Development of Fe-Pd-based ferromagnetic shape memory alloys ...
Transcript of Development of Fe-Pd-based ferromagnetic shape memory alloys ...
Development of Fe-Pd-based ferromagnetic shape memory
alloys by using combinatorial materials science
Dissertation
zur
Erlangung des Grades
Doktor-Ingenieur
der
Fakultät für Maschinenbau
der Ruhr-Universität Bochum
von
Sven Hamann
aus Essen
Bochum 2013
Dissertation eingereicht am: 08.01.2013
Tag der Verteidigung: 20.03.2013
Erster Referent: Prof. Dr.-Ing. Alfred Ludwig
Zweiter Referent: Prof. Dr.-Ing. Gunther Eggeler
Gold is for the mistress - silver for the maid -
Copper for the craftsman cunning at his trade.
"Good!" said the Baron, sitting in his hall,"
But Iron - Cold Iron - is master of them all."
From “Cold Iron” by Rudyard Kipling (1865 - 1936)
“Bella veniunt et abeunt, sed gregarii mei pro aeternitate sunt.”
Executive Summary / Kurzfassung
Im Rahmen der vorliegenden Promotionsarbeit wurde in einem ersten Schritt die binäre
ferromagnetische Formgedächtnislegierung Fe70Pd30 hinsichtlich ihrer strukturellen
Eigenschaften und in Abhängigkeit verschiedener Herstellungsmethoden sowie der Probenart
untersucht. In einem zweiten Schritt wurden die Methoden der kombinatorischen
Materialforschung genutzt, um neue ternäre Fe-Pd-X ferromagnetische
Formgedächtnissysteme mit verbesserten Eigenschaften zu entwickeln. Dazu wurden dritte
Elemente wie Mn und Cu zum Fe-Pd Ursprungssystem hinzulegiert und deren Einfluss auf
das martensitische Umwandlungsverhalten untersucht. Zur Herstellung wurden Fe-Pd(-X)
Dünnschicht-Materialbibliotheken durch Magnetron-Kathodenzerstäubung auf oxidierten Si
Substraten abgeschieden. Neben simultaner Deposition wurden auch keilförmige
Viellagenschichten zur Herstellung vollständiger binärer sowie partieller ternärer
Materialbibliotheken verwendet. Automatisierte Hochdurchsatz-Charakterisierungsmethoden
wurden genutzt, um die Materialbibliotheken hinsichtlich ihrer Zusammensetzung
(energiedispersive Röntgenanalyse), der Struktur (Röntgenbeugung), des Phasen-
umwandlungsverhaltens (temperaturabhängige Widerstandsmessung) und der magnetischen
Eigenschaften (Magneto-optischer Kerr Effekt) zu untersuchen. Vielversprechende
Zusammensetzungsbereiche wurden dabei mit zusätzlichen Untersuchungsmethoden
hinsichtlich ihres strukturellen Transformationsverhaltens (temperaturabhängige
Röntgenbeugung und Transmission-Elektronen-Mikroskopie) und hinsichtlich der
Sättigungspolarisation, Curie-Temperatur und magnetokristallinen Anisotropie analysiert.
In binären polykristallinen Fe-Pd Materialbibliotheken wurde das zusammensetzungs-
abhängige Transformationsverhalten in Abhängigkeit verschiedener Kathodenzerstäubungs-
methoden untersucht und die damit verbundenen strukturellen Unterschiede mit den
Materialeigenschaften korreliert. Im nächsten Schritt wurde eine neue Methode zur
Herstellung von massiven Fe70Pd30 Proben erprobt und die optimalen
Prozessierungsparameter ermittelt. Das Kapitel der binären Proben schließt mit der
Herstellung von einkristallinen Fe-Pd Dünnschichten und deren Eigenschaften sowie
Umwandlungsverhalten hinsichtlich struktureller Modifikationen. Zunächst wurden die
strukturellen und magnetischen Eigenschaften der verschiedenen Phasen in Abhängigkeit der
Zusammensetzung untersucht. Daran schließt sich die Herstellung freistehender einkristalliner
Fe70Pd30 Dünnschichten an, bei denen eine bisher nicht bekannte reversible strukturelle
Transformation von einer kubisch-raumzentrierten über eine tetragonal-flächenzentrierten hin
zu einer kubisch-flächenzentrierten Struktur beobachtet wurde. Der letzte Teil des Kapitels
beschreibt die Herstellung von 1.2 µm dicken freistehenden Fe70Pd30 Schichten und deren
strukturelle Eigenschaften.
Im anschließenden Kapitel wurden zwei neue ferromagnetische Formgedächtnis Systeme mit
verbesserten Materialeigenschaften entwickelt und untersucht. Das Fe-Pd-Mn System zeigte
dabei eine Vergrößerung des Zusammensetzungsbereichs, in dem eine reversible strukturelle
Transformation erfolgt. Im Vergleich zum binären Fe70Pd30 System wurden die bisher
höchsten Transformationstemperaturen beobachtet. Diese Erhöhung der
Transformationstemperaturen wurde auf Basis von atomistischen Berechnungen, welche in
Kooperation mit M. E. Gruner (Universität Duisburg-Essen) durchgeführt wurden, mit dem
Effekt der magnetischen Anregung von antiferromagnetischen Elementen in einer
ferromagnetischen Fe-Pd Umgebung korreliert. Eine Verringerung der Curie-Temperatur und
der Sättigungspolarisation wurde für diesen Zusammensetzungsbereich beobachtet.
Das letzte Kapitel behandelt die Entwicklung und Untersuchung des Fe-Pd-Cu Systems.
Neben einer Erweiterung des Zusammensetzungsbereichs, in dem eine reversible strukturelle
Transformation erfolgte, wurde auch hier eine Erhöhung der Transformationstemperaturen
beobachtet und mit der erhöhten Löslichkeit von Fe in der transformierenden Phase im Fe-Pd-
Cu System begründet. Neben einer Verringerung der Curie-Temperatur und der
Sättigungspolarisation für diesen Zusammensetzungsbereich wurde eine Erhöhung der
spontanen Volumenmagnetostriktion entdeckt und in Fe-Pd-Cu Massivmaterialien bestätigt.
Strukturell modifizierte, einkristalline Fe70Pd30-XCuX (X = 3, 7 at.%) Schichten wurden
hergestellt und hinsichtlich ihrer magnetischen Eigenschaften, in Abhängigkeit der
Tetragonalität/Kubizität der Elementarzelle, untersucht. Dabei wurde festgestellt, dass 3 at.%
Cu, im Vergleich zu binärem Fe70Pd30, eine Erhöhung der magnetokristallinen
Anisotropiekonstante K1 um bis zu 40% erzeugt. Die Fe70Pd27Cu3 Zusammensetzung zeigt
somit den höchsten Wert für die magnetokristalline Anisotropiekonstante K1, welcher jemals
für eine ferromagnetische Formgedächtnislegierung gemessen wurde.
Contents
1. Introduction ..................................................................................................................1
2. Fundamentals................................................................................................................3
2.1 The martensitic transformation................................................................................3
2.2 The thermal shape memory effect............................................................................6
2.3 The ferromagnetic shape memory effects ................................................................8
2.4 Magnetostriction ...................................................................................................10
2.5 The Invar effect.....................................................................................................12
2.6 The Fe-Pd system..................................................................................................13
2.7 Ternary Fe-Pd-X systems......................................................................................23
2.8 Routes for the development of novel Fe-Pd-X alloys.............................................25
2.9 Combinatorial materials science............................................................................28
2.10 Thin film nucleation and growth ...........................................................................31
3. Experimental methods................................................................................................35
3.1 Fabrication and Processing....................................................................................35
3.1.1 Thin film materials libraries...........................................................................35
3.1.2 Epitaxial thin films ........................................................................................40
3.1.3 Bulk samples / Splats.....................................................................................41
3.2 Characterization ....................................................................................................43
3.2.1 Energy-dispersive X-ray analysis (EDX) .......................................................43
3.2.2 Structural analysis by X-ray diffraction .........................................................45
3.2.3 Microstructural analysis by Transmissionen-Electron-Microscopy ................46
3.2.4 Temperature-dependent resistance measurements..........................................47
3.2.5 Magnetic properties, screening and high-resolution measurements ................49
3.2.6 Mechanical properties investigated by nanoindentation .................................50
4. Results and Discussion................................................................................................51
4.1 Binary Fe-Pd Ferromagnetic Shape Memory Alloys..............................................51
4.1.1 Polycrystalline Fe-Pd thin films.....................................................................51
4.1.2 Bulk / Splat Samples .....................................................................................66
4.1.3 Epitaxial Fe-Pd thin films ..............................................................................74
4.2 Ternary Fe-Pd-X Ferromagnetic Shape Memory Alloys......................................100
4.2.1 The Fe-Pd-Mn System.................................................................................100
4.2.1.1 Polycrystalline Fe-Pd-Mn thin films ............................................................100
4.2.2 The Fe-Pd-Cu System..................................................................................119
4.2.3.1 Polycrystalline Fe-Pd-Cu thin films .............................................................119
4.2.3.2 Fe-Pd-Cu Splats ..........................................................................................135
4.2.3.3 Epitaxial Fe-Pd-Cu thin films ......................................................................142
5. Conclusion and Outlook ...........................................................................................160
6. References .................................................................................................................163
Acknowledgments.............................................................................................................174
Curriculum Vitae .............................................................................................................176
List of abbreviations and symbols
a [nm] lattice parameter
Af [K] austenite finish temperature
As [K] austenite start temperature at.% [-] atomic percent bcc [-] body centered cubic bct [-] body centered tetragonal c [nm] lattice parameter c/a [-] tetragonality of unit cell CPA [-] coherent potential approximation CSM [-] continuous stiffness method d [nm] lattice spacing DC [A] direct current DFT [-] density functional theory ∆Η [J] latent heat DSC [-] differential scanning calorimetry E [J] energy e/a [-] valence-electron ratio EBSD [-] electron backscatter diffraction EDX [-] energy-dispersive X-ray analysis EELS [-] electron energy loss spectroscopy
EK [J] magnetocrystalline anisotropy energy fcc [-] face centered cubic fct [-] face centered tetragonal FIB [-] focussed ion beam FSMA [-] ferromagnetic shape memory alloy FWHM [-] full-width-at-half-maximum GGA [-] generalized gradient approximations GGA [J] Gibb's free energy H [Oe] magnetic field
HA [Τ] anisotropy field HAADF [-] high-angle annular dark field
HC [Τ] magnetic coercivity HRTM [-] high-resolution TEM
HS [T] saturation field ICDD [-] international centre for diffraction data ICSD [-] inorganic crystal structures database
JS [T] saturation polarization K [kJ/m³] magnetocrystalline anisotropy constant
kB [J/K] Boltzmann constant KKR [-] Korringa-Kohn-Rostoker
L10 [-] tetragonal phase
L12 [-] cubic austenite phase M [T] magnetization of a martensitic variant MAE [-] magnetocrystalline anisotropy energy
M f [K] martensit finish temperature MF [-] misfit MFIS [-] magnetic field induced strain MIM [-] magnetic field induced martensite MOKE [-] magneto-optical Kerr effect
MS [K] martensit start temperature N [-] demagnetization factor NRA [-] nuclear reaction analysis PCA [-] principal component analysis PLD [-] pulsed laser deposition PPMS [-] physical property measurement system PVD [-] physical vapour deposition Q [J] activation energy for grain boundary movement R(T) [Ohm] temperature-dependent resistance
R0 [nm] initial grain size RBS [-] Rutherford backscatter diffraction Ref. [-] reference RF [Hz] radio frequency
rFe [nm] covalent atomic radius of Fe
rMn [nm] covalent atomic radius of Mn
rcr [nm] critical nucleus radius
s [-] shape parameter in Kuz’min’s equation SAED [-] selected area electron diffraction SEM [-] scanning electron microscope SMA [-] shape memory alloy SPR-KKR [-] spin polarized relativistic Korringa-Kohn-Rostoker code SQUID [-] superconduction quantum interference device STEM [-] scanning transmission electron microscopy T [K] temperature t [s] time
TC [K] Curie temperature TEM [-] transmission electron microscopy URQ [-] ultra-rapid-quenching VASP [-] Vienna Ab-initio Simulation Package VSM [-] vibrating sample magnetometer WDX [-] wavelength-dispersive X-ray analysis XRD [-] X-ray diffraction XRD(T) [-] temperature-dependent X-ray diffraction Z1 [-] tetragonal phase
α [°] angle α [-] proportional factor β [°] angle ∆Τ [K] thermal hysteresis ε [%] strain φ [°] angle γ [J] interface energy ϕ [°] angle ϑ [°] angle λ [nm] wavelength
λ(hkl) [-] saturation magnetostriction in (hkl) direction
λS [-] saturation magnetostriction θ [°] angle σ [MPa] stress
σbl [MPa] blocking stress
σmag [MPa] magnetic field induced stress
σtbm [MPa] stress for twin boundary movement ψ [°] angle Γ [-] center of the Brillouin zone
1
1. Introduction
Over the last two decades the development of new miniaturized actuators and sensors
increased rapidly.1 This progress is significantly caused by the ongoing discovery of new and
the enhancement of already known material systems that exhibit functional effects. Especially
the miniaturization benefits from these functional effects by using the material itself as
actuating or sensing device. Beside the well-known class of conventional shape memory
alloys (SMAs), such as Ni-Ti2, ferromagnetic shape memory alloys (FSMAs), such as Ni-Mn-
Ga3 and Fe70Pd304, became recently the object of increased scientific interest, due to their
advantageous properties.
The Fe-Pd system is a very promising candidate to be implemented as a functional material
into miniaturized systems, since it exhibits many physical effects that can be exploited for
functional application. Because of its high magnetocrystalline anisotropy, that is oriented
perpendicular to the sample surface, when fabricated in a thin film design, it is interesting as a
magnetic recording material.5,6 In contrast to non-ferromagnetic materials, Fe-Pd shows an
anomalous low thermal expansion over a defined temperature range, what is commonly
known as Invar effect.7,8 Furthermore, magnetostricition is observed for Fe-Pd, showing a
significant length change when exposed to a magnetic field.9 When the Fe70Pd30 composition
is quenched from high temperatures, a face-centred cubic austenite phase can be stabilized at
ambient temperature. Upon cooling, the Fe70Pd30 composition undergoes a reversible
martensitic transformation from a face-centred cubic austenite phase to a face-centred
tetragonal martensite phase.4 This martensitic transformation is highly composition dependent
and occurs for Fe70Pd30 slightly below 293 K. With increasing Fe content the martensitic
transformation is shifted to higher temperatures. Due to a twinned microstructure of the face-
centred tetragonal phase, martensitic variants are formed. These variants are separated by twin
boundaries and have a preferred direction of magnetization. This gives rise to a macroscopic
strain up to 3% originated by the reorientation of martensitic variants, when an external
magnetic field is applied. This effect is known as magnetic-field induced strain and can be
utilized in microsystems to work for example as micro-gas valve, as reported in literature.10
Thus, the magnetic-field induced strain effect could be implemented in several micro- and
nano-electro-mechanical-systems to enable friction free and non-vibrating movements.
Most of these effects were observed for single-crystal samples, as reported in literature.11,12
Since the fabrication of single crystals is time and cost consuming, alternative fabrication
2
methods have to be established to open the way for implementing Fe-Pd as active material
into miniaturized application. Especially thin films are promising for miniaturized application
and allow a rather cost effective fabrication. In contrast to bulk samples, thin films are
strongly influenced by substrate and surface constraints. This can alter significantly the
materials properties when transferred from bulk to thin film geometry. Nevertheless, several
issues still hamper a broad implementation of binary Fe-Pd into technical devices. Especially
the low transformation temperatures, the metastability of the transforming austenite phase and
the insufficiently high magnetocrystalline anisotropy need to be enhanced. Further the high
amount of a cost-expensive element like Pd constrains any mass production. Thus, new
ferromagnetic shape memory alloys based on the Fe-Pd system are of high interest, if they
overcome the previously mentioned constraints. This can be accomplished by the addition of
third elements into Fe-Pd in order to develop novel and enhanced materials. Nevertheless
there is only sparse information about ternary Fe-Pd-X materials so far, and no defined rules
for developing were identified which would allow an alloy design. The impact of Fe content
and valence-electron number e/a as dominating parameters, controlling the temperature shift
of the martensitic transformation, is still not clarified and needs a comprehensive
investigation. This can be achieved by investigating ternary Fe-Pd-X alloys that allow to
separate the different parameters and to gain a deeper insight into the composition-structure-
property relationship.
3
2. Fundamentals
Within the following chapter, the fundamentals of the Fe70Pd30 FSMA are presented and the
different properties in this alloy, fundamental for the magnetic field-induced strain, are
pointed out. In the first part the martensitic transformation is presented, followed by a
description of the conventional shape memory effect. Then the actuation effects occurring in
FSMAs are shown and the prerequisites for showing these effects are pointed out. The next
subsection emphasizes on magnetostriction and explains the fundamentals of this effect. The
causation of the anomalous reduction of thermal expansion coefficients, also known as Invar
effect, which is occurring in Fe-based alloys, is presented in the following subsection. The
next section introduces the Fe-Pd system with the different structural phases that can occur
and their intrinsic properties. A further part examines the change of intrinsic properties of
ternary Fe-Pd-X alloys and gives an overview of the state-of-the-art literature. Subsequently a
route for the successful development of new Fe-Pd-based FSMAs with enhanced intrinsic
properties is presented. The last section presents the procedure that was used to develop new
Fe-Pd-based FSMAs.
2.1 The martensitic transformation
A martensitic transformation is defined as a displacive, first-order diffusionless
transformation from a high-temperature phase (austenite) having a highly symmetric structure
to a low temperature phase (martensite) with a lowered symmetric structure.13 This
transformation was observed for the first time by Adolf Martens in 1887, when he
investigated hardening mechanisms in steel.14 During a martensitic transformation, the atoms
move only fractions of the lattice distance without any local changes in chemical composition
when the structure transforms from the austenite to the martensite phase. Due to the small
displacement of the atoms in the lattice, well-defined correlations exist between the austenite
and the martensite phase.15 The martensitic transformation is a first order exothermic
transformation.16 A change from the austenite to the martensite phase is induced, when the
difference in Gibb’s free energy between both phases (∆GA-M) is higher than the energy
needed for elastic deformation of the crystal lattice (∆GD), formation of new bounding
surfaces (∆GS) and the amount of internal friction (∆GF):
4
0)GGG()GG(G FSDMartensiteAusteniteMA >∆+∆+∆+−=∆ − (1)
In Figure 1 the martensitic transformation is depicted in terms of Gibb’s free energy as a
function of temperature for the austenite and the martensite phase. Upon cooling the
transformation starts from the high-temperature austenite phase at the martensite start
temperature (Ms). This temperature is shifted to lower values (∆T) than the equilibrium
temperature T0, since a driving force is needed to overcome ∆GD, ∆GS and ∆GF. The thermal
hysteresis between these two states is originated by the amount of ∆GD, ∆GS and ∆GF.
Figure 1: Temperature-dependence of Gibb’s free energy for both the austenite as well the martensite phase
(adapted from Ref. 17). To induce the transformation from the high-temperature (austenite) to the low-
temperature (martensite) phase an undercooling is needed to overcome ∆GD, ∆GS and ∆GF.
In dependence on the crystallographic compatibility between austenite and martensite phase,
the energy for elastic deformation ∆GD can become negligible, leading to a decrease of the
undercooling ∆T and thus of the thermal hysteresis. When the temperature reaches the
martensite finish temperature (Mf), the structure has changed fully to the martensite phase. If
the temperature is increased again, the structure changes back from the martensite to the
austenite phase. Due to the reversibility of this process, this kind of transformation is defined
as a thermoelastic transformation.
The change of the crystal lattice within a martensitic transformation can be described by three
stages.18 When a sample undergoes a martensitic transformation, the atoms are sheared along
a neutral habit plane (Figure 2 a) to b)) distorting the austenitic matrix.19 By lattice invariant
shearing the martensite adapts the austenite matrix in order to compensate these shearing
stresses. To reduce these stresses the lattice undergoes either a gliding (c)) or a twinning (d))
5
movement. When the lattice undergoes a gliding (b) to c)), atomic bindings are broken and
new bindings are formed, leading to an irreversible change of the structure.16
Figure 2: Martensitic transformation from the austenite (a)) to the martensite phase by a shearing mechanism
(b)) that can be accompanied by gliding and twinning (c) and d)) for stress relaxation in the martensite (adapted
from Ref. 15).
When a lattice is twinned to reduce shearing stresses, crystal variants (martensitic variants)
are adjusted along a boundary (twin boundary) in a reversed imaged manner. The amount and
kind of bindings of a twin boundary are equal to atoms that do not belong to this boundary.
This is why twin boundaries have a low energy, are highly mobile and can be easily moved
through a crystal. In ordered alloys twinning is the energetically favoured mechanism for
stress relaxation, because all atomic bindings and additionally the degree of order are retained.
Deformation of the lateral dimensions of a sample by twin formation is reversible and
therefore the prerequisite for the shape memory effect. General to all shearing mechanisms a
rotation of the lattice is needed to keep an invariant habit plane. Concluding this, the
martensitic transformation can be defined as a shear-dominated reversible phase
transformation where the nearest atomic neighbours are preserved. Due to the shear
mechanisms, a weakening of the lattice near the martensitic transformation temperature can
6
be observed giving rise to unusual behaviour of a sample’s intrinsic properties like Young´s
modulus, resistivity and thermal expansion.20
Due to the lower symmetry of the martensite lattice, external macroscopic strains can be
compensated better than in the austenite state. Thus besides temperature also strain can be
used as the driving force to induce a martensitic transformation. When a sample in the
austenite state is strained, the lattice starts to transform into the martensite state at a critical
and temperature-dependent stress. The differential change of transformation temperature dMS
by application of a differential stress dσ is described by the Clausius-Clapeyron
equation:13,21,22
ε⋅∆−=σT
H
dM
d
S
(2)
In this equation ∆H defines the latent heat, which is released within the first-order
transformation. The strain of a sample is represented by ε and the temperature by T.
Materials undergoing a martensitic transformation exhibit an anomalous mechanical
behaviour at temperatures near the transformation like lattice softening.23 When a sample is
just above Af it has an austenitic structure. On the application of an external stress the
martensite phase can form because of a stress-induced martensitic transformation following
the Clausius-Clapeyron relation. In the martensite phase external stresses are compensated by
the development of martensitic variants having their long axis aligned to the external stress
field. This effect allows for high strains up to 10% without significant increase of the external
stress and is defined as superelasticity.24 Since the temperature is above Af the martensite
phase is unstable and will transform back into the austenite phase when the external stress is
removed.
2.2 The thermal shape memory effect
The thermal shape memory effect is defined as the ability of a material to recover its original
shape after deformation, when heated to a certain temperature. This is correlated to the
thermoelastic martensitic transformation. At the beginning the material is in the high-
temperature austenitic state (Figure 3 a)). Upon decreasing the temperature it transforms into
the low-temperature martensite state while keeping its overall shape. Due to spatial
restrictions the lattice change is compensated by developing martensitic twins (Figure 3 b)).
These twins are connected by twin boundaries that are highly mobile and allow the material to
7
easily deform when being in the martensite state. If the material is deformed by external
forces martensitic variants compensating most of the stress are preferably formed (Figure
3 c)).
Figure 3: Conventional shape memory effect where a cubic austenite phase a) is transformed into a tetragonal
martensite phase showing a twinned structure b) upon cooling. The different variants are connected by twin
boundaries that allow for an easy deformation of the crystal c). When heated, the crystal transforms back to the
austenite phase - c) to a) - recovering its primary shape.25
This is conducted without any breaking of atomic bonds, leading to a defined relationship for
all atoms within both phases. This gives rise to the shape recovery when the material is heated
to its high symmetric austenite state subsequently. When the temperature is lowered the
material transforms again into the martensite phase without any further shape change. Thus
this effect is defined as one-way-effect. By using appropriate thermal processes in accordance
with mechanical treatment, the material can be trained by inducing dislocations within the
structure. This provides a further shape change within the material when undergoing a
martensitic transformation by heating and cooling. This effect is defined as two-way-effect
and can be used as actuating mechanism for conventional shape memory alloys.
8
2.3 The ferromagnetic shape memory effects
In general two different actuating effects occur within FSMAs. Both effects exhibit a
reversible straining of a sample upon applying an external magnetic field accompanied by a
change in crystal structure. The straining can reach values up to 10% making this more than
ninety-times higher than conventional magnetostriction (TbDyFe26 λS= 1100⋅10-6). Compared
to conventional SMAs, where the actuation is induced by temperature changes, magnetic
fields can be switched faster up into the kHz range. This is only limited by the sonic speed
that restricts the movement of twin boundaries in an alloy.27
For the first effect a martensitic transformation from a weak to a strong magnetic moment
phase is induced by applying an external magnetic field at temperatures near the
transformation. This is correlated to a decrease of free energy during transformation
originated by the magnetic contribution to Gibb’s free enthalpy.28 The effect is known for Ni-
Mn-Ga and Ni-Mn-In alloys, where the martensite phase is induced by magnetic fields and is
defined as magnetic field induced martensite (MIM).29
Figure 4: Schematic of the magnetic field induced strain effect where the orientation of martensitic variants is
changed by an external magnetic field. a) Without a magnetic field a random distribution of martensitic variants
appears. The direction of magnetic moments in the different variants is depicted by a small red arrow. By
applying an external magnetic field (large red arrow in the lower left in b)) the martensitic variants having their
magnetic moment aligned along the external magnetic field are favoured and grow at the expense of others. (the
angle between the easy axes of different variants was chosen to be < 90°). c) With increasing external magnetic
field more variants are aligning along by twin boundary movement (green arrows). d) When the external
magnetic field reaches a critical value all variants are aligned into one direction giving rise to an overall straining
of a sample (adapted from Ref. 30).
In magnetic anisotropic alloys the magnetic moment within the lattice cell is aligned along
certain crystallographic axes. Thus especially in the low symmetry martensite phase the
different variants have a magnetic moment that is oriented to a specific-axis in the lattice (in
9
the case of Fe70Pd30 it is the shorter c-axis of the fct unit cell).57 Without an external magnetic
field, the different variants are oriented to the crystal lattice. If an external magnetic field is
applied, the martensitic variants and thus the crystal lattice start to align along the field
direction. This gives rise to a magnetic field induced strain (MFIS) effect as depicted in
Figure 4. The MFIS appears only in magnetic materials that have a martensite phase like Ni-
Mn-Ga, Fe70Pd30 and Fe3Pt. The main prerequisite for this effect is a sufficient
magnetocrystalline anisotropy that couples the magnetic moment to a specific-axis of the
crystal lattice. The magnetocrystalline anisotropy generates a direction-dependent
magnetization behaviour of the crystal lattice where the material can be magnetized easily in
certain crystallographic directions (easy axis) while other directions are hard to magnetize
(hard axis). In the following all aspects for magnetic field induced twin boundary motion is
declared for a tetragonal structure where c describes the short and easy axis and a defines the
long axis that is hard to magnetize. In general the magnetocrystalline anisotropy energy EK in
alloys having an uniaxial symmetry can be written as:31
)(cos)(cosK)(sinK)(sinKK),,(E 223
42
210K β⋅α+φ+φ+=βαφ (3)
The parameters φ, α, β define the angles of the magnetization to the c-axis ([001]-direction)
and the a-axes ([100]-and [010]-direction) within the unit cell. The maximum
magnetocrystalline anisotropy energy for uniaxial structures is defined as KU = K1 + K2. This
is the case when sin2φ = 1, which describes the orientation of magnetization aligned to the
hard a-axis.32 When an external magnetic field is applied to two adjacent martensitic variants
having their easy axis oriented at an angle of ϑ = 90° to each other, the maximum energy
difference ∆Emag between is defined as:33
U2
U21max,mag K)(sinKH)MM(E =ϑ⋅+⋅−=∆ (4)
The maximum strain that can be reached is defined by the lattice parameters a and c of the
tetragonal unit cell and can be described by ε0 = 1 − c/a. Further the energy difference
∆Emag,max is proportional to the magnetic field induced stress σmag by:34
0magmagmag0
magEE ε⋅∆=σ⇔∆=
εσ
(5)
10
If the maximum magnetic field induced stress σmag applied to the martensitic variant is higher
than the stress σtbm (twin boundary movement) required for moving the twin boundary
between two adjacent variants, a reorientation of the variants appears. This is expressed by:35
tbm0
Umag
K σ≥ε
=σ (6)
This gives rise to a full reorientation of all martensitic variants having their easy c-axis
aligned to the external magnetic field.36 If the c-axis is the short axis in the tetragonal crystal
this effect decreases the outer dimensions of a sample giving rise to the actuation effect. In
order to achieve this MFIS easy movable twin boundaries are one prerequisite for FSMAs.
Materials exhibiting the MFIS have to combine a sufficiently high magnetocrystalline
anisotropy KU with easily movable twin boundaries. Mostly known alloys are: Ni2MnGa3,
Fe3Pt37, Fe70Pd304 and La1.99Sr0.01CuO4
38 but also pure elements like Tb and Dy39 exhibit the
MFIS at very low temperatures of about 4.2 K. The MFIS is highly sensitive to structural
defects and grain boundaries that can pin the twin boundary movement and therefore prevent
any straining. Thus, the maximum achievable straining in polycrystalline materials by the
MFIS is reduced. Up to now single crystals37,40 (Ni2MnGa: 9.5 %; Fe70Pd30: 3 %) and foams41
(Ni2MnGa: 8.7 %) show the highest straining up to several percent for the MFIS. The two
main systems are Ni2MnGa and Fe70Pd30 that both combine a high magnetocrystalline
anisotropy (Ni2MnGa: KU= 160 kJ/m3 at 300 K; Fe70Pd30: KU= 180 kJ/m3 at 77 K) with easy
movable twin boundaries (Ni2MnGa: ε = 0.06; Fe70Pd30: ε = 0.07) at a maximum magnetic
field induced stress (Ni2MnGa: σmag = 2.6 MPa; Fe70Pd30: σmag = 2.3 MPa).34,42
2.4 Magnetostriction
In general magnetostriction is defined as the elastic straining of a sample’s volume or shape
originated by magnetism. The effect is based on the spin-orbit coupling in magnetic materials,
which efficiently couples the magnetism directly to the electron spin and lattice orientation. It
is well known that the magnetic properties of metals and alloys depend on the arrangement
and separation of atoms in the lattice. Therefore, the volume of a material in the magnetically
ordered state is different from that it would have, if no spontaneous ordering of spins would
occur. This ordering of spins provides a repulsive force that results in an increase of the
crystal unit cell volume.43
11
Such an effect is defined as spontaneous volume magnetostriction where the volume of a
material increases without changing the shape. Following Kakehashi et al.50 the spontaneous
volume magnetostriction is a prerequisite for the Invar effect, which influences strongly
thermal expansion and magnetic properties of a material system.
When the electron spins are aligned by an external magnetic field, the electron orbitals change
their shape due to the spin-orbit coupling. Because of the direct coupling of magnetism to the
electron spin and lattice orientation, the crystal lattice changes its distances. This gives rise to
an overall shape deformation and straining of a sample. Such a shape change at constant
volume is defined as Joule’s magnetostriction.
Figure 5: Macroscopic model of the magnetostrictive effect: a) The material is in a paramagnetic state with a
random orientation of magnetic dipoles. b) For temperatures below TC the material changes to a ferromagnetic
state accompanied by an alignment of the magnetic dipoles into areas with uniform magnetic moment (magnetic
domains). c) By applying a magnetic field all areas with uniform magnetic moment align, giving rise to an
overall shape change. (adapted from Ref. 44)
In Figure 5 the principle of Joule’s magnetostriction for an isotropic sample is presented.
When a magnetic material is above its transition temperature (Curie temperature) it is in a
paramagnetic state, with a random distribution of magnetic dipoles (Figure 5 a), dipoles are
depicted by coloured bean-shaped symbols). By lowering the temperature the paramagnetic
state changes to a ferromagnetic state, leading to the formation of regions with aligned
magnetic moments (magnetic domains). In order to minimize the stray field energy the
magnetic domains are aligned in a way where the magnetic flux lines are closed (Figure 5 b)).
If an external magnetic field is applied the magnetic domains align according to this field
leading to a straining of the crystal lattice and thus giving rise to a straining of the sample
(Figure 5 c)). The shape change of the sample in a certain direction is defined by λ(hkl) = dl/l
12
with a positive algebraic sign “+” for elongation and a negative “−“ for contraction. The
averaged magnetostriction constant for a polycrystalline cubic material is defined by:
111100S 5
3
5
2 λ+λ=λ (7)
Here, λ100 and λ111 are the saturation magnetostriction in (100) and (111) direction of a cubic
lattice. In general nearly all magnetic elements and materials show magnetostriction, although
this effect is rather small for pure Fe (λS = -3.6⋅10-6) it reaches values for lanthanides like Tb
and Dy (λS = -3.6⋅10-6 and λS = -3.6⋅10-6) at cryogenic temperatures.
Like all ferromagnetic materials, disordered Fe70Pd30 exhibits a Joule’s magnetostriction
effect. The averaged magnetostriction constant λS = 5⋅10-5 is relatively high in comparison to
the values for pure Fe.9
2.5 The Invar effect
The Invar effect was first observed in 1896 by the Swiss physician Charles Édouard
Guillaume in a Fe65Ni35 alloy, when he tried to develop new materials for length and mass
standard application.45 In 1920 Guillaume was awarded the Nobel price for this discovery.
This effect denotes the anomalous behaviour of thermal expansion coefficients around room
temperature where the volume expansion of an alloy can be zero or even become negative.
This is caused by a superposition of normal thermal expansion due to anharmonic terms in the
interatomic potential and a volume contraction due to the change of the magnetic state. This
effect is well known to appear in Fe-rich alloys, such as Fe-Ni, Fe-Pt and Fe-Pd but it was
also found in several other alloys.46 This effect is not fully understood so far, although
different approaches have been made to explain it. In general, the main theories correlate this
behaviour to two concurrent magnetic states.47 The two-γ-state model was developed by
Weiss et al. in 1963 for the face-centred cubic γ-Fe and reports on the coexistence of two
different ferromagnetic states: The γ1 state is a low energetic state having a small volume
(a = 0.354 nm) and a weak magnetic moment (0.5 µB); the γ2 state is a high energetic state
with a large volume (a = 0.364 nm) and a strong magnetic moment (2.8 µB).48 If the energy
difference between these two states is small enough, the γ1 state can be activated at moderate
temperatures. Due to the smaller volume of the γ1 state the occupation of this state counteracts
to the normal thermal expansion and thus leads to an overall anomalous decrease of the
13
thermal expansion coefficients. Another explanation of the Invar effect in Fe-rich alloys was
suggested by Kondorsky and Sedov.49 They suggested a latent antiferromagnetism where an
antiparallel alignment for the Fe atoms within a ferromagnetic matrix is energetically
favoured. If the Fe content is increased to a certain amount, a break down of the exchange
coupling and thus of the spontaneous magnetization appears. These magneto-elastic
anomalies are correlated to changes of the exchange coupling in dependence on volume and
pressure. Therefore it is energetically favoured to increase the volume at low temperatures
when the exchange coupling energy is minimized. Cooling this alloy leads then to an increase
of volume. When it is heated the magnetic exchange coupling is weakened due to
spontaneous spin fluctuations and accordingly the volume decreases again. This is
superimposed by thermal expansion and leads to anomalous behaviour of thermal expansion
coefficients. Kakehashi et al. correlate this to a gradual transition from a strong to a weak
magnetic state in caused by thermal excitations.50
The Invar effect was found to appear in disordered Fe70Pd30 alloys as reported by Schlosser.51
Matsui et al.52 found an anomalous behaviour of thermal expansion of Fe68Pd32 over a
temperature range from 4.2 K to 800 K indicating a reduction of the thermal expansion
coefficient α [K -1] when increasing the temperature > 400 K. This anomalous behaviour of
thermal expansion becomes maximal when a sample is near a structural instability like a
martensitic transformation as reported by Nakayama et al..53
2.6 The Fe-Pd system
The thermodynamic stable structural phases of the Fe-Pd system in dependence on
composition and temperature at constant ambient pressure are presented in the phase diagram
in Figure 6. Further the metastable phase diagram close to the Fe70Pd30 composition is shown
and the thermoelastic transformation from a face centred cubic (fcc) to a face centred
tetragonal (fct) structure in this compositional region of interest is explained and compared
with recent theoretical investigations. To gain a fundamental understanding of phase
formation a detailed understanding of the equilibrium phase diagram of the binary Fe-Pd
system is required. This diagram depicts the equilibrium phases in dependence on
composition and temperature. When lowering the temperature below the Liquidus-line the γ-
phase forms, that is stable over the complete compositional range. This phase has a fcc
structure with a random (disordered) distribution of Fe as well as Pd atoms in the lattice. The
existence region of the γ-phase lies between the Liquidus-line (the transition from solid to
14
liquid appears between 1577 K and 1828 K in dependence on composition) and reaches down
to temperatures of about 1173 K. Since this thesis focuses on the thermoelastic fcc to fct
transformation of samples near the Fe70Pd30 composition, the equilibrium as well the
metastable phase formation will be described. When considering an infinitesimal slow cooling
process the phase formation can be described as a sequence of thermal equilibrium states. At a
temperature of 1033 K the fcc Fe70Pd30 γ-phase decomposes into an α-Fe phase (bcc: body
centred cubic structure) with a Pd content < 3.5 at.% and a fcc γ-phase with decreased Fe
content according to the lever rule.54 Upon cooling along the γ-phase boundary the Pd content
increases up to a value of 46 at.% at 873 K.
Figure 6: Equilibrium phase diagram for the binary Fe-Pd system.55
When the temperature is lowered further, the Pd content in the bcc α-Fe phase decreases < 1
at.% (stabilized boundary solubility due to mixing entropies) and the fcc γ-phase stabilizes at
a Fe50Pd50 composition. This Fe50Pd50 phase has a L10 in layer ordered structure, that can be
described by a slightly distorted fcc structure. This slight distortion of the fcc structure
perpendicular to the layer planes results in a fct structure. A sequence of Fe and Pd layers
forms the L10 structure, where the lattice parameter in the layer plane is smaller than the value
for the out of plane direction. The phase formation in a bulk body is controlled by the
diffusion coefficients of Fe atoms. Since the diffusion coefficients and thus the diffusion
speed is limited, the formation of metastable phases can be achieved by rapid quenching to
15
low temperatures. When a high-temperature equilibrium phase is quenched to low
temperatures there is still a significant driving force for decomposition and formation of new
phases. This is prohibited by a significant reduction of diffusion kinetics at low temperatures
that avoids decomposition and allows for the existence of metastable phases.
The interest on disordered Fe70Pd30 is based on the appearance of a reversible martensitic
transformation from an austenitic fcc (in the following termed as metastable/transforming fcc
phase) to a martensite fct phase. This fct phase occurs below room temperature and can be
achieved by quenching the sample from the γ-phase to temperatures < 573 K. Figure 7 a) to c)
shows three metastable phase diagrams describing the sequence of the different structural
phases as a function of Pd content and temperature. All diagrams show the presence of a
reversible martensitic transformation from a high-temperature fcc phase to a low-temperature
martensite fct phase. The transformation temperatures increase with increasing Fe content
until the fct structure transforms irreversibly into the bct structure.
Figure 7: Metastable phase diagrams for quenched alloys around the composition Fe70Pd30. Presented are the
structures as a functions of the Pd content and temperature. a) The full phase transformation ranging from a bcc,
over a bct and a fct to a fcc structure is reported by Sugiyama et al.56. The transformation from fcc to fct was
reported to be reversible, while the fct to bct transformation was found to be irreversible. b) Cui et al.57 reports
on an extension of the transforming region up to Pd contents of 29.3 at.% where samples still have a fct structure
and undergo a martensitic transformation. c) Matsui et al.58 describes the sequence from bcc (α), over fct (γ’) to
fcc (γ) as a function of composition and temperature. Further the variation of lattice parameters ratio c/a is
correlated to the different phases. This implies that the c/a ratio can be regarded as a parameter to describe this
structural transition through the different phases.
This dependency for other Fe-based alloys like Fe-Ni59 and Fe-Pt60 is often described in
literature by the valence-electron to atom ratio (e/a). In Figure 7 a) the structural sequence for
Fe-Pd foil samples with a thickness of about 40 µm is depicted. These samples were
fabricated by melting and subsequent homogenisation annealing at 1373 K for 60 h followed
by dicing and rolling into foils. To obtain the disordered fcc phase the foils were solution
16
treated at 1173 K for 1 h and afterwards quenched in iced water. Due to the sample
preparation and treatment a polycrystalline structure with crystal lattice defects like
dislocations was observed. Further the existence range of the fct phase is increased to higher
Pd values. For Figure 7 b) single-crystal samples with a Fe70Pd30 composition were fabricated
using the Bridgman method from arc-melted buttons. This crystal was diced into several
pieces for investigation. Due to the fabrication method no hints for crystal defects were
observed. Here, the existence range for the fct phase is smaller than in a). The samples
presented in diagram c) were fabricated using a plasma jet furnace followed by a
homogenization annealing at 1423 K for 7 days encapsulated in a quartz glass tube. After
annealing the samples inside the glass tube were quenched in iced water to prevent
decomposition of the γ-phase. All diagrams show that the transformation from the fcc to fct
phase occurs around 300 K for the Fe70Pd30 composition. When the temperature is further
decreased the sample changes its structure irreversibly from a fct to a bct structure without
recovery of the fct phase upon subsequent heating. This irreversible transformation cannot be
used for the shape memory effect and thus needs to be avoided.61 The existence range of the
different structural modifications of the parental fcc phase differ slightly and thus also the
transformation temperatures in the presented diagrams. This is correlated to the different kind
of samples (single crystal, polycrystalline bulk and foils) and the processing routes. Especially
sample impurities and quenching rate affect significantly the developing microstructure as
reported by Matsui et al..58 This can also be determined by thermal hysteresis width that
differs for all three different sample types. The single-crystal sample from Cui et al.57 shows a
thermal hysteresis of about 5 K while this increases to about 10 K for the bulk samples
fabricated by Matsui et al. and reaches a maximum value of approximately 20 ± 10 K for the
foils. The thermal hysteresis is controlled among other factors by the crystal lattice misfit
between the austenite and the martensite phase. Therefore the single-crystal samples show the
smallest misfit between martensite and austenite indicated by a small amount of energy that
has to be given into the system to transform. In contrast the Fe-Pd foils need a significantly
higher amount of energy to undergo this transformation. This is related to the high amount of
structural defects, originated by the fabrication process, that constrain the martensitic
transformation. Stress effects originated by structural defects increase the transformation
temperatures, as mentioned in the previous chapter. This is observed for the bulk and foil
samples that exhibit higher transformation temperatures than the single-crystal samples. As
mentioned in Chapter 2.1 the stress state of a sample significantly affects the transformation
temperatures. Kato et al.22 report on an increase of temperatures for the fcc to fct
17
transformation in dependence on stress by a factor of dσ/dT = 4.8 MPa/K. The amount of Fe
in a sample depends significantly on the kind/geometry of sample. It was found that thin film
samples generally allow to increase the content of Fe atoms in the fct phase. This is different
when compared to bulk samples that form Fe-rich precipitates at lower Fe contents. Sugimura
et al.61 reported on a Fe72.5Pd27.5 thin film sample in a single fct phase.
Figure 8: The three martensitic variants that can be formed upon transformation from the fcc austenite to the fct
martensite phase.57 The variants develop by contracting one of the base vectors (v1, v2 and v3). For the fct
structure the long axes are termed as a and the short as c-axis.
Starting from the fcc structure there are three different variants that can be formed upon
transformation to the fct phase. In Figure 8 these martensitic variants are shown. According to
this model the cubic austenite transforms into the tetragonal martensite by contracting the
lattice along one of the catersian axes and expanding the others correspondingly. In the
following, the short axis is defined as c-axis while the two longer axes are termed as a-axes.
In Fe70Pd30 single crystals the lattice parameters are determined to be afcc = 0.3756 nm for the
fcc austenite and afct = 0.3822 nm and cfct = 0.363 nm for the fct martensite structure.57 The
degree of tetragonality of the fct martensite phase can be calculated by calculating the c lattice
parameter over the a lattice parameter. From the above mentioned lattice parameters Cui et
al.57 determined c/a = 0.95 while Matsui et al.58 extrapolated a value of c/a = 0.914.
Considering Matsui’s value a complete alignment of martensitic variants by an external
magnetic field (ferromagnetic shape memory effect, see Chapter 2.3) can lead to a maximum
strain of 6%. Stern et al.62 calculated the maximum strain to be around 6.8% by using
electronic structure calculations. The Curie temperature TC for Fe70Pd30 is determined to be
18
around 700 K and depends strongly on composition. Thus both the austenite as well as the
martensite are ferromagnetic with a saturation polarization of JS,fcc= 1.357 T and
1.508 < JS,fct < 1.885 T.57,62 The magnetic easy axis in the fcc austenite phase is aligned along
the [111] direction. The fct martensite phase has the magnetic easy axes aligned along the
[100] and [010] (a-axes) directions and the magnetic hard axis aligned to the c-axis of the
tetragonal unit cell.57
As shown in Figure 7 c) the sequence of different structures, developing in dependence on Pd
content and temperature is given by the c/a ratio of the fcc unit cell. This is expressed by the
Bain path formalism,63 which describes a continuous distortion of a unit cell from a fcc
austenite, over a fct and bct martensite to a bcc structure. According to Bain all these phases
can be defined as body centred tetragonal structures having different cbct- to abct-axis ratios.
Figure 9: Bain path distortion of a body-centred cubic (bcc) cell (“blue atoms”), with a c/a ratio 1, that can be
defined between two face-centred cells (“black atoms”). The bcc cell changes continuously its lattice parameters
c and a upon a tetragonal distortion. Starting from a bcc the lattice distorts to a body-centred tetragonal structure
with c/a = 1.02 upon elongation of the c-axis. When further distorted a face-centred tetragonal (fct) cell with
c/a = 1.33 develops, as defined by the black atoms. Finally the black atoms form a face- centred cubic (fcc) unit
cell with an c/a = 1.41 (adapted from Ref. 63).
Thus the structural sequence from a fcc to a bcc structure can be defined as a continuous
tetragonal distortion where the c/a ratio decreases from fcc to bcc. In Figure 9 the variation of
a unit cell upon tetragonal distortion along the Bain path is depicted. To understand the Bain
path formalism, there are two related systems based either on a face centred or a body-centred
lattice. The fcc austenite phase has a c/a ratio of cfcc/afcc=1 when using the face-centred
assumption (defined by “black atoms” in Figure 9). This can be also described by using the
19
body centred assumption, where 41.12a/c bctbct == (“blue atoms” in Figure 9). Within this
body centred assumption cbct=cfcc, while the abct axis is aligned along the face diagonal of the
face-centred cell. From geometry fccfccbct c2/1a2/1a ⋅=⋅= can be derived for a body
centred cell within a face centred lattice. The degree of tetragonal distortion can be calculated
from the face- to the body- centred lattice by fccfccbctbct a/c2a/c ⋅= . Within this thesis the
body-centred assumption will be used in the following to describe the degree of tetragonal
distortion and related results. Along the Bain path the c-axis is elongated continuously while
the a-axis is compressed to keep a nearly constant unit cell volume. Both, the fct
(cbct/abct = 1.33) as well as the bct (cbct/abct = 1.02) structure only differ in the degree of
tetragonal distortion. Besides the bcc and the fcc phase only the fct and the bct structures are
stable at T < 200 K. For temperatures T > 200 K several intermediate states between the bct
and the fcc phase with 1.02 < c/a < 1.41 were observed. As described in Chapter 2.1, a
martensitic transformation is a first order transformation accompanied by a step-like change
of intrinsic properties like lattice parameter, magnetic saturation polarization and electrical
resistance. The martensitic transformation appearing in alloys around the Fe70Pd30
composition is of weak first order, since the c/a ratio changes continuously from a fcc to a fct
phase.9,64 At the beginning the c/a ratio increases rapidly when starting to decrease the
temperature to induce the martensitic transformation from the high-temperature fcc austenite
to the low-tempertaure fct martensite phase. This steep increase of the c/a ratio stagnates
when the structure is completely transformed into the fct phase but keeps on increasing
further when the temperature is decreased. Additionally the Fe70Pd30 based alloys exhibit a
small hysteresis in the range of 2-5 K being related to the rather weak first order
transformation.
This weak first order transformation from a fcc to a fct structure around the Fe70Pd30
composition can be interpreted in terms of a band-Jahn-Teller instability, which is originating
from a degeneracy in the electronic band structure at the Γ-point.65,66 This degeneracy gives
rise to an energy gain (∆E) of approximately 14 meV/atom when the structure is changed
from a high symmetric fcc to a lower symmetric fct structure. The energy difference ∆E
between all different structures (bcc to fcc) along the Bain path is rather small allowing to
easily change the structure at finite temperatures initiated by such an energy gain.
Calculations of the energy gain for the single elements Fe and Pd along the Bain path
(c/abct ratio) confirm a stabilization at the known crystal structures (bcc for Fe and fcc for Pd).
20
Figure 10: Calculated energy landscape as a function of the c/abct ratio (from bcc to fcc) along the Bain path for
Fe (dark grey curve), Pd (grey curve) and Fe70Pd30 (red curve). The pure elements show a distinct energy
minimum at their equilibrium structures (bcc for Fe and fcc for Pd). For Fe70Pd30 a flattening of the energy
landscape (red curve) appears with a minimum around the fct structure.67
When ∆E is calculated for the Fe70Pd30 composition, the energy landscape flattens
significantly with an energy minimum around the fct structure as presented in Figure 10.67
This indicates that only a small amount of energy is needed to alter the structure from fcc to
fct. Figure 11 shows the total energy as a function of the tetragonal distortion of a lattice
along the Bain path. The upper curve is similar to Figure 10 and demonstrates the energy
landscape for Fe70.4Pd29.6 having an energy minimum around the fct structure. To investigate
the impact of relaxation upon tetragonal distortion on the total energy, a 108 atom supercell
was distorted at different c/abct ratios. The location of atoms was not fixed to predefined
position as in Figure 10. Instead a Conjugate-Gradient-Algorithm was used to calculate the
shifting of atomic locations with the lowest energy simulating the relaxation of the lattice
upon tetragonal distortion. Compared to the ideal positions, an energy gain ∆E ≥ 29 meV was
determined for all c/abct ratios upon relaxation of the lattice. The energy landscape of the
relaxed lattice is depicted by the lower curve in Figure 11. Now the energy minimum is
located at the bcc structure, defining a new structural ground state for this system. The energy
difference between bcc and fcc state is increased significantly to about 24 meV.65 This shows
that relaxation mechanisms have a significant impact on the structural ground state for
Fe70.4Pd29.6. Since all ab initio simulations were calculated at T = 0 K only estimations can be
given at finite temperatures. Especially thermal fluctuations and the impact of entropy for
finite temperatures can alter this energy profile leading to a different structural ground state.
21
Figure 11: Upper curve: Total energy as a function of the c/abct ratio for fixed atomic positions. A rather flat
energy landscape appears with an energy minimum around the fcc structure. Lower curve: When allowing the
atoms to displace their position and to relax the lattice at different tetragonal distortions, a different picture
appears. Compared to fixed atomic positions the lower curve gains ≥ 29 meV in energy. Due to the movement of
atoms the bcc structure defines the new ground state with an energy difference of 24 meV between bcc and fcc
structure.65
For magnetic materials, either in bulk or in thin film geometry, the exchange integral and thus
the magnetic state and properties correspond directly to the ratio of interatomic distance a to
the radius r as known from the Bethe-Slater formalism.68 Therefore the lattice parameter can
be used to vary a material’s magnetic properties. In bulk material the lattice parameters of the
crystal structure can only slightly be varied. However, thin films can be grown epitaxially
where the film orientation is defined by the substrate, achieving a nearly single-crystal like
quality. Further the in-plane lattice parameter can be adapted by the growing film from the
substrate even if its equilibrium lattice parameter is different. When a thin film adapts the
lattice parameters of a substrate by straining itself, although the equilibrium parameter differs
significantly, strained coherent film growth occurs. Coherent film growth with strains in the
range of some percent is usually restricted to very thin layers (∼5 nm), since a large amount of
elastic energy is required to avoid relaxation of the strained lattice. For soft materials this
amount of elastic energy is reduced allowing to coherently grow strained films up to high
thicknesses.69 Therefore materials that show lattice instabilities are well suited regarding
22
strained coherent growth. SMAs belong to this class of materials showing lattice instabilities
like a softening of the elastic constants near the martensitic transformation.
Figure 12: a) The shift of Fe70Pd30 (002) peak determined by XRD indicates the increase of the out-of-plane
lattice constant of the unit cell and thus the tetragonal distortion. Both the bcc and the fcc structure defining the
boundaries of the Bain path are depicted by dotted lines. b) The different stages of tetragonal distortion c/abct
along the Bain path between bcc (bottom: “red atoms”) and fcc (top: “dark blue atoms”). c) Variation of the
Fe70Pd30 unit cell as a function of substrate lattice spacing of the different buffer layers. The in-plane lattice
parameter abct (“open squares”) is identical to the substrate lattice spacing since it follows a straight line. The
c/abct values (“filled squares”) indicate a constant volume of the unit cell over the Bain path.67
A first attempt was reported by Godlevsky and Rabe70 who predicted the possibility to induce
a cubic to tetragonal distortion with c/a ratios from 0.95 to 1.25 in Ni2MnGa. Buschbeck et al.
adapted this idea to the Fe-Pd system and fabricated 50 nm thick, strained grown epitaxial
thin films at the Fe70Pd30 composition. These films were grown on single-crystal MgO
substrates with different buffer layers on top to adjust the in-plane lattice parameter. This
allowed to strain the Fe70Pd30 unit cell over the full range of the Bain path with
1.09 ≤ c/abct ≤ 1.39. Using XRD a shifting of the out-of-plane lattice parameter (shift of (002)
lattice peak indicated by black arrow) evidences the change of the c/abct-ratio on different
buffer layers as depicted in Figure 12 a). To clarify this, the tetragonal distortion of the unit
cell is sketched in Figure 12 b), with a bcc unit cell (bottom: red atoms) at c/abct = 1.09
towards a fcc unit cell (top: blue atoms) with decreasing in-plane lattice parameter. From pole
figure measurements the in plane lattice parameter was determined for all films on different
buffer layers to calculate the c/abct ratio value. In Figure 12 c) both the c/abct ratio and the in-
plane lattice parameter are presented as a function of the substrate lattice spacing (and thus of
the different buffer materials). The in-plane lattice parameter abct (open squares) is identical to
the substrate lattice spacing attesting the strained coherent growth of the film. The c/abct
23
values (filled squares) as determined from in-plane lattice parameter measurements indicate a
constant volume of the unit cell over the full Bain path as marked by the dotted line. This
shows that the volume does not change upon tetragonal distortion. Magnetic investigations
revealed a strong variation of the magnetic properties upon tetragonal distortion of a Fe70Pd30
unit cell. Due to the structural variations the magnetocrystalline anisotropy constants K1 and
K3 change significantly. Also the Curie temperature TC shows an increase to 830 K with
decreasing c/abct. These results indicate that it is possible to adjust the crystallographic
structure (c/abct ratio) for a defined composition in the Fe-Pd system by strained epitaxial
growth. Especially the magnetic properties can be adjusted significantly in dependence on the
c/abct ratio allowing to tailor materials properties.
2.7 Ternary Fe-Pd-X systems
The Fe70Pd30 FSMA system exhibits several advantageous properties like the MFIS, high TC
and JS. Therefore this material is promising for sensor- and actuator-application. When
compared to Ni2MnGa, Fe-Pd is more ductile, has a higher magnetocrystalline anisotropy and
higher saturation polarization. For full implementation into technical devices there are several
properties that have to be tailored and further optimized. Especially an increase of
transformation temperatures and stability range of the fct martensite phase is required. A
promising route to enhance these properties is the alloying of third elements into Fe70Pd30.
This was reported in literature, where several third elements were added to Fe70Pd30 and the
change of properties was investigated. The effects of Co and Ni addition to Fe-Pd on the fcc-
fct transformation temperature and magnetic properties were investigated by Tsuchiya et al.71
who determined slightly increased values for saturation polarization compared to those for
Fe70Pd30. Co addition was reported to shift the fcc-fct transformation temperature slightly
higher, while Ni (isoelectronic to Pd) addition shifts it lower. These results suggest that the
relative stability between fcc and fct phases is affected by the electronic structure. Figure 13
shows the martensitic transformation temperature as a function of the e/a ratio. With
decreasing e/a ratio a shift of the martensitic transformation to higher temperatures is
observed. All ternary Fe-Pd-X with X = Co and Ni compositions can be correlated to a
polynomial function as indicated by a black line. Further investigation on Fe-Pd-Co and Fe-
Pd-Ni performed by Vokoun et al.72 and Sánchez-Alarcos et al.73, however, did not find
similar results. They reported a decrease of the fct-fcc transformation temperature upon
addition of Co into Fe70-XPd30CoX with X=2.02; 3; 4.1; 6.25 at.%. The non-reversible
24
transformation from the fct to bct phase as described in Chapter 2.6 can be decreased to lower
temperatures upon addition of Co.
Figure 13: Martensitic transformation temperature as a function of the e/a ratio. A slight increase of
transformation temperature is observed for Fe68Pd29.9Co3.1.71
A decrease of the magnetostriction values in the fcc austenite phase was also identified due to
Co addition. Fabrication of Fe-Pd nanocrystalline particles with Co additions was reported by
Kovacs et al.74 to increase magnetic anisotropy and reduce formation temperature of the fcc
austenite phase. The addition of Ni into Fe-Pd results in a decrease of transformation
temperature accompanied by a lowered Curie temperature.71 Lin et al. reported an increase of
magnetostriction due to the addition of Ni into Fe70-XPd30NiX with X= 2; 4; 6 and 8 at.% when
compared to Fe70Pd30.75 Further they found, that a doping of Fe-Pd with Ni prevents the
decomposition of the fcc austenite phase when annealed at 673 K for 100 h and thus
stabilizing the parental phase against decomposition.76 An increase of anisotropy in Fe-Pd by
alloying with Pt was envisaged in Stern et al.77, but could not be verified experimentally.
Alloying Fe-Pd with Pt (isoelectronic to Pd) was found to decrease transformation
temperatures.78 Takeuchi et al. reported the combinatorial fabrication of Fe-Pd-Ga thin
films.79 Magnetic measurements were performed for this system and showed results, which
correlated well with the magnetic properties of the known sub-systems Fe-Pd and Fe-Ga.
Structural investigations of aged Fe66Pd30Rh4 were reported by Lin et al. using TEM, and
XRD measurements.80 They revealed a decrease of transformation temperature to T = 170 K
and the existence of a monoclinic intermediate structure between the fcc austenite and the fct
martensite structure. The alloying of Mn into Fe-Pd was performed by Alarcos et al.
motivated by the ferromagnetic-antiferromagnetic transition as described by the Bethe-Slater
25
curve, which can enhance magnetic properties (increased magnetocrystalline anisotropy and
saturation magnetization).81 They found a significant increase of transformation temperatures
for Fe69.4-XPd30.6MnX (x = 0; 1; 2.5 and 5) compositions. All these reports are based on
experimental investigations giving only partial insight how a single element alters the
properties within the Fe-Pd system. For the Fe-Pd-Co and Fe-Pd-Ni system a defined
variation of transformation temperature was correlated to the e/a ratio without describing
significant trends that can be used to systematically control the system’s properties. A
systematic theoretical prediction to enhance the Fe-Pd alloy was given by Opahle on the basis
of DFT calculations.82 In addition to their important correlation of the formation of the fct
phase in the Fe-Pd system to the band-Jahn-Teller effect, the authors pointed out that
increasing the minority spin density of states at the Fermi level should destabilise the fcc
austenite. This is regarded as the origin for the formation of the fct martensite at higher
temperatures. Therefore they suggested to optimize the Fe content in order to systematically
shift the martensitic transformation temperature. This is also observed from experimental
results by various groups, where a shift of the martensitic transformation can be directly
linked to the Fe content in Fe-Pd-X with X = Co, Ni and Pt.
2.8 Routes for the development of novel Fe-Pd-X alloys
One of the aims of this thesis is to investigate how different fabrication methodologies and
sample designs affect the materials properties of the Fe-Pd system. The second aim is the
development of novel Fe-Pd-X FSMAs with improved properties. These properties are
defined by benchmark values in the following of this thesis.
As presented in the previous chapter, there is only sparse information on ternary Fe70Pd30
based FSMAs. There were less than 25 papers published within that field of research so far.
Thus, the subject of this thesis is a systematic approach to clarify how the properties of binary
Fe70Pd30 can be improved by the addition of third elements. Further a comprehensive
understanding will be provided in order to allow for a tailoring of properties regarding future
application. A fundamental prerequisite to enhance Fe70Pd30 is therefore to define which
criteria are important and which benchmark values have to be reached:
a) a high martensitic transition temperature Ms > 350 K
b) a reversible martensitic transformation ∆T < 5 K
26
c) a high saturation polarization JS > 1 T
d) a high magnetocrystalline anisotropy at RT K1 > 100 kJ/m3
e) a high Curie temperature TC > 550 K (TC > MS)
f) highly movable twin boundaries (MFIS of several %)
g) a high blocking stress (σbl > 10 MPa)
A high martensitic start temperature (Ms) for Fe-Pd-X alloys is important to assure the
occurrence of the martensite phase that exhibits the MFIS, around operating temperature.
Many applications require a thermal stability of the implemented functional material in a
temperature regime up to 350 K, as for example required in the automotive sector. The
thermal hysteresis ∆T needs to be small in order to allow for a fast switching without
additional energy consumption (by varying the temperature) between austenite and martensite
phase. The saturation polarization JS contributes to the overall energy product of an alloy that
limits the maximum energy transduction from magnetic to mechanical energy output. This
value needs to be increased because it allows the use of low external magnetic fields to obtain
the maximum energy input at the anisotropy field HA = 2 K/JS and thus enhancing its
efficiency. As an intrinsic material property, the anisotropy constant K represents the
maximum energy density which can be supplied by an external magnetic field.83 The Curie
temperature TC defines the upper limit where the alloy is ferromagnetic. This value should not
interfere with the martensitic transformation and thus needs to be significantly higher than
MS. Further the ferromagnetic to paramagnetic transition has to be shifted to high values away
from operating temperature. This is related to continuous decrease of magnetization with
temperature that breaks down when the Curie point TC is reached. This lowers JS and the
maximum energy output significantly and thus leads to a decrease of efficiency around the
operating temperature regime. The maximum achievable strain by Fe-Pd-based FSMAs is
another important criterion to be fulfilled. Conventional SMAs exhibit high mechanical
strains but cannot be operated at high frequencies, since these are limited by the material’s
ability to loose heat. In contrast to this, high-frequency actuation is performed by materials
with rather low strains like piezoelectric materials. FSMA could combine these two important
criteria by allowing for high actuation frequencies (up to the kHz regime) and high strains at
the same time and therefore are of high interest for many actuation applications. An increase
of the maximum MFIS in Fe-Pd-based alloys from 3.5% to higher values, as known for
Ni2MnGa, is therefore required. The blocking stress σbl defines the force per area unit below
27
which any twin boundary movement is blocked and presents the upper limit of mechanical
actuation load.
Regarding these benchmarks, a fundamental question arises, asking which elements should be
added into Fe-Pd to enhance the alloy and fulfil these prerequisites. The elements Ni and Co
are ferromagnetic around room temperature but do not enhance significantly the properties, as
reported in literature. A promising candidate to enhance intrinsic properties can be achieved
by altering the magnetic coupling and thus lattice parameters in terms of the amount of
tetragonal distortion (c/a ratio) in the fct phase. When adding antiferromagnetic elements into
Fe-Pd, a modification of the lattice by altering the exchange interaction can be achieved. As
mentioned before, the magnetic exchange interaction depends strongly on interatomic
distances. Magnetic inhomogeneities introduced in this way can alter the magnetism in
FSMAs and affect the lattice parameters. This influences the martensitic transformation and
thus can be used to significantly adjust the FSMA properties. Magnetic excitations can couple
to the system at finite temperatures, when spins of Fe and antiferromagnetic elements flip
against their preferred orientation at T = 0 K. Such effects were reported previously for
related systems, where magnetic inhomogeneities caused antiparallel spin ordering.84 A
promising candidate element to improve the properties of Fe70Pd30 is elemental Mn, which
can couple ferromagnetically as well as antiferromagnetically with Fe. Antiferromagnetic Mn
has a higher orbital moment when compared to Fe and could further increase the overall
orbital moment when added into Fe-Pd. Nevertheless, a general correlation of martensitic
transformation and composition within ternary Fe-Pd-X alloys was not identified up to now.
Several publications report a correlation of transformation temperatures with e/a ratio. This
was shown for Fe-Pd-Ni and Fe-Pd-Co alloys but did not reveal conclusive results for other
Fe-Pd-based alloys like Fe-Pd-Mn, Fe-Pd-Cu. Fe-Pd-Rh, Fe-Pd-Ga and Fe-Pd-Pt. To prove
that the e/a ratio is the decisive factor controlling the transformation temperatures, the
addition of Cu is promising. Cu is in a fcc structure under ambient conditions and allows to
increase the e/a ratio while keeping for example the Fe content constant and thus varying one
parameter without changing the other.
Regarding all these facts, this thesis focuses on investigating the impact of Mn and Cu as third
elements added into Fe-Pd in order to enhance intrinsic properties.
28
2.9 Combinatorial materials science
Optimization of materials based on prior knowledge is a process that was performed since the
beginning of using tools and arms in mankind history. Due to technical and scientific
restrictions, the optimization of materials was a sequential and therefore tedious procedure.
Nowadays, rational planning and automation allow to accelerate novel developments in
materials science. This route focuses on detailed knowledge regarding relation of intrinsic
properties of materials to a set of their performance properties. Usually this knowledge was
obtained from experimental and simulation data. The first use of combinatorial techniques for
solid state materials can be ascribed to Boettcher et al.85 in 1955, while Kennedy et al.86
developed the first thin film composition spreads using co-evaporation in the late 1960’s. In
1970 Hanak introduced an integrated materials-development workflow that defines four key
aspects including a) a complete compositional mapping of a multicomponent system in one
experiment, b) simple rapid non-destructive compositional analysis, c) testing of properties by
a scanning device and d) computer data processing.87 This workflow can be considered as a
first attempt to establish a combinatorial approach for the development of new materials.
Xiang and Schultz reinitiated combinatorial routes in materials science in 1995.88 In the
combinatorial approach multiple samples with different compositions are synthesized and
investigated by high-throughput characterization methods. This was successfully performed to
discover new and to enhance known functional89, magnetic90, catalytic91 and optical92
materials, as well as polymers93. The advantage of combinatorial materials science can be
understood when calculating the number of possible Fe-Pd-X-Y systems. There might be a
possible number of 15 elements (alkaline earth metals, transition metals, lanthanides, non-
metals and metalloids) that could be promising to be added into the binary Fe70Pd30 system to
enhance the materials properties. This yields a number of 15 different Fe-Pd-X ternary
systems, each consisting of 5050 samples with different compositions when using a
compositional increment of 1 at.% per sample. Extending this to quaternary Fe-Pd-X-Y
systems the number of different systems increases to 105 with more than 500.000 samples per
system. This can be further extended when considering not only the composition of the
starting materials but also the different conditions for processing. For a simple catalytic
material the number of experimental runs is rapidly increased to several millions (see Figure
14).94
29
Figure 14: Examples of diversity of materials compositions, process factors, and operation conditions applicable
for combinatorial screening. A) Dependence of the number of possible systems on the number of components.
Red: systems investigated up to now.95 B) Factors and their levels for one-step synthesis of diphenylcarbonate.96
C) Hyperspace of features of materials and measurements in chemical sensors.97 (Figure from Ref. 94)
All these considerations indicate the need for a highly automated approach to effectively find
new and enhance known materials by fabricating materials libraries within one single process.
These libraries have to be investigated by high-throughput methods in order to rapidly
characterize materials properties of interest. Correlating these properties a deeper
understanding for the structure-property-relationship in materials are aimed to be achieved in
a significantly reduced time frame.
In general, there are different types of materials libraries: In discrete libraries every sample is
locally separated and has an individual composition. In continuous libraries the samples are
not separated and a compositional gradient occurs over the whole substrate and appears even
over the sample. Diffusion couples can be regarded as a predecessor of continuous libraries in
bulk materials and were introduced by Zhanpeng98, Miller99 and Goldfarb100. Thin film
composition spreads were fabricated by co-deposition from different sources like
evaporation86 and sputter-deposition.99 Further multilayer deposition101 of wedge-type layers
was performed to induce a thickness and thus a compositional gradient across a substrate.
30
After fabrication, automated high-throughput characterization is usually applied to gain
information about the materials library. Properties to analyse are composition of the samples
by energy- (EDX) and wavelength- dispersive X-ray analysis (WDX), Rutherford backscatter
diffraction (RBS), nuclear reaction analysis (NRA) and Auger-electron spectroscopy. The
structure can be determined by X-ray diffraction (XRD) and electron backscatter diffraction
(EBSD). Further magnetic properties are revealed by scanning superconducting quantum
interference device (SQUID) and magneto-optical Kerr effect (MOKE) measurements.
Mechanical properties can be investigated by using nanoindenter mappings. Next to these
common methods there are specialized techniques like optical methods for Hydrogen storage
materials102, photoluminescence of materials92, Seebeck microprobe103 and scanning-tip
microwave near-field microscope104. Further possibilities for high-throughput characterization
and processing of thin film materials libraries are devices fabricated by micro-electro-
mechanical system routes. In general these devices can be separated into active and passive
devices. Passive devices are shadow mask structures used during thin film deposition
processes and cantilever substrates90,105 that allow measuring a thin film’s stress states.
Further devices are micro-tensile106,107 and bulge test108,109 structures to determine mechanical
properties in thin film samples. Active devices have incorporated sensor or actuator structures
that allow for a processing of thin film samples deposited onto. Arrays of micro-hotplates110
can be used to in situ characterize a thin film sample under varying annealing parameters
(heating and cooling rates up to 104 K/s) way beyond what can be realized with a
conventional apparatus. Nanocaloric devices allow to determine transformation enthalpies of
nano-sized films in situ during annealing.111
After characterization, the materials properties have to be correlated in order to allow for an
understanding of structure-property relationship. A first attempt was reported by Hanak87,
who used computer processing in tabular, graphical and functional forms as part of the
“multi-sample-concept”. But his attempt lacked of resource capabilities, since computers were
not able to handle huge amounts of data in those days. Today data analysis is performed by
using dedicated software with various capabilities for cross-referencing and correlation
functions. Typical data correlation and mining software is developed by numerical computing
environments like MATLAB, ORIGIN and LabVIEW. The main application for such
software lies in the area of multi-property correlation, pattern recognition and clustering
analysis to identify areas of similarity. As an example, the XRDSuite package112 software
developed in the group of Prof. I. Takeuchi are here mentioned. This software allows
correlating spectral data with several integer data sets for many thousands of samples. Further
31
cluster analysis and principal component analysis (PCA) are implemented within to facilitate
the laborious analysis of analyzing large sets of spectral data, by sorting patterns by their
similarity into discrete groups and subsequently deducing the representative basis X-ray
patterns.113
2.10 Thin film nucleation and growth
In a thin film deposition process atomic, molecular or ionic species are generated and
transported from a source to a substrate through a medium. In a sputter deposition process this
medium is the gas phase from which the atomic particles condensate on the substrate. When
the atomic species have a sufficiently small kinetic energy, they physiosorb onto the substrate
surface in the moment of condensation. Since the atomic species are not in equilibrium with
each other, they diffuse on the surface until they interact and form clusters. These clusters
decay directly or continue to grow until reaching a critical radius where they are
thermodynamically stable. These thermodynamically stable clusters are then defined as
nuclei, from which the thin film growth starts. The critical radius rcr is defined by:114
vcr G
2r
∆γ−= (8)
The specific surface free energy is γ and the volume free energy is denoted by ∆Gv. After
formation the film growth starts at these nucleation sites.
Figure 15: Schematic of film nucleation during thin film deposition. Depicted are atoms depositing out of the
plasma onto a substrate. Surface diffusion allows the atoms to conglomerate and to form nuclei. Blue arrows
indicate the energies involved in the deposition and conglomeration process. (Adapted from Ref. 115)
Young’s equation can be used to describe the different types of film growth in dependence on
surface energy γsv of a solid matter, surface tension energy γfv of a film nucleus droplet of the
32
deposited material, interface energy γfs of this droplet and angle θ between solid matter and
droplet (Figure 15):116
fv
fssv)cos(γ
γ−γ=θ (9)
According to this equation, there is a restricted number of angles that can appear. In
dependence on the angle θ between solid matter and droplet of deposition material there are
different types of film growth. When θ > 0° and γsv < γfs + γfv the depositing atoms can diffuse
over the substrate and conglomerate to three-dimensional islands. With increasing number of
arriving atoms these islands grow and new islands are formed. In the initial state the substrate
is not fully covered. With increasing thickness the islands start to coalesce and a continuous
film is formed. This kind of growth is defined as Volmer-Weber growth117 as depicted in
Figure 16 a). When θ ≈ 0° several monolayers are formed before island growth occurs.
Figure 16: Different kinds of thin film growth. a) Volmer-Weber growth is defined by formation of islands that
grow in size (from 1 to 4) until they coalesce. b) Stranski-Krastanov growth forms monolayers (1 to 2) before
island growth starts (3). These islands increase in size with ongoing deposition (3 to 4). c) The formation of
monolayer after monolayer (1 to 4), each covering the full substrate surface is described by Frank-van-der-
Merwe growth. This figure is adapted from Ref. 116.
This growth mechanism is named after Stranski-Krastanov118 and is depicted in Figure 16 b).
When the angle between surface and film nucleus is θ = 0° the film grows monolayer per
monolayer on the substrate fully covering the surface before a further layer is growing (see
Figure 16 c)). This case is defined as Frank-van-der-Merwe growth119 where γsv ≥ γfs + γfv is
fulfilled. Next to the angle between film nucleus and solid matter, saturation of the vapour
phase affects the film growth further. When the vapour phase is saturated with atoms/atomic
clusters over a certain level, layer growth is observed, since the amount of Gibb’s free energy
to form a two-dimensional nucleus is lower than for a three-dimensional nucleus. Stranski-
Krastanov growth can occur when there is a higher misfit between forming film and substrate,
33
although the interface energy would prefer a layer growth. In this case a first atomic layer is
growing which adapts to the substrate’s lattice parameter. When the film thickness increases,
elastic energy for adapting the lattice parameter is needed. If the adhesion energy for
conglomeration of the atoms is lower than the amount of elastic energy, three-dimensional
island growth occurs. Further the substrate temperature affects significantly the kind of film
growth and the film morphology. When the temperature is low the diffusion of atoms on the
substrate surface is restricted leading to an increase of small film nuclei. At higher
temperatures the atoms have enough energy to diffuse over long distances of the substrate
surface giving rise for less but larger film nuclei. According to Thornton’s structure zone
model120 a highly defect rich microstructure is formed when the substrate temperature Ts is
below a third of the source material’s melting temperature Tm. This model was developed for
polycrystalline thin film growth at high deposition rates. In this temperature range, shadowing
is the major effect leading to overgrowing of small by larger grains and thus to a high amount
of interfaces and pores. When the substrate temperature during deposition is increased to
0.3 < Ts/Tm < 0.5 surface diffusion of atoms is mainly occurring leading to a columnar grain
growth. At higher temperatures an increase of the grain size is observed.
In most cases thin film growth starts at different locations on the substrate that coalesce with
increasing amount of deposition materials to a closed film with a polycrystalline
microstructure. Thin films often show a preferred orientation/texture of the microstructure.
Many materials with a fcc (bcc) structure show a distinct (111) fibre-texture ((110) fibre-
texture). This is originated by the fact, that it is energetically favoured to have the lattice
planes with the highest number of atoms oriented to the film plane.121 Next to substrate
temperature, also the deposition rate, mutual inter-atomic reactions and the structure of the
substrate affect the film growth. When nucleation sites are forming on a surface having
similar lattice parameters, the coating atoms adapt to the structure of the surface. It is
energetically favoured for the nucleus to adapt to the periodicity of the underlying substrate
because the surface potential energy shows minima at the substrate’s atomic positions. When
the substrate’s lattice parameter is adapted by the film nucleus pseudo-morphed growth
occurs. This kind of growth is usually restricted to very thin films because there is often a
misfit between the substrate and the thin film lattice parameter. This misfit results in an
amount of elastic energy incorporated in the film structure and is often released at higher film
thicknesses by the formation of misfit dislocations.
34
The presence of a defined relationship between crystal structure of substrate and thin film
material is defined as epitaxial growth. This relationship is defined by the lattice planes and
directions that are parallel in film and substrate:
(HKL) Substrate || (hkl)Film ; [UVW] Substrate || [uvw]Film (10)
In dependence on the difference in crystal structure and lattice parameters of substrate and
thin film material there can be a misfit. This misfit depends on the lattice parameter of
substrate dSubstrate and of the thin film dFilm and is defined by:
Substrate
SubstrateFilm
d
ddMF
−= (11)
The difference in crystal structures and the misfit can give rise to a film growth where the unit
cell is rotated at 45° to the substrate unit cell. This can occur when a fcc structure is forming
on top of bcc unit cells.122 Epitaxial growth can be achieved only in a small regime where the
substrate temperature is above a critical temperature and the amount of coating atoms is
sufficiently low. The atoms in a deposition process need a certain amount of energy to diffuse
over the surface and a certain amount of time to adapt to the substrate structure before further
atoms come in.
35
3. Experimental methods
3.1 Fabrication and Processing
The methods used for the fabrication and the processing of all samples, especially the
fabrication of binary and ternary materials libraries performed in different sputter systems, are
explained in the following. A short description of the experimental planning prior to the
deposition of materials libraries is pointed out. Epitaxial growth of thin films is presented in
the next section and followed by the fabrication and processing of bulk samples.
3.1.1 Thin film materials libraries
The materials libraries were fabricated by physical vapour deposition (PVD) using an ultra-
high vacuum combinatorial sputter system (CMS 600/400LIN) from DCA, Finland and a CS
730 S cluster system from Von Ardenne. 4-inch Si (100) wafers (single side polished, prime
quality) with a 1.5 µm thermal SiO2, working as diffusion barrier, were used as substrates for
all materials libraries. After deposition a lift-off process was performed to structure the thin
film materials library into discrete 3 mm x 3 mm samples. This patterning allowed to
distinguish between the different samples and further to define every sample to a specific
location on the Si/SiO2 substrate. The sputter system consists of a load lock, a central
distribution chamber (CDC), a mask storage chamber and two sputter chambers. Within this
work, both sputter chambers were used to fabricate the materials libraries. The sputter
chamber K1 allows for a sequential deposition, while the second chamber K2 enables the
simultaneous deposition from elemental or alloy targets. In both chambers the substrate is
located on a rotatable and height-movable substrate holder with an integrated SiC heater (293
K ≤ T ≤ 1273 K). To pre-clean the substrate and to enhance adhesion of the thin films, a bias
voltage can be applied to the manipulator. The K1 sputter chamber is equipped with six
magnetron cathodes (three working in DC and three in RF mode) on a movable arm (Figure
17).
36
Figure 17: Schematic of the sputter chamber (K1) for sequential deposition of wedge-type layers (drawn by D.
Grochla). The inset depicts three ignited sputter cathodes (the plasma colours indicate different elements to be
deposited). Two wedge-type layers were already deposited in an angle of 120° apart. The movable arm locates
the sputter cathode having green plasma above the Si/SiO2 substrate in order to deposit the next wedge-type
layer.
Target-substrate distance was set to 87.5 mm. A main shutter is placed directly below the
magnetron cathodes to control the deposition time during the process and to avoid
unintentional deposition. Four independently controlled shutters (4D shutters) are located
above the substrate and are placed 90° apart. These shutters can be used to create various
wedge-type layers and shield different parts of the substrate during deposition. To fabricate a
wedge-type layer a main shutter shields the substrate during deposition and retracts slowly
(moving speed between 1 to 5 mm/s) in order to vary the amount of material deposited across
the substrate. Different retraction lengths can be used to vary the area and thickness of the
wedge-type layer over the substrate. This allows controlling the film thickness of every wedge
layer down to zero thickness. The K2 sputter chamber has five 45° tilted magnetron cathodes
(four working in RF and one in DC mode) mounted in a circular geometry (similar to the
geometrical mounting in Figure 18). The focal point of each cathode is pointing to the middle
of the substrate having a target-substrate distance of 185 mm. All materials libraries were
prepared by sputtering from elemental targets (100 mm diameter, purity 99.99%) that were
pre-cleaned by sputtering prior to deposition in order to remove surface layers on the targets.
The typical base pressure before deposition in both chambers was lower than 4.8·10-5 Pa. As
deposition-gas Ar (6N) at a pressure of 0.67 Pa and under a flow of 20 sccm was used. All
37
depositions in K1 and K2 were started by fabricating a homogeneous Fe layer (d ≈ 10 nm) to
enhance adhesion between substrate and thin films and were ended by depositing a
homogeneous Pd film (d ≈ 10 nm) to avoid oxidation in air.
Figure 18: Schematic of the confocal mounting of the sputter cathodes used for fabrication of materials libraries
(geometrical mounting similar to K2) by co-deposition (drawn by D. Grochla). The sputter cathodes are
confocally angled towards the substrate. The inset depicts three different cathodes during the co-deposition
process (the plasma colours are correlated to different elements to be deposited). The slopes of the elemental
wedges are fixed by the angle of the different cathodes. The angle between the elemental wedges is fixed due to
the position of the cathodes relative to the substrate.
The film thickness for all samples was set between 100 nm to 700 nm. Next to depositions
carried out in the DCA sputtering system, binary Fe-Pd materials libraries were fabricated
from a Fe70Pd30 alloy target (purity 99.95%) using a CS 730 S Von Ardenne cluster system.
This system has two sputtering chambers each equipped with four sputtering cathodes for
targets with 75 mm and 150 mm in diameter. The target to substrate distance during
deposition was around 50 mm. Further a load lock, a distribution chamber and a high-
frequency etching chamber is attached to the sputtering system. The typical base pressure
before deposition in both chambers was lower than 6·10-5 Pa. As deposition-gas Ar (6N) at a
pressure of 0.1 Pa and under a flow of 25 sccm was used. An aperture located below the
sputtering cathodes was used to protect the substrate from unintended deposition. Due to the
variation in plasma density below the target during sputter deposition, the composition of
deposited thin films differed from the nominal target composition across the substrate.
38
Binary composition spreads were fabricated in both sputter chambers (K1 and K2). To
fabricate a binary composition spread the amount of each element has to be increased
continuously over the diameter of the substrate in a wedge-like fashion. Performing this for
two different elements having an angle of 180° apart from each other, a gradual change in
composition and thus a binary composition spread is fabricated. After deposition of the first
wedge the substrate was rotated 180° and a wedge from the second element was-deposited
opposing to the first wedge (Figure 19 a)). Both wedges have a thickness < 15 nm at the
thicker end to allow to form homogenous thin films by subsequent adequate annealing. This
procedure was repeated until the desired film thickness was achieved. Binary spreads by co-
deposition were fabricated in K2 where simultaneous deposition from two elemental targets
was performed. By changing the deposition-rates, the area covered in the binary composition
diagram, is varied. The overall thickness in this process is controlled by deposition time and
sputter rate. To fabricate binary thin film materials libraries with small variation in
composition, a Fe70Pd30 alloy target was used. Sputter deposition was performed at ambient
temperature (293 K) without heating the substrate, although the temperature increases by a
maximum value of ∆T = 10 K due to the sputtern deposition process. The deposition process
was carried out under static conditions without any relative movements between target and
substrate. An aperture was used to shield the substrate during ignition and was retracted when
the target reached working condition. The deposition rates were used to control the final
thickness of the materials library by deposition time.
Figure 19: a) Fabrication of a binary materials library spread by two opposing wedge-type thin film layers. Next
to the region, where both wedge-type layers are overlapping (green: A-B), there are additional pure elemental
regions located at the upper (light green: B) and lower part (carmine red: A) of the substrate. b) Alignment of the
three different wedge-type layers to fabricate a full ternary materials library. The colour indicates the different
deposition materials. All wedge-type layers are oriented to each other by an angle of 120°. The Si wafer
substrate is sketched by black lines and shows the areas of overlapping wedges. The triangular shaped area
depicted by A-B-C defines the region where the ternary materials library is located. The edge regions (A-B, A-C
and C-B) depict locations on the wafer where only binary compositions are deposited.
39
Ternary composition spreads were fabricated by depositing three wedges of different
elements. In K1 the deposited wedge-type layers have an angle of 120° to each other. After
fabrication of a 10 nm thick homogenious Fe adhesion layer, the first wedge was-deposited by
retracting the main shutter over a defined distance (70 mm) across the substrate.
After this, the substrate was rotated 120° and the next wedge was deposited. Another rotation
of the substrate at 120° allowed to fabricate the last wedge. This procedure generates a full
ternary materials library including all binary compositions as depicted in Figure 19 b). To
fabricate thicker films this process was repeated until the final thickness was achieved. It
allows to cover the full area of all ternary compositions by one materials library. In K2
simultaneous deposition from three elemental targets was used to fabricate ternary Fe-Pd-X
materials libraries. Due to the geometric arrangement of the sputter cathodes in K2, the
elemental wedges cannot be aligned at 120°. Nevertheless, this method was used to fabricate
materials libraries covering the region of interest in ternary composition diagrams.
Figure 20: Ternary composition diagram showing the difference between calculated (blue triangles) and
experimentally determined (black squares) compositions of a materials library indicating a difference < 5 at.%.
All the presented techniques for preparation are very effective to fabricate materials libraries
consisting of a large number of single samples with different compositions. To control the
area within the ternary composition diagram, the composition in the centre of the substrate
was determined. This only allowed to determine the centre of the area in the composition
diagram, that is covered by the materials library. It did not give any information about the
variation of composition from neighbouring samples (step size of composition) and also did
not allow to determine the compositions at the border of the materials library. Thus only little
40
information about the compositional range, covered by the fabricated materials library, was
known. In order to enhance process control a model was developed to calculate the
composition for each sample of the materials library in advance. Therefore the variation in
thickness for all three elemental thin film wedges was determined and the composition was
calculated by adding the different film thicknesses. Stripes of every element were deposited at
defined deposition parameters (retraction speed of main shutter, cathode power, Ar flow and
pressure) to determine the distribution of film thickness across the substrate. These film
thicknesses as a function of location on the substrate were then used to calculate the
composition for all samples by using the specific molar volume. All calculations were
performed using Origin software from OriginLab Corporation and self-written templates to
speed up and facilitate experimental planning. In Figure 20 the calculated and measured
compositions (energy-dispersive X-ray analysis) for all samples of a materials library are
presented. This shows that the difference in composition between calculation and
measurement is < 5 at.%, allowing to adequately target the compositional region covered by
the materials library.
After deposition, all materials libraries were annealed in a furnace (Schmetz IU 54 1F) under
N2 atmosphere at 80 kPa and then quenched to room temperature in N2 overpressure (40 kPa)
with a cooling rate of approximately 15 K/s to achieve the metastable transforming phase and
to avoid decomposition. After optimizing the annealing process (1073 K < annealing time <
1173 K; 15 min < dwell time < 70 min) to achieve transforming samples in a single fcc phase
state by using binary Fe-Pd samples, the ternary materials libraries were annealed at 1123 K
for 30 min.
3.1.2 Epitaxial thin films
In order to further investigate selected compositions showing enhanced intrinsic properties,
epitaxial growth of Fe-Pd(-X) thin films was performed in close collaboration with S. Fähler
and S. Weiss at the IFW Dresden. Epitaxially grown thin films are advantageous to determine
intrinsic materials properties like the Curie temperature (TC), anisotropic saturation
polarization, magnetocrystalline anisotropy and crystal structure. These properties are hardly
to investigate in polycrystalline thin films. The thin films were deposited in a Bestec sputter
system equipped with four 2-inch and one 4-inch sputter cathode. The base pressure prior
deposition was p < 5⋅10-7 Pa and the thin films were deposited from Fe, Pd and Cu (purity
99.99%) 2-inch and one 4-inch Fe69Pd31 alloy target (purity 99.99%). The deposition was
41
carried out by using Ar (6N) at a pressure of 0.8 Pa and under a flow of 5.8 sccm. In
dependence on targeted composition the applied direct current (DC) power was set to 60 W
for Fe70Pd30, 120 W for Fe, 90 W for Pd and 50 W for Cu. The substrate to target distance
was approximately 90 mm. To investigate the variation of intrinsic properties as a function of
different c/a ratios (ratio of the long to the short axis), substrates with different lattice
parameters were used (discussed in detail in Chapter 4.2.3). Therefore epitaxial metallic
buffer layers (Cr (c/a=1.09), Au (c/a=1.12), Rh (c/a=1.33), Ir (c/a=1.39) and Cu (c/a=1.54))
serving as growth surface for the strained growth of Fe-Pd-X films were deposited at 573 K
onto MgO(100) single crystals of 10 mm x 10 mm x 0.15 mm in size. The nominal sequence
for all epitaxial grown thin films is: MgO / 50 nm buffer / 50-1000 nm Fe-Pd-X thin film.
These buffer layers were deposited by sputter deposition from 2-inch targets (Cr, Au and Cu)
as well as pulsed laser deposition (Rh and Ir). To promote epitaxial growth of buffer layers
having a fcc crystal structure (Au, Rh, Ir and Cu) a 5 nm thick Cr layer was-deposited onto
the (100) MgO substrate. Buffer layers of elements having a bcc crystal structure directly
grow epitaxially without the need for any intermediate layer to promote growth.
3.1.3 Bulk samples / Splats
Bulk samples of selected compositions, identified by thin film screening, were fabricated for
further investigation. The fabrication as well as the annealing for all splats was performed by
I. Claussen in the group of Prof. S. Mayr at the University of Göttingen. In order stabilize the
metastable Fe70Pd30 phase in the bulk samples, dedicated fabrication methods are needed. The
splat-quenching or ultra-rapid-quenching (URQ) technique123,124 allows to rapidly solidify
materials from the liquid phase at high cooling rates. Thus it is well suited to fabricate bulk
samples showing the metastable Fe70Pd30 phase after fabrication. The targeted composition
was weighted from elemental slugs of the different elements (purity 99.95%), allowing for a
compositional accuracy of approximately 0.01 at.%. Weighted slugs were then alloyed to
produce small ingots with a mass from 1 g to 2 g. These slugs were located on a water-cooled
Cu plate in an arc melting apparatus with a base pressure of 3⋅10-1 Pa. Prior to the fabrication
process the arc melter chamber was purged six times with Ar (purity 99.998%) to prevent the
samples from contamination and oxidation. The source material (slugs) was then melted
under an Ar atmosphere 6⋅105 Pa to fabricate the ingot. For homogenization the ingots were
remelted six times from both sides. Subsequently, the ingot was divided into samples of about
42
0.18 g each, which were remelted by using the arc melter to form small spheres for splat-
quenching.
An URQ apparatus from Edmund Bühler GmbH was used to fabricate the bulk
samples/splats. The small spheres were located on a ceramic sample holder within the middle
of a high-frequency coil in this apparatus. The process chamber was pumped and purged by
Ar (purity 99.998%) six times to avoid contamination and a process pressure of 5⋅105 Pa Ar
was set. Then a high-frequency current was applied to the coil leading to an inductive melting
and a levitation of the liquefied sphere. When the material was completely melted the
electrical current was switched off and the liquefied sphere fell down.
Figure 21: Schematic of the splat-quenching process. a) A high-frequency coil inductively generates circling
currents leading to levitation and melting of a sphere. b) When the sphere is fully liquefied (yellow depicts a
glowing sphere) the current is switched off and the falling droplet is splatted between two Cu pistons creating a
thin foil sample (adapted from Bühler GmbH125).
On this way the sphere passed two photo sensors that activated two water-cooled Cu pistons
to splat the liquefied sphere (Figure 21). The resultant splats have diameters between 20 and
30 mm and thicknesses between 50 to 60 µm and are compositionally isotropic within the
resolution of the energy-dispersive X-ray analysis. Due to the high cooling rates in the range
of 106 K/s the microstructure of the splats shows a high defect-density and thus internal
stresses. To heal out these defects and to reduce the internal stresses, selected compositions
were annealed subsequently. Selected splats were encapsulated into quartz tubes under an Ar
(purity 99.998%) atmosphere of 6⋅105 Pa and annealed at temperatures between 873 and
1123 K for different dwell times ranging from 15 min up to 1 h. For rapid quenching the
quartz tubes were dropped into water after annealing. Slower cooling rates were performed by
43
cooling the quartz tubes at ambient air (≈ 0.5 K/s) or keeping them in the tube furnace
(< 0.2 K/s) until the furnace reached ambient temperature (293 K).
3.2 Characterization
The methodologies to investigate the intrinsic and extrinsic properties of the samples are
presented in the following, which starts with a description of the energy-dispersive X-ray
analysis that was used to measure the composition of all samples. This is followed by Monte
Carlo simulations determining the penetration depth of the electron beam into a thin film
sample. Structural investigation methods are explained in the next subsection explaining the
different diffraction techniques and the importance of measurement parameters for the quality
of results. The high-throughput investigation technique to determine samples undergoing a
martensitic transformation is shown in the following subsection. Magnetic screening as well
as high resolution techniques are presented in the subsequent section. The last subsection
emphasizes on the determination of mechanical properties by nanoindentation.
3.2.1 Energy-dispersive X-ray analysis (EDX)
In order to investigate the chemical composition, all samples were measured by energy-
dispersive x-ray analysis (EDX). Samples were located inside a scanning-electron-microscope
(Leo 1430 VP or Jeol JSM 5800LV) equipped with an Oxford Instruments Inca EDX system
(polymer window, Si-Li detector with 30 mm2 area, Peltier-cooled). This system allows to
quantitatively determine elements from B to U. All samples were measured at an acceleration
voltage of 20 kV, 600 x magnification (integrating the measurement over an area of 600 µm
by 400 µm) and at a process time = 5 (giving a minimum energy resolution of 132 eV)
resulting in 20-30% dead time (Inca-EDX is a dead time corrected system) during
measurement. In order to provide the highest accuracy the measurement time was set to 60 s,
resulting in spectra with more than 250.000 counts. It has to be mentioned that longer
measurement times do not enhance the accuracy of composition measurement due to a
saturation of the signal-to-noise ratio in the EDX spectra (Bremsstrahlung and
background).126 Prior to every EDX-analysis a Co standard was measured at the same
parameters in order to determine the beam-current and thus correction coefficients for the
later EDX analysis. Fe70Pd30 and Fe60Pd30Mn10 alloy-standards (error < 0.05 at.%,
composition for both was quantified by inductively coupled plasma analysis) were measured
44
before, allowing to calibrate the Inca-EDX system. All this led to a compositional error
< 0.2 at.% for all samples. This procedure was applied to all samples (polycrystalline thin
films, epitaxially grown thin films and bulk samples).
Figure 22: Trajectories of inelastically scattered electrons in a 0.5 µm thin film for different elements ( a) Fe, b)
Pd, c) Mn and d) Cu). The normalized X-ray yield originated from L and K shells in dependence on penetration
depth is depicted (different colours correspond to excited shell).
Monte Carlo simulations were performed by using CASINO software127,128 (simulated was
the inelastic scattering of 1000 electron trajectories until the energy of each trajectory was
below 50 eV due to energy loss by collision using a Mott by Interpolation129,130 physical
model) to ensure a sufficient penetration depth of the electron beam at an acceleration voltage
of 20 kV into the sample yielding compositional information through the whole thickness. In
Figure 22 a) to d) the distribution of inelastic scattered electrons in a 0.5 µm thick Fe, Pd, Mn
and film (bulk density values were chosen) are shown. Presented are cross-sections of 0.5 µm
thick thin films on a SiO2 diffusion barrier (1.5 µm) on top of a Si substrate and the
trajectories of electrons (blue) in dependence on depth and lateral dimension. On the left side
in each figure the normalized X-ray yield as a function of depth over the full film thickness is
presented. The different colours correspond to the excited X-rays of the different shells of the
different elements. Due to the elemental mass of Fe, Mn and Cu a high penetration depth of
45
electrons can be observed in Figure 22 a) and b). Here the generation of characteristic X-rays
for elemental analysis occurs over the full thin film thickness including parts of the SiO2/Si
substrate. Heavier elements like Pd decrease this penetration depth significantly. For a Pd film
(Figure 22 b)) X-rays are mainly generated from a depth of about 0.5 µm, while a small
fraction of electrons fully penetrates the Pd film. These results confirm a sufficient
information depth for EDX analysis. Due to the fact that Fe-Pd-based alloys are investigated
in this work, the penetration depth of electrons and the generation of characteristic X-rays is
sufficiently high to quantitatively measure the composition of Fe-Pd-X (X = Mn and Cu)
films.
3.2.2 Structural analysis by X-ray diffraction
The crystallographic structure of thin film and bulk samples was investigated by X-ray
diffraction (XRD) in Bragg-Brentano-geometry using a Bruker-AXS system (GADDS area
detector, Cu-Kα radiation, 0.01° resolution, spot size 0.5 mm, integration time 300 s), a
PANalytical X’Pert PRO MPD (Pixel detector and Szintillation counter, Cu Kα and Co Kα
radiation, 5 mm mask, programmed divergence slit on both incident and diffracted beam path,
0.04 rad soller slit, step size 0.013° for phase analysis and 0.0065° for lattice parameter
determination). Pole-figures were measured in a four-circle set-up with Cu Kα radiation using
a Phillips X’Pert system. Calibration of the systems was performed on an Al2O3 standard
(Bruker) and a Si standard (PANalytical). Temperature-dependent X-ray diffraction, XRD(T),
was used to verify and analyze martensitic transformations and transition temperatures. For
the Bruker system XRD(T) measurements were conducted using an evacuated Be dome to
prevent ice formation on the samples (chamber pressure < 1 Pa). The temperature was varied
in the range from 253 K to 373 K (10 K steps) and measured using a NiCr-Ni thermocouple
(error < 0.5 K). For the PANalytical system XRD(T) was conducted by using an Anton Paar
TTK 450 (temperature range: 120 K to 723 K at a chamber pressure < 0.15 Pa). Further
structural investigation on selected samples was performed using synchrotron-based X-ray
microdiffraction at the 2-BM beam line at the Advanced Photon Source at Argonne National
Laboratory (photon energy 15 keV, image-plate detector MAR 345, beam size 15 µm x 15
µm refined by a set of 30 Be compound refractive lenses). To observe the martensitic
transformation the samples were measured between 300 K and 450 K for 20 s per spot to
obtain sufficient diffracted intensity. Prior to the measurement the system was calibrated
46
using a LaB6 standard (National Institute of Standards and Technology). The Fit2d
software131 was used to extract the lattice parameters from integrated diffraction patterns.
In order to determine lattice parameters the d-spacing was determined by fitting of diffraction
peaks using a Cauchy function and then calculated by using Bragg´s law.132 The coherence
length of X-rays in the film was investigated by means of a modified Scherrer formula.133 The
full-width-at-half-maximum (FWHM) of a diffraction peak was determined by fitting with a
Lorentz function. Micro-stress analysis was performed using the FWHM of a Gaussian
function fitted to the diffraction peak according to Ref. 134. In order to determine the residual
stress in the thin film samples the sin2(ψ) method134,135 was used. The lattice parameter was
determined by a Pearson VII peak fitting136 and then plotted as a function of sin2(ψ). From the
slope of the d = F(sin2(ψ)) function the stress state and values were determined.137,138
Phase analysis and visualization of the diffraction data for the materials libraries was
conducted by using the MATLAB-based “XRDsuite” software package.139 To determine
samples with similar structural phases cluster analysis140 and non-negative matrix
factorization141 were used to reduce the complexity of the data. The identification of the
different phases was performed by using several databases like inorganic crystal structures
database (ICSD), international centre for diffraction data (ICDD), Pauling Files binaries and
Pearson's Crystal database. The “CaRIne” software package was used in order to calculate
XRD patterns of crystal structures predicted by density functional theory (DFT) simulations.
3.2.3 Microstructural analysis by Transmissionen-Electron-Microscopy
For further crystallographic investigations the most promising samples were selected to be
investigated by means of transmission electron microscopy (TEM). A Zeiss Libra 200
(CRISP) TEM was used for high resolution imaging and electron energy loss spectroscopy
analysis (EELS). Further a FEI Tecnai F20 G2 TEM, equipped with an EDX (EDAX) system,
a high-angle annular dark field (HAADF) detector and a scanning transmission electron
microscopy (STEM) unit was used. Cross-sectional TEM-lamellae were prepared by a 5 kV
Ga-Ion focused ion beam (Zeiss XB1540 Workstation (FIB)) and afterwards polished with
500 eV Ar-ions using a low voltage Ar-gun (PHI-AES970). Further sample preparation was
accomplished by cutting TEM lamellas out of the sample using an FEI Quanta 3D focused ion
beam system. Compositional depth profiles of selected film samples were determined by a
scanning Auger microscope (PHI AES970) having an error < 1 at.% for compositional
analysis.
47
3.2.4 Temperature-dependent resistance measurements
Temperature-dependent resistance measurements were carried out for the rapid investigation
of thermoelastic transformations (hysteretic behaviour of resistance as a function of
temperature). Other established methods used to investigate structural transformations are
differential scanning calorimetry (DSC)142 and temperature-dependent X-ray diffraction
(XRD(T))143.
Figure 23: a) Overview of the high-throughput test stand for resistivity and magneto-optical Kerr-effect
(MOKE) investigations. b) 4 pin probe head for resistivity screening measuring on a 4-inch Si wafer located on
the hot/cold chuck. The close-up view depicts 4 pins touching a sample of a materials library. c) 4 pin screening
probe head for magneto-resistance measurement between the poles of an electromagnet. d) 5x4 pin probe head
for continuous resistance measurement.
The DSC method requires a tedious sample preparation and a significant amount of mass that
is usually only ensured for films with a thickness in the µm range. XRD(T) can only be
performed on samples with a maximum lateral size of 9 mm x 14 mm. Thus, both methods
are moderately suited for thin film samples and do not allow for a high-throughput
investigation of full materials libraries. The change of electrical resistance in a material is
correlated to its changes in crystal lattice, crystal phases (precipitates) and boundaries (grain
boundaries and lattice imperfections).144 Further this method is very well suited to investigate
thermoelastic transformations in thin films.145,146,147,148 By comparing temperature-dependent
resistance measurement (R(T)) with XRD and DSC measurements it was found, that
transformation temperatures determined by all three methodologies agree for Fe-Pd-based
48
FSMAs.149 The resistance-screening of materials libraries was performed by using a Keithley
Multi Source Meter 2000 at a constant source current of 0.1 A. Therefore the materials library
was placed into a custom built fully automated high-throughput test-stand on a heatable and
coolable wafer chuck (Figure 23).150
A four-point probe head was used to determine the electrical resistance of each sample of the
materials library at ambient temperature. To gain an average value of the sample´s resistance
each sample was measured three times and the arithmetic middle-value was taken. To
determine the appearance of a martensitic transformation in the samples, R(T) measurements
from 170 K to 470 K (heating/cooling rate: 5 K/min) were performed. Generally this
measurement was arranged in a two-step procedure.
Figure 24: R(T) curves are presented for different samples where the heating is indicated by red and the cooling
cycle by blue coloured dots. a) The “linear” R(T) graph shows a metallic behaviour without hints for a
thermoelastic transformation. b) Graph of a screening measurement with a slight S-shape indicating the presence
of a thermoelastic transformation. c) Single mode R(T) measurement shows a hysteretic shape and indicating the
presence of a thermoelastic transformation. Further the transformation temperatures as well as the width of the
thermal hysteresis are depicted.
In the first step the screening for samples undergoing a martensitic transformation was
performed by using the four-point probe, which was automatically positioned to predefined
locations across the wafer in order to record the resistance of each point at a constant
temperature. After measuring the full materials library, the temperature was varied by a step
of 5 K, and the measurement sequence across the wafer was repeated. To identify the
occurrence of a martensitic transformation, individual R(T) curves were plotted for each
sample distinguishing between heating and cooling curve. Besides the well-known metal
behaviour (Figure 24 a)), where the resistance increases linearly with increasing temperature,
also hysteretic shaped R(T) curves were observed for various samples (Figure 24 b)). Selected
samples were further investigated by single-mode measurements using a 5 x 4 pin probe head,
where the sample´s resistance was measured continuously while the temperature was varied
(heating/cooling rate: 5 K/min). The resulting R(T) curves were used to determine the
49
transformation temperatures. Transformation temperatures were determined by applying the
tangent method on the data gained by single-mode measurements (Figure 24 c)). In the
following the abbreviations for martensite-start-: Ms; martensite-finish-: Mf; austenite-start-:
As and austenite-finish-: Af - temperatures were used. The width of thermal hysteresis is
abbreviated by: ∆H.
3.2.5 Magnetic properties, screening and high-resolution measurements
For automated magnetic characterization of thin film materials libraries, a custom-designed
water-cooled electromagnet is integrated into the high-throughput test-stand (Figure 23),
supplying a maximum magnetic flux density of about 0.3 T between its pole shoes. The
magnet is attached to an adjustable frame, so the gap between the materials library and the
pole shoes is kept constant at approximately one millimeter, giving enough space to move the
materials library underneath the magnet. A magneto-optical Kerr-effect (MOKE) system was
used, consisting of a 5 mW multimode laser diode (λ = 670 nm), a polarizer, an analyzer and
a photodiode to measure the intensity of the reflected light. To measure the magnetic field
during the measurement, a Hall sensor positioned between the pole shoes was used. The
measurement signal of the photodiode is read out using a lock-in amplifier to detect small
intensity changes. A special probe head was designed to fit in the gap between the pole shoes,
thus enabling magnetoresistance measurements (resistance is measured while the magnetic
field is cycled). Both MOKE and magnetoresistance measurements can be performed as
temperature-dependent screening measurements across the materials library. This set-up was
used to identify ferromagnetic samples within the materials library. Although no quantitative
magnetization results can be achieved by this method, a qualitatively investigation of the
samples was possible by distinguishing samples that are hard and soft-magnetic.
Quantitative magnetization measurements were conducted using a Physical Property
Measurement System (PPMS, Quantum Design) equipped with a temperature-dependent
vibrating sample magnetometer (VSM) unit. For this measurement, the materials library was
cut into 4.5 mm x 4.5 mm pieces and selected samples were glued to a quartz sample holder.
Saturation polarization JS was extracted from in-plane hysteresis measurement, performed in a
field range from -0.2 T to 0.2 T (2.5 mT/s-1 sweep rate) in the martensite and the austenite
state. In order to determine the saturation polarization the saturated magnetization-signal (m)
was divided by the volume (V) of the sample (JS= µ0MS = m/V) previously determined by
multiplying lateral dimensions and film thickness. For epitaxially grown thin films JS was
50
determined in-plane in <100> and <110> and out-of-plane in <001> direction in order to
determine the magnetocrystalline anisotropy constants. The Curie temperature, TC, was
determined from in-plane temperature-dependent magnetization measurements (stabilizing the
temperature every 10 K and re-calibrating the position every 20 K) in an applied magnetic
field (near the saturation polarization JS) of the sample) using Kuz’min’s fit.151,152
3.2.6 Mechanical properties investigated by nanoindentation
Investigation of the mechanical properties (Young´s modulus and hardness) was performed
by using a MTS Nanoindenter XP equipped with a Berkovich indenter. The measurements
were performed at ambient and elevated temperatures (353 K) using a heat stage. Single
samples were located on a stage inside the nanoindenter and each sample was measured up to
49 indents (array of 7 x 7, ∆x = ∆y = 50 µm) per composition for statistical purposes. To
eliminate large translations of the motorized stage and to avoid associated drift effects, a
serpentine positioning of the indents across the samples was conducted. Further the
penetration depth and loading rates (strain rate: 0.05 s-1) were chosen to allow for precise
measurements and reduce substrate influence.153 The continuous stiffness method (CSM) was
applied for all measurements and from the recorded load-displacement (P-h) curves, hardness
and Young’s modulus were determined. 154,155 All P-h curves were previously corrected for
residual zero-point errors and thermal drift. The values for Young’s modulus and hardness
were determined from an indentation depth ranging from 50 nm to 70 nm. The calibration of
the indenter area function (projected contact area versus contact depth) and of the ordinary
sample-size value of the machine compliance was performed by using a fused quartz sample
as reference material.156
51
4. Results and Discussion
Within this chapter the results for binary Fe70Pd30 and ternary Fe-Pd-X samples in different
states are presented, as revealed by using the experimental methods described in Chapter 3.
This starts with section 4.1 where the results from binary Fe70Pd30 FSMAs are presented and
discussed. Section 4.1.1 emphasizes on the martensitic transformation in binary Fe70Pd30
samples having a polycrystalline microstructure. This is followed by investigating the
feasibility of the splat-quenching method to stabilize metastable phases and to develop
transforming Fe-Pd bulk samples. Finally the results for epitaxially grown Fe-Pd thin films,
which were investigated in close collaboration with various partners, are briefly described.
Chapter 4.2 presents the results from the different ternary Fe-Pd-X systems. In section 4.2.1
the Fe-Pd-Mn system is investigated in terms of phase stability and magnetic excitation due to
addition of an antiferromagnetic element. The last section within this chapter presents the
results revealed for the Fe-Pd-Cu system and shows how Cu can significantly improve the
materials properties.
4.1 Binary Fe-Pd Ferromagnetic Shape Memory Alloys
4.1.1 Polycrystalline Fe-Pd thin films
Several of the presented results were measured by H. Brunken as part of his diploma thesis,
which was supervised by the author of this thesis. The deposition of binary polycrystalline
FSMA Fe-Pd films was performed in a first attempt to investigate if thin films in the as-
deposited state undergo a martensitic transformation. Therefore these films have to be in the
transforming austenitic fcc phase after deposition on the Si/SiO2 substrate. Since the
transforming fcc austenite phase is metastable and forms only at temperatures above 1033 K,
rapid cooling is required to avoid decomposition when cooling a sample from T > 1033 K to
ambient temperature (300 K). Such high cooling/quenching rates can be obtained by physical
vapour deposition (PVD) techniques. The atoms and atomic clusters ejected from the target
escape with energies up to 10 to 50 eV, equivalent to a temperature between 104 and 105 K.157
When these highly energetic particles deposit on a substrate at ambient temperature (300 K),
high quenching rates in the range of 104 K/s are achieved. Such high quenching rates in a
52
PVD process allow control of the structural state in thin film samples. These structural states
range from amorphous to nano-crystalline to fully crystalline structures that can be
thermodynamically stable as well as metastable.
Figure 25: State of matter as a function of energy and temperature.157
In Figure 25 the state of matter as a function of energy and temperature is presented. The
location of fixed atoms on lattice sites (deposited thin film) and independently moving
electrons and ions (plasma state) on both scales indicate the significant difference in energy as
well as temperature. The energy of sputtered atoms is about 100 times the energy of
evaporated atoms. This additional energy provides sputtered atoms with a higher surface
mobility on the substrate and thus can facilitate phase formation. To prove that as-deposited
thin films undergo a martensitic transformation, Fe-Pd binary films were deposited from a
Fe70Pd30 alloy target and elemental Fe and Pd targets mounted in a confocal geometry. Co-
deposition as well as deposition from an alloy target allow for homogeneous mixing on the
atomic scale over several millimetres in the as-deposited thin film. Since significant
compositional gradients are absent, long range diffusion is avoided. The absence of long
range diffusion avoids changes in composition at the nanoscale after annealing of the thin film
when compared to the moment of film nucleation. Thus, only crystallization is needed to form
the metastable fcc austenite phase that undergoes a martensitic transformation. In Figure 26
53
the compositional variation across a 4 inch Si/SiO2 substrate of thin films deposited from an
alloy target is shown.
Figure 26: Colour-coded compositional variation of a) Fe and b) Pd content across a 4 inch Si/SiO2 substrate in
thin films, deposited from a Fe70Pd30 alloy target.
Due to small differences in magnetic flux density below the target, provided by permanent
magnets inside the magnetron sputtering cathode, the composition varies over the substrate.158
The Fe content ranges from approximately 69 to 73 at.% and the Pd content from 27 to 31
at.%. A binary Fe-Pd composition spread, fabricated by co-deposition is presented in Figure
27. The Fe (Pd) content varies over a broader range from 33 to 77 at.% (23 to 67 at.%) when
compared to the films deposited from an alloy target.
Figure 27: Binary Fe-Pd composition spread fabricated by co-deposition from elemental targets. a) The
variation in Fe content across a 4 inch Si/SiO2 substrate is depicted by colour coding. b) The compositional
variation of Fe (black circles) and Pd (red squares) is presented as a function of the Y position (starting from the
flat of the Si wafer and following the black arrow with dotted line).
54
The crystal structure forming during the deposition process was examined by XRD for both,
the alloy and co-deposited materials libraries. Both the co-deposited as well as the thin films
deposited from an alloy target, did not reveal the austenitic fcc phase.
Figure 28: XRD linescan of a co-deposited binary Fe-Pd composition spread in the as-deposited state. The
colour coding indicates the diffraction peak height. Around the Fe70Pd30 composition undergoing a martensitic
transformation, a broad (111) peak of the stable Fe50Pd50 phase appears. Further elemental Fe precipitation can
be observed correlated to the (110) α-Fe peak. With decreasing Fe content the elemental Fe precipitation
disappears and the stable fcc Fe50Pd50 phase shifts to lower angles indicating a decrease in lattice parameter a due
to an increased amount of larger Pd atoms at the expense of smaller Fe atoms.
An XRD linescan along the gradient of a co-deposited composition spread in the as-deposited
state is depicted in Figure 28. Presented is the intensity of diffraction peaks by colour coding
as a function of the 2θ angle for different compositions from Fe71Pd29 to Fe24Pd76. Around the
Fe70Pd30 composition that undergoes a martensitic transformation, no clear indication of the
austenitic fcc phase is present. A small broad diffraction peak can be observed around 2θ =
40.5° originating from the (111) fcc Fe50Pd50 lattice plane. Further, there is a small peak at 2θ
= 44.5° that is correlated to elemental α-Fe. Around the Fe50Pd50 composition the stable Fe-
Pd fcc phase is formed and the elemental Fe precipitation disappears. Further it can be seen
that the (111) fcc diffraction peak shifts to lower angles with decreasing Fe content. This
indicates an increase of the (111) lattice plane distance and thus of the lattice parameter, a,due
to an increased amount of larger Pd atoms at the expense of smaller Fe atoms. The absence of
the fcc austenite phase in the as-deposited state can be correlated to the fact that the sputtered
55
atomic clusters do not have sufficient energy for formation of the metastable fcc phase. In a
further experiment co-deposition of Fe-Pd films on a substrate heated to 573 K was
performed. It was investigated if the additional amount of energy, provided by thermal heat
from the Si/SiO2 substrate, would facilitate formation of the metastable fcc phase.
Figure 29: XRD diffraction patterns of the crystal structure formed for co-deposited thin films on a Si/SiO2
substrate heated to 573 K. Colour coding indicates peak intensities of the XRD patterns in dependence on the Fe
content.
The crystal structures formed were examined by XRD linescan measurements and are
depicted in dependence on the Fe content in Figure 29. The intensity of diffraction peaks is
shown by colour coding as a function of the 2θ angle at different compositions from Fe77Pd23
to Fe33Pd67. Around the Fe70Pd30 composition, a broad peak between 2θ = 41.2° (the (111)
peak originates from the stable Fe50Pd50 fcc phase) and 2θ = 41.7° (the (111) peak is
correlated to the fct martensite phase that transforms into the austenite phase upon heating)
appears. This proves that next to the stable Fe50Pd50 fcc phase also small amounts of the
Fe50Pd30 fcc austenite phase are present in the Fe70Pd30 film. Furthermore, there is a small
peak at 2θ = 44.5° that is correlated to elemental α-Fe precipitates. Around the Fe50Pd50
composition the stable Fe-Pd fcc phase is formed and the elemental α-Fe precipitation
disappears. At Fe contents < 62 at.% both the (200) stable Fe-Pd as well as the metastable
Fe70Pd30 fcc peak appear, hinting at a multiphase structure in the thin films. Additionally the
thin film samples show hints of the existence of the Fe-Pd bct phase at Fe contents < 60 at.%
56
as indicated by the (002) bct Fe-Pd peak. The XRD results for the co-deposited binary
composition spread prove that heating the Si/SiO2 substrate during deposition allows to
partially achieve the desired fcc austenite phase within the thin films. Nevertheless the
substrate temperature of 573 K is still well below the existence value for the fcc austenite
phase of 1033 K as determined from the phase diagram. Thus, partial decomposition into the
stable Fe50Pd50 phase and elemental α-Fe precipitates occur within the thin films. Based on
the results, a subsequent heat treatment at T > 1033 K after deposition is mandatory to form
the fcc austenite phase and to achieve thin film samples undergoing a martensitic
transformation. Thus, Fe-Pd materials libraries fabricated from an alloy target and by co-
deposition from elemental targets were annealed after deposition at temperatures between
1073 K < T < 1173 K and for different dwell times of 20 min < t < 60 min followed by a
subsequent quenching of the samples at cooling rates > 15 K/s. Structural analysis of the
samples showed that annealing of thin film samples at 1123 K for 30 min, revealed a fcc
austenite single-phase without any precipitation. Annealing at lower temperatures led to an
increase of the dwell time needed to change the film structure from the stable fct phase to the
metastable fcc austenite phase, and to dissolve the α-Fe precipitates.
Figure 30: XRD linescan of a binary Fe-Pd co-deposited composition spread annealed at 1123 K for 30 min.
Due to the annealing treatment a single fct martensite phase, that should transform into the fcc austenite phase
upon heating, appears at the Fe70Pd30 composition. Due to the distinct diffraction peak having a small FWHM
width, the existence of additional precipitation phases can be ruled out.
57
At higher annealing temperatures, the amorphous SiO2 diffusion barrier crystallized and
diffused into the thin film. Although the dwell time was reduced to form the fcc austenite
phase at T > 1123 K, the thin film sample developed additional Fe-Si phases. Concluding all
these findings, the optimum heat treatment parameters were identified as 1123 K for 30 min.
Thus, all binary Fe-Pd materials libraries were annealed at these parameters in the following.
In Figure 30 a XRD binary line scan of a co-deposited materials library annealed at 1123 K
for 30 min is presented. Around the Fe70Pd30 composition, only one significant peak with a
small FWHM appears at 2θ = 41.7°. This peak is correlated to the (111) lattice of the
metastable fct martensite that transforms into the fcc austenite phase upon heating. The
absence of the (200) and (002) fct martensite peaks at 2θ(200) = 47.3° and 2θ(002) = 50.8° can
be explained by the low signal to noise ratio that covers these weak peaks in the background
of the diffraction spectra. With increasing Fe content, the fct martensite phase changes to the
bcc phase as indicated by the developing (110) bcc peak. When the Fe content is lowered < 69
at.%, the fct phase changes to the fcc austenite phase indicated by the appearance of the (200)
fcc peak. A decrease in transformation temperatures below ambient temperature (293 K) and
with decreasing Fe content corresponds well to the metastable phase diagram presented in
Figure 7. With further decreasing Fe content, the (111) fcc Fe70Pd30 peak broadens and shifts
towards the position of the (111) peak of the stable Fe50Pd50 phase. Between 38 at.% < Fe <
45 at.% the fcc stable phase changes to the Fe-Pd bcc phase as indicated by a splitting of the
(200) fcc into the (110) and the (002) bcc peak. Below 38 at.% of Fe the Fe-Pd fcc phase
appears again.
After verifying that the thin film samples fabricated from an alloy target and by co-deposition
have a transforming fct/fcc single-phase structure without precipitates at ambient temperature,
it was investigated if these samples undergo a martensitic transformation. Therefore
automated temperature-dependent electrical resistance screening measurements were made on
these films, as described in section 3.2.4. Figure 31 shows the electrical resistance values as a
function of temperature for thin films deposited from an alloy target a) and by co-deposition
b). All thin films were annealed at 1123 K for 30 min followed by a rapid quenching to
ambient temperature. The R(T) measurements in both diagrams show an S-shaped curve
where the transformation temperatures are determined using the tangential method. Due to the
vanishing thermal hysteresis in some measurements, one value was defined for Mf and As as
well as Ms and Af. This vanishing can be correlated to the second-order-like character of the
martensitic transformation in Fe70Pd30. It is noted that the martensitic transformation shifts
towards higher temperatures with increasing Fe content. This behaviour follows the
58
metastable phase diagram in Figure 7. Samples with a Fe content outside of 70 at.% < Fe < 73
at.% do not exhibit a martensitic transformation, as indicated by a linear R(T) relationship.
Figure 31: R(T) measurements on thin films deposited from an alloy target a) and by co-deposition b) that were
annealed at 1123 K for 30 min followed by quenching. The R(T) measurements in both diagrams show an S-
shaped curve where the transformation temperatures are determined using the tangential method. Due to the
vanishing thermal hysteresis Mf and As as well as Ms and Af were set equal. The martensitic transformation shifts
towards higher temperatures with increasing Fe content. Samples with a composition outside of the interval of
70 at.% < Fe < 73 at.% do not exhibit a martensitic transformation.
Since the R(T) measurement is an indirect method to determine the presence of a martensitic
transformation, additional XRD(T) measurements were performed to prove if these films
undergo a martensitic transformation.
Figure 32: XRD(T) measurements of a Fe70.8Pd29.2 thin film deposited from an alloy target and annealed at 1123
K for 30 min, followed by quenching. At 248 K the sample is in the fct martensitic phase indicated by the
presence of the (111) fct and the weak (200) fct peak. Upon heating a shift of the (200) fct peak towards higher
angles is observed. At 292 K the (200) fcc peak starts to grow in intensity and is fully developed at 333 K. This
structural behaviour was found to be reversible and thus evidences the occurrence of a martenstic transformation.
59
In Figure 32 the XRD(T) of a Fe70.8Pd29.2 thin film between 248 K < T < 353 K is presented.
At 248 K the sample is in the fct martensite phase, showing a well-developed (111) fct peak at
2θ = 41.7° and a weak (200) fct peak at 2θ = 47.7°. Due to the signal-to-noise ratio, the (002)
fct peak at 2θ = 50.8° disappears in the background of the measurement. Further peaks are
originated by a Be dome that was covering the sample during measurement to prevent ice
formation due to humidity. When the temperature is increased, the (200) fct peak shifts
towards higher 2θ angles. At 292 K the (200) fcc peak at 2θ = 48.8° starts to grow in intensity
until reaching 333 K. This vanishing of the (200) fct peak and the development of the (200)
fcc peak upon heating was found to be reversible when repeated several times. Due to the
absence of any additional peaks and the reversibility of this behaviour, the sample is in a
single-phase state and exhibits a martensitic transformation. The transformation temperatures
determined from XRD(T) correlate well with the results from R(T) measurement. These
findings prove that it is not possible to achieve polycrystalline thin films exhibiting a
martensitic transformation in the as-deposited state fabricated from an alloy target or by co-
deposition. Thus, a heat treatment at 1123 K for 30 min followed by rapid quenching is
required to achieve samples in the transforming fct/fcc phase. Since the films have to be
annealed, also multilayer films can be used to fabricate transforming thin films near the
Fe70Pd30 composition that exhibit a martensitic transformation.
Figure 33: Binary composition spread fabricated using wedge-type multilayer thin films deposited from
elemental Fe and Pd targets. A larger compositional interval was covered by this method, since shutters were
used for the deposition. The Fe (Pd) content varies from 5 at.% to 85 at.% (15 at.% to 95 at.%).
Therefore wedge-type multilayer thin films were fabricated from elemental Fe and Pd targets
as described in Chapter 3.1.1. The thin films were fabricated as binary composition spread
and the variation in composition is depicted in Figure 33. The Fe (Pd) content varies between
5 at.% to 85 at.% (15 at.% to 95 at.%) with an almost linear increase over the wafer as shown
60
in Figure 33 b). Using the multilayer fabrication method a larger compositional variation
along the binary composition spread can be achieved when compared to co-deposited films.
After annealing at 1123 K for 30 min and subsequent quenching, the multilayer thin films
show a similar structural sequence (bct-fct-fcc) with increasing Fe content like the co-
deposited binary composition spread in Figure 30.
Figure 34: a) R(T) screening measurements of films fabricated by multilayer deposition indicates the occurrence
of a martensitic transformation. With increasing Fe content a shift of the martensitic transformation towards
higher temperatures is observed. The Fe75Pd25 thin film does not show any hints of a structural transformation. b)
XRD(T) on the Fe71Pd29 thin film proves that this sample exhibits a martensitic transformation. The
transformation temperatures determined by R(T) and XRD(T) correlate well.
Thin films having a transforming fct/fcc structure were further investigated by R(T) and
XRD(T) to determine transformation temperatures. R(T) measurements presented in Figure
34 a) indicate the S-shaped slope in electrical resistance. The transformation temperatures
61
were again determined by using the tangential method. With increasing Fe content, a shift of
the martensitic transformation to higher temperatures is observed. The Fe75Pd25 sample only
shows a linear increase of electrical resistance with temperature, indicating the absence of a
structural transformation. XRD(T) was used to further prove if samples that show an S-shaped
R(T) relationship undergo a martensitic transformation. XRD(T) measurements on the
Fe71Pd29 thin film are depicted in Figure 34 b). The structural change of this film is similar to
the thin film deposited from an alloy target. This sample is in the martensite phase at 293 K as
determined from the (111), the (200) and the (002) fct peaks. With increasing temperature,
both the (200) and the (002) fct peaks shift towards the (200) fcc peak position at 2θ = 48.8°.
When the film is around 309 K, a first weak (200) fcc diffraction peak appears at 2θ = 48.8°
and grows with increasing temperature. The transformation temperatures determined by R(T)
and XRD(T) measurements correlate well.
Figure 35: a) An annular dark field photograph of a Fe70Pd30 lamella fabricated by focussed ion beam (FIB)
milling and observed by a transmission electron microscope (TEM) at a magnification of x59.000 is shown. The
dark area on the left of this pictograph depicts the SiO2 on top of the Si/SiO2 substrate while the area on the right
in light grey colours presents the Fe70Pd30 film. A patterned structure on the surface of the lamella is observed
within an area marked by a red ellipse. This pattern is correlated to a martensitic twin structure. High-resolution
TEM was used to magnify the area in the ellipse. b) At high magnification a tweed like surface structure can be
observed that is related to a twinned martensitic structure within this sample (red frame corresponds to the red
ellipse). Selected-area electron diffraction in c) over the area depicted by a white ellipse in b) was used to verify
the martensite structure and to determine the lattice parameters. (FIB milling was performed by A. Sehrbrock
and TEM investigations were carried out by S. Irsen)
62
To gain deeper insight into the microstructure of Fe-Pd thin films undergoing a martensitic
transformation as determined by R(T) and XRD(T) measurements, TEM was used. Further,
TEM investigations were conducted to clarify if the transforming Fe-Pd thin films are in a
single austenite/martensite phase without precipitates and if the elemental Fe and Pd
multilayer dissolve during the annealing process. Therefore a Fe70Pd30 thin film fabricated by
multilayer wedge-type deposition was milled using the focussed ion beam (FIB) technique to
produce a TEM lamella. In Figure 35 a) an annular dark field pictograph of a Fe70Pd30 lamella
fabricated by focussed ion beam (FIB) milling and observed by TEM is presented. The dark
area on the left in this pictograph shows the SiO2 parts of the Si/SiO2 substrate in the lamella.
The area in light grey colours located on the right of this pictograph presents the Fe70Pd30
film. It has to be noted that the microstructure does not show any indications of a multilayer
microstructure nor precipitation. This proves that the thin elemental Fe and Pd layers fully
dissolved during the annealing. When observing the microstructure in a) stripes can be
observed within the area marked by a red ellipse. These features can be interpreted as a
martensitic twin structure in this thin film lamella. High-resolution TEM magnifying the area
inside the ellipse in b) confirms that the tweed-like surface structure corresponds to a nano-
twinned martensitic structure within this sample (red frame corresponds to the red ellipse).
Selected-area electron diffraction in c) over the area depicted by a white ellipse in b) shows
the diffraction pattern of a twinned fct martensite structure observed from the (111) direction.
Both the nanostructure high-resolution pictograph as well as the diffraction pattern are
reported in literature for a twinned martensitic structure.159
Table 1: Fabrication and annealing parameters of all thin film samples investigated within this chapter. The film
thickness varies between 500 nm and 700 nm for films deposited from an alloy target and 450 nm to 700 nm for
the composition spreads.
Magnetic measurements were additionally performed on several samples fabricated by the
different sputtering methods. These measurements revealed that the Curie temperature,
63
saturation polarization and magnetic coercivity are similar to literature values and no obvious
trend was determined in dependence on the fabrication technique.
The fabrication and processing parameters of all thin films investigated within this chapter are
summarized in Table 1. The first and second columns describe the compositional variation of
the samples and the sputter deposition technique. The following two columns present the
deposition parameters and the annealing conditions. The final column shows the film
thickness variation within the samples.
Figure 36: Martensite start temperature Ms as a function of the Fe and in dependence on the sputtering method.
Co-deposited thin films show significant higher transformation temperatures in comparison to films fabricated
from an alloy target or to wedge-type multilayers from elemental targets. The martensitic transformation is
lowest for thin films fabricated as wedge-type multilayers.
When comparing the transformation temperatures of all single-phase Fe-Pd thin films heat
treated at 1123 K for 30 min followed by quenching, significant differences can be observed.
In Figure 36 the martensite start temperature Ms for three different compositions is presented
as a function of the Fe content and in dependence on the fabrication method. All values for Ms
are considerably higher than reported in literature for bulk samples and single crystals.57 This
is related to stress induced by the difference in thermal expansion coefficient of thin film (in
this case Fe-Pd) and substrate (in this case Si).22,61 Such thin film stress shifts the martensitic
transformation to higher temperatures according to the Clausius-Clapeyron equation as
described in Chapter 2.1. As depicted in Figure 36, also the different methods (co-deposition,
deposition from a Fe70Pd30 alloy target and wedge-type multilayer deposition) do show
differences in Ms at similar Fe contents. The martensitic transformation is shifted to lowest
temperatures for thin films that were fabricated as wedge-type multilayer from elemental
64
targets. The films deposited from a Fe70Pd30 alloy target show a slightly increased Ms while
the co-deposited films have the highest transformation temperatures.
Figure 37: a) Coherence length as a function of the Fe content and for the different fabrication methods (open
black symbols). The correlation coefficient R2, a measure for the peak fitting quality, is depicted for all methods
by red filled symbols. The highest coherence lengths between 70 nm and 80 nm are observed for thin films
fabricated as wedge-type multilayer. Co-deposited and from an alloy target fabricated films show a significantly
decreased coherence lengths between 20 nm and 30 nm. In b) the micro-stress inside the thin film is depicted as
a function of Fe content and fabrication technique (open black symbols). The correlation coefficient again
indicates a high fit quality (filled red symbols). It has to be noted, that the multilayer thin films show the lowest
values of micro-stress.
This can be interpreted in a way that the stress state in thin films is significantly related to the
fabrication methods. To clarify the impact of fabrication method on the thin film
microstructure, the coherence length and the micro-stress inside the thin films was
investigated by analysing the measured XRD patterns. The coherence length as a function of
the Fe content and for the different deposition methods is shown in Figure 37 a) by open
black symbols. The different symbols correspond to the different fabrication methods. Further
65
the correlation coefficient R2 of the Lorentzian peak fitting is presented in red colour and by
filled symbols. All R2 are > 0.98, indicating a high peak fitting quality and thus a small error
for measuring the coherence length. Thin films deposited as multilayer exhibit the highest
values in coherence lengths ranging from 70 nm to 80 nm. Since the coherence length can be
interpreted as a measure for the grain size, it can be concluded that the films fabricated as
wedge-type multilayers have the largest grain size compared to films deposited by the other
techniques. The micro-stress state in these films is again presented as a function of Fe content
in Figure 37 b). The correlation coefficient R2 again ensures a high peak fitting quality and
thus a small error for the micro-stress measurements. The micro-stress inside the co-deposited
thin films is highest among all other fabrication methods, while it is lowest for the
multilayered thin films. Although the micro-stress state differs from the residual stress in thin
films, it can be generally considered as a measure for the residual stress distribution in thin
films.160 All these findings indicate that the microstructure in transforming thin films is highly
affected by the fabrication method, even after annealing. To understand this, the phase
formation during the annealing process has to be examined. Thin films fabricated by co-
deposition and from an alloy target already in the as-deposited state have a homogeneous
mixing of Fe and Pd atoms on the nanometre scale over several millimetres in lateral
dimension. Although nanometre sized grains of different phases already form during
deposition, there are no significant compositional gradients at the nanoscale in the thin film
samples. Thus, long range diffusion is avoided during annealing. This is different in thin films
fabricated as wedge-type multilayers. Due to the multilayer structure of pure elemental Fe and
Pd layers, there are high compositional gradients that accelerate the diffusion process when
the thin films are annealed. Within these elemental Fe and Pd layers, nanometre-sized grains
are forming. When this thin film is annealed Fe-Fe and Pd-Pd bonds are breaking to decrease
the compositional gradient according to Fick’s law161 and new Fe-Pd bonds are forming. The
energy of mixing by breaking elemental bonds and the formation of new heterogeneous bonds
is an exothermic process. This process is known to release energy in terms of heat especially
in multilayered thin films where a self-propagating exothermic reaction is often observed.162
This exothermic process induces an additional amount of thermal energy into the grain
growth during annealing and thus leads to a significantly increased grain size in thin films
fabricated as multilayers. Thin films with larger grain sizes do have less boundary interfaces
than films with small grains. Since the presence of boundary interfaces in thin films strongly
affects and can even increase the residual stress state201, it can be concluded that the co-
deposited thin films have the highest micro-stress and thus the highest transformation
66
temperatures. In contrast to the multilayered thin films where the large grain size and the
reduced amount of interface boundaries provides a reduced micro-stress state.
4.1.2 Bulk / Splat Samples
All presented splat samples in this part were fabricated and annealed by I. Kock from the
Georg-August-Universität Göttingen. Binary Fe70Pd30 bulk samples were fabricated using the
splat-quenching technique as described in section 3.1.3. This method is well-suited to
maintain metastable structures within bulk samples and to avoid segregation into their
equilibrium phases. Especially for the Fe-Pd system, this method is therefore quite promising
to fabricate transforming Fe-Pd FSMAs.
Figure 38: Pictures of an as-splatted sample: a) Front view optical photograph. b) Cross-sectional view
indicating a highly columnar microstructure even in the as-splatted state. (Figure originally published in
Ref.163)
Due to the high quenching rate, the material crystallizes in the desired fcc austenite phase,
which can transform to the martensite fct phase upon cooling. When the liquid droplet is
rapidly solidified by the Cu pistons of the splat-quencher, the material flows from the centre
to the sides giving rise to a round shaped geometry. In general, the splats have a diameter of
20 mm and a thickness of around 60 µm as depicted in Figure 38 a). The droplet starts to
solidify at the coldest location which is the centre of the Cu pistons. This crystallization
process induces a stress gradient over the splat as indicated by a slight bending which is
present after fabrication. Because of the heat sink of the Cu pistons, the microstructure shows
a columnar growth orthogonal to the sample surface as observed by a light microscope in
Figure 38. Induced by the splat-quenching process, there is a large amount of stress and
defects incorporated in the sample, which can pin the martensitic phase boundaries.
Therefore, the samples were annealed at various temperatures (Tan= 873 K, 973 K, 1073 K,
1173 K and 1273 K) for 15 min, in order to heal out defects and improve the transformational
67
behaviour. In Figure 39 the microstructure for three different samples is shown. To
investigate the grain shape and size, trenches were milled into the samples and lamellas were
fabricated using FIB in order to be investigated by SEM and TEM. Figure 39 a) shows an
overview on the trench milled into an as-splatted sample. Both the overview and the close-up
indicate a highly columnar grain shape with a width of about 180 nm. When the sample is
annealed at 1073 K for 15 min the columnar shape of the grains becomes more distinct and
the widths increase to about 440 nm. The inset shows a TEM picture of a lamella where
martensitic twins can be observed as fishbone-like structures on the polished grain surface.
The twin structure indicates the presence of the martensite phase and thus a possible
transforming sample.
Figure 39: a) SEM overview of a trench milled into an as-splatted sample using a FIB (FIB was performed by
H. Brunken). b) Close-up view of the as-splatted sample showing a columnar growth with a column width of
about 180 nm. c) and d) Upon annealing at 1073 K for 15 min the columnar grains can be distinguished better
due to the increased column widths of about 440 nm. The inset shows a TEM lamella of three grains with a
martensitic twin structure on the polished surface. In e) and f) TEM pictures of a splat annealed at 1273 K are
shown. The column width is increased further to about 640 nm. (TEM investigations were performed by D.
König and the figures were originally published in Ref. 163)
When the splat is annealed at higher temperatures of 1273 K for 15 min, the grain widths
increase further (Figure 39 e) and f)) to about 640 nm. In order to investigate if the splats
68
undergo a martensitic transformation and to determine the transformation temperatures, R(T)
measurements were carried out.
Figure 40: a) Normalized R(T) measurements as a function of temperature for splats annealed at different
temperatures. The black dots in the inset define the transformation temperatures determined by using the
tangential method for a splat annealed at 1273 K. b) Temperature-dependent magnetization measurements of all
samples under a constant external magnetic flux density of 0.03 T. In both diagrams the curves are offset for
clarity. (Magnetization measurements were carried out by I. Claussen and both figures were originally published
in Ref. 163)
Figure 40 a) shows R(T) measurements of Fe70Pd30 splats for both the as-splatted and the
annealed state. The as-splatted as well as the samples annealed at 873 K and 973 K show only
a slight non-linearity in the R(T) curve. The hysteretic shape of the R(T) curves is barely
visible due to the low signal to noise ratio. Nevertheless, the transformation temperatures
were determined using the tangential method as depicted in the inset by light grey lines.
Splats annealed at higher temperatures indicate a distinct non-linearity in the R(T) curve. In
contrast to R(T) measurements of conventional SMAs164, these samples show only weak
changes in the R(T) curve with vanishing hysteresis widths. Therefore it is not possible to
distinguish between Ms and Af or Mf and As temperatures. Thus, only one temperature is
determined for the start (Af and Ms) and the end (As and Mf) of the martensitic transformation.
From this graph the transformation temperatures increase with increasing Tan.
To further prove the presence of a martensitic transformation, temperature-dependent
magnetization measurements M(T) at a constant magnetic flux density of 0.03 T were carried
out. In Figure 40 b) the magnetization curves as a function of temperature for the as-splatted
as well as the annealed samples are shown. At temperatures > 291 K the magnetization does
not vary significantly with temperature because the splats are in the austenitic state. When the
temperature is lowered the magnetization starts to decrease rapidly at the beginning of the
martensitic transition. Measurements down to 200 K (not shown here) indicate a continuous
69
decrease of the magnetization. A hysteretic M(T) curve is observed for samples annealed at
873 K, 1073 K, 1173 K and 1273 K. The hysteresis extends to lower temperatures and does
not close even in the cycles that were measured down to 50 K. Since there is no distinct
change in the M(T) curves at low temperatures, it was not possible to determine values for Mf
and As. At a higher temperature regime, Ms and Af were determined by using the tangential
method. Splats annealed at 1173 K and 1273 K show a distinct decrease in magnetization and
a hysteretic M(T) curve progression. The sample annealed at 973 K does not indicate a
transformation, although the one annealed at 873 K clearly transforms. The decrease of
electrical resistance and magnetization with decreasing temperature, as observed for all
samples, can be correlated to a change in band structure and especially in the density of
electronic d-states near the Fermi energy as reported by Opahle et al..66 An alternative
explanation for this decrease of magnetization and resistance is the increase of magnetic
anisotropy energy in the martensite state with decreasing temperature.32 Under a low
magnetic flux density of 0.03 T, the individual magnetic moments rotate out of the direction
of the applied magnetic field and into the direction of the magnetic easy axis when the
magnetic anisotropy energy is sufficiently high. This leads to a demagnetization of the sample
and a change in magnetic domain structure, which influences the electronic scattering at
domain walls.165
To identify the phases appearing within the splats, XRD(T) measurements were performed.
Figure 41 presents the XRD diffraction patterns at different temperatures. At 298 K, the splat
annealed at 1073 K shows the (111) and the (200) fct martensite peaks of the transforming
Fe70Pd30 phase. When the temperature is increased, the (200) peak of the fct phase shifts to
higher 2θ angles and the (200) fcc austenite peak appears and grows in intensity. The
transformation is reversible as confirmed by cycling the temperature during XRD
measurement. The martensitic transformation is observed for all splats by the (200) fct peak
shift and the development of the (200) fcc peak. Even the sample annealed at 973 K shows a
slight shifting of the (200) fct and a development of the (200) fcc peak. Samples in the as-
splatted state and annealed at 873 K and 973 K show next to the (111) fct/fcc peak one broad
peak at the (200) fct and (200) fcc peak positions at low temperatures. With increasing
temperature this broadened peak shifts more and more to the (200) fcc peak position while the
austenitic state appears. As-splatted samples and those annealed at 873 K and 973 K show
additional phases due to a decomposition of the austenite phase into Fe50Pd50 and α-Fe. The
(110) peak of the α-Fe phase (splats annealed at 873 K and 973 K) and a (111) Fe50Pd50 peak
(as-splatted and annealed at 873 K and 973 K) occurs for various samples.
70
Figure 41: XRD(T) diffraction patterns of a) as-splatted and for 15 min annealed samples at b) 873 K, c) 973 K,
d) 1073 K, e) 1173 K and f) 1273 K. The spectra are offset for clarity. (Figure originally published in Ref. 163)
The samples annealed at 873 K and 973 K show a weak (200) fcc peak. This is correlated to
the decreased amount of the Fe70Pd30 phase due to a decomposition of the transforming phase
when annealed at temperatures < 1073 K. Splats annealed at 1073 K and higher show a
distinct (200) fcc peak when they are heated over 279 K. It has to be noted that the position of
the fct peaks are slightly different in comparison to literature values.143 This can be explained
by the temperature dependence of the lattice parameters for Fe70Pd30 reported by Cui.57 At
248 K these splats are in a martensitic state but show a small amount of austenite remainders.
At temperatures above 293 K, the splats annealed at temperatures > 973 K transform into the
austenitic state without any martensite remainders. Splats annealed at ≤ 973 K are in an
intermediate state, where both the fct martensitic and the fcc austenitic structure occur even at
low temperatures (248 K). From the occurrence of the peaks during heating, the
71
transformation temperatures (only Ms/Af and Mf/As), were determined with an error <5 K. In
Figure 43 the transformation temperatures determined by R(T), M(T) and XRD(T) are
presented as a function of the annealing temperature Tan. As described in the text previously,
it was not possible for several measurements to distinguish between Ms and Af or As and Mf
for several measurements. Therefore, Ms/Af and As/M f are combined to one value for R(T)
and XRD(T). As depicted in Figure 43 the As/M f curve shows a steeper increase than the
Af/Ms curve. A line as a guide to the eye was drawn through the experimental data
accentuating an Arrhenius-like increase of the Af/Ms temperatures with increasing Tan.
Figure 42: Transformation temperatures of the Fe70Pd30 splats determined by different measurement methods
and annealed for 15 min at different temperatures Tan. (Figure originally published in Ref. 163)
All three measurement methods indicate distinct changes in crystal structure, resistance and
magnetization of the splats upon phase transformation. The transformation temperatures
determined by different methods match within the error margins and rise with increasing Tan.
Only the sample annealed at 973 K shows significant deviations in all measurements, since
only a weak transformation is observed. The XRD diffraction patterns for this sample show
the presence of Fe50Pd50 and α-Fe precipitates while the EDX analysis of this sample
indicates slight inhomogeneities in the Pd content. This is assumed to be caused by some
irregularities during the splatting or annealing process.
In order to gain an understanding about the increase of transformation temperatures with
increasing Tan, a closer examination of parameters influencing the transformation behaviour is
needed. From literature it is known that the transformation temperatures are mainly
influenced by the Pd content57, the defect density166 and the stress in the sample.22 These
properties depend on compositional homogeneity, quenching rate, grain size and shape, and
72
the presence of precipitates. Moreover, it is necessary to distinguish between factors affecting
the temperature T0 at which austenite and martensite phase are in thermodynamic equilibrium
and other factors shifting the actual transformation temperatures via the necessary
undercooling ∆T = (T0 - Ms) that acts as a driving force. In the present experiments, care was
taken to ensure adequate homogeneity of the Pd content in the splats. However, the possibility
of slight deviations in Pd content (< 0.3 at.%) cannot be ruled out by EDX. Since the
thermodynamic equilibrium temperature T0 strongly depends on Pd content57 (≈ 20 K/at.%),
the transformation temperatures can be shifted by several degrees, independent of any effects
originated by the annealing process. Furthermore, the defect density and distribution in the
samples annealed at different temperatures is not identical. Due to the fabrication process of
splat-quenching, there are differences between inner and outer parts of the splat. A high defect
density provides many nucleation sites for the martensite phase166, leading to high
transformation temperatures. These defects can heal out during the annealing process leading
to a decrease in Ms while Tan increases. On the other hand, defects can hinder the movements
of the phase boundaries through the material. When precipitates have formed during the
annealing process for samples annealed at Tan < 1073 K, the phase boundaries can be pinned.
This increases the necessary undercooling and the material transforms at lower
temperatures.167 Since the As and Mf curve increases much more distinctly with increasing Tan
than the progress of Ms and Af and there is hardly any hysteresis between heating and cooling,
the amount of the martensite phase is assumed to increase by movement of the phase
boundaries rather than by nucleation of new martensite nuclei.
The transformation temperatures can be further affected by the stress state inside the sample.
This is well known as stress-induced martensite.22 However, as the splats are not attached to a
rigid substrate, most of the stress incorporated during the splat-quenching process can relax
via bending or expansion of the sample. But there can also be contributions to the intrinsic
stress by lattice distortions at defects in rapidly solidified polycrystalline samples. The most
obvious property that should show distinct changes in the samples annealed at different Tan is
the grain size as depicted in Figure 39. The dependence of transformation temperatures on
grain size was discussed in several works. A strong increase of the transformation
temperatures with grain size was reported by Seki et al.168 and Kang et al.169. An increase of
grain size with Tan is expected according to R = R0 + α⋅t⋅exp(-Q/kBTan) giving rise to a shift of
transformation temperatures towards higher values.170 Here, the initial grain size is R0, the
activation energy for grain boundary movement is Q and α is defined as a proportionality
factor. The data in Figure 42 supports this model, when considering a short annealing time
73
like 15 min. Deviations from the Arrhenius-like behaviour can be explained by variations in
the initial grain size due to the splat-quenching and slight deviations below the resolution
limit of EDX anaylsis, in the Pd content. As mentioned above, the martensite phase is
assumed to increase by growth rather than by nucleation. In a coarse-grained structure, there
are less grain boundaries pinning the phase boundaries, which leads to an increase of
transformation temperatures, especially of Mf, as the elastic energy stored during
transformation is smaller.171 XRD measurements at 313 K in the austenite phase were further
examined to clarify if the defect density and the intrinsic stress decreases upon annealing. The
(111) fcc peaks were fitted using a Gaussian function and the FWHM was used to calculate
the intrinsic stress as a function of Tan.
Figure 43: Intrinsic stress (black squares) as a function of annealing temperature Tan determined from XRD
measurements of splats in the austenite phase. The black line through the squares serves as guide to the eye. The
red circles represent the coefficient of correlation (R2) of the fit function to describe the Gaussian shape of the
(111) diffraction peak. An increase of intrinsic stress with Tan > 973 K is observed.
Figure 43 presents the change in intrinsic stress for the different splats. To calculate the
amount of stress, a Young’s modulus of 25 ± 5 GPa was used for the austenite phase.172 This
distribution indicates a maximum in intrinsic stress for the as-splatted sample. With
increasing Tan the intrinsic stress decreases due to out healing of crystal defects and a decrease
of interfaces like grain boundaries. For the splat annealed at 973 K the intrinsic stress shows a
minimum originated by the presence of equilibrium phases like Fe50Pd50 and α-Fe
precipitates. When Tan is increased to T > 973 K the amount of intrinsic stress increases again,
but is still below the value for the defect-rich structure of the as-splatted sample. Concluding
this, it is assumed that the martensitic transformation is shifted to higher temperatures with
74
increasing Tan due to the increase in grain size and changes in stress state and defect density
during the annealing process.
Summarizing all the presented findings it can be stated, that splat samples represent the bulk
counterpart to thin films within the Fe-Pd system. The splat-quenching technique avoids
decomposition of the fcc phase and reveals transforming bulk samples. These samples are
well-suited to investigate the change of materials properties when up-scaling promising thin
film compositions in newly developed Fe-Pd-X systems.
4.1.3 Epitaxial Fe-Pd thin films
The application of FSMA alloys in technical devices requires highly crystalline materials in
order to allow for the MFIS. This is only given when restraints like grain boundaries and
lattice defects are absent because they hamper twin boundary motions. Further there are
several intrinsic properties that are difficult to measure in polycrystalline thin film and bulk
samples, like magnetocrystalline anisotropy constants, magnetic coercivity HC and remanence
BR. These properties are important to investigate when discovering new Fe70Pd30-based
materials. Single crystals are ideal samples, since they allow determination of all intrinsic
materials properties without affecting the measurement results by external properties like
microstructure and scaling effects. But single-crystal samples are difficult to fabricate and
suffer from expensive fabrication and processing routes. This becomes important when such
materials, which exhibit their actuating/sensing effects only in the single-crystal state, have to
be incorporated into miniaturized devices. Thus, the fabrication of single crystalline thin films
also known as epitaxial thin film growth, is promising. Thin films, fabricated by epitaxial
growth using different methods like pulsed laser deposition, e-beam evaporation and sputter
deposition, were investigated in this chapter. All samples were fabricated by collaboration
partners from the IFW Dresden (group of S. Fähler for pulsed laser deposited films), Georg-
August-University of Göttingen (group of Prof. S. Mayr for thin films fabricated by e-beam
evaporation) and Christian-Albrechts-University Kiel (group of Prof. E. Quandt for thick
freestanding films) and investigated at the Ruhr-Universität Bochum. All data on samples
fabricated by other groups, were investigated and/or analyzed by the author of this thesis.
Substrate attached epitaxial Fe-Pd thin films
In a first attempt, epitaxial Fe-Pd thin films were fabricated using pulsed laser deposition to
investigate the effect of post-annealing on the thin film properties. The aim was to improve
75
the films’ microstructure by annealing of defects and relaxation of stress, in order approach
single-crystal-like behaviour. Investigations of phase stability, texture and phase
transformation behaviour are performed and correlated with magnetic properties. Fe-Pd films
with different Pd contents were deposited at room temperature on MgO (100) single-crystal
substrates using PLD from elemental targets by J. Buschbeck from IFW Dresden. Further
information about the fabrication can be found in Ref. 173. Fe-Pd films with 19 at.% < Pd <
37 at.% were annealed in the stability region of the fcc phase under vacuum at p = 10 -5 mbar
encapsulated in quartz tubes. After annealing at 1223 K for 10 min, the samples were cooled
by convection by taking the quartz tubes out of the furnace.
Figure 44: XRD patterns of epitaxial thin films with different Pd contents before and after annealing. Upon
annealing the diffraction peaks shift and the FWHM becomes smaller and more distinct, indicating a reduction in
defects and a structural transformation. The (110) bct diffraction peak was calculated for a c/abct - ratio = 1.06 as
observed in as-deposited films. Stars (*) denote diffraction peaks of Fe-rich bcc precipitates that occur in as-
deposited films. (XRD measurements on the as-deposited films were performed by J. Buschbeck and the figure
was originally published in Ref. 211)
76
The samples cooled from 1223 K to 673 K within 2 min leading to a cooling rate of ≈ 4.6 K/s
By this procedure, a decomposition of the alloy is avoided. Due to the chemically inert
behaviour of MgO, interfacial reactions or intermixing are not observed even at higher
temperatures. In Figure 44 XRD results, measured in Bragg-Brentano geometry, are presented
for as-deposited as well as annealed thin films at different compositions. In dependence on
composition, the position of the diffraction peaks changes considerably. A comparison with
peak positions expected from bulk material suggests that at 33 at. % Pd (Figure 44 a)), the fct
martensite transformed to the fcc austenite upon annealing. A fct to fcc-transformation due to
heat treatment is also observed in films with 29 at.% Pd and 37 at.% Pd content. According to
the intense (200) fcc peak, heat treatment of the film with 28 at. % Pd results mainly in a
transformation from bct to fcc phase after annealing (Figure 44 b)). A slightly increased
intensity observed between the (200) fcc diffraction peak and the (110) bct peak indicates that
minor amounts of martensitic phases are present. High stress in as-deposited fct films causes
the formation of misoriented fractions by deformation twinning.173,201 This is accompanied by
the occurrence of the (111) peak. After annealing, the (111) diffraction peak vanishes. At the
same time, the coherence length of thin films with Pd ≥ 28 at.% increases from the as-
deposited state (3 to 8 nm) upon annealing (41 to 45 nm) as shown in Table 2. This can be
explained by grain growth and healing out of defects when annealed at high temperatures, and
the absence of martensitic twins due to the formation of the fcc austenite phase. Shifting of
peak positions in the diffractograms of the Fe75Pd25 film suggest that the bcc phase
transformed into a bct martensite after annealing (Figure 44 c)). This phase transformation is
also observed in Fe81Pd19 films and is accompanied by a rather surprising change in texture
from a (200) orientation to (110) orientation. The existence of the bct martensite phase was
checked by measuring the tetragonal distortion of the unit cell in the 4-circle diffractometer.
From the position of the 2θ diffraction angle of the (110) diffraction and the tilted (101) peak,
the c/abct-ratio = 1.09 was determined. Due to splitting up into variants, the formation of a
martensitic phase can counteract an increase in coherence length by grain growth. Formation
of the bct martensite thus explains why only minor changes in coherence length are observed
after annealing of films with Pd ≤ 25 at.%. Epitaxial, Fe-rich precipitates being present in as-
deposited films are denoted as stars (*) in Figure 44 and vanish after annealing. As expected
by the Fe-Pd phase diagram, this phase is dissolved in the matrix upon annealing at
temperature above 1173 K. Structural data of all thin films in the as-deposited and the
annealed state is summarized in Table 2.
77
Table 2: Summarized structural data of Fe-Pd thin films at different Pd contents in the as-deposited and
annealed state, as determined by XRD. The normal lattice vector refers to the direction of the unit cell that is
closest to the substrate normal. (Table originally published in Ref. 211)
The change in texture of the thin films was investigated by pole figure measurements (Figure
45). As an example, pole figure measurements on the Fe67Pd33 film in the as-deposited and
the annealed state are presented for samples with ≥ 29 at.% Pd content. In the as-deposited
state, fourfold-symmetric intensities of the epitaxial orientation were measured at an angle of
ψ = 55°. Additional intensities were observed in the centre of the pole figure and around
ψ = 70°. These further texture components are caused by deformation twinning. After
annealing the intensities in the pole figure sharpened significantly around ψ = 55°. This
indicates the formation of a fcc structure having a sharp texture and a high coherence length,
as previously determined by Bragg-Brentano measurements. The epitaxial orientation
relationship for this thin film in the annealed state is MgO (001)[100] || Fe-Pd fcc (001)[100].
The crystal axes of the Fe-Pd fcc and the MgO unit cell are aligned in parallel. Besides the
epitaxial orientation, no additional texture components are observed. Misoriented fractions
being present in the as-deposited state vanish. When the Pd content is reduced, as presented
for the Fe72Pd28 film, the pole figure exhibits sharp intensities with fourfold-symmetry in the
as-deposited and the annealed state. Due to the low mismatch between MgO substrate and
Fe72Pd28 film, the bct phase grows in a single variant state.
78
Figure 45: Pole-figures measured on thin films at different Pd contents in the as-deposited and annealed state. In
a) the Fe67Pd33 film has a fct structure in the as-deposited state. The intensities originate from nonepitaxially
oriented texture components that vanish upon annealing leading to a full epitaxial fcc oriented structure in b). c)
Upon annealing the Fe72Pd28 film, the intensity maxima in the pole figures shift along the ψ direction, indicating
a phase transformation from the bct to the fcc phase. e) The Fe75Pd25 film exhibits a bcc structure in the as-
deposited state that transforms to a bct martensite state by annealing (f)). This bct martensite phase has a twinned
microstructure as depicted by a pronounced splitting of the intensities. The strong MgO (200) substrate intensity
located in the centre of the bcc (110) and the bct (101) pole figure is covered by a white spot for a better
understanding of the crystal orientation in the thin films. (Pole figure measurements were performed by J.
Buschbeck and the figure was originally published in Ref. 211)
No additional splitting of the intensities was observed. After annealing, the intensities remain
sharp but shift significantly from ψ =47° to ψ =55°. Together with the results from Bragg-
Brentano measurements, this evidence reveals the transformation from epitaxial bct phase to
epitaxial fcc phase. The epitaxial orientation relationship of this fcc phase to the substrate is
identical to the fcc phase in films with higher Pd content (as mentioned for the Fe67Pd33 film).
79
When the Pd content is reduced further, the intensities of the Fe75Pd25 film are sharp in the as-
deposited state. Upon annealing, the diffraction peaks in the pole figure measurement broaden
and split in intensity between ψ ≈ 53° and 60°. This behaviour can be explained by the
formation of a bct martensite phase with a twinned microstructure. In agreement with Bragg-
Brentano XRD, the pole figure measurement of the (110) bct diffraction peak verifies the
alignment of (110) bct lattice planes in parallel to the substrate plane. A single intensity is
observed in the centre of this pole figure (see Figure 46).
Figure 46: Illustration of the formation of the bct martensite phase with (110)bct planes orienting in parallel to
the substrate upon annealing. a) Starting from two fcc unit cells sitting on top of each other, the Bain
transformation results in formation of a bct unit cell with an in-plane aligned c-axis. b) As expected for this
alignment of the (110)bct planes a central diffraction peak is observed in the corresponding pole figure
measurement. (Figure originally published in Ref. 211)
The formation of this special alignment of the bct martensite unit cells can be understood in
terms of a phase transformation induced by annealing of the sample. Annealing was
performed in the stability region of the disordered fcc phase. According to the phase diagram,
the bcc phase in the as-deposited films transforms to the fcc phase at high temperatures.
Considering the annealing temperature of 1223 K, it is likely that this transformation is not
martensitic but derives from diffusion processes. The epitaxial orientation relationship of this
fcc phase is expected to be similar to the fcc phase at higher Pd contents: MgO (001)[100] ||
Fe-Pd fcc (001)[001]. During cooling, the high temperature fcc austenite phase transforms
into the bct martensite phase. Due to symmetry-breaking, three different alignments of the bct
unit cell on the MgO substrate are possible:
80
One with c-axis perpendicular to the film plane:
MgO (001)[100] || Fe-Pd bct (001)[110]
Two orientations with their c-axis oriented into the film plane:
MgO (001)[100] || Fe-Pd bct (110)[001]
MgO (001)[100] || Fe-Pd bct (110)[001]
Variant 1) is formed via the usual way of illustrating the Bain transformation with two
neighbouring fcc unit cells sitting beside each other on the substrate.67 In this case the
substrate plane is parallel to the a-b-plane of the forming bct unit cell. A (002) bct diffraction
peak belonging to variant 1), however, cannot be identified in XRD (Figure 44). In addition
this orientation should give intensities around ψ =45° in the (101) pole figure measurement
(Figure 45 f)). However, only a weak intensity is observed. Thus, there is no indication that
significant amounts of variant 1) are present in the film. Formation of variants 2) and 3) is
schematically illustrated in Figure 46. In this case, two fcc unit cells sitting on top of each
other have to be considered. According to the Bain transformation, a bct variant forms having
its (110) plane aligned in parallel to the film plane (Figure 46 a)). Formation of these variants
also explains the high (110) intensities in Figure 44. Since the (110) plane of these bct
martensite variants lies in the film plane, their c-axis does as well. Due to fourfold-symmetry
of the substrate, a second variant must form with its c-axis rotated by 90° in the film plane.
Thus, the Bain transformation explains the unusual texture change from (200) orientation in
the as-deposited state to (110) orientation in the bct martensite upon annealing. However,
since the Bain transformation is only a simplified model that does not consider a habit plane,
it does not explain the tilt and rotation of the crystal axes involved in the martensitic
transformation as depicted in Figure 45. The observed orientation of the bct martensite
forming after annealing is different to the growth texture of the bct phase in the as-deposited
state of Fe72Pd28 samples, where the cbct-axis was aligned perpendicular to the film plane in a
single variant state. Pole figures of defect rich, deformation twinned, fct films in the as-
deposited state (≥ 29 at.% Pd), exhibit considerable improvement in texture after annealing,
due to recrystallization (Figure 45 b)). The driving force for recrystallization is the reduction
of defects (grain boundaries and other interfaces) and stress (elastic energy) within the
microstructure. During the recrystallization process, defect-poor regions in the microstructure
grow at the expense of defect-rich regions. In the as-deposited films, only regions close to the
substrate interface (≈ 20 nm thickness) have an epitaxial orientation to the substrate.201 During
annealing, misoriented texture components vanish while epitaxy is maintained.
81
Figure 47: Schematic illustration of texture improvement by recrystallization. The as-deposited fct-film exhibits
a heterogeneous microstructure consisting of two layers: a) epitaxial orientation at the interface to the substrate
and b) a misoriented, deformation-twinned layer on top. During annealing, the defect-rich, misoriented layer is
overgrown by the epitaxial layer. In result, a fully epitaxial, single-crystal-like film is obtained. (Figure originally
published in Ref. 211)
This suggests that the epitaxial region at the substrate interface grows at the expense of the
defect-rich, deformation-twinned part of the film. This process is schematically illustrated in
Figure 47, showing the vanishing of misoriented areas in the thin film’s structure upon
annealing. According to the metastable phase diagram, films that exhibit a fcc structure at
room temperature should transform into martensitic phases at lower temperatures. XRD(T)
was performed on an annealed Fe72Pd28 thin film sample within an interval from 275 K to
175 K in steps of 25 K during cooling (Figure 48). Besides peaks originated by the substrate
and sample holder, only the (200) fcc diffraction peak is observed until 200 K. When the
sample is cooled below 200 K, the intensity of the (200) fcc peak decreases and a new
diffraction intensity appears at lower angles. This peak belongs to the (200) fct plane of the
martensite phase. This indicates the presence of a fcc-fct transformation upon cooling of the
sample. When the temperature is decreased further, the (200) fct peak shifts towards smaller
angles, indicating a decrease of the unit cell’s c/a-ratio. The length of the a-axes are
determined from the centre of the (200) fct peak. It is assumed that the volume of the unit cell
is constant during martensitic transformation (Vfct = Vfcc = 0.05336 nm3). Accordingly, the c/a
ratio of the fct martensite changes from c/a = 0.95 at 200 K to c/a = 0.94 at 125 K. Besides the
shift, a broadening of the (200) fct peak is observed with decreasing temperature. The
tetragonal distortion of the fct martensite strongly depends on the undercooling from the
martensite start temperature.64 It is known, that martensitic transformations can be constrained
by the interface to the substrate.174
82
Figure 48: Phase transformation from fcc to fct determined for an annealed Fe72Pd28 film by XRD(T)
measurements. Diffraction peaks from the sample holder are indicated by *. (Figure originally published in Ref.
211)
Considering this, there are two possible explanations for the peak broadening of the (200) fct
peak: 1) An inhomogeneous tetragonal distortion forms in the film during cooling. The peak
width suggests that c/a ratios ranging from approximately 0.95 to 0.93 coexist in the film at
125 K. A possible reason for this could be the hindering of the martensitic transformation at
the substrate interface. This corresponds to the fact that the fcc austenite diffraction peak does
not fully vanish until 125 K, revealing that a significant fraction of residual austenite remains
in the film. 2) A temperature-dependent reduction of the variant width due to decreasing c/a
ratio. Splitting up into smaller variants allows for compensation of elastic energy that is
caused by the tetragonal distortion. Similar reduction in the variant size has been observed
before, as a consequence of a decreasing c/a ratio in fct single crystals.64 Further XRD(T)
measurements on a Fe67Pd33 thin film sample did not reveal a phase transformation. This is
known in literature, and is correlated to the fact that the increased Pd content in this sample
shifts the martensitic transformation to lower temperatures and thus stabilizes the fcc phase
down to 125 K (see Figure 7).
Magnetic properties were investigated by hysteresis measurements at room temperature by
using a PPMS. In Figure 49 the hysteresis curves measured on Fe75Pd25, Fe72Pd28 and
Fe67Pd33 thin films are presented as examples.
83
Figure 49: Structure-dependent shape of magnetic hysteresis curves measured from the annealed films at room
temperature. (Figure originally published in Ref. 211)
The shape of the hysteresis curves can be interpreted as a magnetic fingerprint of the phases
present in the Fe-Pd films. While the fcc films exhibit a low HC and low saturation field, the
bct film exhibits a significantly higher coercivity and is harder to saturate. Magnetically softer
behaviour of the fcc phase is expected by its high crystal symmetry. In contrast, the shape of
the hysteresis curve measured from the bct phase was rather unexpected. As-deposited bct
martensite films can be saturated easily and exhibit a lower coercivity when compared to
films with a fct structure at Pd contents ≥ 29 at.%. Although the phases are similar (both
structures differ slightly in tetragonal distortion c/a ratio), this can be explained by the
significant difference in texture. While the bct phase in as-deposited films grows in a single
variant, untwinned state with c-axis out-of-plane, the bct phase forming after annealing has its
c-axis oriented in the plane and is in a twinned state. Thus, there are two possible
contributions that can explain the different shape of the hysteresis curves: 1) Pinning of
magnetic domain walls at twin boundaries and 2) the alignment of both hard and easy
magnetization axes along the measurement direction in annealed films. The different shape of
the hysteresis curves enable tracking of martensitic transformations. In contrast to XRD,
magnetic measurements can be performed down to lower temperatures. Hysteresis
measurements at 400, 200, and 50 K are depicted in Figure 50. Both the Fe72Pd28 and Fe71Pd29
compositions exhibit indications of a magnetic transition from the magnetization
measurements. From temperature-dependent hysteresis measurements, both, coercivity HC
and saturation field HS were extracted and are summarized in Figure 51. HS was taken from
the applied field where J reaches 90% of the polarization at 1 T.
84
Figure 50: Hysteresis curves of annealed thin films at different temperatures. The shape of the hysteresis loop is
independent of temperature for Fe67Pd33 (fcc at RT) and Fe75Pd25 (bct at RT). For Fe72Pd28 and Fe71Pd29 the
observed temperature-dependent change in shape of the hysteresis loop indicates a fcc to bct transformation.
(Figure originally published in Ref. 211)
Coercivities and saturation fields of the Fe75Pd25 and Fe67Pd33 films having a bct and fcc
structure at room temperature, exhibit an almost linear, temperature-dependent behaviour.
Thus, there is no evidence for a phase transformation in these samples. The slight, continuous
increase in HC and HS observed with decreasing temperature for the Fe75Pd25 sample is
expected for a normal anisotropic ferromagnetic material. In contrast to this, both films with
intermediate Pd content (Fe72Pd28, Fe71Pd29) exhibit a strong increase in coercivity and
saturation field starting below 250 K. Both films are in the fcc phase at room temperature.
From XRD it is known that the Fe72Pd28 film transforms to the fct martensite phase during
cooling. As expected by a tetragonal distortion, the hysteresis measurements reveal a
transformation from fcc to phases with increased magnetocrystalline anisotropy. In agreement
with the metastable phase diagram, the transformation is shifted to lower temperatures at
higher Pd content. The magnetic behaviour indicates that the martensitic transformation
proceeds further below the temperature of 125 K than was accessible in XRD. At 50 K, the
magnetic properties of both films are comparable to those of the more Fe-rich sample
exhibiting the bct phase. Formation of bct martensite is expected from the metastable phase
diagram and supported by XRD(T). The gradual decrease in tetragonal distortion that was
observed could finally result in the formation of the bct martensite at lower temperatures.64
85
Figure 51: Saturation field HS and coercivity field HC extracted from the hysteresis loop measurements in Figure
50. Films with highest and lowest Pd content only show minor changes in the magnetic properties. In Fe72Pd28
and Fe71Pd29 thin films indications of the phase transformation from fcc to bct are observed. (Figure originally
published in Ref. 211)
From literature it is known that pulsed laser deposited Fe-Pd films can exhibit a
heterogeneous microstructure in the as-deposited state. This microstructure is usually
originated by stress-induced deformation twinning.201 Recrystallization during postannealing
enables significant reduction in defects. As illustrated in Figure 52, annealing also has
considerable impact on the phase stability. Stress-induced fct martensite is present in as-
deposited films with ≥ 29 at. % Pd173. After annealing, the stress within the thin films is
released and the samples transform into the fcc austenite state. At low Pd content the bct
martensite phase is observed. Compared to bulk material, the bct phase is stabilized at lower
Pd contents. Interface and elastic energies stored in as-deposited fct films provide a large
driving force for grain growth. Since stress and defect density at the substrate interface are
highest201, the recrystallization starts from there. During annealing, the epitaxial fraction of
the as-deposited film overgrows the entire misoriented, deformation twinned part of the film
and deformation twins vanish. Single-crystal-like misorientation-free texture is obtained in
the films. Formation of the fcc phase shows that by recrystallization significant stress
reduction is achieved. Defect and stress reduction in annealed films result in a significantly
increased coherence length. Due to epitaxial (100) growth, both, the fcc and the fct phase can
be clearly identified by sharp, well separated (200) fcc and (200) fct peaks. The peak shift
enables easy following of the transformation path from fcc-fct. Comparable to experiments on
86
Ni-Mn-Ga174, the presented results indicate that the rigid substrate interface stabilizes the
austenite in Fe-Pd.
Figure 52: Comparative illustration of the phase stability ranges of Fe-Pd at room temperature in bulk material,
as-deposited films and annealed films. In bulk material the existence-range of fct is very narrow. Due to stress in
as-deposited films the epitaxial fct-phase is stabilized up to high Pd content. After annealing this stress is relaxed
and the fcc phase is observed. At low Pd contents either bcc or bct phases are observed. White areas have not
been investigated. (*) Indicates a compositional region where XRD measurements indicate a fcc phase with
minor amounts of martensite. (Figure originally published in Ref. 211)
It is observed that both bct martensite and fct martensite in the annealed films preferably form
variants with their c-axis aligned in the plane, corresponding to a two-variant state. In a
martensitic transformation, variants are formed to compensate the lattice distortion. In an
unconstrained state, equal fractions of all three possible variants would be expected. For the
present case, the different thermal expansion coefficients of film and substrate may result in a
stressed state. Since the thermal expansion coefficient of common metals exceeds those of
oxides, cooling from the annealing temperature results in tensile stress (this argument neglects
the invar anomaly around room temperature). Tensile stress may result in a variant selection,
but they should rather favour alignment of the unit cells with their a-axes in the plane and c-
axis out-of-plane. This, however, contradicts to the present observations. Since thin films are
investigated, it can be also suspected that the observed preference of particular variants is an
effect of the substrate, constraining the martensitic transformation. Indeed, recent model
calculations indicate that during a cubic to tetragonal transformation, the substrate constraint
may result in an alternating alignment of variants having parallel c-and a-axes in the film
plane while an a-axis is aligned perpendicular to the film plane.175 To enable the formation of
an invariant plane, this transformation requires the presence of two habit planes. Between the
habit planes and the interface to the substrate some residual austenite remains. This was
calculated for the transformation from cubic austenite to tetragonal martensite in Ni-Mn-Ga
and a similar scenario can be expected in Fe-Pd. The temperature dependence of the magnetic
coercivity and saturation field suggests that magnetocrystalline anisotropy increases during
87
martensitic transformation, from fcc to fct and bct structure. In general, an increase in
anisotropy can be expected due to the symmetry breaking of the formerly cubic unit cell.
However, strained epitaxial film growth showed that the magnetocrystalline anisotropy is
reduced if the tetragonal distortions of the unit cell exceeds those of the fct phase.67 To
explain this discrepancy, the microstructure of the bct martensite has to be considered. The
bct martensite exhibits a twinned microstructure. Hence a large number of pinning sites are
present, which commonly results in an increased coercivity. When considering that stronger
tetragonal distortion favours splitting up of the martensite into smaller variants, pinning can
explain the increasing coercivity with decreasing c/a-ratio. Moreover, decreasing the c/a-ratio
increases the twinning angle and the tilt of the crystal axes of the martensite unit cell with
respect to the former austenite unit cell.174 Due to the splitting of the martensite into variants,
a broad orientation distribution of the easy magnetization axes forms in the martensite,
explaining the roundish shape of the magnetic hysteresis measured for the bct martensite.
Concluding the presented results it is shown that annealing is suitable to achieve significant
reduction in defects and homogenization of microstructure in pulsed laser deposited epitaxial
Fe-Pd films. Single-crystal-like fcc films are obtained. Due to recrystallization, deformation
twins that were present in the as-deposited state completely vanish. Because of the epitaxial
(100) orientation, transformation from fcc austenite to fct martensite was investigated in detail
in XRD experiments. Residual austenite is observed even at temperatures of 50 K, below the
start of the martensitic transformation along with a complex formation of martensitic variants.
Both observations show that the substrate constrains the transformation from fcc to fct phase.
Different to bulk material, magnetic measurements indicate a continuous transformation from
fct to bct martensite at low temperatures. According to our results annealing of pulsed laser
deposited thin films is a promising route to obtain single-crystal-like epitaxial Fe-Pd films.
Freestanding Fe70Pd30 epitaxial thin films
When a film is grown epitaxially, the substrate underneath is constraining twin boundary
movement in the thin film lattice. Therefore it is important to release epitaxially-grown films
from the underlying substrate to allow twin boundary movement and therefore the MFIS.
Epitaxial growth of Fe70Pd30 thin films was performed by T. Edler using electron-beam
evaporation on MgO (100) substrates heated to T = 963 K. The structural relationship for the
thin film on the substrate is defined by MgO(001)[100] || Fe70Pd30fcc(001)[100] as depicted in
Figure 53.
88
Figure 53: Epitaxial relationship of the Fe70Pd30 thin film deposited on a (100) oriented MgO substrate.
Key parameters that have to be adjusted in this process are composition and the quality of the
crystal lattice in terms of defects and stress state within the thin film after deposition. Further
information about fabrication and processing can be found in Ref. 159. In order to release
these thin films from the substrate, two routes to etch the substrate were developed and
performed by Edler et al..176 The approach used to release the Fe70Pd30 thin films from the
substrate uses a saturated sodium bicarbonate solution that dissolves the MgO substrate at a
low rate of ≈ 80 nm/h. This procedure allows releasing epitaxially-grown Fe70Pd30 thin films
perfectly from the underlying MgO substrate without affecting composition, structure or
surface. XRD performed after fabrication as well as after releasing the film from the MgO
substrate revealed single crystalline thin films without a high crystal lattice defect density.
Figure 54 a) shows XRD spectra measured in Bragg-Brentano geometry for the substrate-
attached Fe70Pd30 thin film. Next to the substrate peaks of the (200) and (400) MgO planes,
there are additional peaks from Cu Kβ and W Lα radiation (blue text) diffracted on the MgO
planes, since the Cu X-ray radiation source is not monochromatic. A further peak comes from
the stage of the X-ray diffraction system. The Fe70Pd30 thin film shows only the presence of
the (200) fcc peak indicating a highly textured growth of the film on the underlying MgO
substrate. After releasing the thin film, the MgO peaks disappear, leaving only the (200) fcc
Fe70Pd30 peak. In Figure 54 b) a pole figure measurement of the (111) fcc planes of the
Fe70Pd30 thin film after MgO etching is presented to further prove if epitaxial growth is
present.
89
Figure 54: a) Comparison of θ/2θ XRD measurements before (red line) and after (green line) MgO etching
confirming the structural integrity of thin films after etching. The blue annotations indicate additional diffraction
peaks from Cu Kβ and W Lα radiation diffracted on the MgO planes and from the XRD stage. b) Pole figure of
the (111) fcc Fe70Pd30 planes proving the high degree of crystallinity after film release. (Pole figure measurement
was performed by T. Edler and originally published in Ref. 177)
Due to the four-fold symmetry and the significantly small FWHM of the (111) peaks, a high
degree of crystallinity and no structural changes after MgO etching is confirmed. In order to
investigate if the samples undergo a martensitic transformation, XRD(T) measurements were
performed on substrate-attached as well as on freestanding films in the as-deposited state.
These measurements show a variation of the fcc lattice parameter with temperature but do not
indicate a martensitic transformation. Since the Invar effect was observed to occur in the Fe-
Pd system, Figure 55 presents the lattice parameter of the fcc unit cell as a function of
temperature for both the substrate-attached and the freestanding Fe70Pd30 thin films in the as-
deposited state. For T > 300 K the latter remains constant, while the former increases with
temperature, although at a slightly reduced slope. This is considered to be a substrate effect,
originated by thermal expansion of the MgO substrate while the Fe70Pd30 thin film is attached
on top. Since calculations predict a very soft material and therefore a low resistance to
straining67, the thin film is expanded by the thermal expansion of the underlying MgO. This is
further corroborated by calculating the thermal expansion coefficient from the change in
lattice constant. The thermal expansion coefficient determined from the slope of the curve for
T > 300 K is αMgO+Fe70Pd30 = 12⋅10-6 K-1 which is in good agreement with the literature value
for MgO (αMgO = 14⋅10-6 K-1).
90
Figure 55: Lattice parameters as a function of temperature for the substrate-attached and the freestanding
samples in the as-deposited state. Dashed lines serve as guide for the eye. The Invar effect is observed as
stagnation of thermal expansion of the lattice parameter for T > 300 K in the freestanding sample. (Figure
originally published in Ref. 177)
While the substrate constraints are considered to hinder any structural changes of the lattice,
defect pinning was identified to constrain any martensitic transformation in as-released thin
films. Therefore, an annealing treatment with subsequent quenching was performed to heal
out crystallographic defects and thus to reduce pinning sides. The freestanding thin film
sample was sealed in a quartz tube under an Ar atmosphere of 600 mbar and annealed for 30
min at 1173 K. Subsequent quenching in water kept the sample in the metastable fcc phase
and avoided decomposition. Using XRD(T), a fractional thermoelastic fcc to bcc
transformation of the freestanding sample upon cooling was identified. Figure 56 presents
XRD(T) patterns measured at different temperatures. Above 278 K a single (200) fcc phase
occurs, consistent with the pole figure measurement. Additional peaks at 44.5° and 50.6°
originate from the sample holder and at 47.8° from a supporting Si plate. The step-like feature
at 47° is caused by a Ni filter that was used to weaken the Cu Kα2 and Kβ radiation. Below
278 K the (110) bcc peak starts to grow and increases in intensity upon cooling down to 128
K. At 228 K, small amounts of the fct phase occur, as indicated by the presence of the (200)
fct peak in the diffraction pattern. With decreasing temperatures down to 128 K, this peak
also increases in intensity. By heating the sample again up to 400 K, both the (110) bcc at 278
K and the (200) fct peak at 253 K decrease reversibly and disappear. At low temperatures
only small fractions transform from the fcc to the fct phase as inferred from the (200) fct peak
intensity. Thus a fractional martensitic transformation is considered to occur.
91
Figure 56: XRD(T) results of an annealed freestanding Fe70Pd30 sample. A reversible fractional fcc to bcc
transformation upon cooling was observed by the appearance of the (110) bcc peak. By further cooling down to
253 K, the (200) fct peak occurs, indicating a fractional martensitic fcc to fct transformation.
Since the (200) peak of the fcc phase occurs at all temperatures, it is reasonable to assume that
parts of the freestanding film transform simultaneously and reversibly from bcc to fcc and
from fct to fcc. The slight shift observed for all peaks with changing temperature is attributed
to the thermal expansion of the sample. It would have been interesting to measure the
structural changes at even lower temperatures, but this was not possible, since the Anton Parr
TTK 450 was not able to lower the temperature < 120 K. Instead, magnetic measurements
were performed to investigate the magnetic properties down to 25 K in order gain hints for
further structural changes below 120 K. These measurements were preformed by T. Edler
using a Quantum Design superconducting quantum interference device (SQUID)
magnetometer. The thin films were investigated upon cycling of the temperature several times
between 25 K and 400 K under a constant magnetic flux density of 0.01 T as presented in
Figure 57. The results from SQUID measurements confirm the findings of the XRD(T). The
temperature-dependent magnetization M(T) follows the Curie-Weiss behaviour for the as-
deposited freestanding thin film, while the annealed sample shows a more complex scenario.
First the magnetization increases during cooling (from 1 to 2), but changes its slope around
258 K as determeind by the tangential method. This coincides well with a fractional structural
change from fcc to fct around 253 K as determined by XRD(T). During the martensitic
transformation, the magnetic anisotropy increases significantly as reported by Kakeshita and
Fukuda.178 This anisotropy counteracts the external magnetic field, and thus leads to a
92
deviation from a Curie-Weiss behaviour towards lower values. When the sample is heated
again (from 2 to 3), the magnetization follows the cooling curve and again changes its slope
around 258 K, where the martensitic transformation occurs.
Figure 57: M(T) of as-deposited (blue squares) and annealed (red circles) freestanding samples measured in a
constant external magnetic flux density of 0.01 T. (Measurements were performed by T. Edler and the figure was
originally published in Ref. 177)
When the sample is heated above room temperature for the first time, quenched-in defects
(such as vacancies) can heal out. This reduces the pinning sites in the sample, and thus leads
to an increase in magnetization when compared to the initial cooling curve. As the
temperature is decreased again from 400 K to low temperatures (from 3 to 4), the slope
follows the curve of the as-deposited film. As observed for the initial cycle, the magnetization
shows a defined kink at 253 K as determined by the tangential method. When the temperature
is increased again, the curve fully follows the cooling curve. The Curie temperatures for the
freestanding film in the as-deposited and annealed states were examined according to
Kuz’min’s model.151 The parameters β = 1/3 and p = 5/2 were kept constant while the shape
parameters s, the Curie temperatures and the magnetizations at 0 K were fitted. The Curie
temperatures found for both the as-deposited (TC = 643 K) and annealed (TC = 625 K) films
agree well with bulk samples reported in the literature.52 The two fractional, thermoelastic
transformations from fcc to fct and fcc to bcc structure, as observed by XRD(T) in Figure 56,
can be understood using the Bain-path formalism as mentioned in Chapter 2.6 Figure 9. Using
the (200) fcc peak, a lattice parameter of c = 0.3788 nm for the fcc unit cell is observed at 400
K. The (110)-planes of the bcc cell correspond to the (200)-planes of the fcc cells and are thus
93
parallel to the substrate surface. At 128 K a lattice parameter of a = 0.2950 nm for the bcc cell
and of c = 0.3844 nm for the fct cell is observed. The fcc to bcc transformation in bulk Fe-Pd
is generally considered to be non-thermoelastic and irreversible.179 However, the
transformation behaviour of a Fe-Pd sample not only depends on the chemical composition,
but is also affected by its microstructure and other constraints. For example, the
transformation temperature of polycrystalline bulk Fe69.9Pd30.1 samples with a grain size of
about 2.6 µm is roughly 40 K lower than a sample having a grain size of 6.1 µm. A single-
crystal-like value of 292 K was found for a grain size of 266 µm.168 The microstructure after a
martensitic transformation in bulk material is determined by boundary conditions from the
matrix in which the martensite forms. To dissipate the total elastic energy, a twinning of the
martensite structure occurs. This elastic and twin-boundary energy counteracts the tendency
to form the martensitic phase, and leads to a hysteresis in the martensitic transformation.171 In
this sense, open surfaces can be considered as an ultimate way of relieving all constraints - in
particular if they are accompanied by a strongly reduced dimension normal to the surface. A
freestanding film can easily accommodate stress load by bending, expanding or stress
relaxation at the surfaces. Once a new structure has formed over the whole film thickness, all
transformation stresses (apart from the interface to the untransformed portion of the film) are
considered to be relaxed. Thus, the martensitic transformation does not stop in the fct phase,
but continues along the Bain path to the final bcc structure. Nevertheless, the boundary of this
transformation will be accompanied by large lattice strains, as it moves through the film. This
boundary can be pinned at defects, such as remaining dislocations, which originate from film
growth180 and have not completely healed out during annealing; this explains the remaining
fcc structure. A stress-induced transformation into the fct structure can be envisaged at this
interface.22 The reverse bcc transformation is related to the fact that the fcc phase retains a
significant volume fraction in the film. When the temperature is increased, these areas can act
as nucleation sites for the fcc phase and may greatly facilitate this transformation. Such
behaviour is not observed in bulk samples, as this would cause a large straining of the
surrounding matrix of the transforming area. However, in the freestanding film, this can,
again, be easily accommodated at the film surfaces. In conclusion, a fractional transformation
from the fcc structure at 400 K to a bcc structure at 278 K and a fct structure at 253 K,
respectively, was observed in freestanding epitaxially grown films. Even at temperatures as
low as 128 K the fcc, bcc and fct structures coexist. These reversible transformations can be
interpreted with the Bain-path formalism: due to the flat energy landscape between the
different crystal structures in materials showing a Bain path behaviour181, an adequate driving
94
force can induce such a fcc to bcc transformation, while free surfaces can provide a relaxation
source for the strains of the matrix.
Substrate attached Fe70Pd30 thick films
Ultra-thin films in the nanometre range are very challenging to be implemented into
microsystems, since their small thickness makes handling complicated. Further the MFIS can
be destabilized by this thickness due to a decrease in blocking stress making significantly
thicker films more desirable. Fabrication of active Fe-Pd films in the micrometre range would
provide a route towards application for this class of materials. Although epitaxial growth of
thin films is usually restricted to nanometre thicknesses, this approach was extended to obtain
freestanding Fe70Pd30 films of micrometer thickness, thus fulfilling both key requirements for
the integration into microsystems as well as prerequisites for FSMA films, that is, martensitic,
ferromagnetic at RT, freestanding, and single-crystalline-like.
Figure 58: Schematic of the orientation relationship of MgO substrate, Cr adhesion layer, Au buffer and
Fe70Pd30 film.
These films were fabricated by C. Bechtold from Christian-Albrechts-Universität Kiel and
investigated in close collaboration with the Ruhr-Universität Bochum. Fe70Pd30 films with a
thickness of 1.2 µm were fabricated by sputter deposition at very low deposition rates in the
range of 0.024 nm/s on a Au-buffer layer (d = 50 nm), deposited on a Cr adhesion layer (d = 5
nm) on (001) oriented MgO, polished single-crystal substrates. For further information on the
fabrication process refer to Ref. 182. Due to the coherent epitaxial growth on a Au buffer
layer deposited onto a MgO substrate, the Fe70Pd30 films are deposited in a bct martensitic
95
structure. The structural relationship between the substrate, the Au buffer layer and the
Fe70Pd30 film is:
MgO(001)[100] || Au(001)[100] and Au(001)[100] || Fe70Pd30fcc(001)[110]
as depicted in Figure 58. XRD measurements in Bragg-Brentano geometry reveal the (200)
Au diffraction peak of the buffer layer as well as the Fe70Pd30 (002) reflection. Assuming a
constant volume of the Fe70Pd30 unit cell compared to the cubic austenite, the lattice
parameters were calculated to be a = 0.287 nm and c = 0.321 nm (±0.001 nm), which
constitutes a c/a = 1.12. XRD(T) indicate no significant change in the crystal structure in a
temperature range between 150 and 375 K as presented in Figure 59 a). A peak shift of the
(200) Au and the (200) Fe70Pd30 diffraction peak dependent on temperature is observed. This
is correlated to the thermal expansion with temperature in both layers. Pole figure
measurements of the (101) diffraction peak at 2θ = 42.15° reveal a four-fold symmetry (see
Figure 59 b)). The maximum intensity is obtained at an average of 47.278° at ϕ = 45°. The
peak in ϕ direction is sharp with a small FWHM, indicating a well-oriented growth of the bct
unit cell rotated by 45° compared to the edges of the MgO cell. The increased FWHM in the
ψ direction indicates a relaxation of the lattice. Furthermore, no hints of relaxation such as
twinning in the film structure is found. All these results indicate the high quality of the single-
crystal structure in the Fe70Pd30 film even at such a high film thickness of 1.2 µm.
Figure 59: a) XRD(T) measured in Bragg-Brentano geometry reveals the (200) growth of the Au buffer layer
and the (002) growth of Fe70Pd30. A martensitic transformation of the bct Fe70Pd30 phase is not observed. b) The
four-fold symmetry of the (101) pole figure indicates epitaxial growth without misorientations. (Pole figure
measurement was performed by C. Bechtold and the figure was originally published in Ref. 182)
96
To gain a deeper insight into the microstructure of the film, TEM investigations were carried
out by B. Erkatal from Christian-Albrechts-University Kiel. A lamella sample was prepared
by FIB milling. Cross-sectional TEM and electron diffraction investigations revealed an
entirely epitaxial heterostructure. In Figure 60 a) a 2 µm wide cross-section bright-field TEM
image of the Fe70Pd30 film is presented. No signs of columnar growth or small-angle grain
boundaries are observed in the film’s cross-section, proving the rather perfect single-
crystalline structure. The bending contours observed in TEM are caused by the FIB milling
due to differences in lamella thickness. Compositional investigations performed by using
EDX in scanning mode (STEM) verified the Fe70Pd30 composition. Selected area electron
diffraction (SAED) patterns of the Fe70Pd30 film and MgO are shown in Figure 60 b) and c).
Based on the slight bending of the sample, only marginal readjustments of the zone axis
orientation were needed to observe the same SAED pattern over the complete area of the
cross section. A SAED pattern taken along the [100] zone axis of the film confirmed the
presence of the bct structure in the Fe70Pd30 film. The high-resolution TEM (HRTEM) image
of the rough interface between the Fe70Pd30 film and the Au buffer layer is depicted in Figure
60 d) and indicates an orientation relationship as described in Figure 58.
Figure 60: a) Cross-sectional bright field TEM image of Fe70Pd30 (thickness ≈ 1.2 µm) on a Au buffer layer
(thickness 50 nm) and MgO(100). b) Electron diffraction pattern of Fe70Pd30 (location: white circle in (a)). c)
MgO diffraction pattern revealing identical lattice distances of MgO(020) and Fe70Pd30 (110). d) HRTEM image
of the interface between the thick Fe70Pd30 film and Au (location: white box in (a)), showing the epitaxial growth
of the Fe70Pd30 film. (TEM measurements were performed by B. Erkatal and the figure was originally published
in Ref. 182)
97
The Young’s modulus of the epitaxially grown Fe70Pd30 film was determined by
nanoindentation under the guidance of J. Pfetzing-Micklich (Figure 61 a)). Averaging over 25
measurements gave an average value of 140 (±10) GPa. To release the Fe70Pd30 film from the
substrate, wet chemical etching of the sacrificial Au buffer layer was performed using an
aqueous solution of potassium iodide and iodine. Owing to shadowing effects during the
sputtering of Fe70Pd30, access to the Au buffer layer was blocked at the vertices. Therefore the
5 mm x 5 mm MgO substrates were broken into 1 mm x 5 mm strips. The small contact
surface between the buffer layer and the etchant results in long etching times of up to 36 h.
KI/I 2 is highly selective and does not significantly attack the Fe70Pd30 film during etching as
determined by the absence of etching pits using optical microscopy. XRD measurements on
the released, freestanding films show only a slight lattice relaxation compared to the films
restricted by the rigid substrate. The out-of-plane lattice parameter changed from 0.321 nm to
0.318 nm (±0.001 nm).
Figure 61: a) Nanoindentation depth profile of the Fe70Pd30 film deposited on a single crystalline MgO substrate.
(nanoindentation was performed under the guidance of J. Pfetzing-Micklich) b) Tensile tests show a reversible
deformation of 1% (εmax ≈ 1.5%, εirr ≈ 0.4%). (Tensile testing was performed by C. Bechtold and the figure was
originally published in Ref. 182).)
Mechanical properties of the released films were further investigated with a uniaxial tensile
testing machine. Since the film thickness is very small for tensile testing, the fixation within
the Al2O3 clamping jaws is critical. Careful clamping of the film ensured no slip was observed
up to strains of 1.5% (Figure 61 b)). A maximum strain of 1.5% was obtained at a stress level
of 150 MPa. In total, a plastic deformation of less than 0.4% is observed. The slopes of
unloading and loading branches coincide well, proving an elastic behaviour in these regions.
From the linear slope of the first unloading, the Young’s modulus was estimated to be ≈ 15
98
GPa. This is contrary to the nanoindentation measurements and indicates an unusual
elastically soft behaviour. The difference in mechanical properties determined by tensile
testing and nanoindentation can be explained by differences in mechanical loading and
deformation of the material. While in tensile testing uniaxial mechanical loading is applied
along [110] bct, a multiaxial stress state is present in the material under the tip of the
nanoindenter. The difference between the two measurements reflects the high mechanical
anisotropy that is characteristic for SMA.183 Nanoindentation measures an average of the
Young’s modulus in epitaxial films while tensile testing probes a particular direction.
Moreover, significant plastic deformation occurs in particular in front of the Berkovich tip.
By this deformation, the tetragonal lattice that has been forced by the coherent epitaxial
growth will get lost. The plastically deformed material at the nanoindenter tip is thus expected
to have significantly different mechanical properties compared to the rest of the material. All
these results prove that the constraints of an epitaxial interface allow coherent epitaxial
growth of Fe70Pd30 films up to large thicknesses as required for microsystems. Epitaxial
growth to such high thicknesses and without relaxation of the crystal structure implemented
by the seed layer is related to the mechanically soft behaviour originating from the
structural/martensitic instability. This is confirmed by stress-strain measurements of a
freestanding, 1.2 mm thick film. The observed Young’s modulus of 15 GPa is more than one
order of magnitude smaller than values known for common metals. Neither releasing the film
from the substrate nor heating it up to 375 K and straining resulted in a reverse transformation
to austenite. Since a stabilizing substrate is absent, alternative reasons for the absence of a
reverse transformation have to be considered. During a forward transformation, the tetragonal
martensite has a lower free energy than the cubic austenite below Ms. In order to minimize the
elastic energy at the habit plane, differently aligned unit cells form that are connected by twin
boundaries. Since one lattice constant of the martensite is shorter and one longer compared to
the austenite, this is geometrically possible. The reverse transformation commonly nucleates
at some residual austenite or at twin boundaries, where length compensation is possible. The
present film, however, was grown artificially, directly in the martensitic state. Considering a
diffusionless transformation from a single variant martensite to cubic austenite, the austenite
has no degree of freedom to compensate the tetragonal deformation. Hence the absence of the
reverse transformation (within the parameter ranges examined) can be attributed to the
different origin of the martensite stabilized in the film. The development of such artificial
structures as presented, are promising for micro-sensors probes. In these materials, length
changes can result in a significant change of magnetic properties and therefore can be
99
detected even without contact. Since a tetragonal distortion of Fe70Pd30 films already resulted
in strong changes of the anisotropic magnetic properties184, the further orthorhombic
distortion during uniaxial strain is expected to give even higher effects. Compared to
superplastic stress-strain curves of FSMA within the martensitic state, the present film reveals
a reversible behaviour, not requiring an external restoring system, and almost linear behaviour
is observed over a range of > 1% strain, which is beneficial for unambiguous data evaluation.
The key advantage, however, is that a single variant state is kept and the magnetic properties
are not averaged over different variant orientations.
100
4.2 Ternary Fe-Pd-X Ferromagnetic Shape Memory Alloys
4.2.1 The Fe-Pd-Mn System
4.2.1.1 Polycrystalline Fe-Pd-Mn thin films
The addition of Mn into Fe-Pd is quite promising as explained in Chapter 2.8, since it
preferentially couples antiferromagnetically to the surrounding Fe atoms and thus allows
utilization of the interdependence of the structural stability of the lattice and magnetism.
Investigation of Fe-Pd-Mn FSMAs bulk samples revealed that small Mn additions favourably
suppress the fct-bct transition that is known to be irreversible in bulk samples (see Chapter
2.6). Further, it was also reported that the addition of Mn also shifts the martensitic
transformation to higher temperatures.73
To clarify the impact of Mn on the structural and magnetic properties when added into Fe-Pd,
several Fe-Pd-Mn materials libraries were fabricated by co-deposition on Si/SiO2 substrates.
After deposition these materials libraries were subsequently annealed at 1123 K for 30 min
under N2 atmosphere followed by quenching. The high-throughput mapping techniques
mentioned in Chapter 3.2 were used to identify a compositional region of interest. Favourable
materials properties for new FSMAs were defined in Chapter 2.8 with respect to
transformation temperatures and magnetic properties. Thus, promising new FSMA thin film
samples have to undergo a martensitic transformation at an increased temperature when
compared to the binary Fe70Pd30 alloy. Further, these samples have to be in a ferromagnetic
state at T ≥ 550 K and in a single martensite phase. An example of the the identification of
this region of interest is shown for one materials library, and the findings are then summarized
for all three libraries fabricated, processed and investigated under identical conditions within
this thesis.
High-throughput characterization for identifying novel FSMAs
In a first step resistance mappings were performed across the materials libraries at 300 K. This
allowed identification of areas with similar values in electrical resistance. It has to be noted, that the
electrical resistance correlates with the conducting medium. This can lead to significantly different
values in electrical resistance for samples with similar materials properties when there is a difference
in film thickness. The film thickness of all samples in the presented materials library varies between
350 nm and 500 nm and therefore only changes its value in a range ≤ 30%. This variation in film
101
thickness is negligible, since the resistance mapping is used to get a coarse overview of boundaries
connecting different phases. Different phases as well as single- and multiple phase regions exhibit
different amounts of boundary interfaces within the samples and thus shift electrical resistance inthese
samples to higher values.
Figure 62: a) Electrical resistance mapping at 300 K of a Fe-Pd-Mn materials library showing 3082 colour-
coded resistance values. b) Partial section of a ternary composition diagram with colour-coded resistance values.
Lines are added for clarity to distinguish between the different regions. (Figure originally published in Ref. 185)
For this mapping 3082 points were measured at a step size of 1.5 mm in x- and y-directions. Each
point was measured three times for better statistics. In Figure 62 the variation in electrical resistance
(colour-coded) is presented as a function of the location on the substrate in a) and as a function of
composition in b). For clarity only the region of interest is displayed colour-coded, while the areas
with significantly higher and lower resistance are displayed in grey. During measurement, contact
problems between the sample and the pins can occur, leading to measurement values in the range of
MΩ: these points also have been excluded (shown in grey). For displaying the resistance data (3082
values) in a ternary composition diagram (301 compositions measured), the results of the EDX
measurements were interpolated. This is valid because the measurement points are close enough to
each other such that the concentration gradients between them are almost linear. By interpolating the
data, each resistance value pairs up with a composition and can be displayed in the ternary
composition diagram using colour-code. The results indicate the presence of different regions
with significantly different values in electrical resistance. Although the electrical resistance
changes only from 0.15 Ohm to 0.28 Ohm, distinct areas with similar values can be identified.
In the next step, R(T) measurements were performed to identify thin film samples that
undergo a martensitic transformation. This was conducted using the mapping mode in a
temperature range of 273 K to 423 K with temperature steps of 5 K.
102
Figure 63: a) Visualization of a Fe-Pd-Mn materials library showing the categorized results of R(T)
measurements. In b) R(T) results are summarized in the ternary composition diagram, where the region
containing transforming samples is encircled by a black line. Sample showing no transformation
(Fe51.1Pd42.8Mn6.1) are shown in c) while sample exhibiting a non-linearity and thus a martensitic transformation
(Fe68.5Pd26.8Mn4.7) are exemplarily presented for a specific composition in d). (Figure originally published in Ref.
185)
The results are presented in Figure 63: R(T) curves of samples that do not undergo a
martensitic transformation have a quasi-linear R(T) relationship (see Figure 63 c)), while
those exhibiting a martensitic transformation show an S-shaped non-linearity as presented in
Figure 63 d). As described, R(T) measurement curves give detailed information about the
transformation temperatures and temperature-hysteresis of transforming samples. The
analysis is time-consuming, since generally this has to be done manually. However, to get a
quick overview of the transforming sections of the materials library, the curves were
categorized to a scale ranging from 0 (non-transforming) to 1 (transforming). To do so the
smoothed second derivative of each curve was taken as an indication if a transformation
occurs. From the derived curve, the difference between maximum and minimum was taken
103
and collected for each curve before normalizing all values. The thin film samples in the area
depicted by a black line in Figure 63 b) were identified to undergo a martensitic
transformation. To prove if the samples which show a martensitic transformation are
ferromagnetic, MOKE measurements were carried out. Using the MOKE setup described in
Chapter 3.2, 301 compositions from the Fe-Pd-Mn materials library were measured at 300 K
(measurement grid: 4.5 mm). The magnetic flux of the electromagnet was cycled between -
300 mT and +300 mT.
Figure 64: Colour-coded visualization of the Fe-Pd-Mn materials library showing 301 categorized MOKE
measurement points. b) The same results shown in a section of the ternary composition diagram. The
ferromagnetic region is encircled by a black line. The following graphs depict different types of MOKE
measurement curves. A non-ferromagnetic sample (Fe63.2Pd19.7Mn17.1) is shown in c). Measurement curves for
ferromagnetic samples with detectable (Fe72.4Pd15.8Mn11.8) and well-defined magnetic hysteresis
(Fe82.1Pd12.3Mn5.6) are shown in d) and e). (Figure originally published in Ref. 185)
While the non-ferromagnetic points show no hysteresis curve (Figure 64 c)), the magnetic
sections of the materials library display hysteresis curves with step-like changes in MOKE
intensity (Figure 64 d) and e)). To compare the resulting data, the measurement curves were
automatically categorized defining a scale ranging from 0 (no hysteresis) to 1 (well-
pronounced hysteresis). This was done by taking again the second derivative of the original
curve to analyze the change in intensity of the MOKE signal. Since the different measurement
104
curves exhibit different noise levels, the resulting curve was divided by a factor proportional
to this noise level. The difference between maximum and minimum values of this data was
collected for each curve and finally normalized. These values correspond well with the quality
of the curves. The screening of the magnetic properties of the Fe-Pd-Mn materials library
using MOKE revealed ferromagnetic and non-ferromagnetic regions in the composition
diagram as visualized in Figure 64. On the Fe-rich side of the phase diagram (Fe > 80 at.%)
all samples are ferromagnetic, while in the Fe-poor region (Fe < 55 at.%) and the region with
Mn content > 12 at.% no ferromagnetic behaviour is observed. The remaining area shows
both ferromagnetic and non-ferromagnetic sections depending on the Pd- and Mn content.
The R(T) measurements revealed a region in the phase diagram where samples undergo a
martensitic transformation. In the room temperature resistance map of the materials library,
the same section can be is identified as well because it exhibits a higher resistance compared
to the surrounding points. In particular, those points are of interest which reveal both,
magnetic and phase-transformation behaviour. There is not a complete overlap of these zones,
but rather a small section includes samples of all regions as shown in Figure 65.
Figure 65: Superposition of all measurement results in the Fe-Pd-Mn materials library defining a region of
interest. All thin film samples within this region are ferromagnetic at 300 K and undergo a martensitic
transformation. (Figure originally published in Ref. 185)
This area lies in the range of (60 - 66 at.%) Fe, (6 - 8 at.%) Mn and (28 - 32 at.%) Pd. This
suggests that for Fe-Pd-Mn FSMAs Mn has to be added to the Fe70Pd30 system at the expense
of Fe (keeping the Pd content approximately constant) in order to achieve a transforming
phase, similar to Fe70Pd30.
The presence of a martensitic transformation determined by the R(T) measurement was
finally proven by using XRD(T). This is presented in Figure 66 where a Fe68Pd29Mn3 sample
was investigated between 300 K and 398 K. At 300 K the sample is in the fct martensite
phase indicated by the presence of the (111) and the absence of the (200) fcc diffraction peak.
105
Figure 66: XRD(T) measurement to verify that thin films showing an S-shaped non-linearity in the R(T)
screening undergo a martensitic transformation.
The (200) and the (002) fct peaks cannot be detected, since the signal to noise ratio is very
low in this measurement, hiding both diffractions in the background of the spectra. With
increasing temperature the (200) fcc peak of the austenite phase appears at 333 K and
develops in intensity fully at 363 K. This growing and disappearing of the (200) fcc peak was
found to be reversible, indicating that thin films with an S-shaped R(T) curve undergo a
martensitic transformation.
Structural investigation of all Fe-Pd-Mn materials libraries was conducted by XRD
measurements. Figure 67 presents samples that are in a single martensite/austenite phase
(depicted by filled red and blue circles) within the compositional region of interest. This
region is located within multi-phase areas. Besides the phases already known from the binary
Fe-Pd alloy (Fe50Pd50 with α-Fe precipitates and bcc Fe-Pd) new Mn-rich phases appear. At
Fe contents < 57 at.% a fcc structure, similar to the Fe50Pd50 phase and elemental Pd and Mn
precipitates (open green circles) appear. By increasing the amount of Fe > 67 at.%, the Fe-Pd
fcc/fct phases decompose and precipitates form. As described before, FSMAs have to be in a
single-phase state to show the MFIS, since defects such as grain boundaries hamper twin
boundary motion.
106
Figure 67: Diagram of the metastable phases occurring in the Fe-Pd-Mn thin film system. Besides multi-phase
regions, a small single-phase region is present (filled red and blue circles correspond to samples in a single
martensite/austenite phase at 300 K). (Figure originally published in Ref. 190)
Therefore all further investigations concentrate on Fe-Pd-Mn thin films being in the fcc/fct
single-phase state as depicted in Figure 67. The metastable phase diagram for the Fe-Pd-Mn
system indicates that the stability of the phases is preserved by trend, where the amount of Mn
increases at the expense of Fe for all single-phase samples. This is not unexpected since it is
known that Fe and Mn form several intermetallic compounds.186 Furthermore, an increase of
the lattice spacing was found in all fct/fcc Fe-Pd-Mn single phases when compared to binary
Fe70Pd30. This is related to the larger covalent atomic radius of Mn and the increased
spontaneous volume magnetostriction in Fe-Mn alloys.218 In addition, it is known from
literature that the crystal structure as well the intrinsic properties in Fe-Mn alloys vary
significantly dependent on the composition and annealing procedure.218 This is in good
agreement with previously reported results from literature.73 Although the solubility of Mn in
Fe-Pd was reported to be < 2.5 at.% it was found that Mn contents up to 10 at.% are soluble if
the Fe content is reduced accordingly.73
107
In Figure 68 the martensite start temperature Ms is correlated to the different crystal phases
determined within the Fe-Pd-Mn materials libraries. The compositional range in which single-
phase Fe-Pd-Mn thin films undergo a martensitic transformation follows the line in the phase
diagram, where the Mn content is increased at the expense of Fe content. The values
determined for Ms are the highest transformation temperatures reported for any binary and
ternary Fe-Pd-X FSMAs so far.
Figure 68: Phase diagram of single-phase Fe-Pd-Mn thin films with Ms transformation temperature depicted by
colour-code.
Magnetic properties of Fe-Pd-Mn thin films
Further changes in the magnetic properties are observed within the Fe-Pd-Mn single-phase
region when compared to Fe70Pd30. This is caused by the different magnetic alignments of Fe
and Mn in the Fe-Pd matrix. In Figure 69 the analyzed data revealed by magnetic
measurements for two thin films is presented as an example. Figure 69 a) shows the magnetic
hysteresis curves for a Fe66.1Pd30.1Mn3.8 and a Fe60Pd31Mn9 thin film in the martensitic (50 K)
and the austenitic (400 K) state. It can be seen, that both samples do not show a significant
difference in HC but do in JS. Further, the martensite phase has a significantly higher
saturation polarization JS than the austenite phase as explained previously for the binary
Fe70Pd30 alloy. In Figure 69 b) the normalized magnetization J/J50 K as a function of
temperature between 50 K and 400 K and under a constant magnetic flux density of µ0H =
0.05 T is depicted for both samples.
108
Figure 69: a) Magnetic hysteresis measurement for the martensite (50 K) as well as for the austenite (400 K)
state of a Fe66.1Pd30.1Mn3.8 (open circles) and a Fe60Pd31Cu9 (filled squares) thin film. b) Normalized
magnetization J/J50 K as a function of temperature and under a constant magnetic flux density of µ0H = 0.05 T is
presented. The data points are fitted using the Kuz’min equation for the Fe66.1Pd30.1Mn3.8 (open circles and red
curve) and the Fe60Pd31Cu9 (filled squares and blue curve) thin film.
Using the Kuz’min relation a fit (red: Fe66.1Pd30.1Mn3.8; blue: Fe60Pd31Mn9) was applied
through the data points and extrapolated to the temperature, where J/J50 K approaches zero.
Both the JS and TC values revealed for single-phase samples are presented in Figure 70. Here,
the saturation polarization as a function of Fe content (colour coding indicates the Mn
content) for both the martensitic (squares, determined at 50 K) and the austenitic structure
(triangles, determined at 400 K) is shown in a). The saturation polarization decreases with
increasing Mn content from JS = 1.56 T for Fe66.7Pd28.8Mn4.5 and reaches a minimum at
JS = 1.22 T for Fe61.6Pd30Mn8.4 in the martensitic fct phase (determined at T = 50 K). At Mn
contents > 8.4 at.% JS increases again slightly to JS = 1.29 T for the Fe60Pd31Mn9 thin film.
This behaviour originates from the addition of antiferromagnetic Mn at the expense of
ferromagnetic Fe. The antiferromagnetic interactions are considered to be strong and thus an
increase of TC with decreasing Fe content is observed. This is shown in Figure 70 b) where TC
and the shape fitting parameter s as a function of Fe content are presented.
109
Figure 70: a) JS as a function of Fe content for single-phase samples in both the martensite (50 K) and austenite
phase (400 K). Highest values for JS are identified for the highest Fe content. b) TC and shape parameter s as a
function of Fe content, show a significant decrease with increasing Fe content. The Mn content is indicated by
colour-coding.
This behaviour is not unusual, since the corresponding trend in the binary alloy exhibits an
increase of TC with decreasing Fe content until it reaches its maximum value for the Fe50Pd50
phase.187 Nevertheless, it can be concluded that the addition of Mn generally lowers TC when
compared to the binary Fe70Pd30 alloy. The addition of 9 at.% Mn at the expense of Fe content
lowers TC (TC for Fe70Pd30 = 700 K to Fe60Pd31Mn9 = 499 K) by about 28 %.
Transmission electron microscopy
TEM was performed to investigate the existence of a single ternary phase in the sample
Fe66.1Pd30.1Mn3.8. The prepared lamella consist of the following layers: SiO2 substrate layer,
the homogeneous Fe-Pd-Mn layer and a Pt-protection layer deposited prior to FIB-
110
preparation. Figure 71 a) shows the HRTEM image of a selected region of the
Fe66.1Pd30.1Mn3.8 layer. A red coloured square indicates a region where a diffraction pattern
was calculated using a fast Fourier transformation. The difference in colour across the
overview photograph in Figure 71 a) is related to the orientation contrast of differently
orientated grains in the lamella. The diffraction pattern taken from the region is depicted by a
red square in Figure 71 b) and reveals the occurrence of lattice planes with different distances.
The different distances are ascribed to the following lattice planes: d(002) fct = 0.189 nm,
d(111) fct = 0.22 nm and d(111) fcc = 0.223 nm.
Figure 71: a) High resolution transmission electron microscope (HRTEM) overview of a Fe66.1Pd30.1Mn3.8 film.
The inset b) shows a fast Fourier transformation conducted over a regions depicted by a red square. The
diffraction patterns in this region revealed a mixture of the martensite fct and the austenite fcc phases. (FIB
preparation was performed by C. Zamponi and TEM by B. Erkatal from the group of Prof. L. Kienle at the
Christian-Albrechts-University of Kiel)
The determined lattice spacings confirm that the presented lamella of a Fe66.1Pd30.1Mn3.8 thin
film has a microstructure consisting of both the martensite as well as the austenite phase.
Since this measurement was performed at ambient temperature (300 K) the martensite phase
would be expected judging from the increased transformation temperatures compared to
Fe70Pd30. Nevertheless, the sample where the lamella was extracted is a thin film. This kind of
111
sample usually exhibits high residual stresses due to the difference in thermal expansion
between thin film and substrate. This can lead to a significant increase in transformation
temperatures due to stress-induced martensite as mentioned previously. After extraction of a
lamella out of the surrounding thin film, the residual stress can partially relax through a
bending of the lamella, allowing it to transform back into the austenite phase as shown here.
Further a local heating of the observed region by the electron beam cannot be excluded. This
can lead to increased observation temperatures than the ambient value of 300 K and thus can
induce the martensitic transformation in the lamella. The determined lattice distances are
significantly higher compared to the distances reported for the binary Fe70Pd30 alloy. This is
in good accordance with the synchrotron measurements that revealed an increase of the lattice
parameters upon addition of Mn at the expense of Fe, which are presented in the following
section. Compositional analysis by EDX further revealed no decomposition or separation of
Mn and Fe-Pd. Thus it can be summarized that Mn is homogeneously dissolved in the Fe-Pd
matrix and no composition or phase segregation was detected within the analyzed regions.
Stress effects in Fe-Pd-Mn thin films
To further clarify how the addition of Mn varies the lattice parameters within the Fe-Pd-Mn
thin films, temperature-dependent synchrotron radiation measurements were performed.
Further residual stress measurements using the sin2(ψ) method were used to determine the
crystal structure and the stress state of Fe-Pd-Cu thin films in the martensite (300 K) as well
as in the austenite (393 K) state. In Figure 72 the area detector diffractograms of a
Fe66.1Pd30.1Mn3.8 thin film at two different temperatures are presented as examples. The area
diffractograms depict the colour-coded intensity (blue for low and white for high intensity) in
terms of X-ray counts as a function of d-spacing and ψ angle. At an angle of ψ = 90°, the ψ-
vector points to the thin film surface at an angle of 90° and thus is parallel to the thin film
normal vector. This is indicated by a dashed white line, that divides the upper half (90°-160°)
to one half-sphere and the lower part (20°-90°) to the other half-sphere above the sample.
Instead of intensity rings known for polycrystalline samples, intensity spots appear. These
spots originate from texturing and correspond to the specific d-spacings of different phases. In
Figure 72 a) the crystal structure of a Fe66.1Pd30.1Mn3.8 thin film in the martensite phase at
300 K is presented. The martensite phase can be identified by the presence of the (220), (002),
(200) and (111) fct diffraction peaks. Upon heating to 393 K the thin film transforms into the
austenite phase by developing a strong (200) fcc peak.
112
Figure 72: Area detector diffractograms of a Fe66.1Pd30.1Mn3.8 thin film at different temperatures. a) At 300 K the
thin film is in the martensite phase, proved by the presence of the (200) and (002) fct peaks. When heating the
sample to 393 K in b), the sample transforms into the austenite phase and the (200) fcc peak appears. The red
stars denote diffraction intensities originated from the Si/SiO2 substrate. (Synchrotron measurements were
performed by H. Brunken)
In order to verify the lattice spacings determined by TEM, the lattice spacings in the
martensite (fct (111) and fct (200) diffraction peaks at 300 K) and austenite (fcc (111)
diffraction peak at 393 K) state were determined. The variation of lattice spacings in
113
dependence on the Fe content (Mn content is colour-coded) is depicted in Figure 73 for a
Fe66.1Pd30.1Mn3.8 thin film. The values for both the (111) fct martensite (open circles) and
(111) fcc austenite (open squares) lattice spacings correspond to the left scaling (black
arrows). Values describing the variation of the (200) fct martensite (open triangles) lattice
spacing correspond to the right scale (dark blue). From this data it can be easily seen that the
lattice parameters increase upon addition of Mn, which originates from the significantly
higher covalent atomic radius compared to Fe (rMn = 0.135 nm < rFe = 0.126 nm).188
Figure 73: Shown is the lattice spacing revealed from the different diffraction peaks in the martensite (circles:
fct (111) and triangles: (200) diffraction peaks at 300 K) and the austenite (squares: fcc (111) diffraction peak at
393 K) state as a function of the Fe content for a Fe66.1Pd30.1Mn3.8 thin film. The values for (111) peaks
correspond to the scaling on the left (black). Values determined from the fct (200) correspond to the right scale
(dark blue).
It has to be mentioned that the lattice spacing of the fcc (111) diffraction peak was measured
at a higher temperature (393 K) compared to TEM performed at ambient temperature (300 K).
Since the lattice parameters especially in Fe-Pd-based systems depend on temperature, slight
differences have to be expected. Further, stress effects can alter the lattice parameters,
especially when considering that the TEM lamella can partially relax while the thin film was
substrate attached during synchrotron radiation measurements. Thus it can be concluded, that
values determined from synchrotron measurements correlate well with values determined
from TEM diffraction on a Fe66.1Pd30.1Mn3.8 thin film within the error margins.
In order to determine the residual stress state of this thin film sample in the austenite phase,
the lattice parameter afcc of the thin film was determined from the (111) fcc peak at different
ψ-angles. The variation of lattice parameter with tilt angle afcc = sin2(ψ) is shown in Figure
114
74 a). From the slope of this relationship the stress state and the strain in this thin film can be
determined. The data points are fitted by using a linear function and the slope of this function
is used to determine the residual strain. The positive slope indicates a tensile residual strain
that appears for thin films showing Invar-related anomalies with smaller thermal expansion
coefficient than the substrate they were deposited onto. In Figure 74 b) the FWHM and c) the
intensity of the (111) fcc diffraction peak as a function of sin2(ψ) is shown to ensure no
texture related errors. Since the sin2(ψ) method only reveals the strain of the sample, the
Young’s modulus needs to be determined further for every thin film.
Figure 74: Lattice parameter as a function the ψ-angle for a Fe66.1Pd30.1Mn3.8 thin film. a) A linear function was
fitted through the data points to determine the slope. The positive slope indicates a tensile strain state in this
sample. For every lattice spacing afcc in a) that corresponds to a diffraction peak, the FWHM b) and the intensity
c) for this peak was determined.
This was performed by nanoindentation measurements of the Fe-Pd-Mn thin films. The
measurements were conducted at ambient and elevated temperatures (353 K) to determine the
Young’s modulus in the austenite phase. Examples of the nanoindentation measurements are
depicted in Figure 75 a) to c) for a Fe66Pd30Mn4 thin film sample. In Figure 75 a) the load on
the sample as a function of the displacement starting from the thin film surface is shown. In
Figure 75 b) the data evaluated to measure the Young’s modulus are presented.
115
Figure 75: Examples of measurement curves for a) load, b) Young’s modulus and c) hardness as a function of
displacement of the Berkovich tip into the thin film surface.
The area defined by hatched stripes depicts the indentation depth that was used for Young’s
modulus measurement and describes a displacement ranging from 50 nm to 70 nm. The
values for thin film hardness are illustrated in Figure 75 c). Here, the hardness values again
were calculated from indentation depth between 50 nm to 70 nm. The choosen indentation
depth follows the rule that up to an indentation depth of one-tenth of the overall film
thickness, the influence of the substrate is negligible.
Figure 76: Summary of nanoindentation measurements at 353 K for Fe-Pd-Mn single-phase thin films.
Presented are the a) Young’s modulus and the b) thin films’ hardness as a function of Fe content with colour-
coded Mn content. With decreasing Fe content (and increasing Mn content) a lowering of Young’s modulus and
hardness are observed. Further data from binary Fe70Pd30 reference thin films that were fabricated and processed
under identical conditions are included in both diagrams.
Nanoindentaion measurements are summarized for single-phase Fe-Pd-Mn thin films in
Figure 76. In Figure 76 a) the Young’s modulus values as a function of Fe content (Mn is
colour-coded) are illustrated. From the distribution of the data, it can be stated that the
116
Young’s modulus increases with increasing Mn content. The values for the binary Fe70Pd30
reference thin films that were fabricated and processed under identical conditions, are marked
by a hatched area around 140 GPa. Since the values are significantly higher than the value for
the underlying Si/SiO2 substrate, a significant influence of the Si substrate can be excluded.
Hardness values revealed from nanoindentation measurements are shown in Figure 76 b) and
follow a similar trend as observed for the variation Young’s modulus with Mn content.
The determined values for the Young’s modulus were used to calculate the amount of residual
stress in the Fe-Pd-Mn single-phase thin films. In Figure 77 the residual stress as a function of
the Mn content is depicted for samples with a Pd content between 28.7 at.% and 30.5 at.%.
The colour-code indicates the martensite start temperature Ms. A significant increase in
residual tensile stress is observed with increasing Mn. The values for low Mn content > 7 at.%
exhibit high tensile stress values, while Ms is in the range of 356 K to 364 K. Upon decrease
of Mn content, the amount of tensile stress lowers significantly while Ms increases. At a Mn
content of 6.7 at.% the amount of tensile stress is lowered to 658 MPa although the
martensitic transformation is shifted to the highest value of Ms = 379 K.
Figure 77: Presented is the amount of residual stress as a function of Mn content. To separate the impact of Pd
on thermal expansion behaviour and thus the stress state, only thin film samples with a small variation in Pd
(28.7 at.% < Pd < 30.5 at.%) content are compared. Upon addition of Mn a significant increase of tensile stress is
observed. The change of Ms temperature (colour-coded) does not follow a well-defined trend and has average
values for samples with highest tensile stress amount.
Thin films with a 4.5 at.% < Mn < 6 at.% exhibit Ms temperatures > 350 K. Nevertheless, the
lowest value for Ms = 336 K is found for a Fe66.1Pd30.1Mn3.8 thin film with a stress of
117
339 MPa. From this data it can be concluded that no obvious correlation between tensile
stress amount and transformation temperatures exists.
Variation of martensitic transformation in the Fe-Pd-Mn system
In order to identify the dominant parameter that can be used to describe the shift of
transformation temperatures, the martensite start temperature Ms of all samples in a single-
phase state is plotted as a function of e/a ratio and the Fe content. The amount of Mn in each
sample is indicated by the colour-code of the symbols. The data is presented in a in a three-
dimensional diagram. For clarity, only the projections of data points into the three base planes
are depicted. Ms (coloured triangles) and the e/a ratio (coloured circles) as a function of Fe
content do not show a simple analytic trend.
Figure 78: Martensite start temperatures, Ms as a function of e/a ratio and Fe content presented in a three-
dimensional graph with colour-coded Mn content. Due to the addition of Mn, Fe content and e/a ratio are two
independent variables. When looking at the martensite start temperature Ms as a function of the e/a ratio, a linear
dependency becomes visible. In the other two projections no clear dependency is present. (Figure originally
published in Ref. 190)
However, an approximately linear behaviour is observed when Ms is depicted as a function of
the e/a ratio. It should be kept in mind that the e/a ratio is independent of the Fe content when
one or more additional elements are added into Fe-Pd alloys. At e/a = 8.51 (Fe64.3Pd29Mn6.7)
the highest transformation temperature in the system is Ms = 379 K and reaches down to a
value of 335 K at e/a = 8.56 (Fe66.1Pd30.1Mn3.8). The description of Ms in terms of the e/a ratio
is common in literature, especially for Fe-based systems. Although this relation was mostly
118
applied to binary alloys such as Fe-Pd, Fe-Pt and Fe-Ni,189 where the Fe content and the e/a
ratio are related to each other, this dependence holds also for some ternary Fe-Pd-X (X = Ni
and Co) alloys including the Fe-Pd-Mn system.71 After identification of the decisive factor
that describes the change of the martensitic transformation, the origin of the increased
transformation temperatures has to be clarified.
Figure 79: Energy gain due to displacements from the ideal lattice positions compared for different ternary
compositions Fe70.4Pd25X4.6, where X represents the third component. Further information can be found in Ref.
65. The choice of antiferromagnetic elements such as X = Cr and Mn lead to considerable energy gain compared
to binary Fe70.4Pd29.6. (Figure adapted from Ref. 190)
A first hint comes from ab initio calculations performed by M. E. Gruner presented in Figure
79. Here, a significant energy gain (approximately 27 meV/atom) due to displacements from
the ideal lattice positions is found after addition of 5 at.% of Mn. This is significantly smaller
if the lattice sites are occupied by (ferromagnetic) Fe and underlines that antiferromagnetic
Mn causes local structural distortions (even at low temperatures) which significantly impact
the system when compared to other ferromagnetic contributions like Fe, Co, and Ni.191
However, the influence of the magnetoelastic effects at finite temperature are not easy to
predict, since - on the one hand - a magnetically excited Mn atom might still show a parallel
alignment to the average magnetization, while - on the other hand - the thermal stability of the
ferromagnetic phase itself is compromised by large amounts of Mn. A shift of martensitic
transformation due to thin film stress effects is not considered, since Ms does not distinctly
scale with the stress amount.
By summarizing all the shown findings an enhancement of properties due to the addition of
Mn into Fe-Pd occurs. The Fe-Pd-Mn system exhibits the highest transformation temperatures
119
up to 379 K measured for Fe-Pd-based FSMAs so far. The single-phase region in the
compositional diagram is extended, allowing to alloy up to 9 at.% of Mn in Fe-Pd without
segregation of the fct/fcc phase. Considering the magnetic properties, neither TC nor JS are
drastically lowered. From theoretical predictions an energy gain due to displacements from
the ideal lattice positions upon addition of Mn was reported, explaining the increased amount
of energy needed to induce the martensitic transformation. Finally the e/a ratio was found to
be the dominating parameter, describing the change of the martensitic transformation.
4.2.2 The Fe-Pd-Cu System
4.2.3.1 Polycrystalline Fe-Pd-Cu thin films
In order to investigate the materials properties without significantly affecting the magnetic
properties Cu was added into Fe-Pd as described in Chapter 2.8. The change of the e/a ratio at
constant Fe content, the variation of the unit cell volume and the impact of Cu on the
stabilization of the metastable and disordered Fe70Pd30 phase against decomposition was
investigated. Due to the higher number of valence electrons of Cu in comparison to Fe and
Pd, the e/a ratio is supposed to increase. Nevertheless if the Fe content is increased at the
expense of Pd, also e/a ratios below 8.6 can be achieved. This allows for a variation of the Fe
content while keeping the e/a ratio constant and thus enables the identification of the decisive
factor that characterizes the martensitic anomaly. The colvalent atomic radius of Cu lies
between those of Fe and Pd and thus decreases the unit cell volume, if the Cu content
increases at the expense of Pd. Finally, Cu with its full d-shell is known to be an austenite
stabilizer eventually allowing stabilization of the high temperature fcc austenite phase.
In order to gain a deeper understanding of basic magnetic and magnetoelastic properties, first-
principles calculations in the framework of density functional theory (DFT) were carried out.
The theory results by M. E. Gruner (University of Duisburg-Essen) were then compared to
experimental results. The calculations were carried out with the Korringa-Kohn-Rostoker
Green-function method (KKR) in connection with the atomic sphere approximation (ASA)
and the coherent potential approximation (CPA) for the description of configurational
disorder, as implemented within the fully relativistic Spin Polarized Relativistic Korringa-
Kohn-Rostoker (SPR-KKR) package from Ludwig-Maximilians-Universität Munich.192,193
The calculations employed the generalized gradient approximations (GGA) for the description
of the exchange correlation potential in the formulation of Perdew, Burke and Ernzerhoff.194
Separate calculations were carried out for all compositions exhibiting martensitic
120
transformations as determined in the experiments. For each composition, the equilibrium
lattice constants for the austenite ferromagnetic and magnetically disordered configuration
were determined; the latter was simulated by mixing Fe-species of different spin orientations
by means of CPA, in the spirit of the disordered local moment approach.195,196 Furthermore,
spin and orbital magnetic moments were obtained assuming a ferromagnetic spin
configuration as the ground state. In the angular momentum expansion, orbitals were included
up to lmax = f states; 1639 irreducible k-points were used for the Brillouin zone integration. In
a separate calculation, magnetic exchange-parameters were determined for use within a
Heisenberg-model.197 From this, the Curie temperatures within the mean-field approximation
were obtained, which allows a comparison of compositional trends between theory and
experiment.
Structural behaviour of Fe-Pd-Cu thin films
The measured compositions of the ternary Fe-Pd-Cu materials libraries are depicted in a
partial composition diagram in Figure 80 a) (600 samples). Open grey circles represent
compositions measured by EDX, while black lines indicate the calculated e/a ratios of the
respective samples. Filled circles represent all samples showing a thermoelastic non-linearity
in the R(T) measurement, with the martensite start temperature being indicated by colour-
coding. The composition area investigated, is centered around Fe70Pd30 (from Fe40 to Fe94)
with variation of Cu from 1 to 18 at.%. In order to identify compositions showing phase
transformations, all compositions were investigated by R(T) screening measurements.
Figure 80: a) Partial Fe-Pd-Cu composition diagram showing the compositional area covered by two materials
libraries. Constant e/a ratios are defined by black lines. Grey circles denote the fabricated samples of the
materials libraries. Transforming samples are indicated by filled circles, with colour-coding indicating the
martensite start temperature Ms; b) example R(T) curve of the transforming sample Fe71.8Pd26.6Cu1.6. The
determination of transformation temperatures by the tangential method is indicated. (Figure originally published
in Ref. 199)
121
Figure 80 b) shows an example R(T) curve from a Fe71.8Pd26.6Cu1.6 thin film, which indicates
a phase transformation due to its non-linear S-shaped R(T) curve, from which transformation
temperatures were determined by the tangential method.144 In order to assign the observed
non-linearities in the R(T) curves to a martensitic transformation, XRD(T) was performed.
Figure 81 shows an example of XRD(T) data for the sample Fe71.8Pd26.6Cu1.6. At 253 K, (111)
and (200) fct reflections are observed, as expected for a polycrystalline martensitic structure
comparable to binary Fe70Pd30.61 No additional peaks are observed, which indicates the
absence of large precipitates. With increasing temperature, the (111) fct peak does not change;
instead the (200) fct peak decreases while the (200) fcc peak starts to grow, starting at
approximately 293 K. With further increasing temperature, the (200) fcc peak continues to
increase in intensity up to 343 K. For temperatures above 343 K, no further growth of the
(200) fcc peak is observed. Thus, martensitic transformation temperatures deduced from the
R(T) measurement are confirmed by the structural analysis and the transformation
temperatures determined by these methods showed no significant differences.
Figure 81: XRD(T) of the sample Fe71.8Pd26.6Cu1.6 measured between 253 K and 373 K. (Figure originally
published in Ref. 199)
In order to determine the structure of the samples, room temperature XRD mapping of the Fe-
Pd-Cu samples was performed. Figure 82 a) shows the ternary Fe-Pd-Cu composition
diagram, indicating all investigated compositions. Transforming compositions are colour-
coded according to their martensite start temperature, Ms, while the symbols represent the
respective phases. The squares in Figure 82 a) define single-phase compositions showing a
non-linearity in the R(T) measurement. Diamonds designate multi-phase samples where the
122
transforming Fe70Pd30 phase dominates but additional precipitate phases occur and are thus
not considered any further. The dark grey triangles represent samples having a Fe50Pd50
structure with additional Fe-Cu and Pd-Cu precipitates. For Fe-concentrations above 72.5
at.% only a bcc structure and precipitate phases (α-Fe, Fe-Cu and Pd-Cu) are present (grey
circles). Open squares indicate samples having a Fe70Pd30 austenite phase with additional Fe-
Cu and Pd-Cu precipitates present at larger Cu contents, and not showing an S-shaped R(T)
measurement curve.
Figure 82: a) Ternary diagram of the Fe-Pd-Cu system showing the distribution of phases (symbols) and the
transforming compositions (Ms colour-coded). The arrow indicates a constant Fe content. Diffractograms along
this composition are summarized in b). For Cu > 6 at.% the transforming fct/fcc structure decomposes into a bcc
structure with Fe8Cu2 precipitates. (Figure originally published in Ref. 199)
123
All single-phase compositions are located in a zone with Fe content between 69 and 72 at.%
and having a Cu content from 1 to 5 at.%. The distribution of transforming single-phase
samples in the ternary compositional diagram suggests an increase of Cu at the expense of Pd.
Samples with Cu contents <5 at.% transform up to a maximum Fe content of 72 at.%. For
binary films an expansion of the transforming region from Fe70Pd30 to Fe72Pd28 was
previously observed.143 However, in the ternary system, a considerably larger maximum
Ms = 359 K for compositions around Fe71.8Pd26.6Cu1.6 was observed. In order to understand
how Cu alloying into Fe-Pd affects the structure, a “binary cut” was defined through the
ternary composition diagram (arrow in Figure 82 a)). Along this line, Cu and Pd contents
vary, while the Fe content is constant at 70 at. %. Figure 82 b) illustrates the structural change
of these samples, in the form of a colour-coded (red = high intensity, blue = low intensity) top
view of the diffraction patterns. A martensitic structure with (111) and (200) peaks is
observed at room temperature up to 6 at.% Cu. No additional peaks originating from
precipitate phases are observed. For Cu contents > 6 at.%, a (110) bcc phase occurs in
addition to the transforming fct/fcc phase. At 8 at.% Cu, the (110) Fe8Cu2 peak appears. Due
to an increase of Cu at the expense of Pd content, a slight shift of the (111) fct/fcc peak is
observed over the whole compositional range as shown in Figure 82 b).
While austenitic Fe70Pd30 has a lattice constant of a = 0.3756 nm57, alloying with small
amounts of Cu should change this. The lattice constant a of the austenite fcc phase was
determined from the (200) fcc peaks of selected Fe-Pd-Cu samples using XRD at 373 K.
Figure 83: Calculated and experimentally determined lattice constants for fixed Fe contents (70 to 72 at.%) in
dependence on Cu. Inset: Lattice constants a of the fcc cubic austenite phase (measured at 373 K) for single-
phase transforming Fe-Pd-Cu samples within the composition diagram. (Lattice parameters calculated by M. E.
Gruner and originally published in Ref. 199)
124
Results are shown in Figure 83 for experimental (black) and calculated (red) lattice
parameters for selected Fe-Pd-Cu samples with a constant Fe content (Fe70: squares; Fe71:
triangles; Fe72: circles). The inset shows the locations of the experimental samples in the
ternary composition diagram. Since the covalent atomic radius for Cu (0.128 nm) lies
between that of Fe (0.126 nm) and Pd (0.137 nm), a decrease of lattice constants with addition
of small amounts of Cu at the expense of Pd should occur.188 As expected, the lowest lattice
constant is observed for samples with the smallest Pd and highest Cu content. However, the
lattice parameters of the Fe-Pd system are known to deviate from Vegard´s law198 and at
elevated temperatures the Invar-behaviour of Fe-Pd leads to a significant reduction of the
thermal expansion coefficient in a broad range of temperature. In fact, the calculated values
for a (determined for the ferromagnetic ground state) are slightly higher than the lattice
constants measured experimentally. Since calculations represent the state at 0 K and
measurements were performed at 373 K, this is opposite to what one would expect from
normal thermal expansion and thus another indication for the presence of Invar anomalies in
these films. This observation is of importance for the interpretation of the influence of
substrate-induced stress on the martensitic transformation and will be discussed in the
following.
Transmission electron microscopy
TEM was used to confirm the existence of a single ternary phase in the sample
Fe69.5Pd26.7Cu3.8. The prepared lamella (not shown) consist of the following layers: SiO2
substrate layer, a thin FeOx intermediate layer, the homogeneous Fe-Pd-Cu layer and a Pt-
protection layer deposited during FIB-preparation. The thin oxidized Fe layer at the film-
substrate interface is related to diffusion effects during annealing and has a thickness of
approximately 40 nm. The grain structure of the layer, revealed by diffraction contrast,
verified a homogenous chemical composition in the Fe-Pd-Cu layer. Figure 84 a) shows the
HRTEM image of a selected region of the Fe-Pd-Cu layer. The martensitic twin structure can
be seen in the lattice image as well as on the Fast Fourier Transform (inset b) in Figure 84).
125
Figure 84: High resolution transmission electron microscope (HRTEM) image showing twinned lattice planes
of a Fe69.5Pd26.7Cu3.8 film. The inset shows the fast Fourier transform revealing the twinned microstructure. (FIB
preparation was performed by A. Sehrbrock and TEM by S. Irsen at caesar Bonn and the figure was originally
published in Ref. 199)
Summarizing these results, Cu is homogeneously dissolved in the Fe-Pd matrix. In the
investigated area, no composition or phase segregation was detected. The results were
confirmed by Auger-electron spectroscopy depth profiles.
Magnetic properties of Fe-Pd-Cu thin films
Since determination of magnetic anisotropy is hampered by the polycrystalline microstructure
of the films, the experimental investigations concentrate on the other key properties, the
saturation polarization, JS and the Curie temperature TC. In Figure 85 magnetic measurements
for a Fe71Pd27.7Cu1.3 (open circles) and a Fe70.2Pd25.5Cu4.3 (filled squares) thin film are
presented as examples. Figure 85 a) shows the magnetic hysteresis curve in the martensitic
(blue symbols; at 243 K) and the austenitic (red symbols; at 393 K) state of both samples. As
explained in the previous chapter, the martensite phase shows a significantly higher saturation
polarization JS than the austenite phase.
In Figure 85 b) the normalized magnetization J/J50 K as a function of temperature between
315 K and 500 K and under a constant magnetic flux density µ0H = 0.05 T is depicted for
both samples.
126
Figure 85: a) Magnetic hysteresis measurement for the martensite (243 K) as well as for the austenite (393 K)
state of a Fe71Pd27.7Cu1.3 (open circles) and a Fe70.2Pd25.5Cu4.3 (filled squares) thin film. b) Normalized
magnetization J/J50 K as a function of temperature and under a constant magnetic flux density of µ0H = 0.05 T is
presented. The data points are fitted using Kuz’min’s equation for the Fe71Pd27.7Cu1.3 (open circles and red curve)
and the Fe70.2Pd25.5Cu4.3 (filled squares and blue curve) thin film.
Again using the Kuz’min relation, a fit function (red: Fe71Pd27.7Cu1.3; blue: Fe70.2Pd25.5Cu4.3)
was applied through the data points and extrapolated to the temperature, where J/J50 K
approximates zero. The determined values for both JS and TC are shown in Figure 86. Figure
86 a) presents the saturation polarization as a function of Cu content (colour coding indicates
the Fe content) for both the martensitic (squares, determined at 243 K) and the austenitic
(triangles, determined at 393 K) structure.
Figure 86: a) Saturation polarization for both the martensite fct and austenite fcc phase determined at 243 K and
393 K respectively, in comparison to the calculated ground state total (spin + orbital) magnetic moments, for
selected compositions along the path marked in Figure 82 a). The straight line through the experimental data
illustrates the deviations from a linear behaviour. b) Normalized values of Curie temperature TC in dependence
on Cu content determined by calculation and experiments. (Values for JS and TC were calculated by M. E.
Gruner and originally published in Ref. 199)
127
Error bars indicate a maximum variance < 5% of JS. Generally, the saturation polarization for
Fe-Pd-Cu samples is about 10% lower than for Fe70Pd30 (1.5 T for martensite and 1.35 T for
austenite phase). The higher JS of the martensite phase compared to the austenite reflects the
expected decrease of spontaneous magnetization with increased thermal fluctuations. For both
martensite and austenite, JS has a general tendency to decrease with increasing Cu content,
which is in agreement with the composition dependence of the ground state magnetic
moments obtained from ab initio calculations (performed by M. E. Gruner). However, for Cu
contents from 5 to 6 at.% the experimental JS values increase abruptly to a maximum of 1.1 T
for the austenite and 1.35 T for the martensite phase. Although a direct evidence of large
precipitates was not observed with XRD, the rather unexpected increase of JS indicates that
the formation of small precipitates below the sensitivity of XRD already starts at 5 at.% Cu.
Below 5 at.% the homogeneous solution of Cu, which has a filled d-shell and thus shows only
a negligible induced spin-polarization on the order of 0.08 µB, decreases magnetization. At Cu
contents between 5 and 6 at.%, decomposition of the Fe-Pd-Cu sample into a Fe70Pd30 phase
with a high JS and Cu-rich precipitate phases results in the overall increase of magnetization.
When the Cu content increases to values > 6 at.% the volume fraction of non-magnetic
precipitates increases and thus leads to a decrease in JS. This agrees well with observations for
the Fe-Pd-Cu splats with a Cu content > 5 at.% as shown in the following. Figure 86 b) shows
the experimentally determined and calculated Curie temperatures, TC, as a function of the Cu
content, which are in both cases presented relative to TC of binary Fe70Pd30. Experimentally,
TC was extrapolated from a fit of the temperature dependence of saturation curves to
Kuz’min’s parameterization.151,152 Because accurate results cannot be achieved by a
simultaneous fit of the shape parameters s and p and the critical exponent of the
magnetization β, while determining the spontaneous magnetization M0 and TC by
extrapolating the magnetization curve, the values were restricted to s = 1, p = 5/2 and β = 1/3
as determined by Buschbeck et al..67 The choice of a large value s=1 (also valid for other 3d
transition metals) is consistent with the largely decreased spin-wave stiffness of Fe-Pd alloys
reported in the vicinity of the fcc-fct transition.200 It should also be noted, that the calculated
TC values are systematically too large on an absolute scale, since M. E. Gruner used the
mean-field approximation to the Heisenberg-model, which furthermore neglects Invar-typical
longitudinal spin fluctuations. Nevertheless, both approaches show the same trend - a slight
decrease as a function of composition for Cu-concentrations < 4 at.%. This is again expected
from the dilution of the magnetic sites by the Cu-atoms, but it obviously does not lead to a
severe degradation of the magnetic properties. Between 4 and 5 at.% Cu, a kink to lower
128
TC/TC(Fe70Pd30) ratios is observed in the experimental data - similar to the anomaly in the
magnetization presented in Figure 86 b). This kink is again attributed to the decomposition of
the single ternary phase into a multiphase structure for Cu > 5 at.%. The highest TC of 560 K
for a single phase is observed at a composition with 72 at.% Fe. This value is slightly lower
compared to 600 K for Fe70Pd30.123 Samples having a Cu content of 5 at.% exhibit a smaller
TC - but following the previous arguments this is very likely attributed to the formation of a
complex phase mixture. In addition, we expect uncertainties originating from slight changes
of the composition due to the limited accuracy of the EDX measurements and the formation
of interfacial layers.
Decisive for a large MFIS is a strong coupling between the orientation of magnetic moments
and the tetragonal distortion of the martensitic twins, which usually arises from spin-orbit
coupling. This can be quantified in terms of the magnetocrystalline anisotropy energy, which
is connected to the change of orbital moments upon variation of the magnetization direction.
These quantities are difficult to measure in a high-throughput approach and therefore further
experimental characterization of orbital magnetism of the Fe-Pd-Cu system must be left for
future work. Fully relativistic first-principles calculations as employed in this investigation,
however, provide a straight forward approach to grasp the principal trends of related
quantities, such as orbital moments in dependence on composition. Figure 87 provides the
variation of the total orbital moment per atom for all compositions of interest. The obtained
variations are small and at the limits of the methodological resolution. Nevertheless, a few
discernable trends are discussed below.
Figure 87: Total orbital moments µorb of Fe-Pd-Cu as a function of composition obtained from first-principles
calculations. Only moderate variations of µorb are encountered within the martensitically transforming
concentration range. (Calculated by M. E. Gruner and originally published in Ref. 199)
129
The total orbital moment of the binary reference composition Fe70Pd30 is with 0.054 µB/atom
among the largest value for the considered concentration range and thus it could be concluded
that addition of Cu does not significantly affect the magnetocrystalline anisotropy of the
material. Slightly larger values are only obtained for compositions with an Fe content of about
72 at.% and low Cu content. The Fe-species provide the largest orbital moment of about 0.065
µB/atom with an overall variation on the order of 1%. Thus, changes of its fractional
contribution according to composition will provide the most relevant trend. The Pd-atoms
exhibit an orbital moment of about 0.025 µB/atom (with a variation of 5% over all
compositions), which is proportional to the magnitude of the Pd-spin moment. The latter, in
turn, is induced by the surrounding Fe moments. This raises the expectation that a large Fe
content should have a beneficial influence on the magnetocrystalline anisotropy, since it is
commonly traced back to the hybridization of the 3d and 4d electrons in this alloy. Cu
exhibits only a small induced moment, which is connected with a rather tiny orbital
contribution of the order of 0.01 µB/atom.
Stress effects in Fe-Pd-Cu thin films
In comparison to bulk samples, thin films exhibit considerably higher transformation
temperatures, regardless of their composition. This can partly be attributed to the high stress
state of the films due to the different thermal expansion coefficients of thin film and substrate.
It is known that Fe-Pd FSMA thin films on SiO2 and MgO substrates are subject to high
tensile stresses.201,202 These lead to stress-induced martensitic transformations at
systematically increased transformation temperatures. Kato et al.22 estimated a proportionality
factor of 4.8 MPa/K for the Fe-Pd system. From other experiments on Fe70Pd30 thin films
fabricated and processed in a similar way as in the present study, tensile stresses in the range
of 0.2 GPa to 0.3 MPa after annealing were observed.61 These values for tensile stress are
sensitive to the composition and decrease systematically with increasing Fe content.203 This
behaviour was attributed to the Invar effect, which is defiend as the anomalous reduction of
thermal expansion, and is also present in Fe-Pd alloys in this composition range. The Invar
effect originates from a thermodynamic repopulation of electronic configurations, which are
characterized by different local magnetic moments and equilibrium volumes. This effect
becomes increasingly prominent for Fe-rich compositions204 and may even lead to negative
thermal expansion coefficients. The Invar properties of the ternary alloy are - apart from the
changing Fe content - also influenced by Cu addition. As measurements of thermal expansion
coefficients are difficult in film/substrate composites, the spontaneous volume
130
magnetostriction ωs0 was calculated by M. E. Gruner to give a qualitative indication. The
quantity ωs0 is defined as the relative change (VFM-VPM)/VPM between the volume VFM of the
ferromagnetic phase and the volume VPM of the paramagnetic phase, both extrapolated to T =
0. Alternatively, ωs0 can be interpreted as the relative volume expansion induced by the onset
of ferromagnetism below TC, which is measured in comparison to a (hypothetical) reference
alloy which remains paramagnetic at all temperatures.
Figure 88 shows the variation of ωs0 as a function of the composition, demonstrating an
increase of up to 10% with Fe concentration as the dominating trend in the transforming
composition range. As a secondary trend, at constant Fe content, an increase of ωs0 with
increasing Cu concentration is observed, which is opposite in direction with respect to the
corresponding valence electron concentration e/a. This clearly demonstrates that the moment-
volume interaction in the ternary system cannot be consistently interpreted as a function of e/a
alone. It further can be concluded that increasing Fe as well as Cu contents, will enhance the
Invar-typical anomalies and thus further reduce thermal expansion and elastic constants below
TC. Thus, the addition of Cu can be expected to further decrease the tensile stress compared to
a binary film with the same Fe content.
Figure 88: Calculated spontaneous volume magnetostriction ωs0 of Fe-Pd-Cu as a function of composition. A
clear increase of ωs0 of about 10% is encountered for increasing Fe concentrations. (Calculated by M. E. Gruner
and originally published in Ref. 199)
The increase of Invar-related anomalies and the decrease of thin film stress due to the addition
of Cu was further investigated by synchrotron-based measurements. Residual stress
measurements using the sin2(ψ) method were used to determine the stress state of Fe-Pd-Cu
131
thin films in the martensite (300 K) as well as in the austenite (393 K) state. Figure 89 shows
area detector diffractograms at two different temperatures of a Fe71Pd26Cu3 thin film.
Figure 89: Area detector diffractograms of a Fe71Pd26Cu3 thin film at different temperatures. a) At 300 K the
thin film is in the martensite phase proved by the presence of the (200) and (002) fct peaks. When heating the
sample to 393 K in b), the sample transforms into the austenite phase and the (200) fcc peak appears. The red
stars denote diffraction intensities originated from the Si/SiO2 substrate. (Synchrotron measurements were
performed by H. Brunken)
132
The area diffractograms depict the colour-coded intensity (blue for low and white for high
intensity) in terms of X-ray counts as a function of d-spacing and ψ angle. At an angle of ψ =
90° the ψ-vector points to the thin film surface at an angle of 90° and thus is parallel to the
thin film normal vector. This is indicated by a dashed white line, that divides the upper half
(90°-160°) to one half-sphere and the lower part (20°-90°) to the other half-sphere above the
sample. Instead of intensity rings known for polycrystalline samples intensity spots appear.
These spots originate from texturing and correspond to the specific d-spacings of different
phases. In Figure 89 a) the synchrotron diffraction pattern of a Fe71Pd26Cu3 thin film in the
martensite phase at 300 K is presented. The martensite phase can be identified by the presence
of the (220), (002), (200) and (111) fct diffraction peaks. Upon heating to 393 K, the thin film
transforms into the austenite phase and develops a (200) fcc peak. In order to determine the
residual stress state of this sample in the austenite phase, the lattice parameter afcc of the thin
film was determined from the (111) fcc peak at different ψ-angles.
Figure 90: Lattice parameter as a function the ψ-angle for a Fe71Pd26Cu3 thin film. a) A linear function was
fitted to the data. The positive slope indicates a tensile strain state in this sample. For every lattice spacing afcc in
a) that corresponds to a diffraction peak, the FWHM b) and the intensity c) for this peak were determined.
An example of the variation of lattice parameter with tilt angle afcc = sin2(ψ) is shown in
Figure 90 a). From the slope of this curve the stress state and the strain in this thin film can be
133
determined. The data points are fitted by a linear function and the slope of this fit function is
used to determine the residual strain. The positive slope indicates a tensile residual strain that
appears for thin films showing Invar-related anomalies having smaller thermal expansion
coefficient than their substrate. In Figure 90 b) the FWHM and c) the intensity of the (111)
fcc diffraction peak at different ψ angles and thus as a function of sin2(ψ) is shown to ensure
no texture related errors. Since the sin2(ψ) method only reveals the strain of the sample, the
Young’s modulus needs to be determined separately for each thin film. This was performed
by nanoindentation measurements of the Fe-Pd-Cu thin films. The measurements were
conducted at ambient and elevated temperatures (353 K) to determine the Young’s modulus
in the austenite phase. A summary of nanoindentation measurements is depicted in Figure 91.
For these measurements the Young’s modulus was determined for an indentation depth
ranging from 50 nm to 75 nm. This indentation depth follows the rule that up to an
indentation depth of one-tenth of the overall film thickness, the influence of the substrate is
negligible.154 Presented are the a) Young’s modulus and b) the thin film hardness as a
function of Pd content with colour-coded Cu content at 353 K. Figure 91 b) shows hardnesses
values of the thin films revealed from nanoindentation measurements in dependence on Pd
content. Since the error bars in both measurements are quite large, no significant trend can be
observed. As mentioned previously, the amount of Cu increases at the expense of the Pd
content in the Fe-Pd-Cu system. Since Pd and Cu have similar Young’s moduli (EPd = 121
GPa and ECu = 117 GPa) and number of valence electrons (eCu = 11 and ePd = 10) it can be
expected that the Young’s modulus of the Fe-Pd-Cu single-phase does not vary significantly
with changing Cu content. The same behaviour can be expected for the variation of hardness
with composition. All values for the Young’s modulus are higher than the values determined
for a binary Fe70Pd30 reference thin films that was fabricated and processed under identical
conditions. The values determined for the Young’s modulus in Figure 91 a) were used to
calculate the residual stress in the Fe-Pd-Cu single-phase thin films. Figure 92 depicts the
residual stress as a function of the Cu content for samples with a constant Fe content of 71
at.%. The colour-coding indicates the martensite start temperature Ms. A significant decrease
in residual tensile stress is observed with increasing amount of Cu. The values for low Cu
content in the range of 1 at.% exhibit high tensile stress values, while Ms is in the range of
343 K to 353 K. Upon further increase of Cu content, the amount of tensile stress lowers
significantly while Ms increases.
134
Figure 91: Summary of nanoindentation measurements at 353 K for Fe-Pd-Cu single-phase thin films. Presented
are the a) Young’s modulus and the b) thin films hardness as a function of Pd content with colour-coded Cu
content. No distinct trend can be observed with changing composition. Further binary Fe70Pd30 reference thin
films that were fabricated and processed under identical conditions are included in both diagrams.
At a Cu content of 1.6 at.% the amount of tensile stress is lowered to 673 MPa. Thin films
with a Cu content > 1.6 at.% exhibit Ms temperatures > 350 K. The data point with a Cu
content of 3.9 at.% has a tensile stress amount of 667 MPa and shows a significant decrease in
Ms. This decrease can be attributed to the fact that this sample has a Cu content that is close to
the limit (Cu > 4 at.%) where the Fe-Pd-Cu single-phase starts to decompose by forming Cu-
rich precipitates. It cannot be ruled out that this thin film already shows a tendency on the
nanoscale to decompose. However, no hints regarding this were observed.
135
Figure 92: Residual stress as a function of Cu content. To separate the impact of Fe on thermal expansion
behaviour and thus the stress state, only thin film samples with a constant Fe content of 71 at.% are shown. Upon
addition of Cu a significant decrease of tensile stress is observed. The Ms temperature (colour-coded) does not
change significantly and has highest values for the samples with lowest tensile stress.
Considering these findings, it can be concluded that the addition of Cu decreases the thin film
residual stress state. The Ms temperature does not drop for the thin films with smallest
amounts of tensile stress, indicating that the increase in transformation temperatures upon
addition of Cu in Fe-Pd is not controlled by the formation of stress-induced martensite. The
presented tensile stress values are high, when compared to literature values for binary Fe-Pd
thin films.61 Nevertheless, the considerable increase of volume magnetostriction revealed
from ab initio calculations and the decrease of thermal expansion with increasing Cu content
is a good explanation for these high tensile stress values. Further investigations of the thermal
expansion coefficient of these samples would be interesting. This is not possible with these
samples, since the substrate constrains the thermal expansion expansion of the film.
4.2.3.2 Fe-Pd-Cu Splats
Substrate constraints are not present in bulk-like samples. Therefore it is desirable to cross-
check the findings for corresponding systems. In this part Fe-Pd-Cu splats were fabricated,
processed and investigated. All splats were fabricated and annealed by I. Kock from the
Georg-August-Universität Göttingen. The samples were produced by the splat-quenching
technique to obtain thin foils in a metastable phase and to avoid demixing during cooling. It is
known that a high cooling rate is beneficial for the formation of the fct phase.52 The splats
136
were annealed to improve their microstructure and transformation behaviour. The influence of
quenching and annealing on binary Fe70Pd30 splats was described in Chapter 4.1.2. Figure 93
presents room temperature XRD data of different splats with the nominal compositions of
Fe70Pd30, Fe68.4Pd29.3Cu2.3 and Fe70Pd23.8Cu6.2. All splats were annealed at the same
temperature and time (1123 K, 1 h) as used for the thin film materials libraries. In order to
investigate the stability of the metastable phase in the Fe-Pd-Cu splats, cooling rates were
varied and compared with the results from Fe70Pd30 splats. For rapid cooling (> 100 K/s), the
splats, which were sealed in quartz tubes under inert gas, were water-quenched. For slow
cooling, the splats were cooled in air (< 0.5 K/s), while for very slow cooling the splats rested
in the furnace after annealing until room temperature was reached (cooling rate < 0.2 K/s,
temperature after 12 h ~300 K). The binary Fe70Pd30 splats show the transforming Fe70Pd30
phase for both the quenched and air cooled samples, indicated by the (111) and (200) fcc
peaks. No additional precipitate phases are observed. The oven-cooled Fe70Pd30 splat shows
fractions of the Fe50Pd50 and α-Fe phase besides the transforming Fe70Pd30 phase. The
Fe68.4Pd29.3Cu2.3 splats, that were water-quenched and air-cooled, show a single-phase
structure in Figure 93.
Figure 93: XRD patterns of Fe70Pd30, Fe68.4Pd29.3Cu2.3 and Fe70Pd23.8Cu6.2 splats processed with different cooling rates after annealing (Quenched in water, air-cooled and oven-cooled). (Splats were fabricated and annealed by I. Kock and originally published in Ref. 199)
Two additional peaks from Fe50Pd50 and Fe8Cu2 phases occur only for the very slowly cooled
splats, with the highest fraction occurring in the Fe70Pd30 sample. Comparing the (111) fcc
and (111) Fe50Pd50 peaks for the oven-cooled Fe70Pd30 to the Fe68.4Pd29.3Cu2.3 splat, a
137
significant decrease in the (111) fcc peak intensity (and corresponding increase in (111)
Fe50Pd50 peak intensity) is observed. Thus a high fractional decomposition from the
transforming Fe70Pd30 phase into the Fe50Pd50 equilibrium phase with precipitates occurs for
the binary Fe70Pd30 splat upon very slow cooling. By addition of Cu, this decomposition can
be partially suppressed. In contrast, the water-quenched Fe70Pd23.8Cu6.2 splat does not show
the single transforming phase: instead the (111) fcc/fct peak is shifted to higher angles and
shows a broad peak at 42.9° which is correlated to the (110) bcc phase. For the oven-cooled
sample, a further decomposition into precipitate phases (Fe50Pd50 and Fe8Cu2) occurs. Thus
the decomposition of the high-temperature and metastable phase to equilibrium phases is
suppressed by the addition of small amounts of Cu into binary Fe-Pd FSMA even for slow
cooling rates. In contrast, Fe70Pd23.8Cu6.2 splats do not show a single ternary phase due to
decomposition upon cooling.
Figure 94 a) shows the change of lattice constants as a function of temperature during the
martensitic transformation for a water-quenched Fe68.4Pd29.3Cu2.3 splat. Lattice constants were
determined from XRD(T) measurements by peak fitting. The error for these results is smaller
than the size of the symbols presented in Figure 94 a). From 128 K (cfct = 0.360 ± 0.001 nm,
afct = 0.382± 0.001 nm) to 233 K, only the low temperature tetragonal martensite (fct) phase
occurs. Within this range, an increase of c and a decrease of a is observed as the structure
comes close to the transformation temperature. At 230 K the transition from the martensite to
austenite phase starts, along with the appearance of (200), (220) and (311) fcc peaks (not
shown here). With further increasing temperature, the (200), (002), (202) and (311) martensite
peaks decrease and finally disappear at 238 K. When the temperature increases further a
growing of the (200), (220) and (311) fcc austenite peaks is observed, which finishes at 278
K. At 300 K the lattice constant for the austenite phase was determined to be afcc = 0.374 ±
0.001 nm. In Figure 94 b) the linear thermal expansion coefficient as a function of Cu content
for splat samples with different compositions is presented. The colour-code indicates the Fe
content of the splats. The linear thermal expansion coefficient for every Fe-Pd-Cu splat was
calculated from XRD(T) measurements. The variation of lattice parameters afcc of the
austenite phase was determined as a function of temperature between 220 K and 400 K and
fitted by using a linear function.
138
Figure 94: a) Change of lattice constants by the transformation from the low temperature (fct) to the high
temperature phase (fcc), determined by XRD(T) for a water-quenched Fe68.4Pd29.3Cu2.3 splat. b) Linear thermal
expansion coefficients as a function of Cu content for splat samples with different compositions. Colour-coding
indicates the amount of Fe content in the different splats. The addition of Cu produces a significant decrease of
the linear thermal expansion coefficient. (Figure a) originally published in Ref. 199)
The error for the linear thermal expansion coefficient is smaller than the symbol size in Figure
94 b). Starting from a Fe70Pd30 reference splat that was fabricated and processed identically
and is in good accordance with literature values, the thermal expansion coefficient decreases
upon addition of Cu for a Fe68.6Pd29.8Cu1.6 splat. To separate the impact of Fe on the variation
of the thermal expansion the Pd content is kept constant at 29.8 at.% for the following Fe-Pd-
Cu splats. Although the Fe content is decreasing from 68.6 at.% to 67.9 at.% the thermal
expansion coefficient further decreases to a value of 1.4 x 10-6 K-1, which is well below the
value for the Fe70Pd30 splat. These findings further corroborate the previously presented
findings for Fe-Pd-Cu thin films.
139
At 128 K the Fe68.4Pd29.3Cu2.3 splat has a tetragonality of c/a = 0.942, which is higher than
reported for a Fe68.8Pd31.2 single crystal.37 Due to the change of the lattice constants for
Fe68.4Pd29.3Cu2.3, a shape change consisting of an expansion of (a128K - a300K)/a300K = 2.14%
along the a-axis and a contraction of (c128K - a300K)/a300K = 3.74% can be estimated. Thus a
maximum strain caused by conversion of the variants upon cooling from 300 K to 128 K of
approximately 5.88% can estimated. This is slightly lower than the value of about 6% known
for the Fe68.8Pd31.2 single crystal. Since these investigations were performed on polycrystalline
samples, where grain boundaries hinder magnetic field induced strain and the influence of
micro-stress on the transformational behaviour was not investigated, these values just give an
upper estimate for MFIS in Fe68.4Pd29.3Cu2.3.
Variation of martensitic transformation in the Fe-Pd-Cu system
Fe-Pd-Cu thin film materials libraries were fabricated covering a range of 40 < Fe < 95 at.%,
5 < Pd < 55 at.% and 1 < Cu < 18 at.% and exhibiting transformation temperatures of up to
359 K. TEM confirmed the presence of a single ternary phase and showed the occurrence of
nano-twins. XRD investigation of Fe68.4Pd29.3Cu2.3 splats, processed with different cooling
rates after annealing, showed an enhanced stability against decomposition of the high-
temperature metastable phase in comparison to Fe70Pd30. Ab initio calculations and
experiments reveal the same trends for TC and JS, which are not significantly affected by the
dilution of the magnetic sites in the ternary alloy. Slightly decreased orbital moments and an
enhanced volume magnetostriction are predicted from ab initio calculations for Cu-enriched
compositions.
The presented findings deliver a comprehensive overview of the compositional dependence of
the martensitic transformation temperatures. The complete analysis, however, must consider
that a systematic influence by substrate induced-stresses may be present. On the other hand,
the fabrication method used for bulk-like splats is not feasible for producing the large amount
of data necessary to reveal the relevant trends. Therefore, it is worthwhile to combine the
results of both approaches with previous results from literature for the binary Fe-Pd system.
This has been done in Figure 95 a), which summarizes the martensite start temperature, Ms, as
a function of the composition in a three-dimensional diagram. The composition is defined in
terms of the valence electron concentration e/a (see Figure 80 for the whole materials library)
and Fe content. In the introduction, these quantities were identified as the two most influential
parameters determining the physics of the martensitic transformation in the ternary alloy.
Again, the full three-dimensional representation (Figure 95 b)) is difficult to interpret. Thus,
140
the discussion is restricted to projections of the data onto each of the three base planes, which
cover the essential physics.
Figure 95: Martensite start temperatures, Ms, as a function of e/a ratio and Fe content presented in a 3D graph
with colour-coded Cu content. Binary systems are shown by black symbols (squares and triangles). For clarity,
only projections of the 3-dimensional data onto three different planes are depicted in a); the spatial distribution
of the data is given in b). (Circles refer to the Ms obtained from thin film experiments in this study, diamonds
refer to values obtained for splats. For comparison, literature data for binary alloys57,143 are included (black open
squares and black filled triangles). Black lines are guides to the eyes and visualize general compositional trends
for the three kinds of data. Red dotted lines mark the border lines beyond which Cu-rich ternary thin films
decompose into different phases. Outside of this volume, only data points corresponding to single-phase samples
are presented. (Figure originally published in Ref. 199)
These are shown exclusively in Figure 95 a). The bulk samples (69 at.% < Fe < 71 at.% and
8.575 < e/a < 8.613), published by Cui et al.57 show a linear distribution within all three
planes, depicted by a black dashed line. For the Fe-Pd-Cu thin film martensite start
temperatures, a clear linear dependence on both the Fe content and the e/a ratio is observed.
The slope for these ternary thin films is considerably lower than for the binary reference data.
As discussed before, this can be traced back to the tensile stresses, which are largest for the
samples with the smallest Fe content and decrease significantly with increasing Fe content.
Tensile stresses can also be partially responsible for the considerable increase of the
transformation temperatures with respect to bulk material. Splats as bulk references show
much lower transformation temperatures in comparison to Fe-Pd-Cu thin films. For
141
Fe68.4Pd29.3Cu2.3 MS is 334 K for the thin film and 238 K for the splat, while for the binary
composition, MS of splat and bulk systems coincide. If the observed difference of Ms between
film and splat is solely attributed to the stress state of the film, we obtain a value in the range
of 460 MPa using the factor of 4.8 MPa/K determined by Kato et al..22 A maximum Ms of 359
K was observed for a sample with an Fe content of 71.8 at.% and an e/a ratio of 8.58. This is
one of the highest value obtained in the Fe-Pd system so far and significantly higher than the
values obtained for corresponding binary systems,143 demonstrating that the addition of Cu
can substantially improve the relevant FSMA properties. This trend cannot be ascribed to a
stress-related variation of the martensitic transformation temperature, since, as shown before,
the Invar-related reduction of thermal expansion, which partially compensates tensile stresses,
is in fact enhanced in the Cu-doped alloys. The clear linear dependence between Ms and the
compositional parameters vanishes if the Fe content reaches a threshold value (indicated in
Figure 95 by red dotted lines), below which no transforming single component samples were
obtained. Above the critical concentration, the stability region is depicted by a triangle
(marked by the broken black lines on the bottom plane of Figure 95). In comparison to results
concerning binary bulk samples57 and thin films143 the compositional range in which single-
phase transforming samples can be found is significantly extended to lower Fe contents. From
the discussion of the binary alloys and ternary thin films alone, it cannot be decided whether
the Fe content or the valence electron concentration represents the dominating variable which
needs to be optimized in order to improve the FSMA behaviour in a ternary system. However,
a conclusive indication comes from the splat experiments. Here, the Ms value of the ternary
splats coincides with the extrapolation of the data from the work of Cui et al.,57 if plotted
versus the Fe content, while it deviates significantly for larger e/a values from the respective
projection of the bulk values. This interpretation is also supported by the thin film data of
Inoue et al..143 These, however, do not show a well-defined linear behaviour. This may be
related to the variation of the fabrication parameters and the changing film thicknesses of the
samples. Nevertheless, in the range between 70.5 and 71.5 at % Fe, the Ms values of the
binary thin films (film thickness 200 nm) fall almost on top of Fe-Pd-Cu data if plotted
against the Fe content, but deviates for the e/a projection. The films with low Fe content agree
better with the bulk curve because of the higher film thickness (2 and 4 µm).
Based on these observations, it can be concluded, that the addition of Cu is a very promising
route to develop improved Fe-Pd-based FMSAs. Cu extends the stability of the
thermoelastically transforming phase to Fe concentrations up to 72 at.% and therefore is
142
expected to increase the martensitic transformation temperatures to values which are suitable
for technological applications.
4.2.3.3 Epitaxial Fe-Pd-Cu thin films
This chapter presents the fabrication of epitaxial Fe70Pd30-XCuX (X = 3, 7 at.%) thin film
samples and investigates the influence of Cu alloying on the structural and magnetic
properties. The motivation for this was to grasp the idea of coherent epitaxially grown Fe-Pd
thin films on varying buffer layers from Buschbeck et al.67 and S. Kauffmann-Weiss et al.206
and to apply this to new Fe-Pd-based FSMAs found by combinatorial materials science. One
aim was to gain a deeper insight and to further investigate the new Fe-Pd-Cu system with
regard to properties, which are hardly to measure in polycrystalline thin films. Most
promising Fe-Pd-Cu compositions were chosen by the author and fabricated by using
coherent epitaxial growth.
All epitaxial thin film were fabricated and partially investigated at the IFW Dresden in close
collaboration with S. Kauffmann-Weiss in the group of S. Fähler during a scientific stay of
the author. Further all presented findings were developed in close collaboration with
S. Kaufmann-Weiss. The calculated results shown in this chapter were carried out by M. E.
Gruner in the group of Prof. P. Entel from the University of Duisburg-Essen. Findings from
calculated results were further developed in close collaboration with M. E. Gruner.
Single crystals are well suited to determine materials properties such as the
magnetocrystalline anisotropy, which can be affected by finite size, texture and stress effects.
These are difficult to measure in polycrystalline samples. Nevertheless single crystals are
difficult to fabricate and often expensive as large amounts of the material are needed.
Epitaxial films represent the thin-film counterpart to bulk single crystals and allow
determination of the anisotropic magnetic properties. Furthermore, coherent epitaxial growth
on cubic buffer layers with different lattice parameters can be used to adjust the tetragonal
distortion of the martensite as explained in Chapter 2.6. This approach allows the c/abct ratio
to be controlled almost along the entire Bain transformation path between bcc (c/abct = 1) and
fcc (c/abct = 1.41). Through this kind of epitaxial film growth, artificial single-variant states
are realized, allowing K to be measured along the different crystallographic directions (K1 and
K3).67 Coherent epitaxial growth causes the lattice constants of the buffer to alter the in-plane
lattice parameters of Fe70Pd30-XCuX films and thus c/abct ratio. The c/abct was controlled by
depositing Fe70Pd30-XCuX films on Cr (1.07), Au (1.09), Ir (1.31), Rh (1.34) and Cu (1.57)
buffer layers that were previously deposited on MgO substrates. The intended structural
143
orientation relation between substrate, buffer and Fe70Pd30-XCuX thin film is schematically
depicted in Figure 96 for each different buffer material.
Figure 96: Structural orientation relation between MgO substrate, buffer materials and Fe70Pd30-XCuX thin film.
Next to a) Cr as buffer material, all following buffers b) Au, c) Ir, d) Rh and e) Cu have a Cr adhesion layer
below. With varying buffer material the c/abct ratio increases as illustrated by a tetragonal distortion of the
Fe70Pd30-XCuX unit cell.
In order to analyse the crystal structure of Fe70Pd30-XCuX films grown on different buffer
layers, XRD scans in Bragg-Brentano geometry were performed (Figure 97). This method is
only suitable to reveal the out-of-plane lattice parameter. With changing buffer material from
Au to Cu, the position of the (002)fcc diffraction peak of the fcc buffers (blue pentagons)
increases to higher 2θ angles. By using a bcc buffer, such as Cr, the peak position of the (002)
diffraction peak occurs at considerably higher 2θ angles.
This reflects the shift of the XRD peak with the lattice parameter for several buffer layers
with different structures. When varying the buffer material, the position of the (002)
diffraction peak of Fe70Pd27Cu3 (red circles) shifts from 2θ = 70.5° (c/abct = 1.07) to a lower
144
value at 2θ = 60.6° (c/abct = 1.31). For a film on the Cu buffer (c/abct = 1.57), the (002) peak
shifts to 2θ = 52.6°.
Figure 97: Bragg-Brentano scans of Fe70Pd27Cu3 films on different buffer layers. Red circles mark (002)
diffraction peaks of Fe-Pd-Cu films and blue pentagons indicate (002) diffraction peaks of the cubic buffers.
With increasing c/abct ratio, a shift of (002) Fe-Pd-Cu peaks to lower angles is observed. Red dotted lines indicate
the boundaries of the Bain transformation path. (Figure originally published in Ref. 205)
These XRD measurements show the presence of the (002) diffraction peak only in the
Fe70Pd27Cu3 thin films, which indicates a highly textured microstructure. The absence of any
further diffraction peaks suggests a coherent epitaxial growth of the Fe70Pd27Cu3 films. Since
the unit cell volume remains almost constant, this results in an increase of the out-of-plane
lattice parameter due to the shrinking of the in-plane lattice parameter. The in-plane lattice
parameters were measured by conducting 2θ scans of the (101)bct lattice planes under tilted
conditions at a tilt angle ψ. The (101)bct diffraction peaks were then used together with the
measurements in Bragg-Brentano geometry to calculate the in-plane lattice parameter of the
films. The results revealed identical in-plane lattice parameters for the Fe70Pd27Cu3 films and
the underlying buffer material (Figure 98 a)). This indicates a coherent growth in all thin films
deposited on the different buffers and reveals the following relation: dbuffer = abct.
The c/abct ratio for each thin film sample was then calculated by using results from Bragg-
Brentano and tilted XRD measurements. This is shown in Figure 98 b), depicting the
dependency of the tetragonal deformation (c/abct ratio) for all buffer materials. This
deformation behaviour is similar to investigations for thin binary Fe70Pd30 films which are
added for comparison.67,206
145
Figure 98: a) In-plane lattice parameters abct for Fe70Pd30 (rectangle) and Fe70Pd30-XCuX thin films with X = 3, 7
at.% (triangle) are fixed by the substrate’s lattice spacing d of different buffers (as marked at the top). b) The
change in tetragonal deformation (c/abct) with buffer material is smaller for the Fe70Pd30-XCuX system (triangle)
than for the binary Fe70Pd30 (rectangle67,206). The curves illustrate the expected change in deformation at constant
volume of the unit cell. (Figure originally published in Ref. 205)
In contrast to Fe-Pd films, the c/abct ratio of Fe70Pd27Cu3 is found to be slightly smaller,
suggesting that the addition of smaller atoms, such as Cu, reduces the volume of the Fe70Pd30
unit cell. This is in good accordance with results on lattice parameters determined for
polycrystalline Fe-Pd-Cu thin films and presented in Figure 83 in Chapter 4.2.3.1. The
presented thin-film samples with varying c/abct ratio allow investigation of the intrinsic
properties as a function of the tetragonal deformation. The variation of the c/abct ratio is used
in the following, next to the composition, as a parameter to control the magnetic properties.
For a detailed understanding of the anisotropic magnetic properties, the quality of coherent
epitaxial growth and the orientation relationship of thin film unit cells and the underlying
buffer material have to be determined.
The structural quality of the thin film samples was confirmed by pole figure measurements on
the (101)bct diffraction plane. Due to the 4-fold surface symmetry of the (001)-oriented MgO
substrate, only one quadrant of the pole figure measurement for Fe70Pd27Cu3 films on
different buffers is depicted in Figure 99 a).
146
Figure 99: a) (101) bct pole figures of Fe70Pd27Cu3 films on different buffer layers. With increasing c/abct ratio a
shift of the (101) pole towards higher ψ angles is observed. b) Drawing illustrating the geometric relation
between the tetragonal distortion (c/abct ratio) and the tilt angle ψ within a bct unit cell. (Pole figure
measurements were conducted together with S. Kaufmann-Weiss and the figure was originally published in Ref.
205)
The MgO [100] edges are oriented parallel to the edges of the figure. This reveals an
orientation relationship of the bct unit cell of the film and MgO (001) substrate as planned in
Figure 96 and evidences the successful coherent epitaxial growth of Fe-Pd-Cu thin films:
Fe70Pd30-XCuX(001)[110] || fccbuffer(001)[100] || Cr(001)[110] || MgO(001)[100] or
Fe70Pd30-XCuX(001)[110] || bccbuffer(001)[110] || MgO(001)[100].
The corresponding pole figures in Figure 99, illustrated by using custom-made software207,
indicate a high quality epitaxial growth of the films since well-defined and sharp peaks were
observed. The unit cells are rotated by 45° with respect to each other - a shift in ϕ direction of
about 45° - to reduce misfits between different layer materials. These pole figure
measurements can be further used to determine the c/abct ratio, since the tilt angle ψ of the
(101)bct plane is connected to the tetragonal distortion by tan ψ = c/abct (Figure 99 b)). This
relation predicts a tilt angle of the (101) pole at ψ ≈ 45° for a bcc structure (c/abct = 1) and at
ψ ≈ 54° for a fcc structure (c/abct = 1.41). When the buffer material is varied, the ψ angle shift
147
of the poles reflects the structural changes from bcc towards fcc structure. For a film on Cr
buffer the (101) pole is at a tilt angle ψ ≈ 47°. This confirms an almost cubic structure with
c/abct = 1.07. However, an Ir buffer layer shifts ψ to ≈ 52°, giving a c/abct ratio of 1.31 which
corresponds to a fct structure. The pole-figure measurement is therefore a second independent
measurement confirming the c/abct ratios obtained from θ-2θ scans with accuracy in the range
of 1%. A further increased tetragonal distortion is visible for the film on a Cu buffer
(c/abct = 1.57). In Bragg-Brentano scans (Figure 97) the (002) diffraction peak of the
Fe70Pd27Cu3 thin film is observed beyond the limits of the Bain path. This tetragonal
distortion agrees well with c/abct = 1.57 derived from the position of the pole in the (101)bct
pole figure at ψ ≈ 57°. The intensity in this measurement further exhibits a substantial
broadening and splitting in the ϕ direction. This observation agrees well with a report on
highly strained binary Fe70Pd30 films on Cu buffer layers.206 The broadening can be ascribed
to adaptive nanotwinning, which reduces the huge elastic energy induced by the coherent
epitaxial strain through the formation of twin boundaries. The rotation of the tetragonal
variants that form (101)fct twin boundaries leads to the small observed additional intensities.
The same set of XRD measurements was performed for a series of Fe70Pd23Cu7 thin films
with increased Cu content. These measurements are not shown here, since they are practically
identical to the ones obtained for lower Cu content. Concluding these results, no Cu-rich
precipitates occur even at 7 at.% Cu in epitaxially grown films. This is in contradiction to
polycrystalline annealed Fe-Pd-Cu thin films which decompose at Cu contents above 4 at.%.
This behaviour can be attributed to the difference in fabrication process routes. Epitaxial films
are deposited at room temperature without any annealing treatment, avoiding precipitation
compared to annealed films.
The lattice parameters determined for Fe70Pd23Cu7 are equal to those of Fe70Pd27Cu3. No
change in tetragonal deformation with increased Cu content is observed. Accordingly the
symbols in Figure 98 depict both the Fe70Pd27Cu3 and the Fe70Pd22Cu7 composition.
Previously presented experiments on polycrystalline annealed films have shown a decrease of
lattice constant with increasing Cu content, since the atomic radius for Cu (0.128 nm) is
smaller than for Pd (0.137 nm). However, it is known that alloys such as Fe-Pd, which exhibit
the Invar effect, deviate from the rule of mixture of Vegard’s law. This has to be considered
for the Fe-Pd and the second binary Cu-Pd systems.208
Epitaxial Fe70Pd30-XCuX thin films show improved growth behaviour when compared to
binary Fe70Pd30 films. In order to avoid unfavourable relaxation mechanisms, such as (111)fcc
deformation twinning,201 very low deposition rates of 0.024 nm/s were required for epitaxial
148
growth of binary Fe70Pd30 films.182 The present Fe-Pd-Cu films, were grown under a
deposition rate one order of magnitude higher (0.3 nm/s). Nevertheless, the poles in the (101)
pole figure and the diffraction peaks in the Bragg-Brentano measurement reveal a small full
width at half maximum. By using a modified Scherrer equation134, the determined coherence
length is in the range of the overall thickness of the Fe-Pd-Cu films, confirming a high quality
epitaxial growth. This enhancement of crystal growth can be understood in terms of the
structural variation of the total energy.
Total energy in epitaxially grown Fe-Pd-Cu films
Calculations of the total energy were performed by M. E. Gruner from the University of
Duisburg-Essen. These simulations were obtained from first principles within a supercell
description involving a full relaxation of the atomic position which minimizes the interatomic
forces for each tetragonal stage. The disordered arrangement was realized using a 500 atom
supercell and a random distribution of the 340 Fe atoms and 160 Pd atoms, or respectively,
135 Pd and 25 Cu atoms. In order to avoid statistical uncertainties, which hamper a direct
comparison, 25 randomly chosen Pd atoms were exchanged by Cu to model a comparable
ternary distribution, while the remaining elements were left untouched. In addition,
composition-dependent and finite-temperature magnetic properties using the Korringa-Kohn-
Rostoker (KKR) approach as implemented in the Munich SPR-KKR code (version 5.4) were
used.192 Further computational details can be found in Ref. 65 and 190. Here, the disordered
nature of the Fe-Pd alloy was modelled in terms of averaging the electronic scattering
properties within the coherent potential approximation (CPA). The CPA enables the economic
use of small cells but does not allow for structural relaxations. The magnetic exchange
parameters were determined for use within a Heisenberg model following the approach by
Liechstenstein et al.197, starting from the ferromagnetic state. The corresponding Curie
temperatures were estimated within the mean-field approximation. According to Figure 100
a), only a small variation in elastic energy E is involved if the c/abct ratio is varied along the
entire Bain transformation path. In addition to the global minimum at bcc we observed a
second local minimum beyond the Bain path (c/abct > 1.41). Such a local minimum can be
attributed to the formation of a nanotwinned microstructure in the simulation cell and is
further confirmed by pole figure measurements depicted previously. This minimum is
observed for the binary Fe68Pd32 solid solution, but also for ternary Fe68Pd27Cu5, where the
feature is comparatively shallow. Replacing 16% of Pd atoms by Cu increases the overall
valence electron number e/a by 0.05, which tends to stabilize the fcc austenite.209
149
Figure 100: a) Variation of the total energy per atom as a function of the tetragonal distortion c/abct obtained
from ab initio calculations of a Fe68Pd32 (black squares) and Fe68Pd27Cu5 supercell (red triangles). The
calculations were carried out at constant volume of 13.1⋅10-3 nm3/atom and include a full geometric optimization
of the atomic positions. Apart from the bcc ground state minimum at c/abct = 1 (which defines the energy
reference for each composition), in both cases a second local minimum around c/abct = 1.5 (beyond fcc) is
obtained, which corresponds to the appearance of a finely twinned adaptive superstructure in the 500 atom
simulation cell. b) The variation of the average ground state magnetic moment (normalized per atom) for both
configurations as a function of c/abct. (Calculations performed by M. E. Gruner and originally published in Ref.
205)
This is reflected in calculations from M. E. Gruner by the noticeably decreased energy
difference between the bcc ground state at c/abct = 1 and the fcc austenite at c/abct = 1.41.
Interestingly, the energy profile does not change significantly at low deformations for
c/abct < 1.3. Changes appear mainly in the vicinity of the local fcc maximum, which is now
embedded in a flatter energy landscape. A flat profile in the vicinity of the martensitic phase,
however, can be considered beneficial for the formation of the metastable fct phase, which is
required for the MFIS. The decrease in the fcc-bcc energy difference implies a decreased
temperature for the onset of the martensitic transformation, but this can be compensated by
the simultaneous replacement of Pd by Fe, as shown for polycrystalline Fe-Pd-Cu thin films.
The simultaneous substitution of Pd by Cu and Fe therefore opens a way to specifically
design the profile of the binding surface around 41.12a/c bctbct == which determines the
stability range of the fct phase. The total difference in elastic energy between the bcc and fcc
150
state, ∆E, is reduced from 15 to 12 meV atom-1 for Fe68Pd32 and Fe68Pd27Cu5, respectively.
This reduces the driving force for the transformation and all associated relaxation processes
and is thus favourable for strained epitaxial growth.
Magnetic remanence, coercivity and saturation field
The hard magnetic axis in a tetragonal lattice of a bulk Fe70Pd30 single-crystal is aligned along
the c-axis while the magnetic easy directions lie within the basal plane.57,210 Within the easy
plane, not all directions are equivalent and a slight anisotropy is observed, which favours both
[110]bct easy axes.
Figure 101: Determination of magnetocrystalline anisotropy within the basal plane by measuring hysteresis
curves of thin films with different tetragonal deformation ratios a) to e) at 300 K. Solid lines indicate
measurements along the [110]bct direction, broken lines along [100]bct direction. f) Comparison of hysteresis
loops measured in out-of-plane orientation along [001]bct at 300 K. The inset presents a magnification around
µ0H = 0 T. From these measurements, the magnetic constants JR, JS and HC to calculate the magnetocrystalline
anisotropy were determined. (Figure originally published in Ref. 205)
151
To investigate structure-dependent magnetic properties of epitaxial Fe70Pd30-XCuX films,
magnetic hysteresis curves were measured at 300 K in in-plane orientation for thin film
samples having different tetragonal deformation (Figure 101 a) to e)). The magnetization of
the thin film samples differs significantly along different crystallographic in-plane directions
([100]bct and [110]bct) when compared to binary Fe70Pd30 thin films. The most evident
differences occur for c/abct = 1.31, which is located close to the middle of the Bain
transformation path. For this c/abct ratio both crystallographic directions have different values
for remanent polarisation JR and saturation fields HS, while the values for magnetic coercivity
HC do not change significantly. Furthermore, the shape of the magnetization curves and thus
the magnetic characteristic changes with the c/abct ratio. The magnetic hysteresis measured in
the [100]bct direction shows a step-like switching behaviour for the thin film with a low
tetragonal deformation (c/abct = 1.07). In the [110]bct direction a similar switching process
occurs, but additionally a nearly linear increased magnetization at higher fields is observed.
This behaviour originates from a rotation of the magnetization in the thin film during
measurement, indicating that the [110]bct direction is the harder axis within the basal plane,
while JR is reduced in this direction. By increasing the c/abct ratio to 1.31 and higher, an
opposite behaviour occurs. This is correlated to a change in sign of the magnetic anisotropy
within the basal plane.
A summary of the extracted values for magnetic coercivity HC and saturation field HS along
the [110]bct is depicted in Figure 102 a) and b). The trend of µ0HC and µ0HS determined from
in-plane measurements of films with different tetragonal distortions is similar to epitaxial
Fe70Pd30 films (black squares in Figure 102) and epitaxial Fe1-XPdX films deposited on
MgO(100) substrates as reported in literature.67,211 Films having a c/abct ratio close to the bcc
structure exhibit a low HC and HS due to the high crystal symmetry. Thin films with a bct
structure exhibit a slightly increased HC and slightly reduced HS values. Thin films having a
fct structure exhibit a different loop shape with the highest values for HC and HS due to the
large lattice deformation. When the tetragonal distortion comes close to the fcc phase, HC and
HS are reduced because of the high crystal symmetry of the fcc structure. All fabricated
Fe70Pd30-XCuX thin films follow this behaviour. Both values HC and HS approximate to zero
when the tetragonal distortion comes close to both cubic structures bcc (c/abct = 1) and fcc
(c/abct = 1.41). In contrast both values show a maximum value for tetragonal structures. The
HC value decreases and shows only slight differences between both in-plane curves for
Fe70Pd30-XCuX and Fe70Pd30 thin films, distorted beyond the Bain path (c/abct = 1.57, Figure
101 e)).
152
Figure 102: Presented is the in-plane a) coercivity field µ0HC || [100]bct and b) saturation field µ0HS || [100]bct in
dependence on c/abct ratio. c) The out-of-plane saturation field µ0HS || [001]bct is extracted from magnetic
hysteresis measurements in Figure 101. Black rectangles represent the results for Fe70Pd30 films, red circles
Fe70Pd27Cu3 films and blue triangles Fe70Pd23Cu7 films. All lines are visual guides for the eye. (Figure originally
published in Ref. 205)
This behaviour will be discussed together with the magnetic anisotropy constants K1 and K3
in the following.
To obtain the magnetic anisotropy constant K1, out-of-plane magnetic hysteresis curves along
the [001]bct direction were measured (Figure 101 f)). Along the in-plane directions, all thin
films are magnetically saturated at magnetic flux densities of 0.1 T. In contrast to this, flux
densities > 1 T are required to saturate the same thin films in out-of-plane hysteresis
measurements. The shape anisotropy dominates the magnetization behaviour for an ideal thin
film (demagnetization factor N = 1). When neglecting magnetocrystalline anisotropy,
magnetic saturation is expected at µ0HS = JS. The difference between µ0HS and JS will be used
later to calculate the magnetic anisotropy constant K1. When increasing the tetragonal
deformation to c/abct < 1.41, the HS value decreases (Figure 102 a)).
For all measurements a small hysteresis is observed (see inset in Figure 101 f)). This
switching process is attributed to a slight angular misalignment of the thin film sample normal
with respect to the field during measurement. For a pinning-controlled magnetic coercivity
mechanism, HC exhibits a Kondorsky-like strong increase when the direction of the magnetic
field approaches the hard direction.212 At the same time the remanence-to-saturation ratio
decreases. As a detailed analysis requires a more accurate control of the tilt angle, the HC
153
value is not used for the out-of-plane direction here. Values for HC and HS along the [110]bct
and [001]bct directions from the in-plane and out-of-plane hysteresis curves were extracted for
Fe70Pd27Cu3 and Fe70Pd23Cu7 films, following same trends (see blue triangles in Figure 102).
Variation of Curie temperature and spontaneous polarization
Previous experiments have indicated that the degree of tetragonal distortion of the lattice
significantly affects the Curie temperature.67 For all epitaxial Fe70Pd30-XCuX thin films,
temperature-dependent magnetization curves were measured in-plane along the [110] bct
direction. An applied magnetic flux density in the range of ≈ 1 T in this direction is sufficient
to saturate the thin film sample. To avoid the destructive thermal decomposition of the
metastable alloy,213 the temperature-dependent magnetization curves were only measured up
to 400 K, which is below the values for TC. As in the previous sections, the TC values of the
thin films were determined from an extrapolation of the magnetization curves using
Kuz’min’s empirical fit.151,152 The results are presented in Figure 103 a).
Figure 103: Variation of a) experimental and b) calculated Curie temperature TC in dependence on c/abct and
composition. With increasing c/abct < 1.41, TC decreases. For c/abct = 1.57, similar values as at c/abct = 1.11 are
observed. The theoretical Curie temperatures are obtained using the mean-field approximation to the Heisenberg
model and corrected by a factor 0.75. The dashed lines in b) represent results for the adaptive nanotwinning
concept. These data agree well with the experimental values. (Figure originally published in Ref. 205)
Black rectangles represent the values for Fe70Pd30, red circles Fe70Pd27Cu3 and blue triangles
Fe70Pd23Cu7. Both the Fe70Pd30 as well as the Fe70Pd30-XCuX systems show a similar variation
of TC with c/abct. The values for TC decrease monotonically as the c/abct ratio increases up to
154
1.41. The value of TC = 652 K at c/abct = 1.39 approximates the value of 600 K reported for
the bulk fcc phase (c/abct = 1.41).52 The change in TC by addition of Cu is within the range of
experimental error. Thus, compared to other FSMAs, such as Ni-Mn-Ga (TC < 370 K), the
Curie temperature TC in Fe-Pd-based systems is significantly higher.214
In order to determine the compositional and structural trends with regard to TC and the ground
state JS independently, ab initio calculations were performed by M. E. Gruner. The
calculations corroborate the experimental findings. Values for TC are obtained from a classical
Heisenberg model which is parameterized with first-principles magnetic exchange constants.
In the present case, the mean-field approximation can already give a reasonable estimate of
the structural and compositional trends governing the magnetic transformation. Nevertheless,
it has to be considered that this approach systematically overestimates critical temperatures
(typically by 20-30%) due to neglecting spin fluctuations. Furthermore, the induced nature of
the Pd moments, which vary in magnitude according to the field of the surrounding localized
Fe moments, is also not taken into account.215 However, a comparison with a numerically
exact Monte Carlo treatment of the Heisenberg model216 demonstrates that for a heuristic
prediction these shortcomings can be compensated sufficiently by correcting TC
systematically with an empirical factor of 0.75. The quantitative values for TC that are
calculated along the Bain path between bcc and fcc (Figure 103 b)) agree with the values from
experiments. However, no significant variation on TC in dependence on the Cu content (3
at.% and 7 at.%) is observed.
For tetragonal distortions beyond the Bain path at c/abct > 1.41, the experimentally determined
TC value increases again. The TC value of 822 K at c/abct = 1.56 is similar to values for a
structure with c/abct = 1.11. In contrast, the calculations exhibit only a small increase in TC
(solid line in Figure 103 b)). These assume a single variant with c/abct = c/abox > 1.41
according to the substrate constraint. The agreement is much better if an energetically more
favourable nanotwinned fct configuration is assumed, where the individual twins with
c/abct < 1.41 form a microstructure which suits the substrate constraints. From a simple
geometric consideration an approximated relation between the epitaxial constraint due to the
substrate c/abox and the tetragonality of the twins c/atwin has been derived:206
2
twinbox a
c21
a
c−
+= (12)
By using this relation to extrapolate TC for c/abct > 1.41 beyond fcc, a reasonable agreement
with experimental values is obtained for the thin films deposited on a Cu buffer layer (dashed
line in Figure 103 b)).
155
Another important intrinsic parameter for the FSMAs is the high spontaneous magnetic
polarisation JS. Figure 104 a) shows how JS depends on the tetragonal distortion for different
Cu contents. These values were extracted from the deformation-dependent changes in the
magnetic hysteresis curves in Figure 101 f). According to the structural variation of TC, a
minimum of JS = 1.19 T, taken at a constant temperature of 300 K, at fcc (c/abct = 1.41) is
observed. Near to the bcc (c/abct = 1.09) and at huge strains, JS reaches values of 1.76 T. When
compared to binary films, both Fe70Pd27Cu3 and Fe70Pd23Cu7 films reach only 80% of the JS
of Fe70Pd30. This is consistent with previously shown findings for polycrystalline Fe-Pd-Cu
thin films in Chapter 4.2.3.2. Nevertheless, this value for JS is still significantly higher than
the value JS = 0.76 T obtained for the most prominent FSMA Ni-Mn-Ga.217 While the main
trends are identical, where JS decreases with increasing Cu content having a minimum value
at the fcc structure, the experimental values obtained at 300 K close to TC exhibit a
significantly stronger variation with both composition and c/abct ratio than the results from the
ab initio calculations at T = 0 K (Figure 104 b)).
Figure 104: a) Spontaneous polarization JS determined for different c/abct ratios and compositions at 300 K. JS
for Fe70Pd30-XCuX is approximately 80% of the value for Fe70Pd30. Lines are guides for the eye. b) Dependence of
calculated ground state total (spin + orbital) magnetic moments on c/abct and composition at 0 K. The dashed
lines represent results for the adaptive nanotwinning concept. (Figure originally published in Ref. 205)
The ground state magnetic moments vary only by a few percent while the changes found
experimentally are almost one order of magnitude larger. The larger reduction for the fcc
phase around 300 K is expected, since the fcc phase exhibits the lowest TC. This might also be
156
correlated to longitudinal spin fluctuations, which can occur in Invar materials such as fcc
Fe70Pd30 218,219 and are known to reduce the values for TC in the austenite phase further.
Magnetocrystalline anisotropy energy
As mentioned before, one of the key intrinsic magnetic properties of FSMAs is their
magnetocrystalline anisotropy energy (MAE). For the MFIS, the MAE represents the
maximum energy input possible by applying an external magnetic field. Hence the MAE
limits the energy available to move twin boundaries or conduct external work. In general the
MAE of a tetragonal lattice can be described by the following equation:220
MAE = K1 sin2(α) + K2 sin4(α) + K3 sin4(α)cos(4β) (13)
Here Ki are the anisotropy constants, α is the angle between the magnetization direction and
the c-axis and β is the angle between the field and the a-axes within the basal plane of a
tetragonal lattice. In most compounds without rare-earth elements, higher-order terms (K2)
can be neglected. For the Fe70Pd30 system, this was reported by Cui et al..57 Thus, it is
sufficient to consider here only K1, which describes the work required to magnetize the
sample along the hard magnetization c-axis, and K3, which characterizes the 4-fold anisotropy
within the basal plane. In the fabricated Fe70Pd30-XCuX thin films, the c/abct ratio is controlled
by strained epitaxial film growth. This allows determination of Ki for all distortions along the
Bain path at one temperature (300 K), which is not possible in bulk samples. Due to the
martensitic transformation and the temperature dependency of the c/aratio, it is not possible to
separate the influence of T and c/a in the bulk. From previously shown hysteresis
measurements on Fe70Pd30-XCuX thin films (Figure 101), the [001]bct direction was determined
to be the hard magnetization axis while the [100]bct and [110]bct directions form the easy
plane.
From the present thin film experiments K1 can be determined from the measurements of HS
along the hard [001]bct direction. The shape anisotropy was considered by using a
demagnetization factor N = 1 of an ideal infinite film, and the anisotropy field HA can be
calculated by µ0HA = µ0HS - NJS. The anisotropy constant is then converted by:
2
JHK SA
1⋅−= (14)
In Figure 105 a) K1 is plotted as a function of the c/abct ratio and the Cu content. The
tetragonal deformation of a Fe-Pd unit cell results in the formation of an easy plane,
157
corresponding to K1 < 0. The maximum effect is observed around c/abct = 1.34, corresponding
to the fct structure, which exhibits the MFIS in bulk. For binary fct Fe70Pd30 the value was
found to be K1 = -1.6 x 105 J/m3. This agrees well with literature values reported for fct single
crystals (open rectangles in Figure 105 a)) and DFT calculations (stars in Figure 105 a)).67,42
In contrast, Fe70Pd30-XCuX films exhibit an increased absolute value of magnetocrystalline
anisotropy of K1 ≈ -2.4 x 105 J/m3. The absolute values for K1 at 300 K of Fe70Pd27Cu3 and
Fe70Pd23Cu7 even exceed the magnetocrystalline anisotropy constants reported for any other
FSMAs like the Ni-Mn-Ga system: K1 = 1.65 x 105 J/m3 for a 10 M single-variant single-
crystal and K1 = 1.7 x 105 J/m3 for the 14 M structure.214 A larger MAE of similar magnitude
is only observed for Ni-Mn-Ga at significantly lower temperatures217,221 according to the
considerable variation of the MAE with temperature in uniaxial magnets222. When changing
the tetragonal deformation close to highly symmetric cubic structures (bcc and fcc), K1 is
reduced for all compositions. At c/abct > 1.41, the magnetocrystalline anisotropy increases
again, but does not reach the values of the fct structure.
Figure 105: Magnetocrystalline anisotropy constants a) K1 and b) K3 as a function of c/abct for various Cu
contents (solid symbols, T = 300 K). Also shown are the values for a bulk Fe68.8Pd31.2 single-crystal (open
rectangle42) and calculations for disordered Fe70Pd30 (stars, T = 0 K67). The errors for K1 are shown by error bars
and for K3 they are in the range of the symbol size. All lines are guides for the eye. (Figure originally published
in Ref. 205)
This is again consistent with the presence of adaptive nanotwins in the film.206 For a very
small martensitic variant width it is no longer possible to form a complete 90° domain wall at
twin boundaries since the exchange energy favours a parallel alignment of magnetization.
158
According to the random anisotropy model of Herzer, the critical length for this is the
magnetic exchange length lexch.223 This exchange length lexch is in the order of 18-85 nm (and
depends on the composition) for fct Fe-Pd67, which far exceeds the width of adaptive
nanotwins206. Due to the 4-fold symmetry of the basal plane a deviation from an idealized
easy plane behaviour can be observed in films (Figure 101) and bulk samples.57,210 K3 is a
measure of the anisotropy within this basal plane and defines the work that is necessary to
magnetize along the respective directions of the bct unit cell:
2K3 = W[100]bct - W[110]bct (15)
The magnetocrystalline anisotropy constant K3 can be extracted from the area enclosed by the
magnetic hysteresis curves (Figure 101) measured along both the [110]bct and [100]bct
directions. These values of K3 are two orders of magnitude smaller than for K1 (Figure 105
b)) and change sign within the Bain path. For c/abct values close to bcc, positive values are
observed. Films with c/abct ratios close to fcc have negative K3 values. In between these
values no significant differences were observed for Fe70Pd30 and Fe70Pd30-XCuX films. As for
|K1|, |K3| exhibits a maximum of 1.8 x 105 J/m3 for the fct structure (c/abct = 1.34).
The presented findings and analysis of the structure and magnetism in ternary Fe-Pd-Cu
epitaxial thin films suggests that the addition of small amounts of Cu is able to significantly
enhance the functional properties of the Fe70Pd30 FSMA. The combination of thin film
experiments and ab initio simulations yield an insight into the “frozen stages” of the
martensitic transformation process within the limits of the Bain path, which is enforced by the
epitaxial relation to selected buffer materials. This is in particular favourable for a detailed
analysis of the anisotropic magnetic properties since the measurements are affected neither by
a magnetically-induced reorientation nor by the continuous variation of the tetragonal
distortion with temperature, which both occur in bulk samples. One main result is that Cu
flattens the energy landscape, which suppresses common relaxation mechanisms and thus
allows for a much better film quality in combination with faster growth. More importantly, an
increase of the magnetocrystalline anisotropy constant K1 by ≈ 40% is obtained, which is a
substantial improvement of a key property necessary for the MFIS. The values obtained for
the fct structure even exceed those reported for the prototype Ni-Mn-Ga system, which makes
the Fe-Pd-Cu system of significant interest for microsystems with a high energy density. A
minor drawback is the reduction of the spontaneous polarization JS by about 15-20% due to
the addition of Cu.
159
XRD measurements indicate that the use of room temperature deposition techniques prevents
decomposition, which must otherwise be expected for more than 5 at.% Cu. However, upon
increasing the Cu content from 3 to 7 at.% no variation in the trend of magnetic properties
was observed. This might in turn be taken as an indication for the presence of structural or
compositional inhomogeneities at the nanoscale (as for example a tendency towards a slight
L10 short-range order with antiphase boundaries). Recent Mössbauer experiments on Fe-Pd-
Cu splats with <1 at.% Cu suggest some chemical short-range order.224 Moreover, first-
principles calculations of Fe-rich Fe-Pd predict a certain preference for forming a layered type
of order (L10 or Z1),225,226 while the cubic Fe3Pt-type L12 order appears to be ruled out for
energetic reasons65. In the martensitic state, ageing behaviour resulting in symmetry-
conforming short-range order at this length scale does not necessarily inhibit shape memory
applications227 and could tentatively induce a kind of beneficial two-way behaviour. This
aspect, however, cannot be probed for the films tested with the presented methods and
requires further investigation.
160
5. Conclusion and Outlook
In the presented work, combinatorial fabrication and high-throughput characterization
methods were employed to develop new Fe-Pd(-X) ferromagnetic shape memory alloys
(FSMAs) with improved properties. Broad regions of the full binary and ternary Fe-Pd(-X)
compositional diagram were fabricated as thin film materials libraries by wedge-type and
confocal thin film deposition and were investigated by high-throughput characterization:
namely, energy dispersive X-ray analysis (EDX) for composition, temperature-dependent
resistance (R(T)) and magneto-optical Kerr effect (MOKE) measurements for characterization
of phase transformation and magnetic properties. Further X-ray diffraction (XRD) and
synchrotron measurements for structural, and nanoindentation at elevated temperatures for
mechanical, properties were investigated. Promising thin films were further investigated by
non-high-throughput methods such as vibrating sample magnetometry (VSM), temperature-
dependent X-ray diffraction (XRD(T)) and transmission electron miocroscopy (TEM) for
detailed analysis.
Polycrystalline Fe-Pd thin film materials libraries revealed transforming composition regions
in dependence of sputter deposition method. Lowest transformation temperatures were found
for films, fabricated by wedge-type deposition. This was correlated to the exothermic energy
of mixing during annealing and the decrease of internal stress due to a reduction of interfaces
in the microstructure.
The splat quenching method was identified to be well-suited for fabricating single-phase
Fe70Pd30 bulk samples that exhibit a martensitic transformation. Annealing at different
temperatures resulted in a shift of transformation temperatures and was correlated to the
columnar grain growth and the difference in grain sizes.
Epitaxial Fe-Pd thin films attached to a substrate revealed different structural phases in
dependence on composition without undergoing a martensitic transformation in the as-
deposited state. After annealing a reversible transformation was identified in a Fe72Pd28 thin
film.
Annealed freestanding Fe70Pd30 thin films showed an unusual reversible martensitic
transformation from the bcc over fct to the austenite fcc phase. This sequence of phases was
theoretically expected from the Bain path formalism but was never observed before for Fe-Pd.
Epitaxial Fe70Pd30 films were fabricated and investigated with regard to demonstrate the
feasibility of fabricating freestanding 1.2 µm thick films that can be implemented as
sensing/actuating devices in microsystems.
161
The Fe-Pd-Mn system was developed as a new FSMA showing improved properties. A
broadening of the transforming single-phase region was identified in the composition
diagram, allowing to alloy up to 9 at.% of Mn at the expense of Fe content. Transformation
temperatures up to 379 K were measured in Fe-Pd-Mn thin films, exceeding all values known
for Fe-Pd-based FSMAs from literature so far. Magnetic properties such as Curie temperature
TC and saturation polarization JS in this system were found to be only slightly lowered
compared to the Fe-Pd system. Stress measurements allowed to exclude the formation of
stress-induced martensite as the cause for the increased transformation temperatures. From
theoretical predictions an energy gain due to displacements from the ideal lattice positions
upon addition of Mn was reported. This explained the increased amount of energy in terms of
temperature that is needed to induce the martensitic transformation. Finally the e/a ratio was
found to be the dominating parameter, describing the change of martensitic transformation
temperatures.
The Fe-Pd-Cu system indicated a broadening of the transforming single-phase region, where
up to 4 at.% of Cu can be added at the expense of the Pd contents. Increased transformation
temperatures were determined for this system, while TC and JS are not significantly lowered.
In accordance to ab initio results, an increase of spontaneous volume magnetostriction and
thus a decrease of thermal expansion was postulated. This was confirmed by synchrotron
experiments revealing a decrease of residual thin film stress and a decrease of thermal
expansion coefficient in Fe-Pd-Cu splats with increasing Cu content. Finally, the Fe content
was found to be the decisive factor controlling the shift of martensitic transformation. The
addition of Cu also enabled to increase the Fe content to values above the limit known for the
binary Fe-Pd system, without the formation of precipitates. Epitaxial Fe70Pd30-XCuX (X = 3
and 7 at.%) thin films deposited on different buffer layers, varying the in-plane lattice
parameter, revealed a significant change of magnetic properties in dependence of the c/a-ratio
in the unit cell. Finally, the highest value for the magnetocrystalline anisotropy constant
K1 = 2.4 x 105 J/m3 at 300 K was identified for c/abct = 1.34. This value increases by about
40% due to the addition of Cu and is higher than any other value reported for FSMAs so far.
The discovered results in the new Fe-Pd-Mn and Fe-Pd-Cu thin film systems exceed the
benchmark values defined in Chapter 2.8. Both systems showed martensitic start temperatures
with Ms > 350 K without increasing the thermal hysteresis to values ∆T > 5 K. The saturation
polarization was lowered in both systems, but never exceeded JS < 1 T in the martensite
phase. The Curie temperature was found to vary between 460 K < TC < 550 K and thus is
smaller than the benchmark value. The magnetocrystalline anisotropy constant K1 can be
162
considered as the most important value for the MFIS. Here, the benchmark value of
K1 = 1.0 x 105 J/m3 was significantly exceeded for a Fe70Pd27Cu3 thin film. All other
benchmark values were not proven and will be subject for future work.
Furthermore it has to be mentioned, that no distinct rules were identified for enhancing Fe-Pd-
based FSMAs upon addition of further elements from the present findings. It was found, that
the shift of martensitic transformation is controlled by different dominating parameters in
both ternary systems. While the transformation temperatures vary with Fe content in the Fe-
Pd-Cu system, the Fe-Pd-Mn system is controlled by the e/a ratio. This indicates that the
impact of the third element on the martensitic transformation, structure and magnetic
parameters is tremendous and alters the materials properties significantly.
All presented findings indicate the advantage of combinatorial fabrication and
characterization methodologies to develop new FSMAs with enhanced properties. The impact
of third elements on the composition-structure-property relationship can be determined up to a
certain extent in polycrystalline thin film materials libraries. However, key properties for
FSMAs like the magnetocrystalline anisotropy cannot precisely be measured in
polycrystalline thin film samples. Therefore the combination of different sample designs such
as materials libraries, epitaxial films and bulk samples is well-suited for an accelerated
understanding, followed by a tailored development of new materials with improved
properties. This approach can be further extended to other materials that undergo reversible
structural or magnetic transformations and are of high interest for industrial application, such
as Fe- and Co-based magnetocaloric materials.
163
6. References
1 Janocha H 1999 Adaptronics and Smart Structures: Basics, Materials, Design and Applications (Berlin:
Springer) 1st ed 2 Otsuka K and Ren W 2005 Physical metallurgy of Ti-Ni-based shape memory alloys Prog. Mater Sci. 50 pp 511-678 3 Ullakko K, Huang J K, Kanter C, Kokorin V V and O’Handley R C 1996 Large magnetic field induced strains in Ni2MnGa single crystals Appl. Phys. Lett. 69 p 1966 4 James R D and Wuttig M 1998 Magnetostriction of martensite Philos. Mag. A 77 p 1273 5 Gehanno V, Marty A, Gilles B and Samson Y 1997 Magnetic domains in epitaxial ordered FePd (001) thin films with perpendicular magnetic anisotropy Phys. Rev. B 55 18 pp 12552-12555 6 Chen M and Nikles D E 2002 Synthesis of spherical FePd and CoPt nanoparticles J. Appl. Phys. 91 8477 7 Duplessis R R, Stern R A and MacLaren J M 2004 Phase stability criteria for cubic and orthorhombic Fe3Pt and Fe3Pd J. Appl. Phys. 95 6589 8 Jessen K 1962 Über die Invar-Eigenschaften der Eisen - Palladium-Legierungen Ann. Phys. 464 5-6 pp 313-315 9 Matsui M and Adachi K 1983 Magnetostriction of Fe-Pd Invar J. Magn. Magn. Mater. 31-34 pp 115-116 10 Okazaki T, Tanaka M, Ogasawara N, Furuya Y, Saito C and Imaizumi N 2009 Development of Magnetic-Field-Driven Micro-Gas Valve Mater. Trans. 50 3 pp 461-466 11 Chopra H D and Wuttig M 1995 Precursor Shear Elastic Pseudo-Isotropy in Near Second Order Martensitic Phase Transitions J. Phys. IV 05 C8 pp 157-160 12 Muto S, Oshima R and Fujita F E 1990 Elastic softening and elastic strain energy consideration in the f.c.c.-f.c.t. transformation of Fe-Pd alloys Acta Metall. Mater. 38 4 pp 685-694 13 Otsuka K. and Wayman C M 1999 Shape memory materials (Cambridge Univ Press) 1st ed 14 Martens A 1887 Stahl und Eisen 7 pp 235-242 15 Haasen P 1994 Physikalische Metallkunde (Springer) 3rd ed 16 Wayman C and Duerig T 1990 Introduction to Martensite and Shape Memory, in Engineering Aspects of
Shape Memory Alloys (Elsevier Butterworth-Heinemann Oxford) ed T Duerig, K Melton, D Stöckel and C Wayman
17 Hornbogen E 1985 The effect of variables on martensitic transformation temperatures Acta Metall. 33 pp 595-601
18 Bowles J S and Mackenzie J K 1954 The crystallography of martensite transformations Acta Mater. 2 pp 129-234
19 Lieberman D S, Wechsler M S and Read T A 1955 Cubic to orthorhombic diffusionless phase change-experimental and theoretical studies of AuCd J. Appl. Phys. 26 4 pp 473-484
20 Nagasawa A 1976 Nature of Martensitic phase transition in β phase alloys J. Phys. Soc. Jpn. 40 4 pp 1021-1026
21 Wollants P, Bonte M and de Roos J R 1979 A thermodynamic analysis of the stress-induced martensitic transformation in a single crystal Z. Metallkd. 70 113
22 Kato H, Liang Y and Taya M 2002 Stress-induced FCC/FCT phase transformation in Fe-Pd alloy Scripta Mater. 46 471
23 Kästner J, Petry W, Shapiro S M, Zheludev A, Neuhaus J, Roessel T, Wassermann E and Bach H 1999 Influence of atomic order on TA1[110] phonon softening [4] and displacive phase transition in Fe72Pt28 Invar alloys EPJB 10 641
24 Oikawa K, Saito R, Anzai K, Ishikawa H, Sutou Y, Omori T, Yoshikawa A, Chernenko V A, Besseghini S, Gambardella A, Kainuma R and Ishida R 2009 Elastic and Superelastic Properties of NiFeCoGa Fibers Grown by Micro-Pulling-Down Method Mater. Trans. 50 4 pp 934-937
25 Miyazaki S, Fu Y Q and Huang W M 2009 Thin Film Shape Memory Alloys: Fundamentals and Device Applications (Cambridge: University Press) 1st ed
26 Du Trémolet De Lacheisserie E 1993 Magnetostriction, Theory and Applications of Magnetoelasticity (Boca Raton, Ann Arbor, London, Tokyo: CRC Press) 1st ed
27 Krevet B and Kohl M 2010 Thermodynamic modelling of ferromagnetic shape memory actuators Mater. Sci. Forum 635 p 175 28 Omori T, Watanabe K, Umetsu R Y, Kainuma R and Ishida K 2009 Martensitic transformation and magnetic
field-induced strain in Fe-Mn-Ga shape memory alloy Appl. Phys. Lett. 95 8 082508 29 Chmielus M, Witherspoon C, Wimpory R C, Paulke A, Hilger A, Zhang X, Dunand D C and Müllner P 2010
Magnetic-field-induced recovery strain in polycrystalline Ni-Mn-Ga foam J. Appl. Phys. 108 123526 30 Quandt E and Wuttig M 2004 Comparison of joule and twin induced magnetostriction Bremen Actuator Conf.
Proc. pp 359-362
164
31 Bozorth R M 1993 Ferromagnetism (Piscataway, New Jersey: IEEE Press) 1st ed 32 Kakeshita T, Fukuda T and Takeuchi T 2006 Magneto-mechanical evaluation for twinning plane movement
driven by magnetic field in ferromagnetic shape memory alloys Mater. Sci. Eng. A 438-440 12-17 33 O’Handley R C, Murray S J, Marioni M A, Nembach H and Allen S M 2000 Phenomenology of giant
magnetic-field-induced strain in ferromagnetic shape-memory materials J. Appl. Phys. 87 pp 4712-4717 34 Heczko O, Scheerbaum N and Gutfleisch O 2009 Magnetic Shape Memory Phenomena In Nanoscale
Magnetic Materials and Applications (Heidelberg: Springer) ed J P Liu, E Fullerton, O Gutfleisch, D J Sellmyer pp 399-439
35 Heczko O 2005 Magnetic shape memory effect and magnetization reversal J. Magn. Magn. Mater. 290-291 787-794
36 Ullakko K, Huang K, Kokorin V V and O’Handley R C 1997 Magnetically controlled shape memory effect in Ni2MnGa intermetallics Scripta Mater. 36 pp 1133-1138
37 Kakeshita T and Fukuda T 2002 Giant magnetostriction in Fe3Pt and FePd ferromagnetic shape-memory alloys Mater. Sci. Forum 394-395 p 531
38 Lavrov A N, Komiya S and Ando Y 2002 Antiferromagnets: Magnetic shape-memory effects in a crystal Nat. Mater. 418 p 385
39 Rhyne J J, Foner S, McNiff E J and Doclo R 1968 Rare earth metal single crystals. I. high-field properties of Dy, Er, Ho, Tb, and Gd J. Appl. Phys. 39 p 892
40 Sozinov A, Likhachev A A, Lanska N and Ullakko K 2002 Giant magnetic-field-induced strain in NiMnGa seven-layered martensitic phase Appl. Phys. Lett. 80 p 1746
41 Chmielus M, Zhang X X, Witherspoon C, Dunand D C, Müllner P 2009 Giant magnetic-field-induced strains in polycrystalline Ni-Mn-Ga foams Nat. Mater. 8 p 863
42 Kakeshita T, Fukuda T and Takeuchi T. 2006 Magneto-mechanical evaluation for twinning plane movement driven by magnetic field in ferromagnetic shape memory alloys Mater. Sci. Eng. A 438-440 pp 12-17
43 Hausch G 1973 Magnetovolume effects in invar alloys: Pressure dependence of the curie temperature Phys. Status Solidi A 16 2 371-376
44 Cullity B D and Graham C D 2008 Introduction to Magnetic Materials (Piscataway, New Jersey: IEEE Press) 2nd ed
45 Guillaume C E 1897 Les Comptes Rendus de l'Académie des sciences 125 pp 235-238 46 Kaufman L, Clougherty E V and Weiss R J 1963 The lattice stability of metals-III. Iron Acta Metall. 11 p 323 47 Wassermann E F 1989 New Developments on the Invar-Effect. Phys. Scr. T. 25 pp 209-219 48 Weiss R J 1963 The Origin of the ‘Invar’ Effect Proc. Phys. Soc. 82 p 281 49 Kondorsky E I and Sedov V L 1960 Effect of the Polarity of the III-V Intermetallic Compounds on Etching J.
Appl. Phys. 31 331 50 Kakehashi Y 1989 Magnetovolume effect and finite-temperature theory of magnetism in transition metals and
alloys Physica B 161 pp 143-152 51 Schlosser W 1971 Occurrence of Invar Anomalies in Binary Alloys of Fe with Ni, Pd, Pt, and Mn J. Appl.
Phys. 42 1700 52 Matsui M, Shimizu T, Yamada H and Adachi K 1980 Magnetic Properties and thermal expansion of Fe-Pd
Invar Alloys J. Magn. Magn. Mater. 15-18 pp 1201-1202 53 Nakayama T, Kikuchi M and Fukamichi K 1980 Young's modulus and the ∆E effect of Fe-Pd Invar alloys J.
Phys. F. Metal. Phys. 10 715 54 Haasen P 1994 Physikalische Metallkunde (Berlin: Springer) 3rd ed 55 Burton B P 1990 Fe-Pd (Iron-Palladium), Binary Alloy Phase Diagrams (ASM Alloy Phase Diagrams Center
2nd ed) ed T B Massalski vol 2 pp 1749-1751 56 Sugiyama M, Oshima R and Fujita F E 1984 Martensitic transformation in the iron-palladium alloy system
Mater. Trans. 25 9 pp 585-592 57 Cui J, Shield T W and James R D 2004 Phase transformation and magnetic anisotropy of an iron-palladium
ferromagnetic shape-memory alloy Acta Mater. 52 1 pp 35-47 58 Matsui M, Yamada H and Adachi K 1980 A New Low Temperature Phase (fct) of Fe-Pd Invar J. Phys. Soc.
Jpn. 48 6 pp 2161-2162 59 Matsui M and Chikazumi S 1978 Analysis of Anomalous Thermal Expansion Coefficient of Fe-Ni Invar
Alloys J. Phys. Soc. Jpn. 45 pp 458-465 60 Sumiyama K, Shiga M, Morioka M and Nakamura Y 1979 Characteristic magnetovolume effects in Invar type
Fe-Pt alloys J. Phys. F 9 p 1665-1677 61 Sugimura Y, Cohen-Karni I, McCluskey P and Vlassak J J 2005 Stress Evolution of Sputter-Deposited Fe-Pd
Shape-Memory Thin Films J. Mater. Res. 20 9 pp 2279-2287 62 Stern R A, Willoughby S D, MacLaren J M, Cui J, Pan Q and James R D 2003 Fe3 Pd ferromagnetic shape
memory alloys J. Appl. Phys. 93 p 10
165
63 Bain E C and and Dunkirk N Y 1924 The nature of martensite Trans. Am. Inst. Min. Metall. Eng. 70 p 25 64 Oshima R 1981 Successive Martensitic Transformations in Fe-Pd Alloys Scr. Metall. 15 829-833 65 Gruner M E and Entel P 2011 Impact of local lattice distortions on the structural stability of Fe-Pd magnetic
shape-memory alloys Phys. Rev. B 83 214415 66 Opahle I, Koepernik K, Nitzsche U and Richter M 2009 Jahn-Teller-like origin of the tetragonal distortion in
disordered Fe-Pd magnetic shape memory alloys Appl. Phys. Lett. 94 7 072508 67 Buschbeck J, Opahle I, Richter M, Rössler U K, Klaer P, Kallmayer M, Elmers H J, Jakob G, Schultz L and
Fähler S 2009 Full tunability of strain along the fcc-bcc bain path in epitaxial films and consequences for magnetic properties Phys. Rev. Lett. 103 21 216101
68 Slater J C 1930 Atomic shielding constants Phys. Rev. 35 pp 509-529 69 van der Merwe J H 1963 Crystal Interfaces. Part I. Semi-Infinite Crystals J. Appl. Phys. 34 123 70 Godlevsky V V and Rabe K M 2001 Soft tetragonal distortions in ferromagnetic Ni2MnGa and related
materials from first principles Phys. Rev. B 63 134407 71 Tsuchiya K, Nojiri T, Ohtsuka H and Umemoto M 2003 Effect of Co and Ni on martensitic transformation and
magnetic properties in Fe-Pd ferromagnetic shape memory alloys: Structural and functional control of materials through solid-solid phase transformations in high magnetic field Mater. Trans. JIM 44 12 pp 2499-2502
72 Vokoun D, Wang Y W, Goryczka T and Hu C T 2005 Magnetostrictive and shape memory properties of Fe-Pd alloys with Co and Pt additions Smart Mater. Struct. 14 p 261
73 Sánchez-Alarcos V, Recarte V, Pérez-Landazábal J I, Gómez-Polo C, Chernenko V A and González M A 2008 Reversible and irreversible martensitic transformations in Fe-Pd and Fe-Pd-Co alloys Eur. Phys. J. Special Topics 158 p 107
74 Kovacs A, Sato K and Hirotsu Y 2007 Strong perpendicular magnetic anisotropy of Fe-Pd nanocrystalline particles enhanced by Co addition J. Appl. Phys. 101 033910
75 Lin Y C, Lin C F, Yang J B and Lee H T 2011 Microstructures and magnetostriction of two-phase Fe66-Pd30-Ni4 high-temperature ferromagnetic shape memory alloys J. Appl. Phys. 109 07A912
76 Lin Y C and Lin C F 2012 Effects of Ni addition on the magnetostriction and microstructures of Fe70-xPd30Nix high-temperature ferromagnetic shape memory alloys J. Appl. Phys. 111 07A902
77 Stern R A, Willoughby S D, Ramirez A, MacLaren J M, Cui J, Pan Q and James R D 2002 Electronic and structural properties of FePd-Pt ferromagnetic shape memory alloys J. Appl. Phys. 91 10 pp 7818-7820
78 Wada T, Tagawa T, Taya M 2003 Martensitic transformation in Pd-rich Fe-Pd-Pt alloy Scripta Mater. 48 pp 207-211
79 Takeuchi I and Wuttig M 2007 High-throughput screening of magnetic properties of quenched metallic-alloy thin-film composition spreads J. Appl. Surf. 254 pp 734-737
80 Lin Y, Lee H T, Jen S U and Chen Y T 2007 Magnetic structure of an Fe-Pd-Rh alloy J. Appl. Phys. 101 09N514
81 Sánchez-Alarcos V, Recarte V, Perez-Landazabal J I, Gonzalez M A, Rodriguez-Velamazan J A 2009 Effect of Mn addition on the structural and magnetic properties of Fe-Pd ferromagnetic shape memory alloys Acta Mater. 57 pp 4224-4232
82 Opahle I, Koepernik K, Nitzsche U and Richter M 2009 Jahn-Teller-like origin of the tetragonal distortion in disordered Fe-Pd magnetic shape memory alloys Appl. Phys. Lett. 94 072508
83 Heczko O 2005 Magnetic shape memory effect and magnetization reversal J. Magn. Magn. Mater. 290 p 787 84 Liot F and Abrikosov I A 2009 Local magnetovolume effects in Fe65Ni35 alloys Phys. Rev. B 79 014202 85 Boettcher A, Haase G and Thun R 1955 Strukturuntersuchung von Mehrstoffsystemen Durch Kinematische Elektronenbeugung Z. Metallkde. 46 5 pp 386-400 86 Kennedy K, Stefansky T, Davy G, Zackay V F and Parker E R 1965 Rapid Method for Determining Ternary-Alloy Phase Diagrams J. Appl. Phys. 36 3808 87 Hanak J J J 1970 The “multiple-sample concept” in materials research: Synthesis, compositional analysis and
testing of entire multicomponent systems Mater. Sci. 5 pp 964-971 88 Xiang X D, Sun X, Briceno G, Lou Y, Wang K A, Chang H, Wallace-Freedman W G, Chen S W and Schultz
P G 1995 Combinatorial Approach to Materials Discovery Science 268 pp 1738-1740 89 Takeuchi I, Lauterbach J and Fasolka M J 2005 Combinatorial materials synthesis Mater. Today 8 pp 18-26 90 Takeuchi I, Famodu O O, Read J C, Aronova M A, Chang K S, Craciunescu C, Lofland S E, Wuttig M,
Wellstood F C, Knauss L and Orozco A 2003 Identification of novel compositions of ferromagnetic shape-memory alloys using composition spreads Nat. Mater. 2 pp 180-184
91 Tuinstra H E and Cummins C H 2000 Combinatorial Materials and Catalyst Research Adv. Mater. 12 pp 1819-1822
92 Wang J, Yoo Y, Gao C, Takeuchi I, Sun X, Chang H, Xiang X D and Schultz P G 1998 Identification of a Blue Photoluminescent Composite Material from a Combinatorial Library Science 279 pp 1712-1714
166
93 Hoogenboom R, Meier M A R and Schubert U S 2003 Combinatorial Methods, Automated Synthesis and
High-Throughput Screening in Polymer Research: Past and Present Macromol. Rapid. Comm. 24 15-32 94 Potyrailo R, Rajan K, Stoewe K, Takeuchi I, Chisholm B and Lam H 2011 Combinatorial and High-
Throughput Screening of Materials Libraries: Review of State of the Art ACS Comb. Sci. 13, 579-633 95 Jansen M 2002 A concept for synthesis planning in solid-state chemistry Angew. Chem. 41 pp 3746-3766 96 Cawse J 2001 Experimental Strategies for Combinatorial and High-Throughput Materials Development Acc.
Chem. Res. 34 pp 213-221 97 Göpel W 1998 Chemical imaging: I. Concepts and visions for electronic and bioelectronic noses Sens.
Actuators B 52 pp 125-142 98 Zhanpeng J A 1981 A study of the range of stability of σ phase in some ternar systems Scand. J. Metall. 10 pp
279-287 99 Miller N C and Shirn G A 1967 Co-sputtered Au-SiO2 cermet films Appl. Phys. Lett. 10 pp 86-88 100 Goldfarb I, Zolotoyabko E, Berner A and Shechtman D 1994 Novel sample preparation technique for the
study of multicomponent phase diagrams Mater. Lett. 21 pp 149-154 101 Dahn J R, Trussler S, Hatchard T D, Bonakdarpour A, Mueller-Neuhaus J R, Hewitt K C and Fleischauer M
2002 Economical Sputtering System To Produce Large-Size Composition-Spread Libraries Having Linear and Orthogonal Stoichiometry Variations Chem. Mater. 14 pp 3519-3523
102 Gremaud R, Broedersz C P, Borsa D M, Borgschulte A, Mauron P, Schreuders H, Rector J H, Dam B and Griessen R 2007 Hydrogenography: An Optical Combinatorial Method To Find New Light-Weight Hydrogen-Storage Materials Adv. Mater. 19 pp 2813-2817
103 Wu K H, Hung C I, Ziolkowski P, Platzek D, Karpinski G, Stiewe C and Mueller E 2009 Improvement of spatial resolution for local Seebeck coefficient measurements by deconvolution algorithm Rev. Sci. Instrum. 80 pp 105104-105108
104 Chang H, Gao C, Takeuchi I, Yoo Y, Wang J, Schultz P G, Xiang X D, Sharma R P, Downes M and Venkatesan T 1998 Combinatorial synthesis and high throughput evaluation of ferroelectric/dielectric thin-film libraries for microwave applications Appl. Phys. Lett. 72 pp 2185-2187
105 Woo N C, Ng B G and van Dover R B 2007 High-throughput combinatorial study of local stress in thin film composition spreads Rev. Sci. Instrum. 78 072208 5
106 Sharpe W N, Yuan B, Vaidyanathan R and Edwards R L 1996 New test structures and techniques for measurement of mechanical properties of MEMS materials. in Microlithography and metrology In Micromachining II, ed M T Postek and C Friedrich pp 78-91
107 Hemker K and Sharpe W N 2007 Microscale Characterization of Mechanical Properties Annu. Rev. Mater. Res. 37 pp 93-126
108 Gally B J, Abnet C C, Brown S and Chui C 2000 Wafer scale testing of MEMS structural films In Thin films-stresses and mechanical properties VIII, ed R Vinci, O Kraft, N Moody, P Besser and E Shaffer pp 195-200
109 Gaspar J, Schmidt M, Held J and Paul O 2007 Reliability of MEMS Materials: Mechanical Characterization of Thin-Films Using the Wafer Scale Bulge Test and Improved Microtensile Techniques Material Research Society Symposium (Boston USA) ed D A LaVan, M G DaSilva, S M Spearing and S Vengallatore vol 1052 DD01 supplement: 02 pp 15-20
110 Hamann S, Ehmann M, Thienhaus S, Savan A and Ludwig A 2008 Micro-hotplates for high-throughput thin film processing and in situ phase transformation characterization Sens. Actuators A 147 pp 576-582
111 McCluskey P J, Zhao Z W, Kfi O and Vlassak J J 2011 Precipitation and thermal fatigue in Ni-Ti-Zr shape memory alloy thin films by combinatorial nanocalorimetry Acta Mater. 59 13 pp 5116-5124
112 Takeuchi I, Long C J, Famodu O O, Murakami M, Hattrick-Simpers J, Rubloff G W, Stukowski M and Rajan K 2005 Data management and visualization of x-ray diffraction spectra from thin film ternary composition spreads Rev. Sci. Instrum. 76 pp 62223-62228
113 Long C J, Hattrick-Simpers J, Murakami M, Srivastava R C, Takeuchi I, Karen V L and Li X 2007 Rapid structural mapping of ternary metallic alloy systems using the combinatorial approach and cluster analysis Rev. Sci. Instrum. 78 072217
114 Lüth H 1993 Solid Surfaces, Interfaces and Thin Films (Berlin: Springer) 5th ed pp 88-131 115 Ohring M 2001 Material Science of Thin Films (Waltham, Massachusetts: Academic Press) 2nd ed 116 Bauer E 1958 Phänomenologische Theorie der Kristallabscheidung an Oberflächen Z. Kristallogr. 110 372 p
1820 117 Volmer M and Weber A 1926 Keimbildung in übersättigten Gebilden Z. Phys. Chem. 119 277 118 Stranski J N and Krastanov L 1938 Zur Theorie der orientierten Ausscheidung von Ionenkristallen
aufeinander Ber. Akad. Wiss. Wien 146 797 119 Frank F C and Van der Merwe J H 1949 One-Dimensional Dislocations. I. Static Theory Proc. R. Soc.
London A 198 205 120 Thornton J A 1977 High rate thick film growth Annu. Rev. Mater. Sci. 7 1 pp 239-260
167
121 Walton D, Rhodin T N and Rollins R W 1963 Nucleation of silver on sodium chloride J. Chem. Phys. 38
2698 122 Chang C A 1992 Metal alloy-films containing body-centered cubic and face-centered cubic metals - Relation
between structures and lattice spacings J. Appl. Phys. 72 p 4463 123 Jones H 1973 Splat cooling and metastable phases Rep. Prog. Phys. 36 11 p 1425 124 Duwez P, Willens R H and Klement W J 1960 Continuous Series of Metastable Solid Solutions in Silver-
Copper Alloys J. Appl. Phys. 31 1136 125 Edmund Bühler GmbH: Rascherstarrungstechnologie, Splat Cooling, April 2010 http://www.edmund-
buehler.de/i-rascherstarrungstechnologie.pml 126 Oxford Instruments Inca Manual 2009 127 Casino software available at www.gel.usherb.ca/casino 128 Drouin D, Couture A R, Joly D, Tastet X, Aimez V and Gauvin R. 2007 A Fast and Easy-to-use Modeling
Tool for Scanning Electron Microscopy and Microanalysis Users Scanning 29 pp 92-101 129 Czyzewski Z, MacCallum D O, Romig A and Joy D C 1990 Calculations of Mott scattering cross section
J. Appl. Phys. 68 7 pp 3066-3072 130 Drouin D, Hovington P and Gauvin R 1997 A New Monte Carlo Code in C Language for Electron Beam
Interaction - Part II: Tabulated Values of the Mott Cross Section Scanning 19 pp 20-28 131 http://www.esrf.eu/computing/scientific/FIT2D/ 132 Cullity B D 1956 Elements of X-Ray Diffraction (Massachusetts: Addison-Wesley Publishing Company, Inc.)
1st ed 133 Warren B E X-Ray Diffraction (Boston: Addison-Wesley Publishing Co.) 1st ed pp 251-254 134 Spieß L, Schwarzer R, Behnken H and Teichert G 2005 Moderne Röntgenbeugung (Wiesbaden: B. G.
Teubner) 1st ed pp 263-284 135 Birkholz M 2006 Thin Film Analysis by X-Ray Scattering (Weinheim: Wiley-VCH) 1st ed 136 Prevéy P S 1986 The use of pearson VII distribution functions in X-ray diffraction residual stress
measurement Adv. X-Ray Anal. 29 pp 103-111 137 Noyan I C and Cohen J B 1987 Residual Stress Measurement by Diffraction and Interpretation ed B Ilschner
and N J Grant (Berlin: Springer) 1st ed 138 Cheng W and Finnie I 2007 Residual Stress Measurement and the Slitting Method (Berlin: Springer) 1st ed 139 Takeuchi I, Long C J, Famodu O O, Murakami M, Hattrick-Simpers J, Rubloff G W, Stukowski M and Rajan
K 2005 Data management and visualization of x-ray diffraction spectra from thin film ternary composition spreads Rev. Sci. Instrum. 76 62223-62228
140 Long C J, Hattrick-Simpers J, Murakami M, Srivastava R C, Takeuchi I, Karen V L and Li X 2007 Rapid structural mapping of ternary metallic alloy systems using the combinatorial approach and cluster analysis Rev. Sci. Instrum. 78 072217
141 Long C J, Bunker D, Li X, Karen V L and Takeuchi I 2009 Rapid identification of structural phases in combinatorial thin-film libraries using x-rays diffraction and none-negative matrix factorization Rev. Sci. Instrum. 80 103902-103906
142 Cho H, Kim H Y and Miyazaki S 2005 Fabrication and characterization of Ti-Ni shape memory thin film using Ti/Ni multilayer technique Sci. Tech. Adv. Mater. 6 pp 678-683
143 Inoue S, Inoue K, Fujita S and Koterazawa K 2003 Fe-Pd Ferromagnetic Shape Memory Alloy Thin Films Made By Dual Source DC Magnetron Sputtering Mater. Trans. 44 2 298
144 Lo Y C and Wu S K 1993 A study of B2-B19-B19´ two-stage martensitic transformation in a Ti50Ni40Cu10 alloy, Acta Metall. Mater. 41 3 pp 747-749
145 Faulkner M G, Amalraj J J and Bhattacharyya A 2000 Experimental determination of thermal and electrical properties of Ni-Ti shape memory wires Smart Mater. Struct. 9 pp 632-639
146 Uchil J 2002 Shape memory alloys - characterization techniques Pramana Indian Academy of Sciences 58 1131-1139
147 Wu S K, Lin H C and Lin T Y 2006 Electrical resistivity of Ti-Ni binary and Ti-Ni-X (X = Fe, Cu) ternary shape memory alloys Mater. Sci. Eng. A 438-440 pp 536-539 148 Wan D and Komvopoulos K 2005 Thickness Effect on Thermally Induced Phase Transformations in
Sputtered Titanium-nickel Shape-memory Films J. Mater. Res. 20 1606 149 Mohanchandra K P, Ho K K and Carman G P 2004 Electrical Characterization of NiTi Film on Silicon
Substrate J. Intell. Mater. Syst. Struct. 15 387 150 Thienhaus S, Zamponi C, Rumpf H, Hattrick-Simpers J, Takeuchi I and Ludwig A 2006 High-throughput
characterization of shape memory thin films using automated temperature-dependent resistivity measurements Mater. Res. Soc. Symp. Proc. 894 0894-LL06-06.1
151 Kuz’min M D 2005 Shape of Temperature Dependence of Spontaneous Magnetization of Ferromagnets: Quantitative Analysis Phys. Rev. Lett. 94 107204
168
152 Kuz’min M D, Richter M and Yaresko A N 2006 Factors determining the shape of the temperature
dependence of the spontaneous magnetization of a ferromagnet Phys. Rev. B 73 100401R 153 Pharr G M, Strader J H and Oliver W C 2009 Critical issues in making small-depth mechanical property
measurements by nanoindetation with continuous stiffness measurement J. Mater. Res. 24 653 154 Oliver W C and Pharr G M 1992 An improved technique for determining hardness and elastic modulus using
load and displacement sensing indentation experiments J. Mater. Res. 7 1564 155 Oliver W C and Pharr G M 2004 Measurement of hardness and elastic modulus by instrumented indentation:
Advances in understanding and refinements to methodology J. Mater. Res. 19 3 20 156 Pfetzing J 2010 Nanoindentation von NiTi-Formgedächtnislegierungen PhD thesis (Institute for Materials,
Ruhr-Universität Bochum, Germany) 157 Sree Harsha K S 2006 Principles of Physical Vapor Deposition of Thin Films Elsevier Science & Technology 158 Kaufmann S, Buschbeck J, Heczko, Schultz L and Fähler S 2008 Epitaxial growth of Ni-Mn-Ga:
Consequences of Magnetron configuration on martensitic behaviour Proc. ICOMAT (Santa Fe) 159 Kühnemund L, Edler T, Kock I, Seibt M and Mayr S G 2009 Epitaxial growth and stress relaxation of vapor-deposited Fe-Pd magnetic shape memory alloy films New J. Phys. 11 113054 160 De Keijser T H, Langford J I, Mittemejer E J and Vogels A B P 1982 Use of the Voigt function in a single-line method for the analysis of X-ray diffraction line broadening J. Appl. Cryst. 15 pp 308-314 161 Fick A 1855 Ueber diffusion Ann. Phys. 94 pp 59-86 162 Gavens A J, Van Heerden D, Mann A B, Reiss M E and Weihs T P 2000 Effect of intermixing on self-
propagating exothermic reactions in Al/Ni nanolaminate foils J. Appl. Phys. 87 3 pp 1255-1263 163 Kock I, Hamann S, Brunken H, Edeler T, Ludwig A and Mayr S G 2010 Development and characterization of Fe70Pd30 ferromagnetic shape memory splats Intermetallics 18 pp 877-882 164 Zarnetta R, Savan A, Thienhaus S and Ludwig A 2007 Combinatorial study of phase transformation
characteristics of a Ti-Ni-Pd shape memory thin film composition spread in view of microactuator applications Appl. Surf. Sci. 254 743 8
165 Viret M, Samson Y, Warin P, Marty A, Ott F and Søndergård E 2000 Anisotropy of domain wall resistance Phys. Rev. Lett. 85 3962
166 Guenin G and Gobin P F 1982 A localized soft mode model for the nucleation of thermoelastic martensitic transformations: Application to the β → 9r transformation Metall. Mater. Trans. 13 1127 34
167 Dondl P and Bhattacharya K 2007 The effect of precipitates on the evolution of a martensitic phase boundary Proc. Appl. Math. Mech. 7 1151207
168 Seki K, Kura H, Sato T, Taniyama T 2008 Size dependence of martensite transformation temperature in ferromagnetic shape memory alloy FePd J. Appl. Phys. 103 063910
169 Kang S, Lee Y, Lim Y, Nam J, Nam T and Kim Y 2008 Relationship between grain size and martensitic transformation start temperature in a Ti-30Ni-20Cu alloy ribbon Scripta Mater. 59 1186 9
170 Levine R D 2009 Molecular Reaction Dynamics (Cambridge: University Press) 1st ed 171 Wollants P, Roos J, Delaely L 1993 Thermally- and Stress-Induced Thermoelastic Martensitic
Transformations in the Reference Frame of Equilibrium Thermodynamics Progr. Mater. Sci. 37 227 88 172 Kock I 2010 Phasenbildung, Phasenübergang und mechanische Eigenschaften des Funktionsmaterials Eisen-
Palladium PhD thesis (I. Physical Institut, Georg-August-Universität Göttingen) 173 Buschbeck J, Lindemann I, Schultz L and Fähler S 2007 Growth, structure, and texture of epitaxial Fe100-xPdx
films deposited on MgO(100) at room temperature: An x-ray diffraction study Phys. Rev. B 76 205421 174 Buschbeck J, Niemann R, Heczko O, Thomas M, Schultz L and Fähler S 2009 In situ studies of the
martensitic transformation in epitaxial Ni-Mn-Ga films Acta Mater. 57 2516 175 Onisan A T, Bogdanov A N and Rössler U 2010 Domain models for ferromagnetic shape-memory materials
Acta Mater. 58 13 4378-4386 176 Edler T and Mayr S G 2010 Film lift-off from MgO: Freestanding single crystalline Fe-Pd films suitable for
magnetic shape memory actuation - and beyond Adv. Mater. 22 4969 177 Edler T, Hamann S, Ludwig A and Mayer S G 2011 Reversible fcc ↔ bcc transition in freestanding epitaxially grown Fe-Pd ferromagnetic shape memory films Scripta Mater. 64 pp 89-92 178 Kakeshita T and Fukuda T 2006 Magnetic field-control of microstructure and function of materials exhibiting
solid-solid phase transformation Sci. Technol. Adv. Mater. 7 350 179 Oshima R and Sugiyama M 1982 Martensite transformations in iron-palladium alloys J. Phys. Colloq. 43 C4 180 Kühnemund L, Edler T, Kock I, Seibt M and Mayr S G 2009 Epitaxial growth and stress relaxation of vapor-
deposited Fe-Pd magnetic shape memory films New J. Phys. 11 113054 181 Gruner M E, Adeagbo W A, Zayak A T, Hucht A and Entel P 2010 Lattice dynamics and structural stability
of ordered Fe3Ni, Fe3Pd and Fe3Pt alloys using density functional theory Phys. Rev. B 81 064109
169
182 Bechtold C, Buschbeck J, Lotnyk A, Erkartal B, Hamann S, Zamponi C, Schulz L, Ludwig A, Kienle L,
Fähler S and Quandt E 2010 Artificial single variant martensite in freestanding Fe70Pd30 films obtained by coherent epitaxial growth Adv. Mater. 22 1-4
183 Wagner M F X and Windl W 2008 Lattice stability, elastic constants and macroscopic moduli of NiTi martensites from first principles Acta Mater. 56 6232
184 Buschbeck J, Opahle I, Fähler S, Schultz L and Richter M 2008 Magnetic properties of Fe-Pd magnetic shape memory alloys: Density functional calculations and epitaxial films Phys. Rev. B 77 174421
185 Thienhaus S, Hamann S and Ludwig A 2011 Modular high-throughput test-stand for the versatile screening of thin film materials libraries Sci. Technol. Adv. Mat. 12 054206 186 Acet M, Schneider T, Gehrmann B and Wassermann E F 1995 The Magnetic aspects of the martensitic transformations in Fe-Mn Alloys J. Phys. IV Colloque C8 p 379 187 Klimars S, Hesse J and Huck B 1985 Mössbauer spectroscopy on iron rich Fe-Pd alloys including fcc invar J. Magn. Magn. Mater. 51 183 188 Periodic table 2006 (Illinois, Skokie: Sargent-Welch Scientific Company) 189 Acet M, Wassermann E F, Andersen K, Murani A and Schärpf O 1997 The anomalous temperature dependence of the paramagnetic response of Fe-rich fcc Fe-Ni Europhys. Lett. 40 93 190 Kauffmann-Weiss S, Hamann S, Gruner M E, Buschbeck J, Ludwig A, Schultz L and Fähler S 2012 Understanding the Magnetic Shape Memory System Fe-Pd-X by Thin Film Experiments and First Principles Calculations Adv. Eng. Mater. 14 8 pp 724-749 191 Gruner M E, Hamann S, Brunken H, Ludwig A and Entel P 2012 Compositional trends and magnetic excitations in binary and ternary Fe-Pd(-X) magnetic shape memory alloys J. Alloys Compd. <doi:10.1016/j.jallcom.2012.02.033> 192 Ebert H 2009 Lecture notes in physics: Electronic structure and physical properties of solids 535 191 (Berlin:
Springer Verlag) ed H Dreysse 193 Ebert H <http://olymp.cup.uni-muenchen.de/ak/ebert/SPRKKR> 194 Perdew J P, Burke K and Ernzerhoff M 1996 Generalized gradient approximation made simple Phys. Rev.
Lett. 77 3865 195 Gyorffy B L, Pindor A J, Staunton J, Stocks G M and Winter H J 1985 A first-principles theory of
ferromagnetic phase transitions in metals J. Phys. F: Met. Phys. 15 1337 196 Khmelevkyi S, Turek I and Mohn P 2003 Large negative magnetic contribution to the thermal expansion in
iron-platinum alloys: quantitative theory of the Invar effect Phys. Rev. Lett. 91 037201 197 Liechstenstein A I, Katsnelson M I, Antropov V P and Gubanov V A 1987 Local spin density functional
approach to the theory of exchange interactions in ferromagnetic metals and alloys J. Magn. Magn. Mater. 15-18 1201
198 Okamoto H 1993 Phase Diagrams of Binary Iron Alloys (Materials Park/Ohio: ASM International) 199 Hamann S, Gruner M E, Irsen S, Buschbeck J, Bechtold C, Kock I, Mayr S G, Savan A, Thienhaus S, Quandt E, Fähler S, Entel P and Ludwig A 2010 The ferromagnetic shape memory system Fe-Pd-Cu Acta Mater. 58 pp 5949-5961 200 Kuang J P, Kontani M, Matsui M and Adachi K 1988 Electronic, phonon and magnon specific heats of FePd
alloys Physica B 149 209 201 Edler T, Buschbeck J, Mickel C, Fähler S and Mayr S G 2008 Mechanisms of stress generation and relaxation
during pulsed laser deposition of epitaxial Fe-Pd magnetic shape memory alloy films on MgO New J. Phys. 10 063007
202 Kock I, Edler T and Mayr S G 2008 Growth behavior and intrinsic properties of vapor-deposited iron palladium thin films J. Appl. Phys. 103 046108
203 Kim H J, Han J H, Kaiser R, Oh K H and Vlassak J J 2008 High-throughput analysis of thin-film stresses using arrays of micromachined cantilever beams Rev. Sci. Instrum. 79 045112
204 Khmelevskyi S and Mohn P 2004 Magnetostriction in Fe-based alloys and the origin of the Invar anomaly Phys. Rev. B 69 140404R
205 Kauffmann-Weiss S, Hamann S, Gruner M E, Schultz L, Ludwig A and Fähler S 2012 Enhancing magnetocrystalline anisotropy of the Fe70Pd30 magnetic shape memory alloy by adding Cu Acta Mater. 60 pp 6920-6930 206 Kauffmann-Weiss S, Gruner M E, Backen A, Schultz L, Entel P and Fähler S 2011 Magnetic Nanostructures by Adaptive Twinning in Strained Epitaxial Films Phys. Rev. Lett. 107 206105 207 Kauffmann A 2011 <CorrVert http://sco.ifw-dresden.de> 208 v Steinwehr H E 1967 Ursachen der Abweichungen von der Vegardschen Regel Z. Kristallogr. 1967 125 360 209 Nishiyama Z 1978 Martensitic transformation (New York: Academic Press) 210 Kakeshita T and Fukuda T 2006 Energy evaluation for twinning plane movement under magnetic field in ferromagnetic shape memory alloys Int. J. Appl. Electromagn. Mech. 23 45
170
211 Buschbeck J, Hamann S, Ludwig A, Holzapfel B, Schultz L and Fähler S 2010 Correlation of Phase Transformations and Magnetic Properties in Annealed Epitaxial Fe-Pd Magnetic Shape Memory Alloy Films J. Appl. Phys. 107 113919 212 Kondorsky E J 1940 A mechanism of magnetic hysteresis in heterogeneous alloys Phys (USSR)2 161 213 Buschbeck J, Heczko O, Ludwig A, Fähler S and Schultz L 2008 Magnetic Properties of epitaxial Fe-Pd films measured at elevated temperatures J. Appl. Phys. 103 07 B334 214 Straka L and Heczko O 2003 Magnetic anisotropy in Ni-Mn-Ga martensites J. Appl. Phys. 93 8636 215 Polesya S, Mankovsky S, Sipr O, Meindl W, Strunk C and Ebert H 2010 Finite-temperature magnetism of FexPd1−x and CoxPt1−x alloys Phys. Rev. B 82 214409 216 Gruner M E, Entel P, Minar J, Polesya S, Mankovsky S and Ebert H J 2012 Electronic and magnetic trends in martensitically transforming Fe-Pd alloys Magn. Magn. Mater. <http://dx.doi.org/10.1016/j.jmmm.2012.02.081> 217 Tickle R and James R D 1999 Magnetic and magnetomechanical properties of Ni2MnGa J. Magn. Magn. Mater. 195 627 218 Pepperhoff W and Acet M 2001 Engineering materials (Berlin: Springer Verlag) vol VIII 219 Wassermann E F 1990 Ferromagnetic materials (Amsterdam: Elsevier) ed K H J Buschow and E P Wohlfarth vol 5 p 237 220 Buschow K H J and de Boer F R 2004 Physics of magnetism and magnetic materials (New York: Kluwer) vol 97 221 Klaer P, Eichhorn T, Jakob G and Elmers H J 2011 Microscopic origin of magnetic anisotropy in martensitic Ni2MnGa Phys. Rev. B 83 214419 222 Callen E R and Callen H B 1960 Anisotropic magnetization J. Phys. Chem. Solids 16 310 223 Herzer G J 1992 Nanokristalline soft magnetic materials Magn. Magn. Mater. 112 258 224 Claussen I, Brand R, Hahn H and Mayr S G 2012 Relaxation scenarios in Fe-Pd and Fe-Pd-Cu ferromagnetic shape memory splats Scripta Mater. 66 163 225 Barabash S V, Chepulskii R V, Blum V and Zunger A 2009 First-principles determination of low-temperature order and ground states of Fe-Ni, Fe-Pd, and Fe-Pt Phys. Rev. B 80 220201 226 Chepulskii R V, Barabash S V and Zunger A 2012 Ab initio theory of phase stability and structural selectivity in Fe-Pd alloys Phys. Rev. B 85 144201 227 Ren X and Otsuka K 1997 Origin of the Rubber-like Behavior in Metal Alloys Nature 389 579
171
Publications
Peer reviewed publications on ferromagnetic shape memory alloys
1) Kauffmann-Weiss S, Hamann S, Gruner M E, Schultz L, Ludwig A and Fähler S
2012 Enhancing magnetocrystalline anisotropy of the Fe70Pd30 magnetic shape
memory alloy by adding Cu Acta Mater. 60 pp 6920-6930
2) Kauffmann-Weiss S, Hamann S, Gruner M E, Buschbeck J, Ludwig A, Schultz L and
Fähler S 2012 Understanding the Magnetic Shape Memory System Fe-Pd-X by Thin
Film Experiments and First Principles Calculations Adv. Eng. Mater. 14 8 pp 724-749
3) Gruner M G, Hamann S, Brunken H, Ludwig A and Entel P 2012 Compositional
trends and magnetic excitations in binary and ternary Fe-Pd(-X) magnetic shape
memory alloys J. Alloys Compd. (accepted)
4) Thienhaus S, Hamann S and Ludwig A 2011, Modular high-throughput test-stand for
the versatile screening of thin film materials libraries, Sci. Technol. Adv. Mat. 12
054206
5) Lai Y W, Hamann S, Ehmann M and Ludwig A 2011 High-throughput
characterization of stresses in thin film materials libraries using Si cantilever array
wafers and digital holographic microscopy Rev. Sci. Instrum. 82 063903
6) Edler T, Hamann S, Ludwig A and Mayr S G 2011 Reversible fcc ↔ bcc transition in
freestanding epitaxially grown Fe-Pd ferromagnetic shape memory films Scripta
Mater. 64 1 pp 89-92
7) Hamann S, Gruner M E, Irsen S, Buschbeck J, Bechtold C, Kock I, Mayr S G, Savan
A, Thienhaus S, Quandt E, Fähler S, Entel P and Ludwig A 2010 The ferromagnetic
shape memory system Fe-Pd-Cu Acta Mater. 58 pp 5949-5961
8) Bechtold C, Buschbeck J, Lotnyk A, Erkartal B, Hamann S, Zamponi C, Schulz L,
Ludwig A, Kienle L, Fähler S and Quandt E 2010 Artificial single variant martensite
in freestanding Fe70Pd30 films obtained by coherent epitaxial growth Adv. Mater. 22
24 p 2668
172
9) Buschbeck J, Hamann S, Ludwig A, Holzapfel B, Schultz L and Fähler S 2010
Correlation of Phase Transformations and Magnetic Properties in Annealed Epitaxial
Fe-Pd Magnetic Shape Memory Alloy Films J. Appl. Phys. 107 113919
10) Kock I, Hamann S, Brunken H, Edler T, Mayr S G and Ludwig A 2010 Development
and characterization of Fe70Pd30 ferromagnetic shape memory splats Intermetallics 18
pp 877-882
11) Ludwig A, Zarnetta R, Hamann S, Savan A and Thienhaus S 2008 Development of
multifunctional thin films using high-throughput experimentation methods
Int. J. Mater. Res. 99 10 pp 1144-1149
Proceeding papers on ferromagnetic shape memory alloys
1) Hamann S, Gruner M E, Thienhaus S, Savan A and Ludwig A 2011 Combinatorial
Development of Fe-Pd-X thin film systems with improved intrinsic properties 3rd
International Conference on Ferromagnetic Shape Memory Alloys Proceeding,
(Dresden Germany) ed S Fähler 18.-22.07.2011 pp 140-141
2) Bechtold C, Buschbeck J, Lotnyk A, Erkartal B, Hamann S, Zamponi C, Schulz L,
Ludwig A, Kienle L, Fähler S and Quandt E 2010 Development Towards MSM
Active FePd Thick Films Actuator 2010 Conference Proceedings (Bremen Germany)
ed H Borgmann pp 299-302
3) Hamann S, Savan A, Thienhaus S and Ludwig A 2008 Combinatorial Development
of Fe-Pd-Mn Ferromagnetic Shape Memory Thin Films Actuator 2008 Conference
Proceedings (Bremen Germany) ed H Borgmann pp 271-274
Other peer reviewed publications
1) Hamann S, Brunken H, Salomon S, Meyer R, Savan A and Ludwig A 2013 Synthesis
of Au micro-wires by selective oxidation of Au-W thin film composition spreads
Sci. Technol. Adv. Mat. 14 1 p 015003
2) Grochla D, Siegel A, Hamann S, Buenconsejo P J S, Kieschnick M, Brunken H,
König D and Ludwig A 2013 Time and space resolved high-throughput
173
characterization of stresses during sputtering and thermal processing of Al-Cr-N thin
films J. Phys. D: Appl. Phys. 46 8 084011
3) Meyer R, Hamann S, Ehmann M, Thienhaus S, Jaeger S, Thiede T, Devi A, Fischer R
A and Ludwig A 2012 Micro-gradient-heaters as tools for high-throughput
experimentation ACS Comb. Sci. 14 10 pp 527-578
4) König D, Buenconsejo P J S, Grochla D, Hamann S, Pfetzing-Miklich J and Ludwig
A 2012 Thickness-dependence of the B2-B19 martensitic transformation in nanoscale
shape memory alloy thin films: zero hysteresis in 75 nm thick Ti51Ni38Cu11
Acta Mater. 60 1 pp 306-313
5) Zimmer C M, Schubert J, Hamann S, Kunze U and Doll T 2011 Nanoscale
photoelectron ionisation detector based on lanthanum hexaboride Phys. Status Solidi A
- Applications and Materials Science 208 6 pp 1241-1245
6) Meyer R, Hamann S, Ehmann M, König D,Thienhaus S, Savan A and Ludwig A
2011 Small Scale Deposition of Thin Films and Nanoparticles by Micro-Evaporation
Sources J. Microelectromech. S. 19 pp 1264-1269
7) Lai Y W, Koukourakis N, Gerhardt N C, Hofmann M R, Meyer R, Hamann S,
Ehmann M, Hackl K, Darakis E and Ludwig A 2010 Integrity of micro-hotplates
during high-temperature operation monitored by digital holographic microscopy,
J. Microelectromech. S. 19 5 pp 1175-1179
8) Hamann S, Ehmann M, Thienhaus S, Savan A and Ludwig A 2008 Micro-hotplates
as processing and characterization platforms for combinatorial materials science and
high-throughput experimentation Sensor. Actuat. A-Phys. 147 pp 576 - 582
174
Acknowledgments
The presented results, findings and interpretations were performed at the Centre for Advanced
European Studies and Research (caesar), Bonn in the beginning and finished at the Institute
for Materials, Faculty of Mechanical Engineering at the Ruhr-University of Bochum from
February 2007 to December 2012.
I would like to express my gratitude to my doctoral adviser Prof. Dr.-Ing. Alfred Ludwig, for
giving me the opportunity to perform my PhD thesis and for his trust and encouragement
during this time. Further I have to thank Prof. Dr.-Ing. Alfred Ludwig especially for allowing
me to have insights into many other fields of scientific work next to the topic of my PhD.
Additionally I have to thank many current and former members of his group, who helped me
and my scientific research in many ways and provided a pleasant working atmosphere.
Especially gratefulness goes to the following colleagues: Sigurd (for technical support in
many ways), Alan (for helping in many ways but especially sputtering), Michael Ehmann (for
MHPs and GHs, we got quite a few papers out), Hayo (for company and more starting from
the 1st semester), Dennis (for fruitful discussions and everything else), Robert (for his critical
considerations and being my cigarette break buddy), Alexander (for technical help and many
more things), Dario (for helping and cheering me up) and Pio (for inspiring discussions and
ideas). Further thanks go to all students that supported my scientific work.
Thanks also to Prof. Dr. Gunther Eggeler and many of his group members (Janine, Jan, Klaus
and Christoph) for the cooperative atmosphere and scientific support.
Additional gratitude goes to Prof. Dr.-Ing. F. Peters for supervising the process of my Ph. D.
defense.
Further, I would like to acknowledge many colleagues from other research institutions for the
nice and productive cooperation: Markus Gruner for very fruitful discussions bringing
understanding into the dark of experimental findings; Sebastian Fähler for inspiring
discussions and many advices; Jörg Buschbeck and Sandra Kauffmann-Weiss for this great
collaboration with the IFW in Dresden; Anja Backen and Robert Niemann for several fruitful
discussions; Christoph Bechtold and Eckhard Quandt for the nice company at caesar; Burak
and Lorenz for deep insight by TEM; Iris Claussen, Tobias Edler and Stefan Mayr for very
175
fruitful collaboration. Further many thanks to all others that I might have forgotten and
especially to all persons that participated in publishing this huge amount of papers.
Finally, I have to thank Christina for her support (especially during hellweek) and everthing
she did and my parents Geli and Gerd as well as my brother Dennis and his girlfriend Lina. I
additionally want to thank all my other relatives and friends for their moral support by
cheering me up in moments when I was down.
176
CURRICULUM VITAE
PERSONAL INFORMATION
Surname Hamann
Name Sven
Date of Birth 30 August 1979
Place of Birth Essen (Ruhr), Germany
Nationality German
WORK EXPERIENCE
Oct 2006 – present Research associate and lecturer
with Prof. Dr.-Ing. A. Ludwig, Chair for MEMS materials - Institute for Materials - Faculty of Mechanical Engineering, Ruhr-Universität Bochum, Germany
Project “Combinatorial development of Fe-Pd based ferromagnetic shape memory alloys“, part of the Deutsche Forschungsgemeinschaft (DFG)-priority programme SPP 1239
Mar – Apr 2011 Guest researcher
Leibniz Institute for Solid State and Materials Research Dresden (IFW Dresden), Germany
Epitaxial growth of Fe-Pd based ferromagnetic shape memory alloys.
Jan – May 2005 Student research assistant
Chair of Automatic Control and Systems Theory - Faculty of Mechanical Engineering, Ruhr-Universität Bochum, Germany
Programming of the waste-gallery system in the Airbus A380 (industrial project for AEROTEC)
Mar – Sept 2006 Diploma thesis “Die Charakterisierung und Anwendung von Mikro-Heizplatten im Bereich der kombinatorischen Materialforschung von dünnen Schichten“ (Characterization and application of micro hotplates in the field of combinatorial material science of thin films)
Prof. Dr.-Ing. A. Ludwig, Combinatorial Material Science (CMS), centre of advanced european studies and research (caesar), Bonn, Germany
177
EDUCATION
Oct 2003 – Sept 2006 Main Studies “Mechanical Engineering” (focus field Micro-Engineering) and diploma degree “Dipl.-Ing. Mechanical Engineering”
Ruhr-Universität Bochum, Germany
Feb – Aug 2003 Semester abroad
Universidad Politecnica De Valencia, Spain
Oct 2000 – Jan 2003 Basic studies “Mechanical Engineering” and intermediate diploma
University of Duisburg-Essen, Germany
1999 – 2005 Training as reserve officer / reservist lieutenant, including course of studies “Leadership”
Deutsche Bundeswehr (German Army)
1991 – 1999 Secondary School, German university entrance qualification
Helmholtz-Gymnasium, Essen, Germany
ACHIEVEMENTS
23 – 28 AUG 2010 Invited lecture at Joint European Magnetic Symposia (JEMS): “Structural, magnetic and phase transformation properties of Fe-Pd-X thin films”, Krakow, Poland
MAR 2010 “Best presentation” during the evaluation of the last funding period of the Deutsche Forschungsgemeinschaft (DFG)-priority programme SPP 1239
FEB – AUG 2003 ERASMUS-scholarship for an academic semester abroad
2002 – 2006 Scholarship holder of the “Prof. Dr. Koepchen Academic Foundation” (RWE AG, Essen, Germany)
NATIVE LANGUAGE German
LANGUAGES English (fluent), Spanish (advanced knowledge), French (basic knowledge)