Magneto-Optical Investigations of Multiferroic Hybrid ... · 3.2 Ferroelectric domains in BaTiO ......
Transcript of Magneto-Optical Investigations of Multiferroic Hybrid ... · 3.2 Ferroelectric domains in BaTiO ......
TECHNISCHE
UNIVERSITAT
MUNCHEN
WALTHER - MEISSNER -
INSTITUT FUR TIEF -
TEMPERATURFORSCHUNG
BAYERISCHE
AKADEMIE DER
WISSENSCHAFTEN
Magneto-Optical Investigations of
Multiferroic Hybrid Structures
Diploma Thesis
Matthias Brasse
Advisor: Prof. Dr. Rudolf Gross
Munich, October 2009
Abstract
Multiferroic hybrid structures, which hold the promise of an electric field control of fer-
romagnetism, are investigated by means of the magneto-optical Kerr effect.
As part of this thesis, a magneto-optical Kerr setup allowing spatially resolved Kerr
imaging has been established. This setup is presented, characterized and its capabilities
in large scale magnetic domain imaging are demonstrated.
The studied multiferroic hybrid structures consist of a ferromagnetic thin film and a
ferroelectric substrate and have mostly been prepared during this thesis. Nickel and iron-
cobalt have been chosen as ferromagnetic constituent and either piezoelectric actuators
based on lead circonate ceramics or barium titanate single crystals serve as ferroelectric
substrate. In these hybrid structures magnetization control by means of electric fields
is achieved by making use of the piezoelectric and the magnetoelastic effect. In all pre-
pared samples the mechanical strain dependence of the magnetic hysteresis is studied
extensively. On top of that, differences in the evolution of the magnetic microstructure
in dependence of the applied strain are observed and pointed out. In the piezoelectric
actuator based samples it is found that the strain coupling is strongly affected by the
used fabrication technique and by the choice of the ferromagnetic constituents. However,
in both, the actuator based hybrids as well as the barium titanate composites, magne-
toelastic coupling unambiguously influences the anisotropy of the attached ferromagnetic
thin film. Moreover, magneto-optical experiments are successfully performed regarding
electric field induced magnetization switching. Magnetization control by means of electric
fields is therefore established for all investigated samples.
In addition, experiments aiming at local magnetization control have been performed.
Local strain control is achieved by a modification of the actuator based hybrid struc-
tures and allows to control the magnetization orientation selectively. Applying the
magneto-optical Kerr technique, local magnetization switching is confirmed. Addition-
ally, magnetotransport measurements are applied to correlate the observed domain struc-
ture with anisotropic magnetoresistance data. Good agreement between both experiments
is achieved. Finally a novel sample design consisting of patterned nickel electrodes on
piezoelectric substrates is proposed and investigated. First experiments prove that local
strain and magnetization control in these samples is also possible.
Taken together, these results show that an electric field control of ferromagnetism is
possible via the elastic channel both on macroscopic (mm) as well as on microscopic (µm)
length scale.
I
Contents
Abstract I
1 Introduction 1
2 Theoretical Basics 5
2.1 Ferroics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.1 Ferromagnetism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.2 Ferroelectricity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2 Multiferroics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2.1 Piezoelectric Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.2 Magnetoelastic Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3 Magneto-Optical Kerr Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3.1 Phenomenology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3.2 Microscopic Origin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3.3 Macroscopic Description . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3.4 Quantitative Formulation . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3.5 Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3 Materials 21
3.1 Ferromagnetic Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.1.1 Nickel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.1.2 Iron-Cobalt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2 Piezoelectric Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2.1 Lead Zirconate Titanate . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2.2 Barium Titanate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2.3 Lithium Niobate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4 Magneto-Optical Kerr Effect with Spatial Resolution 27
4.1 The Existing Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.2 Kerr Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.3 Image Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.4 Resolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.5 Characterization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
III
IV Contents
4.5.1 CCD Noise and Signal-to-Noise Ratio . . . . . . . . . . . . . . . . . . 34
4.5.2 Magneto-Optical Contrast . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.5.3 Sensitivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.6 Imaging of Thin Ferromagnetic Films . . . . . . . . . . . . . . . . . . . . . . 39
4.7 Quantitative Kerr Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5 Magneto-Optics on Ferromagnetic/Ferroelectric Hybrids 45
5.1 Ferromagnetic Thin Films on Piezoelectric Actuators . . . . . . . . . . . . . 45
5.1.1 Sample Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5.1.2 Strain Control of the Magnetic Anisotropy . . . . . . . . . . . . . . . 47
5.1.3 Voltage Dependence of the Magnetic Hysteresis . . . . . . . . . . . . 49
5.1.4 Domain Evolution at Constant Strain . . . . . . . . . . . . . . . . . . 52
5.1.5 Strain Induced Birefringence . . . . . . . . . . . . . . . . . . . . . . . 58
5.1.6 Extraction of the Magnetic Contribution to Birefringence . . . . . . 61
5.1.7 Strain Induced Magnetization Switching . . . . . . . . . . . . . . . . 62
5.2 Ferromagnetic Thin Films on BaTiO3 Substrates . . . . . . . . . . . . . . . . 64
5.2.1 Sample Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.2.2 M(H) Measurements at Constant Strain . . . . . . . . . . . . . . . . 66
5.2.3 Magnetization Switching at Constant Magnetic Field . . . . . . . . . 68
5.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
6 Local Magnetization Control 71
6.1 Ferromagnets on Piezoelectric Actuators . . . . . . . . . . . . . . . . . . . . . 71
6.1.1 Sample Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
6.1.2 Local Polarization Control . . . . . . . . . . . . . . . . . . . . . . . . . 73
6.1.3 Local Strain Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
6.1.4 Local Magnetization Switching . . . . . . . . . . . . . . . . . . . . . . 76
6.1.5 Magnetotransport Measurements . . . . . . . . . . . . . . . . . . . . . 79
6.2 Patterned Ni Electrodes on Piezoelectric Substrates . . . . . . . . . . . . . . 88
6.2.1 Sample Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
6.2.2 Polarization and Strain Control . . . . . . . . . . . . . . . . . . . . . . 90
6.2.3 Strain Induced Shift of the Coercive Field . . . . . . . . . . . . . . . 94
6.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
7 Conclusion and Outlook 97
Bibliography 101
List of Figures
2.1 Hysteresis of ferromagnets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Coordinate system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3 Multiferroics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.4 Hysteresis of piezoelectric materials . . . . . . . . . . . . . . . . . . . . . . . . 11
2.5 Phenomenology of the magneto-optical Kerr effect . . . . . . . . . . . . . . . 13
2.6 Microscopic origin of the magneto-optical Kerr effect . . . . . . . . . . . . . 14
2.7 Geometry of the Kerr effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.8 Calculation of the MOKE signal . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.1 Cubic and tetragonal unit cells of BaTiO3 . . . . . . . . . . . . . . . . . . . . 23
3.2 Ferroelectric domains in BaTiO3 . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.1 Kerr effect schematic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.2 Photograph of the established Kerr setup . . . . . . . . . . . . . . . . . . . . 29
4.3 Photograph of LED and focusing lens . . . . . . . . . . . . . . . . . . . . . . 30
4.4 Image processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.5 Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.6 Magneto optical contrast . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.7 Contrast with /4 plate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.8 Sensitivity per pixel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.9 Domains Co . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.10 Domain evolution in CrO2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.11 Quantitative Kerr microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
5.1 Ferromagnet on piezoelectric actuator . . . . . . . . . . . . . . . . . . . . . . 46
5.2 Strain induced anisotropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
5.3 Voltage dependent hysteresis of Ni and Fe50Co50 . . . . . . . . . . . . . . . . 50
5.4 Voltage dependent hysteresis of Ni . . . . . . . . . . . . . . . . . . . . . . . . 52
5.5 Domain evolution in Fe50Co50 on piezoelectric actuator . . . . . . . . . . . . 53
5.6 Inhomogenous strain distibution . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.7 Domain evolution in Ni at Vp = −30V . . . . . . . . . . . . . . . . . . . . . . . 55
5.8 Domain evolution in Ni at Vp = +30V . . . . . . . . . . . . . . . . . . . . . . . 56
5.9 Domains in Ni on actuator at Vp = 0V . . . . . . . . . . . . . . . . . . . . . . 57
5.10 Illustration of the actuator deformation . . . . . . . . . . . . . . . . . . . . . 58
V
VI List of Figures
5.11 Reference measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.12 Strain induced birefringence . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.13 Extraction of magnetic contribution to birefringence . . . . . . . . . . . . . . 61
5.14 Magnetization switching in Fe50Co50 cemented onto actuator . . . . . . . . 62
5.15 Magnetization switching in Ni evaporated onto actuator . . . . . . . . . . . 64
5.16 Ni on BaTiO3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.17 Hysteresis curves of Ni on BaTiO3 for different applied voltages . . . . . . . 67
5.18 Domain evolution in Ni on BaTiO3 at E = +400 V/mm . . . . . . . . . . . . 67
5.19 Domain evolution in Ni on BaTiO3 at E = −60 V/mm . . . . . . . . . . . . . 68
5.20 Magnetization switching in Ni on BaTiO3 at 0H = 7.3 mT . . . . . . . . . 69
5.21 Magnetization switching in Ni on BaTiO3 at 0H = 8.0 mT . . . . . . . . . 69
6.1 Sample preparation for local magnetization control . . . . . . . . . . . . . . 72
6.2 Polarisation of single PZT monolayers . . . . . . . . . . . . . . . . . . . . . . 73
6.3 Local strain generation in evaporated samples . . . . . . . . . . . . . . . . . 74
6.4 Local strain generation in cemented samples . . . . . . . . . . . . . . . . . . 75
6.5 Local magnetization control in FeCo . . . . . . . . . . . . . . . . . . . . . . . 76
6.6 Local magnetization switching in Ni . . . . . . . . . . . . . . . . . . . . . . . 77
6.7 Local magnetization switching in Ni for different contact schemes . . . . . . 79
6.8 Sample preparation for magnetotransport measurements . . . . . . . . . . . 80
6.9 Magnetotransport at different applied voltages . . . . . . . . . . . . . . . . . 82
6.10 Local magnetization control and MTR . . . . . . . . . . . . . . . . . . . . . . 84
6.11 Magnetotransport and local magnetization control . . . . . . . . . . . . . . . 85
6.12 Correlation of magnetotransport and MOKE . . . . . . . . . . . . . . . . . . 86
6.13 Ni on PZT plate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
6.14 Ni electrodes on BaTiO3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
6.15 Polarization and strain control in Ni/PZT composites . . . . . . . . . . . . . 92
6.16 Strain control on PZT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
6.17 Strain induced coercivity shift in Ni on PZT . . . . . . . . . . . . . . . . . . 94
List of Tables
2.1 Overview of effects in multiferroic material systems . . . . . . . . . . . . . . 10
3.1 Magnetoeleastic constants of Ni and Fe50Co50 . . . . . . . . . . . . . . . . . . 22
3.2 Piezoelectric coefficients of PZT, BaTiO3 and LiNbO3 . . . . . . . . . . . . 23
5.1 Overview of prepared samples. . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
6.1 Overview of prepared samples for local magnetization control. . . . . . . . . 90
VII
Chapter 1
Introduction
I was led some time ago to think it very likely, that if a beam of plane-
polarized light were reflected under proper conditions from the surface of
intensely magnetized iron, it would have its plane of polarization turned
through a sensible angle in the process of reflection.
(John Kerr, 1877)
This is the very first sentence of John Kerr’s publication on the discovery of the ro-
tation of the plane of polarization by reflection from a magnetic surface [1]. At that
time, John Kerr could not foresee the impact of his findings onto the research regarding
magnetic media. As of today, we know that his discovery paved the way to modern
magneto-optic experiments. The magneto-optical Kerr effect (MOKE), named after his
discoverer, became a powerful tool for probing surfaces of bulk ferromagnets or magnetic
thin films. Reasons for the success of magneto-optical methods are that they allow a fast
and non-destructive investigation of ferromagnetic samples and thus are widely used for
their characterization. Nowadays, Kerr signals can be recorded even on the time scale of
femtoseconds by applying pump-probe experiments and thus give information about mag-
netization dynamics [2, 3]. In addition, the magneto-optical Kerr effect allows the direct
observation of magnetic domains [4]. The introduction of the digital difference technique
in which non-magnetic background is digitally subtracted, together with the CCD camera
revolution represented important stimuli for the magneto-optical Kerr technique, since
from then on processes in the magnetic microstructure could be investigated vigorously
[5, 6, 7, 8]. In comparison to other domain observation methods like magnetic force mi-
croscopy (MFM), scanning electron microscopy (SEM) or tunneling electron microscopy
(TEM), the MOKE technique is the only one that allows direct evaluation of the mag-
netization vector. Moreover, in terms of the necessary financial investment it generally
beats its counterparts as well. As a consequence of these advantages, the magneto-optical
Kerr effect is today’s standard technique in domain imaging applications.
The significance of the magnetic substructure in research and applications has risen
steadily and plays an important role in the development of modern spintronic or mag-
1
2 Chapter 1 Introduction
netoelectronic devices. In contrast to conventional electronic devices, which utilize only
the charge degree of freedom, spintronic devices also take advantage of the electron spin
degree of freedom and thus represent a unification of electric and magnetic properties.
In the history of magnetism and the field of spintronics research and also regarding tech-
nological applications the discovery of the giant magneto resistance (GMR) by A. Fert
and P. Grunberg [9, 10] and of the tunneling magneto resistance (TMR) by M. Julliere
[11] have certainly represented mile-stones. Today’s hard disk drives reading heads and
magnetic sensor technology most often rely on one of those effects.
However, the scalability of magnetoelectronic devices to smaller dimensions turned out
to be the critical endeavor for future relevance in technology. Today’s nanofabrication
techniques allow to structure nanoscale systems, but the critical point is to control the
magnetic microstructure. In most applications, the magnetization of ferromagnets is
controlled by means of magnetic fields. Although magnetic fields are the natural way to
control the magnetization orientation, in small scale systems they need to be of sufficient
strength to achieve the desired magnetization control. Large magnetic fields implicate
stray fields which finally prevent ultimate scaling to state of the art nanodevices. More-
over, power consumption rises with the magnetic field strength.
As a consequence, the search for other possibilities of magnetization control began.
Electric fields are a promising alternative, because they are easy to implement on even
nanometer length scales, power efficient and fully switchable. To allow for magnetization
control, efficient coupling between the electric field and the ferromagnetic order parameter
is essential. Therefore, multifunctional material systems enabling magnetization control
by means of electric fields have been under vigorous research and numerous schemes have
been reported. One promising approach is the usage of intrinsic multiferroic materials
that unite electric and magnetic phases and thus allow a direct magnetization control by
electric fields [12, 13, 14, 15, 16]. In addition, heterostructures in which magnetic and
electric phases share a common interface also yield a promising alternative [17]. Last,
the establishment of multifunctional ferromagnetic and ferroelectric hybrid structures
represent another approach to aim for an electric field control of the magnetization [18,
19, 20, 21, 22]. All the above proposals rely on efficient coupling of the ferrolectric and
ferromagnetic phases, which is generally achieved with the so-called magnetoelectric or
magnetoelastic effects. More important, these approaches in principle facilitate scaling
to smaller dimensions.
From this matter of fact, the central motivation for this work arose. Multiferroic hybrid
structures have been prepared and studied in the framework of this thesis. The investi-
gation has concentrated upon magnetization control by means of electrically-controlled
mechanical strain. Moreover, the scalability and the possibility of local magnetization
control in the prepared samples was analyzed. The main instrument of characterization
was a magneto-optical Kerr setup that allows spatially resolved magnetic domain imag-
ing. As part of this thesis, an existing MOKE setup has been extended and has proven
to be a very valuable method in the observation of magnetic microstructure as already
3
mentioned above.
The outline of this thesis is the following:
Chapter 2 briefly reviews and summarizes the characteristics of ferroic systems, in
particular ferromagnetism and ferroelectricity. Then the concept of multiferroics is in-
troduced and the piezoelectric and magnetoelastic effects are reviewed in the context of
multiferroic systems. Finally the physical principles regarding the magneto-optical Kerr
effect are introduced.
Basic properties of the materials used for this thesis are presented in Chapter 3. Nickel
and iron-cobalt serve as ferromagnetic constituents, whereas lead zirconate titanate, bar-
ium titanate and lithium niobate are reviewed as the piezoelectric parts of the multiferroic
hybrid structures.
In Chapter 4 the magneto-optical Kerr setup established during this thesis is presented.
First, the setup and its components are described before its characterization in terms of
resolution, sensitivity and contrast is presented. Finally imaging experiments on thin
ferromagnetic films are shown to demonstrate the capabilities of the new low resolution
Kerr microscope.
Experiments on ferromagnetic/ferroelectric hybrid structures are presented in Chap-
ter 5. Hybrid structures consisting of piezoelectric actuators and ferromagnetic thin films
as well as composites made of barium titanate and nickel thin films are discussed. Both
types of samples are studied regarding their capabilities in magnetoelastic coupling and
magnetization control by means of electric fields.
Chapter 6 deals with the scalability of hybrid structures allowing electric field controlled
magnetization switching. The possibility of local magnetization control in hybrids based
on piezoelectric actuators is discussed. Moreover a novel approach for local magnetization
control is presented, in which micron size patterned ferromagnetic electrodes are strained
by piezoelectric substrates.
Finally, in Chapter 7 a brief conclusion is given and ideas for future applications and
experiments are proposed.
Chapter 2
Theoretical Basics
In this chapter the theoretical background, that is of relevance for this work, is re-
viewed. At first, basics of ferromagnetism and ferroelectricity as ferroic systems are
discussed. Then, characteristics of systems uniting ferromagnetic and ferroelectric prop-
erties, so-called multiferroics, are introduced. At last, the physical principles underlying
the magneto-optic sample characterization on the basis of the Kerr effect are reviewed.
2.1 Ferroics
Ferroics is the generic name given to the study of ferromagnets, ferroelectrics and ferro-
elastics. All ferroic materials have in common that they exhibit spontaneous long range
order characterized by an order parameter X ∈ M,P,", where M is the magnetization
in ferromagnets, P the polarization in ferroelectrics and " the strain in ferroelastics. It is
important to note, that this ordering is spontaneous which means that it occurs in absence
of external fields. Moreover, ferroics undergo a phase transition from a state with no order
(X = 0) above a critical temperature Tc to a state with long range order (X ≠ 0) below
Tc. The critical temperature Tc is referred to as Curie temperature. The application
of external fields Y ∈ H,E,1 results in a response of the order parameter. In linear
approximation the response is governed by X = Y, where denotes the susceptibility
tensor.
For the scope of this thesis, especially ferromagnetism and ferroelectricity are of im-
portance. Therefore a short review of both is given in the following.
2.1.1 Ferromagnetism
The order parameter in ferromagnets is the magnetization M, which is defined as the
quantity of magnetic moment per unit volume V [23]
M =1
V∑
j
j (2.1)
1H denotes a magnetic field, E an electric field and mechanical stress
5
6 Chapter 2 Theoretical Basics
The microscopic origin of the magnetic moment is the coupling of the classical orbital
angular momentum L of an electron with its spin S. In ferromagnets, magnetic moments
strongly interact with each other. This interaction is caused by the quantum mechanical
exchange mechanism, which is primarily responsible for the ordering of magnetic moments
in ferromagnets.
As mentioned above, the magnetization responds to external magnetic fields as de-
scribed in linear approximation by [24]
M = MH, (2.2)
where M denotes the magnetic susceptibility. In this work, magnetic fields will be
described by the magnetic induction B, which is related to M by
B = 0(H +M), (2.3)
where 0 = 4 × 10−7 VsAm denotes the vacuum-permeability.
Apart from those basic principles, ferromagnetism involves many fundamental con-
cepts. For the framework of this thesis, the concept of magnetic hysteresis, domain
formation and magnetic anisotropy are of importance and thus they are summarized
briefly.
Domains
A magnetic domain describes a region within a ferromagnet which has a uniform mag-
netization [25]. In absence of external fields, a ferromagnet is generally composed of
numerous domains with different magnetization orientations. Domains with different
magnetization orientations are separated by domain walls. The formation of magnetic
domains is energetically favorable compared to a single domain state. In the resulting
domain pattern the magneto-static stray field energy is balanced against the energy of
the domain walls, anisotropy and the term arising from the exchange mechanism [26].
In bulk ferromagnets, generally two types of domain walls exist: Bloch and Neel walls
[27]. The difference between both wall types is the mode of rotation of the magnetization
vector within the domain wall. If the rotation of the magnetization vector is within the
domain wall plane (perpendicular to the domain wall plane), the domain wall is referred
to as a Neel wall (Bloch wall).
In thin ferromagnetic films many different domain configurations and domain wall
types are present. An extensive study can be found in [5]. Generally, the domain wall
structure is strongly influenced by the film thickness. In particular, for thin films with
in plane anisotropy Neel type walls are preferred, since the magnetization rotates within
the domain plane and thus minimizes the demagnetization energy term (see below).
2.1 Ferroics 7
H
M
Mr
Hc
Ms
Figure 2.1: Illustration of a ferromagnetic hysteresis curve.
Hysteresis
One general feature of ferromagnetic substances is that the magnetization responds hys-
teretic upon the application of external magnetic fields [23] as depicted in Fig. 2.1. In the
demagnetized state at zero magnetic field the ferromagnet consist of domains and exhibits
no net magnetization. Increasing the field, domains which are energetically favored grow
at the expense of unfavorable domains. For large external magnetic fields (H >>Hc), the
magnetization is saturated and the ferromagnet is in a single domain state. Decreasing
the external field to H = 0 the macroscopic magnetization is reduced to the remanent
magnetization Mr. This behavior is due to the formation of domains in order to minimize
the free energy of the ferromagnet. At the coercive field H = −Hc, the macroscopic mag-
netization has diminished to zero as a consequence of domain formation. The individual
domains still exhibit a magnetization, however the average over all domains results in
M = 0. A further decrease of the magnetic field yields a uniform magnetic state again.
Magnetic Anisotropies
The direction of the external magnetic field H and the magnetization M do not necessarily
coincide in a ferromaget. The reason therefore are anisotropies, which can be induced
by intrinsic properties of the material, like shape or crystal structure, or by extrinsic
contributions like stresses. In order to determine the orientation of the magnetization
a free energy model was introduced [28, 29], which accounts for all intrinsic anisotropy
contributions in the following form:
Ftot = Fstat + Fdemag + Faniso (2.4)
In this approach Fstat denotes the Zeeman energy, Fdemag stands for the demagnetization
term and Faniso represents crystalline anisotropies. In equilibrium, the magnetization
always resides in local minimum of the free energy density. The minima (maxima) in the
8 Chapter 2 Theoretical Basics
z
y
H
M
x
Figure 2.2: Coordinate system relative to the sample [30].
free energy correspond to magnetically easy (hard) directions.
A coordinate system has been introduced [30] to describe the orientation of both M
and H with respect to the sample (cf. Fig. 2.2). Thus the following parametrization in
polar coordinates is achieved:
M = (M,Θ,Φ) (2.5)
H = (H,, ) (2.6)
Capital Greek letters describe the orientation of the magnetization and lower case
letters the orientation of the external magnetic field. In terms of this parametrization
the contributions to the free energy density can be expressed as follows:
• The Zeeman term Fstat describes the magnetostatic energy of the magnetization M
in an external magnetic field H. It is given by [31]
Fstat = −0H ⋅M. (2.7)
In terms of the parametrization above it becomes
Fstat = −0HM(sin Θ sin Φ sin sin + cos Θ cos + sin Θ cos Φ sin cos) (2.8)
Without consideration of further contributions to the free energy, the magnetization
would always be aligned parallel to the external magnetic field.
• The demagnetization term Fdemag takes the anisotropy arising from the sample
shape and geometry into account and is therefore often denoted as shape anisotropy.
For thin films, which is appropriate for all samples used in this thesis, it yields [32]:
2.2 Multiferroics 9
Fdemag =0
2M2 sin2 Θ cos2 Φ (2.9)
Usually, the shape anisotropy causes the magnetization to lie within the film plane
of the ferromagnetic sample since a magnetically hard axis perpendicular to the
film surface is induced.
• The crystalline anisotropy Faniso arises from the crystal structure of the ferromag-
netic sample. However, in this thesis only polycrystalline thin films have been
prepared. In this case, no net crystalline anisotropies are present [19] and thus this
energy term can be omitted.
All in all, in polycrystalline thin ferromagnetic films two intrinsic contributions to the
free energy are present, the Zeeman energy and the demagnetization term. Consequently,
considering ideal growth conditions the ferromagnetic thin film is fully isotropic in the
film plane but exhibits a magnetically hard axis out-of-plane.
2.1.2 Ferroelectricity
In ferroelectric materials the order parameter is the polarization P. The response of the
polarization upon the application of an electric Field is governed by
P = 0EE, (2.10)
where 0 = 8.85 × 10−12 AsVm denotes the permittivity of free space and E the electri-
cal susceptibility. Ferroelectric materials exhibit a spontaneous polarization below the
Curie temperature. In accordance to ferromagnets, their polarization can be reversed
by external electric fields and it also shows a hysteretic behavior. Another similarity
to ferromagnetism is the fact that domain formation also plays an important role in
ferroelectricity (cf. Chapter 3).
2.2 Multiferroics
So far, ferroelectrics and ferromagnets have been introduced as examples of ferroic ma-
terials. They both have in common that they exhibit spontaneous long range order.
Materials that unite more than one ferroic order parameter simultaneously are called
multiferroics. In Fig. 2.3 the relationship between the individual order parameters M, P
and " and the corresponding fields H, E and is represented by different colors. How-
ever, in multiferroic materials there are also cross-links, since each field may influence all
order parameters and the resulting effects are listed in Tab. 2.1.
For the framework of this thesis, the magnetoelastic and the piezoelectric effect are of
importance and therefore their main properties are reviewed briefly.
10 Chapter 2 Theoretical Basics
E
P
M
H
N S
M
P+ – + –
+ – + –
ε
ε
σ
Figure 2.3: Relation between the different ferroic order parameters and the external fields in multiferroic
systems [12].
Y→X EffectH→ P magnetoelectric effectE→M converse magnetoelectric effectH→ " magnetostrictive effect →M magnetoelastic effect → P piezoelectric effectE→ " inverse piezoelectric effect
Table 2.1: Overview of effects in multiferroic material systems. The external field Y induces a response
of the order parameter X in multiferroics.
2.2.1 Piezoelectric Effect
Piezoelectricity is the ability of some materials to generate an electric field or electric
potential in response to an applied mechanical stress [33]. Materials exhibiting the direct
piezoelectric effect also show the inverse piezoelectric effect, which describes the defor-
mation of the piezoelectric material upon the application of external electric fields. It is
important to note, that all ferroelectric materials show piezoelectricity, however, there
are piezoelectric materials, like e.g. quartz, which do not exhibit ferroelectricity. For
the scope of this thesis, especially the inverse piezoelectric effect is of importance. The
resulting elastic strain " can be correlated with the applied electric field E by [33, 34]
" = dpE − sE, (2.11)
where sE stands for the coefficient of elasticity (inverse of the Young’s modulus), de-
notes the mechanical stress and dp is the piezoelectric strain coefficient. The piezoelectric
effect is fundamental for the functioning of piezoelectric actuators and other piezoelec-
tric materials, which will be introduced in Chapter 3. In essence, one has to distinguish
between the longitudinal (d33-effect) and the transversal (d31-effect) mode of the piezo-
2.2 Multiferroics 11
electric substrate. The longitudinal mode describes the deformation of the piezoelectric
material parallel to the external field, whereas the transversal mode describes the defor-
mation perpendicular to it. The two effects can be expressed as [35]:
"3 = d33E3 − sE333 (2.12)
"⊥ = d31E3 − sE11⊥. (2.13)
As piezoelectric materials are generally also ferroelectric, the strain response to the
electric field is non-linear but hysteretic [36, 37]. The resulting hysteresis curve for the
mechanical strain is schematically depicted in Fig. 2.4. Due to its shape the hysteresis
curve is also denoted as butterfly curve. The origin of the hysteresis is the inverse piezo-
electric effect together with the switching and movement of ferroelectric domain walls
[36]. The minima in the hysteresis curve correspond to the ferroelectric coercive field
indicating the polarization reversal [36]. As a consequence of the hysteretic behavior, the
resulting strain at a certain electric field is not uniquely defined but depends upon the
field history. Moreover, also at vanishing electric field piezoelectric materials generally
exhibit remanent strain. For the matter of application, piezoelectric materials are usu-
ally used in a semi-bipolar field regime which comprises only the positive or the negative
hysteresis branch.
ε
E
Figure 2.4: Strain-field hysteresis of piezoelectric materials.
2.2.2 Magnetoelastic Effect
The magnetoelastic effect, also known as converse magnetostrictive effect, describes the
effect of stress onto the magnetization of a ferromagnet [31]. The applied stress induces
strains within the ferromagnet. In presence of strains, the free energy approach introduced
12 Chapter 2 Theoretical Basics
in Eq. (2.4) has to include the energy contribution Fmagel arising from magnetoelastic
coupling. Consequently the free energy Ftot of a polycrystalline ferromagnet under applied
stress is given by
Ftot = Fstat + Fdemag + Fmagel. (2.14)
Considering only strains along the x, y and z directions, shear strains can be neglected
in polycrystalline samples, since they average out. Thus the resulting strain is purely
uniaxial and is manifested in the magnetoelastic contribution Fmagel as [31, 34]
Fmagel =Kmagel,x sin2 Θ sin2 Φ +Kmagel,y cos2 Θ +Kmagel,z sin2 Θ cos2 Φ (2.15)
with
Kmagel,i =3
2(c12 − c11)"i, i ∈ x,y, z . (2.16)
In this equation denotes the magnetostrictive constant, c11, c12 represent the elastic
moduli of the ferromagnet and "i stands for the strains along the corresponding directions
in Voigt notation [38].
Strains can arise for instance from the lattice mismatch between substrate and thin film.
However, in this thesis strain will be induced on purpose with the help of piezoelectric
substrates. Thus, considering the principles presented above it has to be emphasized
that the magnetoelastic effect in principle opens the avenue for strain control of the
magnetization orientation.
2.3 Magneto-Optical Kerr Effect
The interaction of light with solids depends primarily on the electronic structure of the
solids. Magneto-optical effects come into play if the interaction additionally depends on
the magnetic state of the material. The most prominent magneto-optical effects are the
Faraday and the Kerr effect. Generally they differ in the manner of the observation. If
the magneto-optical effect is observed in transmission, it is denoted as Faraday effect.
Contrarily, the Kerr effect describes magneto-optical effects observed in reflection. In the
framework of this thesis a magneto-optical Kerr setup was extended and then used for
characterization of ferromagnetic samples. Therefore a short review of phenomenology,
origin and quantitative description of the magneto-optical Kerr effect (MOKE) will be
given in the following.
2.3.1 Phenomenology
The magneto-optical Kerr effect is based on the interaction of polarized light with a
magnetized sample and was first observed in 1877 [1]. The phenomenology of the Kerr
2.3 Magneto-Optical Kerr Effect 13
fk
Mtan(hk)=b/a
sample
EiErincident
lightreflectedlightb
aθ1
Figure 2.5: Phenomenology of the magneto-optical Kerr effect. Upon reflection off a magnetized sample
the polarization plane of the incident light (angle of incidence 1) is rotated by an angle 'k.
The state of polarization changes from linearly polarized to elliptically polarized, which is
characterized by the ellipticity k.
effect is pointed out with the help of Fig. 2.5. If you regard a linearly polarized light
beam, it undergoes two transformations upon the reflection off a magnetized sample.
First, its plane of polarization is rotated by the so-called Kerr angle 'k. In addition,
the state of polarization changes from linearly polarized to elliptically polarized. The
new state of polarization can be characterized by the ellipticity k which is defined as
the inverse tangent of the ratio of the principle ellipse axes tan k = b/a. All in all, both
transformations are combined to the complex Kerr angle Ψk with [39]
Ψk = 'k + ik (2.17)
The complex Kerr rotation depends upon the magnetization of the sample as will be
demonstrated in following. Therefore the magneto-optical effect constitutes a perfect
measure of the magnetization and is widely used for characterization of ferromagnetic
samples and domain observation.
2.3.2 Microscopic Origin
In order to understand the microscopic origin of the magneto-optical Kerr effect, quantum
mechanics has to be applied. In the following a simple explanation for the interaction
of light with a 3d ferromagnet based on quantum mechanical transition rules is given
[40]. An extensive treatment of this problem can be found in [41, 42]. Let us consider
optical transitions between a doubly degenerate dxz,yz level (l = 2, ml = ±1) and a pz level
above the Fermi energy. Due to the exchange mechanism in ferromagnets, the d levels for
spin-up and spin-down electrons are separated by the exchange energy EEx (cf. Fig. 2.6).
Moreover, the spin-orbit coupling lifts the degeneracy of the dxz,yz level for both spin
directions. They split into a d(x+iy)z level with ml = +1 and a d(x−iy)z level with ml = −1.
Optical transitions have to satisfy the selection rules for electric dipolar transitions. For
14 Chapter 2 Theoretical Basics
d↑xz d↑
yzd↑
(x+iy)z
d↑(x-iy)z
p↑z
d↓yzd↓
xz
d↓(x-iy)z
d↓(x+iy)z
p↓z
++--
-
S.O.
S.O.
E
absorption spectrum +
↓
↓↑
↑
ESO
ESO
EEx
EF
Figure 2.6: Microscopic origin of the magneto-optical Kerr effect in ferromagnets. The d energy levels
split due to exchange mechanism (EEx) and the spin-orbit coupling (ESO). On the left, the
optical dipole transitions for right- (+) and left- (-) circularly polarized light are depicted.
The corresponding absorption spectra versus photon energy are sketched on the right.
circularly polarized light they are given by [43]
Δl = ±1 Δm = ±1. (2.18)
The resulting transitions and the absorption spectra are depicted in Fig. 2.6. For
right-circularly (+) polarized light transitions from d(x−iy)z to pz are allowed, whereas left-
circularly (-) polarized light can excite electrons from d(x+iy)z to pz. During the transition,
the spin is conserved. Evidently, the absorption spectra is different for left- and right-
circularly polarized light. Thus, linearly polarized light, which is a superposition of
(+) and (-) circularly polarized light, is transformed into elliptically polarized light. In
addition, it can be shown that the different absorption behavior for (+) and (-) circularly
polarized light causes a phase shift between the two components [40]. This phase shift is
the origin of the Kerr rotation.
Of course, the absorption spectra drawn in Fig. 2.6 are exaggerated and in reality the
absorption curves for (+) and (-) are broader and they overlap. However, with the help
of this simple picture it is pointed out that the Kerr effect in ferromagnets arise from the
simultaneous occurrence of exchange splitting and spin-orbit coupling.
2.3.3 Macroscopic Description
Macroscopically the interaction of light with solids is determined by the complex index
of refraction n = nR+ inI. The index of refraction is connected with the dielectric tensor "
2.3 Magneto-Optical Kerr Effect 15
and the permeability tensor by n2 = " ⋅ . In the optical regime, the permeability tensor
can be set equal to one [44] and thus the index of refraction yields:
n2 = " (2.19)
For isotropic media and in absence of external fields, the diagonal elements of " are
identical and the off-diagonal elements are zero [44]. If the media is magnetized, how-
ever, the dielectric tensor yields asymmetric off-diagonal elements. Classically, " can be
calculated assuming a Lorentz-oscillator model [39, 45]. For a cubic material neglecting
higher order terms it is given by [5, 39]
" = "
⎛⎜⎜⎝
1 −iQmz iQmy
iQmz 1 −iQmx
−iQmy iQmx 1
⎞⎟⎟⎠
, mi =Mi
∣M∣, i ∈ x,y, z (2.20)
where mi denotes the projections of the components of M along the x, y and z direction
(direction cosines). " is the dielectric constant in absence of a magnetization. The
material constant Q is the so-called Voigt or magneto-optical constant.
Inserting " into Eq. (2.19) yields the index of refraction. The calculation involves
Maxwell’s equations with proper boundary conditions and can be found in [42, 46]. Two
solutions, n(+) and n(−), are obtained for right- and left-circularly polarized light, respec-
tively. In linear approximation they are given by [47]
n(±) ≈√"(1 ±
1
2Qk ⋅m) , (2.21)
where k is the unit vector along the propagation direction of the dielectric displacement
vector D. Exhibiting two different indices of refraction, the material is birefringent in
presence of a magnetization.
From this matter of fact, two consequences can be deduced. First, birefringence usually
results in different propagation velocities for (+) and (-) circularly polarized light [44] in
the medium. This effect is called linear magnetic birefringence and it causes the charac-
teristic rotation of the polarization plane of the incident light [47]. Second, the indices of
refraction are complex (Q is generally complex) and thus the media absorbs electromag-
netic radiation. However, usually the imaginary parts of n(+) and n(−) differ [47] which
results in a different absorption behavior for (+) and (-) polarized light. Hence, linearly
polarized light is transformed into an elliptical state of polarization. This transformation
is denoted as magnetic dichroism.
As the magneto-optical Kerr effect arises from the interaction of light with a magnetic
material, its information depth is restricted by the penetration depth of light in the probed
medium. In metals, information depths of about 20 nm have been reported [48, 49]. As a
consequence, magnetic properties of surfaces of bulk samples as well as of thin films can
be investigated by the Kerr effect.
16 Chapter 2 Theoretical Basics
M M
M
polar longitudinal transversal
(a) (b) (c)
Figure 2.7: Illustration of the Kerr effect geometries with respect to the plane of incidence. (a) The
magnetization is oriented perpendicular to the sample surface which is referred to as the
polar Kerr effect. (b) The magnetization is parallel to the the sample surface and parallel to
the plane of incidence causing the longitudinal Kerr effect. (c) In the transversal geometry
the magnetization is within the film plane but perpendicular to the plane of incidence.
2.3.4 Quantitative Formulation
So far, qualitative aspects of the Kerr effect have been discussed. This section is dedicated
to the presentation of quantitative expressions for the complex Kerr rotation Ψk. At
first, one distinguishes between three different effects depending on the orientation of
the magnetization with respect to the sample surface and the plane of incidence. The
resulting geometries are depicted in Fig. 2.7. The polar effect is characterized by an
out-of-plane magnetization direction (cf. Fig. 2.7(a)). In the longitudinal geometry the
magnetization is oriented within the film plane and parallel to the plane of incidence as
displayed in Fig. 2.7(b). In contrast, if the magnetization is oriented perpendicular to
the plane of incidence in Fig. 2.7(c), the transversal Kerr effect occurs.
For the following analysis, the electric field of the incident and reflected light is split
into two components E = Ess+Epp. Es denotes the component perpendicular to the plane
of incidence (s-polarized), Ep denotes the component parallel to the plane of incidence
(p-polarized). The electric field after the reflection off the sample surface is denoted as
E′. Reflection of electromagnetic waves at surfaces and interfaces can be described by the
Fresnel coefficients, which are denoted rss, rsp, rps, rpp [44]. Therefore E′ can be calculated
with the help of the reflection matrix R as
(E′
s
E′
p
) = (rss rsp
rps rpp)(
Es
Ep) . (2.22)
The complex Kerr rotation is determined by the amplitude ratio Ψk = E′
p/E′
s [50, 51].
For s-polarized light (Ep = 0) it simplifies to Ψsk = rps/rss, for p-polarized light (Es = 0) the
complex Kerr rotation yields Ψpk = rsp/rpp assuming small angles in both cases [42, 47].
The Fresnel coefficients occurring in this scattering matrix are tabulated [42, 52, 53].
However, for an individual analysis of the different Kerr effect geometries, it is more
convenient to express R in terms of rp, rl, rz, which denote the scattering matrices for the
2.3 Magneto-Optical Kerr Effect 17
polar, longitudinal and transversal geometry. Thus the reflection matrix can be rewritten
as [52, 53]
R =m2pr
p [Q/mp] +m2l r
l [Q/ml] +m2t t
t [Q/mt] . (2.23)
The square brackets imply that Q has to be replaced by Q/mi with (i ∈ p, l, t)
in the individual scattering matrices. mp, ml, mt again denote the projections of the
magnetization onto the corresponding axes and they yield m2p +m
2l +m
2t = 1.
For the polar Kerr effect the components of rp are given by [52, 53, 54]
rpss =
n1 cos 1 − n2 cos 2
n1 cos 1 + n2 cos 2
(2.24)
rppp =
n2 cos 1 − n1 cos 2
n2 cos 1 + n1 cos 2
(2.25)
rpsp = r
pps =
in1n2Q cos 1
(n1 cos 1 + n2 cos 2)(n2 cos 1 + n1 cos 2), (2.26)
where n1 and n2 are the complex indices of refraction (for the magnetic material the
index of refraction is approximated by n2 ≈√" [42]) and 1, 2 denote the incident and
refracted angles according to the law of Snellius. Hence, the Kerr rotation arising from
the polar magnetization component can be calculated for s- and p-polarized light as
Ψs,polark =
rpps
rpss=
in1n2Q cos 1
(n2 cos 1 + n1 cos 2)(n1 cos 1 − n2 cos 2)(2.27)
Ψp,polark =
rpsp
rppp
=in1n2Q cos 1
(n1 cos 1 + n2 cos 2)(n2 cos 1 − n1 cos 2). (2.28)
Consequently, the largest Kerr rotation in both cases is obtained for 1 = 0°, which
corresponds to perpendicular light incidence.
For the longitudinal Kerr effect, the Fresnel coefficients of rl are tabulated as [50, 52, 53]
rlss =
n1 cos 1 − n2 cos 2
n1 cos 1 + n2 cos 2
(2.29)
rlpp =
n2 cos 1 − n1 cos 2
n2 cos 1 + n1 cos 2
(2.30)
rlsp = −r
lps =
in1n2Q cos 1 tan 2
(n1 cos 1 + n2 cos 2)(n2 cos 1 + n1 cos 2), (2.31)
and the Kerr rotation follows as
18 Chapter 2 Theoretical Basics
Ψs,longk =
rlps
rlss
= −in1n2Q cos 1 tan 2
(n2 cos 1 + n1 cos 2)(n1 cos 1 − n2 cos 2)(2.32)
Ψp,longk =
rlsp
rlss
=in1n2Q cos 1 tan 2
(n1 cos 1 + n2 cos 2)(n2 cos 1 − n1 cos 2). (2.33)
In contrast to the polar effect, the angle 2 appears in the denominator of the Kerr
rotation in the longitudinal geometry. As a consequence, at perpendicular light incidence
(1 = 2 = 0°), the Kerr rotation vanishes. The maximum Kerr rotation is obtained at an
angle of incidence 0° < 1° < 90°, depending on the angle of refraction 2. Comparing the
Kerr rotation of longitudinal and polar effect, the longitudinal Kerr rotation is smaller
by an order of magnitude than the polar one because of the additional tan 2 term in the
denominator.
At last, the Fresnel coefficients of the scattering matrix for the transversal geometry rt
are [52, 53, 54]
rtss =
n1 cos 1 − n2 cos 2
n1 cos 1 + n2 cos 2
(2.34)
rtpp =
n2 cos 1 − n1 cos 2
n2 cos 1 + n1 cos 2
+ i2n1n2Q cos 1 sin 2
(n2 cos 1 + n1 cos 2)2 (2.35)
rtsp = r
tps = 0. (2.36)
The off-diagonal elements of the scattering matrix for the transversal geometry are zero
and thus the transversal magnetization component does not cause a Kerr rotation of the
incident light. However, the amplitude of the p-polarized light upon reflection depends
on Q. Therefore the intensity change of p-polarized light in the transversal geometry is
a direct measure for the Voigt constant Q.
2.3.5 Detection
Having introduced the scattering matrices in the last paragraph, this section finally deals
with the detection of the magneto-optical Kerr effect. As already mentioned, the Kerr
effect induces a rotation of the polarization plane of the incident light and it also trans-
forms linearly polarized light into elliptically polarized light. In order to measure this
effect, one has to introduce simple polarization optics as depicted in Fig. 2.8. A polarizer
is placed into the illumination path before the sample to polarize the light linearly along
its transmission axes. The orientation of the polarizer is defined by the angle 'p with
respect to the plane of incidence. The reflected light is probed by an analyzer with the
orientation of the transmission axis defined by 'a. This setup allows to measure the real
Kerr rotation 'k only, the ellipticity k is not considered. Inserting a quarter-wave plate
2.3 Magneto-Optical Kerr Effect 19
Ml
Mt
s
p
yayp
polarizer analyzer
sample
Figure 2.8: Basic measurement principle and setup of the polarization components for MOKE. The
angles 'p and 'a define the transmission axis of the polarizer and analyzer with respect to
the plane of incidence.
results in a removal of the ellipticity and will be discussed in Chapter 4. The electric field
after transmission through the polarizer and incident upon the ferromagnetic film can be
expressed as
E = E0 sin'ps +E0 cos'pp. (2.37)
The interaction of the electric field with the ferromagnet is governed by the scattering
matrix introduced in Eq. (2.23). Assuming that the magnetization has only components
within the film plane Ml and Mt (cf. Section 2.1.1), the scattering matrix simplifies to
[55]
R =m2l r
l [Q/ml] +m2pt
t [Q/mt] . (2.38)
After the reflection off the sample surface the components of the electric field E′ are
given by [45, 55]
E′
s =m2l r
lspE0 cos'p + r
lssE0 sin'p (2.39)
E′
p = (m2tr
tpp +m
2l r
lpp)E0 cos'p +m
2l r
lpsE0 sin'p.
The reflected light travels to the analyzer where only the component parallel to the
analyzer transmission axis will pass. The electric field component E′′ along the analyzer
transmission axis can be calculated as
E′′ = E′
s sin'a +E′
p cos'a. (2.40)
Inserting the expressions obtained in Eq. (2.39), E′′ is finally determined as [45, 55]
20 Chapter 2 Theoretical Basics
E′′ = E0 m2l r
lps sin('p − 'a) + r
lss sin'p sin'a + (m2
trtpp +m
2l r
lpp) cos'p cos'a . (2.41)
The final intensity is the absolute value of the electric field calculated in Eq. (2.41) and
it is proportional to the measured signal at the detector. Analyzing the above expression,
it contains terms with the longitudinal magnetization component as well as with the
transversal magnetization component. However, if the orientation of the polarizer is
chosen such that it transmits s-polarized light only ('p = 90°), terms containing mt drop
out and the final intensity just depends on ml [45, 55]:
I/I0 = ∣rlss∣
2sin2'a + ∣m2
l rlps∣
2cos2'a + [rl
ssm2l r
l∗ps + r
l∗ssm
2l r
lps] sin'a cos'a. (2.42)
The ∗ in Eq. (2.42) denotes the complex conjugate.
In summary, the main result of the analysis above is that the resulting signal de-
pends only upon the longitudinal magnetization component if the polarizer angle is set
to 'p = 90°. Thus, using s-polarized incident light enables the measurement of the lon-
gitudinal magnetization component. In the experiments performed during this thesis
s-polarized incident light has been used exclusively.
Chapter 3
Materials
In the previous chapter the physical principles of ferromagnetism, ferroelectricity and
magneto-optic effects have been introduced. This chapter deals with the most relevant
properties concerning the materials that have been used in the framework of this thesis.
Firstly, nickel and iron cobalt will be introduced as the ferromagnetic constituents before
the basic properties of the piezoelectric materials lead zirconate titanate, barium titanate
and lithium niobate will be summarized.
3.1 Ferromagnetic Materials
The performance of the prepared hybrid structures is greatly influenced by the choice
of the ferromagnetic material. As the interplay between the ferroelectric phase and the
ferromagnetic phase is governed by the magnetoelastic coupling coefficient of the ferro-
magnet, this parameter is of great importance for successful magnetization control by
means of electric fields. The 3d-transition element nickel (Ni) and the binary alloy iron-
cobalt (Fe50Co50) have been chosen as the ferromagnetic constituents for the prepared
hybrid structures. In the following their most relevant physical properties, saturation
magnetization Ms, magnetostrictive constants 100, 110, and elastic moduli c11, c12
will be given. The magnetostrictive constants ijk describe the maximal elongation or
contraction = 23l/l of the ferromagnet in the direction [ijk] of the external magnetic
field upon magnetization from the demagnetized state to saturation [31] for cubic ma-
terials. For this thesis only polycrystalline ferromagnetic films have been prepared. In
this case, the magnetostrictive constant can be averaged as = 25100 +
35110 [31]. The
elastic moduli represent the stress per strain ratio cij =i"j
of the corresponding material
according to Hooke’s law [38].
3.1.1 Nickel
Nickel is a very common ferromagnet that most people have encountered in form of coins
in everyday life. It has a fairly high Curie temperature of TNi = 627 K [56] and crys-
tallizes in the face-centered-cubic (fcc) structure with room temperature lattice constant
21
22 Chapter 3 Materials
Material 100 × 106 110 × 106 × 106 c11 − c12 (1011 N/m2)
Ni −45.9 −24.3 −32.9 0.9Fe50Co50 119.3 41.3 72.5 1.2
Table 3.1: Magnetoelastic constants for bulk Ni taken from [59] and Fe50Co50 [31, 60] at room temper-
ature.
aNi = 3.524 A [57]. The bulk saturation magnetization of Ni is Ms = 411 kA/m [58]. Its
bulk magnetoelastic constants can be found in Tab. 3.1.
3.1.2 Iron-Cobalt
The equiatomic iron-cobalt alloy (Fe50Co50), also known as permendur, has been used as
ferromagnetic constituent apart from nickel. At room temperature, it crystallizes in the
body-centered-cubic (bcc) structure [61] with lattice parameter aFeCo = 2.849 A [62]. Due
to its large saturation magnetization of Ms = 1950 kA/m [26], permendur is often used
in transformers or generators when a high flux density is needed. Its Curie point is at
TNi ≈ 1200 K [63]. The magnetoelastic constants are given in Tab. 3.1.
3.2 Piezoelectric Materials
Apart from the ferromagnetic material, the piezoelectric constituent also plays a decisive
role for successful magnetization control in multifunctional hybrid structures. Therefore
three different possible candidates have been used in this thesis: lead zirconate titanate,
barium titanate, lithium niobate. Especially their converse piezoelectric response upon
the application of electric fields is of relevance for this work. In the following the used
materials will be introduced briefly.
3.2.1 Lead Zirconate Titanate
Lead zirconate titanate (Pb [ZrxTi1−x]O3, 0 < x < 1), abbreviated as PZT, is a ferro-
electric material which is commonly used in electroceramics or piezoelectric devices. For
x < 0.5, the ferroelectric phase crystallizes in the perovskite structure with a tetragonal
unit cell in analogy to barium titanate (cf. Fig. 3.1). For smaller portions of titanium
the rhombohedral unit cell predominates [64]. The Curie temperature ranges within
Tc(x = 0) = 500 K and Tc(x = 1) = 250 K also depending on the portion of titanium. The
ferroelectricity below the Curie point arises from the spontaneous displacement of the
Ti4+/Zr4+-ion out of the symmetry center of the unit cell.
In the framework of this thesis, two types of commercially available piezoelectric ac-
tuators made from PZT ceramics have been used. First, PSt 150/2×3/5 actuators [65]
3.2 Piezoelectric Materials 23
Material d31 (10−12 C/N) d33 (10−12 C/N)
PZT (PSt 150/2×3/5) −290 640PZT (PZT-5H) −274 593BaTiO3 −33.4 90LiNbO3 −0.9 6.0
Table 3.2: Piezoelectric coefficients of PZT ceramics (in PSt 150/2×3/5 stacked actuators [65] and PZT-
5H piezoelectric plates [66]), single crystal BaTiO3 [67] and single crystal LiNbO3 [68] at
room temperature.
T ≥ Tc T1 < T < Tc
Ba2+
O2-
Ti4+
Ps ≠ 0Ps = 0
cubic tetragonal
6 possibledirections of Ti4+
displacement
a a
a c
Figure 3.1: Illustration of the cubic and tetragonal unit cell of BaTiO3. The spontaneous polarization
is caused by the displacement of the Ti4+ ion [34].
built from stacks of PZT have been applied (cf. Chapter 5). In addition, PZT-5H piezo-
electric plates [66] have been used (cf. Chapter 6). The piezoelectric properties of the
ceramics employed in these particular actuators have been summarized in Tab. 3.2. In
the conventional mode of operation, both types of piezoelectric actuators rely on the d33-
and the d31-effect (cf. Chapter 2).
3.2.2 Barium Titanate
Barium titanate (BaTiO3) is a ferroelectric ceramic which is often used in dielectric
capacitors because of its relatively large relative permittivity of "/"0 ≈ 4100 [69] at room
temperature. It exhibits several crystalline phase transitions at distinct temperatures
[70]. Above the Curie temperature of Tc = 393 K [71] BaTiO3 is in the paraelectric
state with a cubic crystalline structure. Upon cooling below the Curie point it becomes
ferroelectric. At room temperature, the underlying crystal structure is tetragonal with
the corresponding lattice parameters a = b = 0.3993 nm and c = 0.4034 nm [72]. BaTiO3
24 Chapter 3 Materials
E=0
E>Ec
90° domain walls 180° domain wall
P
P=0c
c
a
(a)
(b)
Figure 3.2: (a) Schematic illustration of the multi-domain state in BaTiO3 in absence of an electric field.
The a-domains are tilted with respect to the orientation of the c-domains. (b) Application
of an electric field results in the formation of a single domain state [34, 73].
crystallizes in the perovskite structure with a unit cell as depicted in Fig. 3.1 for the
cubic and tetragonal phase. The ferroelectricity in the tetragonal phase results from the
displacement of the Ti4+-ion from its central position. BaTiO3 exhibits further phase
transitions at lower temperatures, but those are omitted since they are not of relevance
for the experiments in this thesis, which were all done at room temperature. Crystalline
BaTiO3 is chosen as substrate material because it has a relatively large piezoelectric
coefficient as given in Tab. 3.2.
The substructure of BaTiO3 in the tetragonal phase is dominated by ferroelectric do-
mains. The displacement of the Ti4+-ion out of the symmetry center can be along six
directions. As always two of these directions are equivalent by crystal symmetry, this
behavior gives rise to the formation of three sorts of domains at room temperature:
• c-domains with polar axis along (001)
• a1-domains with polar axis along (100)
• a2-domains with polar axis along (010)
In absence of external electric fields, the crystal resides in an equilibrium multidomain
state consisting of a- and c-domains as depicted in Fig. 3.2(a) [71]. The domains are
separated by 180° and 90° domain walls. Since the lattice parameters a and c are different,
the a-domains are tilted with respect to the orientation of the c-domains. Upon the
application of an external electric field with E > Ec the crystal enters a single domain
state as schematically depicted for the (001) direction in Fig. 3.2(b) [34, 73].
3.2 Piezoelectric Materials 25
3.2.3 Lithium Niobate
Lithium niobate (LiNbO3) is of central importance in integrated and guided wave optics
as well as in filter devices and resonators [74]. Below its ferroelectric Curie temperature
of Tc = 1483 K LiNbO3 crystallizes in a distorted hexagonal close-packed configuration
[74]. Table 3.2 lists the piezoelectric coefficient relevant for this thesis. In contrast to
PZT and BaTiO3, LiNbO3 is a pyroelectric material. Thus, the spontaneous polarization
cannot be reversed by external electric fields [75].
Chapter 4
Magneto-Optical Kerr Effect with
Spatial Resolution
The establishment of spatially resolved magneto-optical Kerr imaging was a substantial
aim of this thesis, since it offers a powerful tool to investigate magnetic microstructure.
Therefore an existing magneto-optical setup [45] was extended to allow for spatial reso-
lution. In the following an overview of the new Kerr imaging setup and its components
is given. Moreover, a characterization of the system’s resolution and sensitivity is shown.
At last the results of Kerr imaging on thin ferromagnetic (FM) films are presented and
the possibility of quantitative Kerr microscopy is discussed in brief.
4.1 The Existing Experimental Setup
In Fig. 4.1 a schematic drawing of a conventional Kerr setup can be seen. Such a system
has been built up at the WMI by M. Pelkner [45]. The basic features of this experimental
setup is reviewed in this section before turning to Kerr effect measurements with spatial
resolution. In the existing experimental configuration the longitudinal as well as the polar
Kerr effect can be recorded. The transverse Kerr effect, however, cannot be monitored
since the magnetic field cannot be applied perpendicular to the plane of incidence in
the current setup. For the purpose of this thesis the longitudinal configuration is usually
desired and therefore the angle of light incidence is chosen to be 40°, which is the maximum
angle allowed by the setup geometry and has proven to give the best results in terms of
sensitivity of all allowed incident angles [45]. The basic components of the conventional
Kerr setup are:
• Light source
As light source a AlGaInP semiconductor diode is used. It emits coherent light
with a characteristic wavelength of 670 nm and power of 7 mW.
• Polarizers
The polarizers are Glan-Thompson polarizers made of calcite. Glan Thompson
polarizers assure a very high extinction ratio of " = 10−6, which is fundamental
27
28 Chapter 4 Magneto-Optical Kerr Effect with Spatial Resolution
sample
electromagnet
polarizer analyzer
objective lens
focusing lens
light source detector
Figure 4.1: Experimental setup for basic Kerr effect measurements.
to achieve good contrast conditions in Kerr microscopy as will be discussed in
Section 4.5.2.
• Electromagnet
The electromagnet consists of two coils which are mounted onto an iron yoke and
connected in series. It allows the application of a magnetic field along one direction.
The power supply is a bipolar Bob 20-20 manufactured by Kepco.
• Hall probe
To measure the magnetic field a Hall probe of the type 475 DSP fabricated by
Lakeshore is used. An analog output enables to control the power supply of the
electromagnet and thus a very exact magnetic field adjustment is possible.
• Detector
Light detection is carried out by a Si - photo diode with integrated signal amplifi-
cation. The active area of the diode has a diameter of 2.5 mm.
4.2 Kerr Microscopy
In the original Kerr setup presented above, lenses were used to focus the laser light onto
the sample collimating the laser spot to a diameter of about 0.5 mm on the sample
surface. Thus the region probed was quite small and no information about the spatial
arrangement of magnetic domains was retrieved. But in addition to that, the magneto-
optical Kerr effect allows to go beyond and visualize magnetic domain structures with
microscopy techniques [76]. This can be achieved with a Kerr microscope. During this
4.2 Kerr Microscopy 29
LED
CCD camera
objective lens
analyzerquarter wave plate
polarizer
slit aperture
Hall probe
electromagnetsample holder
Figure 4.2: Photograph of the Kerr microscopy setup which was built up during this thesis.
thesis the existing experimental setup was partly redesigned and extended to enable
spatial resolution. Therefore some major modifications had to be made, as depicted in
Fig. 4.2.
Detector
At first the detector had to be exchanged and a Luca (S) EMCCD (Electron Multiplying
Charged Coupled Device) camera manufactured by Andor was installed. The camera
has 658 × 496 active pixels with a size of 10 µm × 10 µm resulting in an image area of
6.58 mm × 4.96 mm. Each pixel has a bit depth of 14 and the maximum read out rate
is 27 frames per second. The camera is fan-cooled and the active sensor is additionally
cooled by a Peltier element to reduce readout noise. It is connected via USB 2.0 and
can be controlled with LabVIEW during measurements. Long- and short-pass filters
are mounted onto the camera in order to absorb light of a wavelength < 600 nm and
> 665 nm, respectively. Thus, noise arising from stray light falling onto the camera
sensor is minimized.
Light source
Secondly the laser diode had to be replaced by a LED (Light Emitting Diode). The
problem with laser illumination in microscopy and imaging is its coherence resulting in
strong interference fringes and speckle. Even digital image subtraction cannot remove
this problem, since the necessary stability of all components is not given. Consequently a
high power LED, Luxeon LXHL-LD3C fabricated by Lumitronix is used. The electrical
30 Chapter 4 Magneto-Optical Kerr Effect with Spatial Resolution
lensLED
Figure 4.3: Photograph of the mounted LED and the focusing lens.
power is 3 W and the luminous flux is specified as 90 − 140 lm, which is equivalent to
a radiometric power of around 0.5 − 0.8 W. The LED emits red light with a dominant
wavelength of = 627 nm a spectral half-width of Δ = 20 nm. The angle of radiation
is 130°. To ensure thermal stability the LED is mounted onto an aluminum block (cf.
Fig. 4.3).
A focusing lens is directly placed and mounted on top of the diode to collimate the
light to a radiation angle of around 3°. In order to prevent the emission of stray light,
an aluminum tube covers the diode and lens. Replacing the laser diode with a LED is
advantageous, because it emits incoherent light. However, the collimation of the emitted
light in the current setup is not perfect and thus the light intensity is certainly lower com-
pared to laser illumination. However, further improvements in the illumination including
improved collimation are currently under way.
Slit aperture
Between LED and polarizer a slit aperture is placed in the illumination path (cf. Fig. 4.2).
A correct illumination aperture is important in Kerr microscopy, since too small apertures
result in disturbing diffraction fringes [5]. However, a large aperture causes light beams
to hit the polarizers at small angles which are not lying in the plane of incidence. The
consequence is an increased background intensity due to depolarization effects and thus
the magneto-optical contrast is reduced [5, 77]. Therefore, the slit aperture should be
oriented parallel to the plane of incidence for the longitudinal Kerr effect. The home-
built slit aperture integrated in the Kerr setup allows to adjust the slit size for optimal
illumination and contrast conditions.
4.3 Image Processing 31
Quarter wave plate
A rotatable achromatic quarter wave plate manufactured by Thorlabs is placed between
the sample and the analyzer in the illumination path. The wavelength bandwidth goes
from 450 nm to 800 nm. The ellipticity of the reflected light can be removed by proper
adjustment of the wave plate which finally improves the imaging contrast (cf. Section 4.5).
LabVIEW measurement program
The imaging experiments are automated with LabVIEW measurement software. The
central LabVIEW program developed in the framework of this thesis “MOKELuca.vi”
synchronizes all measurement devices and contains the camera readout and data handling.
Moreover, programs for image post-processing, e.g. contrast enhancement, have been
implemented.
Adjustment procedure
The proper adjustment of all components is of great importance for good contrast con-
ditions. At first the slit aperture should be placed properly in order to assure a homoge-
neous illumination of the sample. Moreover a good focus of the sample onto the camera
is necessary and is achieved by careful adjustment of lens and camera positions. The
objective lens can be tilted to reduce the image distortion arising from the oblique light
incidence [5]. The polarizer is placed such that the transmission axis is perpendicular to
the plane of incidence (s-polarized) to be sensitive only on the longitudinal component
of the magnetization as discussed in Chapter 2. The quarter wave plate and the analyzer
are then iteratively adjusted to achieve extinction before opening the analyzer by the
desired angle (cf. Section 4.5).
4.3 Image Processing
Digital image processing and contrast enhancement techniques are a substantial tool to
visualize magnetic domains. The detection of magnetic microstructure in a recorded
raw image without digital image processing is difficult (cf. Fig. 4.4(a)). At first, the
probed material has to exhibit an as large as possible Kerr rotation in order to obtain
strong contrast between domains of opposite magnetization direction. Moreover surface
imperfections and irregularities can produce a non-magnetic contrast which is then su-
perimposed with the desired magnetic contrast. At last, the illumination should be very
homogeneous within the field of view to assure good visibility. In sum, these issues impede
direct detection of magnetic domains.
In order to reduce those problems the technique of difference imaging was proposed
[7, 8, 78]. The standard procedure is to subtract two images digitally to obtain a resulting
image containing only contrast of magnetic origin. Therefore, an image of the sample
32 Chapter 4 Magneto-Optical Kerr Effect with Spatial Resolution
1 mm
(a) (b)
Figure 4.4: Domain pattern of a polycrystalline Co thin film before (a) and after digital contrast en-
hancement (b).
in a magnetically saturated state is acquired and stored in memory as reference. Then
this reference image is subtracted from images taken at other magnetic field strengths.
Ideally the resulting difference image contains only contrast which is truly of magnetic
origin.
In Fig. 4.4 a real image and a difference image of the same domain pattern are de-
picted. On the one hand a very weak domain pattern can be seen in the real image (cf.
Fig. 4.4(a)). The contrast between the domains of opposite magnetization is very weak.
The illumination of the sample is also not perfectly homogeneous and small speckle exist.
On the other hand the difference image (cf. Fig. 4.4(b)) displays a very clear domain
pattern. The contrast between black and white domains has improved significantly. Fur-
thermore the imaging problems with speckle are reduced. The whole process of image
subtraction is done in real time in the LabVIEW measurement program in order to di-
rectly observe the domain structure. Finally the difference image is saved and can then
be edited by another LabVIEW program or a commercial image processing software for
further contrast enhancement. It should be mentioned that the time between the acqui-
sition of the reference image and the acquisition of the image with the desired domain
pattern should not be too large. Drift effects, caused for instance by thermal fluctuations,
can have a negative influence on the visibility of domains.
4.4 Resolution 33
a
D
sample lens camera
g b
Figure 4.5: Schematic drawing of the imaging setup.
4.4 Resolution
The resolution is a crucial quantity in imaging and microscopy experiments. Therefore,
a short estimation of the resolution of the established Kerr imaging setup is given. Here
the resolution depends primarily on the imaging components, the system geometry and
the quality of the detector. Simple wide angle objective lenses image the sample onto
the CCD camera as seen schematically in Fig. 4.5. In the thin lens approximation the
imaging can be described by [44]
1
f=
1
g+
1
b, (4.1)
where f is the focal width of the lens, g and b the distance from sample to lens and from
lens to detector, respectively. The achievable magnification is limited by the geometry of
the setup, since the minimum distance from object to lens is about 15 cm. Usually the
objective lens and CCD are adjusted to have transverse magnifications between one to
three. In this configuration decent contrast and brightness conditions are given. Then,
assuming a transverse magnification of three and considering the pixel size of 10 µm ×
10 µm, a nominal resolution of around 3 µm can be achieved. But so far the wave nature
of light and the phenomenon of diffraction was neglected. The diffraction-limited spatial
resolution of a microscope is governed by [44]
dmin = 0.61 ⋅
NA, (4.2)
where dmin is the minimum distance between two points on the sample that can be
resolved. NA is the numerical aperture which is defined as NA = n ⋅ sin with the index
of refraction n = 1 for air and the angle as defined in Fig. 4.5. Assuming a lens radius
of D2 = 1 cm and a working distance of g = 20 cm the numerical aperture is approximately
given by NA ≈ 5 ⋅ 10−2. Inserting the wavelength = 627 nm you obtain a minimum
34 Chapter 4 Magneto-Optical Kerr Effect with Spatial Resolution
distance of dmin ≈ 8 µm and thus a lower resolution than the nominal resolution calculated
above. This reveals that our experimental setup is limited by the achievable numerical
aperture. In reality the resolution is probably even worse, since this consideration holds
only true for ideal imaging. No interfering effects like aberrations in the tilted objective
lens have been taken into consideration. Thus the current setup is capable of doing low
resolution Kerr microscopy [5] with a magnification of order 1.
Higher resolutions could be achieved by using a microscope objective and an eyepiece
instead of the objective lens. Especially a smaller working distance allowing the usage of
microscope objectives should be accompanied by a major improvement in resolution. But
this feature was not realized during this thesis since a total redesign of the experimental
setup would have been necessary. We therefore refrained from aiming at huge spatial
resolution and rather focused on the modifications described above, as the typical samples
studied had lateral dimensions of the order of mm. State of the are high resolution Kerr
microscopes are usually based on conventional polarization microscopes and in principle
they can achieve resolutions up to the limit of optical microscopy [79].
4.5 Characterization
There are many parameters in image recording devices that influence the quality of light
detection and the resulting image. Specifically in Kerr microscopy a large contrast be-
tween domains of opposite magnetization direction (black and white domains) is neces-
sary to visualize domain patterns. Usually the limiting factor is noise. Therefore, a short
overview of noise sources and a characterization of the setup is given.
4.5.1 CCD Noise and Signal-to-Noise Ratio
At first some basics of CCD camera noise are reviewed. The signal-to-noise ratio S/N is
determined by the ratio of the generated signal electrons to the number of noise electrons.
The signal S depends upon the number of photons Np impacting onto a camera pixel
during the exposure time and upon the quantum efficiency of the device:
S = Np ⋅ (4.3)
Regarding noise in light detection, there are three major sources that contribute to the
system noise [80]:
• Dark noise Ndark arises from thermally generated charges in the sensor. This noise
component can be reduced by cooling down the light sensor. Our CCD camera is
cooled to −20° C and the nominal dark current is 0.5 e−
s⋅pixel .
• Readout noise Nreadout is an inherent property of the sensor and originates from
the process of amplification and conversion of photoelectrons into a voltage. The
readout noise is specified to be between 1 and 15 e−
s⋅pixel at maximum frame rate.
4.5 Characterization 35
• Photon noise or shot noise Nphoton results from the inherent statistical variation in
the arrival rate of photons incident on the CCD and it follows the Poisson statistics.
Therefore, photon noise can be quantified as Nphoton =√Np ⋅ .
Taking everything into consideration the signal-to-noise ratio can be expressed as fol-
lows [80]:
S
N=
Np ⋅ √N2
photon +N2readout +N
2dark
(4.4)
A comparison of the noise contributions reveals, that shot noise dominates over dark
noise and readout noise.
Dark noise can be estimated as follows: The full well capacity of our CCD camera,
which is the number of electrons that can be held in one potential well of one pixel,
is given by 26000 electrons. Thus the number of counts arising from the nominal dark
current of 0.5 e−
s⋅pixel can be estimated by
Ndark ≈0.5 e−
s⋅pixel
26000 e−
214 counts
≈ 0.03counts
pixel ⋅ s(4.5)
where the dynamic range of 214 counts was inserted.
The maximum readout noise Nreadout can be approximated analogously:
Nreadout ≈15 e−
s⋅pixel
26000 e−
214 counts
≈ 10counts
pixel ⋅ s(4.6)
Shot noise, however, is determined by the number of incident photons and thus by the
number of counts in the end. During operation the number of counts detected by the
CCD is usually around 10000 resulting in a shot noise of
Nphoton ≈√
10000counts
pixel ⋅ s= 100
counts
pixel ⋅ s. (4.7)
This short noise evaluation points out, that signal detection is limited by shot noise
due to the large light intensities [5]. Neglecting dark noise and readout noise, Eq. (4.4)
simplifies to
S
N=√Np ⋅ ∝
√texp ⋅Apixel ⋅ (4.8)
where Apixel is the active pixel area and texp the exposure time [81]. Thus, an increase
in exposure time or pixel size via binning improves the signal-to-noise ratio. The accumu-
lation and averaging of several images per field point is another method of improving the
signal-to-noise characteristics. Of course, the dynamic range of the camera or unwanted
drift effects set limits to long exposure times. Other sources of noise in the imaging
36 Chapter 4 Magneto-Optical Kerr Effect with Spatial Resolution
M
s
p
ya
polarizer analyzer
sample
fk-fk
incidentlight
reflectedlight
(a)
- 4 0 - 2 0 0 2 0 4 08 0
8 2
8 4
MOKE
signa
l (arb.
units)
µ0 H ( m T )
S m o
(b)
Figure 4.6: (a) Schematic illustration of polarized light amplitude vectors before and after reflection
from the sample, which is composed of domains with opposite magnetization direction.
Definition of the Kerr angle 'k and the analyzer angle a. (b) Definition of the absolute
magneto-optical signal Smo.
system are fluctuations in the intensity of the light source or in the optical path. Those
noise contributions are not discussed here, since they can be hardly quantified and have
proven not to have substantial influence onto the imaging quality.
4.5.2 Magneto-Optical Contrast
The figure of merit in Kerr microscopy is the magneto-optical contrast. In the preceding
section parameters like exposure time and binning and their effect on the S/N ratio have
already been explained. Albeit the magneto-optical contrast depends also crucially on the
settings of polarizer and analyzer as defined in Fig. 4.6(a). Ideally the polarizer transmits
only s-polarized light. Then the intensity I(±M) of the black and white domains with
respect to the total intensity I0 reflected by the sample is given by the law of Malus
[82, 83]
I(±M) = I0 ⋅ sin2( a ± 'k) + Ir. (4.9)
Ir denotes the non-zero residual transmitted intensity through the analyzer in extinc-
tion, which arises from depolarization effects and finite illumination apertures. a is the
angle of which the analyzer is opened from extinction and 'k stands for the Kerr rotation.
The relative magneto-optical signal smo can be expressed as the difference between the
two intensities I(±M):
smo = I(+M) − I(−M) = I0 sin 2 a sin 2'k (4.10)
In [45, 84] it is is shown how to extract the analyzer angle of optimum signal by
4.5 Characterization 37
minimizing the normalized signal Ik =smo
Iwith respect to a. The resulting analyzer angle
of optimum signal is found to be at 'a =√Ir/I0. However, the best visibility of domains
is rarely given by the criterion of optimum signal as defined above. First, if the analyzer
angle a is too small, the light intensity reaching the camera is not sufficient and domains
will be barely visible. Second, the relative magneto-optical signal increases in first order
with 2a (cf. Eq. (4.9)) which corresponds to an increase of the visible contrast. Third, the
increase in signal is accompanied by an increase in photon noise. All in all, good domain
visibility rather requires a large signal-to-noise ratio. The absolute magneto-optical signal
as seen in Fig. 4.6(b) is given by Smo = A ⋅Np ⋅smo, where A is a constant of proportionality
depending on the quantum efficiency and the amplification and conversion process of the
detected photons [5, 84]. Shot noise is then given by Nphoton =
√12ANp (I(+M) + I(−M)).
Applying Eq. (4.4) in the limit of high light intensities we obtain the following expression
for the Smo/N ratio [5]:
Smo
N=√ANp ⋅
I(+M) − I(−M)√
12 (I(+M) + I(−M))
(4.11)
Inserting Eq. (4.9) and minimizing with respect to a gives [5]
tan opta =
4
¿ÁÁÀ'2
k + Ir/I0
1 + Ir/I0
, (4.12)
where opta denotes the analyzer angle of maximum contrast visibility. Assuming an
extinction ratio of Imin/I0 ≈ 10−6 according to the specifications of the Glan Thompson
polarizers and a Kerr rotation e.g. for Fe50Co50 of 'k ≈ 0.03° [45] the angle of optimal
contrast is given by opta ≈ 1.9°. Although this result can only be a rough estimation, it
points out that the analyzer has to be opened beyond extinction. In reality the extinction
ratio of the polarizers is probably larger because of depolarization effects or imperfections
in the illumination path. Also noise contributions arising from fluctuations in the light
source or in the optical path have not been taken into consideration. In the experiment
the analyzer is usually opened as far as necessary to obtain good imaging results without
being in camera saturation.
An additional element that enhances the magneto-optical contrast is the inserted quar-
ter wave plate. It removes the ellipticity of the reflected light and thus enhances the signal
amplitude. Comparing hysteresis loops obtained on an identical sample with and without
quarter wave plate the magneto-optical contrast doubles (cf. Fig. 4.7) without changing
noise characteristics.
4.5.3 Sensitivity
Having explained the fundamentals of noise and magneto-optical contrast, the sensitivity
of the established Kerr setup regarding the minimal resolvable Kerr rotation is analyzed
38 Chapter 4 Magneto-Optical Kerr Effect with Spatial Resolution
- 6 0 - 4 0 - 2 0 0 2 0 4 0 6 0- 6
- 3
0
3
6
S m o w i t h o u tλ/ 4 p l a t e
MOKE
signa
l (arb.
units)
µ0 H ( m T )
S m o u s i n gλ/ 4 p l a t e
Figure 4.7: Comparison of the magneto-optical signal obtained by hysteresis measurements with (red
curve) and without (black curve) quarter wave plate.
and compared to the original MOKE setup described in [45]. The procedure is as follows:
First, the signal-to-noise ratio of a single pixel of the CCD camera is extracted. Using this
result, the minimum resolvable Kerr rotation of the Kerr imaging system is approximated.
In magneto-optical Kerr effect measurements the relevant signal is Smo as depicted in
Fig. 4.6(b). Noise, however, sets an ultimate limit on the camera’s sensitivity. The fol-
lowing analysis regarding the camera sensitivity was done using a polycrystalline Ni-film
with a thickness of 100 nm (sample M090224A). A M(H) hysteresis loop was recorded
with an exposure time texp = 1 s, an analyzer angle of a = 2° and four accumulations per
field point. This means, that four single images per field point were taken and averaged.
The resulting M(H) curve of one example pixel can be seen in Fig. 4.8(a).
To determine the signal Smo and the noise N , mean and standard deviation of 45
measurement points on the lower and upper branch of the hysteresis loop were calculated
for this particular pixel. The same evaluation was also done for the sum signal of several
pixels. The calculated signal-to-noise ratios are plotted in Fig. 4.8(b) as a function of the
number of pixels Npixel that were summed. As expected, Smo/N increases with√Npixel.
A function fit (f = Afit
√Npixel) with the fit parameter Afit leads to an average signal-to-
noise ratio per pixel of Afit = S/N ≈ 2. By integrating over many pixels, the total S/N
ratio can be improved notably. For instance, if the signal of all camera pixels (658× 496)
is summed up, the resulting integral signal-to-noise ratio yields S/N ≈ 1140, assuming
a constant signal-to-noise ratio per pixel of 2. Often only small sample regions are of
interest regarding an M(H) characterization. Then, the usual procedure is to define a
region of interest (ROI). Hence, the hysteresis curve M(H) is obtained by integrating
over all pixels which belong to the ROI.
4.6 Imaging of Thin Ferromagnetic Films 39
- 3 0 - 2 0 - 1 0 0 1 0 2 0 3 0
1 2 0
1 2 4
1 2 8MO
KE sig
nal (a
rb. un
its)
µ0 H ( m T )(a)
0 3 0 6 0 9 0 1 2 00
5
1 0
1 5
2 0
2 5
S mo/N
Νp i x e l
m e a s u r e m e n t d a t a f i t d a t a
(b)
Figure 4.8: (a) Signal of a single pixel in a M(H) measurement. (b) Smo/N ratio in dependence of the
number of integrated pixels Npixel. The black curve represents the measured data, the red
curve is a fit curve f = Afit
√Np with fit parameter Afit.
To estimate the minimum Kerr rotation which is resolvable by the novel Kerr setup,
the Kerr rotation of the used Ni-film has to be calculated first. For a wavelength of
= 627 nm, the refractive index of Ni is given by n = 1.98 + i3.74 [85] and the magneto-
optical constant yields Q = 0.0005 − i0.0002 [86]. Applying Eq. (2.32), the real part of
the Kerr rotation is given by ∣'k,Ni∣ ≈ 0.02°. Then, the minimum resolvable Kerr rotation
Δ'k is given by the ratio of 2 ⋅ 'k,Ni and S/N [84]. Inserting S/N = 1140 from above for
an integrally recorded hysteresis loop, the minimum resolution finally yields Δ'k ≈ 0.15′′.
Comparing this sensitivity with the one of the original MOKE setup (Δ'k ≈ 1.8′′ [45]),
the resolution of the novel Kerr imaging setup in terms of the Kerr rotation is superior
by an order of magnitude.
4.6 Imaging of Thin Ferromagnetic Films
In this section imaging experiments on thin ferromagnetic films are presented in order
to visualize the capabilities of the low resolution imaging system. A cobalt (Co) and a
chromium dioxide (CrO2) thin film are chosen as examples. In both cases reference images
at large negative saturation fields were taken and the technique of difference imaging was
applied to obtain optimal magnetic contrast.
Cobalt thin film
A polycrystalline Co thin film (sample M090512A) with a thickness of 50 nm was prepared
on a Silicon (100) oriented substrate by electron beam evaporation at a base pressure of
approximately 1.9×10−7 mbar. The coercive field was determined to be 0Hc = 1.81 mT.
40 Chapter 4 Magneto-Optical Kerr Effect with Spatial Resolution
(a) (b)
(c) (d)
(e) (f)
1 mm
Figure 4.9: Domain pattern of a polycrystalline Co thin film at (a) 0H = 1.73 mT, (b) 0H = 1.78 mT,
(c) 0H = 1.81 mT, (d) 0H = 1.83 mT, (e) 0H = 1.86 mT, (f) 0H = 1.91 mT.
4.6 Imaging of Thin Ferromagnetic Films 41
1mm e.a.(a) (b) (c) (d)
Figure 4.10: Domain evolution in an epitaxial CrO2 thin film with a thickness of 200 nm. Domains
nucleate in form of stripes.
The imaging was done with an exposure time of texp = 1 s and one image accumulation
per field point. In Fig. 4.9 the domain pattern is depicted.
In Fig. 4.9(a) and Fig. 4.9(f) the image color is homogeneously dark gray or white,
respectively. Thus the Co thin film is in a single domain state at the corresponding
magnetic fields. In contrast, Fig. 4.9(c)-(e) reveal domain patterns which are called
zigzag walls [5]. This kind of domain wall typically exist in polycrystalline ferromagnetic
films of corresponding thickness. Higher resolution imaging would probably allow to see
the internal structure of the domain wall, which is a cross-tie wall [5].
CrO2
Chromium dioxide is a half-metallic ferromagnet and is a very promising candidate for
spintronic applications because its spin polarization is 100% [87]. The magneto-optic
investigations were done on an epitaxial (110)-oriented CrO2 thin film grown on a titani-
umdioxide substrate (TiO2) with a film thickness of 200 nm. The film was prepared by
chemical vapor deposition of the group of A. Gupta, University of Alabama, Tuscaloosa,
USA. In Fig. 4.10 the evolution of magnetic domains can be seen, with the easy axis
being oriented parallel to the external magnetic field.
Domain wall nucleation and motion dominates the easy axis magnetization reversal,
as already observed in [88]. Stripe-like domains characterize the pattern of the mag-
netic substructure. The formation of stripe domains has also been observed in other
ferromagnetic films [5].
42 Chapter 4 Magneto-Optical Kerr Effect with Spatial Resolution
Imag
e inte
nsity
(arb.
units)
M a g n e t i z a t i o n a n g l e
c a l i b r a t i o n
(a) (b)
Figure 4.11: (a) Calibration of image intensities to the corresponding magnetization directions for quan-
titative Kerr microscopy. (b) Domain pattern of a Co thin film quantitatively evaluated.
4.7 Quantitative Kerr Microscopy
So far, the obtained Kerr microscopy images were only analyzed from a qualitative point
of view. Apart from that, it is also possible to do a quantitative determination of mag-
netization directions [89, 90]. The proposed method is based on a combination of the
longitudinal and the transverse Kerr effect and works for thin films where the magne-
tization lies in the film plane. In essence, an identical domain pattern is recorded in
longitudinal and transverse configuration. In addition, calibration experiments for neg-
ative and positive magnetization saturation for both measurement configurations have
to be performed. The information of both effects is then digitally combined. Longitu-
dinal Kerr effect measurements with s-polarized incident light are only sensitive to the
longitudinal component of the magnetization. For a quantification of the longitudinal
component at first two images in negative and positive saturation have to be recorded
and their image intensities serve as calibration measurements (cf. Fig. 4.11(a)). Since
the longitudinal magnetization component is the projection of the total magnetization
onto the longitudinal direction, every color value can be assigned to a longitudinal mag-
netization component by fitting a cosine function to the calibration points. The same
procedure would be necessary for the transverse magnetization component. But in the
current setup no calibration measurement for the transverse Kerr effect can be done, since
the used electromagnet only allows magnetic fields along one direction. Therefore, it is
not possible to determine the transverse magnetization direction quantitatively.
In Fig. 4.11(b) an example for a quantitatively evaluated image obtained by the new
Kerr microscope setup is depicted. Small image regions that have a uniform color are
analyzed and the average image intensity is calculated. Then a magnetization direction
can be assigned. As seen by the color code in Fig. 4.11(b), the transverse direction cannot
4.8 Summary 43
be assigned uniquely. For illustration purposes, one of the two possible magnetization
orientations is chosen as represented by the red arrows.
4.8 Summary
Summing up, spatially resolved magneto-optical Kerr imaging has been established at the
WMI in the framework of this thesis. The new setup allows Kerr imaging with a spatial
resolution on the order of 10 µm and with an integral sensitivity in terms of the Kerr
rotation of about 0.15′′. Therefore, large scale domain patterns of typical ferromagnetic
thin film samples prepared at the WMI can be studied.
Chapter 5
Magneto-Optics on
Ferromagnetic/Ferroelectric Hybrids
The magnetoelastic, magnetoelectric and piezoelectric effect introduced in Chapter 2 can
be used to achieve an electric field control of ferromagnetism. Many different approaches
involving different kinds of ferroelectric and ferromagnetic materials have been proposed
to this end [12, 13, 14, 16, 17]. Among the most promising candidates are novel hybrid
structures consisting of ferroelectric substrates and ferromagnetic thin films [18, 21, 91].
The concept and functionality of these hybrids has already been discussed extensively
elsewhere [19, 20, 34, 92]. However, the magnetic microstructure of the ferromagnetic
constituent has not been subject to any investigation so far. Consequently, the magneto-
optical Kerr setup allowing to resolve local magnetic substructure established in this
thesis (cf. Chapter 4) has been used to study these hybrid structures intensively.
In this chapter ferromagnetic films affixed to ferroelectric substrates are investigated
by means of Kerr microscopy. Hybrids consisting of Ni or Fe50Co50 films deposited onto
piezoelectric actuators are discussed and compared to bilayers of BaTiO3 substrates and
ferromagnetic thin films. In both cases, the voltage dependence of the magnetic hysteresis
and the corresponding domain evolution of the ferromagnetic films shall be studied in
detail. Moreover, experiments demonstrating the magnetization switching induced by
strain are presented.
5.1 Ferromagnetic Thin Films on Piezoelectric Actuators
Polycrystalline Ni thin films on piezoelectric actuators have been studied by means of
ferromagnetic resonance spectroscopy and SQUID magnetometry in [34, 92]. In these
papers it was shown that a rotation of the magnetic easy axis by 90° upon changing the
polarity of the piezo-voltage is possible. Moreover, it was reported that the application
of electric fields allows to switch the magnetization and also to adjust the magnetization
direction reversibly within a range of approximately 70° in nickel thin films.
Following up on these results this section deals with the magneto-optic investigation of
the same kind of hybrid structures. At first, the preparation of the samples is presented
45
46 Chapter 5 Magneto-Optics on Ferromagnetic/Ferroelectric Hybrids
dominantelongationaxis
piezoelectricactuator
FM
Vp
x
y
z glass isolation
electrodes
PZT
(a)
2 mm(b)
Figure 5.1: (a) Schematic view of the prepared samples and illustration of the on stack isolating design of
the actuator. The ferromagnet is deposited onto the face of the actuator [92]. (b) Photograph
of a Ni film cemented onto piezoelectric actuator.
before dealing with measurement results.
5.1.1 Sample Preparation
Commercially available Piezomechanik PSt 150/2×3/5 actuators [65] serve as piezoelectric
substrates. They are built in an on stack isolating design as illustrated in Fig. 5.1(a):
100 µm thick PZT active layers are embedded by metal electrodes with a thickness of
10 µm forming a multilayer structure with dimensions x × y × z = 2 × 3 × 5 mm3. Due to
the interdigitated contact scheme, the direction of the electric field as well as the electric
polarization within adjacent PZT monolayers alternates. The actuator has a nominal
expansion of Δyy = 0.13% along the dominant axis of elongation (d33 effect) in the semi-
bipolar range of −30 V ≤ Vp ≤ +150 V. As shipped, the actuator’s face is covered with a
50 µm thick polymer coating. This coating has to be removed via polishing prior to the
deposition of the ferromagnetic films onto the actuator. Two different techniques have
been used to deposit the ferromagnet onto the actuator.
Evaporation of Ni Thin Films onto Piezoelectric Actuators
Prior to the evaporation process, the coating-free face of the actuator has to be cov-
ered with an isolating PMMA (polymethylmethacrylate) layer. Therefore, PMMA 950K
is spin-coated onto the actuator’s face. After that, the sample is baked at 110° C for
approximately 30 minutes. This process is repeated twice in order to obtain a result-
ing PMMA thickness of approximately 140 nm. Then a polycrystalline Ni thin film is
deposited onto the PMMA covered actuator by electron beam evaporation at a base pres-
sure of around 1×10−7 mbar and with deposition rates of 1−2 A/s. A list of the prepared
samples and further details about material, film thickness and deposition technique can
5.1 Ferromagnetic Thin Films on Piezoelectric Actuators 47
Sample Material Deposition technique Thickness of FM film (nm)
M090213P Ni Direct evaporation 100M090713A Ni Direct evaporation 100M090715A Ni Direct evaporation 120M090220Aa Ni Cementation 50M090220Ac Ni Cementation 50B080623Ad Fe50Co50 Cementation 20
Table 5.1: Overview of prepared samples.
be found in Tab. 5.1.
Cementation of Ni and Fe50Co50 Thin Films onto Actuators
In an alternative preparation procedure ferromagnetic films are first evaporated onto MgO
(100) (magnesium oxide) substrates under the same conditions as described above. After
the evaporation the MgO substrates underneath the ferromagnetic film are polished down
to a thickness of approximately 50 µm and then cemented onto the face of the actuator
using Vishay M-Bond 600 epoxy [30] (cf. Fig. 5.1(b)).
Comparison of the Fabrication Techniques
The evaporation technique yields the advantage that the thickness of the PMMA buffer
layer can be adjusted in such a way that the strain transmission from the substrate onto
the ferromagnet is almost perfect. Moreover, the surface of the PMMA buffer layer is very
flat and smooth which is essential for high quality of the ferromagnetic thin film regarding
its magnetic and optical properties. In contrast, in samples prepared by the cementing
technique, the actual strain onto the MgO is about 70% smaller than the nominal stress
exerted by the actuator [34]. The reduction of strain transmission can be attributed to
the use of epoxy and strain relaxation in the MgO substrate. A further drawback is the
fact that constant thickness and uniformity of the applied epoxy throughout the sample
cannot be guaraneed. Thus the strain transmission might vary across one and the same
sample and it most likely also differs from sample to sample. However, the cementing
technique offers significant advantages compared to the evaporation technique as it allows
to structure well defined Hall-bars into the ferromagnet.
5.1.2 Strain Control of the Magnetic Anisotropy
In this section the phenomenology of strain-induced anisotropy in polycrystalline ferro-
magnetic films on piezoelectric actuators will be reviewed, which has been discussed in
detail in [92]. The free energy density approach introduced in Chapter 2 is applied in
48 Chapter 5 Magneto-Optics on Ferromagnetic/Ferroelectric Hybrids
Vp > 0 V
y
x
M
(a)
Vp < 0 V
y
x
M
(b)
Figure 5.2: Deformation of the actuator and attached ferromagnetic film upon applying a voltage
Vp ≠ 0 V. The dotted contours depict the relaxed actuator at Vp = 0 V and the red ar-
row indicates the orientation of the induced easy axis within a Ni thin film. (a) Tensile
strain along y induces a magnetic easy axis along x. (b) Compressive strain along y induces
a magnetic easy axis along y [92].
order to deduce the orientation of the magnetic easy and hard axis upon deformation of
the actuator.
Applying voltages Vp > 0 to the actuator results in an expansion along its dominant
elongation axis y (cf. Fig 5.2(a)). This is accompanied by a contraction along the two
orthogonal axes. Thus tensile strain along y is exerted onto the attached ferromagnetic
film, whereas the strain along x is compressive. In contrast, for Vp < 0 the situation is
inverted as depicted in Fig. 5.2(b). In the following the strains along x, y and z are
expressed as "1, "2 and "3 in Voigt matrix notation [38]. Shear strains can be neglected
since they average out in polycrystalline films. According to elastic theory the strains
along x- and y-axis, "1 and "2, are linked by the Poisson ratio
= −"1
"2
(5.1)
with = 0.45 [65]. The out-of-plain strain "3 is given by [19]
"3 = −c12
c11
("1 + "2) (5.2)
where c11 and c12 are the elastic moduli of the chosen ferromagnetic material (cf.
Chapter 3). The strain "2 along the y-axis can be calculated assuming a linear expansion
of the actuator:
"2 =Δz
z
Vp
180 V(5.3)
According to Eq. (5.1)-(5.3), all strains in the ferromagnetic film can be calculated for
5.1 Ferromagnetic Thin Films on Piezoelectric Actuators 49
any given piezo-voltage Vp assuming perfect strain transmission through the buffer layer.
In order to determine the equilibrium orientation of the magnetization as a function of
Vp the free energy ansatz introduced in Eq. (2.14) is chosen:
Ftot = Fstat + Fdemag + Fmagel (5.4)
In absence of external magnetic fields, the Zeeman term vanishes leaving the demag-
netization term and the magnetoelastic contribution. For thin films it can be assumed
that the shape anisotropy dominates over the magnetoelastic energy causing the magne-
tization to remain only within the film plane. However, no further anisotropy is induced
by the demagnetization term. Consequently the free energy density simplifies to
Ftot = Fmagel =Kmagel,x sin2 Θ +Kmagel,y cos2 Θ, (5.5)
where the polar coordinates introduced in Fig. 2.2 have been applied considering
Φ = 90°. Inserting the strains calculated above into the expression for Kmagel,x and Kmagel,y
(cf. Eq. (2.16)) and minimizing Eq. (5.5) with respect to Θ yields the orientation of the
magnetically easy and hard axes. The following results have been obtained for Ni [92]
and Fe50Co50, respectively:
• Ni: Applying tensile strain along y (Vp > 0) the easy axis is oriented along x (cf.
Fig. 5.2(a)). For compressive strain along y (Vp < 0) it is aligned parallel to the
y-direction (cf. Fig. 5.2(b)).
• Fe50Co50: Tensile strain along y (Vp > 0) induces an easy axis that is oriented along
y. For compressive strain along y (Vp < 0) the easy axis points along the x-axis.
In sum, the response of the magnetic anisotropy onto the applied strain is opposite in
the two materials. This behavior originates from the opposite sign of the magnetostrictive
coefficients Ni = −32.9⋅10−6 and FeCo = 72.5⋅10−6 (cf. Chapter 3). In the following section
it will be checked whether the theoretical prediction of the easy axis orientation can be
corroborated with magnetic hysteresis measurements at constant strain.
5.1.3 Voltage Dependence of the Magnetic Hysteresis
The influence of strain onto the magnetic hysteresis is analyzed in this section
following [92, 93, 94]. At first, measurements of ferromagnetic films cemented onto ac-
tuators are presented. Then ferromagnets evaporated directly onto actuators are inves-
tigated. All measurements were done with the imaging setup including LED and CCD
camera described in Chapter 4. If not stated otherwise, the hysteresis curves were ob-
tained by integrating over all pixels of the CCD camera (cf. Chapter 4).
50 Chapter 5 Magneto-Optics on Ferromagnetic/Ferroelectric Hybrids
-10 -5 0 5 10
-1
0
1
M/M
s
µ0H (mT)
Vp=+150 V Vp=0 V Vp=-30 V
Vp
H||x
Ni
(a)
-40 -20 0 20 40
-1
0
1
M/M
s
µ0H (mT)
Vp=+150 V Vp=0 V Vp=-30 V
Vp
H||y
Fe50Co50
(b)
Figure 5.3: Hysteresis curves at different applied voltages Vp for (a) Ni at H∥x (sample M090220Ac) and
(b) Fe50Co50 at H∥y (sample B080623Ad).
Ferromagnetic Films Cemented onto Piezoelectric Actuator
In Fig. 5.3(a) three M(H) curves of Ni (sample M090220Ac) at different strain states
can be seen. The external magnetic field was prepared at 0H = −60 mT, then swept
to 0H = 60 mT and cycled back to 0H = −60 mT again. It was oriented along the
x-axis (cf. Fig. 5.1(a)). The piezo-voltage remained constant during one sweep. The
three hysteresis measurements were recorded at Vp = −30 V, Vp = 0 V and Vp = +150 V,
respectively.
The hysteresis curves obtained for the Fe50Co50 sample (B080623Ad) are depicted in
Fig. 5.3(b). The magnetostrictive constants of Ni and Fe50Co50 have different signs (cf.
Chapter 3). Therefore, the dependence of the orientation of the easy axis on the applied
strain is inverted in Fe50Co50 compared to Ni as explained above. To allow for a direct
comparison with the hysteresis curves shown for Ni, the results for the Fe50Co50 sample
being rotated by 90° are shown. Thus H pointed along the y-axis. The measurement
procedure was analogous to the Ni sample above.
According to the theory of magnetoelasticity presented above, the magnetic easy axis
should be aligned parallel to the external magnetic field for voltages Vp > 0 in both mate-
rial systems considering their orientation as described above. However, this expectation
cannot be corroborated without ambiguity by the measurement results. Generally, a
rectangular hysteresis loop indicates a magnetic easy axis, whereas a harder axis is char-
acterized by a smooth and S-shaped loop [32]. But the measured hysteresis curves are
equally smooth and S-shaped for both Vp = +150 V and Vp = −30 V. Thus a clear de-
termination of the orientation of hard and easy axes upon applied voltage is not possible
considering the measurement results for Ni and Fe50Co50.
5.1 Ferromagnetic Thin Films on Piezoelectric Actuators 51
However, the coercive field in both samples is shifted to larger magnetic field values for
Vp = 150 V and to slightly smaller fields for Vp = −30 V compared to the relaxed state for
Vp = 0 V. The absolute difference is approximately 2 mT for Ni as well as for Fe50Co50.
Thus a larger relative shift of the coercive field for different strain states can be observed
in the Ni sample.
Two arguments can be given to support this observation. First, the coercivity can be
calculated for a single domain particle using a Stoner-Wohlfarth approach [32, 95]. In
this simple approximation the coercive field is given by Hc = 3/Ms, where denotes
the applied stress and Ms the saturation magnetization. The magnitudes of the magne-
tostrictive constants can be estimated as ∣FeCo∣ ≈ 2 ⋅ ∣Ni∣ and the values for saturation
magnetization yield MFeCos ≈ 4 ⋅MNi
s (cf. Chapter 3). Assuming that the applied stress
onto Ni and FeCo is the same, the relative shift in the coercive force is expected to be
approximately twice as large for Ni compared to Fe50Co50. Second, Fe50Co50 has a larger
coercive field than Ni and is therefore the harder magnet. As a consequence Fe50Co50
has a larger resistivity against demagnetization caused by external fields or applied strain
than Ni [26].
The results shown above were reproducible and have been observed for all the samples
prepared by the cementing technique. Therefore, the presentation of results obtained
with other samples is omitted.
Ferromagnetic Films Evaporated onto Piezoelectric Actuators
All the samples prepared by direct evaporation showed qualitatively the same results
and therefore the following study is restricted to the results obtained with the sample
M090713A. The corresponding M(H) curve at different strain states can be seen in
Fig. 5.4, with the external magnetic field being aligned along the x-axis. In this sample,
the Vp = +30 V-hysteresis curve is clearly rectangular, whereas for Vp = −30 V the loop
is smooth and S-shaped. According to [32], these results indicate that a magnetic easy
axis is induced parallel to the magnetic field for Vp = +30 V whereas the x-direction
is magnetically harder for Vp = −30 V. This observation is in full agreement with the
theoretical predictions made in Section 5.1.2.
The absolute shift in the coercive field from the Vp = −30 V curve to the Vp = +30 V
curve is about 5 mT and also the magnetic remanence has clearly increased in the same
voltage range.
Comparison
In samples prepared by the cementing technique, the determination of the easy axis
orientation on the basis of the magnetic hysteresis loops at different applied voltages Vp
is difficult. In contrast, the evaporated samples permit a clear distinction between the
magnetic easy and hard loop. Moreover, the shift in the coercive field is more pronounced
in samples prepared by direct evaporation than in samples prepared by cementation.
52 Chapter 5 Magneto-Optics on Ferromagnetic/Ferroelectric Hybrids
-20 -10 0 10 20
-1.0
-0.5
0.0
0.5
1.0
M/M
s
µ0H (mT)
Vp=+30 V Vp=0 V Vp=-30 V
Vp
Ni
H||x
Figure 5.4: Hysteresis curve of Ni evaporated onto actuator at different applied voltages (sample
M090713A).
Thus, the magneto-elastic coupling must be a lot more efficient and homogeneous in the
evaporated samples, meaning that the strain exerted on the ferromagnetic film is larger.
In order to confirm these findings the evolution of the magnetic microstructure of both
types of samples is investigated in the following section.
5.1.4 Domain Evolution at Constant Strain
So far, the analysis was restricted to the voltage dependence of the magnetic hysteresis
loops, which have been obtained by integration over all camera pixels. Thus information
about local magnetic substructure was neglected. However, Kerr imaging allows to di-
rectly observe processes in the local magnetic microstructure of thin films. In order to get
a more detailed understanding about the strain dependence of ferromagnetic hysteresis,
the domain evolution of the ferromagnetic thin films will be investigated for different
voltages applied at the actuator.
Ferromagnetic Films Cemented onto Piezoelectric Actuator
This paragraph deals with the domain evolution at different strain states of ferromagnets
cemented onto actuators. In particular, results obtained with sample B080623Ad will be
shown. The corresponding M(H) curve was presented in Fig. 5.3(b).
At first, the magnetic field was prepared again at 0H = −60 mT before it was swept
to 0H = 60 mT. At each field point a Kerr image was recorded and the difference
imaging technique was applied to enhance the magnetic contrast (cf. Chapter 4). The
domain evolution of Fe50Co50 for Vp = +150 V and Vp = −30 V is depicted in Fig. 5.5
5.1 Ferromagnetic Thin Films on Piezoelectric Actuators 53
1 mm
µ0H = 13.9 mT µ0H = 16.4 mT µ0H = 17.9 mT
ROI
H||y
V p = +
150
VV p =
-30
V
(a) (b) (c)
(d) (e) (f)
Figure 5.5: Comparison of the domain evolution in Fe50Co50 at Vp = +150 V and Vp = −30 V at (a)/(d)
0H = 13.9 mT, (b)/(e) 0H = 16.4 mT, (c)/(f) 0H = 17.9 mT. Definition of a region of
interest (ROI) in (a) for quantitative evaluation (see text).
by means of six images at different magnetic field values. H was oriented along the y-
direction, which corresponds to the easy axis for positive Vp. The images in the upper
row (cf. Fig. 5.5(a),(b),(c)) were taken with an applied piezo-voltage of Vp = +150 V,
the ones in the bottom row at Vp = −30 V (cf. Fig. 5.5(d),(e),(f)). At an external
field of 0H = 13.9 mT domains have already nucleated and propagated at the left and
right edges of the sample simultaneously for both voltages (cf. Fig. 5.5(a) and (d)).
Going to larger magnetic fields, the domain pattern starts to differ for the different strain
states. At 0H = 16.4 mT the magnetic microstructure for Vp = +150 V (cf. Fig. 5.5(b))
exhibits basically the same characteristics as in Fig. 5.5(a). However, for Vp = −30 V,
the magnetization in center regions of the sample has switched (cf. Fig. 5.5(e)). At
even larger fields (0H = 17.9 mT) the magnetization has switched almost completely
for Vp = −30 V and the sample is uniformly magnetized except for some small regions at
the edges (cf. Fig. 5.5(f)). In contrast, for Vp = +150 V, the domain wall motion is not
finished yet, since there is a region in the sample center with black domains. Thus, the
magnetization still points along the original direction (cf. Fig. 5.5(c)).
54 Chapter 5 Magneto-Optics on Ferromagnetic/Ferroelectric Hybrids
- 4 0 - 2 0 0 2 0 4 0- 1
0
1
H | | y
V p = - 3 0 V
w h o l es a m p l e
s a m p l ec e n t e r( R O I )M/
M s
µ0 H ( m T )(a)
- 4 0 - 2 0 0 2 0 4 0- 1
0
1
H | | y
V p = + 1 5 0 V
w h o l es a m p l e
s a m p l ec e n t e r( R O I )M/
M s
µ0 H ( m T )(b)
Figure 5.6: Comparison of the M(H) hysteresis loops evaluated for the sample center (red curve) and
the total sample (green curve) at (a) Vp = −30 V and (b) Vp = +150 V.
Obviously, strain influences the domain evolution as expected from the M(H) curves
in Fig. 5.3. However, a noticeable feature is the fact that at 0H = 13.9 mT the domain
patterns for Vp = +150 V and Vp = −30 V do not differ significantly, because in both
cases the edge regions show white domains. Going to larger magnetic fields, the domain
structure remains more or less unchanged over a magnetic field range of approximately
4 mT for Vp = +150 V. In contrast, for Vp = −30 V the magnetization switches within
the sample center in the same field range. This behavior indicates that the left and right
edge regions of the sample switch at a certain magnetic field regardless the applied strain.
It is only in the sample center where the magnetization switching differs for positive and
negative Vp. As a consequence, the strain transmission at the left and right edge regions
must be different from the strain transmission in the sample center.
To confirm this assumption, the sample center was evaluated separately in the M(H)
hysteresis measurement at Vp = +150 V and Vp = −30 V (cf. Fig. 5.3(b)) by defining a
region of interest (ROI) as depicted in Fig. 5.5(a). The resulting hysteresis curve can be
seen in Fig 5.6. For comparison, the hysteresis curve of the whole sample is also depicted.
The M(H) loop evaluated in the sample center clearly differs from the hysteresis curve of
the total sample for Vp = −30 V (cf. Fig 5.6(a)) as well as for Vp = +150 V (cf. Fig 5.6(b)).
This measurement result proves the assumption that strain is not distributed uniformly
in the ferromagnetic film. As mentioned in Section 5.1.1, reasons might be inhomogenous
strain transmission arising from imperfect pre-polishing of the actuator or non-uniform
epoxy distribution.
5.1 Ferromagnetic Thin Films on Piezoelectric Actuators 55
referenceimage
H||x (a) (b)
(c)
-40 -20 0 20 40
-1
0
1M
/Ms
µ0H (mT)
Vp=-30 V
Figure 5.7: Domain evolution in Ni at Vp = −30 V. The images are obtained at an external magnetic
field of (a) 0H = −9.7 mT, (b) 0H = 3.4 mT, (c) 0H = 16.9 mT. The magnetization
reversal is characterized by a coherent magnetization rotation.
Ferromagnetic Films Evaporated onto Piezoelectric Actuators
Figure 5.7 and Fig. 5.8 depict the domain evolution in Ni evaporated onto an actuator
(sample M090713A) at piezo-voltage of Vp = −30 V and Vp = +30 V, respectively. In both
cases H was aligned parallel to x and the measurement procedure was equal to the one
described above. Unfortunately this particular sample exhibits a rough surface which
arises from an imperfect polishing of the piezoelectric actuator prior to the evaporation
of the metal, although the most fine-grained abrasive paper was used. Nevertheless it is
shown here, since this sample yields good domain visibility during magnetization reversal.
Generally, the same results have been obtained with other evaporated samples.
Let us first consider the situation for the hard axis magnetization loop, i.e. for
Vp = −30 V (cf. Fig 5.7). The image color changes gradually from black to white across
the whole sample with increasing magnetic field as seen in Figs. 5.7(a),(b),(c). Hence,
the sample is uniformly magnetized throughout the process of magnetization rotation
56 Chapter 5 Magneto-Optics on Ferromagnetic/Ferroelectric Hybrids
-40 -20 0 20 40
-1
0
1
M/M
s
µ0H (mT)
Vp=+30 V
H||xH||xH||x (a)
(b)
(c)
(d)
referenceimage
Figure 5.8: Domain evolution in Ni at Vp = +30 V. The images were obtained at an external magnetic
field of (a) 0H = 6.6 mT, (b) 0H = 6.8 mT, (c) 0H = 6.9 mT, (d) 0H = 7.2 mT. The
magnetization reversal is dominated by domain nucleation.
5.1 Ferromagnetic Thin Films on Piezoelectric Actuators 57
and the magnetization reversal resembles a coherent magnetization rotation as predicted
for an ideal single domain state by Stoner-Wohlfarth theory [32, 95].
The situation is different for positive voltages applied to the actuator, e.g. Vp = +30 V
(cf. Fig 5.8). At 0H = 6.6 mT, domains start to nucleate in form of stripes distributed
uniformly within the sample (cf. Fig. 5.8(a)). Then, at larger magnetic fields, the white
domains propagate into the region between the “stripes” as depicted in Fig. 5.8(b),(c).
Finally the magnetization reversal is completed as depicted in Fig. 5.8(d). These Kerr
images prove, that domain nucleation and domain wall motion play an important role
in the process of magnetization reversal along the easy axis [5] in samples prepared by
direct evaporation.
(a) (b)
Vp
P
+-
FM
(c)
Figure 5.9: (a) Domain nucleation in Ni on actuator at Vp = 0 V. (b) Domain nucleation in Ni on
actuator at Vp = 0 V after the polarization of PZT was inverted. (c) Schematic illustration
of positive (red) and negative (blue) electrodes embedded in the piezoelectric actuator. In
(a), nucleation is primarily on top of the positive electrodes. In (b) nucleation on top
of negative electrodes dominates. Note that measurements were done at different sample
alignments.
The nucleation pattern observed in Fig. 5.8 for an applied piezo-voltage of Vp = +30 V
was examined more closely, since the regularity of the distance between the stripes was
remarkable. For all positive as well as for small negative applied voltages Vp (depending
on the remnant strain of the actuator) the striped characteristic was found. In Fig. 5.9(a),
the domain pattern for Vp = 0 V and H∥x is depicted. Domains start to nucleate on top of
every second metal electrode embedded in the piezoelectric actuator. This is evident from
Fig. 5.9(a) in which the electrodes are schematically represented by red and blue lines.
Moreover, in the positive polarization state of the PZT based actuator, those electrodes
are the positive ones (red color, cf. Fig. 5.9(c)), regardless of the history of the strain
state (cf. Fig. 5.9(c)). After domains have nucleated on top of the positive electrodes, a
nucleation center on top of the negative electrodes (blue color) can be found, although
this behavior is not as distinct. Thus, the region on top of the actuator-electrodes must
be the preferred center of nucleation. Imaging results at Vp = 0 V and H∥x after a
58 Chapter 5 Magneto-Optics on Ferromagnetic/Ferroelectric Hybrids
polarization reversal of the actuator can be seen in Fig. 5.9(b). Note that the images
were obtained after a readjustment and thus the location of red and blue electrodes in the
two neighboring images differs. This time the domain nucleation on top of the negative
electrodes dominates.
Several effects could in principle account for this nucleation pattern. First, intrinsic
regular surface roughness of the actuator could induce domain nucleation centers. How-
ever, optical microscopy and scanning electron microscopy measurements of untreated
piezoelectric actuator stacks did not reveal any surface roughness or intrinsic imperfec-
tions which are related to the surface region of the electrodes. Therefore, inherent sur-
face effects attributing to the actuator can be excluded. Second, electric stray fields
generated by the electrodes can be ruled out since the stripes appear at Vp = 0 V.
FMPMMA
piezoelectricactuator (side view)
relaxed
deformed(Vp>0)
PZT
electrodes
y
z
x
Figure 5.10: Schematic illustration of the actua-
tor, PMMA and FM film deforma-
tion.
Finally, the most promising explanation is
that the strain induces surface imperfections
in the PMMA buffer layer as schematically
illustrated in Fig. 5.10. An elongation of the
piezoelectric actuator along y for Vp > 0 is
accompanied by a contraction along x and
z governed by the Poisson ratio = 0.45
[65] of the PZT ceramics. The electrodes
embedded in the PZT ceramic are made of
platinum (Pt) [96]. Assuming perfect strain
coupling at the PZT-electrode interfaces, the
electrodes are comparably deformed along
the x- and z-direction. However, the corre-
sponding dilation of the electrodes along y is
determined by the Poisson ratio of platinum
= 0.38 [97], which is slightly smaller than
the one of PZT. As a consequence, the rel-
ative elongation of the electrodes along y is slightly larger (cf. Eq. (5.1)). Thus, the
strain exerted onto the affixed PMMA buffer layer is not totally homogeneous resulting
in small deformations in the PMMA as schematically depicted in Fig. 5.10. These strain-
induced surface imperfections might trigger the preferential domain nucleation on top
of the actuator-electrodes. However, the asymmetry in the domain nucleation pattern
regarding the poling of the electrodes cannot be explained.
5.1.5 Strain Induced Birefringence
The effect of strain onto the magnetic hysteresis and onto the domain evolution has been
discussed so far. Hence, this and the following sections focus onto the voltage control
of the magnetization orientation at constant magnetic field. The obvious measurement
sequence to prove the possibility of magnetization control would be to record Kerr im-
5.1 Ferromagnetic Thin Films on Piezoelectric Actuators 59
ages as a function of the applied piezo-voltage at constant magnetic bias field [92]. If
the magnetization switched by means of the voltage change, the strain control of the
magnetization would be successful.
A
B
µ0HVp,A
-60 mT
Vp
Vp,B
Figure 5.11: Sequence for refer-
ence measurement.
However, at first a reference measurement was done to
rule out strain induced contrast changes in the Kerr im-
ages. The measurement sequence is schematically depicted
in Fig. 5.11. The magnetic field was set to 0H = −60 mT in
order to assure a magnetic saturated state since H >> Hc.
Then the piezo-voltage was cycled starting from Vp,A to Vp,B
at constant magnetic field. Due to the large external mag-
netic field, the voltage sweep should have no influence onto
the state of magnetization because in this regime the Zee-
man term dominates over the magnetoelastic contribution
to the free energy (cf. Chapter 2). For Kerr microscopy, a
reference image was recorded in the beginning at Vp,A and
then the difference imaging technique (cf. Chapter 4) was
applied.
The result of the reference experiment can be seen in Fig. 5.12. In (a) a Fe50Co50 film
cemented onto the actuator is depicted. The image was obtained at Vp,B = −30 V after a
voltage sweep from Vp,A = +120 V. A rectangular area in the sample center is certainly
darker than the rest of the sample. Only beneath this region of the Fe50Co50 sample the
strain was applied, whereas beneath the left and the right edges no strain was exerted
(see Chapter 6 for further details on local strain generation). Obviously, the variation
of strain caused changes in the contrast pattern of the Kerr images. This effect cannot
be attributed to a magnetization rotation, because the ferromagnetic film is uniformly
magnetized along the external magnetic field. Furthermore, one can observe ripples and
speckle in the image which can be attributed to surface unevennesses generated by the
application of the local strain.
Ferromagnetic films evaporated onto actuators show a similar behavior (cf. Fig 5.12(b)).
This time, strain was applied to the whole ferromagnetic film and the voltage at the ac-
tuator was cycled from Vp,A = +60 V → Vp,B = −30 V. A pattern of regular horizontal
stripes can be clearly seen. These stripes can again be identified as the electrodes embed-
ded in the actuator beneath the ferromagnetic film (cf. 5.1.3). Once more, every second
electrode dominates the pattern.
Contrast changes in the Kerr images were recorded upon sweeping Vp, although the fer-
romagnetic film was saturated. Thus, the detected contrast patterns cannot be attributed
to magnetoelastic effects. It is the applied strain that induces the image patterns as ob-
served in Fig. 5.12. Two aspects may explain this phenomenon. First, the applied stress
might induce surface effects which alter the intensity of the reflected light. A variation
of light intensity changes of course the contrast in the difference imaging technique and
60 Chapter 5 Magneto-Optics on Ferromagnetic/Ferroelectric Hybrids
1mm H||y
µ0H=-60 mT
(a)
1mm H||x
µ0H=-60 mT
(b)
Figure 5.12: Visualization of strain induced birefringence in (a) Fe50Co50 cemented onto actuator
(sample B080623A) and (b) Ni evaporated onto actuator (sample M090213P). Measure-
ments were done at constant magnetic field 0H = −60 mT. Vp was cycled from (a)
Vp,A = −30 V → Vp,B = +120 V and from (b) Vp,A = +60 V → Vp,B = −30 V, respectively.
The images correspond to “Vp,B − Vp,A” in both cases.
thus might explain the observed contrast patterns. Second, stress induces variations
of the refractive indices of the MgO or PMMA buffer layer [98, 99, 100]. Assuming a
penetration depth of light comparable to the film thickness, the measured effect could
arise from strain induced birefringence in the MgO or the PMMA. Since the reference
images are recorded at different strain states than the images in Fig. 5.12, strain induced
birefringence might be responsible for the anomalous contrast patterns. In order to rule
out strain induced birefringence effects, equivalent measurements with a sample having
an intermediate buffer layer between the MgO/PMMA and the ferromagnetic film, for
instance gold, would be necessary. If the gold buffer layer had a sufficient thickness, no
birefringence in the MgO or PMMA would occur due to the limited penetration depth of
light (cf. Chapter 4). However, this experiment was not accomplished during this thesis.
All in all, it is evident that the applied stress induces changes in the contrast of the
recorded images. Regardless of the true origin of this effect, it will be denoted as stress
induced birefringence in the following. Thus, the measurement sequence proposed in
the beginning of this section cannot be used, since the magnetic Kerr rotation would be
superimposed by a stress induced signal. It would be impossible to draw conclusions about
the magnetic properties of the sample. Therefore, another method has been devised, as
described in the next Section 5.1.6.
5.1 Ferromagnetic Thin Films on Piezoelectric Actuators 61
- 4 0 - 2 0 0 2 0 4 0
- 1
0
1 V p = + 3 0 VV p = - 3 0 V
DBA
M/M s
µ0 H ( m T )
C
(a)
A B
CD
Sequence Sequence
~
C
B,B~A,A
µ0H
Vp
VpC
VpB
µ0H
Vp
VpC
VpB
(b)
Figure 5.13: Schematic illustration of the proposed measurement sequences in order to extract the mag-
netic effects contributing to birefringence. (a) Hysteresis loop with indicated measurement
points. (b) Illustration of sequence ◻ and sequence .
5.1.6 Extraction of the Magnetic Contribution to Birefringence
In order to extract the Kerr rotation originating from the magnetic nature of the sample,
another measurement sequence has been proposed. It is denoted as sequence ◻ and is
illustrated in Figs. 5.13(a),(b).
• A: Preparation of the sample at 0H = −60 mT and on a magnetically easy loop
(e.g. V Ap > 0 V for H∥x in Ni) .
• B: Sweep H to measurement field at constant voltage V Ap = V B
p . Usually the mea-
surement field yields H ≲Hc.
• C: Sweep Vp to desired value V Cp at constant magnetic field.
• D: Sweep magnetic field back to start value 0H = −60 mT at constant piezo-voltage
V Cp = V D
p .
• Image subtraction: Subtract image obtained at point B from the image obtained at
point A (B−A) and the image obtained at point D from image at point C (C−D).
Applying this measurement sequence, the image subtraction is always among images of
equal strain state. Therefore, the resulting contrast in the difference images contains only
information of magnetic origin. This method allows the investigation of strain induced
magnetization reversal by means of Kerr microscopy.
In order to check for reversible magnetization switching sequence was proposed as
sketched in Fig. 5.13(b). After the sweep of Vp to the desired value V Cp , it is cycled back
to the original value on the easy axis loop (denoted as V Bp ). In the end the difference
images B −A and B − A are generated and compared.
62 Chapter 5 Magneto-Optics on Ferromagnetic/Ferroelectric Hybrids
5.1.7 Strain Induced Magnetization Switching
Having introduced sequence ◻ and sequence the magnetization switching in ferromag-
netic films upon changing the voltage at the actuator is analyzed. At first cemented
samples are studied before turning to the evaporated samples.
Magnetization Switching in Ferromagnetic Films Cemented onto Actuators
(a) (b)
(c) (d)
1 mm H||y
Vp=0 V
Vp=-30 V Vp=+150 V
Vp=+150 V
Figure 5.14: Magnetization switching in Fe50Co50 (sample B080622Ad) cemented onto actuator by
changing the applied voltage Vp at 0H = 18 mT. The difference images (a) B − A at
V Bp = +150 V, (b) C−D at V C
p = 0 V, (c) C−D at V Cp = −30 V and (d) B−A at V B
p = +150 V
are displayed. The corresponding parts of sequence ◻ and are orange-colored.
In principle all prepared samples have shown the same characteristics and therefore
only the results of the Fe50Co50 sample B080623Ad are presented. Preparation of the
sample was at 0H = −60 mT and V Ap = +150 V and the external magnetic field was
aligned along the y-axis, which corresponds to the easy axis. The measurement field was
5.1 Ferromagnetic Thin Films on Piezoelectric Actuators 63
chosen to be 0H = 18 mT, because it is slightly below the coercive field (cf. Fig. 5.3(b)).
Figure 5.14(a) displays the difference image B−A of the Fe50Co50 sample at V Bp = +150 V.
At the left and right sample edges the switching of M has already taken place, whereas
in the sample center M still points into its original direction. Thus the magnetization
switching in those areas has to be attributed to the magnetic field sweep from point A
to point B. As discussed above (cf. Section 5.1.4), the strain transmission at the edge
regions of the sample can be neglected anyways. In Fig. 5.14(b) the difference image
C−D at V Cp = 0 V can be seen. The magnetization has been switched also in the sample
center except for a region on the upper edge. A complete magnetization switching has
been achieved after a voltage sweep to V Cp = −30 V as depicted in Fig. 5.14(c). The
Fe50Co50 film is uniformly magnetized since only white domains are visible. Therefore,
the magnetization in the sample center was switched by means of the applied voltage
only.
Moreover, sequence was applied to check the reversibility of the magnetization switch-
ing. Therefore, the voltage was cycled V Bp = +150 V → V C
p = −30 V → V Bp = +150 V still at
the same magnetic field. The difference image B−A is depicted in Fig. 5.14(d). Evidently,
the original magnetization configuration seen in Fig. 5.14(a) cannot be reproduced with
a second voltage cycle. This behavior indicates, that the magnetization switching is not
reversible. However, a more quantitative analysis presented in Chapter 6 will reveal, that
reversible magnetization switching is indeed partly possible.
Magnetization Switching in Ferromagnetic Films Evaporated onto Actuators
In Fig. 5.15 the voltage control of M in evaporated ferromagnetic films (sample M090713P)
is depicted. The magnetic field was applied along the x-direction and the images were
obtained at a measurement field of 0H = 8 mT, which is slightly below the corresponding
coercive field for Vp = +30 V (cf. Fig. 5.4). The image B −A (cf. Fig. 5.15(a)) contains
only black domains indicating that no domain nucleation has taken place yet. Sweeping
the voltage from V Bp = +30 V to V C
p = 0 V changes the magnetic state. The resulting
magnetic microstructure is visible in Fig. 5.15(b), which is the corresponding difference
image C − D. Domain nucleation has started on top of the positive piezo-electrodes. A
consecutive measurement sequence with a voltage cycle to V Cp = −8 V yields the image
C − D depicted in Fig. 5.15(c). The stripe pattern is more distinct and domains have
started to propagate into the region in between the stripes. Going to V Cp = −30 V as
depicted in Fig. 5.15(d) the magnetization has switched completely and the sample is
uniformly magnetized again. Thus magnetization switching within the whole thin film
by means of a variation in the piezo-voltage has been achieved.
The reversibility of this process was also checked by applying sequence . However, in
analogy to the cemented samples a reversible control could not be detected and therefore
the presentation of the corresponding Kerr image is omitted.
64 Chapter 5 Magneto-Optics on Ferromagnetic/Ferroelectric Hybrids
1 mm H||x
Vp=+30 VB
(a)
Vp=0 VC
(b)
Vp=-8 VC
(c)
Vp=-30 VC
(d)
Figure 5.15: Magnetization switching in Ni (sample M090213P) evaporated onto actuator by changing
the applied voltage Vp at 0H = 8 mT . (a) Difference image B-A at V Bp = +30 V; (b)
Difference image C-D at V Cp = 0 V; (c) Difference image C-D at V C
p = −8 V; (d) Difference
image C-D at V Cp = −30 V.
Discussion
To sum up, voltage-induced magnetization switching has been reported for ferromagnetic
films cemented onto actuators as well as for evaporated films in agreement with results
from [92]. Furthermore, it became evident, that the process of magnetization switching
is accompanied by domain nucleation in samples prepared by the evaporation technique.
However, reversible control of the magnetization orientation could not be confirmed by a
qualitative analysis of the Kerr images. A more quantitative study presented in Chapter 6
will demonstrate, that reversible M control is indeed possible depending on the external
magnetic field.
5.2 Ferromagnetic Thin Films on BaTiO3 Substrates
In the previous section, hybrid structures based on ferromagnetic thin films and piezo-
electric actuators have been studied. In these hybrids, MgO or PMMA buffer layers
separate the ferroelectric and the ferromagnetic material. The coupling between the two
phases is governed by the magneto-elastic effect (cf. Chapter 2). This section deals with
another multiferroic heterostructure consisting of a BaTiO3 substrate and a ferromag-
netic thin film. Thus, a direct phase boundary between both constituents is present. The
coupling between the ferromagnet and the ferroelectric is attributed to the converse mag-
netoelectric effect - the effect of an electric field onto the magnetization (cf. Chapter 2).
However, in heterostructures, the converse magnetoelectric effect is primarily caused by
strain coupling at the interface [15, 101, 102, 103, 104]. In essence, there are mainly two
contributions to the internal stress at the interface [34]: First, the conventional piezo-
electric effect which describes the macroscopic deformation of the ferroelectric phase as
5.2 Ferromagnetic Thin Films on BaTiO3 Substrates 65
a function of the electric field. Second, considerable local strains arise from changes in
the ferroelectric domain structure. As explained in Chapter 3, a- and c- domains coexist
in a BaTiO3 single crystal at room temperature. Since the lattice constants a and c are
not equal, changes in the ferroelectric domain structure cause strains in the clamped fer-
romagnetic film [105]. A comparison of both contributions [34] revealed that the strain
induced by domain reorientation in BaTiO3 dominates over the strain induced by the
linear piezoelectric effect. In addition to the sources of strain, the magnetoelastic effect
in the ferromagnetic constituent is fundamental for the strain coupling at the interface
of the ferromagnet/BaTiO3 heterostructure.
The following treatment and analysis of the ferromagnet/BaTiO3 composite struc-
ture are in analogue to the ferromagnet/actuator samples above. The magnetoelectric
coupling in those bilayers is investigated using the magneto-optical Kerr effect. At first,
characteristic features of the sample are pointed out before the effect of an applied electric
field onto the magnetic hysteresis and the corresponding domain evolution is discussed.
Finally, magnetization control by means of electric fields is studied.
5.2.1 Sample Characteristics
Au
Ni
E
VBTO
BaTiO3
(a)
- 4 0 0 - 2 0 0 0 2 0 0 4 0 0- 6 0
- 3 0
0
3 0
6 0
I (nA)
E ( V / m m )(b)
Figure 5.16: (a)Schematic illustration of the Ni/BaTiO3 sample setup allowing the application of out-
of-plane electric fields [34]. (b) I(E) curve of Ni on BaTiO3 with out-of-plane oriented
electric field.
The samples were fabricated by Stephan Geprags and consisted of an approximately
500 µm thick BaTiO3 single crystal substrate on top of which a polycrystalline ferromag-
netic film with a thickness of 50 nm was deposited by electron beam evaporation. Ni
and Fe50Co50 were chosen as ferromagnetic constituents again. However, the following
analysis is restricted to the results of one Ni/BaTiO3 sample (NB5), since all investigated
samples showed similar characteristics. The lateral dimensions of this particular sample
66 Chapter 5 Magneto-Optics on Ferromagnetic/Ferroelectric Hybrids
are 5×5 mm2. The substrate side of the sample is covered by a sputtered gold layer which
was finally attached to a copper piece using silver glue (cf. Fig. 5.16(a)). In order to
control the polarization of the BaTiO3 substrate, an electric field along the out-of-plane
direction (001) was applied. The gold layer and the polycrystalline Ni film served as
electrodes. In this configuration, the application of a large electric field E > Ec ideally
results in a uniform ferroelectric state only consisting of c-domains (cf Chapter 3).
In Fig. 5.16(b) the resulting I(E) curve for an out-of-plane oriented electric field is
depicted. At E ≈ +50 V/mm and at E ≈ +210 V/mm the curve exhibits peaks whereas
at E ≈ −80 V/mm and at E ≈ −160 V/mm minima are present. These extrema can
be referred to as the displacement currents indicating the polarization reversal at the
coercive field of the ferroelectric BaTiO3. Evidently two positive as well as two negative
extrema are observed. This unusual behavior is attributed to the imperfect substrate
quality. Grains and small cracks in the single crystal BaTiO3 substrate can be seen by
the naked eye. Therefore, the electric field distribution within the BaTiO3 might be
inhomogeneous which can explain the observed data. The asymmetry of the observed
coercive fields can be attributed to the use of different electrode materials [45]. For the
magneto-optical investigation the sample was mounted such that a 2 × 2 mm2 large and
crack-free section of the surface was imaged onto the CCD camera.
5.2.2 M(H) Measurements at Constant Strain
Following the procedure of Section 5.1.3 the M(H) hysteresis curves for different applied
electric fields were recorded. A corresponding analysis of Ni on BaTiO3 can be found in
[45]. As above, the magnetic field was swept and the voltage at the BaTiO3 substrate
VBTO remained constant. The Kerr signal of a 2 × 2 mm2 section of the sample was
monitored by integrating over all camera pixels.
In Fig. 5.17 the resulting M(H) loops for three different applied electric fields,
E = −60,0,+400 V/mm, can be seen. Evidently, the coercive field is shifted by approxi-
mately 1.5 mT from the E = −60 V/mm curve to the E = +400 V/mm curve [34, 45]. Also
the remanence increases with growing electric field. These results resemble strongly the
behavior of the Ni/piezoactuator samples. However, the distinction between the magnet-
ically easy and hard axis upon application of strain is not as unique as in the evaporated
ferromagnet/actuator hybrids (cf. Fig. 5.4). As mentioned, the strain is caused by a
combination of linear piezoelectric effect and changes in the domain structure of the fer-
roelectric BaTiO3 substrate. As a consequence, strain coupling at the interface of the
Ni/BaTiO3 composite results in changes of the magnetic anisotropy in the Ni film.
In accordance to Section 5.1.4, the magnetic substructure during the magnetization
reversal was investigated. In Fig. 5.18, domain images at E = +400 V/mm are de-
picted. A few scratches on the sample surface which arise from sample cleaning are
evident. Fig. 5.18(a) depicts the domain pattern at 0H = 8.3 mT. The magnetization
switching has already started at the upper left sample edge. Going to larger magnetic
5.2 Ferromagnetic Thin Films on BaTiO3 Substrates 67
- 1 0 0 1 0- 1
0
1
E = 0 V / m m
E = - 6 0 V / m mM/
M s
µ0 H ( m T )
E = + 4 0 0 V / m m
Figure 5.17: Hysteresis curve of Ni on BaTiO3 for E = −60,0,+400 V/mm. Remanence and coercivity
increase with growing electric field.
fields, 0H = 8.6 mT, causes the domain walls to propagate into the sample center (cf.
Fig. 5.18(b)). The characteristic feature of the domain boundaries is their zig-zag struc-
ture as already mentioned in Chapter 4 [5]. Also Fig. 5.18(c) displays a similar domain
wall structure at 0H = 9.0 mT.
In contrast, the images displayed in Fig. 5.19 exhibit a very rough surface pattern in
form of vertical stripes. The origin of this behavior is the multi-domain state of the
BaTiO3 substrate at E = −60 V/mm . Referring to Fig. 3.2, the tilting of the a-domains
with respect to the c-domains results in surface unevenness. One would expect, that
the difference imaging technique averages out these surface effects. However, as already
mentioned in Chapter 4, small fluctuations in the light intensity or the components are
1 mm H
(a) (b) (c)
Figure 5.18: Domain evolution in Ni on BaTiO3 at E = +400 V/mm. The images are obtained at an
external magnetic field of (a) 0H = 8.3 mT, (b) 0H = 8.6 mT, (c) 0H = 9.0 mT.
68 Chapter 5 Magneto-Optics on Ferromagnetic/Ferroelectric Hybrids
(a) (b)
Figure 5.19: Kerr images of Ni on BaTiO3 at E = −60 V/mm. The images are obtained at an external
magnetic field of (a) 0H = −7.0 mT and (b) 0H = −7.9 mT. The origin of the surface
unevenness can be attributed to the multi-domain structure of the BaTiO3 substrate.
always present and thus cause imperfections in the difference imaging technique. As a
consequence, domain visibility is strongly suppressed in this case. Fig. 5.19(a) depicts
the image at 0H = −7.0 mT in comparison to the image at 0H = −7.9 mT depicted
in Fig. 5.19(b). Although a change in the overall contrast is present, a distinct domain
pattern cannot be determined.
5.2.3 Magnetization Switching at Constant Magnetic Field
Up to now, the electric field dependence of the M(H) curve has been studied. The more
interesting question is however, if the magnetization can be controlled by means of the ap-
plied out-of-plane electric field only. In Section 5.1.7 the successful voltage control of the
magnetization in a ferromagnet/actuator hybrid structure was demonstrated. Following
this approach the bilayer sample consisting of Ni/BaTiO3 was studied for its capability
in magnetization control via electric fields. Therefore, sequence ◻ and (cf. Fig. 5.13)
were applied again to make sure that only magnetic contrast is visible.
The sample was prepared at 0H = −100 mT and EA = +400 V/mm. Then the mag-
netic field was swept to the measurement field at point B, which was slightly below the
coercive field (cf. Fig. 5.17). At constant magnetic field the electric field was cycled
to EC = −60 V/mm. In the end, the difference images B − A and C − D as defined in
Section 5.1.6 are compared.
In Fig. 5.20 the corresponding results for a magnetic field of 0H = 7.3 mT at point B
and C can be seen. The difference image B−A (cf. Fig. 5.20(a)) is uniformly black. This
is to be expected as Vp = const and H <Hc. The difference image C−D after the voltage
sweep to EC = −60 V/mm is depicted in Fig. 5.20(b). As above, the surface roughness
induced by the domains of the BaTiO3 substrate dominates the image. But apart from
that, also domain nucleation centers have been generated by the electric field sweep. In
5.2 Ferromagnetic Thin Films on BaTiO3 Substrates 69
EB =+400 V/mm
(a)
EC =-60 V/mm
(b)
EB =+400 V/mm
(c)
Figure 5.20: Magnetization switching in Ni on BaTiO3 by changing the out-of-plane electric field at
0H = 7.3 mT. (a) Difference image B−A at EB = +400 V/mm. (b) Difference image C−D
at EC = −60 V/mm. (c) Difference image B −B at EB = +400 V/mm.
Fig. 5.20(c), the difference image B − B acquired in a consecutive sequence is shown.
The image does not display any surface inhomogeneity because the substrate is in the
single domain state at EB = +400 V/mm. But the most striking feature is the fact that
the second electric field sweep from EC = −60 V/mm at C back to EB = +400 V/mm
at B has enhanced the magnetization switching, as the difference image B −B is mainly
composed of white domains. Consequently the magnetization switching at this magnetic
field is not reversible at all, since the initial domain state cannot be re-established. In
contrast, further domain wall propagation is induced by the second electric field cycle.
EB =+400 V/mm
(a)
EC =-60 V/mm
(b)
EB =+400 V/mm
(c)
Figure 5.21: Magnetization switching in Ni on BaTiO3 by changing the out-of-plane electric field at
0H = 8.0 mT. The difference images (a) B − A at EB = +400 V/mm, (b) C − D at
EC = −60 V/mm and (c) B −B at EB = +400 V/mm are depicted. The magnetization has
been switched upon the electric field cycle exclusively.
Accordingly Fig. 5.21 displays the results for a measurement field of 0H = 8.0 mT at
point B and C. In analogue, the difference image B − A (cf. Fig. 5.21(a)) is uniformly
70 Chapter 5 Magneto-Optics on Ferromagnetic/Ferroelectric Hybrids
black and thus no magnetization reorientation has taken place yet. After an electric field
sweep from EB = +400 V/mm to EC = −60 V/mm, the resulting magnetic state contains
almost only white domains as can be seen in Fig. 5.21(b), which depicts the difference
image C−D. Therefore, a complete magnetization switching induced by the electric field
sweep has been achieved. In a consecutive measurement, sequence was applied to check
the reversibility of the magnetization switching. In agreement to the measurement above,
no reversible M control is detected in the difference image B − B (cf. Fig. 5.21(c)). In
contrast, the resulting magnetic state consists only of white domains, whereas the initial
state has black domains. Thus the magnetization was switched completely by means of
the electric field sweep in the investigated sample sector, however, no reversible M control
can be achieved.
In a nutshell, the above experiments prove, that magnetization control by means of an
electric field is also possible in multiferroic heterostructures consisting of Ni and BaTiO3.
The converse magnetoelectric effect enables to switch the magnetic state at magnetic
fields close to the coercive field of the ferromagnetic constituent. No reversible M control
has been found, in contrast, the second voltage cycle applied in sequence enhances the
magnetization switching.
5.3 Summary
All in all, electric field control of the magnetization has been imaged by means of spa-
tially resolved Kerr effect. Multifunctional hybrid structures consisting of ferromag-
net/piezoactuator have been prepared and investigated. It has been reported, that the
magnetization of the ferromagnetic thin film can be switched upon the application of a
piezo-voltage and the observed data is in agreement with the findings of [92]. In addition,
electric field controlled ferromagnetism was also established in multiferroic heterostruc-
tures consisting of ferromagnetic thin films and piezoelectric BaTiO3 substrates.
Chapter 6
Local Magnetization Control
In the previous chapter the potential of electric field controlled ferromagnetism in mul-
tiferroic hybrid structures was demonstrated. By means of Kerr microscopy the mag-
netization switching induced by applied electric fields was visualized. However, the re-
sults shown were all obtained in mm-scale “large” samples. To be relevant for eventual
technological application, scaling to smaller dimensions is mandatory. Therefore, it is
necessary to demonstrate a local control of the magnetization. Several concepts based
on magnetostriction have been proposed in order to aim for local magnetization con-
trol [106, 107, 108]. Usually they combine electrostrictive grids with magnetostrictive
materials to propose a voltage controlled non-volatile magnetic random access memory.
However, an effective coupling of ferromagnetic and piezoelectric properties in hybrid
structures at small scales is the key point to achieve local magnetization control. There-
fore, two architectures for ferromagnetic/piezoelectric heterostructures are proposed and
investigated in this chapter. Magneto-optic methods are again applied in order to charac-
terize them. In the first part of this chapter locally controllable hybrid structures based
upon the piezoelectric actuators introduced in Chapter 5 are discussed. Their potential
for local magnetization switching is presented. Moreover, results of magnetotransport
measurements will be shown and correlated to the magneto-optical data. After that, a
novel design allowing local magnetization control is established. This architecture re-
lies on micron-sized electrodes that are patterned onto different piezoelectric substrates.
Local polarization and magnetization control in the novel hybrid structure is studied.
6.1 Ferromagnets on Piezoelectric Actuators
In Chapter 5 multifunctional hybrid structures consisting of a piezoelectric actuator and
a ferromagnetic thin film were introduced and their potential for magnetization control
via electric fields was demonstrated. These hybrids can be modified in order to allow for
local magnetization control. The concept and functionality of this approach is presented
in the following.
71
72 Chapter 6 Local Magnetization Control
FM
x
y
z
electrodes
electrodes
(a)
(b)
(c)
1 mm 5 mm
Figure 6.1: (a) Schematic illustration of a locally contacted actuator with affixed ferromagnetic film. (b)
Photograph of a ferromagnet on actuator with local contacts. (c) Assembly of the actuator
onto a L-shaped brass part mounted onto a chip carrier in order to allow for free expansion
and contraction of the actuator.
6.1.1 Sample Preparation
The hybrid structures enabling local magnetization control are based on the piezoelectric
actuators which have been introduced in Chapter 5 (cf. Fig. 5.1). As mentioned, the
actuator consists of 100 µm thick PZT piezo-active layers which are separated by 10 µm
thick metal electrodes. Those electrodes are connected by a conducting band on the
two sides of the actuator. Every second electrode is isolated by sintered glass to assure
alternating electric fields. In order to enable local control of the PZT active layers the
actuator has to be modified as can be seen schematically in Fig. 6.1(a). At first, the con-
ducting band electrode on the two actuator sides has to be removed by careful polishing.
The glass isolation, however, is maintained. Then the electrodes are contacted one by
one by bond wires as depicted in Fig. 6.1(b). This allows an independent control of each
PZT layer embedded by electrodes. The actuator with affixed ferromagnet is mounted
on a L-shaped brass part (cf. Fig. 6.1(c)) in order to assure free expansion and contrac-
tion. Finally the brass part is adjusted onto a chip carrier and thus the electrodes can
be connected one by one with the pins of the carrier to provide an independent voltage
supply of several actuator electrodes. Since the actuator consists of 20 electrode pairs,
usually only about five electrode pairs are contacted selectively since the number of bond
pads is limited. The remaining electrodes of each side face are connected to a common
contact pad. Thus the remaining part of the actuator can also be voltage-controlled and
6.1 Ferromagnets on Piezoelectric Actuators 73
Vp
P
A
(a)
- 8 0 0 - 4 0 0 0 4 0 0 8 0 0- 1 5 0
0
1 5 0
P
I (nA)
E p ( V / m m )
P
(b)
Figure 6.2: (a) Schematic illustration of the polarization control within one monolayer of PZT. (b)
Measurement of the I-Ep curve of a single PZT layer. The distinct peaks can be referred
to as the displacement current indicating a polarization reversal in the corresponding PZT
monolayer.
thus exert homogeneous stress.
The samples M090213P, B080623Ad and M090220Ac (cf. Tab. 5.1) were modified
accordingly to allow for local control of the piezoelectric actuator. On the basis of these
three samples the following measurement results were obtained.
6.1.2 Local Polarization Control
Having contacted PZT layers of the actuator locally, it has to be checked whether it is
possible to really control their polarization selectively without affecting adjacent PZT
monolayers. This characteristic is fundamental to also exert locally confined strain. Usu-
ally the actuator is operated in the semi-bipolar voltage range −30 V ≤ Vp ≤ +150 V.
In this voltage regime the inherent polarization P of the PZT monolayers is alternating
due to its particular contact scheme (cf. Fig. 6.2(a)). Contacting two adjacent elec-
trodes of the multilayer PZT actuator as schematically depicted in Fig. 6.2(a) enables
to control the polarization of the embedded PZT layer. The resulting I(Ep) curve can
be seen in Fig. 6.2(b). The voltage was cycled from Vp = 80 V → −80 V → +80 V
(−800 Vmm ≤ Ep ≤ +800 V
mm) and the current was recorded. Two distinct peaks at around
Ep = +420 Vmm and Ep = −420 V
mm , respectively, can be seen. Those can be referred to
as the displacement currents, which indicate a polarization reversal. Thus it is possible
to control the polarization of a single PZT layer. To check whether the polarization di-
rection of neighboring layers is affected by a reversal of P in the embedded layer, their
polarization was checked before and after the reversal again using the I(Ep) technique
just described. It turned out that the polarization direction of adjacent PZT layers is not
74 Chapter 6 Local Magnetization Control
x
y
Vp=+28 Vµ0H=-60 mT
1 mm(a)
Vp=-30 Vµ0H=-60 mT
(b)
Vp=+28 Vµ0H=-60 mT
(c)
Figure 6.3: Voltage cycle Vp = +30 V → −30 V → +30 V at selectively contacted actuator for 0H =
−60 mT (sample M090213P). A reference image was recorded at Vp = +30 V. Kerr images
at (a) Vp = +28 V, (b) Vp = −30 V (the red and blue lines indicate the positive and negative
electrodes) and (c) Vp = +28 V are depicted.
affected by a polarization reversal in a particular PZT monolayer. Thus the polarization
of each PZT layer can be controlled individually.
6.1.3 Local Strain Control
Apart from the polarization, the more important aspect is local strain control. In order
to examine local strain generation, strain induced birefringence was exploited. This
phenomenon was observed in Chapter 5 and it implicates contrast changes in the Kerr
images induced by surface or birefringence effects.
At first a sample with an evaporated Ni film was investigated (sample M090213P)
for local strain generation. Therefore, a voltage Vp was applied to two electrodes em-
bedding one PZT monolayer, whereas the remaining actuator electrodes were supplied
with a constant voltage Vp,r = +30 V. The two electrodes, to which Vp was applied, are
depicted as red and blue lines in Fig 6.3(b). Furthermore, a large magnetic field was
applied, 0H = −60 mT, to assure magnetic saturation of the ferromagnetic film affixed
to the locally contacted actuator. Then the voltage at the two electrodes was cycled
from Vp = +30 V to Vp = −30 V and back. A reference image was recorded in the begin-
ning at Vp = +30 V and the difference imaging technique (cf. Chapter 4) was applied. In
Fig. 6.3(a) the image obtained at Vp = +28 V (with reference image subtracted) is de-
picted. Since the strain state had not changed appreciably with respect to the reference
image, no contrast pattern is present. Contrarily, at Vp = −30 V (cf. Fig. 6.3(b)) a clear
contrast pattern in form of a stripe can be seen. Evidently, the stripe appears mainly
6.1 Ferromagnets on Piezoelectric Actuators 75
y
x
Vp=-20 Vµ0H=-60 mT
1 mm
(a)
Vp=+150 Vµ0H=-60 mT
(b)
Vp=-20 Vµ0H=-60 mT
(c)
Figure 6.4: Ferromagnetic film cemented on actuator undergoes a voltage cycle Vp = −30 V → 150 V →
−30 V at three locally contacted electrode pairs at 0H = −60 mT (sample B080623Ad).
The reference image was recorded at Vp = −25 V. (a) Vp = −20 V, (b) Vp = +150 V, the
blue lines indicate the electrodes connected to the common ground, the red lines represent
the sweeping electrodes and the electrodes represented by the white lines were on a constant
potential Vp,r = −30 V. (c) Vp = −20 V.
between the locally contacted actuator-electrodes to which Vp was applied. Therefore,
the contrast of the Kerr image has changed in the region on top of the embedded PZT
layer. This contrast pattern disappeared again after the back sweep to Vp = +28 V (cf.
Fig. 6.3(c)) indicating that the actuator is in the initial strain state again.
The same experiment has also been done with a ferromagnetic film cemented onto a
piezoelectric actuator (Fe50Co50, sample B080623Ad). This time, the contact scheme was
slightly different. Six adjacent PZT monolayers were exposed to Vp, the remaining ones
were supplied with a constant voltage Vp,r. The corresponding electrodes are represented
by the lines indicated in Fig. 6.4(b). The red electrodes were connected to the high output
of the source-meter and therefore they will be referred to as positive electrodes. The blue
ones will be denoted as negative electrodes. The white electrodes were supplied with a
constant voltage Vp,r = −30 V during the measurement. The voltage at the red electrodes
was cycled from Vp = −30 V → +150 V → −30 V . Once more, the external magnetic field
was sufficiently large (0H = −60 mT) in order to achieve a uniform magnetized state
throughout the measurement. The reference image was recorded at the beginning of the
cycle at Vp = −25 V and the difference imaging technique was applied. Fig. 6.4(a) was
obtained at Vp = −20 V and does not reveal any contrast pattern, since the strain state
was not altered significantly with respect to the reference state. In contrast, a distinct
contrast pattern is present in Fig. 6.4(b), depicting the difference image obtained at
Vp = +150 V. Evidently, there is a strain induced contrast pattern in the sample center,
which corresponds to the individually controlled PZT layers. Hence, locally confined
birefringence is induced by the applied strain. The contrast pattern vanishes again when
76 Chapter 6 Local Magnetization Control
H||y Vp=+150 VB
(a)
Vp=-30 VC
(b)
Vp=+150 VB
(c)
Figure 6.5: Local magnetization control in Fe50Co50 cemented onto actuator (sample B080623Ad). The
red and blue lines indicate the actuator-electrodes which were locally contacted. The mea-
surement was done applying sequence ◻ and at 0H = 18 mT. (a) Difference image B−A
at V Bp = +150 V, (b) C −D at V C
p = −30 V, (c) B − A at V Bp = +150 V.
going back to the initial strain state (cf. Fig. 6.4(c)). Thus local and reversible strain
variation is also possible in ferromagnets cemented onto piezoelectric actuators.
6.1.4 Local Magnetization Switching
In Chapter 5 the voltage controlled switching of M in a ferromagnetic film was demon-
strated. Having achieved local strain control by contacting the PZT layers individually,
local magnetization switching within small regions of the ferromagnet should be possible.
In the following experiments on the hybrid structures allowing local strain control are
presented. The measurement procedure and notation is identical to the one applied in
Chapter 5, where sequence ◻ and (cf. Fig. 5.13) were introduced in order to extract
the Kerr rotation.
Ferromagnet cemented onto actuator
Both prepared samples, B080623Ad and M090220Ac, yield similar results, and therefore
the following analysis is restricted to the results obtained on the Fe50Co50 based sample
(B080623Ad), which can be seen in Fig. 6.5. The magnetic field was oriented along the
y-axis, which corresponds to the easy axis for Vp > 0. The field and actuator preparation
was at 0H = −60 mT and at V Ap = +150 V, respectively. The actuator electrodes that
underwent the voltage cycle are depicted again as red lines in Fig. 6.5, whereas the blue
electrodes were connected to ground. The remaining part of the actuator was supplied
with a constant voltage of Vp,r = +150 V during the measurement sequence.
Then sequences ◻ and were applied in order to investigate the magnetization switch-
ing. In Fig. 6.5(a), the difference image B−A at V Bp = +150 V is depicted. The originally
6.1 Ferromagnets on Piezoelectric Actuators 77
black domains have already started to be switched into white domains due to the mag-
netic field sweep to 0H = 18 mT , especially at the sample edges. After the magnetic
field sweep the voltage was cycled at the denoted electrodes. The obtained difference
image C−D at V Cp = −30 V can be seen in Fig. 6.5(b). Further domain wall propagation
has taken place upon the voltage sweep. Evidently, the locally confined strain induced
a local magnetization switching. However, the application of local strain did not result
in complete magnetization switching, since there are still edge regions with the original
magnetic state. Also the magnetization switching was not only confined to the region
where the PZT layers are cycled but it also occurred also in adjacent parts of the ferro-
magnet. The reversibility of this switching process was measured applying sequence .
The corresponding difference image B− A can be seen in Fig. 6.5(c). As in Chapter 5, no
reversible magnetization switching is observed.
Ferromagnet evaporated onto actuator
H||x
Vp=+60 VB
(a)
Vp=-30 VC
(b)
Vp=+60 VB
(c)
Figure 6.6: Local magnetization control in Ni evaporated onto actuator (sample M090213P). The red
and blue lines indicate the locally contacted actuator-electrodes which embed four PZT
layers in total. Note that the nearest-next electrodes (yellow) were not contacted to prevent
direct contact of regions with opposing strain. The remaining electrodes were supplied with
Vp,r = +60 V. The measurement was done applying sequence ◻ and at 0H = 8 mT. (a)
B −A at V Bp = +60 V, (b) C −D at V C
p = −30 V, (c) B − A at V Bp = +60 V.
Nickel films evaporated onto actuators have also been studied for local magnetization
switching (sample M090213P). The number of PZT monolayers, that underwent a voltage
cycle, varied between two and six and the corresponding electrodes are again depicted as
insets in Fig. 6.6 and Fig. 6.7. Note that in this particular measurement the two nearest-
next electrodes (depicted as yellow lines) were not contacted, in order to prevent a direct
78 Chapter 6 Local Magnetization Control
contact of regions with opposing strain and thus possible sample damage. The remaining
electrodes were again supplied with a constant voltage Vp,r = +60 V, but they are omitted
for clarity. The magnetic field was aligned along x, which is the easy direction for Vp > 0 V.
Sequence ◻ and were applied again and the sample was prepared at 0H = −60 mT
and V Ap = +60 V before the magnetic field was swept to point B. The subsequent voltage
cycle yielded V Bp = +60 V → V C
p = −30 V.
Figure 6.6 displays the results of a measurement with four selectively contacted PZT
layers for 0H = 8 mT. The difference image B − A depicted in Fig. 6.6(a) reveals
that H < Hc, since no domain nucleation has occurred yet apart from a small region
near the image bottom. A voltage cycle at the particular electrodes to V Cp = −30 V
results in domain nucleation as visible in Fig. 6.6(b). Evidently, the domain nucleation
is limited to the region of the cycled electrodes, which are indicated by the red and blue
lines. Therefore, the local strain generation has successfully inverted the magnetization
direction without affecting the magnetic state of the remaining ferromagnetic film. As
already observed in Chapter 5, the dominating nucleation center is on top of the metal
electrodes. To check reversibility, sequence was also applied and the difference image
B−A was obtained (cf. Fig. 6.6(c)). The initial magnetic state could not be re-established
after a second voltage cycle.
To corroborate these findings, the same experiment was performed when two or re-
spectively six PZT layers were selectively contacted. The measurement configuration
and procedure was identical to the one above. For two contacted PZT layers, the imag-
ing results in terms of difference images B−A and C−D are depicted in Figs. 6.7(a),(b).
They were obtained at a magnetic field of 0H = 8.5 mT. Apart from a small region near
the image bottom no domains have nucleated in B − A. In contrast, in C − D domains
have nucleated on top of the selectively contacted red electrode due to the corresponding
voltage sweep. Also the region on top of the neighboring non-contacted electrodes (yel-
low) is partly affected and domain nucleation is visible, probably due to strain coupling
in the ferromagnetic film or the PMMA buffer layer. In the remaining ferromagnetic film
no further domain nucleation is observed. A similar behavior is found in Figs. 6.7(c),(d),
which display the difference images B − A and C − D for six controlled PZT layers at
0H = 7.0 mT. Local magnetization switching is observed upon the voltage sweep at the
selectively contacted electrodes.
Discussion
In summary, local magnetization switching was successfully achieved in both types of
hybrid structures. The magnetization switching in evaporated samples is more distinct
than in the samples prepared by cementing technique and also the local confinement of
the magnetization switching is superior. However, as already observed in Chapter 5,
the nucleation of domains happens preferably on top of the electrodes embedded in the
actuator underneath the ferromagnetic film.
6.1 Ferromagnets on Piezoelectric Actuators 79
(a)
H||x
(c)
(b)
(d)
Vp=-30 VC
Vp=+60 VB
Figure 6.7: Local M control in Ni for different contact schemes. (a),(b) Difference images B − A and
C−D for voltages V Bp = +60 V → V C
p = −30 V at 0H = 8.5 mT with two selectively contacted
PZT layers. (c),(d) Difference images B−A and C−D for voltages V Bp = +60 V → V C
p = −30 V
at 0H = 7.0 mT with six selectively contacted PZT layers.
6.1.5 Magnetotransport Measurements
In the previous section the possibility of local magnetization control in hybrid systems
was demonstrated by means of Kerr microscopy. In order to corroborate these findings,
magnetotransport (MTR) measurements were applied to this ferromagnetic/piezoelectric
hybrid system. In the following a short introduction into the theory of the anisotropic
magneto resistance will be given before dealing with MTR measurements on integrally
and locally strained Ni thin films.
Anisotropic Magneto Resistance
The anisotropic magneto resistance (AMR) is a phenomenon which implies that the
resistance of a ferromagnet depends on the orientation of the magnetization with respect
80 Chapter 6 Local Magnetization Control
Vp
IV
H
y
xVp
I V
H
(a) (b) (c)I||H I H
2 mm
Figure 6.8: (a) Photograph of the ferromagnetic/piezoelectric hybrid system with contacts to allow four-
point MTR measurements. Schematic illustration of the measurement configuration and
contact contact scheme of the (b) I∥H and (c) I ⊥H orientation.
to the electric current. In thin ferromagnetic films, the magnetization generally lies within
the film plane (cf. Chapter2). The resistivity for a current flowing perpendicular to the
magnetization is given by ⊥ which differs from ∥ denoting the resistivity of a current
parallel to the magnetization [109]. In 3d transition elements and alloys this anisotropy
originates from the non spherical-symmetrical charge distribution of the outer atom shell
[109]. Consequently, spin-orbit coupling between the conducting electrons and the shell
electrons is asymmetric with respect to the orientation of the electrical current. All in
all, the result is an anisotropic scattering of the conduction electrons, which depends
on the orientation of current with respect to the magnetization direction. Whenever the
material is composed of several magnetic domains, this is taken into account by averaging
over all magnetization orientations. However, spin scattering occurring at domain walls
[110] will not be accounted for.
Assuming a current density j that encloses an angle of with the magnetization M,
the longitudinal and transversal resistivity (long and trans) in polycrystalline samples
can be expressed as follows [111]:
long = ⊥ + (∥ − ⊥) cos2( ) (6.1)
trans = (∥ − ⊥) sin cos (6.2)
In general, the resistivities fulfill the relation ∥ > ⊥, which holds also true for Ni
with ∥ − ⊥ = 0.16 µΩcm [109]. All in all, MTR measurements possess the potential
to make quantitative statements about the magnetization orientation. Therefore, this
technique was applied to the prepared hybrids in order to confirm the data obtained by
the magneto-optical analysis.
In Fig. 6.8 the contact scheme allowing for MTR measurements on a Ni/actuator hy-
6.1 Ferromagnets on Piezoelectric Actuators 81
brid structure is depicted. The Ni film was contacted on top with four contacts (cf.
Fig. 6.8(a)), whereas the actuator was prepared as described above. This contact scheme
enables a four-point resistance measurement. In Fig. 6.8(b), the current is oriented par-
allel to the external magnetic field. As the current is driven between two contacts, the
voltage is measured at the two remaining ones. In contrast, the current is oriented per-
pendicular to H in Fig. 6.8(c). During the MTR measurements a constant bias current
of Ibias = 10 mA was set and the voltage Vlong was recorded. The geometry of the present
four-point AMR configuration is not well enough defined to allow for a quantitative mea-
surement of . Therefore, the resistance Rlong =VlongIbias
is used to characterize the measure-
ments. It is assumed that the geometry of the current flow remains constant regardless
the applied strain = const ⋅R and thus Eq. (6.1) remains valid and can be expressed in
terms of Rlong, R∥ and R⊥. MTR measurements on a micro-patterned Hall-bar would be
the ultimate solution to imply a defined geometry, but this was not achieved during this
thesis. The transversal resistance Rtrans was not recorded. Moreover, the dominant axis
of elongation of the piezoelectric actuator (y-axis) was always oriented perpendicular to
H as depicted in Figs. 6.8(b),(c). Thus the easy axis pointed along x for Vp > 0 V.
Magnetotransport and Integral Strain
Having briefly explained the AMR theory and the sample preparation, measurement
results will be shown in the following. At first integral strain is applied to the Ni film
as described in Chapter 5. In Fig. 6.9(a) results with current aligned parallel to the
external field can be seen, whereas Fig. 6.9(b) displays the result of I ⊥H. In both cases,
the voltage Vlong was recorded in dependence of the applied magnetic field for different
applied voltages Vp. The obtained results were in accordance with [112].
At first, the behavior in magnetic saturation (0H ≈ ±60 mT) for both current orien-
tations is analyzed, which directly reflects the characteristic of the AMR. The measured
resistance for I∥H is about Rlong ≈ 403 mΩ, whereas for I ⊥ H the resistance yields
Rlong ≈ 355 mΩ. This difference can be easily understood by AMR theory. In magnetic
saturation, the magnetization is aligned along the external magnetic field. Thus, for I∥H,
M is aligned parallel to I, but for I ⊥H, M is perpendicular to I. Therefore, the observed
resistance difference can be understood, since R∥ > R⊥.
In both current configurations the resistance curves for different applied voltages Vp
do not overlap for H >> Hc. The origin of this behavior cannot be the AMR effect
since the magnetization is saturated and points along the external field. Drift or offset
effects can also be excluded, since this phenomenon was observed in several measurements
and analogous in [112]. The magnitude of this shift in resistance also depends on the
orientation of the current. For I∥H, the change in the resistance at 0H = ±60 mT is
given by ΔR/R = (RVp=+60 V −RVp=−30 V)/RVp=+60 V ≈ −0.2%. In analogy, for I ⊥ H one
obtains ΔR/R ≈ 0.5%. The observed behavior can be attributed to the piezoresitive effect,
i.e. the so-called piezoresistance [113, 114, 115]. It describes the change of the electrical
82 Chapter 6 Local Magnetization Control
- 6 0 - 4 0 - 2 0 0 2 0 4 0 6 03 8 8
3 9 2
3 9 6
4 0 0
4 0 4
R long (m
Ω)
µ0 H ( m T )
I | | H V p = - 3 0 V
V p = 0 V
V p = + 6 0 V
(a)
- 6 0 - 4 0 - 2 0 0 2 0 4 0 6 0
3 5 6
3 6 0
3 6 4
3 6 8
V p = 0 VV p = + 6 0 V
V p = - 3 0 V
R long (m
Ω)
µ0 H ( m T )
I H
(b)
Figure 6.9: Magnetotransport measurement on a Ni thin film at different applied voltages Vp with (a)
I∥H and (b) I ⊥H.
resistance upon the application of mechanical stress. In contrast to the piezoelectric effect,
the applied stress only causes a change in the resistance, but no electrical potential is
generated. The resistance change due to the piezoresistive effect can be expressed as
ΔR
R=K ⋅
ΔL
L(6.3)
where K is a gauge factor and ΔLL the nominal expansion or contraction [113, 114]. The
gauge factor K implies the change of the cross-sectional area as well as eventual resistivity
changes caused by the deformation ΔLL . For evaporated Ni films, a gauge factor of K = 2.7
at room temperature was reported [116]. Moreover, the nominal stroke of the piezoelectric
actuator along the dominant axis of elongation (y-axis) is (ΔLL
)y= 1.3 × 10−3 [65]. Using
the Poisson relation with the given Poisson ratio = 0.45 (cf. Chapter 5), the nominal
deformation along x is given by (ΔLL
)x= − ⋅ 1.3 × 10−3 ≈ −0.6 × 10−3 . Inserting this data
into Eq. (6.3), the nominal resistance change for the I∥H configuration can be calculated
as ΔR/R ≈ −0.2% which is in agreement with the measurement data. In the transverse
configuration, the theoretical piezoresistance is given by ΔR/R = 0.4%, which is slightly
smaller than the observed value of ΔR/R ≈ 0.5%. For the purpose of exact quantification
of the piezoresistance, measurements at larger magnetic fields than 0H = ±60 mT are
indispensable. This quantification was not done in this thesis due to the limited field
range of the home-built electromagnet. Nevertheless, the presented data reveals clearly,
that the observed effect can be attributed to piezoresistance.
The resistance curves exhibit minima (maxima) for I∥H (I ⊥H) at the coercive fields
of the Ni thin film. For I∥H, the components of M perpendicular to the current grow and
thus Rlong decreases until the coercive field is reached. Then Rlong goes up again since
6.1 Ferromagnets on Piezoelectric Actuators 83
the components of M perpendicular to the current decrease. The same argumentation
holds true for the I ⊥ H configuration. Thus the extrema in the resistance curves are a
consequence of the AMR.
Another striking feature is the variation of the resistance curves for different ap-
plied piezo-voltages Vp. In both measurement geometries equivalent patterns can be
seen. The minimum, or maximum respectively, is more pronounced at Vp = −30 V
than at Vp = +60 V. Putting this into numbers, the anisotropic magneto resistance
(AMR = (Rmax −Rmin)/Rmin) × 100%) for I∥H yields AMRVp=−30 V ≈ 3.6% and
AMRVp=+60 V ≈ 0.8%, respectively. Similar magnitudes are obtained for I ⊥ H with
AMRVp=−30 V ≈ 3.6% and AMRVp=+60 V ≈ 0.3%.
Assuming a simple single domain Stoner-Wohlfarth model, this feature can be ratio-
nalized as follows. Applying Vp = +60 V to the actuator induces a magnetic easy axis
along x as discussed in Chapter 5. According to Stoner-Wohlfarth [95], the magnetization
reversal along the easy axis is characterized by a flip of the magnetization by 180° at the
coercive field [32]. As a consequence, there should be no change in the resistance ideally,
since no magnetization component perpendicular to the field is present. However, this
argument holds only true for perfect alignment of current parallel (perpendicular) to the
magnetic field, which is never given in reality. Thus the extrema in the resistance curve
for Vp = +60 V are a consequence of misalignment of I and H. Concerning the hard axis
magnetization reversal (Vp = −30 V) the Stoner-Wohlfarth theory predicts a coherent
magnetization rotation. Thus the transverse magnetization component increases steadily
until the maximum is reached at the coercive field before it decreases again. According
to AMR theory Rlong has a minimum (maximum) for I∥H (I ⊥ H) at the coercive field
which is more pronounced than for Vp = +60 V.
The Stoner-Wohlfarth approach enables to explain the observed strain dependent char-
acteristics in the resistance curves. However, in Chapter 5 Kerr imaging proved that the
easy axis magnetization loop is dominated by domain nucleation, whereas coherent mag-
netization rotation was found for the hard axis magnetization reversal. Thus the above
explanation using Stoner-Wohlfarth theory holds basically only true for the Vp = −30 V
curve. The magnetization reversal for Vp = +60 V as observed in Chapter 5 is abrupt and
accompanied by the formation of stripe domains. However, averaging over all domains
the magnetization component perpendicular to H remains also small compared to the
hard axis magnetization reversal and therefore a smaller extremum is expected. Hence,
considering the above explanations the resistance curves are in good agreement with the
Kerr images obtained in Chapter 5.
Magnetotransport and Local Strain
In the preceding section, the piezoelectric actuator was integrally strained and magne-
totransport was recorded simultaneously. In this paragraph, the concept of local strain
generation described above (cf. Fig. 6.1) is implemented together with four-point MTR
84 Chapter 6 Local Magnetization Control
measurements (cf. Figs. 6.8(b),(c)). The aim of this experiment is to correlate the
Kerr imaging results obtained in Section 6.1.4 for hybrids prepared by the evaporation
technique with magnetotransport. Therefore, eight PZT monolayers were selectively con-
tacted, whereas the remaining part of the piezoelectric actuator was held on a constant
potential. The resulting configuration is schematically depicted in Fig. 6.10. The area
comprised by the red box represents the eight selectively contacted PZT layers and thus
the region of local strain variation. The current geometries I∥H and I ⊥ H are also dis-
played with respect to the region of local strain. A corresponding domain image indicating
the locally controlled PZT layers can be seen in Fig. 6.7(d).
Vp
HI||
I
Figure 6.10: Local mag-
netization
control and
MTR.
The resistance curves of Ni for different applied local
strains and for both current orientations are depicted in
Fig. 6.11. Comparing these curves with the MTR results ob-
tained with a collectively strained actuator (cf. Fig. 6.9),
it can be noted that the characteristic features are again
present. However, the dependence of the anisotropic magneto
resistance on the applied voltage is not as distinct as above.
Calculating AMR = (Rmax −Rmin)/Rmin) × 100% for I∥H yields
AMRVp=−30 V ≈ 1.6% and AMRVp=+60 V ≈ 1.3% respectively. For
I ⊥ H the AMR can be calculated as AMRVp=−30 V ≈ 1.1% and
AMRVp=+60 V ≈ 0.7%. This behavior is of course an indication for
the generation of local strain whose influence onto magnetotrans-
port is inferior compared to the application of integral strain. For
I∥H, the resistances for Vp = −30 V and Vp = +60 V in magnetic saturation disperse no-
ticeable in comparison to the I ⊥ H configuration. The origin of this behavior are most
likely unwanted drift effects in addition to the already mentioned piezoresistance.
In Section 6.1.4 the domain nucleation and evolution in Ni on actuator under local
applied stress was discussed. Therefore, sequences ◻ and were applied as introduced
in Chapter 5. In order to compare and correlate MTR measurements with the observed
magnetic microstructure, sequence ◻ and were applied again and Vlong was recorded
simultaneously to the Kerr images at the measurement points B, C and B (cf. Fig. 5.13)
of the respective sequence. In order to minimize the statistical error, the voltmeter
was read out ten times per measurement point. The sample was prepared at A with
0H = −100 mT and V Ap = +60 V. The measurement field at B and C was iteratively
increased from 0H = −20 mT to 0H = 20 mT in each consecutive measurement sequence.
The corresponding voltage was cycled from V Bp = +60 V to V C
p = −30 V.
In Figs. 6.12(a),(b) the MTR measurement results for I∥H and I ⊥ H are depicted.
The resistances Rlong at B, C, and B (denoted as RB, RC and RB) are displayed for both
configurations. For I∥H, RB decreases with increasing magnetic field until the coercive
field is reached before it increases again. This behavior is in analogy to the resistance
curve for Vp = +60 V shown in Fig. 6.11(a). Below the coercive field the voltage cycle
6.1 Ferromagnets on Piezoelectric Actuators 85
- 6 0 - 4 0 - 2 0 0 2 0 4 0 6 03 9 4
3 9 6
3 9 8
4 0 0R lon
g (mΩ
)
µ0 H ( m T )
I | | H V p = - 3 0 VV p = 0 V
V p = + 6 0 V
(a)
- 6 0 - 4 0 - 2 0 0 2 0 4 0 6 0
3 6 8
3 7 0
3 7 2
R long (m
Ω)
µ0 H ( m T )
I H V p = - 3 0 V
V p = 0 V
V p = + 6 0 V
(b)
Figure 6.11: Magnetotransport measurement on a Ni/piezoactuator hybrid structure with (a) I∥H and
(b) I ⊥H. The actuator is strained locally by a variation of the applied voltage Vp.
from B to C certainly increases the M component perpendicular to H as observed in
the Kerr images depicted in Figs. 6.7(c)(d). As a consequence the decrease of RC with
increasing magnetic field is more distinct than for RB. Above the coercive field, both
voltage curves overlap again. Accordingly the behavior of RC and RB for I ⊥ H can be
explained.
RB does not differ from RB at first glance, regardless the current orientation. However,
the Kerr image shown in Fig. 6.6(c) suggested, that local magnetization switching induced
by a voltage cycle is not reversible. Therefore, it was expected that RB equals RC, which
is in contradiction to the observed data.
In order to investigate this behavior more closely, the difference of Rlong at C and
B (ΔRlong = RC-RB) and at B and B (ΔRlong = RB-RB) was derived from the MTR
measurement data and is represented by the closed symbols in Figs. 6.12(c),(d). Moreover,
the obtained Kerr images at points B, C and B were evaluated quantitatively as also done
in Section 4.7. Therefore, a region of interest (ROI), which comprises the region of strain,
was defined first (cf. Fig. 6.12(e)). Then the intensity of all pixels belonging to the ROI
was summed in order to get the integral MOKE signal. This evaluation yields an effective
M averaged over all local domains, which then can be described with a Stoner-Wohlfarth
model. Of course this approach does not reflect the real magnetic microstructure, but
nevertheless this model has proven to describe the transport data accurately as outlined
above. Using the relation cos = M/Ms [45, 117] the angle between magnetization and
external field can be determined. Obviously, this angle equals the angle between M
and I for I∥H. For all obtained images at B, C and B this evaluation was done and the
result can be seen in Fig. 6.12(f), where cos2( ) is plotted against the magnetic field.
Let us first consider the cos2( ) graphs for B and B depicted in Fig. 6.12(f). For
0H < 3.5 mT and 0H > 11.5 mT the curves are basically identical. This behavior indi-
86 Chapter 6 Local Magnetization Control
(b)
(d)
(f)
(a)
(c)
M
ROI
(e)
-20 -10 0 10 20396
398
400
402
Rlo
ng (m
Ω)
µ0H (mT)
I||H
RC
RB
RB
-20 -10 0 10 20
367
368
369
370
RB
RB
RC
Rlo
ng (m
Ω)
µ0H (mT)
I H
-20 -10 0 10 200.0
0.2
0.4
0.6
0.8
1.0
γ
H,I
M
BC
B
cos2 γ
µ0H (mT)
A B
CD
A,A B,B
C
-20 -10 0 10 20
-4
-2
0
2
4
I||H
RB-RB
RC-RB
∆R
long
(mΩ
)
µ0H (mT)
calculationMTR measurement
-20 -10 0 10 20
-2
-1
0
1
2
calculationMTR measurement
I H
∆R
long
(mΩ
)
µ0H (mT)
RC-RB
RB-RB
Figure 6.12: Results of MTR measurements in locally strained Ni applying sequence ◻ and for (a) I∥H
and (b) I ⊥ H. Plot of ΔRlong for (c) I∥H and (d) I ⊥ H. The closed symbols represent
the MTR measurement data, the open symbols represent the data calculated from the
Kerr images (see text). (e) Difference image obtained by Kerr microscopy. The ROI for
quantitative evaluation of Kerr images is depicted. The magnetization vector represents
the averaging over the whole magnetic substructure within the ROI. (f) Plot of cos2( )
versus 0H for B, C and B.
6.1 Ferromagnets on Piezoelectric Actuators 87
cates, that the magnetization rotation induced by Vp is reversible in this field regime. For
3.5 mT < 0H < 11.5 mT the curves differ, which means that the magnetization cannot
be rotated reversibly in this field range. This characteristic is in agreement with the qual-
itative analysis of the Kerr images shown in Section 6.1.4 which yielded no reversibility
regarding magnetization control by strain at magnetic fields close to the coercive field.
Thus reversible magnetization rotation is not possible within field range of 0Hc ± 4 mT.
Comparing the cos2( ) graphs for B and C (cf. Fig. 6.12(f)), the voltage sweep from B
to C induces a magnetization rotation away from the original state for 0H < 15 mT in the
depicted range, since both curves do not overlap in this field regime. For 0H > 15 mT,
the Zeeman energy dominates and thus the magnetization is aligned along the external
field regardless the applied strain.
With the knowledge of cos2( ), ΔRlong = RC-RB and ΔRlong = RB-RB can be calculated
according to Eq. (6.1) as:
RC −RB = (R∥ −R⊥) ⋅ (cos2 C − cos2 B) (6.4)
RB −RB = (R∥ −R⊥) ⋅ (cos2 B − cos2 B) (6.5)
The difference R∥−R⊥ can be approximated from the MTR data shown in Fig. 6.11. For
I∥H, R∥ is obtained directly as R∥ = Rlong(±60 mT) ≈ 401 mΩ (considering Vp = 0 V). For
the matter of simplicity, R⊥ is taken as the minima in the resistance curve at the coercive
field which yields R⊥ = Rlong(±10 mT) ≈ 395 mΩ. As a consequence, R∥ −R⊥ ≈ 6 mΩ. In
accordance, for I ⊥H the difference R∥ −R⊥ yields R∥ −R⊥ ≈ 3 mΩ.
ΔRlong = RC-RB and ΔRlong = RB-RB calculated from the MOKE data is displayed
with open symbols in Figs. 6.12(c),(d).
Comparing the calculated curves for ΔRlong (open symbols) with the measured curves
(closed symbols) for both current orientations, the general shape of both, calculated and
measured curve, is similar for RC-RB as well as for RB-RB. However, there are also several
discrepancies.
Regarding RC-RB, the piezoresistive effect has not been taken into account in the
calculated data, since the Kerr effect measurements are not sensitive onto piezoresistance.
This explains the difference of calculated and measured curve in magnetic saturation,
which is especially distinct in the I ⊥ H configuration. Apart from that, the shape of
both curves is basically equivalent except that the two extrema at the coercive field are
much more pronounced in the curve calculated from the MOKE data.
Accordingly, concerningRB-RB the discrepancy in the extrema around the coercive field
is even larger. Comparing the shape of both, measured and calculated curves, similarities
are present. However, from a quantitative point of view their features are very different.
Concluding, it is possible to reconstruct the general shape of the ΔRlong curve from
the MOKE measurement data especially for RC-RB, however there are still major dis-
88 Chapter 6 Local Magnetization Control
crepancies, which are attributed to the simplifications that have been made regarding
the calculation of the ΔRlong data. First, the approximation of R∥ −R⊥ from the MTR
measurement curve in Fig. 6.11 is surely a considerable simplification. For the matter of
correctness, a proper simulation of the MTR data of Fig. 6.11 would certainly yield more
correct results for R∥ −R⊥. Second, the definition of the ROI in Fig. 6.12(e) is random
and must not be the right choice. The actual current paths are certainly more compli-
cated but this is not reflected by the choice of the ROI. Third, the Stoner-Wohlfarth
approach to average over the magnetic substructure in the ROI and thus neglecting any
domain structure of the ferromagnet might be to simple at this point. It has been shown
that domains play also a significant role in magnetotransport [110]. Finally, the varia-
tion of strain in locally confined regions might influence the current flow such that the
assumption of constant current geometry is erroneous.
Nevertheless, the above analysis has proven, that local magnetization control in
Ni/actuator hybrids can be corroborated by MTR measurements. A further important
result of this paragraph is that the possibility of reversible/irreversible control of M de-
pends upon the magnetic field. These findings are in agreement with the Kerr images of
Section 6.1.4.
6.2 Patterned Ni Electrodes on Piezoelectric Substrates
In this section a novel approach for local magnetization control is presented. So far,
magnetization switching in non-patterned ferromagnetic films was demonstrated. Piezo-
electric actuators served as substrates. In the following the direct control of a patterned
ferromagnet will be analyzed. In particular, the piezoelectric actuator used so far is
replaced by other piezoelectric substrates. The aim of this section is to show that in
principle anisotropy and magnetization control is possible in small scale samples if an
effective coupling between ferroelectric and ferromagnetic properties is achieved.
6.2.1 Sample Preparation
In the preceding experiments Ni has proved to be a promising candidate for hybrid
structures allowing voltage controlled ferromagnetism. Although its magnetostrictive
constant is smaller than the one of Fe50Co50 (cf. Chapter 3), the strain induced effects
on the magnetic properties were more distinct in Ni than in Fe50Co50 as discussed in
Section 5.1.3. Therefore, Ni was chosen to serve as the ferromagnetic constituent for the
novel hybrid structure design. However, different kinds of piezoelectric substrates, namely
PZT, LiNbO3 and BaTiO3, have been investigated on their effectiveness in magnetoelastic
coupling.
6.2 Patterned Ni Electrodes on Piezoelectric Substrates 89
V+
V-
200 µm
V+
V-
z
xy
(a) (b)
(c)
(d)
1 cm
Figure 6.13: (a) Schematic view of an untreated PZT-5H plate. (b) Illustration of patterned Ni electrodes
deposited onto a PZT plate. (c) Photograph of contacted Ni electrodes on a PZT disk. (d)
Microscope image of Ni electrodes on PZT. Surface imperfections are clearly visible.
Ni on PZT
PZT exhibits a very high inverse piezoelectric effect (cf. Chapter 3) and is thus an
ideal candidate for application in piezoelectric/ferromagnetic hybrids. Therefore, PZT-5H
plates fabricated by Morgan Electroceramics [66] were used as substrate. They come with
a diameter of 10 mm and a thickness of 0.35 mm. The top and bottom face is covered
by a metal electrode which allows to apply an electric field along z in the conventional
mode of operation as depicted in Fig. 6.13(a). Thereby the dominant axis of elongation
and also the intrinsic polarization is aligned along the z-axis. For the purpose of this
thesis, the original top and bottom electrodes had to be removed prior to the deposition
of ferromagnetic electrodes onto one face of the substrate. Therefore, the faces of the
PZT disks were polished with an abrasive paper with smallest available graining (5 µm)
first. Additionally the substrate was polished with a 1 µm diamond suspension. However,
even after an extensive polishing process the surface of the PZT plate was still very rough
and porous as can be seen in Fig. 6.13(d). Even polishing in the TUM crystal laboratory
did not bring any major improvements.
Nevertheless, Ni electrodes were deposited onto the x-y-face of the plate as seen
schematically in Fig. 6.13(b). Fabrication of this electrode configuration allows to create
electric fields within the x-y-plane and thus it can be assumed that also lateral polariza-
90 Chapter 6 Local Magnetization Control
Sample Substrate Electrode geometries l∣w∣s
M090522A PZT 1000∣50∣50, 1000∣100∣100, 1000∣200∣100M090716A PZT 1000∣50∣50, 1000∣100∣100, 1000∣200∣100BTOM1 BaTiO3 1000∣50∣50LNOM1 LiNbO3 1000∣50∣50, 1000∣100∣100, 1000∣200∣100
Table 6.1: Overview of prepared samples for local magnetization control.
tion components are generated. Consequently lateral stress can be exerted by making use
of the d33 effect. The geometry of the electrode configuration is very important, since it
determines the device performance and stress distribution [118]. On the one hand a large
spacing between the electrodes improves the homogeneity of the electric field and thus
the strain between the electrodes. On the other hand large spacings require large applied
voltages to efficiently produce strain. Three different electrode geometries with length l,
width w and spacing s have been designed and fabricated; l∣w∣s in µm: I) 1000∣50∣50, II)
1000∣100∣100, III) 1000∣200∣100.
Firstly, optical lithography is used to generate the desired mask pattern onto one x-
y-face of the plate. After the lithography, polycrystalline Ni is deposited onto the PZT
plate via electron beam evaporation at a base pressure of approximately 1 × 10−7 mbar
and with a deposition rate of 1-2 A/s up to a final thickness of 100 nm. The photoresist is
removed by doing a lift-off in acetone under exposure to short ultrasound pulses. At last
the thus obtained electrodes are bonded onto contact pads as depicted in Fig. 6.13(c).
Details about the prepared samples can be found in Tab. 6.1.
Ni on BaTiO3 and LiNbO3
Besides PZT, also BaTiO3 and LiNbO3 were used as piezoelectric substrates. The prop-
erties of these two materials have been reviewed in Chapter 3. Compared to the PZT
plate, these substrates did not need a special treatment prior to the deposition of Ni apart
from cleaning. The lithography and evaporation process was equivalent to the method
presented above. Exemplary, the result for BaTiO3 can be seen in Fig. 6.14. The sample
was mounted onto a chip carrier and the electrodes were bonded (cf. Fig. 6.14(a)). In
contrast to PZT, the surfaces of the used BaTiO3 and LiNbO3 substrates are smooth and
thus the quality of the patterned Ni electrodes is superior (cf. Fig. 6.14(b)).
6.2.2 Polarization and Strain Control
In order to achieve local magnetization control via magnetoelastic coupling, the ability
of local polarization and strain generation in the substrates has to be verified. In the
present electrode structure it is assumed that the electric field penetrates laterally onto the
adjacent surface layers of the piezoelectric substrate and is distributed inhomogeneously.
6.2 Patterned Ni Electrodes on Piezoelectric Substrates 91
200 µm
(a) (b)
Figure 6.14: (a) Photograph of Ni electrodes patterned on a BaTiO3 substrate. The sample is mounted
onto a chip carrier. (b) Microscopy image of Ni electrodes on BaTiO3.
It has been shown for similar electrode designs, that three main regions regarding the
electric field distribution have to be considered [118, 119]: First, the electric field is
homogeneous and parallel to the surface between the patterned electrodes and close to
the surface. Second, at the electrode edges the field concentration is largest. Finally,
below the electrodes there is a region of depletion, since the electrodes are equipotential
areas. Having these considerations in mind the prepared samples were investigated upon
local polarization control.
In Fig. 6.15(a) an exemplary contact scheme of the electrodes is depicted. During the
measurement voltages were applied to the electrodes and the current was measured using
the standard two point technique. The result of an INi measurement of Ni on PZT in
a 1000∣50∣50 geometry can be seen in Fig. 6.15(b). Apart from the virgin curve, dis-
tinct peaks can be seen at VNi,pol ≈ ±100 V which correspond to displacement currents.
These peaks were reproducibly observed in consecutive cycles. Neglecting any inhomo-
geneous electric field distribution at the electrode edges and assuming a uniform electric
field within the PZT plate in between the patterned electrodes, the corresponding electric
field values can be estimated as Epol = VNi,pol/d. d = 4×50 µm denotes the spacing between
the contacted electrodes excluding the width of the three inner electrodes. Thus lateral
polarization reversal of the PZT plate can be achieved with an electric field of approxi-
mately Epol ≈ ±500 V/mm, which is in good agreement with the value Epol ≈ ±420 V/mm
measured for the PZT in the PSt 150/2 × 3/5 piezoelectric actuators in Section 6.1.2.
Equivalent measurements were done with different contact schemes and in different
electrode geometries for PZT. It was found that the spacing of PZT between the contacted
electrodes has a great influence on the measurement of the displacement current, as
expected. In general, the following observation was made. If the spacing d between the
contacted electrodes was greater than 200 µm, no displacement currents were found. This
92 Chapter 6 Local Magnetization Control
VNi
P
A
(a)
- 2 0 0 - 1 0 0 0 1 0 0 2 0 0
- 1 0
0
1 0
I (nA)
V N i ( V )
P
P
(b)
Figure 6.15: (a) Schematic illustration of poling and field generation in Ni electrodes patterned on piezo-
electric substrates. (b) I(VNi) curve obtained with patterned Ni electrodes on PZT in the
1000∣50∣50 geometry contacted as depicted in (a). The distinct peaks at VNi,pol = ±100 V
correspond to the displacement currents indicating that lateral polarization control is pos-
sible (sample M090716A).
is likely a consequence of the limited voltage supply by the source meter.
Doing the same measurements on Ni electrodes patterned on LiNbO3 and BaTiO3, no
displacement currents were found for any of the investigated electrode geometry. In case
of LiNbO3 no displacement current is expected, since LiNbO3 is a pyroelectric material
(cf. Chapter 3). For BaTiO3, an in plane coercive field of Hc ≈ 50 V/mm was reported
in [73]. Although the geometry of the electric field distribution in the present electrode
pattern is undefined, there should be regions between the electrodes with a uniform
electric field [118, 119] and thus one would expect polarization reversal and displacement
current. Thus the observed behavior for BaTiO3 cannot be explained.
Besides the state of polarization the control of strain is of fundamental interest. The
electric field distribution discussed above results in the generation of inhomogeneous
strains. The strain is expected to be uniform between the electrodes. Maximum strains
should be generated at the electrode edges.
To study the strain variation, a similar approach as in Chapter 5 was chosen. Strains
were detected by means of Kerr imaging due to strain induced birefringence in the PMMA
or MgO buffer layer underneath the ferromagnet. Here, no buffer layer is present. How-
ever, the refractive indices of all used piezoelectric substrates change as a function of the
applied electric field as a consequence of the linear electro-optic effect (Pockels effect)
[120, 121, 122, 123, 124]. Thus strain variation in the substrate should be manifested
by contrast changes in the Kerr images as a function of the electric field. Therefore, a
large magnetic field (0H = −50 mT) was applied to the PZT based sample (M090716A)
to exclude any magnetic contribution to the detected birefringence. The Ni electrodes
6.2 Patterned Ni Electrodes on Piezoelectric Substrates 93
-150 -100 -50 0 50
-3.0
-1.5
0.0
1.5
referenceimage
image (I)I (
nA)
VNi (V)
image (II) 1000|200|100
µ0H=-50 mT
VNi
H
(a)
(I)
(b)
(II)200 µm
(c)
Figure 6.16: (a) Measured I(VNi)-curve of Ni electrodes on PZT for 0H = −50 mT in the 1000∣200∣100
geometry and contact scheme as depicted in the figure inset. The reference image was
recorded at VNi = −160 V, image (I) at VNi = +30 V and image (II) at VNi = −160 V. (b)
Difference image (I), the red and blue lines indicate the contacted electrodes, the white
lines the non-contacted floating electrodes. The observed contrast shows that the strain
state has changed. (c) Difference image (II) at VNi = −160 V exhibits no contrast pattern.
The original strain state is again reached.
in the 1000∣200∣100 geometry were contacted as schematically illustrated in the inset of
Fig. 6.16(a). Then a I(VNi) cycle starting from VNi = −160 V was monitored. Simul-
taneously, Kerr images were recorded applying the difference imaging technique. The
resulting I(VNi)-curve is depicted in Fig. 6.16(a) indicating the voltages to the corre-
sponding difference images depicted in Fig. 6.16(b),(c). The red and blue lines in the
images represent the contacted electrodes, the white lines the floating electrodes respec-
tively. Sweeping the applied voltage from VNi = −160 V → +30 V the state of exerted
strain should change. This effect is manifested in Fig. 6.16(b), because a clear contrast
pattern in between the contacted electrodes is visible, whereas in the image part on the
right side the contrast has not varied significantly. Thus the applied voltage has gener-
ated strain locally in the PZT substrate. After the return sweep, the contrast pattern
vanishes completely again (cf. Fig. 6.16(c)), since present and original strain state are
equivalent. This observation proves that this electrode pattern onto the PZT substrate
is capable of local strain generation. The same characteristics were observed in BaTiO3
based samples. However, the LiNbO3 substrate did not reveal any local strain genera-
tion. This behavior is attributed to the low d33 coefficient of LiNbO3 compared to PZT
or BaTiO3 (cf. Chapter 3). Unfortunately, the Kerr technique does not straightforwardly
allow to quantify the strain and thus no strain comparison regarding different electrode
geometries or different piezoelectric substrates has been done.
94 Chapter 6 Local Magnetization Control
-10 0 10
-1
0
1
VNi=
+180
V
VNi= -30 V
M/M
s
µ0H (mT)
Hc for V↑
VNi
H1000|50|50
(a)
-10 0 10
-1
0
1
VNi= -160 V
M/M
s
µ0H (mT)
VNi= +30 V
VNi
1000|200|100H
Hc for V↑
(b)
Figure 6.17: M(H) hysteresis curve of Ni on PZT under applied voltages VNi at the patterned elec-
trodes for different geometries and contact schemes. (a) l∣w∣s: 1000∣50∣50, contacted elec-
trodes frame three non-contacted electrodes. (b) l∣w∣s: 1000∣200∣100, VNi applied between
neighboring electrodes (sample M090716A). The strain induced shift of the coercive field
is indicated in both figures.
6.2.3 Strain Induced Shift of the Coercive Field
Having confirmed the ability of local strain generation in PZT and BaTiO3 substrates
mediated by patterned electrodes, the influence of strain onto the magnetic properties
of the patterned Ni electrodes is analyzed in the following. The applied experimental
technique is again magneto-optic spectroscopy. In Chapter 4 the nominal resolution of
the Kerr microscope was calculated as 8 µm. Consequently it should be possible to resolve
the designed electrode patterns with a minimum width and spacing of 50 µm.
At first, Ni on PZT in the 1000∣50∣50 geometry was investigated. Two electrodes were
contacted embedding three floating electrodes in between as schematically depicted in
the inset of Fig. 6.17(a). The measurement procedure was equal to the one described
in Chapter 5. At constant voltage VNi the M(H) was recorded. The magnetic field was
oriented perpendicular to the stripe electrodes (cf. Fig. 6.17(a)). In order to evaluate
the Kerr rotation properly, a region of interest was defined which comprised the two
contacted electrodes as well as the electrodes in between.
The resulting hysteresis curves for VNi = +180 V and VNi = −30 V are shown in
Fig. 6.17(a). Evidently there is a small shift of approximately 1 mT in the coercive
field, which was confirmed in several measurements. Thus, this shift has to be attributed
to the applied strain.
The same result was obtained for a 1000∣200∣100 structure on PZT depicted in
Fig. 6.17(b). This time neighboring electrodes were contacted and the voltage range
(VNi = −160 V and VNi = +30 V) was on the negative branch of the piezoelectric hystere-
sis. The magnetic field orientation was again perpendicular to the stripe electrodes as
6.3 Summary 95
depicted in the figure inset. Once more, there is a shift in the coercive field in the M(H)
curves for different applied voltages, which has approximately the same magnitude as
above but is different in sign.
This sign difference regarding the shift of the coercive field observed in Fig. 6.17(a)
and Fig. 6.17(b) can be rationalized by considering the ferroelectric hysteresis of the PZT
substrate as discussed in Chapter 2 and schematically depicted in Fig. 2.4. The M(H)
measurement of the 1000∣50∣50 patterned electrodes (cf. Fig. 6.17(a)) was prepared on
the positive hysteresis branch, whereas in the 1000∣200∣100 geometry (cf. Fig. 6.17(b)) the
PZT substrate was prepared on the negative hysteresis branch. Thus the sign difference
of the shift of the coercive field is consistent with the expected strain characteristic of the
piezoelectric substrate.
To conclude, the obtained results prove, that on application of voltages the coercive
field shifts. Thus, magnetoelastic coupling is indeed possible to some degree in the novel
Ni/PZT hybrid structures. However, in BaTiO3 and LiNbO3 based hybrids no coercivity
shift in dependence of the applied voltage VNi could be detected and therefore the data is
not shown here. This might be reasoned by their lower d33 coefficient compared to PZT
(cf. Chapter 3). Kerr imaging was also applied, but the structure size of the electrodes is
so small, that even if the highest possible magnification is used no statement about the
magnetic substructure can be made and thus the presentation of Kerr images is omitted.
Nevertheless, the novel design of patterned electrodes onto PZT disks has proven to
work in principle. In order to improve the efficiency of the magnetoeleastic coupling,
more sophisticated methods for sample preparation would be necessary, like for example
an interdigital electrode design. Moreover, the choice of the piezoelectric substrate also
plays a decisive role. PZT has proven to be the most promising candidate.
6.3 Summary
In this chapter the feasibility of local magnetization control by means of electric fields
in multiferroic hybrid structures was demonstrated. Hybrids based on piezoelectric ac-
tuators were modified and allowed local magnetization switching as observed with the
Kerr technique. This result was corroborated by magnetotransport measurements. Fi-
nally, a novel hybrid architecture consisting of patterned nickel electrodes on piezoelectric
substrates was proposed and first experiments proved its potential for local strain and
magnetization control.
Chapter 7
Conclusion and Outlook
The aim of this thesis was to investigate multiferroic hybrid structures focusing on mag-
netization control by means of electric fields on macroscopic as well as on microscopic
scales. This topic was motivated by the great fundamental and technological interest in
electric field controlled ferromagnetism at room temperature, which would substantially
facilitate scaling of spintronic devices to small dimensions. Magneto-optical Kerr imaging
was applied to study and illustrate the processes within the magnetic substructure of the
prepared multifunctional hybrid structures.
Establishment of Spatially Resolved Kerr Imaging
An existing magneto-optical Kerr setup was redesigned and extended to allow for spatially
resolved magneto-optical Kerr measurements. Additional components like an adequate
CCD camera, a LED-based light source and a quarter wave plate have been integrated
and the imaging procedures optimized in order to observe magnetic domains. With the
novel Kerr setup lateral resolutions on the order of 10 µm can be achieved considering
magnifications on the order of one. Therefore, it can be denoted a low resolution Kerr
microscope. As shown in Chapter 4, a characterization of the setup concerning imaging
contrast, noise and sensitivity has been implemented. For optimal contrast conditions
the difference imaging technique was applied and the settings of the polarization optics
has been discussed. The system’s sensitivity is limited by photon noise. A determination
of the minimum resolvable Kerr rotation in integral M(H) measurements yielded the
value of Δ'k ≈ 0.15′′ and thus beats the sensitivity of the original MOKE setup [45].
Moreover, the capabilities of the established setup in imaging of thin ferromagnetic films
were demonstrated and also the possibility of quantitative evaluation of domain patterns
was discussed.
Electric Field Control of Ferromagnetism
In Chapter 5 hybrid structures consisting of a ferromagnetic thin film and of either a
piezoelectric actuator or a BaTiO3 single crystal have been studied intensively by means
97
98 Chapter 7 Conclusion and Outlook
of the established magneto-optical Kerr setup. The influence of strain onto the magnetic
hysteresis was analyzed and it was found that the ferromagnetic properties of all sam-
ples exhibit strain dependence. The origin of this strain coupling is the magnetoelastic
effect and its strength depends crucially on the strain coupling of the piezoelectric and
ferromagnetic material.
Therefore, two different preparation techniques of the actuator/ferromagnet hybrids
have been compared by means of voltage dependent M(H) measurements, domain evo-
lution and magnetization switching. It was found that strain coupling in samples pre-
pared by direct evaporation of ferromagnet onto piezoelectric actuator is superior than
in samples prepared by the cementing technique. Apart from that, the capability of
electric field controlled magnetization switching in these large scale samples was investi-
gated. However, the application of strain also induces contrast changes in the Kerr images
resulting from strain induced birefringence or surface effects in the buffer layer of the ac-
tuator/ferromagnet hybrid structure. To exclude strain mediated contrast changes, two
measurement sequences (sequence ◻ and sequence ) have been proposed and successfully
applied to the actuator based samples as well as to the BaTiO3 based heterostructures.
In sum, all prepared samples allowed magnetization switching by means of electric field
sweeps which has been detected and visualized with the Kerr technique.
Local Magnetization Control
The feasibility of local magnetization control has been discussed in Chapter 6. Actuator
based hybrid structures were manipulated to allow for local strain control. This modifi-
cation proved to enable local magnetization control by applying local strains generated
by the piezoelectric effect. As a consequence, local switching of the magnetization by
means of electric fields has been observed. Moreover, magnetotransport measurements
were applied simultaneously to MOKE measurements and they confirmed magnetization
rotation induced by local strain. However, the scalability of the electric field controlled
ferromagnet is limited by the geometry of the piezoelectric actuator.
In order to achieve substrate-independent scaling to smaller dimension a novel archi-
tecture consisting of patterned ferromagnetic electrodes on piezoelectric substrate was
proposed and first samples were fabricated. PZT, BaTiO3 and LiNbO3 have been used
as piezoelectric substrates and nickel as the ferromagnetic constituent. First experiments
proved, that local strain generation is possible in the novel design and also successful mag-
netoelastic coupling has been observed by means of magnetic hysteresis measurements.
Outlook
Beyond the outlined achievements, further interesting implementations, applications and
fundamental studies exist. In the following a brief outlook on possible future projects is
99
given:
Although the established magneto-optical setup allows domain imaging and character-
ization of large scale samples prepared at the WMI, there are further issues that could be
implemented to improve the system. First, the light source and collimation optics cur-
rently integrated in the setup should only be a temporary solution, since homogeneous,
collimated illumination of sufficient intensity is one of the key-points to successful domain
visibility. Second, as mentioned in Chapter 4 the resolution is limited by diffraction, be-
cause simple wide angle objective lenses are used for imaging. In order to achieve higher
resolution together with considerable image brightness the use of microscope objectives
and eyepieces is indispensable. A new design of the setup involving smaller working dis-
tances, microscope optics and a well collimated incoherent light source would certainly
increase the resolution by at least an order of magnitude and thus allow high resolution
Kerr microscopy.
From the experimental point of view this thesis brought up some further ideas to aim
at the realization of electric field controlled ferromagnetism on all scales. Regarding the
actuator/ferromagnet hybrid structures, I can think of two major experiments that could
not be achieved in the framework of this thesis. First, a closer examination of the strain
induced contrast pattern presented in Chapter 5 would be appropriate. This anomalous
contrast pattern arises from birefringence effects in the buffer layer or from surface effects.
Inserting an intermediate buffer layer of sufficient thickness, like a gold thin film, would be
sufficient to rule out birefringence effects and thus clarify this open question. Concerning
local magnetization control experiments on actuators with integrated magneto-transport
measurements, well defined Hall-bar structures could finally resolve all doubts about local
magnetization control, since quantitative transport measurements or angular dependent
magnetoresistance measurements [125] would be possible.
The new architecture consisting of patterned ferromagnetic electrodes on piezoelectric
substrates proposed in Chapter 6 proved to be a substantial alternative for local magne-
tization control. PZT turned out to be the most promising candidate for the piezoelectric
constituent. However, surface quality of the used PZT plates was unsatisfactory. There-
fore, experiments using lead magnesium niobate-lead titanate (PMN-PT) single crystals
as piezoelectric substrates are planned. PMN-PT has piezoelectric coefficients which are
comparable to the ones of PZT, but the surface quality is superior. Thus the quality
of the magnetic thin films would be improved and an amplified coupling between the
piezoelectric and the ferromagnetic phase is expected.
In conclusion, the concept of multiferroic hybrid structures aiming at an electric field
control of ferromagnetism is still under vigorous research and it offers many open ques-
tions and unsolved applications. However, the great potential of these multifunctional
hybrid structures for fundamental research and possible technological applications has
100 Chapter 7 Conclusion and Outlook
been pointed out.
Bibliography
[1] J. Kerr, Phil. Mag. 3, 321 (1877).
[2] G. P. Zhang, W. Huebner, G. Lefkidis, Y. Bai, and T. F. George, Nature
5, 499 (2009).
[3] Y. Zhu, X. Zhang, T. Li, X. Huang, L. Han, and J. Zhao, Appl. Phys. Lett.
95, 052108 (2009).
[4] M. Lambeck, IEEE T. Magn. 4, 51 (1968).
[5] A. Huber and R. Schafer, Magnetic Domains, Springer Verlag, 1998.
[6] S. Dey, M. Bowman, and A. Booth, J. Phys. E-Scient. Instr. 2, 162 (1969).
[7] C. Fowler and E. Fryer, J. Opt. Soc. Am. 44, 256 (1954).
[8] F. Schmidt, W. Rave, and A. Hubert, IEEE T. Magn. 21, 1596 (1985).
[9] G. Binasch, P. Grunberg, F. Saurenbach, and W. Zinn, Phys. Rev. B 39,
4828 (1989).
[10] M. Baibich, J. Broto, A. Fert, F. van Dau, F. Petroff, P. Eitenne,
G. Creuzet, A. Friedrich, and J. Chazelas, Phys. Rev. Lett. 61, 2472 (1988).
[11] M. Julliere, Phys. Lett. 54, 225 (1975).
[12] N. A. Spaldin and M. Fiebig, Science 309, 391 (2005).
[13] L. W. Martin, S. P. Crane, Y.-H. Chu, M. B. Holcomb, M. Gajek,
M. Huijben, C.-H. Yang, N. Balke, and R. Ramesh, J. Phys.: Cond. Mat.
20, 434220 (2008).
[14] W. Eerenstein, N. Marthur, and J. Scott, Nature 442, 759 (2006).
[15] W. Eerenstein, M. Wiora, J. L. Prieto, J. F. Scott, and N. Mathur,
Nat. Mater. 6, 348 (2007).
[16] T. Lottermoser, T. Lonkai, U. Amann, D. Hohlwein, J. Ihringer, and
M. Fiebig, Nature 430, 541 (2004).
101
102 Bibliography
[17] C. Binek and B. Doudin, J. Phys.: Cond. Mat. 17, 39 (2005).
[18] M.-T. Bootsmann, S. Dokupil, E. Quandt, T. Ivanov, N. Abedinov, and
M. Lohndorf, IEEE T. Magn. 41, 3505 (2005).
[19] A. Brandlmaier, S. Geprags, M. Weiler, A. Boger, M. Opel, H. Huebl,
C. Bihler, M. S. Brandt, B. Botters, D. Grundler, R. Gross, and
S. T. B. Goennenwein, Phys. Rev. B 77, 104445 (2008).
[20] S. Goennenwein, M. Althammer, C. Bihler, A. Brandlmaier,
S. Geprags, M. Opel, W. Schoch, W. Limmer, R. Gross, and M. Brandt,
phys. stat. sol. (RRL) 2, 96 (2008).
[21] J. Lee, S.-C. Shin, and S.-K. Kim, Appl. Phys. Lett. 82, 2458 (2003).
[22] C. Bihler, M. Althammer, A. Brandlmaier, S. Geprags, M. Weiler,
M. Opel, W. Schoch, W. Limmer, R. Gross, M. S. Brandt, and S. T. B.
Goennenwein, Phys. Rev. B 78, 045203 (2008).
[23] R. Gross, Magnetism and Spintronics, in Lecture Notes, 2005.
[24] S. Blundell, Magnetism in Condensed Matter, Oxford University Press, 2003.
[25] P. Weiss and G. Foex, Le Magnetisme, Armand Colin, Paris, 1926.
[26] R. C. O’Handley, Modern Magnetic Materials: Principles and Applications,
John Wiley & Sons, Inc, 2000.
[27] L. D. Landau and E. Lifshitz, Phys. Z. Sowjetunion 8, 337 (1935).
[28] L. D. Landau and E. M. Lifshitz, Course of theoretical physics V: Statistical
physics, Pergamon Press, 2nd edition, 1959.
[29] S. Goennenwein, Two-Dimensional Electron Gases and Ferromagnetic Semicon-
ductors: Materials for Spintronics, PhD thesis, TU Munchen, 2003.
[30] A. Brandlmaier, Magnetische Anisotropie in dunnen Schichten aus Magnetit,
Master’s thesis, Walther-Meissner-Institut, TU Munchen, 2006.
[31] S. Chikazumi, Physics of Magnetism, John Wiley & Sons, Inc, 1964.
[32] A. H. Morrish, The Physical Principles of Magnetism, New York: IEEE Press,
2001.
[33] A. Arnau, Piezoelectric Transducers and Applications, Springer Verlag, 2004.
[34] M. Weiler, Magnetization Control in Multiferroic Heterostructures, Master’s
thesis, Walther-Meissner-Institut, TU Munchen, 2007.
Bibliography 103
[35] Piezoline, Technical report, Piezosystems Jena GmbH, 2007.
[36] D. Damjanovic, The Science of Hysteresis, Elsevier, 2005.
[37] T. Morita, Y. Kadota, and H. Hosaka, Appl. Phys. Lett. 90, 082909 (2007).
[38] J. Nye, Physical Properties of Crystals: Their Representation by Tensors and
Matrices, Oxford Science Publication, 2000.
[39] M. J. Freiser, IEEE T. Magn. 4, 152 (1968).
[40] P. Bruno, Y. Suzuki, and C. Chappert, Phys. Rev. B 53, 9214 (1996).
[41] P. N. Argyres, Phys. Rev. 97, 334 (1955).
[42] M.-A. Schroeder, Magnetooptische Kerr-Effekte im VUV an Eisen und Eisen-
systemen, PhD thesis, Universitat Hamburg, 2000.
[43] F. Schwabl, Quantenmechanik, Springer Verlag, 2002.
[44] W. Zinth and H. J. Koerner, Physik III, Oldenbourg Verlag, 1998.
[45] M. Pelkner, Aufbau und Charakterisierung eines Spektrometers fur magnetoop-
tischen Kerr-Effekt, Master’s thesis, Walther-Meissner-Institut, TU Munchen, 2008.
[46] M. Faraday, Phil. Trans. Roy. Soc. 136, 1 (1846).
[47] R. Vollmer, Lineare und nichtlineare Magnetoooptik an ultradunnen ferromag-
netischen Schichten und Vielfachschichten. , in 30. IFF Ferienkurs Julich, 1999.
[48] G. Traeger, L. Wenzel, and A. Hubert, phys. stat. sol. (a) 131, 201 (1992).
[49] A. Hubert and G. Traeger, J. Magn. Magn. Mater. 124, 185 (1993).
[50] C. C. Robinson, J. Opt. Soc. Am. 53, 681 (1963).
[51] M. Mansuripur, The physical principles of magneto-optical recording, Cambridge
University Press, 1995.
[52] R. P. Hunt, J. Appl. Phys. 4, 1652 (1938).
[53] Z. J. Yang and M. R. Scheinfein, J. Appl. Phys. 74, 6810 (1993).
[54] C. C. Robinson, J. Opt. Soc. Am. 54, 1220 (1964).
[55] J. M. Florczak and E. D. Dahlberg, J. Appl. Phys. 67, 7520 (1990).
[56] C. Kittel, Introduction to Solid State Physics, Oldenbourg Verlag, 1991.
104 Bibliography
[57] N. W. Ashcroft and D. N. Mermin, Solid State Physics, Oldenbourg Verlag,
2001.
[58] H. Danan, A. Herr, and A. J. P. Meyer, J. Appl. Phys. 39, 669 (1968).
[59] E. W. Lee, Rep. Prog. Phys. 18, 184 (1955).
[60] A. Clark, J. Restorff, M. Wun-Fogle, D. Wu, and T. Lograsso, J. Appl.
Phys. 103, 07B310 (2008).
[61] J. W. Shih, Phys. Rev. 46, 139 (1934).
[62] Landolt-Bornstein, Group III, Vol. 6, Springer Verlag, 1971.
[63] W. Martienssen and H. Warlimont, Springer Hanbook of Condensed Matter
and Materials Data, Springer Verlag, 2005.
[64] E.Sawaguchi, Journal of the Physical Society of Japan 8, 615 (1953).
[65] Low voltage co-fired multilayer stacks, rings and chips for actuation, Technical
report, Piezomechanik GmbH, Munich, 2006.
[66] D. Berlincourt and H. H. A. Krueger, Properties of Piezoelectricity Ceram-
ics, TP-226, Technical report, Morgan Electro Ceramics.
[67] M. Zgonik, P. Bernasconi, M. Duelli, R. Schlesser, P. Gunter, M. H.
Garrett, D. Rytz, Y. Zhu, and X. Wu, Phys. Rev. B 50, 5941 (1994).
[68] R. T. Smith and F. S. Welsh, J. Appl. Phys. 42, 2219 (1971).
[69] W. J. Merz, Phys. Rev. 76, 1221 (1949).
[70] A. von Hippel, Rev. Mod. Phys. 22, 221 (1950).
[71] F. Jona and G. Shirane, Ferroelectric Crystals, Pergamon Press, Oxford Uni-
versity Press, 1962.
[72] L. Shebanow, phys. stat. sol. A 65, 321 (1981).
[73] D. Pantel, Dielektrische und magnetische Eigenschaften multifunktionaler
Dunnschichtstrukturen, Master’s thesis, Walther-Meissner-Insitut, TU Munchen,
2008.
[74] R. S. Weis and T. K. Gaylord, Appl. Phys. A 37, 191 (1985).
[75] A. Sonin and B. Strukow, Introduction to Ferroelectricity, Vieweg Verlag, 1974.
[76] C. A. Fowler and E. M. Fryer, Phys. Rev. 94, 52 (1954).
Bibliography 105
[77] D. Treves, J. Appl. Phys. 32, 358 (1961).
[78] K. Shirae and K. Sugiyama, J. Appl. Phys. 53, 8380 (1982).
[79] H. Rohrmann and H. Hoffmann, Thin Solid Films 175, 273 (1989).
[80] Digital Camera Fundamentals, Technical report, ANDOR Technology.
[81] SNR - signal-to-noise-ratio, Technical report, The Cooke Corporation - PCO. Imag-
ing.
[82] C. Ballentine, R. Fink, J. Araya-Pochet, and J. Erskin, Appl. Phys. A
49, 459 (1989).
[83] S. Bader and J. Erskine, Ultrathin Magnetic Structures - Magneto-Optical
Effects in Ultrathin Magnetic Structures, Springer Verlag, 2005.
[84] D. A. Allwood, G. Xiong, M. D. Cooke, and R. P. Cowburn, J. Phys. D:
Appl. Phys. 36, 2175 (2003).
[85] E. Palik, Handbook of Optical Constants of Solids, Academic Press, New York,
1985.
[86] T. Herrmann, K. Ludge, W. Richter, K. G. Georgarakis, P. Poulopou-
los, R. Nunthel, J. Lindner, M. Wahl, and N. Esser, Phys. Rev. B 73,
134408 (2006).
[87] K. P. Kamper, W. Schmitt, G. Guntherodt, R. J. Gambino, and R. Ruf,
Phys. Rev. Lett. 59, 2788 (1987).
[88] K. B. Chetry, M. Pathak, P. LeClair, and A. Gupta, J. Appl. Phys. 105,
083925 (2009).
[89] W. Rave, R. Schafer, and A. Hubert, J. Magn. Magn. Mater. 65, 7 (1987).
[90] W. Rave and A. Hubert, IEEE T. Magn. 26, 2813 (1990).
[91] K. Dorr and C. Thiele, phys. stat. sol. (b) 243, 21 (2006).
[92] M. Weiler, A. Brandlmaier, S. Geprags, M. Althammer, M. Opel,
C. Bihler, H. Huebl, M. S. Brandt, R. Gross, and S. T. B. Goennenwein,
N. J. Phys. 11, 013021 (2009).
[93] L. Callegaro and E. Puppin, Appl. Phys. Lett. 68, 1279 (1996).
[94] L. Callegaro, E. Puppin, and A. Vannucchi, Rev. Sci. Instrum. 66, 1065
(1994).
106 Bibliography
[95] E. C. Stoner and E. P. Wohlfarth, Phil. Trans. R. Soc. A 240, 599 (1948).
[96] L. Pickelmann, Piezomechanik GmbH, private communication.
[97] G. Kaye and T. Laby, Table of physical and chemical constants, Longman,
London, UK, 1993.
[98] K. Vedam and E. D. D. Schmidt, Phys. Rev. 146, 548 (1966).
[99] S. R. P.B. Bowden, Phil. Magazine 22, 463 (1970).
[100] F. Ay, A. Kocabas, C. Kocabas, and A. Aydinli, J. Appl. Phys. 96, 7147
(2004).
[101] S. Shastry, G. Srinivasan, M. I. Bichurin, V. M. Petrov, and A. S.
Tatarenko, Phys. Rev. B 70, 064416 (2004).
[102] C. Thiele, K. Dorr, O. Bilani, J. Rodel, and L. Schultz, Phys. Rev. B 75,
054408 (2007).
[103] C.-W. Nan, M. I. Bichurin, S. Dong, D. Viehland, and G. Srinivasan, J.
Appl. Phys. 103, 031101 (2008).
[104] L. Qiao and X. Bi, Appl. Phys. Lett. 92, 214101 (2008).
[105] L. Chen and A. L. Roytburd, Appl. Phys. Lett. 90, 102903 (2007).
[106] K. Schroder, J. Appl. Phys. 53, 2759 (1982).
[107] V. Novosad, Y. Otani, S. Kim, K. Fukamichi, J. Koike, K. Manuyama,
O. Kitakami, and Y. Shimada, J. Appl. Phys. 87, 6400 (2000).
[108] S. Parkin, K. Roche, M. Samant, P. Rice, and R. Beyers, J. Appl. Phys.
85, 5828 (1999).
[109] T. R. McGuire and R. I. Potter, IEEE T. Magn. 11, 1018 (1975).
[110] M. Viret, D. Vignoles, D. Cole, J. M. D. Coey, W. Allen, D. S. Daniel,
and J. F. Gregg, Phys. Rev. B 53, 8464 (1996).
[111] D. Thompson, L. Romankiw, and A. Mayadas, IEEE T. Magn. 11, 1039
(1975).
[112] A. Brandlmaier, M. Brasse, M. Opel, G. Woltersdorf, R. Gross, and
S. Goennenwein, Electric field control of remanent magnetization in multifunc-
tional hybrids, DPG Dresden, 2009, 2009.
[113] G. C. Kuczynski, Phys. Rev. 94, 61 (1954).
Bibliography 107
[114] R. L. Parker and A. Krinsky, J. Appl. Phys. 34, 2700 (1963).
[115] D. K. Bagchi and B. D. Cullity, J. Appl. Phys. 38, 999 (1967).
[116] E. Klokholm, J. Vac. Sci. Technol. 10, 235 (1973).
[117] W. Gil, D. Gorlitz, M. Horisberger, and J. Kotzler, Phys. Rev. B 72,
134401 (2005).
[118] C. Bowen, A. Bowles, S. Drake, N. Johnson, and S. Mahon, Ferroelectrics
228, 257 (1999).
[119] W. Beckert and W. Kreher, Comp. Mater. Sci. 26, 36 (2003).
[120] S. Ducharme, J. Feinberg, and R. Neurgaonkar, IEEE J. Qu. Electr. 23,
2116 (1987).
[121] Y.Liu, Z. Chen, C. Li, D. Cui, Y. Zhou, G. Yang, and Y. Zhu, J. Appl.
Phys. 81, 6328 (1997).
[122] C. Land and P. Thacher, Proc. IEEE 57, 751 (1969).
[123] C. E. Land, J. Am. Ceram. Soc. 72, 2059 (2005).
[124] L. Arizmendi, phys. stat. sol. (a) 201, 253 (2004).
[125] W. Limmer, M. Glunk, J. Daeubler, T. Hummel, W. Schoch, R. Sauer,
C. Bihler, H. Huebl, M. S. Brandt, and S. T. B. Goennenwein, Phys.
Rev. B 74, 205205 (2006).
Acknowledgment
At this point I want to thank all the people who contributed to this thesis and supported
me during my time as diploma student at the Walther-Meissner-Institut:
Prof. Dr. Rudolf Gross, for giving me the opportunity to accomplish this thesis at the
WMI and for his supervision of this work. When looking back on this year as diploma
student, I will always remember it as very pleasant and instructive time!
Dr. Sebastian Gonnenwein, the mentor of this thesis, for his grand support and ad-
vice. His richness of ideas together with his infinite enthusiasm were important stimuli
regarding this thesis but also regarding my development as a physicist. I am grateful
for the many hours he spent in the lab together with me to discuss the experiment or
measurement results.
I am especially thankful to Andreas Brandlmaier, who was an excellent guide during
my time at the WMI. He introduced me to all experimental issues and it’s due to his
everlasting patience and motivation that I acquired all the experimental skills necessary
for this work. I also profited from the numerous discussions with him in which his aptitude
for criticism and foresight opened my eyes for novel approaches regarding experiments or
result interpretation.
Stephan Geprags, in first place for the preparation of the BaTiO3 based heterostructures
and his support regarding their measurement and result interpretation. On top of that, I
gratefully appreciate his humorous and friendly personality but also his professional skills
which he was always eager to share with me.
Dr. Matthias Opel, for his encouragement and motivation during the preparation of this
thesis.
All other members of the Magnetiker group, who were without exception always willing
to help concerning experimental problems or any other issue.
Thomas Brenninger, who was always helpful if problems with the EVAP occurred and
owing to his effort I always found excellent conditions for the thin film preparation.
Robert Muller, for the additional components he built for the new MOKE setup.
109
Special thanks to my office mates and “fellow sufferers”, in person Daniel, Michi, Timo,
Chris, Hajo, Themis and Martin, for sharing such a good time in and out of the office. I
wish you all the best for your future.
Last but not least, I want to say thank you to my family and to Martina, who were
always there for me and gave me the best imaginable support. Vielen Dank!
110