification38are reported in
literature.
With respect to the influence of thin film stresses on the
thermal hysteresis, experimental
work indicates a significant reduction for constrained
films38,93 as compared to free-standing thin
films.38,93 However, while the phase transformation properties
of the constrained films were deter-
mined using the curvature method, the free-standing thin films
were characterized using DSC mea-
surements. Thus, the observed differences may be due to
different measurement methods and/or
different thin film thicknesses used.93 Ideally, the same
measurement technique and the same sam-
ple dimensions (film thickness) should be used for
comparison.
An additional controversy reported in literature is concerning a
potential stress gradient in
the SMA thin films, which could influence the shape memory
behavior. While Winzek101 calcu-
lates a homogenous stress distribution in substrate-attached
thin films with a film to substrate
thickness ratio of < 0.1, both Grummon et al.40 and Liu et
al.38 assume a stress gradient through the
film thickness based on the experimental observation that the
intrinsic stresses in sputter-deposited
thin films decrease with increasing thickness100 and the
assumption that a variation of shear stress
from a zero value at the free outer surface to a relative high
value at the film/substrate interface
exists. However, the models based on such assumptions remain
speculative, since no direct expe-
rimental observation of the stress gradients or the
microstructural development during the phase
transformation in constrained films exists.
-
17
2.2.3 Ageing effects of Ni-rich and Ti-rich Ti-Ni thin films
Ageing treatments at elevated temperatures are an effective way
of strengthening Ti-Ni thin films
by precipitate hardening, i.e. by Ni4Ti3 precipitates for
Ni-rich compositions (Fig. 2.13a)36 or by
Ti2Ni precipitates for Ti-rich compositions102. Whereas the
precipitation characteristics of Ni4Ti3
were found to be almost consistent with those reported in bulk
samples36, non-equilibrium phases
were observed in Ti-rich thin films that are not observed in
bulk alloys (Fig. 2.13b-d)102104. When
amorphous Ti51.8Ni48.2 thin films are crystallized at 500 C, the
microstructure of the thin films
changes with annealing duration of 5 minutes, 1 hour and 10
hours in the following sequence: (I)
Guinier-Preston (GP) zones, (II) GP zones and Ti2Ni
precipitates, (III) Ti2Ni precipitates, respec-
tively.102 The Ti2Ni precipitates within the TiNi grains have
the same orientation as the TiNi ma-
trix, and the interface between both phases is partially
coherent.
Fig. 2.13b-d show TEM images of a Ti-rich Ti-Ni thin film
annealed at 500 C / 1 h. In
Fig. 2.13b the GP zones have a diameter ~17 nm, form along the
100 planes of the bcc (B2) phase, are completely coherent and have
a disc / plate-like appearance.104,105 Additionally, spheri-
Fig. 2.13 Microstructure of aged Ti-Ni thin films. (a) TEM
images of precipitates in Ti48.7Ni51.3 thin films aged at 500 C / 1
h after solution treatment at 700 C / 1 h36; (b) bright-field image
of a Ti51.8Ni48.2 thin film annealed at 500 C / 1 h, (c)
bright-field image taken in random orientation, and (d )
high-resolution TEM image in [100] orienta-tion.102 Precipitates
and GP zones are indicated by solid arrows.
-
18
cal Ti2Ni precipitates can be observed and are clearly
distinguishable, if the TEM image are taken
in random orientation, as shown in Fig. 2.13c. A high-resolution
TEM image taken in [100]B2 di-
rection is presented in Fig. 2.13d and reveals the morphology of
the disc-like GP zones and round
Ti2Ni precipitates in more detail. Owing to the existence of the
GP zones, the critical stress for slip
in the parent phase is increased and, as a result, excellent
shape memory properties are obtained for
Ti-rich Ti-Ni shape memory thin films.104 For annealing
temperatures of 600 C and 700 C the GP
zones are lost and only Ti2Ni precipitates are observed, which
tend to distribute at higher annealing
temperatures along the grain boundaries (the equivalent to the
precipitation behavior reported in
bulk samples).106
Next to the observation of GP zones in Ti-rich Ti-Ni thin films,
GP zones were also recent-
ly found in Ti-rich Ti-Ni-Cu thin films fabricated by
co-deposition107109, whereas no GP zones
were found in annealed Ti-Ni multilayer thin films, where the
precipitation behavior of Ti2Ni re-
sembles more closely the bulk-like behavior (Fig. 2.11a-d).
2.2.4 Thickness effects of Ti-Ni thin films
Another fundamental aspect for the application and investigation
of SMA thin films is the critical
thickness needed to yield consistent properties, i.e. to be
comparable to thicker films or even bulk
material. Several groups investigated the lower thickness
boundary in sputtered Ti-Ni thin films for
shape memory application, i.e. the size effect on the
martensitic transformation.110114 Fig. 2.14
shows TEM images of TiNi thin films with varying thicknesses and
the grain size is indicated by
dashed lines. The surface oxide layer, as well as the affected
zone (enriched in Ni) is highlighted in
Fig. 2.14e and the deduced strengthening mechanisms as proposed
by Ishida and Sato are illu-
strated schematically in Fig. 2.14f.
Two kinds of resistances against deformations are considered:
the constraints from neigh-
boring grains and from surface oxide layers. The former effect
increases with increasing thickness,
whereas the latter increases with decreasing thickness.110 In
addition, surface oxidation and the
formation of interdiffusion layers influence the transformation
temperatures and the composition of
the transforming phase owing to the consumption of Ti.110 The
constraints imposed by neighboring
grains was found to saturated, if the film thickness is greater
than the average grain size.
From the above considerations two important conclusions can be
drawn: (I) if the average
grain size and (II) the surface oxide layer thickness are
decreased, the critical thickness boundary
can be significantly lowered. Thus, the multilayer deposition
approach, yielding significantly
smaller grain sizes as compared to co-deposition or sputter
deposition from alloy targets76,77 is
most suitable for investigating thin film shape memory
properties, even for films with thicknesses
below 1 m, if the thin film samples are annealed under UHV
conditions.
-
19
The final film thickness limit or grain size at which a
reversible martensitic transformation is com-
pletely suppressed is found at < 100 nm111 for
substrate-attached thin films and ~50 nm for nano-
crystallites in an amorphous matrix113.
2.2.5 Functional fatigue of SMA thin films.
Stability and fatigue of Ti-Ni thin films have always been
concerns in the development of applica-
tions.2,87,93 The functional fatigue of Ti-Ni film is referred
to the non-durability and deterioration of
the shape memory effect during cycling and results in changes of
physical, mechanical and shape
memory properties115 due to irreversible processes, e.g.
generation of dislocations116120 which
form during the martensitic phase transformation. Thus, the
fatigue of thin films is influenced by
internal (alloy composition, lattice structure, precipitation,
defects and film/substrate interface) and
additionally by external parameters (annealing treatment,
applied stress, stress and strain rates,
heating/cooling rates). Consequently, from the scarce results
published in literature no clear picture
Fig. 2.14 Thickness effect on shape memory behavior of Ti-Ni
thin films. Cross-sectional TEM images of Ti50.0Ni50.0 thin films
with thicknesses of (a) 7, (b) 5, (c) 2, (d) 1 m, respectively; (e)
surface oxide layer and affected zone of Ti50.0Ni50.0 thin film
after heat-treatment at 773 K for 300 s in a vacuum furnace (3 10-5
Pa) with infrared lamps (cross-sectional TEM image); (f)
strengthening mechanisms of thin films.110
-
20
emerges regarding the different contributions of the
aforementioned parameters on the fatigue of
SMA thin films.
However, an initial decrease in the recovery stress () is
consistently reported, as shown in
Fig. 2.15a-b, which is stabilizing upon further cycling.2,87,93
The long term functional stability was
allegedly demonstrated by TiNi Alloy Companys laboratory, where
a microvalve was successfully
tested for more than 50 million cycles (1 % deformation, 1
Hz).121
2.2.6 Current and future developments
The recent developments with respect to the fundamental
understanding of SMA thin films and
their device applications were recently reviewed by Miyazaki et
al.122. Future advances in the field
of SMA thin films will depend on scientific and technological
progress with respect to improved
materials properties and improved manufacturing processes. From
a thin film materials perspective
for actuator applications, higher recovery stresses, higher
transformation temperatures, smaller
thermal hysteresis and a smaller temperature interval of
transformation are desired in order to im-
prove the work output and frequency of operation. A clear
opportunity to improve the shape mem-
ory effect lies thereby in targeting specific thin film textures
using novel processing techniques in
order to optimize the attainable recoverable stress.
Furthermore, ternary and quaternary alloying
additions are suitable for improving the phase transformation
properties (thermal hysteresis, trans-
formation temperatures). From a manufacturing process point of
view, the most challenging prob-
lems are related to keeping the exact stoichiometry of the thin
films and to decrease the necessary
crystallization temperatures in order to decrease the thermal
stresses. New thin film deposition
schemes based on multilayer deposition from elemental targets
are promising in this respect.
Fig. 2.15 Fatigue of SMA thin films. (a) Ms and recovery stress
(rec) as a function of thermal cycles for a Ti-Ni-Cu film. The
strain during the test is ~0.2 %2; (b) recovery stress for
Ti52.0Ni19.8Pd28.2 versus number of cycles N. The values originate
from measurements with 3 K/min, the annealing cycles between were
executed by pulses of electric current.93
-
21
2.3 Combinatorial materials science
Since 1995, when Xiang et al. coined the term combinatorial
approach to materials discovery123,
numerous studies have demonstrated the applicability of the
combinatorial methodology: synthesis
of materials libraries and their high-throughput
characterization for the determination of phase dia-
grams124,125, for the discovery of new or optimized
functional41,126,127, optical128 or catalyst mate-
rials129133 and polymers134. However, the principal idea of a
multiple-sample concept for the
investigation of inorganic materials was already described by
Kennedy et al.135 in 1965 or Hanak136
in 1970 and one of the earliest combinatorial approaches in
material science can be traced back to
Thomas Edison and the year of 1878. At that time, Edison applied
parallel and combinatorial me-
thods for the investigation of suitable filament materials for
the incandescent lamp, as outlined in a
review by Schubert et al..134
2.3.1 Materials libraries
In general, two types of materials libraries can be used in the
combinatorial approach to materials
discovery: discrete libraries sets of samples with individual
compositions (Fig. 2.16a,b) or con-
tinuous libraries a single sample with a continuous
compositional variation (Fig. 2.16c). While
the discrete approach allows rapid screening of large sets of
different materials128, the latter is es-
pecially suited for the determination of
composition-structure-property relationships.
The synthesis of continuous materials libraries of bulk
materials, i.e. diffusion multiples,
and thin films so-called composition spreads were pioneered by
Zhanpeng138 and Kennedy et
al.135, Miller et al.139 and Goldfarb et al.140,
respectively.
Fig. 2.16 Material libraries, (a) A discrete 128-member BaCO3,
Bi2O3, CaO, CuO, PbO, SrCO3, and Y2O3 library prior to sintering.
Each site is 1 mm by 2 mm; the color of each is the natural color
of reflected light from a white light source. BiSrCaCuO and YBaCuO
superconducting films were identified.123 (b) Photograph of the
as-deposited discrete quaternary library under ambient light used
in the search for a blue photoluminescent composite material. The
diversity of colors in the different sites stems from variations in
film thicknesses and the optical indices of refraction.128 (c)
Pho-tograph of an annealed quaternary continuous composition spread
(CCS) for Si, Sn, Co and C.137
-
22
For the fabrication of thin film composition spreads,
co-evaporation135, co-sputtering139 and more
recently, thin-film multilayer approaches140,141 were
introduced. Co-deposition, e.g. the simultane-
ous sputtering of different materials, leads to an atomic mixing
of the materials during deposition,
while the compositional variation (spread) is determined by the
geometric arrangement of the de-
position sources and the deposition rates of the individual
materials. Thus, the attainable composi-
tional variation is limited, as well as the variation of the
composition gradients. In contrast, a
wedge-type multilayer approach, e.g. the sequential deposition
of individual nanoscale, elemental
wedge-type layers requires a subsequent annealing step for the
alloy formation; however, the com-
positional variation and the composition gradients can be
adjusted in a broad range. Additionally,
the wedge-type multilayer approach can be extended in order to
cover large continuous regions of
quaternary or pseudo-quaternary systems, as described in more
detail in section 3.1.1. Due to the
sequential nature of the multilayer approach, the deposition is
slower compared to the
co-deposition approach, thus the achievable film thicknesses are
limited from a practical point of
view.
2.3.2 High-throughput characterization
For the high-throughput characterization of the materials
libraries, various parallel and serial high-
throughput inspection methods are applied. In recent years, the
characterization techniques for the
most fundamental properties, i.e. composition and structure
energy or wavelength dispersive
X-ray analysis (EDX, WDX) and X-ray diffraction were improved
significantly with respect to
their spatial resolution and measurement times, thus enabling
their application to high-throughput
experimentation.
The development of more specialized characterization techniques
for functional properties
of materials range from parallel optical methods for the
characterization of hydrogen storage mate-
rials142, or photoluminescent materials128, to serial methods
for the characterization of magnetic126,
thermoelectric143 or dielectric144 materials. For the latter,
scanning techniques, i.e. scanning squid
microscopy126, a potential Seebeck microprobe143 or a
scanning-tip microwave near-field micro-
scope144 were used, respectively. Furthermore, scanning
characterization techniques have been
developed for the electrochemical screening of materials
libraries, e.g. the scanning droplet cell145
or an alternating current scanning electrochemical microscope
(AC-SECM).146
In general, the possibility to address specific locations on a
materials library (x-y-z posi-
tioning system) combined with an automation of the measurement
is fundamental for high-
throughput experimentation. However, equally important is the
ability to design the materials li-
brary in order to match the composition gradients with the
spatial resolution of the measurements.
Since the latter can vary significantly for different
measurement methods, a technique capable of
-
23
controlling the composition gradients of the materials libraries
is needed, as pointed out above and
outlined in more detail in section 3.1.1.
Additionally, to the scanning techniques,
micro-electromechanical systems (MEMS) offer
powerful tools for the fabrication and processing of materials
libraries as well as for accelerated
materials characterization on planar substrates such as Si
wafers. MEMS can be used for parallel
materials processing, either as passive devices such as shadow
mask structures, or as active devices
such as micro-hotplates.147,148 Microstructured wafers, which
incorporate sensor or actuator struc-
tures such as electrodes149,150, cantilever arrays126,151,152
microtensile test devices153,154 or bulge test
structures155,156 can be used to probe materials properties in
an efficient way. Furthermore, for the
characterization of thin film mechanical properties
nanoindentation can be applied.157 However, the
limitations with respect to substrate effect and machine
compliance when testing thin substrate-
attached materials libraries on the wafer scale need to be taken
into account.158
Within this work, commercially available high-throughput
characterization tools such as
automated EDX and XRD are used next to custom-made
high-throughput test stands for the cha-
racterization of the temperature-dependent resistance and the
stress change of thin-film materials
libraries. Nanoindentation at RT and elevated temperatures will
also be used for the characteriza-
tion of the mechanical properties, as described in more detail
in sections 3.1.2 to 3.1.6.
2.3.3 Data analysis and visualization
Data analysis and visualization capabilities are next to
appropriate high-throughput characterization
techniques a fundamental requirement for the implementation of
the combinatorial materials
science approach. Already in 1970, Hanak used computer data
processing in tabular, graphical and
functional forms as part of his multiple-sample concept.136
However, he and his followers at that
time were lacking tools for data processing and analysis
appropriate for hundreds of data sets. Only
later in the 1990s, computational power and commercially
available software for the primary data
processing tasks: collecting characterization data from analysis
instruments, automated visualiza-
tion of the raw data sets, and the ability to organize and share
measurements results159, was widely
available and enabled the since-then ongoing success of the
combinatorial approach.
Today, along with computer-based data acquisition and
visualization using software, such
as LabVIEW, ORIGIN or MATLAB, respectively, also data mining
techniques have been devel-
oped in order to facilitate the data analysis and will be used
in this work. Two primary functions
are served by data mining techniques: pattern recognition and
classification, both of which form
the foundations for understanding materials
composition-structure-property relationships.160,161 As
an example, cluster analysis and principal component analysis
(PCA) were recently implemented
within the XRDSuite software package162 in order to facilitate
the arduous analysis of analyzing
-
24
large sets of XRD patterns, by sorting patterns by their
similarity into discrete groups and subse-
quently deducing the representative basis X-ray patterns.163
2.3.4 Current and future developments
The recent developments and successes in the field of
combinatorial solid-state chemistry of inor-
ganic material have been recently comprehensively reviewed by
Zhao164 and Koinuma and Takeu-
chi165. Additionally, a number of books focusing on the
synthesis of materials libraries166, their
high-throughput characterization167 or both168, and on data
analysis and visualization169 were pub-
lished over the last years.
Future advances in the field of combinatorial materials science
will depend on the devel-
opment of new characterization techniques for the
high-throughput screening of materials libra-
ries.170 Thus, next to the development of sophisticated methods
for the fabrication of materials li-
braries, the limitations of current high-throughput
characterization tools need to be addressed.
Additional potential for the combinatorial approach lies in
developing the ability to transfer
thin film composition spread results to bulk materials. While
thin film composition spread tech-
niques were found to be especially useful for mapping
composition-structure-property relationships
for thin films, equilibrium phase formation and structural
properties (hardness, elasticity, creep)
were concluded to be more suitably investigated by bulk
diffusion couples.171 Thus, new fields of
applications for the combinatorial thin film approach could
emerge, if desired bulk properties could
be screened using thin film composition spreads and the results
prove to be transferrable to the
bulk material.
Within this work, the latter concept will be pursued, next to
the development of new sputter-
ing schemes for the fabrication of ternary and quaternary thin
film composition spreads and the
development of new high-throughput characterization tools for
the characterization of SMA thin
films, as outlined in the following section in more detail.
-
25
3 Methods
3.1 Combinatorial materials science methodology
The combinatorial materials science / high-throughput
experimentation approach for the
development and discovery of SMA thin films, as well as other
materials, is shown schematically
in Fig. 3.1. It comprises the deposition of thin-film materials
libraries using special magnetron
sputter processes, the automated high-throughput
characterization (screening) of these libraries,
and appropriate visualization and analysis of the data.
3.1.1 Sputter deposition of SMA thin-film materials
libraries
The materials libraries, i.e. continuous composition spreads
used in this work were fabricated by
magnetron sputtering using a dedicated ultra-high vacuum
combinatorial sputtering system (CMS
600/400LIN, DCA, Finland) and 4-inch oxidized Si(100) wafers
with 1.5 m thermal SiO2 as sub-
strates. The sputter system consists of a load-lock,
distribution chamber, mask-storage chamber and
two sputtering chambers. The sputtering chamber used in these
experiments is equipped with three
DC and three RF magnetron cathodes mounted on a moveable arm
above a rotatable substrate
holder with integrated heater, as shown in Fig. 3.2.
Additionally, four intermediate shutters placed
90 apart are located above the substrate, which can be moved
independently and used to partially
shield the substrate or to create wedge-type films. The typical
background pressure before sputter-
Fig. 3.1 Combinatorial materials science / high-throughput
experimentation approach for the development of SMA thin films.
-
26
ing was lower than 2.0 x10-8 Torr. Elemental targets (purity:
99.99% or better) were used and pre-
sputtered in order to minimize contamination. The targets have a
diameter of 4-inch to enable a
sufficient thickness uniformity of the deposited films on the
4-inch substrates. Deposition was done
with Ar (6N) at a pressure of 5 mTorr, a target-to-substrate
distance of 87.5 mm and without inten-
tional heating. The wedge-shaped thickness profiles of the
deposited material were created by one
of the intermediate shutters, which was set to shield the
substrate and then slowly retracted during
the deposition (speed: 1 - 5 mm/s). Alternating wedge-type thin
films of Ti, Ni, and X (X = Cu, Pd,
Ag) were consecutively deposited by switching the position of
the targets above the substrate ac-
cordingly. In all cases the first layer was Ti in order to allow
for good adhesion to the substrate and
the surface layer was the ternary element in order to prevent Ti
surface oxidation during annealing.
Binary composition spreads were fabricated using opposing
wedge-type films, realized by
rotating the substrate by 180 or the use of two opposing
shutters. Due to the degrees of freedom
associated with the wedge-type multilayer approach, both
compositional variation and gradients
can be adjusted in a broad range by selection of appropriate
sputter powers and shutter speeds, re-
spectively. The depositions were controlled by a recipe with an
internal loop repeating the deposi-
tion of each wedge-type film numerous times, in order to build
up the chosen total film thickness.
Fabrication of pseudo-binary composition spreads, compromised of
three elements, with
one kept constant, were deposited using a sequence of opposing
wedge-type films of two elements
and intermediate layers of homogeneous thickness of the third
element. The thickness variation in
Fig. 3.2 Ultra-high vacuum combinatorial magnetron sputter
system for the fabrication of binary, ternary and qua-ternary
thin-film materials libraries using wedge-type multilayer thin
films.
-
27
Fig. 3.3 Schematic of the wedge-type multilayer approach for the
fabrication of ternary materials libraries. (a) individual layers
are rotated relative to each other by 120 and layer thicknesses of
up to 10 30 nm were used (adapted from Goldfarb et al.140). (b) In
order to cover the complete ternary system the wedge-type films are
limited in length (70 mm). Black circle in b represents the contour
of the 4-inch Si(100) substrate. (An animated movie of the sputter
deposition scheme for the fabrication of ternary composition
spreads is available at www.rub.de/wdm.)
the homogeneously deposited film was found to be less than 2 %
within a radius of 2 cm around
the center of the substrate. Apertures were used to increase the
homogeneity, but at a cost of signif-
icantly lower deposition rates.
Ternary composition spreads were realized by rotating the
substrate by 120 each time be-
fore the deposition of the next material (Fig. 3.3a). For
coverage of the complete ternary system,
the movement of the shutter was stopped at an intermediate
position, and thus the length of the
wedge-type layer was intentionally limited. The shutter speed
for all depositions was set to 1 mm/s.
By using a combination of such limited wedges, rotated by 120
relative to each other, the com-
plete ternary and binary systems can be covered in one
experiment on a single substrate, as illu-
strated in Fig. 3.3b. For zoomed-in ternary composition spreads
focused around a certain com-
position region, e.g. Ti50Ni50 wedge-type films extending over
the full length were used. Thereby
the compositional variation and gradients can be adjusted by
variation of sputter power and shutter
speed, respectively. Selection of appropriate deposition
parameters was facilitated by calculating
the resulting compositions at selected positions prior to the
deposition. The calculations were based
on the measured sputter rates of the individual elements and the
deposition time at a given location
on the 4-inch substrate. The total deposition time is the sum of
the exposure time related to the
movement of the shutter calculated using the location and the
shutter speed and/or from a (sub-
sequent) static deposition.
-
28
Quaternary composition spreads were realized using the
multilayer approach of alternating
wedge-type thin films. Sets of opposing wedges were used, with
each set being a combination of a
single wedge and a sequence of three successive, opposing
wedges, as illustrated in Fig. 3.4a. The
second set of opposing wedge-type thin films for the remaining
two materials was rotated by 90
relative to the first set. For the single wedge-shaped thickness
profile ( 30 mm around the sub-
strate center) one of the intermediate shutters was set to
shield the substrate and then slowly re-
tracted (3 mm/s) during the deposition. The sequence of three
successive wedges was realized by
updating the position of the paired shutters, so that one of
them shielded the already covered areas,
whereas the other creates 20 mm wedges at 1 mm/s shutter
speed.
Due to the approach taken, two of the three degrees of freedom
for adjusting and control-
ling the composition range covered by the composition spread are
fixed, namely shutter speed and
deposition time. Therefore, only the sputter power remains for
controlling the composition range
covered within a deposition. The coverage realized using equal
sputter rates of all four elements is
shown in Fig. 3.4b-c (calculated).137 The nine regions created
on the sample (Fig. 3.4a) translate
each to a plane in quaternary composition space. The diagonal
regions lie in the same plane: region
3 is contained in 5, which is contained in region 7. The seven
distinct planes are almost parallel and
are separated by 7 at.% to 10 at.% in component A.
In Fig. 3.5 the measured compositional variations in a
Ti-Ni-Fe-Au quaternary system are
shown, where all sputtering rates were chosen to be equal (Fig.
3.5a,b) or to have a specific ratio of
sputter powers, i.e. 2:2:1:1 for Ti, Ni, Fe, Au (Fig. 3.5c),
respectively. Using equal sputter rates for
all elements, the coverage in the quaternary system can be
maximized, however it will be concen-
trated in the center of the tetrahedron. A selection of a
specific ratio of sputter powers can be used
to shift and confine the covered composition region in certain
parts of the quaternary composition
Fig. 3.4 Scheme of the wedge-type multilayer approach for the
fabrication of quaternary materials libraries. (a) Sets of opposing
wedges were used, whereas each set was a combination of a single
wedge and a sequence of three successive and opposing wedges
(adapted from Chevrier and Dahn137). (b) Composition space covered
within a qua-ternary alloy system (calculated). The two
tetrahedrons are different views of the same data set.137
-
29
space, here onto the Ti-Ni axis. In order to focus on a single
compositional plane out of the quater-
nary composition space, a combination of the wedge-type
multilayers approach for the fabrication
of a ternary composition spread with intermediate homogenous
layers of the fourth element was
used. An example of the achievable compositional variation is
shown for Ti-Ni-Cu-Pd alloys in
section 4.4, Fig. 4.40.
In situ annealing of the multilayer composition spreads at
temperatures of 500 C, 600 C
or 700 C for 1 h (heating rate 50 C/min, temperature stability 2
C) led to alloying of the ele-
mental multilayers via an amorphization reaction and complete
recrystallization, as revealed by
XRD and TEM observations.7680
Fig. 3.6 shows images of an annealed binary, a complete ternary
and a quaternary thin film
composition spread, that nicely compare with the schematic in
Fig. 3.3 and Fig. 3.4. Thus, methods
for the fabrication of a broad range of materials libraries
based on the sputter deposition of wedge-
type thin films were developed and demonstrated.
Fig. 3.6 Continuous thin film composition spreads. (a) binary
composition spread the 301 point measurement grid is indicated (x,
y = 4.5 mm), (b) complete Ti-Ni-Cu ternary composition spread, (c)
Ti-Ni-Fe-Au quaternary com-position spread. White dashed lines in b
and c indicate the ternary composition space (Fig. 3.3) and the 9
distinct re-gions in the quaternary composition spread (Fig. 3.4),
repectively.
Fig. 3.5 Covered composition space within the quaternary alloy
system Ti-Ni-Fe-Au. The two tetrahedra in (a) and (b) are different
views of the same composition data determined by EDX, (sputtering
rates all equal). In (a) the rotation of the tetrahedron is chosen
to highlight the planes corresponding to the different regions of
the composition spread (Fig. 3.4). (c) The relative sputtering rate
ratio was 2:2:1:1 for Ti, Ni, Fe, Au, respectively.
-
30
3.1.2 Compositional analysis energy dispersive X-ray analysis
(EDX)
The compositions of thin-film materials libraries and bulk
samples were characterized by EDX
(Oxford INCA, Si:Li detector, LEO 1430 VP) using bulk standards,
i.e. Ti50Ni40Cu10, Ti50Ni40Pd10,
for calibration of the instrument prior to each measurement.
Thus, an accuracy of 0.5 at.% could
be achieved for the determination of thin film compositions
using measuring times of 90 to 120 s
per measurement point. A typical measurement grid of 301 points
with an x, y spacing of 4.5 mm
is shown in Fig. 3.6a. and at each point the EDX signal was
integrated over an area of 400 m by
600 m. For the EDX characterization of bulk samples, measurement
times of up to 1 h were used,
in order to increase the accuracy to 0.2 at.%.
3.1.3 Structural analysis X-ray diffraction methods (XRD) and
temperature-dependent X-ray diffraction (XRD(T))
Structural properties of thin film and bulk samples were
characterized by XRD at room tempera-
ture (RT) using a Bruker AXS D8 Discover (with GADDS, CuK
radiation, spot size < 1 mm, in-
tegration time 600 s, area detector 2 range from ~26 to ~57),
PANalytical XPert PRO MPD
(Pixel detector, CuK radiation, mono-capillary 0.8 mm,
integration time 600 s, 2-range: ~30
to ~110) and PANalytical XPert PRO MRD (X'Celerator detector,
CuK radiation, spot size
2 mm2, integration time 1800 s, 2-range: ~30 to ~110). The
systems were calibrated using an
Al2O3 standard (Bruker) or Si standard (PANalytical),
respectively. For the temperature-dependent
XRD measurements heating/cooling stages were used: Anton Paar
TTK 450 (temperature range:
-100 C to 450 C, vacuum 0.75 mTorr), Anton Paar DHS 900
(temperature range: RT up to
900 C, Ar atmosphere, 150 Torr overpressure, flow 0.5 l/min) and
a custom-designed Peltier stage
(temperature range: 5 C to 90 C, atmosphere). In order to
monitor the reversible phase transfor-
Fig. 3.7 EDX analysis of Ti-Ni-Cu continuous thin film
composition spreads. (a) distribution of Ni over the wafer, (b) Cu,
(c) Ti. The arrows indicate the gradient direction of the
wedge-type films from the thin to the thick end. A 301 point
measurement grid (x, y = 4.5 mm) as shown in Fig. 3.6 was used for
the automated compositional analysis.
-
31
mation upon heating and cooling by XRD, measurements were
performed with a temperature step
size of 5 K.
Thin film lattice parameters were determined using
synchrotron-based X-ray microdiffrac-
tion at the 2-BM beam line at the Advanced Photon Source at
Argonne National Laboratory. For
this, the wafer was cut into 301 squares (4.5 x 4.5 mm)
consistent with the analysis by EDX. Dif-
fraction measurements were performed at 110 C (i.e. austenitic
state) and at -20 C (i.e. martensit-
ic state). Measurement times of 200 s per spot were used to
obtain sufficient diffracted intensity for
a complete lattice parameter analysis of the thin film samples
using an image-plate detector
(MAR 345). The beam size was focused to 15 x 15 m2 using a set
of 30 Be compound refractive
lenses and the photon energy was set to 15 keV. The system was
calibrated using a CeO2 standard
(National Institute of Standards and Technology). Lattice
parameters were extracted from inte-
grated diffraction patterns using Fit2d software
(http://www.esrf.eu/computing/scientific/FIT2D/).
Visualization and data management of the diffraction data for
ternary alloy systems were
realized using the MATLAB-based XRDsuite software package.162
Implemented methods for the
identification of similar structural phases using cluster
analysis163 and non-negative matrix factori-
zation172 were used to reduce the complexity of the data. For
quaternary alloy systems the Quater-
naryViewer software package137 was used for visualization of
structural data and functional prop-
erties.
For the identification of the detected phases, comparisons were
made to databases of known
phases: inorganic crystal structures database (ICSD), Pauling
Files binaries and Pearson's Crystal
database. Additionally, the CaRIne software package was used in
order to calculate XRD pat-
terns of crystal structures reported in literature, but not
recorded in the databases.
3.1.4 Phase transformation properties of thin films
temperature-dependent resis-tance measurements (R(T))
The R(T) method for the characterization of the phase
transformation of SMAs is based on a
change in the resistivity due to changes in the crystal lattice
and number of internal interfaces (lat-
tice imperfections) and is well established for bulk173175 and
thin film112,176 SMAs. DSC measure-
ments, commonly applied for the characterization of bulk
materials, are likewise suitable for the
characterization of the phase transformation temperatures of
thin films and are generally concluded
to agree well with the R(T) measurements.176 However, the
electrical resistance measurements
were found to be more sensitive with respect to the appearance
of the R-phase and the identifica-
tion of successive transformation steps.175 Additionally, the
R(T) measurements can be conducted
locally, while a standard DSC measurement requires a certain
amount of free-standing material (a
few mg), and thus is not suitable for localized screening of
substrate-attached thin-film materials
-
32
libraries. However, recently introduced parallel
nano-differential scanning calorimeter (PnDSC) a
micro-machined array of calorimetric cells177 could facilitate
the use of calorimetry for the high-
throughput characterization of thin films, as demonstrated for
Ti-Ni-Zr shape memory alloy thin
films178, and provide additional insight into crystallization
kinetics and activation energies.179
R(T) measurements in the temperature range from -40 C to 250 C
(heating/cooling rate
5 K/min) were made using a custom-designed, automated 4-point
probe test stand (Fig. 3.8).180 In
general, two different measuring modes, i.e. screening and
single mode measurements were per-
formed. For the screening of the thin film SMA materials
libraries the 4-point probe was automati-
cally positioned to predefined loca