correlation between the structural, optical, and magnetic

Correlation Between the Structural, Optical, and Magnetic Properties

of CoFeB and CoFeB Based Magnetic Tunnel Junctions

Upon Laser or Oven Annealing

Fakultät für Naturwissenschaften

der Technischen Universität Chemnitz

Dissertation zur Erlangung des akademischen Grades

eingereicht am 7. Januar 2020

vorgelegt von M. Sc. Apoorva Sharma

geboren am 02. Mai 1987 in Bikaner, Indien

Gutachter:

Prof. Dr. Dr. h. c. Dietrich R.T. Zahn

Prof. Dr. Georgeta Salvan

“ Every search begins with beginner’s luck. And every

search ends with the victor’s being severely tested.”

― Paulo Coelho

Bibliografische Beschreibung

Sharma, Apoorva

Correlation Between the Structural, Optical, and Magnetic Properties of CoFeB and

CoFeB Based Magnetic Tunnel Junctions Upon Laser or Oven Annealing

Technische Universität Chemnitz Dissertation, 2020

Diese Dissertation befasst sich mit der Untersuchung der maßgeblichen

Herausforderungen der heutigen TMR-Präparation (tunneling magnetoresistance) für

beispielsweise Magnetfeldsensor- oder auch Speichertechnologie (MRAM – magnetic

random access memory).

Im ersten Teil der Arbeit werden die elektronischen, strukturellen und magnetischen

Eigenschaften der ferromagnetischen Elektrode eines typischen magnetischen

Tunnelkontaktes, z.B. CoFeB, erforscht, wobei spektroskopische Ellipsometrie, magneto-

optische Spektroskopie, Röntgendiffraktometrie und Messverfahren für den spezifischen

elektrischen Widerstand zum Einsatz kommen. Weiterhin wurde der Einfluss der

Temperatur einer thermischen Behandlung auf die optischen und magneto-optischen

Merkmale untersucht, wobei eine starke Korrelation zwischen den beobachteten

spektralen Merkmalen und der Kristallisation von CoFeB nachgewiesen wurde. Die

(magneto-) optische Spektroskopie bietet somit eine zerstörungsfreie und besonders

sensitive Validierungsmethode für die Dünnschichtkristallisation, die in die moderne

CMOS Herstellungstechnologie integriert werden kann.

Der zweite Teil der Arbeit befasst sich mit dem lokalen Tempern unter Verwendung eines

fokussierten Laserstrahls, mit dem Ziel die Referenzmagnetisierung in einem

magnetischen Tunnelkontakt definiert einzustellen und die Wirkung der hierfür

notwendigen thermischen Behandlung auf die übrigen Schichten im Schichtstapel zu

untersuchen. Hierzu wurden zahlreiche Parameter für das laserbasierte lokale Tempern

variiert, um die optimale Austauschfeldstärke im magnetischen Referenzsystem

einzustellen, idealerweise ohne den gegebenen Schichtstapel zu schädigen. Schließlich

wurde der Einfluss des laserbasierten Temperns (als auch des Ofentemperns) auf die

Unversehrtheit der Schichten und Grenzflächen, insbesondere auf die Diffusion

verschiedener Elemente, mittels Röntgen-Photoemissionsspektroskopie untersucht.

Keywords

Exchange bias, Laser irradiation, Magnetic tunnel junction, Magneto-optical Kerr effect

spectroscopy, Magnetometry, Spectroscopic ellipsometry, Spintronic, Tunnel

magnetoresistance and 3d transition metal boride.

Table of content

Chapter 1: Introduction ................................................................................................. 1

1.1 This thesis ............................................................................................................. 2

Chapter 2: Theory.......................................................................................................... 5

2.1 Magnetism ............................................................................................................. 5

2.2 Spintronics ............................................................................................................15

2.3 Optical spectroscopy ............................................................................................18

Chapter 3: Experimental ..............................................................................................31

3.1 Sample preparation ..............................................................................................31

3.2 Annealing ..............................................................................................................32

3.3 Measurement techniques .....................................................................................35

Chapter 4: 3d-transition metal boride layers: structural, electronic, and magnetic

properties ......................................................................................................................46

4.1 Introduction ...........................................................................................................46

4.2 Thick films .............................................................................................................47

4.3 Thin films ..............................................................................................................60

4.4 Conclusion ............................................................................................................65

Chapter 5: Setting exchange bias using laser vs oven annealing techniques .......67

5.1 Introduction ...........................................................................................................67

5.2 Magnetisation reversal of complex MTJs layer stack ............................................69

5.3 Application of FORC analysis: from single CoFeB layer to MTJ layer stack .........81

5.4 Potential of direct-write laser annealing technique................................................84

5.5 Conclusion ............................................................................................................86

Chapter 6: Exchange bias and diffusion processes in laser annealed CoFeB/IrMn

thin films .......................................................................................................................88

6.1 Introduction ...........................................................................................................88

6.2 Magnetometry investigations ................................................................................89

6.3 XPS-depth profiling ...............................................................................................90

6.4 Structural analysis ................................................................................................94

6.5 Topographic characterisation ...............................................................................95

6.6 Conclusion ............................................................................................................96

Chapter 7: Summary and outlook ...............................................................................97

Appendix A .................................................................................................................100

Bibliography ...............................................................................................................101

List of figures .............................................................................................................111

List of tables ...............................................................................................................118

Abbreviations .............................................................................................................119

Erklärung ....................................................................................................................120

curriculum vitae .........................................................................................................121

Scientific contributions .............................................................................................122

Acknowledgements ....................................................................................................126

Introduction Chapter 1

1

Chapter 1: Introduction

High-precision magnetic field sensors have become indispensable in a wide variety of

modern scientific equipment and industrial devices. Their fields of applications range from

the medical and health care sector to the technology of everyday life. For medical

purposes, for example, the magnetic signature of a tracer particle may be tracked in vivo

along all three spatial dimensions as it passes the intestinal tract of a patient, thereby

revealing any anomalies or functional disorders1,2. Also, direct monitoring of the

biomagnetic signals from the heart and the brain has been realised by ultrahigh-precision

magnetic field sensors1,3,4. In automobiles and industrial equipment, to provide another

example, magnetic field sensors are employed for reliable, high-precision, length, angle

and position measurements under challenging environmental conditions5,6. In the field of

power generation and transport, especially within the emerging context of dynamic,

distributed “green” power generation by wind and water, trustworthy contact-free

measurements of electric currents using magnetic field sensors are of great importance7.

The severity of such measurements increases manyfolds in the internet of things (IoT)

applications, where ever so small leakage currents need to be detected by adequate

sensor technologies to save energy8,9. Furthermore, magnetic field sensors have recently

entered the market of mobile communication as their increasingly high sensitivity now

allows smartphone navigation based on the electronic reading of the earth’s natural

magnetic field10.

Aforementioned fields of application, as well as many others, have in common that they

are experiencing a continuous replacement of conventional Hall and anisotropic

magnetoresistance (AMR) sensors by more precise magnetoresistive components. In

particular, the so-called magnetoresistive effects, i.e. the giant (GMR)11 and tunnelling

(TMR)12 magnetoresistance, in which besides the electron charge also its spin is used as

an information carrier, exhibit much higher signal amplitudes and sensitivities compared

to Hall and AMR technology13. The application of the GMR effect already led to

revolutionary progress in hard disk drive technology with such an impact that A. Fert and

P. Grünberg were awarded the Nobel Prize in Physics in 2007. In the last two decades,

the GMR-based technology has been slowly replaced by TMR-based devices, particularly

after the introduction of the first TMR-based read-heads in 2005. Their extremely high

signal output at room temperature, along with their very low energy consumption and their


2

much higher miniaturisation capabilities as compared to the sensors based on the Hall

effect, AMR or GMR effect, promises a significant improvement of sensor quality. The

remarkable achievements in the evolution of the magnetoresistive sensor are shown in

Figure 1.

Figure 1. The milestones in the evolution of magnetoresistive devices.

1.1 This thesis

The research work presented in this thesis was performed as a part of a joint project

between the Semiconductor Physics group at Chemnitz University of Technology (HLPH-

TUC), Back-End-of-Line group at the Fraunhofer Institute for Electronic Nano Systems

(BEOL-Fraunhofer ENAS) and the Laser Microtechnologies at the Laser Institute of the

Hochschule Mittweida (LHM). The project “Interfacial perpendicular magnetic

anisotropy for next-generation monolithic 3D TMR sensors” was financially

supported by the Deutsche Forschungsgemeinschaft with the Grant No.: 282193534. The

objective of this project was the monolithic integration of a three-dimensional magnetic

field sensor system based on TMR in CoFeB / MgO / CoFeB for quantitative

measurements of the magnetic field. The primary emphasis was laid on a higher

sensitivity, minimum signal hysteresis, low overall power consumption as well as a high

degree of miniaturisation. Although the project was tightly knitted among the three project

partners, the work packages can be broadly classified as sample/material

characterisation, microfabrication and laser annealing performed by HLPH-TUC, BEOL-

Fraunhofer ENAS and LHM, respectively.


3

The performance of TMR sensors vastly depends on the thermal treatment for the

crystallisation of the ferromagnetic electrodes as well as the MgO barrier and for setting

the reference magnetisation via exchange bias. The typical method for thermal treatment

involves vacuum oven annealing in the presence of a magnetic field. However, this

inherits some limitations, such as non-selective alignment of the magnetisation direction

of the reference layer for each sensor on a wafer and time consumption. Often, due to the

limited thermal budget, the vacuum oven annealing is incompatible with CMOS

technologies. A laser-based annealing approach overcomes the shortcomings of the

conventional annealing method. This method allows not only the selective magnetisation

alignment of the reference layer of the single sensor but also reduces the annealing time.

This thesis focuses on studying the various aspects of laser annealing on the magnetic

and non-magnetic layers comprised of CoFeB / MgO / CoFeB magnetic tunnel junctions

(MTJ). Properties such as magnetisation, exchange bias, crystallisation and surface

morphology in conjunction with varying laser exposure parameters were investigated.

Additionally, (magneto-) optical spectroscopy methods were used to understand the

electronic structure of the ferromagnetic layers. For benchmarking purposes, studies were

made using the conventional vacuum oven annealing method.

Chapter 2 of this thesis describes the necessary theoretical background required to

explain and understand the results presented in this work. The chapter begins with a brief

overview of the fundamentals of magnetism and explains characteristic features of

ferromagnets, namely, magnetic hysteresis, coercive field, remanence magnetisation etc.

The following sections describe the phenomena that are important for this research, such

as magnetic anisotropy, exchange bias and tunnel magnetoresistance in-depth. It also

sheds light on the advantages of optical spectroscopic techniques and how these can be

exploited to understand material properties. In chapter 3, the sample preparation and

characterisation techniques that have been used in the framework of this thesis are

presented. The basic principle and the technical details for the equipment used are

explained. Additional specifications regarding the individual experiments are discussed

along with the experiment in each following chapter. Chapter 4 presents the optical and

magneto-optical response of the materials investigated in this thesis. Additionally, the

investigation of the crystallisation of CoFeB ferromagnetic layers using (magneto-) optical

methods along with X-ray diffraction as well as electrical four-point probe method is

discussed. Chapter 5 brings a proof-of-concept that the laser annealing is a suitable


4

method for heat-treating the magnetic tunnel junction devices. It also aims to establish the

optimal laser annealing parameters based on the magnetic properties of the MTJ layer

stack and presents the study with different seed layers, which were tested in order to

improve the exchange bias fields. Chapter 6 deals with the more fundamental

investigation of the influence of laser annealing on a reference layer stack. The main focus

here is to characterise and understand the diffusion and the type of migrating species in

the layer stack induced by the laser annealing. Finally, a summary of all the studies is

compiled in chapter 7.

Theory Chapter 2

5

Chapter 2: Theory

This chapter lays the theoretical foundation for the interpretation of the results presented

in this thesis. It also serves the purpose to familiarise the reader with the scientific terms

used throughout this thesis. At first, the fundamentals of magnetism are explained,

followed by a brief review about spintronics and the principles behind the TMR effect,

along with the development road map of TMR sensors. The second part draws attention

toward the light-matter interaction with a focus on (magneto-) optical spectroscopic

techniques for characterisation and analysis of magnetic materials. A section of this

chapter is published (with minor changes) by Royal Society of Chemistry ISBN-13:

9781839162107, ISBN-10:1839162104.

2.1 Magnetism

The term magnetism in ordinary life is associated with the physical property of

ferromagnetic materials to be attracted by a magnetic field. For instance, the spontaneous

magnetisation of the lodestone (magnetite) is already known to humanity for several

centuries. However, it was only in the early 19th century that Hans-Christian Oersted

bridged the magnetism with electricity and triggered more systematic studies of

magnetism. Later on, the discovery of the Faraday and Kerr effect by Michael Faraday

and John Kerr revolutionised the concept of interaction between light and magnetic

materials14. A more detailed description of these effects is presented in section 2.3.5. The

discovery of Faraday and Kerr effect laid the foundation of the famous Maxwell equations

developed by James Clark Maxwell. In the following, the mathematical equations relevant

for this work along with their units, are explained. The units associated with the physical

quantities are shown in appendix A.

The relation between the magnetic induction (B ) and the applied magnetic field (H ) in

vacuum is given by14,15

= 𝜇0 2-1

μo is the magnetic permeability of the vacuum, also known as the magnetic constant.

Theory Chapter 2

6

In order to describe the magnetic field inside matter, a third vector term called

magnetisation (M ) was introduced. The magnetisation M of a material can be defined as

the vector sum of all magnetic moments (m ) located within the material volume (V):

=1

𝑉∑ 2-2

The total magnetic induction will account for the contribution of the magnetisation

accordingly:

= 𝜇0(𝐻 + 𝑀) 2-3

Hence, the relation between the applied magnetic field and the magnetisation is given by

= 𝜒 2-4

where, χ is the magnetic susceptibility, a dimensionless quantity.

Based on the magnetic susceptibility (or permeability), all elements and their alloys can

be classified into three categories, namely as diamagnetic, paramagnetic or magnetically

ordered (ferromagnetic, antiferromagnetic, ferrimagnetic, etc.). For diamagnetic

materials, the magnetic susceptibility is negative (χ < 0). In this case, the atomic

magnetic moments in the absence of the magnetic field are zero and an external magnetic

field H induces atomic magnetic moments, which leads to a reduction of B . On the other

hand, paramagnetic materials have χ > 0; an intrinsic atomic magnetic moment is

already present in the absence of the external magnetic field, but the atomic moments do

not interact with each other. When an external magnetic field H is applied, the magnetic

moments tend to align with the applied field, leading to a linear increase of B with H .

Paramagnetic materials consist of atoms or ions with unpaired electrons, which results in

an individual magnetic moment associated with the atoms or ions. However, these

individual magnetic moments do not interact with each other and have no long-range

ordering, resulting in a zero net magnetisation. Hitherto, for the ease of presentation, the

vector notations for these quantities will be dropped.

Ferromagnetic (FM) materials show magnetic order below a critical temperature, with χ

being several orders of magnitude larger than for the paramagnetic state. Such materials

Theory Chapter 2

7

exhibit magnetic hysteresis and persistent magnetisation at zero fields, the so-called

remanent magnetisation or remanence. Due to these properties, ferromagnetic materials

have gained a lot of attention since centuries and are still an interesting subject for current

research activities. Another peculiar magnetic ordering case is the so-called

antiferromagnetism (AFM), where the neighbouring atomic magnetic moments are

coupled antiparallel to each other, leading to zero remanence. The exchange bias

phenomenon often observed in AFM/FM layered systems represents a significant part of

investigations in this thesis, as the exchange bias not only determines the directional

sensitivity but also improves the signal to noise ratio in magnetic field sensors. A

schematic representation of the classes of magnetically ordered materials (including

cases that are not addressed in this thesis) is shown in Figure 2.

Figure 2. The classification of magnetic materials [Image adapted from14].

2.1.1 Ferromagnetism

As mentioned above, ferromagnetic materials are materials that possess a spontaneous

magnetisation below the Curie temperature. The spontaneous magnetisation is the net

magnetisation that exists inside a magnetised microscopic volume in the absence of an

Theory Chapter 2

8

external magnetic field. The Curie temperature (TC) is the temperature above which

ferromagnetic materials lose their spontaneous magnetisation and become

paramagnetic. Iron (Fe), Cobalt (Co), and Nickel (Ni) are the three elementary

ferromagnetic metals at room temperature. Furthermore, some other rare-earth metals

(4f-elements) also have significantly contributed to magnetism, especially when alloyed

with transition metals. A very prominent example is NdFeB, which is used in permanent

magnets.

Two theories of ferromagnetism have been successful in explaining many of its properties:

the Weiss theory16 and the Stoner band theory of ferromagnetism17. In classical physics,

Weiss postulated that an internal “molecular field” acts in ferromagnetic materials to align

the magnetic moments parallel to each other. Below TC, the molecular field is strong

enough to keep a magnetisation even in the absence of an externally applied field. At high

temperatures, the thermal energy KBT, is larger than the energy of the molecular field,

resulting in the random orientation of the magnetic moments and thus paramagnetic

behaviour with a small but positive susceptibility. This theory has proven to describe the

experimentally observed Curie-Weiss behaviour for the temperature dependence of the

susceptibility for many ferromagnetic materials in the paramagnetic region above the

Curie temperature. However, there are two limitations to the Weiss theory. First, it

assumes that the number of magnetic dipoles below and above the Curie temperature is

the same (ferro to para-magnetic phase transition), while this is in contrast to experimental

evidence. Second, the magnetic dipole moment is an integer number of the Bohr

magneton, which is not observed experimentally18. To explain this discrepancy, the Stoner

band theory of ferromagnetism is needed.

The band theory of ferromagnetism was first proposed by E. Stoner19 and then

independently by J. Slater20–22. It has been successful in explaining non-integer values of

atomic magnetic moments and predicting some aspects of the magnetic behaviour of the

3d metals and alloys evaluated from the imbalance of spin-up and spin-down electrons

by integrating the respective spin-polarised density of states15.

𝑚 = 𝜇𝐵(𝑛↑ − 𝑛↓) 2-5

where, B = 9.27×10-24 J·T-1 is the Bohr magneton and n↑ denotes the number of spinup

and n↓ the number of spindown electrons.

Theory Chapter 2

9

The Stoner model assumes that the magnetisation occurs due to the exchange interaction

between the electrons with the same spin orientation with neighbouring atoms in a crystal

lattice, thus causing an effective imbalance in spin-up and spin-down electron density.

This difference in concentration of spin-up and spin-down electrons results in a band

splitting known as exchange splitting (Δ). The existence of ferromagnetism can then be

expressed by the Stoner criterion15.

𝑆𝐷(𝐸𝐹) ≥ 1 2-6

here, S is the Stoner constant and D(EF) is the density of states.

Thus, the ferromagnetism can be explained by an electron density of states, where the

imbalance due to spin orientations of electrons is addressed by relocating the

corresponding energy densities with respect to each other. The difference between the

upper edge of the 3d band of majority spin electrons and the maximum energy occupied

by an electron at 0K (also known as Fermi energy EF) is the Stoner gap. Figure 3 a shows

the Stoner model for ferromagnetic metals with a 3d shell. The shaded and unshaded

areas represent the occupied and unoccupied states, respectively. The band with higher

occupancy of electrons is called the majority band, and the corresponding electrons are

named majority spin electrons, while the term minority refers to the band with lower

occupancy and the related electrons.

Figure 3. The Stoner model of ferromagnetic metals illustrated for the 3d shell, and nomenclature used for the band description of magnetism (a.). Occupied electron states below the Fermi energy EF are shaded, unoccupied states above EF are shown unshaded. Hext, m and Δ denote the external magnetic field, magnetic moment and exchange splitting, respectively. The calculated density of states of Fe, Co, Ni, and Cu (b.) [Images are taken from15].

Theory Chapter 2

10

2.1.2 Antiferromagnetism

The theory of ferromagnetism in the 3d-transition metals, Fe, Co and Ni, is based on the

postulation of positive exchange interactions between neighbouring atoms/ions. However,

this assumption does not explain the paramagnetic and antiferromagnetic (negative

exchange) behaviour for remaining 3d metals (Sc, Ti, V, Cr, Mn). In the year 1936, Louis

Néel developed the concept of antiferromagnetism14. He observed that below a certain

critical temperature, the atomic magnetic moments in such metals/alloys are arranged

alternately parallel and antiparallel. Above this temperature, the moments are disordered

similar to ferromagnets above TC. This critical temperature was later named after him as

Néel temperature (TN) of antiferromagnets. The Néel temperature is defined as the

transition temperature at which an antiferromagnet becomes paramagnetic. This theory

was later mathematically modelled by Van Vleck23.

A Néel antiferromagnet lattice comprises two intertwined sublattices with equal and

opposite magnetisation, hence resulting in zero net magnetisation. For example, an IrMn3

lattice has two magnetic sublattices A and B, shown in Figure 4, with the magnetisation

MA = -MB. According to the Weiss theory, the inter-sublattice molecular field coupling can

be given as nAB and nBA14. Similarly, the intra-sublattice coupling is given by nAA and nBB.

Therefore, the net molecular field of each sublattice is

𝐻𝐴 = 𝑛𝐴𝐴𝑀𝐴 + 𝑛𝐴𝐵𝑀𝐵 + 𝐻 2-7

𝐻𝐵 = 𝑛𝐵𝐵𝑀𝐵 + 𝑛𝐵𝐴𝑀𝐴 + 𝐻 2-8

here, nAA = nBB, nAB = nBA, and H is the external magnetic field.

In the absence of external field H = 0, the net magnetisation M = MA + MB = 0 and the

magnetisation of each sublattice becomes zero at TN, i.e. MA(TN) = MB(TN) = 0. The

spontaneous magnetisation of each sublattice in the paramagnetic region

(MA = MB = χHB = χHA) for an antiferromagnet can thus be given by:

𝑀𝐴 = (𝐶∗

𝑇) (𝑛𝐴𝐴𝑀𝐴 + 𝑛𝐴𝐵𝑀𝐵 + 𝐻) 2-9

Theory Chapter 2

11

𝑀𝐵 = (𝐶∗

𝑇) (𝑛𝐵𝐵𝑀𝐵 + 𝑛𝐵𝐴𝑀𝐴 + 𝐻) 2-10

where, C∗ = (μ0(n/2)meff2 )/3kB and n is the number of magnetic ions per unit volume with

n/2 on each sublattice14.

Equations 2-9 and 2-10 must have a nontrivial solution in order to fulfil the condition of

spontaneous magnetisation of each sublattice in the absence of a magnetic field. For this,

the determinant of the coefficient of MA and MB must be zero resulting in

[(𝐶∗

𝑇)𝑛𝐴𝐴 − 1]

2

− [(𝐶∗

𝑇)𝑛𝐴𝐵]

2

= 0 2-11

and rearranging the terms on the right-hand side and left-hand sides

(𝐶∗

𝑇)𝑛𝐴𝐴 − 1 = [(

𝐶∗

𝑇)𝑛𝐴𝐵] 2-12

This yields TN = C*·(nAA - nAB). The susceptibility of the antiferromagnet above TN can be

described by adding the sublattice coupling in the Curie-Wiess law.

𝜒 =

𝐶

𝑇 − 𝜃𝑃 2-13

where, θp = C*·(nAA + nAB) and C = 2C*.

The Néel temperature of disordered bulk IrMn3 is around (960 ± 10)K, and increases with

increasing iridium content in the alloy24. Additionally, TN varies from bulk antiferromagnetic

material to thin films. In this we focused on IrMn3 stoichiometry, hence hereafter this

stoichiometry will be referred to as IrMn.

Figure 4. The magnetic lattice structure of IrMn superimposed over the crystal lattice (a.). Magnetic sublattice showing inbound (A-yellow) and outbound magnetisation (B-green) with zero net magnetisation (b.) [images adapted from24].

Theory Chapter 2

12

2.1.3 Magnetic hysteresis

A ferromagnetic material, when subjected to an external magnetic field, results in an

alignment of its magnetisation in accordance with the applied magnetic field direction. The

magnetisation may increase from the remanent magnetisation (Mr) (which also can be

zero if the material was demagnetised before) up to the saturation magnetisation (Ms) at

the maximum when the saturation field (Hs) is applied. If the field is now reduced to zero,

the magnetisation decreases from Ms to Mr. To reduce the magnetisation to zero

(demagnetisation) an external magnetic field in the opposite direction is required, known

as the coercive field or coercivity (HC). The plot of the magnetisation as a function of

applied magnetic field strength for ferromagnetic material results in a closed-loop called

M(H) hysteresis loop, as shown in Figure 5. The suitability of a ferromagnetic material for

a particular application is widely determined by the characteristics shown by its hysteresis

loop (Ms, Mr, HC). For example, a squared shaped hysteresis loop, revealing two stable

states, might be suitable for data storage, whereas materials with small coercive fields

accompanied by a linear evolution of M(H) between remanence and saturation are the

preferred candidates for magnetic field sensor applications.

Figure 5. Exemplary M(H) hysteresis loop of the ferromagnetic material Co60Fe20B20 after 30 min annealing at 600°C: Mr is the remanent magnetisation at H=0; Ms denotes the saturation magnetisation and HC the coercivity.

2.1.4 Magnetic anisotropy

The magnetic anisotropy energy is a measure of the dependence of the ground-state

energy on the direction of the magnetisation. This anisotropy defines the preferential

(easy), the intermediate, and the magnetically hard directions of magnetisation in

Theory Chapter 2

13

materials. The overall magnetic anisotropy is a combination of several contributions.

There are three main contributing anisotropies in the materials studied throughout this

thesis, namely, magnetocrystalline anisotropy, shape anisotropy, and exchange

anisotropy14,15,25.

Magnetocrystalline anisotropy: The magnetocrystalline anisotropy energy is defined as

the energy required to rotate the magnetisation from the easy to a hard direction14. This

occurs due to the spin-orbit coupling, i.e. the coupling between the spin and orbital

momentum of the electrons. When an external field tries to reorient the electron’s

magnetic moment, the orbit of that electron also has to get reoriented, but due to strong

coupling to the crystal lattice, the magnetisation rotation might be hindered.

Shape anisotropy: Shape anisotropy is the result of the demagnetising field (Hd)

generated by the samples due to its own magnetisation14. If the sample is spherical and

a magnetic field is applied, the field will magnetise the sample to the same extent in all

directions. As the magnetisation is uniform, the demagnetisation field is distributed equally

in all directions. A non-spherical sample, for example, a long bar is easier to be

magnetised along the long axis than the short axis, due to the non-uniform demagnetising

field generated by the rod. The shape anisotropy must not be confused with

magnetocrystalline anisotropy as shape anisotropy is not an intrinsic property of the

material and solely depends on the shape of the FM sample.

Exchange Anisotropy: In the year 1956, Meiklejohn and Bean discovered a new form of

anisotropy in magnetic material systems and named it to exchange anisotropy26. They

stated that such anisotropy is the consequence of the interaction between the

ferromagnetic and the antiferromagnetic layers in a FM/AFM bilayer system when cooled

through the Néel temperature in the presence of the static magnetic field. This results in

a shift (or biasing) of the M(H) loop, generally in the opposite direction of the cooling field.

Such a shift is termed as the exchange bias field (HEB). See Figure 6 for a schematic

representation of the effect on the hysteresis loop.

Theory Chapter 2

14

Figure 6. M(H) hysteresis loop of CoFeB / IrMn bilayer at room temperature (RT). The green line represents the M(H) response when the sample is cooled from 200°C to RT in the presence of a magnetic field (120 mT), and the red line is the response of the as-deposited, demagnetised state.

Along with the shift in the hysteresis loop, exchange biased systems exhibit an angular

dependence kud·cosθ resulting in a unidirectional anisotropy in addition to the common

uniaxial anisotropy kua·sin2θ observed in a FM (see Figure 7). Here, kud and kua are the

unidirectional anisotropy and FM uniaxial anisotropy constants. The angle between the

magnetisation and the anisotropy axis is denoted with θ. The energy of such a coupled

bilayer system is the summation of both anisotropy energy per unit area.

𝐸 = −𝜇𝑜𝑀𝐹𝑀𝐻𝑡𝐹𝑀 𝑐𝑜𝑠 𝜃 − 𝜎𝑒𝑥 𝑐𝑜𝑠 𝜃 + 𝑘𝑢𝑎 𝑡𝐹𝑀𝑠𝑖𝑛2 𝜃 2-14

where, MFM is the magnetisation of the ferromagnetic layer with a thickness tFM and

σex = kud·tFM.

In this equation, the first energy term is due to the presence of the external magnetic field,

the second term represents the exchange bias energy, and the last term is related to the

uniaxial anisotropy energy in ferromagnets. Here it must be noticed that the overall

exchange bias field strength varies inversely with the thickness of the ferromagnetic layer

(its magnetisation, to be accurate) for thin films (typically up to 10 nm) and is independent

of the film thickness for thick films14.

In general, for an exchange bias system as the temperature increases and approaches

TN, the exchange bias field strength decreases and vanishes ultimately. However, it is

often observed in thin FM/AFM bilayers that the exchange bias effect disappears already

well below TN. The temperature at which the exchange bias field becomes zero (HEB = 0)

Theory Chapter 2

15

is, therefore, called blocking temperature (TB) and can be significantly lower than TN. For

example, IrMn has TN = 690 K and TB = 540 K27.

Figure 7. Uniaxial magnetic anisotropy induced through the exchange bias effect in CoFeB / IrMn after cooling from 200°C in the presence of 120 mT magnetic field. The coercive fields (a.) and the exchange bias fields (b.) are plotted as determined by angle-resolved longitudinal MOKE magnetometry at room temperature.

The exchange interaction, as well as further related effects, have been vastly investigated

for possible applications in magnetoresistive devices. For instance, the appropriate use

of an AFM material (layer) in a magnetoresistive layer stack assists in tuning the sensitivity

range of a magnetic field sensor and in suppressing the Barkhausen noise. The extensive

experimental research on AFM/FM layers stacks has also stimulated theoretical

descriptions, particularly concerning the microscopic origin of the effect28. In general, the

spins of FM are parallel to each other. However, at the interface to the AFM layer,

localised net moments arise from interfacial roughness, frustrated exchange bonds

(interfacial spins coupled antiferromagnetically), stress and dislocations, giving rise to the

exchange bias27,29,30.

2.2 Spintronics

Spintronics, as the name suggests, is the combination of the term spin and electronics,

whereby both, charge and spin of the electron(/s) are exploited. An often-proposed

scheme of a spintronic device is based on the magnetoresistive effects (MR). In particular,

magnetoresistive devices, based on the giant magnetoresistance (GMR) and tunnelling

magnetoresistance (TMR) effect, are comprised of a non-magnetic layer (conducting in

case of GMR and insulating for TMR) sandwiched between two FM layers.

Theory Chapter 2

16

In the year 1971, P.M. Tedrow and R. Meservey et al. laid the foundation of the TMR

effect by discovering the spin-polarised tunnelling for a Al / Al2O3 / Ni junction31. M.

Julliere, in 1975, proposed the spin conservation theory and presented a quantitative

explanation for the response of the magnetic tunnel junction (MTJ)12. M. Bowen et al.

emphasised the importance of the electronic and crystal structure of the entire electrode/

barrier/ electrode system in the year 2001 and measured a TMR resistance change of

60% at 30K for a Fe(001) / MgO(001) / FeCo(001) MTJ32. This MR response was four

times higher than any other contemporary Al2O3-based MTJ studied until that date.

However, the very low operational temperature limited the concept to find any device

application. Later on, Ikeda et al. presented in 2008 a breakthrough result for the

CoFeB / MgO / CoFeB system with an extraordinary TMR ratio of 603% at room

temperature33.

2.2.1 Tunnelling magnetoresistance effect

Tunnelling magnetoresistance is a magnetoresistive effect that occurs in a magnetic

tunnel junction, which is a multilayer structure consisting of two ferromagnets separated

by an ultra-thin non-magnetic (NM) insulator acting as a tunnel barrier. The charge

transfer through the tunnel barrier occurs due to quantum tunnelling and due to the

regulation of spin-polarised current controlled by the relative orientation of the

Figure 8. The development road map of the magnetoresistive devices [image was taken from34].

Theory Chapter 2

17

magnetisation of two ferromagnetic layers. The current through the two ferromagnetic

electrodes consists of two separate spin current channels for spin-up and spin-down

electrons (Mott`s two-current model35). The two channels encounter a different electrical

resistance in ferromagnetic materials due to the spin polarisation of the FM. The electrode

spin polarisation arises from the imbalance of the density of states of the spin-up and spin-

down electrons near the Fermi level in the ferromagnetic layers and thus from their

magnetic anisotropy in the magnetised state, see Figure 9.

Figure 9. Schematic representation of the tunnel magnetoresistance in the case of two identical ferromagnetic layers separated by a non-magnetic insulating barrier such as MgO. The tunnelling process conserves the spin. When the electronic states on each side of the barrier are spin-polarised, the electrons will more easily find free states to tunnel through the barrier if the magnetisations are parallel (a.) than if they are antiparallel (b.) to each other due to the ratio of the density of states of both electrons (spin-up, spin-down) at the Fermi level. The arrows in green and red show the higher and lower tunnelling probability of spin-polarised electron through a tunnel barrier, respectively. The yellow balls represent the electrons with their intrinsic spin orientation in grey [image redrawn from36].

In an FM/NM/FM structure, the overall electrical resistance depends then on the mutual

magnetisation directions of both FM:

𝑇𝑀𝑅 =2𝑃1𝑃2

1 − 𝑃1𝑃2=

𝑅↑↓ − 𝑅↑↑

𝑅↑↑ 2-15

where, P1 and P2 represent the spin polarisations of two ferromagnetic layers and R↑↓ and

R↑↑ are the resistance of two FM in parallel and antiparallel magnetisation.

Theory Chapter 2

18

2.3 Optical spectroscopy

As discussed previously, the electronic properties of the MTJs ferromagnetic electrodes

play a decisive role in MTJs. Hence it is very important to probe and understand these

properties. There are many experimental techniques to measure the electronic states of

materials, namely: photoelectron spectroscopy (PES), inverse photon-electron

spectroscopy (IPES), electron energy loss spectroscopy (EELS), optical spectroscopy,

etc. Among these methods, the optical spectroscopy techniques hold the high figure of

merit, as they are relatively simple and non-invasive techniques. Optical spectroscopy

techniques are based on the principle of light-matter interaction and can be broadly

classified on the basis of the wavelength of electromagnetic wave used to probe the

material (THz, visible, UV, X-ray etc.). In the present work, two optical spectroscopy

techniques namely, spectroscopic ellipsometry (SE) and magneto-optical Kerr effect

(MOKE) spectroscopy in the NIR (SE only), visible, and UV spectral ranges were used to

assess the electronic as well as magnetic properties of ferromagnetic materials.

Additionally, in this framework, the optical spectroscopies were used to investigate

structural changes in the ferromagnetic layer and their influence on its optical properties.

Both spectroscopic techniques follow the principles of polarimetry, where the sample

under investigation is illuminated with light of known polarisation. The changes in the

polarisation state of the reflected light are recorded, thus providing information about the

electronic states of the sample. The information that can be deduced from these two

methods includes the complex refractive index, the layer thickness, roughness, magnetic

hysteresis, exchange coupling, and crystallinity etc.

2.3.1 Polarisation of light

Light is an electromagnetic wave with the electric and magnetic fields vectors oscillating

perpendicular to each other and also perpendicular to the direction of propagation. It is

sufficient to discuss only the electric field vector to describe the polarisation of the light.

Thus, an electromagnetic wave travelling in the z-direction can be mathematically

represented as a superposition of two orthogonal components of the electric field:

(𝑧, 𝑡) = (𝐸𝑥 + 𝐸𝑦)𝑒𝑖(𝑘𝑧−𝜔𝑡) 2-16

Theory Chapter 2

19

here, Ex and Ey are the components of the electric field projected onto the x- and y- axes.

The quantities X and Y are unit vectors with magnitude one and direction pointing along

their respective axes.

For example, Ey = iEx, where Ex is a real number, means that the y-component of the

electric field is phase-shifted with respect to the x- component by π/2. Substituting the

value in the above equation and considering the real part we obtain:

𝑅𝑒[𝐸(𝑧, 𝑡)] = 𝑅𝑒[𝐸𝑥𝑒𝑖(𝑘𝑧−𝜔𝑡)] + 𝑅𝑒 [𝑒

𝑖𝜋2 𝐸𝑥𝑒

𝑖(𝑘𝑧−𝜔𝑡)]

2-17 = 𝐸𝑥 𝑐𝑜𝑠(𝑘𝑧 − 𝜔𝑡) +𝐸𝑥 𝑐𝑜𝑠 (𝑘𝑧 − 𝜔𝑡 +𝜋

2)

= 𝐸𝑥[𝑐𝑜𝑠(𝑘𝑧 − 𝜔𝑡) − 𝑠𝑖𝑛(𝑘𝑧 − 𝜔𝑡)]

In the above equation, it is clear that the y component lags the x component by a quarter

cycle. Thus, the net electric field vector maintains a constant magnitude and appears to

rotate in a circular pattern in the x-y plane. Various polarisation states with the

corresponding phase differences are shown in Figure 10.

Figure 10. Various polarisation states of light occurring as a result of the various phase difference (𝛿𝑥 −𝛿𝑦 and 𝛿𝑦 − 𝛿𝑥) between the two components of the electric field (along x- and y- axes) of equal amplitude

(𝐸𝑦 = 𝐸𝑥) [image taken from37].

Since Cartesian coordinates x, y and z are interchangeable and are relative directions, it

is standard to define the electric field components respective to the sample plane. The

plane perpendicular to the surface of the sample that contains the vector pointing the

direction of propagation of light is called the plane of incidence. Perpendicular to the

Theory Chapter 2

20

propagation vector is two mutually perpendicular vector components of the electric field

vector of (E), defining the polarisation of light (the x-y plane). These components of E are

therefore parallel and perpendicular to the plane of incidence and are named as “parallel”

(p) and perpendicular “senkrecht” (s), respectively.

2.3.2 Jones formalism representation of polarisation

R. Clark Jones introduced in 1941 a matrix algebra to describe the polarisation of light

and also the influence of optical elements on the light polarisation38. As discussed in the

previous section, the polarisation state can be represented by superimposing two

electromagnetic waves oscillating along x- and y- axes, respectively. The Jones vectors

define the amplitude (E) and phase (ϕ) of the electric field along the x and y axes (p and s).

𝐽 = (𝐸𝑥𝑒

𝑖𝜙𝑥

𝐸𝑦𝑒𝑖𝜙𝑦

) 2-18

Since Jones matrix algebra is applicable only for polarised light, the following discussion

is restricted for linearly polarised s and p waves. A linearly polarised light wave incident

onto a sample is denoted as s (or p) polarised when the electric field vector is

perpendicular (or parallel) to the plane of incidence (x-z). For example, the change in the

polarisation of the reflected light (Jr) from a reflecting sample can be written as the matrix

product of the reflection matrix R to the Jones vector of the incident light Ji.

where, R is the reflection matrix, and r is the reflection coefficient of s and p polarised

light.

𝐽𝑟 = 𝑅 × 𝐽𝑖 2-19

𝑅 = [𝑟𝑠𝑠 𝑟𝑠𝑝𝑟𝑝𝑠 𝑟𝑝𝑝

] 2-20

(𝐸𝑟𝑥𝑒

𝑖𝜙𝑟𝑥

𝐸𝑟𝑦𝑒𝑖𝜙𝑟𝑦

) = [𝑟𝑠𝑠 𝑟𝑠𝑝𝑟𝑝𝑠 𝑟𝑝𝑝

] × (𝐸𝑖𝑥𝑒

𝑖𝜙𝑖𝑥

𝐸𝑖𝑦𝑒𝑖𝜙𝑖𝑦

) 2-21

Theory Chapter 2

21

2.3.3 Light-matter interaction

Often, two physical quantities are used to describe the interaction of light and matter,

namely, the complex refractive index (N) and the dielectric function (ε). Both of them are

complex numbers.

𝑁 = 𝑛 + 𝑖𝑘 2-22

and

휀 = 휀1 + 𝑖휀2 2-23

Here n (real part) is the refractive index of the material and describes the phase velocity

of the light travelling through the material, and k (imaginary part) is called extinction

coefficient, describing the loss of wave energy to the material. Similarly, in the case of the

dielectric function, the real part denotes the dielectric capacity of material and the

imaginary part determines the dielectric losses. These two physical quantities are related

by

휀 = 𝑁2 2-24

In general, the optical response of the material is a direction-dependent property. In an

optically anisotropic material, ε can also be described by a second-rank tensor called

dielectric tensor or permittivity tensor ε.

휀 = (

휀𝑥𝑥 휀𝑥𝑦 휀𝑥𝑧

휀𝑦𝑥 휀𝑦𝑦 휀𝑦𝑧

휀𝑧𝑥 휀𝑧𝑦 휀𝑧𝑧

) 2-25

For a (non-magnetised) medium with a biaxial symmetry, the dielectric tensor can be

reduced to the principal dielectric constants (or principal refractive indices) given as

휀 = (

휀𝑥𝑥 0 00 휀𝑦𝑦 0

0 0 휀𝑧𝑧

) = (

𝑁𝑥𝑥2 0 0

0 𝑁𝑦𝑦2 0

0 0 𝑁𝑧𝑧2

) 2-26

Theory Chapter 2

22

For an optically isotropic (non-magnetised) material εxx = εyy = εzz. The dielectric tensor for

isotropic materials is given by

휀 = (

휀𝑥𝑥 0 00 휀𝑥𝑥 00 0 휀𝑥𝑥

) 2-27

The diagonal components of this tensor can be determined experimentally, e.g. by SE.

Many optical spectroscopy techniques, including SE and MOKE work in reflection

geometry. The diagonal components of the reflection matrix (equation 2-20) are directly

related to the principal component of the dielectric tensor (and to the optical constants)

and are determined by spectroscopic ellipsometry, while the off-diagonal components of

the reflection matrix are related to the off-diagonal components of dielectric tensor and

determined using magneto-optical Kerr effect spectroscopy or by generalized

spectroscopic ellipsometry or Müller matrix ellipsometry. Each of these polarimetry

techniques is discussed in detail in the following sections.

Selection rules

Another way to explain the interaction of light with matter is by taking into account the

absorption of a single photon by a single electron through electronic transitions. Like any

other physical process, the laws of conservation have to be followed in the process of

photons absorption. These conservation laws are often narrated as selection rules39.

Table 1. The selection rules of (magneto-) optical transitions

Energy 𝐸𝑓 − 𝐸𝑖 = ħ𝜔 The energy of the photon absorbed must be equal to the energy difference of the initial and final states

Momentum ħ𝜔

𝑐≈ 0

As a photon has negligible linear momentum compared to the electron, the linear moment of the electron must be conserved → “vertical transition.”

Spin ∆𝑠 = 0 Since photons carry no spins, during an electronic transition, the spin of the electron must be conserved.

Orbital angular momentum quantum number

∆𝑙 = ±1 The total orbital momentum is conserved. As a photon has 𝑙 = 1ħ, it means that the only allowed transition are s to p, p to d, d to f, etc.

Magnetic quantum number

∆𝑚𝑙 = ±1, 0

The total orbital momentum along z-direction (𝑚) must be conserved. The type of the absorbed photon is thus circularly left (-1), or right (+1) polarised, or linearly (0) polarised.

Theory Chapter 2

23

2.3.4 Spectroscopic ellipsometry

When linearly polarised light is incident on a sample surface at an oblique incidence angle,

the reflected light is in general elliptically polarised. In spectroscopic ellipsometry, the

change in the polarisation of the reflected with respect to that of the incident light is

measured. The name “ellipsometry” is related to the analysis of the elliptical polarisation

of the reflected light. Figure 11 shows the principle of ellipsometry.

Figure 11. The schematic of spectroscopic ellipsometry [image redrawn from40]

In a typical rotating compensator ellipsometer, the incident light is linearly polarised. Such

polarisation can be explained as the combination of p and s components of the electric

field oscillating in-phase. The amplitude of each component determines the orientation in

the quadrant. The reflected light from the surface of the sample is elliptically polarised,

which means that both the amplitude and the phase of the p and s components are altered

so that the projection of the net electric field vector forms an ellipse onto a plane

perpendicular to the light propagation direction.

The light interacts with the material following Maxwell’s equations, which in turn, give the

boundary conditions at each interface between two media. The boundary conditions

provide a predictable solution of the interaction of p and s components of the electric field

to the samples and can be used to obtain the Fresnel equations.

𝑟𝑝 =𝐸𝑟𝑝

𝐸𝑖𝑝=

𝑛𝑡 𝑐𝑜𝑠 𝜃𝑖 − 𝑛𝑖 𝑐𝑜𝑠 𝜃𝑡

𝑛𝑡 𝑐𝑜𝑠 𝜃𝑖 + 𝑛𝑖 𝑐𝑜𝑠 𝜃𝑡 2-28

𝑡𝑝 =𝐸𝑡𝑝

𝐸𝑖𝑝=

2𝑛𝑖 𝑐𝑜𝑠 𝜃𝑖

𝑛𝑡 𝑐𝑜𝑠 𝜃𝑖 + 𝑛𝑖 𝑐𝑜𝑠 𝜃𝑡 2-29

Theory Chapter 2

24

𝑟𝑠 =𝐸𝑟𝑠

𝐸𝑖𝑠=

𝑛𝑖 𝑐𝑜𝑠 𝜃𝑖 − 𝑛𝑡 𝑐𝑜𝑠 𝜃𝑡

𝑛𝑖 𝑐𝑜𝑠 𝜃𝑖 + 𝑛𝑡 𝑐𝑜𝑠 𝜃𝑡 2-30

𝑡𝑠 =𝐸𝑡𝑠

𝐸𝑖𝑠=

2𝑛𝑖 𝑐𝑜𝑠 𝜃𝑖

𝑛𝑖 𝑐𝑜𝑠 𝜃𝑖 + 𝑛𝑡 𝑐𝑜𝑠 𝜃𝑡 2-31

where, rp, rs, tp, and ts are the reflection and transmission coefficients for p and s

component θi and θt are the angle of incidence and transmittance, respectively; ni and nt

are the refractive index of incidence and transmission medium, respectively.

Often the SE samples have multiple interfaces, resulting in multiple reflections. The

superposition of these multiple reflections at the interface generates an interference

depending on the relative phase of each reflected light wave (see Figure 12).

Figure 12. The schematic of an optical model for an ambient / thin film / substrate structure, showing the reflected and refracted light at each interface. Using the Fresnel coefficients, the contribution of reflections from each interface can be calculated [image was taken from40].

The detector then measures this change in polarisation in terms of the ellipsometric

parameters Ψ and Δ. These parameters can be related to the Fresnel reflection

coefficients of p and s polarised light, resulting in the fundamental equation for

ellipsometry:

𝜌 =𝑟𝑝𝑝

𝑟𝑠𝑠= 𝑒−𝑖∆𝑡𝑎𝑛𝜓 2-32

Typically, SE is used to determine the optical constants (N) of a material and/or thickness

of thin films. Every material has unique optical constants which are directly related to the

electronic structure of the material. Hence optical constants are considered as a

fingerprint of materials. In ellipsometry, a suitable model describing the sample has to be

Theory Chapter 2

25

developed to fit the experimental data Ψ and Δ. Using Fresnel’s equations, predictive Ψ

and Δ spectra from the model are calculated. If the optical constants or the thickness of

the film are unknown, the estimated values or range are considered as input parameters.

A non-linear regressive approach (for example the Levenberg-Marquardt method) is then

used to find the best match of the calculated Ψ and Δ spectra to the measured ones. The

quantitative comparison of the best match between the calculated spectra and the

measured ones is expressed in terms of Mean Squared Error (MSE). The lower the MSE

is, the better the fit.

𝑀𝑆𝐸𝑁𝐶𝑆 = √1

3𝑛 − 𝑚∑[(

𝑁𝐸𝑖 − 𝑁𝐺𝑖

0.001)2

+ (𝐶𝐸𝑖 − 𝐶𝐺𝑖

0.001)2

+ (𝑆𝐸𝑖 − 𝑆𝐺𝑖

0.001)2

]

𝑛

𝑖=1

2-33

where, n is the number of wavelengths in the measured spectral range, m is the number

of fit parameters in the optical model, and N = cos(2Ψ), C = sin(2Ψ)·cos(Δ),

S = sin(2Ψ)·sin(Δ) are the Mueller matrix components for an isotropic material. The

discussion of the Mueller matrix is beyond the scope of this work and can be found in

reference37. Typical precision in measuring the N, C, and S parameters is ~0.001, hence

this factor is included in the MSE definition, implying that ideal data modelling will have

an MSE of ~1.

2.3.5 Magneto-optical Kerr effect spectroscopy

Michael Faraday, in 1846, discovered a magneto-optical effect, which was later called the

Faraday effect. He observed that a linear polarised light beam changes its polarisation

state after passing through a piece of glass in the presence of a magnetic field. Later,

John Kerr reported a similar effect for the reflected light from a piece of polished iron and

found that this effect is proportional to the magnetisation in the sample41,42. Today, the

characterisation techniques based on the Faraday and magneto-optical Kerr effect

(MOKE) are prominent measurement tools in the field of magnetism.

In MOKE (or Faraday effect) based measurement techniques, the polarisation change of

the reflected (transmitted) light is measured in terms of the complex Kerr (Faraday)

rotation (θ). Practically, θ is measured as the ratio of off-diagonal components to either

one of the diagonal components of the reflection coefficient matrix. (equation 2-34).

Theory Chapter 2

26

𝜃 =𝑟𝑝𝑠

𝑟𝑠𝑠 or 𝜃 =

𝑟𝑠𝑝

𝑟𝑝𝑝 2-34

For the ease of presentation, later only the rp component is considered for the calculations.

However, the mathematical equations remain identical for the rs component, as well.

Additionally, all the MOKE experiments presented in this thesis were conducted with

incident light of p polarisation. In general, the reflection coefficient and the dielectric

function (optical constants) of a material are complex numbers. Therefore, the Kerr

rotation (ΘK) is also expressed in terms of complex numbers. The real part of MOKE (or

Faraday) rotation represents the tilt of the polarisation ellipse (θK or θF) and the imaginary

part denotes its ellipticity (ηK or ηF). η is caused by the difference in the absorption of right

circularly and left circularly polarised light, whereas θ is caused by the difference in the

phase velocity between the left circular and right circular polarised light in a magnetised

medium.

𝛩𝐹 = 𝜃𝐹 + 𝑖𝜂𝐹 2-35.

𝛩𝐾 = 𝜃𝐾 + 𝑖𝜂𝐾 2-36.

Based on the relative orientation of magnetisation and incident plane of light, MOKE can

be classified into three categories: polar (P-MOKE), longitudinal (L-MOKE), and

transversal (T-MOKE). These geometries correspond to the magnetisation aligned normal

to the sample surface (P-MOKE), or in the sample plane but either in the plane of

incidence (L-MOKE) or perpendicular to it (T-MOKE). Figure 13 summarises the basic

geometries of MOKE.

Figure 13. The three geometrical configurations for MOKE, namely polar (a.), longitudinal (b.) and transversal (c.).

In the case of P-MOKE and L-MOKE, the magnetisation lies in the plane of incidence,

which influences the off-diagonal reflection coefficients rps and rsp. In T-MOKE the

Theory Chapter 2

27

magnetisation is perpendicular to the plane of incidence and influences only the diagonal

component reflection coefficient rpp (and/or rss). As a result, only the reflected light intensity

changes as a function of the magnetisation, but no Kerr rotation or ellipticity is measurable

in this configuration.

Phenomenological explanation of MOKE

According to the selection rules described above, a circularly polarised photon can induce

an electronic transition between two states with the same spin polarisation. If the density

of spin-up and spin-down states are not equal, then a different number of left circular

polarised and right circular polarised photons will be absorbed, which in turn results in the

magneto-optical (MO) signal of ferromagnetic materials.

On an atomic scale, the magnetisation in ferromagnetic materials can be seen as a

perturbation that lifts the degeneracy of the electronic states by spin-orbit coupling and

exchange interaction43. Spin-orbit coupling is the interaction between the electron spin

magnetic moment and the orbital angular momentum. This is similar to the Zeeman effect,

but it is caused by the internal magnetic field generated by the electrons orbital motion.

Figure 14 exemplifies these two effects on the energy diagram of a 3d transition metal

ferromagnet. The left side of Figure 14 shows the lifting in the degeneracy of the d and p

electronic states due to the spin-orbit coupling. The energy distance, ESO, between the

split states corresponds to the strength of the spin-orbit interaction. The exchange

interaction takes place among the spins of electrons in two adjacent atoms, resulting in

parallel spin alignment (Pauli principle) in the case of a ferromagnet. The exchange

interaction splits the spin-up and the spin-down electronic states by the exchange energy

Δex, as shown in Figure 14. All possible magneto-optical electronic transitions between p

and d states of a 3d ferromagnet by blue and red vertical lines (depending on the circular

polarisation state of the impinging photon) are shown in Figure 14.

Theory Chapter 2

28

Figure 14. The schematic energy state diagram of a 3d-ferromagnet, showing optical transitions induced by left (blue) and right (red) polarised photons for a system where only spin-orbit coupling is present (left diagram) and for a system where spin-orbit coupling and exchange interaction is present (middle diagram). The notation in the | ⟩ brackets contains the orbital number (𝑙), magnetic number (𝑚),

and spin orientation (↑ or ↓). The right side diagram shows the corresponding absorption spectra of left and right circular polarised light. [adapted from44]

Using ab-initio calculations based on the Kubo linear response formalism45, it was shown

that the presence of spin-orbit coupling and/or exchange interaction in a material yield to

the appearance of off-diagonal elements in the conductivity tensor. For a magnetised

sample, the dielectric tensor can be written as follows:

휀 = 휀𝑥𝑥 (1 𝑖𝑄𝑍 −𝑖𝑄𝑌

−𝑖𝑄𝑍 1 𝑖𝑄𝑋

𝑖𝑄𝑌 −𝑖𝑄𝑋 1) 2-37

here, Q is the so-called magneto-optical Voigt constant, which, for a given value of the

externally applied magnetic field is material-specific. The Voigt constant can have different

values along the x-, y-, and z- space directions.

In this thesis, only the P-MOKE geometry is employed, and in order to keep the

corresponding calculations simple, the magnetisation is placed parallel to the z-axis. By

assuming that Q is parallel to the direction of magnetisation

Theory Chapter 2

29

(

𝑄𝑋

𝑄𝑌

𝑄𝑍

) = 𝑄 ∥ 𝑀 2-38

the Q vector will only have a component along the z-axis. For an isotropic medium with

the magnetisation pointing along the z-direction, equation 2-37 can be reduced to the

following dielectric tensor

휀 = 휀𝑥𝑥 (1 𝑖𝑄𝑍 0

−𝑖𝑄𝑍 1 00 0 1

) 2-39

The refractive index for the left-circularly polarized light (Nl), and right-circularly polarized

light (Nr) can be given as

𝑁𝑟 = 𝑁(1 −𝑄

2⁄ ) and 𝑁𝑙 = 𝑁(1 +𝑄

2⁄ ) 2-40

As mentioned previously, the Faraday effect arises. Thus the complex Faraday rotation

after transmission through a medium of length L:

𝛩𝐹 = 𝜃𝐹 + 𝑖𝜂𝐹 =𝜔𝐿

2𝑐(𝑁𝑟 − 𝑁𝑙) = −

𝜔𝐿

2𝑐𝑄𝑁

2-41

here, c and ω is the velocity and angular frequency of light. Similarly, the Kerr effect which

is measured in the reflection mode the resultant reflection coefficients for the two circular

polarised state of light is

𝑟𝑟 = 1−𝑁(1−

𝑄2⁄ )

1+𝑁(1−𝑄

2⁄ ) and 𝑟𝑙 =

1−𝑁(1+𝑄

2⁄ )

1+𝑁(1+𝑄

2⁄ ) 2-42

Thus, the complex Kerr rotation angle due to these reflection coefficients, i.e. rr an rl is

Θ𝐾 = θ𝐾 + iη𝐾 = arctan (𝑖𝑟𝑟 − 𝑟𝑙𝑟𝑟 + 𝑟𝑙

) ≈ 𝑖𝑟𝑟 − 𝑟𝑙𝑟𝑟 + 𝑟𝑙

2-43

substituting the reflection coefficient from equation 2-42 in 2-43 one obtains the reflection

on a bulk (half space) sample:

Theory Chapter 2

30

Θ𝐾 = θ𝐾 + iη𝐾 = 𝑖𝑁𝑄

1 − 𝑁2 + 0.25𝑁2𝑄2≈ 𝑖

𝑁𝑄

1 − 𝑁2 2-44

In case of the reflection from a thin magnetic single layer on a non-magnetic substrate for

a thickness less than the wavelength d ≪ λ

Θ𝐾 = θ𝐾 + iη𝐾 == −ω

2C·

𝑁2𝑄𝑑

1 − 𝑁2

2-45

For a layer stack samples, a more complex optical model needs to be used and the

relation between θ and Q will change accordingly (see ref.46).

Often, the results are presented in terms of optical conductivity instead of the dielectric

function, as the optical conductivity is amenable to be calculated from the Kubo linear

response formalism (see, e.g. ref.44,45). The relation between ε and σ is given in

equation 2-46

휀 = 1 +4𝜋𝑖

𝜔𝜎

2-46

where, ε is the dielectric tensor and σ is the conductivity tensor. In the case of P-MOKE

of an optically isotropic sample (as the samples considered in this thesis) the conductivity

tensor (σ) is scaled assuming an isotropic sample.

𝜎 = (

𝜎𝑥𝑥 𝜎𝑥𝑦 0

−𝜎𝑥𝑦 𝜎𝑥𝑥 0

0 0 𝜎𝑥𝑥

) 2-47

Due to the symmetry conditions, the components σxz, σyz, σzx, and σzy can be equated to

zero (σxz = σyz = σzx = σzy = 0). The diagonal components (σxx = σyy = σzz) are symmetric

and represent the optical conductivity of the material, whereas the off-diagonal

components (σxy = -σyx) are accountable for MOKE. As mentioned in the previous section

“Light-matter interaction”, the dielectric tensor elements are complex quantities.

Therefore, the conductivity tensor elements are also complex quantities. Knowing the

components of the conductivity tensor, one can write the complex Kerr rotation as47:

𝜃𝑘 + 𝑖𝜂𝐾 =

𝜎𝑥𝑦

𝜎𝑥𝑥√1 +4𝜋𝑖𝜔 𝜎𝑥𝑥

2-48

Experimental Chapter 3

31

Chapter 3: Experimental

In this chapter, the sample preparation and the measurement techniques that were used

in the framework of this thesis are introduced. The basic principles and the technical

properties of the equipment for the fabrication, processing, and characterisation of the

samples are explained. The intention is only to provide a short overview and further details

regarding any particular method can be found in the listed references.

3.1 Sample preparation

The samples investigated in this thesis were prepared by magnetron sputtering. The

CoFeB thin film samples characterised in the chapter “The electrodes: 3d-transition metal

boride” (chapter 4) with the corresponding capping layers were prepared using the

sputtering facility available under the professorship of “Functional Magnetic Materials”,

headed by Prof. Olav Hellwig. The MTJ layer stacks and the IrMn / CoFeB bilayer

discussed in the chapter “Exchange bias” (chapter 5) were deposited at “Singulus

Technologies AG”.

The fundamental principle of sputtering is based on momentum interchange between the

accelerated ions of inert gas and the target material. Typically, argon (Ar) gas serves for

this purpose, which is ionised to Ar+ ion plasma by the application of a high voltage

between the metallic target as cathode and anode in close vicinity. The energised Ar+ ions

strike the target material atoms and provide them with the necessary kinetic energy to

reach the substrate. In the magnetron sputtering a toroidal magnetic field generated by

permanent magnets underneath the cathode surrounds the target (see Figure 15). Due

to the presence of the magnetic field, a dense plasma is confined over the target, resulting

in a higher yield of the ejected target material in comparison to other physical sputtering

methods.

Due to industrial privacy norms, the sputtering equipment used at “Singulus Technologies

AG” is not discussed in this documentation. However, the equipment available at our

institute is discussed in detail and fundamentally similar to the one at Singulus

Technologies. The sputtering chamber is custom made with four confocal configured

sputtering targets. The magnetrons of each target are powered by a DC(/RF) source that

enables to deposit the layers from a single target or from more targets. For the conformal

deposition of films, the substrate is rotated above the target at a distance of around


32

250 mm. This results in a uniform deposition across the substrate wafer with a maximum

diameter of 100 mm. The deposition was conducted under HV conditions, with a base

pressure of ~2×10-4 Pa and argon as the working gas. The argon pressure can be

regulated by the pump rate and the argon flow in the chamber. The pump rate is controlled

with a throttle adjustment at a turbopump. All the depositions were performed at a constant

argon pressure of 0.35 Pa. The nominal deposition rate and the deposited layer thickness

were monitored using a quartz microbalance placed close to the substrate holder. The

resonance frequency of the quartz oscillator shifts due to its increasing mass during the

deposition. The calibration of the quartz crystal microbalance was performed with XRR

(X-ray refractometry) on the deposited samples. The thickness-modulated periodic

oscillations measured in X-ray reflectometry were evaluated with the GenX fitting tool for

extracting the thickness of the sputtered CoFeB film48. The uncertainty of the simulated

thicknesses of the CoFeB films and of the capping layers was in the order of 10% of the

nominal values. This deviation from the nominal thickness is mainly due to the short

sputter time and inherent inaccuracy in measurements and simulation.

Figure 15. Schematic representation of the magnetron sputtering. The argon ions (in red) are responsible for the target etching. The ejected particles (in grey) are sputtered towards the substrate. The direction of the magnetic field used for confining the electrons and the ions close to the target is illustrated by the blue arrows.

3.2 Annealing

In this work, three different types of annealing systems were used for annealing the

samples, namely: in-situ SQUID-VSM system, in-situ macro-MOKE system, and laser


33

irradiation system. The two foremost systems can provide ultra-high vacuum (UHV)

conditions. Each of these methods is discussed in the following section.

3.2.1 In-situ SQUID-VSM annealing

The Quantum Design's Superconducting Quantum Interference Device–Vibrating Sample

Magnetometry (SQUID - VSM) Magnetic Property Measurement System-3 (MPMS-3), is

equipped with a sample holder with a resistive element for heating. This option allows

heating from ambient to 750°C under UHV conditions. The central part of the oven is an

insulated heater stick assembly with a sample space of 5 mm in width which is mounted

at the end of a dedicated sample rod and introduced directly into the sample chamber.

The sample is secured on the oven stick holder with zircar cement (ACA5) in conjunction

with tightly wrapped copper foil around it, following the recipe provided in49.

Figure 16. Oven sample mounting platform with a mark at 66 mm to place the sample on oven heater stick [image was taken from49].

3.2.2 In-situ macro-MOKE annealing

A special heating assembly was designed in the frame of this thesis to anneal larger

samples (3 × 2) cm2 in the presence of magnetic field (up to 500 mT) in UHV environment,

unlike in-situ SQUID-VSM where the sample size is strictly restricted to (1 × 1) mm2. This

heater is assembled in the macro-MOKE spectroscopy system, which not only gives the

freedom of annealing in UHV but also allows for depositions and in-situ MOKE

characterisation. However, in-situ MOKE measurements are at this point beyond the

scope of this work. The heater is carved from a solid copper rod, to ensure homogeneous

heating over the whole sample. The thermal energy is provided with a halogen lamp of

250 W mounted in a cavity, as shown in Figure 17, that enables to reach a maximum

temperature of 600°C. For the precise measurement of the temperature, a type-K


34

thermocouple (chromel–alumel) is placed right below the sample holder and read with an

Omron (E5EN-H) high-resolution temperature controller. To improve thermal conductivity

between the sample holder and sample a thin layer of silver epoxy was used.

Figure 17. CAD modelled isometric view (a.) and transverse view (b.) of the heater assembly developed for the macro-MOKE system.

3.2.3 Laser annealing

Figure 18 shows the sketch of the experimental setup used for the laser annealing at the

Laserinstitut of the Hochschule Mittweida. At a much higher magnification, the heat profile

induced by the laser beam into an exemplary layer stack is sketched. The laser annealing

experiments were performed by Ms. Sandra Busse. The planning of the experiments was

discussed in the DFG-3D-TMR project team. An Nd: YAG laser with an emission

wavelength of 1064 nm was used to locally heat the samples. The laser system is capable

of supplying both continuous wave (CW) and pulsed wave (PW) laser radiation. The laser

beam was first attenuated and then rapidly deflected by using a galvanometer scanner

and focused by an f-theta objective with a focal length of 80 mm. The resulting focal radius

(86%) of the laser beam was 10 µm. For annealing a larger area raster scanning was

performed with 2.5 µm distance between each scanned line, providing an adequate

overlap of consecutive line scans. The external magnetic field was provided by

electromagnets in the Helmholtz coil configuration, generating a maximum magnetic field

of 420 mT. Often the fluence delivered by laser is given in terms of laser intensity, i.e.,

power per unit area (W·cm-2)50. The following formula can be used to convert the laser

power to laser intensity:

𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦 =2𝑃

𝜋 𝑟2

3-1

where, P is the laser power delivered to the sample, r is the radius of the laser beam, and

multiplier 2 is due to the integration assuming a Gaussian laser profile. All the results for

the laser annealing presented in this thesis are thus given in terms of laser intensities.

a. b.

.


35

Figure 18. Schematics of the experimental setup for Nd: YAG laser [image prepared for own publication51].

3.3 Measurement techniques

3.3.1 Structural and topographic characterisation

X-ray diffraction

The atomic stacking has an important implication on the properties of ultrathin and thin

films. For the crystallographic characterisation of the thin films and layer stacks, X-ray

diffractometry (XRD) was used. The thin film thickness was measured by X-ray

reflectometry (XRR). The structural characterisation presented in this work was conducted

using the diffractometer 3000PTS from Seifert-FPM, the Rigaku SmartLab from Rigaku

Corporation, and the KMC-2 Diffraction station52 at Bessy II (Helmholtz-Zentrum Berlin).

Although three different tools were used for the X-ray diffraction measurements, the core

principle of all three types of equipment is the same.

In general, the laboratory X-ray diffraction instrument utilises an X-ray source consisting

of an evacuated tube, in which electrons are emitted from a heated tungsten filament.

These electrons are accelerated by an electric potential (here U = 40 kV at I = 30 mA) to

strike on a water-cooled copper target, producing X-ray radiation with wavelengths

characteristic for the electronic levels in the copper target (mainly copper Kα and Kβ) and

polychromatic background radiation (Bremsstrahlung). For this reason, the instrument is

further equipped with a monochromator crystal to provide beam monochromaticity (with


36

solely Cu-Kα radiation). A scintillation detector mounted on the 2θ circle of the goniometer

is then used to measure the diffracted angular dependent intensity from the sample. The

recorded scan consists of constructive interference peaks from adjacent lattice planes

which fulfil the Bragg equation

𝑛𝜆 = 2𝑑 𝑠𝑖𝑛 𝜃 3-2

where, is the wavelength of the X-ray radiation, d is the interplanar spacing, n is an

integer number and θ is the diffraction angle.

The acquired peak positions from a θ-2θ scan can be compared with reference databases

for particular materials and the relevant Miller indices can be assigned to the peaks. In

this study, the X-ray powder diffraction files (PDF) were used as references, which were

published by the International Centre for Diffraction Data (ICDD). The Bragg peak in a θ-

2θ scan from a particular crystallite can be used to estimate the crystallite size (L) using

the Scherrer formula

𝐿 ≈𝐾𝜆𝐶𝑢

𝛥(2𝜃)𝑐𝑜𝑠𝜃 3-3

With a shape factor of crystallites K ≈ 0.9 considering cubic crystallites, the wavelength of

X-ray radiation Cu ≈ 0.154 nm and Δ(2θ) as the full width at half maximum (FWHM) of

the reflex at θ, given in radians53.

Figure 19. Drawing of the diffractometer showing the general scheme of various goniometers and measurement axes [image taken from54].


37

An X-ray diffractometer can also be employed to perform XRR measurements. In XRR

measurements, a standard θ-2θ scan is performed but at small angles, below 2θ ≈ 8°.

Due to interference effects affecting the reflected X-ray radiation from the sample surface

and its interfaces, oscillations are observed in the reflectance curve plotted against the 2θ

angle along with exponential decay of the intensity. The measured oscillations contain

information about the thickness, density of the material, and interface roughness. A

careful modelling of the measured reflectivity curves can unwrap the above-mentioned

information about the investigated sample. For this purpose, an open-source software

GenX was used48. The general equation for XRR data fitting is shown below

𝑠𝑖𝑛2 𝜃𝑖 =𝜃𝑐

2 + (𝑛𝑖 + ∆𝑛)2𝜆2

4𝑡2

3-4

θi is the observed position of the maximum or minimum of the ith interference fringe, θc the

critical angle for total reflection, ni is an integer, Δn = 0.5 or 0 for the maximum and

minimum, respectively, is the X-ray wavelength, and t is the film thickness.

Atomic force microscopy

The atomic force microscope was developed by Binning and coworkers55, shortly after the

introduction of the scanning tunnelling microscopy. For atomic force microscopy, a sharp

tip prepared on a cantilever is brought in close proximity to the sample surface. The forces

acting between the tip and the sample surface (van der Waals forces, capillary forces,

Pauli repulsion, and Coulomb forces) induce a deflection of the cantilever when the tip is

scanned over the sample. The deflection of the cantilever is detected optically by a split

photodiode, which measures the position of the reflected laser beam from the cantilever

(the scheme is shown in Figure 20). A typical measurement mode is an intermittent mode,

sometimes also referred to as the AC mode due to the oscillation of the tip close to its

resonance frequency and the alternating contact of the tip with the surface, which takes

place right at the turning point of the oscillation. This allows the best possible resolution

while maintaining the tip durability as well as the sample integrity since any friction

between the tip and the sample is, in principle, avoided. When the cantilever and sample

are close to each other during each oscillation, the tip moves through an interaction

potential created by long-range attractive and short-range repulsive forces between the

sample and tip. This causes changes in the amplitude, phase and resonance frequency

of the oscillating cantilever and thus topography, amplitude and phase information


38

(material dependent properties) are collected simultaneously. Further details about atomic

force microscopy can be found, e.g. in55. All the topographical images of the samples

discussed in this work were recorded using an Agilent 5500 scanning probe microscope

enclosed in an isolated acoustic chamber. The atomic force microscope probe consisted

of a silicon tip with an apex diameter of around 10 nm and a cantilever with the resonant

frequency of about 12 kHz and was operated in the AC mode.

Figure 20. A basic principle of atomic force microscopy [image was taken from56].

3.3.2 Optical and magneto-optical spectroscopy

Within the presented work, two spectroscopy methods were employed to characterise the

samples, i.e. spectroscopic ellipsometry (SE) and magneto-optical Kerr effect (MOKE)

spectroscopy.

Variable Angle Spectroscopic Ellipsometry

Spectroscopic ellipsometry measurements were performed using M-2000 ellipsometer

from J. A. Woollam over the same spectral range of 0.7 eV to 5 eV. The complex refractive

index was deduced from the measured Ψ and Δ spectra by the modelling and simulation

tool completeEASE®. For reliance in the modelled results, the spectra are measured for

different angles between 50° - 70° in steps of 5° and different thicknesses of the

investigated layer. The multi-sample analysis with coupled optical constants (n and k) was

performed to fit the model to the data with keeping only the thickness parameter of CoFeB

to vary independently in the close range to XRR measured thicknesses. A similar

procedure was followed for both the annealed and as-deposited samples. The detailed


39

description of optical models used in this thesis is presented in the results section (section

4.2 and 4.3) as the optical modelling approach has to be tailored according to the layer

stack of the investigated samples.

Figure 21. The spectroscopic ellipsometry setup M2000 for measuring the complex ratio (𝜌) of the Fresnel reflection coefficients.

Magneto-Optical Kerr Effect Spectroscopy

The polar Kerr rotation (θK) and ellipticity (ηK) spectra were measured with a home-built

MOKE spectrometer. The spectrometer measures both rotation and ellipticity

simultaneously by means of the polarisation modulation method using a piezo-birefringent

modulator, also known as a photoelastic modulator (PEM). A xenon lamp (75 W) is used

to provide the necessary spectral range from near-infrared (0.7 eV) to ultraviolet (5 eV)

regime. The sample is illuminated with the linear polarised light through a bore in the pole

shoe of the magnet. The reflected light is then guided through the PEM followed by an

analyser (rotated by 45° with respect to the polariser), and a monochromator (see Figure

22). The setup is equipped with an InGaAs diode for the IR spectral region (0.7 eV to

1.4 eV) and a photomultiplier for the visible to UV region (1.5 eV to 5 eV). The spectra

discussed in the thesis were measured at RT in the photon energy from 1.5 eV to 5 eV

with an applied magnetic field ~1.7 T, if not specified otherwise. A detailed discussion on

the extraction of θK and ηK from the measured voltage on the photodetector can be found

in reference57.

The off-diagonal component (εxy = ε1xy + i ε2xy), reflecting the magneto-optical response of

the films, is calculated from the recorded θK and ηK, by designing the layered optical model

using the optical constants of the constituent layer and fitting θK and ηK spectra with a

method described elsewhere58.


40

Figure 22. Schematic diagram of the magneto-optical Kerr effect spectrometer used in this thesis for measuring Kerr rotation (θK) and ellipticity (ηK).

3.3.3 Magnetometry

The magnetometry measurements were performed with the SQUID-VSM MPMS-3 from

Quantum Design and the NanoMOKETM2 system from LOT-Quantum Design GmbH.

Superconducting Quantum Interference Device–Vibrating Sample Magnetometry

The SQUID-VSM magnetometer is the most sensitive magnetic characterisation

technique developed to date. This technique is based on two principles: quantisation of

the magnetic flux in a superconducting loop (with the magnetic flux quantum ∅ = h/2e)

and the usage of a pickup coil linked with parallel Josephson junctions.

In a SQUID-VSM magnetometer, the sample is vibrated through an external magnetic

field. The change in the magnetic flux created by the moving sample is measured with a

superconducting pick-up coil which converts the changes in magnetic flux to a periodic

voltage signal. This voltage is then precisely measured by using two parallel Josephson

junctions. When a constant biasing voltage is applied on the two parallel Josephson

junctions the current flowing through both arms is equal. The magnetic flux created by the

moving sample imbalances the current flowing through the two junctions. An additionally

applied bias voltage is required to reestablish the equilibrium loop59. The required bias

provides the means to measure the magnetic moment of the sample. The magnetisation

is usually specified as the magnetic moment per volume of the measured sample. The

volume of the sample is determined by taking the area of the substrate and the thickness

of the magnetic layer into account. Figure 23 shows the cross-sectional view of the


41

SQUID-VSM. The substrate, the sample holder, and the glue generate a diamagnetic or

paramagnetic background signal, which is usually recorded along with the M(H) loop data.

Such a (dia-) paramagnetic background was subtracted using the data post-processing

software Origin® 2018. Apart from the background signal, the measurements performed

with the MPMS SQUID-VSM reveal some additional artefacts, for instance, the

remanence magnetisation, flux creep being present at low magnetic fields. These

artefacts were tackled by degaussing and/or warming up the superconducting coil

magnets. As the magnetic field in MPMS is not directly measured with a Hall sensor but

via the field-to-current conversion factor (B/I ratio) of the superconducting coil, a regular

calibration is required. Further discussion of these artefacts are beyond the scope of the

present thesis but can be found in detail elsewhere60,61.

Figure 23. A cross-sectional view of the utilised MPMS SQUID - VSM MPMS-3 setup [image taken from62].

Magneto-optical Kerr magnetometry

In order to investigate magnetic the properties of small patterned samples, for instance,

the isolated square patterns of (0.5 × 0.5) mm2 used in the study of various parameters

of laser irradiation, the NanoMOKETM2 magnetometer proved to be beneficial. Such

samples are not suitable for SQUID-VSM measurements since the measured moment is

then averaged over the entire sample. NanoMOKETM2 enables to probe the local

magnetisation with a typical spot size of 100 µm. A schematic image of this setup is

shown in Figure 24.

As discussed in section 2.3.5, the reflected light from a magnetic sample changes its

polarisation state depending on the sample magnetisation. In Nano-MOKETM2, the Kerr


42

rotation angle (θK) is estimated using the nulling method. Here, a linear polarised light

beam is incident on the sample and reflected at an angle of 45°. The rotation angle of the

main axis of polarisation ellipse after reflection with respect to the initial polarisation

direction is estimated as follows: initially, the analyser is rotated to minimise the signal on

the photodetector in the absence of magnetic field; subsequently, a magnetic field of up

to 500 mT is applied in the sample plane using an electromagnet. The rotation of the

polarisation axis induced by the sample magnetised in the applied magnetic field will

cause a change in the light intensity on the photodetector. With this technique, only

relative changes in the magnetisation can be measured, but no absolute values.

Moreover, the sensitivity and information depth is subject to the penetration depth of the

light in metal films. The NanoMOKETM2 system is capable of performing both longitudinal,

transverse and polar MOKE measurements. For the experiments discussed in this thesis,

only the longitudinal configuration of MOKE was utilised.

Figure 24. Top view of the Nano-MOKETM2 magnetometer. The yellow arrows point out the optical components in the optical path of the laser shown with the red dashed line.

3.3.4 First order reversal curves (FORC) method

In order to understand the response of the magnetic field sensor, it is crucial to understand

the magnetic properties of the individual constituent magnetic layers. For instance, the

magnetic reversal (switching field distribution) and offset in the hysteresis loop of the free

layer is strongly influenced by the magnetic state of the pinned layer. The conventional

methods such as M(H) loop measurements (performed by SQUID-VSM or MOKE) are

very powerful but can be ambiguous in the case of a complex multilayer stack. In such a

case, the M(H) loop is a combination of several magnetisation reversals of the various


43

constituent layers. It is desirable to measure the reversal of the individual layers and to

separate the influence on the magnetic properties of the layers coming from the interaction

between the layers. Such shortcomings of the M(H) loop can be removed by using a type

of hysteresis measurement technique called Magnetic First Order Reversal Curves (M-

FORC)63. The M-FORC technique is extensively used by researchers studying

palaeomagnetism; however, its strength was explored only in a few studies in the field of

nano-magnetism and thin film magnetism. In the year 1999, C.R Pike et al. demonstrated

the application of M-FORC for characterising interactions among fine magnetic particles64.

Nowadays the technique is used as a qualitative fingerprinting65,66 method of the

reversible and irreversible switching67,68, interactions69,70, and coercive field

distributions71, etc. for various types of magnetic samples. This technique is not restricted

to the magnetisation measurements but has been modified and implemented to study the

hysteretic magnetoresistance72 (MR-FORC) and magneto-optical73 (MOKE-FORC)

properties of magnetic multilayer samples. This work focuses only on the magnetisation

measurements for recording the reversal curves. For simplicity reasons, hereafter M-

FORC will be referred to as FORC.

Measurement and analysis of FORC diagram

FORC loops are progressively recorded semi-minor loops of the magnetisation as a

function of the applied magnetic field. The sample is first magnetically saturated by

applying an external magnetic field equal to Hs (cf. Figure 5). The magnetic field is then

reduced to a certain reversal field Ha and swept back to Hs in regular field steps of Hb.

This procedure is repeated for several values of Ha generating a series of magnetisation

loops with M(Ha, Hb), known as FORC loops. Often, the values Ha and Hb are chosen to

be regularly spaced, enabling to plot M(Ha, Hb) on an equally spaced grid of Ha and Hb.

The FORC distribution is the second-order derivative of M(Ha, Hb) with respect to Ha and

Hb, respectively (see Eq .3-5). The calculated derivative when plotted as a function of Ha

(x-axis) and Hb (y-axis) generates a contour plot called FORC plot, but it is common

practice to rotate the plot by changing coordinate axes from Ha and Hb to HC = (Hb–Ha)/2

and HI = (Hb–Ha)/2, which represent coercive field and interaction field distribution.

𝜌𝐹𝑂𝑅𝐶 = 𝜕2𝑀(𝐻𝑎, 𝐻𝑏)

𝜕𝐻𝑎𝜕𝐻𝑏

3-5


44

As the FORC plots are the second-order derivative of magnetisation, even minute

changes in the magnetisation will significantly be enhanced. Therefore, very sensitive and

accurate measurements of magnetisation are required. Additionally, a regressive

smoothing factor (SF)74 is implemented in the FORC numerical differentiation algorithm

in order to suppress noise-related artefacts. The FORC results presented in this thesis

were processed with ButterFORC, now called XFORC75. This software was kindly

provided by Dr. Xiang Zhao and was modified to adapt the requirements of the system

studied in thesis work. The FORC loops measured at RT for

Si / SiO2(100 nm) / Ni81Fe19(20 nm) using SQUID-VSM (for Ha = Hb = 0.5 mT) are shown

in Figure 25 a and the FORC contour showing the reversible and irreversible

magnetisation for this sample in Figure 25 b. The maximum (in red) shown in the contour

plot at HC = 2.2 mT (at HI = 0 mT) denotes the shading ‘‘irreversible’’ magnetisation

components of the Ni81Fe19 film. The distribution around this contour presents the

distribution of the coercive field. Also, this ridge is symmetric around HI = 0 mT and

suggests the absence of any exchange bias.

Figure 25. An exemplary first order reversal curve (FORC) recorded for Si / SiO2(100 nm) / Ni81Fe19(20 nm). The reversal magnetic field (Ha) and regular magnetic field (Hb) are shown in red and black solid dots, respectively (a.). The major hysteresis loop (MHL) is shown in the red dashed line. (b.) The FORC distribution calculated from the measured FORC loops. The contour denotes the maxima of the distribution at the “irreversible” located at about HC = 2.2 mT, HI = 0 mT.

Within the course of this thesis, the measurement routines of FORC have been developed

and implemented for the analysis of the reversal behaviour of permalloy layers and, in

addition, for Fe3O4 nanoparticles used as targeted drug delivery systems76. The

advantage of this method will be extended to study the interaction of the free and pinned

layer in MTJ devices as a function of the annealing process.


45

3.3.5 XPS depth profiling

The chemical composition of the layers and interfaces in some of the layer stacks

discussed in this thesis was investigated with XPS depth profiling using a Thermo

Scientific NexsaTM surface analysis system. The samples were iteratively etched with a

built-in monoatomic and gas cluster ion source (MAGCIS), mounted at a 60° angle to the

normal incidence. Monoatomic Ar+ ions were accelerated to a constant energy of 1 keV,

onto a scanning area of (3.0 × 1.5) mm², leading to a mean sputter rate of approximately

0.045 nm·s-1. In a multi-element layer stack, the sputter rate cannot be maintained

constant throughout the depth profile; therefore, a fixed argon etching iteration time of

20 s was chosen, with a subsequent collection of XPS spectra. A dwell time of 5s was

given after each sputtering step to settle down the etched material. The iterations were

repeated until the SiO2 / Si substrate was reached. The XPS spectra were recorded using

the monochromatic Al Kα (1486.6 eV) radiation from the X-ray source, which was focused

on a spot diameter of 300 µm in the middle of the sputtered area. The emitted electrons

were collected at the normal angle at the pass energy of 151 eV, providing an energy

resolution of ~1.4 eV (full width at half maximum (FWHM) of the Ag3d5/2 peak). To prevent

charging of the sample, a dual-source argon-ion/electron flood gun with an electron

energy of ~0.3 eV was used. Ta4f, Co2p, Fe2p, B1s, Ir4f, Mn2p, Ru3d, Si2p, and O1s

XPS core-level spectra were collected in SnapShot mode after each sputtering step.

Finally, using Thermo Scientific AvantageTM acquisition and data analysis software, the

recorded data were fitted, taking into account all the investigated elemental peaks. Those

fits were merged to obtain the distribution profile of elements as a function of the sputtering

time (depth).

Figure 26. Process flow diagram of XPS-depth profiling.

The electrodes: 3d transition metal boride Chapter 4

46

Chapter 4: 3d-transition metal boride layers:

structural, electronic, and magnetic properties

In this chapter, layers based on 3d transition metal boride – CoFeB, which should act as

ferromagnetic electrodes in the MTJ, are investigated with spectroscopic techniques in

order to characterise their optical and magneto-optical response and to extract information

about the underlying electronic structure, conductivity, and magnetic properties.

Additionally, it sheds light on the influence of crystallisation of CoFeB alloy on such

properties. For the ease of presentation, the results are categorised in two sections

dealing with a thick and thin film, respectively. Some of the results presented in this

chapter were published in Physical Review B, DOI: 10.1103/PhysRevB.101.054438 and

Journal of Physics: Condensed Matter, DOI: 10.1088/1361-648x/ab4d2f.

4.1 Introduction

In the last few decades, 3d-transition metal borides have gained certain interest due to

their highly customisable mechanical, electrical, thermal, and magnetic properties

compared to generic 3d-transition metals or alloys77–79. One of such 3d-transition metal

borides is CoFeB, which has received special attention from a fundamental research point

of view, but also in industrial applications. The increasing interest in CoFeB alloys relates

to their atypical properties, such as structurally smooth growth80, soft magnetic

properties81, high spin-polarisation82, and very low Gilbert damping83, which makes them

especially suitable for magnetic tunnel junction devices84. By exploiting the benefits

mentioned above, S. Ikeda et al. presented a milestone improvement in tunnel

magnetoresistance ratio in CoFeB based MTJ of 355% at RT85. The improved TMR ratio

was ascribed to the improvement of the texture of the MgO barrier and the CoFe upon

post-annealing due to amorphous growth of the CoFeB during deposition. In the same

year, D. D. Djayaprawira et al. reported that a 20% inclusion of B in the CoFe maintains

it amorphous upon the deposition, thereby preventing any lattice mismatch issues at the

interface with MgO80. This allows the thin MgO to grow with a well-defined (001) texture

and to serve as a template for CoFeB crystallisation during a post-deposition annealing

process. In 2008, a TMR ratio of ~600% at RT was reported in Co20Fe60B20 / MgO

/ Co20Fe60B20 annealed at 525°C86. On the other hand, it has also been observed that

annealing at higher temperatures, although preferential from a crystallisation point of view,

could induce interlayer diffusion in the MTJs (see chapter 6), resulting in a degradation of


47

the TMR ratio. Therefore, it is interesting to understand the influence of annealing

temperature and composition in the crystallisation of CoXFe(80-X)B20 alloys in detail.

Previous studies mainly focused on the crystallisation of CoFeB alloys using conventional

techniques like X-ray diffractometry (XRD)87, transmission electron microscopy (TEM)88

and also some indirect methods such as resistivity, magnetoresistance89 or magnetometry

measurements90. However, these techniques inherit some limitations regarding the

sample volume required for obtaining a reliable signal (e.g. XRD), are invasive (TEM),

require complex microfabrication processes for realising devices (magnetoresistance

measurements) or provide only an indirect indication of the crystallisation (electrical and

magnetic measurements).

In the first part of this chapter, the dielectric tensor (including the diagonal optical and off-

diagonal magneto-optical components) of Co50Fe50, Co40Fe40B20, and Co60Fe20B20 are

presented, highlighting the changes in the spectra with composition and annealing

temperature. For the ease of modelling the magneto-optical constants, at first thick films

are discussed in section 4.2. These films can be treated as bulk-like, considering the

penetration depth of light in the considered spectral range. The influence of B inclusion

on the dielectric function of Co50Fe50 was modelled, supporting that the major changes in

the optical and magneto-optical spectra of CoXFe(80-X)B20 with annealing temperature are

due to changes in the crystalline structure. These results were further supported by

investigating the crystallisation of the layers using XRD and sheet resistance

measurements. Ellipsometry has proven to be a very sensitive method for investigating

thin films91, also providing the possibility of probing changes in the crystalline structure,

as investigated for Si92, diblock copolymers93 or organic photovoltaic devices94. So far,

only a few studies have regarded the optical and magneto-optical properties of CoFe

alloys95–97 or CoFeB98,99.

4.2 Thick films

Magnetron sputtering was used to deposit 100 nm of CoXFe(80-X)B20 in two stoichiometries

(X = 40 and 60), as well as Co50Fe50 on silicon wafers with native silicon oxide. The layers

were capped with 5 nm of Pt to prevent oxidation of the CoFeB. The deposition was

performed at RT with a base pressure below 2×10-4 Pa and Ar working pressure of

0.35 Pa. The wafers were diced in (1 × 1) cm2 pieces, and each piece was annealed for


48

30 min at temperatures in the range of 300°C to 600°C in steps of 50°C in the macro-

MOKE system. The Co50Fe50 sample served as a standard for crystalline CoFe, and no

further annealing was performed.

The diffractometry measurements (XRD, grazing incidence XRD (GIXRD), and XRR)

were conducted using the SmartLab diffractometer from Rigaku, in order to probe the

crystallisation and thickness of the films. The surfaces of the samples were studied by

atomic force microscopy. The sheet resistance of all samples annealed at different

temperatures was measured with the four-point probe (4-PP) measurement technique.

In order to extract the dielectric function (εxx = ε1xx + i ε2xx) of CoFeB from the measured Ψ

and Δ spectra, an optical model analogous to the physical layer structure was constructed.

Thus, “Si / SiO2(1.8 nm) / CoFeB(tCoFeB) / Pt(tPt) / surface roughness” layered optical

model was built using the reported dielectric functions of Si100, SiO2100, and Pt101 layers.

Additionally, the layer thicknesses determined by XRR (see Table 2) and the surface

roughness determined by AFM were used in the optical model and were kept unchanged

throughout the analysis. The unknown dielectric function of CoFeB was addressed as a

parameterised dielectric layer composed of a Drude oscillator function to account for free-

charge-carrier driven transitions and two Lorentzian oscillators to describe the dispersion

arising from interband transitions. This model was further adjusted in terms of the Drude

and Lorentzian parameters to respond to the structural changes resulting from the

annealing.

Table 2. Nominal and XRR determined thicknesses of the CoXFe(80-X)B20, Co50Fe50, and Pt layer for as the as-deposited samples

Sample ID Nominal thickness XRR thickness

CoFeB Pt CoFeB Pt

Co50Fe50 70 nm 5 nm 61.5±5 nm 3.7±0.3 nm

Co40Fe40B20 100 nm 5 nm 97.4±5 nm 4.9±0.3 nm

Co60Fe20B20 100 nm 5 nm 104.6±5 nm 4.8±0.3 nm


49

4.2.1 Structural properties

Figure 27 presents the XRD θ-2θ scans of the CoFeB samples annealed in vacuum at

different temperatures. The pronounced CoFe(110) reflex observed at 500°C and above

indicates the crystallisation of the films. In accordance with previous studies, the creation

of a crystalline alloy from the initial CoFeB compound occurs while boron diffuses out of

the lattice, resulting in CoFe crystalline grains surrounded by amorphous boron102,103. A

closer look at the diffractograms indicates that Co40Fe40B20 crystallises in a polycrystalline

and polytextured fashion, as the present (110) and (211) peaks correspond to different

crystallographic orientations of the body-centred cubic (bcc) CoFe. Co60Fe20B20, on the

other hand, reveals a strong (110) texture, with more intense (110) and (220) peaks

occurring. The strong (110) texture was also confirmed by additional rocking scan analysis

of the (110) out-of-plane crystallite orientation distribution (see Figure 30 and the related

discussion).

At 550°C and above, a shift and broadening of the Pt(111) peak are found for both

stoichiometries, most probably suggesting a degradation of the Pt layer, possibly due to

alloying or intermixing at the interface with CoFeB. Here, it is worth mentioning that three

stray reflexes at 61°, 43° and 97°, as indicated in the figures are from silicon (400) due to

the Cu-Kβ radiation, Ag(200), and Ag(400) from silver epoxy, respectively.

Figure 27. X-ray diffraction patterns recorded for Si / SiO2(1.8 nm) / CoFeB(100 nm) / Pt(5 nm) before and after annealing under UHV at the indicated temperature for Co40Fe40B20 (a.) and Co60Fe20B20(b.). Additionally, the scan of the as-deposited Si / SiO2(1.8 nm) / CoFe(100 nm) /Pt(5 nm) sample is presented in black in figure 27(a.).The respective reflexes of constituent materials are marked by dotted lines along with the miller indices [powder diffraction file of CoFe (00-049-1567), Pt (00-004-0802), Si (00-027-1402), and Ag (00-004-0783) from the International Centre for Diffraction Data (ICDD)].


50

The observation of a distinct out-of-plane (110) texture formation for the Co60Fe20B20 film

is further confirmed by additional rocking scans (omega scans at fixed detector position

2θCoFe(110)), providing an estimate for the crystal misalignment, the so-called mosaicity.

Figure 28 shows the rocking scan profiles for the Co40Fe40B20 and Co60Fe20B20 film, which

exhibit a clear prefered out-of-plane (110) texture (or in general, crystalline phase)

formation for the latter composition, once annealed at 500°C or higher. The transition from

the amorphous to the (110) textured structure occurs very suddenly, as confirmed by the

dramatic shape change of the rocking scan profile from 450°C to 500°C. After an initial

FWHM of about 12° after annealing at 500°C, the (110) texture of Co60Fe20B20 improves

further to an FWHM of below 9° after annealing at 600°C. In contrast, the Co40Fe40B20

rocking scan reveals no peak and only an angular independent increase in the rocking

scan intensity at 600°C can be observed, confirming random, polycrystalline CoFe

crystallisation. Therefore the omega scans of the samples annealed at the temperatures

below 550°C were not measured as no peak is expected for these samples.

Figure 28. Rocking curve measured at the CoFe(110) reflex for the Co40Fe40B20 (a.) and Co60Fe20B20 (b.) films. The inset in (b.) shows the FWHM of particular Gaussian fits of the obtained peaks.

The vertical coherence lengths, corresponding to the crystallite size in the normal direction

to the sample surface, were calculated from the FWHM of the CoFe(110) peak using the

Scherrer formula (see section 3.3.1). For the investigated films, a maximum crystallite

size of around (25 ± 2) nm is obtained for the annealing temperature of 600°C, as shown

in Figure 29 a, which is consistent with previously reported studies on 100 nm thick CoFeB

films88. As detected by cross-section scanning electron microscopy (SEM) studies

performed in a trench made by focused ion beam (FIB) (see Figure 29 b), the CoFeB alloy

does not fully crystallise within the 30 min applied annealing steps. The crystallisation


51

starts from the top interface with Pt and expands for (25 – 30) nm, in agreement with the

vertical coherence length of the crystallites determined from XRD. No remarkable

differences in the SEM micrographs were observed for Co40Fe40B20 and Co60Fe20B20;

thus, exemplary images for Co40Fe40B20 only are presented.

Figure 29. The CoFe crystallite sizes (vertical coherence lengths) calculated using the Scherrer equation for Co40Fe40B20 and Co60Fe20B20 (a) determined from XRD shown in Figure 27. SEM micrograph collage of Si / SiO2(1.8 nm) / Co40Fe40B20(100 nm) / Pt(5 nm) before (above) and after annealing (below) recorded in the FIB trench at 36° stage tilt (b.).

To investigate near-surface changes in the CoFeB layers further and to avoid the intense

peak from the silicon substrate, GIXRD was performed at a fixed angle of incidence of

Ω = 1°, shown in Figure 30. All the aforementioned CoFe peaks plus the CoFe(200),

which was hidden by the Si substrate peak for symmetric θ-2θ scans, are present after

annealing at 500°C, 550°C and 600°C for Co40Fe40B20, as well as for Co50Fe50 in the as-

deposited state (see Figure 30 a), reasserting the polycrystalline nature of the CoFe alloy

in this composition. On the contrary, for Co60Fe20B20 shown in Figure 30 b, none of the

CoFe peaks are detected, as the Bragg condition is not fulfilled for any detector angle due

to the pronounced CoFe(110) texture and the fixed incident angle of Ω = 1° (cf. Figure 27

and Figure 28, as well as the corresponding discussion). The Pt passivation layer

deposited on the top of Co40Fe40B20 layer exhibits polycrystallinity, although annealing at

temperatures larger than 500°C induces obvious peak shifts in both, symmetric and

grazing incidence XRD scans. On the other hand, the Pt layer on top of Co60Fe20B20 reveal

pronounced (111) texture as seen from the symmetric θ-2θ scans shown in Figure 27 with

comparably large Pt(111) and even obvious Pt(222) peaks, whereas in grazing incidence

only (111) and (220) peaks occur, but no Pt(200). The peaks from diagonal lattice planes

occur due to the pronounced (111) texture of Pt, further considering a random in-plane


52

crystal arrangement (fibre texture), the grazing incident angle and particular mosaicity of

the crystallites. Comprehensively summarising all structural measurements introduced

before, it seems that the Pt texture serves as a template for the CoFe crystallisation taking

place at high-temperature annealing, as it was clearly shown that the CoFe crystallisation

starts from the top interface with Pt and the observed differences in the CoFe texture were

also found in Pt, whose crystal structure further shows distinct distortion for annealing

above 500°C with converging Pt and CoFe peak positions.

Figure 30. The grazing incidence diffractogram recorded for CoFeB thick films before and after annealing at various temperature for two stoichiometries: Co40Fe40B20 (a.) and Co60Fe20B20 (b.). The respective reflexes of the constituent materials are marked by dotted lines along with the Miller indices.

4.2.2 Optical properties

Spectroscopy ellipsometry

Figure 31 shows the evolution of Ψ and Δ spectra for the sample with Co40Fe40B20 and

Co60Fe20B20 annealed at different temperatures. For clarity, only a selection of the data is

shown. It can be seen in the Ψ spectra for both sets of samples that no significant change

occurs up to 400°C, suggesting that no structural changes have occurred so far. However,

upon annealing at 450°C, sharpening of spectral feature in Ψ spectra between 1.0 eV to

2.5 eV can be observed for both sets of samples, which becomes more prominent with a

further rise in annealing temperature. Similarly, a gradual increase in the amplitude of Δ

spectra can be observed for all the samples. In comparison to the XRD results, where the

first noticeable evidence of crystallisation is visible only at 500°C for the two sets of

samples, this suggests higher sensitivity of the SE with respect to structural changes.

Furthermore, the comparison of Ψ and Δ spectra measured for as-deposited Co50Fe50

and annealed Co40Fe40B20 and Co60Fe20B20 shows a great resemblance of the spectral


53

feature developing between 1.0 eV to 2.5 eV upon annealing, suggesting the

crystallisation of CoXFe(80-X)B20 into a CoFe phase.

Figure 31. The evolution of and (inset) spectra recorded for the Co40Fe40B20 (a.) and Co60Fe20B20 (b.)

thick films before and after annealing at various temperatures. For comparison, the and spectrum of Co50Fe50 is plotted along with the Co40Fe40B20 spectra in (a.).

Using the optical model discussed in section 4.2, the complex dielectric functions (ε1xx and

ε2xx) were determined for the two investigated CoFeB stoichiometries and Co50Fe50. For

the ease of discussion, the spectra can be divided into two main regions: (i) the near-

infrared (NIR) region below 1.0 eV (NIR), accounting for intraband transitions, and (ii) the

visible and ultraviolet (UV) region above 1.0 eV, related mainly to interband contributions.

The NIR region of the spectrum is described by a Drude type contribution, related to the

free-electron absorption in Co50Fe50, and will be discussed in more detail later in this

section.

In the case of Co50Fe50, the Drude contribution is followed by a broad structure centred at

around ~1.5 eV in the ε1xx spectra (corresponding feature at ~2.0 eV in ε2xx spectra), see

Figure 32. This feature has been previously ascribed to the hybridisation of p and d

orbitals, resulting in direct transitions of occupied d- and unoccupied p-states in CoFe

alloys with a bcc crystalline phase96,97.

In order to understand the influence of B inclusions on the optical properties of the

Co50Fe50, the complex dielectric function of (Co50Fe50) + B was simulated. For this

purpose, the Bruggemann effective medium approximation approach was used to

calculate the optical constants of the mixed material with the host matrix of Co50Fe50

containing B inclusions. In this approach, 15% of the (Co50Fe50) + B film volume is

assumed to be a spherical inclusion of B into the metallic Co50Fe50. However, it should be


54

noted that this is only a coarse approximation of the actual situation; since previous

studies suggested that B migrates to CoFe grain boundaries or to the neighbouring

layers102,103. The most obvious change induced to the dielectric function spectra of

Co50Fe50 by the B inclusion is visible in the ε2xx spectrum, namely a decrease of the

absolute values. Since B is a non-metallic material, its addition to the metallic Co50Fe50

increases the dielectric losses. Consistently, the values of ε1xx increase, indicating an

increase in the relative permittivity of CoFe-B. A slight broadening of the spectral features

is also observed, but rather negligible when compared with the changes in the features of

the dielectric functions of the CoFe-B alloys extracted from the experimental ellipsometry

spectra before and after annealing, see Figure 33. The good correspondence between

the simulated complex dielectric function of B incorporated in CoFe and the dielectric

function determined for the Co40Fe40B20 indicates that during the crystallisation process

CoFe crystallites are formed and B migrates outside the crystallites, i.e. to the grain

boundaries. This scenario is in line with the results of previous studies of the local

structure of CoFeB102 and of crystalline CoCrPt-B alloys used for recording media in hard

disk drives104.

Figure 32. The complex dielectric function (ε1xx & ε2xx) spectra of the Co50Fe50 (red), Co40Fe40B20 (blue) and B (grey) 105, together with the simulated ε1xx & ε2xx of (Co50Fe50)+B with 15 % B inclusion(yellow). More detailed information about the B inclusion is given in the text.

The dielectric functions of the as-deposited CoFeB alloys present only weak and very

broad spectral features, which gradually become more pronounced with an increase in

annealing temperature, as shown in Figure 33. The characteristic spectral feature of

Co50Fe50 at ~1.5 eV occurs in the ε1xx spectra for the samples annealed at 450°C. This

suggests that 450°C is the onset temperature for crystallisation. As the optical

spectroscopy has an information depth limited to a few 10 nm, the changes visible in the


55

spectra at 450°C indicate that the crystallisation takes place near the surface

(CoXFe(80- X)B20 / Pt interface), as supported by scanning electron microscopy (SEM)

images (cf. Figure 29 b) demonstrating nucleation at the Pt interface. Noticeably, the

pronounced CoFe reflex was observed in XRD scans starting at 500°C, indicating that the

optical spectroscopy allows probing the incipient phase of crystallisation with very small

crystallites. In fact, a remarkable resemblance of the dielectric function of the Co40Fe40B20

after annealing at 600°C and the as-deposited Co50Fe50 is found, which is consistent with

the similarities in the crystalline structure observed with XRD. This suggests that at 600°C

B diffuses completely out of the CoFe crystallites. The systematic decrease in ε2xx with

annealing temperature is furthermore consistent with a greater ordering within the films

due to crystallisation. The characteristic spectral feature in ε1xx spectra of Co60Fe20B20 is

red-shifted relative to Co40Fe40B20, probably due to the difference in the stoichiometric

composition. Additionally, comparing the amplitudes of the ε2xx spectra (mostly <1.0 eV)

of the two stoichiometries, it is evident that the Co40Fe40B20 has lower dielectric losses

due to the free electrons in comparison to Co60Fe20B20. This, in turn, implies that

increasing Co concentration increases the charge carrier concentration, which is

consistent with the empirical finding that the resistivity of Co is almost half of that of

Fe105,106.

Figure 33. The annealing temperature dependent evolution of ε1xx (a.) and ε2xx (b.) spectra for Co40Fe40B20 (solid line) and Co60Fe20B20 (dashed line), and CoFe (black).

The classical Drude equation defines the free charge carrier concentration contribution to

the dielectric function, which in its mathematical form is equivalent to a Lorentzian

oscillator positioned at 0 eV:


56

휀(𝐸) = −ℏ

휀0𝜌(𝜏𝑠𝐸2 + 𝑖ℏ𝐸) 4-1

where, ε0 is the vacuum dielectric constant, ℏ is the reduced Planck constant, τs is the

mean scattering time of the free carriers between successive collisions and ρ is the optical

resistivity. The analysis of the Drude contribution to the dielectric function allows deriving

the optical resistivity (ρ) and scattering time (τs) of the investigated films. The resistivity

and the scattering time ultimately relate to the ordering state of the films, according to the

Fuchs size-effect theory107.

These parameters are shown in Figure 34 for both CoFe-B stoichiometries. The resistivity

remains barely unchanged until 400°C, followed by a maximum at 450°C and a

subsequent decrease with increasing annealing temperature. It should be noted that this

evolution cannot be explained by the B diffusion since reference108 showed that the

migration of B starts already at 200°C (see section 6.3). The decrease in resistivity can

be ascribed to an increase in ordering and decrease in the number of defects, which, in

fact, is consistent with the increase in the crystallite size derived from the XRD

measurements. The presence of a maximum at 450°C relates very likely to a temperature

of nucleation of the crystallites, where the resistivity increases due to the formation of

grain boundaries and defects, originating from the low level of ordering of the crystal. The

poly-textured phase in Co40Fe40B20 (in contrast to the well-oriented phase in Co60Fe20B20)

could arguably also explain the difference in the resistivity and scattering times between

the two alloys, since more mismatched grain boundaries and defects would lead to higher

resistivity due to the shorter mean free path, i.e. longer carrier scattering time.

Sheet resistance (R) measurements were conducted on all the samples in order to

investigate the influence of annealing on the electrical properties of the layers. The

change in sheet resistance of the CoFeB samples with annealing temperature is shown

in Figure 34 b. Up to 400°C, no significant change in the sheet resistance is found. Above

this temperature, a monotonous decrease with increasing temperature is observed,

consistent with the changes observed for the optical resistivity parameter calculated from

the Drude model (cf. Figure 34 a). The maximum at around 450°C is not very obvious as

the entire samples (including the cap layer and 100 nm thick mostly amorphous CoFeB)

are probed electrically, although crystalline CoFe occurs in the upper part only and optical


57

spectroscopy, as a surface-sensitive measurement technique, probes that crystalline

volume exclusively. Given the increase in crystallite size and ordering within the films with

the annealing temperature revealed by XRD, a decrease in the charge carrier (electrons)

scattering due to defects and grain boundaries is expected, which results in a decrease

of the sheet resistance. The trend in R for both stoichiometries is noticeably similar to the

change in ρ obtained from the SE measurements.

Figure 34. Drude parameters resistivity (ρ) in black and scattering time (τs) in red colour as a function of annealing temperature for Co40Fe40B20 (solid symbol) and Co60Fe20B20 (empty symbol). The lines in the figure are drawn to guide the eye (a.). Sheet resistance of the Co40Fe40B20 (filled circles in red) and Co60Fe20B20 (unfilled circles in blue) layers passivated with a Pt thin film as a function of the annealing temperature (b.).

Magneto-optical spectroscopy

Figure 35 a shows the measured θK and ηK MOKE spectra of Co50Fe50, in comparison to

θK reported by D. Weller et al. for Co48Fe52109. Even though the amplitude of θK is slightly

lower than previously reported109, the line shape of both experiments resembles closely.

In fact, the present data is closer to the first-principle calculations performed by Maurer et

al. for this CoFe composition110. The off-diagonal dielectric function of Co50Fe50 was

calculated using the procedure discussed in section 3.3.2 and is shown in Figure 35 b as

εxy (h)2, in order to highlight the spectral features111. It is well established by theoretical

studies that the spin polarised density of states of 3d-transition metals and their alloys are

fairly similar, resulting in similar electronic transitions in magneto-optical spectra96. These

spectral features noticed in the optical region of the spectrum can be explained based on

the theoretical predictions by Kwang Joo Kim et al. for Fe3Co and Co3Fe using the tight-

binding linear-muffin-tin orbitals (TBLMTO) method with the local spin density

approximation (LSDA)97,110. They assign the transition at 2.0 eV as originating mainly from

transitions from the occupied minority-spin d triplet states at lower energy into the


58

unoccupied minority-spin p states. These d→p transitions in the minority-spin bands

become possible through p-d hybridisation.

Figure 35. Measured polar Kerr effect rotation (θK) and ellipticity (ηK) spectra of Co50Fe50 in comparison

to the previously published reference109,110 (a.) and calculated xy (h)2 as the function of photon energy (b.).

Figure 36 shows the evolution in θK and ƞK spectra for the CoFeB samples annealed at

different temperatures for the two investigated stoichiometries. Similar to the SE spectra,

no significant changes in θK and ηK spectra were observed up to 400°C. Upon annealing

at 450°C, the characteristic line shape of the θK spectrum starts resembling that of CoFe.

Annealing at higher temperatures results in the enhancement of spectral features at

~2.0 eV and ~4.7 eV. The improvement in the features upon annealing at 450°C or above

is consistent with considerable ordering in the lattice, and again, this development in the

spectra can directly be correlated with the increase in crystalline ordering of CoFe. It can

also be observed that these features are slightly red-shifted for Co60Fe20B20 compared to

Co40Fe40B20, contrary to previous theoretical calculations, where no significant differences

were found on the MOKE spectra due to different CoFe content96. In fact, this shift may

as well be related to the differences found with XRD in the crystalline structure of both

compounds. In this context, we note that besides composition, also the crystalline

environment influences significantly the magneto-optical properties of the material112. In

fact, the larger amplitude of the spectral features of Co40Fe40B20 is furthermore an

indication of higher spin-polarisation (magnetisation) for the lower Co content, which is

consistent with the calculated Slater–Pauling curve113.


59

Figure 36. The polar Kerr effect measured rotation (θK) and ellipticity (ηK) spectra for the 100 nm thick CoFeB films before and after annealing at the indicated temperatures for two stoichiometry Co40Fe40B20 (a. & c.) and Co60Fe20B20 (b. & d.).

The calculated εxy (h)2 as a function of the photon energy for the two investigated

compositions annealed at 450°C and 600°C is shown in Figure 37. Similar to the off-

diagonal dielectric function of CoFe shown in Figure 35 b, the imaginary part of ε2xy (h)2

shows two main features in the measured spectral range, at ~2.0 eV and ~4.5 eV. As

discussed earlier, the features in the εxy spectra are ascribed to transitions from the filled

part of the 3d band to empty hybridised pz states near Fermi energy (spin-down states).

The relative shifts in the positions of the spectral features of Co60Fe20B20 to lower energies

with respect to Co40Fe40B20 suggest a dependence of the DOS on the stoichiometry. Such

a shift can be explained based on the theoretical calculations of Liu et al. related to the

electronic structure of CoFe alloys with different stoichiometry 114. Although the DOS near

the Fermi energy is similar for both Co and Fe, due to the higher electronegativity and a

smaller exchange splitting of Co, the minority-spin orbitals of Co are situated at lower

energies than the Fe orbitals. A higher percentage of Co in the alloy will, accordingly, lead


60

to a shift of the electronic states of the alloy states to lower energies. The change in

amplitude of the spectral features relates, as mentioned previously, to the degree of spin-

polarisation, which is increasing in the CoFe alloys case with increasing Fe content 113.

Figure 37. The calculated xy(h)2 as a function of the photon energy for two stoichiometries, Co40Fe40B20 and Co60Fe20B20, for Si / SiO2(1.8 nm) / CoXFe(80-X)B20(100 nm) / Pt(5 nm) layer stacks annealed at 450°C and 600°C.

4.3 Thin films

The results from the thick films suggest that optical spectroscopy techniques are very

sensitive to structural changes. In this sub-chapter, we will focus on the structural and

optical properties of thin films on the example of Co60Fe20B20.

The Co60Fe20B20 thin films with nominal thicknesses ranging from 10 nm to 20 nm covered

by a 3 nm gold passivation layer were deposited by DC magnetron sputtering on thermally

oxidised silicon substrates. Similar to the thick film samples, the thickness of the

Co60Fe20B20 / Au bilayers was verified using XRR measurements on the as-deposited

samples (see Table 3). The depositions were performed at RT with a base pressure below

2 x 10-4 Pa and a working pressure of 0.35 Pa, while using Ar as process gas. The

samples were annealed for 30 min in vacuum at 350°C and subsequently for 30 min at

400°C. The first annealing step was performed at 350°C as the transformation of the thin

films from amorphous into a crystalline bcc CoFe phase is expected to occur above

325°C115, in contrast to the thick films studied before, where temperature as high as 450°C

was required for initial crystallisation. The characterisation of the samples with the same

methods as used in section 4.2, was performed in the as-deposited state and repeated

after each of the subsequent annealing steps.


61

4.3.1 Structural properties

The X-ray diffractograms recorded using the 3000PTS diffractometer before and after

annealing at 350°C and 400°C are shown in Figure 38. For the as-deposited state of the

sample, only the Au(111) and Si (substrate) peaks at 38.2° and 69.1°, respectively, were

observed, indicating that the Co60Fe20B20 film is amorphous. After annealing at 350°C, no

peak corresponding to the CoFe crystalline phase was detected, suggesting that the

Co60Fe20B20 layer is still in an amorphous phase or the crystallites are not large enough

to be detected by the XRD equipment. A further annealing step at 400°C resulted in the

occurrence of a well distinguishable 2θ peak around 45°, characteristic for the CoFe(110)

crystalline phase. For this peak, a crystallite size of (8.9 ± 2) nm was calculated using the

Scherrer formula.

Figure 38. X-ray diffractogram recorded for Si / SiO2(100 nm) / Co60Fe20B20(20 nm) / Au(3 nm), in the as-deposited state (squares) and after annealing at 350°C (circles) and 400°C (triangles).

Table 3. Nominal and XRR determined thicknesses of the Co60Fe20B20 and Au layer for as the as-deposited samples

Sample ID Nominal thickness XRR thickness

Co60Fe20B20 Au CoFeB Au

1 10 nm 3 nm 10.5±0.2 nm 2.6±0.1 nm

2 15 nm 3 nm 14.8±0.2 nm 3.2±0.1 nm

3 20 nm 3 nm 18.9±0.2 nm 3.6±0.1 nm


62

4.3.2 Optical properties

Spectroscopy ellipsometry

For the SE, an optical model similar to that of thick films was used to derive the optical

constants for each sample. A multi-sample analysis based on a parametric model

consisting of one Drude free electron model and a series of five Gauss-Lorentz oscillators

was employed in order to represent the fine structure of the optical dispersion in the

Co60Fe20B20 in the entire investigated spectral range. In the multi-sample analysis, the

dielectric function of the samples was considered to be identical for the considered range

of thicknesses. This model was then further adjusted to respond to the structural changes

induced by the annealing steps.

Figure 39 shows the evolution of the Ψ and Δ spectra recorded at an angle of incidence

of 65° for Co60Fe20B20(20 nm) / Au(3 nm) before and after annealing at 350°C and 400°C.

Similar to the thick films, noticeable changes can be observed with the increase in

annealing temperature. Significant changes occur already after the 350°C annealing step

in both Ψ and Δ spectra, whereas no reflex corresponding to CoFe(B) was observed in

XRD (cf. Figure 38). Here it should be noticed that the spectral features observed in the

Ψ and Δ spectra for the thin film samples cannot be compared with the ones measured

for the thick film due to interference between reflected light from the surface and interfaces

in the thin film layer stack (CoFeB-Au, CoFeB-SiO2, and Si-SiO2 ).

Figure 39. Experimental SE ψ and ∆ spectra recorded at 65° AOI for Si / SiO2(100 nm) / Co60Fe20B20(20 nm) / Au(3 nm) before and after vacuum oven annealing at 350°C and 400°C.

Similar changes can also be observed in the real (ε1xx) and imaginary (ε2xx) parts of the

dielectric function (diagonal elements of the dielectric tensor) of Co60Fe20B20 obtained


63

from the SE measurements, before and after annealing, as shown in Figure 40 a. For the

as-deposited films, the dielectric function relates well to the typical line shape of the optical

dielectric function reported for amorphous Co20Fe60B20 films with thicknesses between

10 nm and 40 nm grown on Si / SiO2(1000 nm), i.e. with very broad features98. It should

be noted that thicker films of Co60Fe20B20 (100 nm) discussed previously in section 4.2,

exhibit even lower spectral features (see Figure 40 b in comparison to 20 nm thick films).

While for the modelling of thicker films one Drude oscillator and two Gauss oscillators

located at 0.5 eV and 2.2 eV were employed, in the case of the multi-sample analysis

performed for the films thicknesses in the range from 10 nm to 20 nm, additional

oscillators located at 1.4 eV, 3.3 eV, 3.7 eV and 4.6 eV had to be employed in order

achieve lower MSE values of the simulation. This apparent discrepancy between the thick

and thin film optical properties might have several reasons: possible oxidation of the top

layer of Co60Fe20B20 would have a larger contribution to the optical response of the thin

compared to the thicker films. In addition, a granular structure of the film, which is more

pronounced in the case of the thinner films or the related plasmonic effects in the thin Au

overlayer116 or in the Co60Fe20B20 might affect the thin film dielectric function.

Figure 40.The ε1xx (black) and ε2xx (red) components of the complex of the dielectric function of Co60Fe20B20 (10 nm to 20 nm) before (continuous line) and after annealing at 350°C (dashed line) and 400°C (dotted line) (a.). Comparison of the complex dielectric function of as-deposited 20 nm and 100 nm thick Co60Fe20B20 films (b.).

Upon annealing the thin films, two effects become visible: (i) a change in the low energy

slope of the ε2xx and (ii) a sharpening of the spectral features. The low energy slope of ε2xx

initially increases from the as-deposited state to the sample annealed at 350°C and then

it decreases again in the samples annealed at 400°C. The decrease in the slope of ε2xx

with annealing temperature corresponds to the characteristic behaviour of metals upon


64

crystallisation and the corresponding reduction in the dielectric losses117. Liang et al.

reported merely no changes of the Co20Fe60B20 optical constants of a 40 nm film upon

annealing at 300°C for 2 h, whereby 300°C is below the crystallisation temperature of

CoFeB98.

The most significant changes upon annealing of the thin films occur at 2.6 eV and 3.8 eV

in ε1xx. The lineshape of the spectra obtained after annealing resembles the reported one

for Co20Fe60B20 films (2 nm)98. Liang et al. explained such a lineshape by the optical

anomaly occurring when isolated metal islands percolate into a continuous metallic layer.

Extrapolating to the case of the investigated system, such a structure-rich lineshape of

both ε1xx and ε2xx might be consistent with the increasing roughness of the film after

annealing. Slight oxidation of the CoFeB and/or the diffusion of the B towards the film

surface upon annealing (see section 6.3) are factors that might also influence the dielectric

function and can hardly be taken into account by modelling. Nevertheless, the occurrence

of a clear CoFe(110) peak in the X-ray diffractogram of the 20 nm film annealed at 400°C

(see Figure 38) indicates that the observed changes in the dielectric function (Figure 40)

are more likely to be related to the crystallisation of CoFe(B).

The electronic origin of the observed spectral features is discussed in the following. The

broad pronounced feature in ε2xx at ~2.6 eV was previously observed for FeCo alloys in

the bcc phase 97,115 and ascribed to direct interband transitions in the minority-spin bands

between occupied d- and unoccupied p-states118. It is worth mentioning that the optical

response may also be influenced by interband transitions from gold typically found around

2.5 eV116 or by plasmonic effects in Au or even in CoFeB. While the interband transitions

are taken into account by the optical constants for gold119 used in the discussed model,

the plasmonic effects were not considered. Since the complete removal of the Au

signature at 2.5 eV by the used optical model cannot be warranted, any comparison

should be taken with care. The energy range above 3.0 eV is dominated by the typical

interband transitions of the Fe or Co sites, including d-p as well as d-d transitions. In the

context of pure bcc-Fe, the peaks at 2.8 eV and 4.0 eV were ascribed to transitions along

with high symmetry points of the crystal lattice120. It can be assumed that this is also the

case in a CoFeB alloy, suggesting that the observed transitions are likely to be consistent

with the presence of a CoFe bcc crystalline phase (where the boron diffused out of the

crystallites), in agreement with the presence of the CoFe(110) peak in the diffractogram

of the 400°C annealed sample.


65

Magneto-optical spectroscopy

The changes in the Kerr spectra before and after annealing are compared in Figure 41. A

broad feature centred at about 2.5 eV is in this case also very pronounced, becoming

clearly narrower upon annealing. The spectra compare well with previous studies on CoFe

alloys96, suggesting that the B content in the alloy may have a minimal influence on the

magneto-optical response in the studied energy range. The amplitude of the signal

improves with the annealing, similarly to the previously reported case of FePt alloys121.

Following the interpretation proposed by Cebollada et al., the enhancement of the

magneto-optical Kerr effect of an alloy upon annealing can be regarded as a sign of

improvement in the magnetic ordering of the films. Due to the strong optical interference

oscillations in the MOKE spectra of the thin films, the off-diagonal dielectric function for

these samples could not be calculated with any reasonable assumptions in the optical

model.

Figure 41. Comparison of Kerr rotation θK (black) and ellipticity ηK (red) for Si / SiO2(100 nm) / Co60Fe20B20(20 nm) / Au(3 nm) before (continuous line) and after annealing at 350°C (dashed line) and 400°C (dotted line).

4.4 Conclusion

In this chapter, the evolution of the structural, optical, and magneto-optical properties of

optically thick (100 nm) films of CoXFe(80-X)B20 (X = 40 and 60) and optically thin (from

10 nm to 20 nm) Co60Fe20B20 films were discussed.

The 100 nm films passivated with a 5 nm Pt cap layer were annealed between 300°C and

600°C. The structural and electrical properties of the films were assessed by XRD and


66

electrical four-point probe measurements, respectively. The comparison of the

CoXFe(80- X)B20 dielectric function extracted from spectroscopic ellipsometry with that of

Co50Fe50 allowed to identify CoFe specific spectral features and to analyse the impact of

B on the optical properties of the CoFeB alloys. The (magneto-) optical spectroscopic

techniques are proven to be extremely sensitive to structural changes. The analysis of the

Drude component of the dielectric function of CoFeB allowed extracting information

regarding the resistivity and charge carrier scattering time, which is closely related to the

crystalline order in the films. Corroborating the results of spectroscopic ellipsometry, SEM,

and XRD, we can conclude that the nucleation of the crystallisation starts at the interface

between the CoXFe(80-X)B20 and the crystalline Pt capping layer.

The magneto-optical off-diagonal component of the dielectric function of CoXFe(80-X)B20

extracted from the MOKE spectra shows significant changes with the composition of the

alloy as well as with the structural evolution from the amorphous to the crystalline phase.

Similar to the thick film samples, the evolution in the optical and magneto-optical spectra

presented a strong correlation to the crystallisation of CoFeB. The dielectric function of

Co60Fe20B20 thin films showed more spectral features which are absent in 100 nm

samples. These features are the possible outcome of surface oxidation that may have a

greater influence on optical properties in thin films than thick films or because of the

plasmonic enhancement due to the granular gold overlayer at the surface.

This study underlines the utility of spectroscopic ellipsometry and MOKE spectroscopy for

material optimisation in the field of metallic alloys for spintronic applications. It also opens

the possibilities of such (magneto-) optical spectroscopic techniques for non-invasive and

in-situ characterisation method, which is even compatible with contemporary CMOS

microfabrication technology.

Setting exchange bias using laser direct-write technique Chapter 5

67

Chapter 5: Setting exchange bias using laser vs oven

annealing techniques

This chapter presents the proof-of-concept that focused laser irradiation can be used as

a method to set a defined pinned magnetisation direction through the exchange bias effect

in MTJs. The discussion is based on the comparison of the conventional vacuum

annealing to the laser annealing for varying radiation intensity and scanning speed. The

aim is to achieve the optimal laser annealing parameters so as to set a maximum

exchange bias field strength in the given stack. Further, the efficacy of the laser annealing

as a direct-writing technique was tested by writing a text on a simplified FM/AFM layer

stack and visualised using MOKE microscopy imaging to read the text. Some of the results

presented in this chapter were published in IEEE Transactions on Magnetics, DOI:

10.1109/TMAG.2018.2873428.

5.1 Introduction

A tremendous effort has been put into the integration of spintronic devices to the prevailing

CMOS technology. The miniaturisation of the devices has pushed the researchers to

study new materials and find solutions for technological challenges. One among such

challenges is the annealing of the spintronic devices, which is intended not only to improve

the performance of the device but also to assist in tailoring the properties of the device.

For instance, in the case of MTJ sensor devices, annealing and subsequent cooling in the

presence of a magnetic field allows inducing exchange bias in the device, which can be

exploited to tune the sensing range, the directional sensitivity etc. A method that allows

to locally anneal the micron and sub-micron devices can be highly beneficial. Recently,

Albisetti et al. demonstrated an innovative method of using thermally assisted magnetic

scanning probe microscopy (SPM) to pattern and set the exchange bias in an

IrMn / CoFeB bilayer system122. However, this method suffers some severe drawbacks,

including the fact that it is extremely time-consuming due to the use of SPM, which hinders

its industrial application. The laser-based annealing method is a possible method to

generate the heat locally in the exposed region, and with the help of suitable optics, a

wide device area can be covered. This method gives the advantage of setting the

exchange bias locally and in a short period of time, even for large scale wafer

processing51,123.


68

For tailoring the exchange bias in the FM and AFM layers and optimise the magnetic field

cooling parameters for setting a maximum exchange bias, a series of experiments were

conducted. The sample deposition was performed by magnetron sputtering by the

industrial partner “Singulus Technologies AG”. The layer sequence for the sample

investigated is “Si / SiO2(100 nm) / Ta(5) / CuN(30 nm) / Ta(3 nm) / Ni81Fe19(2 nm) /

IrMn(8 nm) / Co40Fe40B20(2 nm) / MgO(1.8 nm) / Co40Fe40B20(2.2 nm) / Ta(10 nm)”. The

samples were either heat-treated within the SQUID-VSM system (conventional annealing)

or locally with focused laser irradiation. The samples were characterised by SQUID or

MOKE magnetometry or, alternatively, by FORC. Details regarding the measurement

techniques are presented in chapter 3. A pictorial representation of the temperature profile

generated due to laser annealing and the selective alignment of magnetisation in

CoFeB / IrMn in a typical MTJ layer stack is shown in Figure 42.

An additional study was conducted with the aim to understand to which extent the seed

layer, in addition to the laser annealing parameters, influences the exchange bias

between CoFeB (pinned layer) and IrMn in the given stack. For this MTJs were prepared

by magnetron sputtering at Singulus technologies AG with the layer stack: Si /

SiO2(100 nm) / Ta(5 nm) / Seed layer / IrMn(8 nm) / Co40Fe40B20(2.3 nm) / MgO(1.8 nm)

/ Co40Fe40B20(2.3 nm) / Ta(5 nm), where the seed layer was chosen to be Ru(5 nm) or

Ni81Fe19(2 nm). The samples were exposed to laser irradiation by scanning square areas

of (0.5 × 0.5) mm2 in a continuous wave (CW) mode at scan speeds varying from

0.5 mm·s-1 to 5000 mm·s-1 and at laser peak intensities in the range from 60 kW·cm- 2 to

800 kW·cm-2. The exposure was later repeated for larger sample areas of (6 × 6) mm2

with scan speeds of 50 mm·s-1, 500 mm·s-1 and 5000 mm·s-1, at 600 kW·cm-2 to

1000 kW·cm-2 laser intensities. Furthermore, the influence of the number of scan

repetitions on the exchange bias field was analysed by repeating the laser annealing at

5000 mm·s-1 scanning speed and 480 kW·cm-2. For comparison, an analogous study with

conventional annealing technique was performed in the SQUID-VSM system for

temperatures ranging from 123°C to 400°C and cooled in the presence of a 120 mT in-

plane magnetic field, the same as used for all laser annealing processes. The coercive

and exchange bias fields for the pinned layer are calculated from hysteresis loops

measured using the NanoMOKETM2 setup.


69

Figure 42. A schematic diagram showing a typical layer stack of the MTJ sensor and the laser-induced temperature profile for setting the exchange bias in the FM/AFM layers. The CoFeB (FM) layer next to IrMn (AFM) is the pinned layer and the upper CoFeB layer is the free layer, both sandwich the MgO tunnel barrier. The Ni81Fe19 and CuN serve as the seed and buffer layer, respectively. The Ta layers work as the passivation (on the top) and adhesion layer (at the bottom).

5.2 Magnetisation reversal of complex MTJs layer stack

The M(H) hysteresis of the “Si / SiO2(100 nm) / Ta(5 nm) / CuN(30 nm) / Ta(3 nm) /

Ni81Fe19(2 nm) / IrMn(8 nm) / Co40Fe40B20(2 nm) / MgO(1.8 nm) / Co40Fe40B20(2.2 nm) /

Ta(10 nm) layer stack is shown in Figure 43. Assuming that the magnetisation of the free

layer (FL) and the exchange biased layer (PL) are aligned in-plane at zero applied

magnetic field, the hysteresis loop of the MTJ stack can be explained qualitatively as

follows. Prior to the annealing (in Figure 43 a), for large magnetic fields applied in the

positive direction (H > 0.2 T) both FL and PL are saturated and aligned in the direction of

the external magnetic field (point a). When the field is reduced close to zero

(0.2 T > H > 0.05 T), the PL magnetisation rotates gradually to some random orientation

with no net magnetisation due to arbitrary coupling to the IrMn layer, resulting in increased

coercivity (between the points a and b), while the FL remains aligned in the direction of

the applied magnetic field. When further decreasing the magnetic field to zero and then

changing the field direction, the FL rotates (across b to c), yet PL remains randomly

oriented. The increment of the magnetic field in the negative direction also rotates the PL

magnetisation in the direction of the field so that both layers align along the magnetic field

(through c to d). A similar explanation can be drawn for sweeping the magnetic field from

negative to positive saturation, where the resulting coercivity of the pinned layer is

enhanced due to the exchange bias effect, whereas the soft-magnetic free layer reveals

small coercivity as seen from the figure.


70

Figure 43. Magnetisation as a function of the in-plane applied external magnetic field on an as-grown sample (a.) and measured along the easy-axis of the bottom pinned layer after 120 mT field cooling from 227°C (b.). The arrow in green and red represent the magnetisation of the free layer and pinned layer, respectively.

Similar to the as-deposited sample, the M(H) hysteresis loop was recorded for the sample

annealed at 227°C and cooled to RT in the presence of a 120 mT in-plane applied

magnetic field, see Figure 43 b. This causes the exchange bias effect at the FM/AFM

interface with the FM magnetisation to be aligned according to the applied magnetic field.

After the annealing the M(H) hysteresis loop can be described as follows: for a large

positive magnetic field, both the CoFeB layers are saturated, and their magnetisation lies

in the direction of the magnetic field (point a). Lowering the field through zero leads to the

rotation of the FL, with a very sharp reversal (through point a and b) and making the FL

and PL anti-parallel to each other. Further increase field toward negative values rotates

the PL creating a parallel arrangement of PL and FL (at point c) again. This reversal is

only possible when the applied field overcomes the exchange bias field induced in the PL

by the annealing. Moreover, the exchange bias reduces the two symmetric hysteresis

present in I and III quadrants in one loop, that is shifted by HEB with the increased coercive

field (HC). This can be attributed to the gradual reversal of PL as a single magnetic layer

due to exchange biasing. Nevertheless, as the magnetisation reversal of the FL and the

PL can be clearly separated, the FL can rotate according to a low external field

independently with much softer magnetic properties. Thus being the basis of high

sensitive magnetoresistive measurement principles on the basis of the GMR and TMR

effect. This loop allows us to determine HEB and HC for the PL alone by measuring the


71

fields at which it switches: as is conventional, HEB is the average of these two fields, whilst

HC is half the difference between them.

5.2.1 Setting the exchange bias by conventional oven annealing

In the first step, the most favourable temperature for setting the exchange bias in the MTJ

stack described in Figure 42 was investigated by thermal treatment in the SQUID-VSM

system. For this, the samples were annealed at different temperatures for 5 min and then

cooled in the presence of 120 mT magnetic fields at a constant cooling rate of 15 K·min- 1.

Figure 44 shows the magnetisation loop measured at room temperature for a pristine

sample and after each annealing process. The saturation magnetisation (Ms) was

observed to increase with annealing, which is the signature of improvement in crystallinity

in the CoFeB layers. As anticipated, with increasing annealing temperature, a prominent

exchange bias effect between the CoFeB and IrMn layer appears, resulting in a stronger

exchange bias field strength and increased coercivity of the PL. A maximum exchange

bias field of HEB = (110 ± 5) mT and coercivity of HC = (24 ± 1.5) mT was noted for the

sample annealed at 227°C and was observed to be unaffected with an increase in

temperature to 327°C. Further, a very sharp reversal centred around H = 0 mT of the FL

can be seen, which verifies that the FL is not influenced by any demagnetising field from

the PL.

Figure 44. SQUID-VSM measured M(H) hysteresis loops recorded at RT for an as-deposited sample and after consecutive field cooling process from different temperatures as indicated in the legend.

This study was further extended to determine the blocking temperature (i.e. the

temperature at which the biasing vanishes, TB), which in turn defines the working

temperature range of the final microfabricated magnetic field sensors. Initially, the sample


72

was annealed to set the exchange bias. M(H) loops were then recorded at RT and further

iteratively after increasing the temperature in steps up to 327°C. With elevated

temperature and in the absence of magnetic field, the thermal energy overcomes the

exchanges anisotropy energy, resulting in a gradual reduction in exchange bias and

coercivity finally going to zero at TB. From Figure 45, it is evident that the HEB linearly

decreases, vanishing completely at TB = (267 ± 20)°C. A similar trend was observed for

HC. The measured TB is significantly lower than the TN of bulk IrMn due to the fact that for

the thin films, TB is the resultant of finite-size scaling124,125.

Figure 45. Temperature dependence of the coercivity HC (a.)and the exchange bias field HEB for the determination of the TB. The sample was initially cooled in 120 mT field in-plane applied magnetic from 327°C to RT before the measurements at increasing temperature.

5.2.2 Setting the exchange bias by laser annealing

Continuous-wave laser annealing

The results from the conventional oven annealing presented in the previous sub-section

have established a baseline for understanding the exchange bias in the MTJ layer stack.

Further experiments with localised annealing on a microscale using laser irradiation were

targeted to achieve similar results as with conventional oven annealing. Samples having

the same layers (Si / SiO2(100 nm) / Ta(5 nm) / CuN(30 nm) / Ta(3 nm) / Ni81Fe19(2 nm)

/ IrMn(8 nm) / Co40Fe40B20(2 nm) / MgO(1.8 nm) / Co40Fe40B20(2.2 nm) / Ta(10 nm)) were

irradiated with a laser beam using the experimental setup described in section 3.2.3. For

these experiments, square areas-of-interest of (0.5 × 0.5) mm2 were heated by raster

scanning with continuous laser irradiation (CW) in the presence of an external magnetic

field of 120 mT, aligned in-plane. The optical images of the laser annealed samples with

test series about the laser scanning speed and different laser intensities are shown in


73

Figure 46. The magnetic characterisation for the irradiated areas was conducted with

MOKE magnetometry measurements in longitudinal geometry.

a.

b.

c.

d.

Figure 46. Optical micrograph of an exemplary sample annealed with CW irradiation with different scanning speed, 50 mm·s-1 (a.), 500 mm·s-1 (b.), 2000 mm·s-1 (c.), and 4000 mm·s-1 (d.). In each square, the different laser power was used as indicated, the patterns, as well as the labels, were created by laser ablation.

Figure 47 exhibits the coercive field and exchange bias field of the PL determined from

the MOKE hysteresis loops as a function of the CW laser intensity for four different scan

speeds (from 50 mm·s-1 to 4000 mm·s-1). In terms of the HEB evolution with the laser

intensity, a general trend can be deduced for all speeds investigated and can be described

in three different regimes. In the first regime (e.g. from 67 kW·cm-2 to 190 kW·cm-2 for the

50 mm·s-1 scanning speed), the total energy donated to the layer system is not sufficient

to reach the necessary temperatures, and thus, not all ferromagnetic domains can be

pinned in the direction of the cooling field. The second region resembles a plateau and

refers to a regime where the donated energy is sufficient to set the HEB completely. This

indicates the range of laser parameters where the Néel temperature of the

antiferromagnet is reached at the IrMn / CoFeB interface, and therefore the area-of-

interest for applications. The third region is characterised by a decay of the exchange bias

field strength, which indicates that the additional energy provided to the system is possibly

introducing degradation of the layers due to damage at the interfaces through distinct

diffusion (discussed in chapter 6) and alloying, or ultimately melt (note the colour change

in Figure 46). This is detrimental to the exchange coupling of the CoFeB / IrMn system

and even more the TMR effect of the whole MTJ. It is not possible to estimate the

temperature induced in the MTJ stack by the different laser intensities at different


74

scanning speeds currently, but the cooperation partners from LHM have started to

simulate the heat profile in the samples.

Figure 47. The coercive (a.) and exchange bias field (b.) of the PL as a function of the applied laser intensity with scanning speed of 50 mm·s-1, 500 mm·s-1, 2000 mm·s-1, and 4000 mm·s-1 for CW laser annealing. For comparison, the ranges of the maximum values of HC and HEB in the case of oven annealing are shown as hashed areas, including the error bars.

From these investigations, the optimal laser intensity parameter window for setting the

exchange bias field is concluded in Table 4. For the slower scanning velocity (50 mm·s-1)

the optimal window for laser intensities is very limited since a slower scanning velocity

implies that the area-of-interest was exposed to radiation for a longer time interval. This

results in a larger heat transfer to the sample and hence higher temperatures.

The maximum HEB values achieved with CW laser annealing are very close to those

achieved with conventional annealing but yet slightly lower. Since the laser annealing is

performed by scanning a focused CW-laser at 5 μm line spacing, the additional heat of

previous scans and the horizontal heat propagation with a distinct broadening of the heat-

affected zone need to be taken into account. Assuming that the heat dissipation through

the substrate is constant, the local thin film temperature profile may vary drastically with

multiple transitions of the Néel temperature, especially for slower scanning.

Table 4. The optimal laser intensity parameter for the various laser scanning speed.

Scanning speed Optimal laser intensity HEB HC

50 mm·s-1 (200 - 300) kW·cm-2 (97 ± 3) mT (28 ± 2) mT

500 mm·s-1 (250 - 450) kW·cm-2 (103 ± 3) mT (26 ± 3) mT

2000 mm·s-1 (300 - 500) kW·cm-2 (98 ± 4) mT (27 ± 2) mT

4000 mm·s-1 (300 - 500) kW·cm-2 (98 ± 4) mT (26 ± 3) mT


75

Pulsed laser annealing

Similar to the CW laser annealing, in this experiment, the sample with the same layer

stack was irradiated with PW laser with varying intensity. In this case, a step distance

between consecutive pulses of 2.5 µm was used, using pulses of 100 ns duration. The

coercive and the exchange bias field strength obtained in the pulsed laser annealed

samples are shown in Figure 48. This plot can also be categorised into three regions,

similar to the results presented in Figure 47. Laser intensities in the range from

500 kW·cm-2 to 900 kW·cm-2 were observed to induce a maximum average exchange

bias field of HEB ≈ (93 ± 5) mT. In this intensity range, the exchange bias field values do

not significantly depend on the laser intensity. In comparison with the CW laser annealed

sample, the exchange bias field values are significantly lower, which is consistent with the

previous studies reposted by Berthold et al.123. This suggests that the PW laser annealing

method has a much higher cooling rate, which in turn results in a lower heat transfer to

the MTJ stack as compared to the CW laser annealing and thus in lower exchange bias

field values.

Figure 48. Coercive field (black symbols, left axis) and exchange bias field strength (red symbols, right axis) as a function of the laser intensity for the samples annealed using PW laser. For comparison, the ranges of the maximum values of HC and HEB in the case of oven annealing (cf. Figure 47) are shown as hashed areas with included error bars.

The coercivity of the pinned layer increases with an increase in intensities up to a value

of HC = (29 ± 4) mT (Figure 48, black symbols.). The HC does not change significantly

(within the error bars) up to a laser intensity of 900 kW·cm-2. In the intensity range from

900 kW·cm-2 to 1750 kW·cm-2, the HC values decrease. Interestingly in comparison to the

CW laser annealing, above 1500 kW·cm-2, the HC increases significantly again. This

increase of the coercivity could be due to alloying effects caused by interdiffusion in the


76

magnetic thin films because of the intense heat generated by the short laser pulses. This

is well supported by the observed decrease in exchange bias field strength.

Further investigation of the structural integrity of these samples with XPS depth profiling

showed signs of significant diffusion of Cu from the CuN layer into the uppermost layers

of the stack, even at the laser intensities for which maximum exchange bias field was

observed (Dr. V. Dzhagan, private communication). For the investigations presented in

the following sections, the CuN layer is skipped from the layer stack design.

5.2.3 Influence of the seed layer on the laser annealing

Figure 49 exhibits the exchange bias field strength obtained at different scanning speeds

as a function of the CW laser intensity for the sample with the Ru and Ni81Fe19 seed layer

in the Si / SiO2(100 nm) / Ta(5 nm) / Seed layer / IrMn(8 nm) / Co40Fe40B20(2.3 nm) /

MgO(1.8 nm) / Co40Fe40B20(2.3 nm) / Ta(5 nm) layer stack. When comparing with a

reference value obtained from the same sample but annealed by vacuum oven annealing

at 280°C, it is evident that comparable or even larger exchange bias fields can be

achieved by laser annealing at broadly all scan speeds by choosing the appropriate range

of laser intensities. The evolution in the exchange bias field with the increasing laser

intensity is similar to the results discussed in section 5.2.2 and can be divided into three

different regimes, namely the incremental, plateau and a decremental regime based on

the trends observed in HEB (cf. section 5.2.2).

Figure 49. The exchange bias field strength as a function of the CW laser intensity at different scanning speeds applied on a (0.5 × 0.5) mm2 sample areas for Ru (a.) and Ni81Fe19 (b.) seed layer. The horizontal dashed bar indicates the maximum exchange bias field strength (including the error bars of the exchange bias value) achieved for vacuum oven annealing at 280°C.


77

The difference in the obtained maximum exchange bias field strength (average value of

the exchange bias field overall laser intensities comprised in the plateau region) between

samples with a Ru or Ni81Fe19 seed layer are shown in Figure 50.

Figure 50. The difference in the exchange bias field strength determined from the measurements shown in Figure 49. In addition, results of oven annealing at different temperatures are shown in the inset. The samples with a Ru seed layer present a larger exchange bias, both for the oven and laser annealing, with more pronounced differences arising in the latter.

As depicted in the inset, the larger exchange bias field strength for the samples with Ru

seed layer is found for both laser and oven annealed samples, with more pronounced

differences after laser annealing, where a maximum of ΔHEB = 18 mT is found (compared

to ΔHEB ≤ 7 mT). However, for a laser scanning speed of 5000 mm·s-1, no difference in

the pinning field could be observed, which is attributed to difficulties in data evaluation as

the MOKE hysteresis loops for those samples show a pronounced non-linear background.

No other reasonable and physical correct explanation could be found.

The reason for the larger exchange bias field strength with the Ru seed layer is an

improved (111)-texture of the IrMn layer126,127 as proved by XRD (see Figure 51). The

IrMn(111) peak intensity is almost doubled compared to the samples with a Ni81Fe19 seed

layer, which is the case for all laser annealed samples as well (not shown). Note that

Ru(002) also contributes to the detected IrMn(111) peak, as the peak centre of gravity

shifts slightly to larger 2θ values, but the peak at 2θ = 41.36° matches perfectly the

literature value for IrMn(111) [powder diffraction file of IrMn (00-029-0687) from the

International Centre for Diffraction Data (ICDD)] in case of the growth on top of a Ni81Fe19

seed layer. The Ru peak is not expected to contribute significantly to the enhancement

on the IrMn(111) peak intensity, especially taking into account the small Ru layer

thickness. Furthermore, the crystallite size determined from the pronounced Kiessig


78

fringes is larger for Ru as given in each figure, although the nominal IrMn layer thickness

is constant for all the samples, supporting the expected epitaxial growth of IrMn on Ru128.

Figure 51. XRD θ–2θ scans of the samples with (a.) Ru and (b.) NiFe seed layer as-deposited and annealed in an oven at 320°C and 400°C. The IrMn(111) orientation (2θ = 41.36°) is represented by the dashed lines, highlighting a small peak shift in (a.) due to the presence of the (002)-textured Ru seed layer (2θ = 42.15°). The given crystallite size was determined from the pronounced Kiessig fringes.

On the other hand, and although the thin film texture does not suffer by annealing at

temperatures up to 400°C, significant changes in the magnetic properties occur for oven

annealing above 320°C. The difference in exchange bias field strength for Ru and NiFe

samples vanishes and the subsequent degradation of the exchange bias (see Figure 50

inset) at these temperatures can be attributed to a degradation of the FM/AFM interface

responsible for the exchange bias effect, as B and even more Mn diffusion (see chapter 6

and 129,130) is likely to occur. Hence, the exchange bias field strength of the studied layer

stacks in general starts to decrease for annealing temperatures larger than about 280°C,

no matter what seed layer material is used.

5.2.4 Influence of the pattern size and number of scan repetitions for laser

annealing

This section investigates systematically the influence of the size of the laser-treated area

on the resulting value of the exchange bias field. For this purpose, the results discussed

in the previous section from the test patterns of (0.5 × 0.5) mm2 on a sample piece are

now compared to the laser annealing on a single sample piece of (6 × 6) mm2 each with

the same layer stacks. Similar to the studies discussed in section 5.2.4, CW laser

annealing was performed at different scanning speeds as well as laser intensities and the


79

magnetisation reversal was measured using NanoMOKETM2, concentrating on the PL

reversal. The change in the laser intensity ΔI = I(0.5 × 0.5) mm2 - I(6 × 6) mm

2 required to induce

the same exchange bias field strength in (0.5 × 0.5) mm2 square patterns compared to

the (6 × 6) mm2 sample pieces is plotted in Figure 52 as a function of the laser scanning

speed.

The fact that the same exchange bias field strength is achieved for lower laser intensities

in the large laser-treated patterns is a consequence of the scanning speed-dependent

local heat input in competition with the heat dissipation to the ambience and the substrate.

Whereas for a scan speed of 5 mm·s-1 an equal exchange bias field strength is achieved

for approximately the same laser intensities, the required intensity decreases with

increasing scanning speed in the larger pattern. As the area increases, more energy is

provided to the sample which is accumulated throughout the whole laser annealing

process, thus, increasing the overall sample temperature, especially for a heat input

happening much faster than the heat dissipation at increasing laser scanning speeds.

Figure 52. Change in the laser intensity I = I(0.5 × 0.5) mm

2 - I(6 × 6) mm2 needed to induce the same exchange

bias field strength in (0.5 × 0.5) mm2 square patterns (Figure 55) and (6 × 6) mm2 square patterns as a function of the laser scanning speed.

This is further supported by the results of an experiment in which the laser scanning was

repeated several times on the same trace (see Figure 53). With an increasing number of

scan repetitions at an exemplary scanning speed of 5000 mm·s-1 and fixed laser intensity

of 480 kW·cm-2, a clear deterioration of the exchange bias field strength is observed,

already after five repetitions.


80

Figure 53. The exchange bias field strength as the function of the number of CW laser annealing

repetitions at 5000 mm·s-1 scan speed and 240 kW·cm-2 intensity.

For comparison, vacuum oven annealing was performed in SQUID–VSM, where similar

dependencies of exchange bias field strength and the temperature (cf. Figure 49) can be

observed with a maximum exchange bias field strength obtained after field cooling from

250°C (heating and cooling rate of 15 K·s- 1). Higher annealing temperatures lead to a

reduction of the exchange bias (see Figure 54), as it is also the case for higher laser

powers (compare Figure 49), concomitant with deteriorating diffusion processes as

discussed earlier. However, in-field vacuum annealing at 300°C repeated for eight times

does not lead to a relevant change of the exchange bias field strength thereafter, as

shown in Figure 54.

Figure 54. MH hysteresis loops of the sample with layer stack Si / SiO2(100 nm) / Ta(5 nm) / Ru(5 nm) / IrMn(8 nm) / Co40Fe40B20(2.3 nm) / MgO(1.8 nm) / Co40Fe40B20(2.3 nm) / Ta(5 nm) as-deposited (black), after one annealing cycle at 250°C (red), followed by nine cycles of annealing at 300°C (showing only iterations #2 and #9, since no significant changes were found)


81

These findings indicate that the degradation of the exchange bias upon repeated

annealing by laser result from intrinsic structural or chemical changes of the IrMn / CoFeB

system, which are not caused by the subsequent reheating, but rather caused by a steady

increase of the sample temperature with each heating cycle. The laser annealing is

accompanied by very large temperature gradients and by time effects due to the scanning

of the laser beam. These effects may induce not only element diffusion and alloying, but

also large stress, probably influencing the magnetic properties of the films. In addition,

note that the laser annealing is performed under ambient conditions, where

contaminations cannot be excluded. However, no reasonable increase in surface

roughness was observed by AFM in any laser annealed sample at those moderate laser

powers with constant exchange bias.

5.3 Application of FORC analysis: from single CoFeB layer to MTJ layer

stack

The in-plane magnetic hysteresis loops of the MTJ stacks measured in the as-deposited

sample and after setting exchange bias (cf. Figure 44.) shows a very small switching field

distribution for the free layer and a gradual switching of the pinned layer due to exchange

biasing. To further understand the switching field behaviour of the individual layers, FORC

analysis was conducted. This sub-chapter summarises the results obtained for a single

CoFeB layer, for an exchange biased CoFeB layer and for the complete MTJ stack, finally.

Single CoFeB layer

Figure 55 shows the magnetisation loops and the FORC distribution of a 100 nm thick

Co60Fe20B20 film passivated with 5 nm Pt in the as-deposited state. This layer mimics the

free layer in MTJ. A single narrow symmetric distribution around the HI axis can be seen

in the FORC plot, indicating the absence of any type of exchange bias or interaction

between the magnetic domains of CoFeB. The FORC distribution is centred at HC = 0 mT.

Such a behaviour is often described as the reversible part of the FORC distibution131 and

ascribed to a superparamagnetic (SP) behaviour131. In this case, it can be explained by

the fact the as-deposited (amorphous) CoFeB layer has very small grains which have a

magnetisation relaxation time similar to or lower than the average measurement time132,

hence resulting in HC→0 mT. The fact that the FORC distribution centred at zero is not

sharp but exhibits a small tail suggests that the film also contains grains which show

typical single domain (SD) switching133. Such grains might consist of CoFe.


82

Figure 55. Magnetisation loop (a.) and contour plot presenting the FORC distribution (b.) of the sample with Si / SiO2(1.8 nm) / Co60Fe20B20(100 nm) / Pt(5 nm) in the as-deposited state. Both measurements were performed by SQUID-VSM.

Exchange biased CoFeB

The FORC distribution of the as-deposited Si/SiO2(100 nm) / Ta(3 nm) / Ru(3 nm) /

IrMn(8 nm) / Co40Fe40B20(5 nm) / Ta(3 nm) layer stack is shown in Figure 56 a. The most

obvious influence of the addition of the AFM layer into the stack on the FORC distribution

is the occurrence of the relocation of the maxima on both HC and HI axis. The fact that the

maxima are not symmetric around the HI = 0 is due to random pinning sites134,135 at the

Co40Fe40B20 / IrMn interface. The main contributions to the FORC distribution are located

along +HI, suggesting that most of the pinned domains have preferred magnetisation

reversal direction along positively applied fields. Similar to the free layer FORC

distribution, there is a finite magnetisation at HC = 0 mT (for HI between 0 mT and 20 mT),

which indicates that the amorphous Co40Fe40B20 grains are small but have strong

interaction due to the exchange bias field.


83

Figure 56. FORC distribution of a Si / SiO2(100 nm) / Ta(3 nm) / Ru(3 nm) / IrMn(8 nm) / CoFeB(5 nm) / Ta(3 nm) layer stack in the as-deposited state (a.) and after setting the exchange bias by cooling from 200°C in the presence of 120 mT magnetic field(b.).

Figure 56 b shows the FORC diagram of the same layer stack after annealing at 200°C

for 30 min, followed by cooling in the presence of an external magnetic field of 120 mT.

In this case, one pronounced maximum with a well defined FORC contour (HI ≈ 40 mT,

HC ≈ 10 mT) can be observed. This clearly shows the influence of exchange bias in

Co40Fe40B20 films134. A second, smaller contour occurs at HI ≈ 30 mT and HC ≈ 0 T,

indicating that although the applied heat treatment and magnetic field were sufficient to

set the exchange bias, the Co40Fe40B20 film still contains some SP domains. This means

that the Co40Fe40B20 is not completely crystalline after 30 min of annealing at 200°C.

MTJ layer stack

Figure 57 shows the magnetic hysteresis loop and the FORC distribution contour of the

MTJ layer stack consisting of Si / SiO2(100 nm) / Ta(5 nm) / CuN(30 nm) / Ta(3 nm) /

Ni81Fe19(2 nm) / IrMn(8 nm) / Co40Fe40B20(2 nm) / MgO(1.8 nm) / Co40Fe40B20(2.2 nm) /

Ta(10 nm) after oven annealing at 227°C for 30 min and subsequent cooling in the

presence of an applied magnetic field (120 mT). As expected, the FORC countors of the

MTJ stack are the superposition of the free and pinned layers reversal. Due to the large

difference between the switching field distribution of the free and pinned layer, the FORC

distribution contour for each layer appears feeble in the unified FORC plot.


84

Figure 57. MH loop (a.) and FORC distribution (b.) recorded at RT for a Si / SiO2(100 nm) / Ta(5 nm) / CuN(30 nm) / Ta(3 nm) / Ni81Fe19(2 nm) / IrMn(8 nm) / Co40Fe40B20(2 nm) / MgO(1.8 nm) / Co40Fe40B20(2.2 nm) / Ta(10 nm)” MTJ layer stack after setting the exchange bias.

The closer look at the individual layer reveals that the FORC distribution of the pinned

layer is centred around HI = -110 mT suggesting the pinning of the CoFeB adjacent to

IrMn with the coercive field distribution concentrated around HC = 10 mT. This FORC

distribution is elongated along the HC direction, indicating a distribution of switching fields

due to the interaction of several exchanged coupled single domains of CoFeB. The FORC

contribution from the free layer is concentrated around HI = 0 mT(at HC ≈ 2 mT) axis,

indicating that the free layer is independent of any coupling due to the exchange-coupled

CoFeB layer. Further, it can be observed that the FORC distribution of the free layer is

not symmetrical with respect to HI = 0 mT. This may be an artefact due to a too large

measurement step compared to the very narrow switching field reversal of the free layer.

5.4 Potential of direct-write laser annealing technique

With the knowledge gained about the laser annealing method, an experiment was devised

to image the magnetic domains in FM/AFM heterostructures on the laser annealed and

unannealed areas. For this, a simpler layer stack consisting of only the pinned layer was

used “Si / SiO2(100 nm) / Ta(3 nm) / Ru(3 nm) / IrMn(8 nm) / CoFeB(5 nm) / Ta(3 nm)”.

The text “HLPH”, abbreviation for “Halbleiterphysik” was written with a CW laser beam of

380 kW·cm-2 intensity at 50 mm·s-1 scanning speed while an applied in-plane magnetic

field of 120 mT under ambient atmosphere. MOKE images were recorded using the

MOKE microscopy at the Leibniz-Institut für Festkörper- und Werkstoffforschung (IFW),


85

Dresden by Dr. Ivan Soldatov with a home-built MOKE microscope136. The hysteresis

loops were derived from the grey contrast of two areas-of-interest recorded with the

sweeping magnetic field. An exemplary MOKE microscopy image captured under an

applied magnetic field of H = 25.5 mT, along with the hysteresis loop measured at the

exposed and unexposed region is shown in Figure 58.

It is clearly visible from the hysteresis loop measured at two different regions that the

application of laser annealing in the presence of a magnetic field leads to a shift of the

hysteresis loop toward positive magnetic fields due to the induced exchange anisotropy,

see Figure 58 b. In order to understand the grey contrast in the MOKE images, the

Meiklejohn and Bean model can be used. A magnetic field < -60 mT is applied parallel to

the sample surface (in-plane), the CoFeB layer is saturated in the direction of the applied

magnetic field, resulting in monochromatic grey images. When the magnetic field is swept

towards positive values (opposite to initial direction) through H = 0 mT, the magnetic

domains of CoFeB follow the direction of the applied magnetic field. However, the

domains which are pinned due to exchange bias at the interface between CoFeB and

IrMn hinder this rotation. This results in a different grey contrast for the laser-treated and

untreated regions, which makes the text visible. To rotate the pinned CoFeB

magnetisation in the laser treated regions, a higher magnetic field has to be applied to

overcome the CoFeB / IrMn exchange interaction, a fact which is observed as a shift of

the hysteresis loop along the positive direction of the applied magnetic field. Finally, when

the applied magnetic field reaches the positive values required to induce the saturation

magnetisation, the MOKE microscopy image shows a monochrome grey tone with no

visible text, similar to the situation when the sample was saturated in negative applied

magnetic fields. When the field is swept from the positive saturation field towards negative

magnetic fields, the magnetisation of the pinned CoFeB regions rotates at the smaller

fields, because in this case, the exchange bias at the FM/AFM interface aids the reversal.

Hence, in the full cycle of magnetic field sweep, the text is visible only under a positive

applied magnetic field. On the other hand, in the negative field, the magnetisation for both

pinned and unpinned CoFeB remains saturated; therefore, no text is visible. Here, it is

worth mentioning that the contrast visible in the MOKE micrographs is due to the

difference in the reversal of laser exposed as compared to the unexposed region. The

magnetic domains of CoFeB are very small due to the pinning at the AFM/FM interface

and are beyond the optical resolution limits.


86

Figure 58. MOKE microscopy image of the text “HLPH” (“Halbleiterphysik”) inscribed with a CW laser beam on the Si /SiO2(100 nm) / Ta(3 nm) / Ru(3 nm) / IrMn(8 nm) / CoFeB(5 nm) / Ta(3 nm) layer stack. The grey contrasts in the image have the pure magneto-optical origin and correspond to different magnetisation orientations in the FM layer. The hysteresis loop measured on the laser exposed (red) and unexposed region (green); the corresponding area-of-interest is marked in the MOKE image.

5.5 Conclusion

In summary, the results presented in this chapter successfully demonstrate that the laser

annealing can be used for the selective alignment of the magnetisation of the pinned layer

in a typical MTJ. The exchange bias was induced at the AFM/FM interface by annealing

using two methods: large area-annealing in a vacuum and localised heating using laser

irradiation, followed by subsequent cooling in the presence of an external magnetic field.

For this, the temperature on one hand, and the laser intensity, scanning speed, and mode

(PW & CW) on the other hand, were systematically varied to set the maximum exchange

bias. The benchmark values for the exchange bias and the coercive field were obtained

by homogenous heating of the layer stack within the SQUID-VSM magnetometer. The

maximum exchange bias field obtained with CW laser irradiation was marginally lower

than the homogenously heated sample (SQUID-VSM) and the exchange bias set with

pulsed laser annealing was ~18% lower than that achieved with CW laser annealing.

It was shown that in-field laser annealing results in comparable exchange bias field

strengths with larger process windows for increasing scanning speed, being applicable

for different layer stacks. For laser intensities within this process window, no deterioration

of the texture, as well as surface roughness, was observed. Significant changes in HEB

and HC were observed when increasing the laser processed area from (0.5 x 0.5) mm2 to

(6 x 6) mm2. This indicates that there is a delicate dependence of the results of the laser


87

annealing process on the heat propagation through the thin film system into the substrate

or heat dissipation to the ambience.

A systematic study of a CoFeB, CoFeB/AFM, and MTJ layer stack using first order

reversal curve measurements present the insight about the irreversible and reversible

magnetisation of a free layer and pinned layer in an MTJ. FORC diagram of the single

CoFeB layer was observed to be concentrated only at one region around HC = 0 mT

(HI = 0mT) indicating that the layer consists of very fine grains of CoFe which show a

superparamagnetic behaviour. For the CoFeB / IrMn layer stack, FORC plots

demonstrate both main distributions due to exchange bias and satellite distribution

associated with CoFeB layer and inhomogeneity at the CoFeB / IrMn interfaces. The

FORC distributions of the MTJ revealed that the coercivities field distribution has two

distinct regions after setting the exchange bias. The first region around HI = 0 mT

produced due to the top CoFeB (FL) whose coercive field distribution is similar to that of

observed for the single CoFeB layer. The distribution further suggests that the FL have

no magnetic interaction to the other constituent layers. The second region is due to the

CoFeB adjacent to the IrMn where the spread of the coercive field is significantly wider

than that of FL and centred at HI = - 110 mT indicating the strong exchange bias at

CoFeB / IrMn interface.

Using MOKE microscopy, the reversal mechanisms of the laser annealed and the

unannealed regions were investigated. Although the magnetic domains of the pinned

CoFeB layer were not visible due to the optical resolution limits, the qualitative information

from the contrast of the recorded images was sufficient to demonstrate the functionality

of laser annealing as a direct-write technique.

Exchange bias and diffusion processes in laser annealed CoFeB/IrMn thin films

Chapter 6

88

Chapter 6: Exchange bias and diffusion processes in

laser annealed CoFeB/IrMn thin films

In the previous chapter, the results suggested that using Ru as a seed layer improves the

exchange bias due to the improved crystalline texture of IrMn. However, the performance

of an exchange biased system strongly relies on the overall integrity of the constituent

layers in the stack. This chapter presents the influence of laser-induced localised

annealing on the magnetic properties and the diffusion processes occurring in the Si /

SiO2(100 nm) / Ta(3 nm) / Ru(3 nm) / IrMn(8 nm) / CoFeB(5 nm) / Ta(3 nm) layer stack,

using X-ray photoemission spectroscopy depth profiling technique. The results are

compared to those obtained by standard vacuum oven annealing and correlated to the

magnetic properties investigated by magneto-optical Kerr effect magnetometry. The

following chapter was previously published in the Journal of Magnetism and Magnetic

Materials, DOI: 10.1016/j.jmmm.2019.165390.

6.1 Introduction

The large TMR yields rely on thermal treatment to crystalise the CoFeB/MgO layers137,138

together with an appropriate boron migration from the CoFeB103,139. In sensor

applications, this procedure is furthermore crucial to set a reference magnetisation

through the exchange bias effect, appearing as a uniaxial magnetic anisotropy in strongly

coupled AFM/FM thin film systems. Although the migration of boron was reported to be

beneficial for achieving a large TMR ratio, the diffusion mechanisms underlying the

thermal treatment of these devices remain yet controversial. For instance, B migrated

from CoFeB during crystallisation was suggested to be incorporated in the MgO barrier,

providing coherent tunnelling for spin-polarised electrons and hence, improving TMR140–

142. On the contrary, other studies reported that the diffusion of B distorts the tunnel barrier

adversely, resulting in the degradation of TMR143,144. Large TMR ratios were also ascribed

to the migration of B into the Ta capping layer with the formation of Ta-B86,140,145,146.

Furthermore, amorphous CoFeB may work as a suitable diffusion barrier for Mn from the

IrMn antiferromagnet, whereas certain crystalline phases of CoFe facilitate the diffusion

of Mn through vacancies in grain boundaries. Previous studies were so far limited to

conventional oven annealing and rapid thermal annealing processes147.


Chapter 6

89

6.2 Magnetometry investigations

The MOKE magnetometry results obtained from CW laser-treated samples with the layer

stack Si / SiO2(100 nm) / Ta(3 nm) / Ru(3 nm) / IrMn(8 nm) / CoFeB(5 nm) / Ta(3 nm) for

various scanning speeds and irradiation intensities are shown in Figure 59 a. As already

discussed in the previous chapter, the values of the exchange bias field as a function of

the laser intensity can be segmented into three regions for each scanning speed. Despite

the similar trend, the absolute values of HEB and HC differ considerably (by a factor of 2)

due to the different FM layer thicknesses. The choice of the increased CoFeB thickness,

in this case, relates to the requirements for the investigations by XPS depth profiling.

Figure 59 b shows HEB and HC as a function of the annealing temperature for a sample

that was thermally treated in the macro-MOKE chamber.

Figure 59. Exchange bias (red) and coercive field (black) strength determined by MOKE magnetometry as a function of laser intensity and laser scanning speed for Si / SiO2(100 nm) / Ta(3 nm) / Ru(3 nm) / IrMn(8 nm) / CoFeB(5 nm) / Ta(3 nm) sample (a.). The hashed areas in red and black represent the values of exchange bias and coercive field (with the error bars) obtained for a sample annealed in vacuum at 200°C. The coercivity (HC) and exchange bias field (HEB) temperature dependence for the oven annealed sample is shown in (b.). The lines in the figure are guide for the eye.

HEB increases as the annealing temperature increases up to a maximum value of

HEB = (63 ± 2) mT at HC = (11.5 ± 0.5) mT for the sample annealed at 200°C. A further

increase in annealing temperature results in a gradual decrease in HEB. The HC values

compared to the as-deposited sample were observed to decrease with increasing

annealing temperature till 400°C and show an increment once again upon annealing at

500°C. The trend in HEB and HC as a function of temperature is similar to that observed

for laser annealed samples as a function of laser intensity. A maximum of


Chapter 6

90

HEB = (58 ± 2) mT and HC = (11.4 ± 0.5) mT was recorded for the sample laser annealed

with 380 kW·cm-2 at a scanning speed of 50 mm·s-1.

6.3 XPS-depth profiling

6.3.1 Laser annealed samples

The XPS depth profiling was performed according to the procedure described in section

3.3.5. For the XPS measurements, a distribution profile of all constituent elements is

compared to the nominal layer stack of the as-deposited sample in Figure 60. Even though

oxygen adsorbed at the surface in the form of tantalum oxide is observed in the first three

Ar+ milling steps, its concentration reduces significantly within the Ta layer, inferring that

the layer stack underneath is preserved and not oxidised due to passivating Ta-O. A small

fraction of B can also be observed in the IrMn layer due to the high energy Ar ion sputtering

induced migration of B atoms. In conclusion, all the elements show a well-defined

distribution in accordance with the nominal layer stack, implying that the layers have sharp

interfaces in the as-deposited state.

Figure 60. XPS depth profile of an as-deposited layer Si / SiO2(100 nm) / Ta(3 nm) / Ru(3 nm) / IrMn(8 nm) / CoFeB(5 nm) / Ta(3 nm) sample. The thickness values provided in the layer stack and the upper part of the graph are the nominal values used for the deposition process.

The XPS depth profile of the laser annealed samples at laser intensities of 120 kW·cm-2,

380 kW·cm-2, and 900 kW·cm-2 for all the scanning speeds is shown in Figure 61. To ease

the presentation of results, only Co, Mn, Ru, and O normalised atomic distributions are

plotted. At 120 kW·cm-2 (Figure 61 a) the distribution profiles of all considered elements


Chapter 6

91

are similar to those of the as-deposited samples, and no change is detectable with the

scanning speed, as expected due to the small induced rise in temperature, being well

below TN. Up to 380 kW·cm-2 laser intensity (Figure 61 b), no diffusion can be identified,

proving that the layers remain intact. The lower HEB after laser annealing compared to

oven annealing, as well as the decrease of HEB with increasing scanning speed shown in

Figure 59 should therefore not be ascribed to significant diffusion of Mn. However, this

cannot be fully clarified with the experimental study performed here, as the depth

resolution of the method does not allow to study changes exactly at the AFM/FM interface

that cause the exchange bias effect. Furthermore, in the particular case of the scanning

speed of 50 mm·s-1, it can be argued that closer study in the range of 250 kW·cm-2 and

380 kW·cm-2 of laser intensities could reveal larger HEB values. At higher laser intensities,

dramatic changes are observed (Figure 61 c).

Figure 61. Distribution profiles of Co, Mn, Ru, and O for the layer stack Si / SiO2(100 nm) / Ta(3 nm) / Ru(3 nm) / IrMn(8 nm) / CoFeB(5 nm) / Ta(3 nm) laser annealed at different laser scanning speeds as indicated in the legend for (a.) 120 kW·cm-2, (b.) 380 kW·cm-2, and (c.) 900 kW·cm-2 laser intensities.

Namely, substantial diffusion of Mn towards the surface for all scanning speeds, as well

as diffusion of Co towards the substrate for 50 mm·s-1 and 500 mm·s-1. For the lowest

laser scanning speed, Ru diffusion occurs too, as well as an increase in average surface


Chapter 6

92

roughness (σavg) from 1.6 nm to 13 nm (RMS surface roughness increases from 0.5 nm

to 1.9 nm) as determined by atomic force microscopy. These results confirm that for laser

intensities ≥900 kW·cm-2, the degradation in HEB and HC is effectively due to the migration

of Mn from the IrMn alloy to CoFeB and finally, intermixing of all the layers. At slower laser

scanning speed, a larger heat load and consequently a larger temperature increase

occurs within the layers, ultimately leading to the diffusion of Mn and Co. The oxygen

distribution profile remains almost unchanged with no considerable diffusion toward the

CoFeB / IrMn / Ru layers for all intensities, thus excluding any oxidation of the layer stack

upon laser irradiation.

Figure 62. Normalised B distribution profile, for all laser annealed samples of the Si / SiO2(100 nm) / Ta(3 nm) / Ru(3 nm) / IrMn(8 nm) / CoFeB(5 nm) / Ta(3 nm) layer stack, indicating a strong migration of B towards the Ta capping layer and the Ru / Ta seed layer.

A closer inspection of the B1s XPS spectra reveals no signature of boron oxide being

present next to elemental boron for laser intensities up to 380 kW·cm-2 for none of the

three scanning speeds. At 900 kW·cm-2 and 50 mm·s-1 scanning speed, a significant peak

of boron oxide at 192.5 eV148 was recorded, see Figure 63 a, which vanishes after a

sputtering time of 100 s (Figure 63 b), meaning that the presence of oxygen in this sample

is limited to the surface. For 500 mm·s-1 a small (two less pronounced peaks at 187 eV

and 192 eV) and for 5000 mm·s-1 no response from boron oxide was observed. Due to

the strong intermixing of all layers and the low intrinsic photoemission cross-section of

B1s core levels, it has to be noted that it was challenging to deconvolute the B oxide peak

and the B metal peak from the background. The lack of B-O peak for 500 mm·s- 1 and

5000 mm·s-1, indicates that the Ta layer works as an effective passivation layer to prevent


Chapter 6

93

the oxygen diffusion in the underlying layer. Hence suggesting vacuum is not mandatory

for the laser annealing.

Figure 63. XPS B spectra recorded after a sputtering time of the 20s for 900 kW·cm-2 laser annealed samples with Si / SiO2(100 nm) / Ta(3 nm) / Ru(3 nm) / IrMn(8 nm) / Co40Fe40B20(5 nm) / Ta(3 nm) layer stack (a.). A clear boron oxide peak was detected at 192.4 eV (dashed line) for the lowest scanning speed only. The B-O peak vanishes after a sputtering time of 100s (b.).

6.3.2 Vacuum oven annealed samples

Pieces of size (6 × 6) mm2 from the same wafer (“Si / SiO2(100 nm) / Ta(3 nm) / Ru(3 nm)

/ IrMn(8 nm) / Co40Fe40B20(5 nm) / Ta(3 nm)”) were vacuum annealed in the macro-

MOKE chamber (P = 10-7 mbar) with an applied magnetic field of 120 mT at 100°C,

200°C, 300°C, 400°C, and 500°C for 30 min. Similar to the laser annealed samples, the

samples annealed in the oven were studied using the XPS depth profile technique, using

identical equipment in the HLPH-TUC group. In this case, binding energy spectra from

0 eV to 1.3 keV were surveyed. For the Ar+ sputtering ion energy of 200 eV was used to

mill (2 × 2) mm2 craters iteratively in 36 cycles with a sputtering time of 90 s. XPS spectra

were then obtained after each sputtering cycle from the centre of the crater with a

monochromatic Al Kα X-ray source and a spot size of 300 µm.

The results for the samples annealed in the presence of H = 120 mT magnetic field at

200°C for 30 min (where maximum HEB is observed see Figure 59 b) are shown in Figure

64 a. Similar to the laser annealed samples, only Co, Mn, and Ru normalised atomic

distributions are plotted. No noticeable diffusion of Mn in Co and Ru for the sample

annealed at 200°C can be observed, indicating the AFM and FM layers remain intact. The

depth profile of B in the layer stack is shown in Figure 64 b. The expected migration of B


Chapter 6

94

towards Ta was observed for temperatures as low as 200°C, whereas its concentration in

the CoFeB layer only starts decreasing significantly above 300°C. Above this

temperature, B was found not only in the Ta cap layer but also in the Ru seed layer,

indicating that the Ta and Ru have a higher affinity to absorb B compared to other

materials in the layer stack.

Figure 64. Depth profile of the atomic concentration of Co, Mn, and Ru calculated from XPS measurements of Si / Si(100 nm) / Ta(3 nm) / Ru(3 nm) / IrMn(8 nm) / CoFeB(5 nm) / Ta(3 nm) after vacuum field annealing at 200°C (a.). The B diffusion profile for as-deposited samples and samples annealed at 200°C, and 500°C (b.).

6.4 Structural analysis

Figure 65 shows the diffraction patterns recorded for the as-deposited state and after CW

laser annealed samples for 380 kW·cm-2 and 900 kW·cm-2 intensities for all three

investigated scanning speeds. The θ-2θ scans (recorded with the Rigaku SmartLab

equipment) show a pronounced IrMn(111) peak around 2θ = 41.5° with a Ru(002)

shoulder peak at higher angles. For none of the measured samples, any crystalline peak

of CoFe was observed, which is consistent with the XPS depth profile study. From the

XPS study it is known that even though the migration of B is visible for all the annealing

parameters, a major concentration of B can still be observed in the CoFeB layer, which

could explain why no CoFe crystalline phase was found in the XRD measurements.

Additionally, as was discussed in the previous section 6.2, the samples irradiated with

900 kW·cm-2 laser intensity at 50 mm·s-1 and 500 mm·s-1 scanning speed showed

intermixing of layers. A similar observation can be made from the XRD scans. The IrMn

peaks are shifted towards higher angles, indicating the disruption of the IrMn crystal

structure.


Chapter 6

95

Figure 65. θ-2θ scans for an as-deposited (yellow) and laser annealed with 380 kW·cm-2 (red), and 900 kW·cm-2 (blue) Si / SiO2(100 nm) / Ta(3 nm) / Ru(3 nm) / IrMn(8 nm) / CoFeB(5 nm) / Ta(3 nm) layer stack for scanning speeds of 50 mm·s-1, 500 mm·s-1, and 5000 mm·s-1. Expected positions for oriented crystalline structures are marked by vertical dashed lines.

6.5 Topographic characterisation

It is often considered that the topography of the layer stack could be used as an indicator

to reflect the FM/AFM interface quality. Hence, the average surface roughness (σavg) can

be used as a qualitative parameter to address the FM/AFM interface roughness. The σavg

as a function of the scanning speed for selected laser annealed samples is shown in

Figure 66. It can be seen that for the investigated scanning speeds, the samples irradiated

with 900 kW·cm-2 have higher roughness as compared to samples exposed to

380 kW·cm-2. For the samples irradiated with 900 kW·cm-2 laser intensity the σavg

decreases with increasing scanning speed. A maximum of σavg = (13 ± 1) nm was

observed for the sample treated at a scanning speed of 50 mm·s-1, where the XPS profile

showed a significant intermixing, i.e. the disruption of the FM/AFM interface. On the other

hand, the samples exhibiting the highest exchange bias (laser-treated at 380 kW·cm-2 and

oven annealed at 200°C, see Figure 59) showed a non-deteriorated surface roughness

σavg comparable to the as-deposited state. It is, thus, possible to conclude that in the

samples with high exchange bias values, the interface between FM/AFM is very smooth.

These results complement the XPS depth profile results very well.


Chapter 6

96

Figure 66. Average surface roughness (σavg) of the Si / SiO2(100 nm) / Ta(3 nm) / Ru(3 nm) / IrMn(8 nm) / CoFeB(5 nm) / Ta(3 nm) layer stack irradiated with 380 kW·cm-2 (red) and 900 kW·cm-2 (yellow) laser intensity for the investigated scanning speeds. The average surface roughness for the 200°C oven annealed samples are shown as dashed blue line. Insets are the 3D topography images of the selected samples (in green: laser annealed samples; in grey: oven annealed samples).

6.6 Conclusion

In this chapter, the effect of different laser annealing parameters on exchange-coupled

FM/AFM layer systems was investigated, using a Si / SiO2 / Ta / Ru / IrMn / CoFeB / Ta

layer stack. The evolution of the magnetic properties determined from MOKE

magnetometry can be well explained by the diffusion process induced by the laser (or

oven) annealing and revealed by XPS depth profiling. The diffusion of B was already

observed below TN for laser as well as oven annealed samples. Increasing laser intensity

and a decreasing laser scanning speed results in the diffusion of Mn towards the surface,

which is the primary reason for the decrease in HEB. At even higher laser powers or slower

scanning speeds, also other materials start to migrate. In comparison to oven annealing,

it was found that laser annealing, despite the shorter time-scale with ultra-fast and

spatially very strong confined temperature gradients, induces comparable diffusion

processes and dependencies, mainly linked to the studied material diffusions (B, Mn, Co,

Ru) and the exchange bias field strength. Finally, and even though the B distribution by

XPS depth profile measurements is difficult to be determined because of its small intrinsic

photoemission cross-section, its profile is a direct proof that exchange bias is independent

on the B migration.

Summary Chapter 7

97

Chapter 7: Summary and outlook

This thesis focuses on investigating the most important prerequisites for an efficient and

sensitive CoFeB based magnetic tunnel junction device, i.e., crystallisation of CoFeB,

setting the exchange bias in FM/AFM layer at the microscopic level using the laser

annealing (direct-write technique), as well as the selective alignment of magnetisation in

the reference layer, otherwise referred as the direct-write technique for the directional

sensitivity.

The crystallisation of CoXFe(80-X)B20 alloys triggered by vacuum oven annealing was

investigated using X-ray diffraction and scanning electron microscopy, as well as optical

and magneto-optical Kerr effect spectroscopy for annealing temperatures ranging from

300°C to 600°C. The transformation of ~100 nm thick CoXFe(80-X)B20 films from amorphous

CoFeB to polycrystalline CoFe was revealed by the sharpening of spectral features

observed in optical and magneto-optical dielectric functions spectra. The influence of B

on the dielectric function was assessed both experimentally and by optical modelling. By

analysing the Drude component of the optical dielectric function, a consistent trend

between the charge carrier scattering time/resistivity and the annealing temperature was

observed, in agreement with the results of electrical investigations by means of the four-

point-probe method. In the case of Co60Fe20B20 thin films, interestingly, more spectral

features were observed in the dielectric spectra after annealing in comparison to the

~100 nm thick films. Though the origin of these additional features in the dielectric spectra

can not be pinpointed, the evolution observed in the spectra with annealing certainly

proves that the (magneto-) optical spectroscopies were very sensitive to the crystallisation

of the investigated layers. Thus this thesis presents a non-destructive, high precision,

swift, and highly sensitive approach to probe the crystallisation of ultra-thin CoXFe(80-X)B20

films based on (magneto-) optical spectroscopy and highlights the advantages of such

spectroscopy techniques for the process and material optimisation in the field of

spintronics. The knowledge gained in this work will be expanded to probe the laser

annealing induced crystallisation of CoFeB in magnetic tunnel junctions, which otherwise

is not possible using the state-of-the-art techniques (XRD) due to the ultrathin films and

multiple layers in the layer stack. It has also promoted the interest in investigating the

dynamic change in the optical properties of the CoFeB during the laser annealing in

collaboration with LHM, Mittweida.

Summary Chapter 7

98

This work gives first insight into the applicability of laser annealing for setting the exchange

bias in magnetic tunnel junctions. It treats some of the challenges concerning the optimal

layer stack and optimal laser annealing parameters for the magnetic tunnel junctions. At

first, the influence of laser intensity, scanning speed, and laser mode (CW & PW) on the

magnetic properties of a typical magnetic tunnel junction layer stack was investigated.

The observed changes in the magnetic properties were compared with similar samples

annealed using a standard vacuum oven, as the benchmark process. It was observed that

with the CW mode laser annealing an exchange bias comparable to the conventional oven

annealing method could be achieved. The optimal laser intensity window for setting the

maximum exchange bias increases with the increase in scanning speed in the

investigated layer stacks with no obvious degradation in structural and magnetic

properties. During the course of this work, the first order reversal curve technique was

applied to investigate the complex magnetisation reversal of a magnetic tunnel junction.

This promising approach will be extended for measuring magnetoresistive first order

reversal curves in collaboration with Fraunhofer ENAS. The combined information from

the magnetisation and magnetoresistance first order reversal curve measurements will

shed new lights on magneto-transport characteristics of the microfabricated magnetic

tunnel junctions.

Finally, the diffusion processes occurring at the interface of CoFeB / IrMn due to laser

annealing were probed using X-ray photoemission spectroscopy depth profiling

technique. These results were compared to those obtained by standard vacuum oven

annealing and correlated to the magnetic properties investigated by magneto-optical Kerr

effect magnetometry. It can be concluded from the depth profiling that the degradation in

exchange bias depends solely on the dealloying AFM material (IrMn) and is independent

of the migration of boron in the layer stack. The (magneto-) optical spectroscopic studies

in combination with the results obtained from depth profiling will help in developing non-

destructive means to probe interlayer diffusion in the microfabricated sensors.

The results of this work have laid the foundation for the future development of the 3D

magnetic field sensor based on the magnetic tunnel junction devices. For such sensors,

the laser annealing has been found to be a promising technique for achieving local and

selective realignment of the magnetisation for directional sensitivity. An important

progress has been made by the combined assessment of the structural, magnetic

Summary Chapter 7

99

properties and by proving the efficacy of optical spectroscopic techniques for optimising

the spintronic device fabrications process.

Appendix

100

Appendix A

Units for magnetic properties

Quantity Symbol Gaussian & cgs

emu

Conversion

factor, C

SI &

rationalised

mks

Magnetic flux density,

magnetic induction B Gauss (G) 10-4 Tesla (T), Wb·m-2

Magnetic flux Φ Maxwell (Mx), G·cm2 10-8 Weber (Wb),

Volt second (V·s)

Magnetic potential difference,

magnetomotive force U, F Gilbert (Gb) 10/4π Ampere (A)

Magnetic field strength,

magnetising force H

Oersted (Oe); Gb·cm-

1 103/4π A·m-1

(Volume) magnetisation g M emu·cm3 h 103 A·m-1

(Volume) magnetisation 4πM G 103/4π A·m-1

Magnetic polarisation,

intensity of magnetisation J, I emu·cm3 4π·10-4 T, Wb·m-2

(Mass) magnetisation σ, M emu·g 1

4π·10-7

A·m2·kg-1

Wb·m·kg-1

Magnetic moment m emu, erg·G 10-3 A·m2, Joule per

Tesla (J·T-1)

Magnetic dipole moment j emu, erg·G 4π·10-10 Wb·m

(Volume) susceptibility Χ, κ dimensionless,

emu·cm3

4π

(4π)2·10-7

Dimensionless

Henry per meter

(H/m), Wb/(A·m)

(Mass) susceptibility Χρ, κρ cm3·g, emu·g 4π·10-3

(4π)2·10-10

m3·kg

H·m2·kg-1

(Molar) susceptibility Χmole,

κmole cm3·mol, emu·mol

4π·10-6

(4π)2·10-13

m3·mol-1

H·m2·mol-1

Permeability μ dimensionless (4π)2·10-7 H·m-1, Wb·(A·m)-1

Relative permeability i μr dimensionless dimensionless

(Volume) energy density,

energy product k W erg·cm3 10-1 J·m-3

Demagnetisation factor D,N dimensionless 1/4π dimensionless

Multiply a number in Gaussian units by “C” to convert it to SI.

Bibliography

101

Bibliography

1. Li, F. & Kosel, J. An efficient biosensor made of an electromagnetic trap and a magneto-resistive sensor. Biosens. Bioelectron. 59, 145–150 (2014).

2. Shen, W., Liu, X., Mazumdar, D. & Xiao, G. In situ detection of single micron-sized magnetic beads using magnetic tunnel junction sensors. Appl. Phys. Lett. 86, 253901–253903 (2005).

3. Pannetier-Lecoeur, M., Parkkonen, L., Sergeeva-Chollet, N., Polovy, H., Fermon, C. & Fowley, C. Magnetocardiography with sensors based on giant magnetoresistance. Appl. Phys. Lett. 98, 153705–153703 (2011).

4. Chaves, R. C., Freitas, P. P., Ocker, B. & Maass, W. Low frequency picotesla field detection using hybrid MgO based tunnel sensors. Appl. Phys. Lett. 91, 102504 (2007).

5. Treutler, C. P. Magnetic sensors for automotive applications. Sensors Actuators A Phys. 91, 2–6 (2001).

6. Malinowski, G., Hehn, M., Montaigne, F., Schuhl, A., Duret, C., Nantua, R. & Chaumontet, G. Angular magnetic field sensor for automotive applications based on magnetic tunnel junctions using a current loop layout configuration. Sensors Actuators, A Phys. 144, 263–266 (2008).

7. van Schalkwyk, J. A. & Hancke, G. P. Energy harvesting for Wireless Sensors from electromagnetic fields around overhead power lines. in 2012 IEEE International Symposium on Industrial Electronics 1128–1135 (IEEE, 2012).

8. Hancke, G., Silva, B. & Hancke, Jr., G. The Role of Advanced Sensing in Smart Cities. Sensors 13, 393–425 (2012).

9. Lee, C., Han, Y., Jeon, S., Seo, D. & Jung, I. Smart parking system for Internet of Things. in 2016 IEEE International Conference on Consumer Electronics (ICCE) 263–264 (IEEE, 2016).

10. Caruso, M. J. Applications of Magnetoresistive Sensors in Navigation Systems. in SAE Technical Paper Series vol. 1 1–8 (1997).

11. Baibich, M. N., Broto, J. M., Fert, A., Van Dau, F. N., Petroff, F., Etienne, P., Creuzet, G., Friederich, A. & Chazelas, J. Giant Magnetoresistance of (001)Fe/(001)Cr Magnetic Superlattices. Phys. Rev. Lett. 61, 2472–2475 (1988).

12. Julliere, M. Tunneling between ferromagnetic films. Phys. Lett. A 54, 225–226 (1975).

13. Joshi, V. K. Spintronics: A contemporary review of emerging electronics devices. Eng. Sci. Technol. an Int. J. 19, 1503–1513 (2016).

14. Coey, J. M. D. Magnetism and Magnetic Materials. vol. 3 (Cambridge University Press, 2010).

15. Stohr, J. & Siegmann, H. C. Magnetism. Springer Ser. Solid State Sci. vol. 152 (Springer Berlin Heidelberg, 2006).

16. Weiss, P. L’hypothèse du champ moléculaire et la propriété ferromagnétique. J. Phys. Théorique Appliquée 6, 661–690 (1907).

17. Stoner, E. C. LXXX. Atomic moments in ferromagnetic metals and alloys with non-

Bibliography

102

ferromagnetic elements. London, Edinburgh, Dublin Philos. Mag. J. Sci. 15, 1018–1034 (1933).

18. Spaldin, N. A. Magnetic Materials. (Cambridge University Press, 2010).

19. Heisenberg, T. Collective electron ferronmagnetism. Proc. R. Soc. London. Ser. A. Math. Phys. Sci. 165, 372–414 (1938).

20. Slater, J. C. The Ferromagnetism of Nickel. Phys. Rev. 49, 537–545 (1936).

21. Slater, J. C. The Theory of Ferromagnetism: Lowest Energy Levels. Phys. Rev. 52, 198–214 (1937).

22. Slater, J. C. Ferromagnetism and the Band Theory. Rev. Mod. Phys. 25, 199–210 (1953).

23. Van Vleck, J. H. Recent developments in the theory of antiferromagnetism. J. Phys. le Radium 12, 262–274 (1951).

24. Tomeno, I., Fuke, H. N., Iwasaki, H., Sahashi, M. & Tsunoda, Y. Magnetic neutron scattering study of ordered Mn3Ir. J. Appl. Phys. 86, 3853–3856 (1999).

25. Cullity, B. D. & Graham, C. D. Introduction to Magnetic Materials. (John Wiley & Sons, Inc., 2008).

26. Meiklejohn, W. H. & Bean, C. P. New Magnetic Anisotropy. Phys. Rev. 102, 1413–1414 (1956).

27. Nogués, J. & Schuller, I. K. Exchange bias. J. Magn. Magn. Mater. 192, 203–232 (1999).

28. Binek, C., Hochstrat, A. & Kleemann, W. Exchange bias in a generalized Meiklejohn–Bean approach. J. Magn. Magn. Mater. 234, 353–358 (2001).

29. Nogués, J., Sort, J., Langlais, V., Skumryev, V., Suriñach, S., Muñoz, J. S. & Baró, M. D. Exchange bias in nanostructures. Phys. Rep. 422, 65–117 (2005).

30. Berkowitz, A. E. E. & Takano, K. Exchange anisotropy — a review. J. Magn. Magn. Mater. 200, 552–570 (1999).

31. Tedrow, P. M. & Meservey, R. Spin Polarization of Electrons Tunneling from Films of Fe, Co, Ni, and Gd. Phys. Rev. B 7, 318–326 (1973).

32. Bowen, M., Cros, V., Petroff, F., Fert, A., Martínez Boubeta, C., Costa-Krämer, J. L., Anguita, J. V., Cebollada, A., Briones, F., De Teresa, J. M., et al. Large magnetoresistance in Fe/MgO/FeCo(001) epitaxial tunnel junctions on GaAs(001). Appl. Phys. Lett. 79, 1655–1657 (2001).

33. Ikeda, S., Miura, K., Yamamoto, H., Mizunuma, K., Gan, H. D., Endo, M., Kanai, S., Hayakawa, J., Matsukura, F. & Ohno, H. A perpendicular-anisotropy CoFeB–MgO magnetic tunnel junction. Nat Mater 9, 721–724 (2010).

34. Zheng, C., Zhu, K., Cardoso de Freitas, S., Chang, J.-Y., Davies, J. E., Eames, P., Freitas, P. P., Kazakova, O., Kim, C., Leung, C.-W., et al. Magnetoresistive Sensor Development Roadmap (Non-Recording Applications). IEEE Trans. Magn. 55, 1–30 (2019).

35. Wilson, A. H. The Electrical Conductivity of the Transition Metals. Proc. R. Soc. London. Ser. A. Math. Phys. Sci. 167, 580–593 (1938).

36. Chappert, C., Fert, A. & Van Dau, F. N. The emergence of spin electronics in data

Bibliography

103

storage. Nat. Mater. 6, 813–823 (2007).

37. Fujiwara, H. Spectroscopic Ellipsometry. (John Wiley & Sons, Ltd, 2007).

38. Azzam, R. M., Azzam, R. M. A.-G., Bashara, N. M. & Bashara, N. M. Ellipsometry and polarized light.

39. Yu, P. Y. & Cardona, M. Fundamentals of Semiconductors. Climate Change 2013 - The Physical Science Basis (Springer Berlin Heidelberg, 2010).

40. What is ellipsometry. https://www.jawoollam.com/resources/ellipsometry-tutorial/what-is-ellipsometry.

41. Kerr, J. XLIII. On rotation of the plane of polarization by reflection from the pole of a magnet. London, Edinburgh, Dublin Philos. Mag. J. Sci. 3, 321–343 (1877).

42. Magneto-Optics. vol. 128 (Springer Berlin Heidelberg, 2000).

43. Huhne, T. & Ebert, H. Fully relativistic description of the magneto-optical properties of arbitrary layered systems. Phys. Rev. B 60, 12982–12989 (1999).

44. Bruno, P., Suzuki, Y. & Chappert, C. Magneto-optical Kerr effect in a paramagnetic overlayer on a ferromagnetic substrate: A spin-polarized quantum size effect. Phys. Rev. B 53, 9214–9220 (1996).

45. Kubo, R. Statistical’Mechanical Theory of Irreversible Processes. I. General Theory and Simple Applications to Magnetic and Conduction Problems. J. Phys. Soc. Japan 12, 570–586 (1957).

46. Fronk, M., Bräuer, B., Kortus, J., Schmidt, O. G., Zahn, D. R. T. & Salvan, G. Determination of the Voigt constant of phthalocyanines by magneto-optical Kerr-effect spectroscopy. Phys. Rev. B 79, 235305 (2009).

47. Ebert, H. Magneto-optical effects in transition metal systems. Reports Prog. Phys. 59, 1665–1735 (1996).

48. Björck, M. & Andersson, G. GenX : an extensible X-ray reflectivity refinement program utilizing differential evolution. J. Appl. Crystallogr. 40, 1174–1178 (2007).

49. Design, Q. Magnetic Property Measurement System, EverCool Dewar Option User’s Manual. System 93 (2002).

50. Černý, R. & Přikryl, P. Chapter 3 Modeling Laser-Induced Phase-Change Processes: Theory and Computation. Semiconductors and Semimetals vol. 75 43–78 (2003).

51. Berthold, I., Müller, M., Klötzer, S., Ebert, R., Thomas, S., Matthes, P., Albrecht, M. & Exner, H. Investigation of selective realignment of the preferred magnetic direction in spin-valve layer stacks using laser radiation. Appl. Surf. Sci. 302, 159–162 (2014).

52. Többens, D. M. & Zander, S. KMC-2: an X-ray beamline with dedicated diffraction and XAS endstations at BESSY II. J. large-scale Res. Facil. JLSRF 2, A49 (2016).

53. Waseda, Y., Matsubara, E. & Shinoda, K. X-Ray Diffraction Crystallography. Springer, Berlin, Heidelberg vol. 46 (Springer Berlin Heidelberg, 2011).

54. ETA-Diffractometer. https://www.helmholtz-berlin.de/forschung/oe/em/mikro/ methoden/eta_en.html.

55. Seo, Y. & Jhe, W. Atomic force microscopy and spectroscopy. Reports Prog. Phys.

Bibliography

104

71, 016101 (2008).

56. Technologies, A. Agilent AFM 5500 Scanning Probe Microscope User’s Guide. Agil. Technol. (2008).

57. Oakberg, T. Magneto-optic Kerr effect. HINDS Instruments 1–6 (2005).

58. Fronk, M. Charakterisierung von dia-, para- undferromagnetischen dünnen Schichten mittelsmagnetooptischer Kerr-Effekt-Spektroskopie. 96 (2010).

59. Buchner, M., Höfler, K., Henne, B., Ney, V. & Ney, A. Tutorial: Basic principles, limits of detection, and pitfalls of highly sensitive SQUID magnetometry for nanomagnetism and spintronics. J. Appl. Phys. 124, 161101 (2018).

60. Quantum Design Application Note 1500-011. Using SQUID VSM Superconducting Magnets at Low Fields. 1–8 (2010).

61. Schulze, C., Gemming, S. & Albrecht, M. Magnetization Reversal in Film-Nanostructure Architectures. (Chemnitz, Techn. Univ., Diss., 2014, 2014).

62. Design, Q. Magnetic Property Measurement System SQUID VSM EverCool Option User’s Manual. Quantum Des. Appl. Note 1–42 (2007).

63. Fang, Y., Zha, C. L., Bonetti, S. & Åkerman, J. FORC studies of exchange biased NiFe in L10 (111) FePt-based spin valve. J. Phys. Conf. Ser. 200, 072002 (2010).

64. Pike, C. R., Roberts, A. P. & Verosub, K. L. Characterizing interactions in fine magnetic particle systems using first order reversal curves. J. Appl. Phys. 85, 6660–6667 (1999).

65. Dumas, R. K., Li, C.-P., Roshchin, I. V., Schuller, I. K. & Liu, K. Magnetic fingerprints of sub-100 nm Fe dots. Phys. Rev. B 75, 134405 (2007).

66. Rahman, M. T., Dumas, R. K., Eibagi, N., Shams, N. N., Wu, Y.-C., Liu, K. & Lai, C.-H. Controlling magnetization reversal in Co/Pt nanostructures with perpendicular anisotropy. Appl. Phys. Lett. 94, 042507 (2009).

67. Davies, J. E., Hellwig, O., Fullerton, E. E., Denbeaux, G., Kortright, J. B. & Liu, K. Magnetization reversal of Co/Pt multilayers: Microscopic origin of high-field magnetic irreversibility. Phys. Rev. B 70, 224434 (2004).

68. Davies, J. E., Wu, J., Leighton, C. & Liu, K. Magnetization reversal and nanoscopic magnetic-phase separation in La1−xSrxCoO3. Phys. Rev. B 72, 134419 (2005).

69. Pike, C. R., Ross, C. A., Scalettar, R. T. & Zimanyi, G. First-order reversal curve diagram analysis of a perpendicular nickel nanopillar array. Phys. Rev. B 71, 134407 (2005).

70. Katzgraber, H. G., Pázmándi, F., Pike, C. R., Liu, K., Scalettar, R. T., Verosub, K. L. & Zimányi, G. T. Reversal-Field Memory in the Hysteresis of Spin Glasses. Phys. Rev. Lett. 89, 257202 (2002).

71. Dumas, R. K., Li, C. P., Roshchin, I. V., Schuller, I. K. & Liu, K. Deconvoluting reversal modes in exchange-biased nanodots. Phys. Rev. B 86, 144410 (2012).

72. Dumas, R. K., Greene, P. K., Gilbert, D. A., Ye, L., Zha, C., Åkerman, J. & Liu, K. Accessing different spin-disordered states using first-order reversal curves. Phys. Rev. B 90, 104410 (2014).

73. Gräfe, J., Schmidt, M., Audehm, P., Schütz, G. & Goering, E. Application of magneto-optical Kerr effect to first-order reversal curve measurements. Rev. Sci.

Bibliography

105

Instrum. 85, 023901 (2014).

74. Harrison, R. J. & Feinberg, J. M. FORCinel: An improved algorithm for calculating first-order reversal curve distributions using locally weighted regression smoothing. Geochemistry, Geophys. Geosystems 9, 1–11 (2008).

75. Zhao, X., Roberts, A. P., Heslop, D., Paterson, G. A., Li, Y. & Li, J. Magnetic domain state diagnosis using hysteresis reversal curves. J. Geophys. Res. Solid Earth 122, 4767–4789 (2017).

76. Dhavale, R. P. P., Waifalkar, P. P. P., Sharma, A., Dhavale, R. P. P., Sahoo, S. C. S. C., Kollu, P., Chougale, A. D. D., Zahn, D. R. T. R. T., Salvan, G., Patil, P. S. S. P. B. B., et al. Monolayer grafting of aminosilane on magnetic nanoparticles: An efficient approach for targeted drug delivery system. J. Colloid Interface Sci. 529, 415–425 (2018).

77. Mohn, P. The calculated electronic and magnetic properties of the tetragonal transition-metal semi-borides. J. Phys. C Solid State Phys. 21, 2841–2851 (1988).

78. Bourourou, Y., Beldi, L., Bentria, B., Gueddouh, A. & Bouhafs, B. Structure and magnetic properties of the 3d transition-metal mono-borides TM–B (TM=Mn, Fe, Co) under pressures. J. Magn. Magn. Mater. 365, 23–30 (2014).

79. Kettler, W., Wernhardt, R. & Rosenberg, M. Electrical resistivity and thermopower of amorphous Fe x Co 80−x B 20 alloys. J. Appl. Phys. 53, 8248–8250 (1982).

80. Djayaprawira, D. D., Tsunekawa, K., Nagai, M., Maehara, H., Yamagata, S., Watanabe, N., Yuasa, S., Suzuki, Y. & Ando, K. 230% room-temperature magnetoresistance in CoFeB∕MgO∕CoFeB magnetic tunnel junctions. Appl. Phys. Lett. 86, 092502 (2005).

81. Hauser, H. & Grössinger, R. Isotropic hysteresis modeling of Fe–Co–B alloys. J. Appl. Phys. 85, 5133–5135 (1999).

82. Huang, S. X., Chen, T. Y. & Chien, C. L. Spin polarization of amorphous CoFeB determined by point-contact Andreev reflection. Appl. Phys. Lett. 92, 242509 (2008).

83. Bilzer, C., Devolder, T., Kim, J.-V., Counil, G., Chappert, C., Cardoso, S. & Freitas, P. P. Study of the dynamic magnetic properties of soft CoFeB films. J. Appl. Phys. 100, 053903 (2006).

84. Devolder, T., Kim, J.-V., Nistor, L., Sousa, R., Rodmacq, B. & Diény, B. Exchange stiffness in ultrathin perpendicularly magnetized CoFeB layers determined using the spectroscopy of electrically excited spin waves. J. Appl. Phys. 120, 183902 (2016).

85. Ikeda, S., Hayakawa, J., Lee, Y. M., Sasaki, R., Meguro, T., Matsukura, F. & Ohno, H. Dependence of Tunnel Magnetoresistance in MgO Based Magnetic Tunnel Junctions on Ar Pressure during MgO Sputtering. Jpn. J. Appl. Phys. 44, L1442–L1445 (2005).

86. Ikeda, S., Hayakawa, J., Ashizawa, Y., Lee, Y. M., Miura, K., Hasegawa, H., Tsunoda, M., Matsukura, F. & Ohno, H. Tunnel magnetoresistance of 604% at 300K by suppression of Ta diffusion in CoFeB∕MgO∕CoFeB pseudo-spin-valves annealed at high temperature. Appl. Phys. Lett. 93, 082508 (2008).

87. Sun, B., Li, G. Q., Zhao, W. X., Shen, Z., Liu, Y. H. & Chen, P. Perpendicular coercive force of thick CoFeB thin films grown on silicon substrate. Mater. Lett. 123, 221–223 (2014).

Bibliography

106

88. Takeuchi, T., Tsunekawa, K., Choi, Y., Nagamine, Y., Djayaprawira, D. D., Genseki, A., Hoshi, Y. & Kitamoto, Y. Crystallization of Amorphous CoFeB Ferromagnetic Layers in CoFeB/MgO/CoFeB Magnetic Tunnel Junctions. Jpn. J. Appl. Phys. 46, 623–626 (2007).

89. Cardoso, S., Cavaco, C., Ferreira, R., Pereira, L., Rickart, M., Freitas, P. P., Franco, N., Gouveia, J. & Barradas, N. P. Characterization of CoFeB electrodes for tunnel junctions. J. Appl. Phys. 97, 95–98 (2005).

90. Lavrijsen, R., Paluskar, P. V., Loermans, C. T. J., van Kruisbergen, P. A., Kohlhepp, J. T., Swagten, H. J. M., Koopmans, B. & Snoeck, E. Magnetism in Co80-xFexB20: Effect of crystallization. J. Appl. Phys. 109, 093905 (2011).

91. Irene, E. A. Applications of spectroscopic ellipsometry to microelectronics. Thin Solid Films 233, 96–111 (1993).

92. Wakagi, M., Fujiwara, H. & Collins, R. W. Real time spectroscopic ellipsometry for characterization of the crystallization of amorphous silicon by thermal annealing. Thin Solid Films 313, 464–468 (1998).

93. Carvalho, J. L., Somers, M. E. & Dalnoki-Veress, K. Ellipsometry as a probe of crystallization in binary blends of a sphere-forming diblock copolymer. J. Polym. Sci. Part B Polym. Phys. 49, 712–716 (2011).

94. Engmann, S., Turkovic, V., Hoppe, H. & Gobsch, G. Revealing the Active Layer Morphology within Complete Solar Cell Devices via Spectroscopic Ellipsometry. J. Phys. Chem. C 117, 25205–25210 (2013).

95. Kim, K. J., Lee, S. J. & Lynch, D. W. Study of optical properties and electronic structure of ferromagnetic FeCo. Solid State Commun. 114, 457–460 (2000).

96. Kumar, M., Nautiyal, T. & Auluck, S. Optical and magneto-optical properties of Fe4−xCox (x = 1–3). Eur. Phys. J. B 73, 423–432 (2010).

97. Kim, K. J., Lee, S. J. & Park, J. M. Electronic structure of DO3-ordered Fe3Co and Co3Fe studied by spectroscopic ellipsometry. J. Magn. Magn. Mater. 241, 6–10 (2002).

98. Liang, X., Xu, X., Zheng, R., Lum, Z. A. & Qiu, J. Optical constant of CoFeB thin film measured with the interference enhancement method. Appl. Opt. 54, 1557 (2015).

99. Kalashnikova, A. M., Pavlov, V. V., Pisarev, R. V., Kalinin, Y. E., Sitnikov, A. V. & Rasing, T. Optical and magnetooptical properties of CoFeB/SiO2 and CoFeZr/Al2O3 granular magnetic nanostructures. Phys. Solid State 46, 2163–2170 (2004).

100. Herzinger, C. M., Johs, B., McGahan, W. A., Woollam, J. A. & Paulson, W. Ellipsometric determination of optical constants for silicon and thermally grown silicon dioxide via a multi-sample, multi-wavelength, multi-angle investigation. J. Appl. Phys. 83, 3323–3336 (1998).

101. Lynch, D. W. & Hunter, W. R. Comments on the Optical Constants of Metals and an Introduction to the Data for Several Metals. in Handbook of Optical Constants of Solids (ed. Palik, E. D.) 275–367 (Elsevier, 1997).

102. Rumaiz, A. K., Woicik, J. C., Wang, W. G., Jordan-Sweet, J., Jaffari, G. H., Ni, C., Xiao, J. Q. & Chien, C. L. Effects of annealing on the local structure of Fe and Co in CoFeB/MgO/CoFeB tunnel junctions: An extended x-ray-absorption fine structure

Bibliography

107

study. Appl. Phys. Lett. 96, 112502 (2010).

103. Fuji, Y., Kaji, S., Hara, M., Higashi, Y., Hori, A., Okamoto, K., Nagata, T., Baba, S., Yuzawa, A., Otsu, K., et al. Highly sensitive spintronic strain-gauge sensor based on a MgO magnetic tunnel junction with an amorphous CoFeB sensing layer. Appl. Phys. Lett. 112, 062405 (2018).

104. Hellwig, O., Margulies, D. T., Lengsfield, B., Fullerton, E. E. & Kortright, J. B. Role of boron on grain sizes and magnetic correlation lengths in recording media as determined by soft x-ray scattering. Appl. Phys. Lett. 80, 1234–1236 (2002).

105. Li, D., Xu, Y. N. & Ching, W. Y. Electronic structures, total energies, and optical properties of -rhombohedral B12 and -tetragonal B50 crystals. Phys. Rev. B 45, 5895–5905 (1992).

106. Ventrua, G. & Risegari, L. Isolation. in The Art of Cryogenics 89–102 (Elsevier, 2008).

107. Fuchs, K. The conductivity of thin metallic films according to the electron theory of metals. Math. Proc. Cambridge Philos. Soc. 34, 100–108 (1938).

108. Sharma, A., Hoffmann, M. A., Matthes, P., Busse, S., Selyshchev, O., Mack, P., Exner, H., Horn, A., Schulz, S. E., Zahn, D. R. T., et al. Exchange bias and diffusion processes in laser annealed CoFeB/IrMn thin films. J. Magn. Magn. Mater. 489, 165390 (2019).

109. Weller, D., Reim, W., Ebert, H., Johnson, D. D. & Pinski, F. J. Correlation between bandstructure and magneto-optical properties of BCC Fex Co1-x. Le J. Phys. Colloq. 49, C8-41-C8-42 (1988).

110. Maurer, T., Sticht, J., Oppeneer, P. M., Herman, F. & Kübler, J. Theory of the magneto-optical Kerr effect of Fe, Ni and some Fe-based alloys. J. Magn. Magn. Mater. 104–107, 1029–1030 (1992).

111. Krinchik, G. S. & Gushchin, V. S. Investigation of inter-band transitions in ferromagnetic metals and alloys by the magnetooptic technique. Sov. Phys. J.E.T.P. 24, 984 (1969).

112. Osgood, R., Riggs, K., Johnson, A. E., Mattson, J., Sowers, C. & Bader, S. Magneto-optic constants of hcp and fcc Co films. Phys. Rev. B - Condens. Matter Mater. Phys. 56, 2627–2634 (1997).

113. Lee, P. H., Xiao, Z. R., Chen, K. L., Chen, Y., Kao, S. W. & Chin, T. S. The magnetism of alloys: First principle calculations. Phys. B Condens. Matter 404, 1989–1992 (2009).

114. Liu, A. Y. & Singh, D. J. General-potential study of the electronic and magnetic structure of FeCo. Phys. Rev. B 46, 11145–11148 (1992).

115. Gupta, R., Gupta, A., Gupta, M., Rajput, P. & Wildes, A. Evolution of structural and magnetic properties of amorphous CoFeB film with thermal annealing. J. Appl. Phys. 114, 63903 (2013).

116. Derkachova, A., Kolwas, K. & Demchenko, I. Dielectric Function for Gold in Plasmonics Applications: Size Dependence of Plasmon Resonance Frequencies and Damping Rates for Nanospheres. Plasmonics 11, 941–951 (2016).

117. Hunderi, O. Optical properties of disordered and amorphous metal films. Thin Solid Films 37, 275–280 (1976).

Bibliography

108

118. Kim, K. J., Leung, T. C., Harmon, B. N. & Lynch, D. W. Calculation of optical properties and self-energy shifts for ferromagnetic Ni, Co and Fe. J. Phys. Condens. Matter 6, 5069 (1994).

119. Johnson, P. B. & Christy, R. W. Optical Constants of the Noble Metals. Phys. Rev. B 6, 4370–4379 (1972).

120. Debnath, N. C., Chowdhuri, M. R. & Chatterjee, S. Electronic energy bands and optical properties of Cr and Fe by the KKR method. J. Phys. F Met. Phys. 15, 1693 (1985).

121. Cebollada, A., Weller, D., Sticht, J., Harp, G. R., Farrow, R. F. C., Marks, R. F., Savoy, R. & Scott, J. C. Enhanced magneto-optical Kerr effect in spontaneously ordered FePt alloys: Quantitative agreement between theory and experiment. Phys. Rev. B 50, 3419–3422 (1994).

122. Albisetti, E., Petti, D., Pancaldi, M., Madami, M., Tacchi, S., Curtis, J., King, W. P., Papp, A., Csaba, G., Porod, W., et al. Nanopatterning reconfigurable magnetic landscapes via thermally assisted scanning probe lithography. Nat. Nanotechnol. 11, 545–551 (2016).

123. Berthold, I., Löschner, U., Schille, J., Ebert, R. & Exner, H. Exchange bias realignment using a laser-based direct-write technique. Phys. Procedia 56, 1136–1142 (2014).

124. Jungwirth, T., Marti, X., Wadley, P. & Wunderlich, J. Antiferromagnetic spintronics. Nat. Nanotechnol. 11, 231–241 (2016).

125. Lombard, L., Gapihan, E., Sousa, R. C., Dahmane, Y., Ducruet, C., Papusoi, C., Prejbeanu, I. L., Nozières, J. P., Dieny, B. & Schuhl, A. IrMn and FeMn blocking temperature dependence on heating pulse width. J. Appl. Phys. 107, 09D728 (2010).

126. Tsunoda, M., Imakita, K., Naka, M. & Takahashi, M. L12 phase formation and giant exchange anisotropy in Mn3Ir/Co–Fe bilayers. J. Magn. Magn. Mater. 304, 55–59 (2006).

127. Dahmani, F. Effect of Seed Layer on Room Temperature Tunnel Magnetoresistance of MgO Barriers Formed by Radical Oxidation in IrMn-Based Magnetic Tunnel Junctions. Jpn. J. Appl. Phys. 51, 043002 (2012).

128. Singh, U., Echtenkamp, W., Street, M., Binek, C. & Adenwalla, S. Local Writing of Exchange Biased Domains in a Heterostructure of Co/Pd Pinned by Magnetoelectric Chromia. Adv. Funct. Mater. 26, 7470–7478 (2016).

129. Cardoso, S., Freitas, P. P., de Jesus, C., Wei, P. & Soares, J. C. Spin-tunnel-junction thermal stability and interface interdiffusion above 300°C. Appl. Phys. Lett. 76, 610–612 (2000).

130. Wang, Y., Zeng, Z. M., Han, X. F., Zhang, X. G., Sun, X. C. & Zhang, Z. Temperature-dependent Mn-diffusion modes in CoFeB- and CoFe-based magnetic tunnel junctions: Electron-microscopy studies. Phys. Rev. B 75, 214424 (2007).

131. Pike, C. R. First-order reversal-curve diagrams and reversible magnetization. Phys. Rev. B 68, 104424 (2003).

132. Roberts, A. P., Pike, C. R. & Verosub, K. L. First-order reversal curve diagrams: A new tool for characterizing the magnetic properties of natural samples. J. Geophys. Res. Solid Earth 105, 28461–28475 (2000).

Bibliography

109

133. Muxworthy, A. R. & Roberts, A. P. First‐Order Reversal Curve (FORC) Diagrams. in Encyclopedia of Geomagnetism and Paleomagnetism 266–272 (Springer Netherlands, 2007).

134. Khanal, S., Diaconu, A., Vargas, J. M., Lenormand, D. R., Garcia, C., Ross, C. A. & Spinu, L. Exchange bias in (FeNi/IrMn) n multilayer films evaluated by static and dynamic techniques. J. Phys. D. Appl. Phys. 47, 255002 (2014).

135. Gilbert, D. A., Ye, L., Varea, A., Agramunt-Puig, S., del Valle, N., Navau, C., López-Barbera, J. F., Buchanan, K. S., Hoffmann, A., Sánchez, A., et al. A new reversal mode in exchange coupled antiferromagnetic/ferromagnetic disks: distorted viscous vortex. Nanoscale 7, 9878–9885 (2015).

136. Soldatov, I. V. & Schäfer, R. Advances in quantitative Kerr microscopy. Phys. Rev. B 95, 014426 (2017).

137. Lee, Y. M., Hayakawa, J., Ikeda, S., Matsukura, F. & Ohno, H. Giant tunnel magnetoresistance and high annealing stability in CoFeB∕MgO∕CoFeB magnetic tunnel junctions with synthetic pinned layer. Appl. Phys. Lett. 89, 042506 (2006).

138. Feng, G., van Dijken, S., Feng, J. F., Coey, J. M. D., Leo, T. & Smith, D. J. Annealing of CoFeB/MgO based single and double barrier magnetic tunnel junctions: Tunnel magnetoresistance, bias dependence, and output voltage. J. Appl. Phys. 105, 033916 (2009).

139. Schuhmann, H., Seibt, M., Drewello, V., Thomas, A., Zbarsky, V., Walter, M. & Münzenberg, M. Controlling boron redistribution in CoFeB/MgO magnetic tunnel junctions during annealing by variation of cap layer materials and MgO deposition methods. arXiv 1405, 1–11 (2014).

140. Mukherjee, S., Knut, R., Mohseni, S. M., Anh Nguyen, T. N., Chung, S., Tuan Le, Q., Åkerman, J., Persson, J., Sahoo, A., Hazarika, A., et al. Role of boron diffusion in CoFeB/MgO magnetic tunnel junctions. Phys. Rev. B 91, 085311 (2015).

141. Lu, Y., Tran, M., Jaffrès, H., Seneor, P., Deranlot, C., Petroff, F., George, J.-M., Lépine, B., Ababou, S. & Jézéquel, G. Spin-Polarized Inelastic Tunneling through Insulating Barriers. Phys. Rev. Lett. 102, 176801 (2009).

142. Stewart, D. A. New type of magnetic tunnel junction based on spin filtering through a reduced symmetry oxide: FeCo/Mg3B2O6/FeCo. Nano Lett. 10, 263–267 (2010).

143. Komagaki, K., Hattori, M., Noma, K., Kanai, H., Kobayashi, K., Uehara, Y., Tsunoda, M. & Takahashi, M. Influence of diffused boron into MgO barrier on pinhole creation in CoFeB/MgO/CoFeB magnetic tunnel junctions. IEEE Trans. Magn. 45, 3453–3456 (2009).

144. Bai, Z., Shen, L., Wu, Q., Zeng, M., Wang, J.-S., Han, G. & Feng, Y. P. Boron diffusion induced symmetry reduction and scattering in CoFeB/MgO/CoFeB magnetic tunnel junctions. Phys. Rev. B 87, 014114 (2013).

145. Kurt, H., Rode, K., Oguz, K., Boese, M., Faulkner, C. C. & Coey, J. M. D. Boron diffusion in magnetic tunnel junctions with MgO(001) barriers and CoFeB electrodes. Appl. Phys. Lett. 96, 262501 (2010).

146. Kozina, X., Ouardi, S., Balke, B., Stryganyuk, G., Fecher, G. H., Felser, C., Ikeda, S., Ohno, H. & Ikenaga, E. A nondestructive analysis of the B diffusion in Ta–CoFeB–MgO–CoFeB–Ta magnetic tunnel junctions by hard x-ray photoemission. Appl. Phys. Lett. 96, 072105 (2010).

Bibliography

110

147. Zhang, Y., Fan, X., Wang, W., Kou, X., Cao, R., Chen, X., Ni, C., Pan, L. & Xiao, J. Q. Study and tailoring spin dynamic properties of CoFeB during rapid thermal annealing. Appl. Phys. Lett. 98, 42506 (2011).

148. Ong, C. W., Huang, H., Zheng, B., Kwok, R. W. M., Hui, Y. Y. & Lau, W. M. X-ray photoemission spectroscopy of nonmetallic materials: Electronic structures of boron and BxOy. J. Appl. Phys. 95, 3527–3534 (2004).

List of figures

111

List of figures

Figure 1. The milestones in the evolution of magnetoresistive devices. ...........................2

Figure 2. The classification of magnetic materials [Image adapted from14]. .....................7

Figure 3. The Stoner model of ferromagnetic metals illustrated for the 3d shell, and

nomenclature used for the band description of magnetism (a.). Occupied electron states

below the Fermi energy EF are shaded, unoccupied states above EF are shown unshaded.

Hext, m and Δ denote the external magnetic field, magnetic moment and exchange

splitting, respectively. The calculated density of states of Fe, Co, Ni, and Cu (b.) [Images

are taken from15]. .............................................................................................................9

Figure 4. The magnetic lattice structure of IrMn superimposed over the crystal lattice (a.).

Magnetic sublattice showing inbound (A-yellow) and outbound magnetisation (B-green)

with zero net magnetisation (b.) [images adapted from24]. ............................................. 11

Figure 5. Exemplary M(H) hysteresis loop of the ferromagnetic material Co60Fe20B20 after

30 min annealing at 600°C: Mr is the remanent magnetisation at H=0; Ms denotes the

saturation magnetisation and HC the coercivity. ............................................................. 12

Figure 6. M(H) hysteresis loop of CoFeB / IrMn bilayer at room temperature (RT). The

green line represents the M(H) response when the sample is cooled from 200°C to RT in

the presence of a magnetic field (120 mT), and the red line is the response of the as-

deposited, demagnetised state. ..................................................................................... 14

Figure 7. Uniaxial magnetic anisotropy induced through the exchange bias effect in

CoFeB / IrMn after cooling from 200°C in the presence of 120 mT magnetic field. The

coercive fields (a.) and the exchange bias fields (b.) are plotted as determined by angle-

resolved longitudinal-MOKE magnetometry at room temperature. ................................. 15

Figure 8. The development road map of the magnetoresistive devices [image was taken

from34]. ........................................................................................................................... 16

Figure 9. Schematic representation of the tunnel magnetoresistance in the case of two

identical ferromagnetic layers separated by a non-magnetic insulating barrier such as

MgO. The tunnelling process conserves the spin. When the electronic states on each side

of the barrier are spin-polarised, the electrons will more easily find free states to tunnel

through the barrier if the magnetisations are parallel (a.) than if they are antiparallel (b.)

to each other due to the ratio of the density of states of both electrons (spin-up, spin-

down) at the Fermi level. The arrows in red and grey show the higher and lower tunnelling

probability of spin-polarised electron through a tunnel barrier, respectively. The yellow

balls represent the electrons with their intrinsic spin orientation and direction of rotation in

grey and blue arrows.[image redrawn from36]. ............................................................... 17

List of figures

112

Figure 10. Various polarisation states of light occurring as a result of the various phase

difference between the two components of the electric field (along x- and y- axes) of equal

amplitude (𝐸𝑦 = 𝐸𝑥) [image taken from37]. ..................................................................... 19

Figure 11. The schematic of spectroscopic ellipsometry [image redrawn from40 ............ 23

Figure 12. The schematic of an optical model for an ambient / thin film / substrate

structure, showing the reflected and refracted light at each interface. Using the Fresnel

coefficients, the contribution of reflections from each interface can be calculated [image

was taken from40]. .......................................................................................................... 24

Figure 13. The three geometrical configurations for MOKE, namely polar (a.), longitudinal

(b.) and transversal (c.). ................................................................................................. 26

Figure 14. The schematic energy state diagram of a 3d-ferromagnet, showing optical

transitions induced by left (blue) and right (red) polarised photons for a system where only

spin-orbit coupling is present (left diagram) and for a system where spin-orbit coupling

and exchange interaction is present (middle diagram). The notation in the | ⟩ brackets

contains the orbital number (𝑙), magnetic number (𝑚), and spin orientation (↑ or ↓). The

right side diagram shows the corresponding absorption spectra of left and right circular

polarised light. [adapted from44] ..................................................................................... 28

Figure 15. Schematic representation of the magnetron sputtering. The argon ions (in red)

are responsible for the target etching. The ejected particles (in grey) are sputtered towards

the substrate. The direction of the magnetic field used for confining the electrons and the

ions close to the target is illustrated as blue arrows. ...................................................... 32

Figure 16. Oven sample mounting platform with grading to place the sample on oven

heater stick [image was taken from48]. ........................................................................... 33

Figure 17. CAD modelled isometric view (a.) and transverse view (b.) of the heater

assembly developed for the macro-MOKE system......................................................... 34

Figure 18. Schematics of the experimental setup for Nd: YAG laser [image prepared for

the own publication49]. .................................................................................................... 35

Figure 19. Drawing of the diffractometer showing the general scheme of various

goniometers and measurement axes [image taken from52]. ........................................... 36

Figure 20. A basic principle of atomic force microscopy [image was taken from54]. ....... 38

Figure 21. The spectroscopic ellipsometry setup M2000 for measuring the complex ratio

(𝜌) of the Fresnel reflection coefficients. ........................................................................ 39

Figure 22. Schematic diagram of the magneto-optical Kerr effect spectrometer used in

this thesis for measuring Kerr rotation (θK) and ellipticity (ηK). ....................................... 40

List of figures

113

Figure 23. A cross-sectional view of the utilised MPMS SQUID - VSM MPMS-3 setup

[image taken from60]. ...................................................................................................... 41

Figure 24. Top view of the Nano-MOKETM2 magnetometer. The yellow arrows point out

the optical components in the optical path of the laser shown with the red dashed line. 42

Figure 25. An exemplary first order reversal curve (FORC) recorded for Si /

SiO2(100 nm) / Ni81Fe19(20 nm). The reversal magnetic field (Ha) and regular magnetic

field (Hb) are shown in red and black solid dots, respectively (a.). The major hysteresis

loop (MHL) is shown in the red dashed line. (b.) The FORC distribution calculated from

the measured FORC loops. The contour denotes the maxima of the distribution at the

“irreversible” located at about HC = 2.2 mT, HI = 0 mT. .................................................. 44

Figure 26. Process flow diagram of XPS-depth profiling. ............................................... 45

Figure 27. X-ray diffraction patterns recorded for Si / SiO2(1.8 nm) /

CoFeB(100 nm) / Pt(5 nm) before and after annealing under UHV at the indicated

temperature for Co40Fe40B20 (a.) and Co60Fe20B20(b.). Additionally, the scan of the as-

deposited Si / SiO2(1.8 nm) / CoFe(100 nm) /Pt(5 nm) sample is presented in black in

figure 27(a.).The respective reflexes of constituent materials are marked by dotted lines

along with the miller indices [powder diffraction file of CoFe (00-049-1567), Pt (00-004-

0802), Si (00-027-1402), and Ag (00-004-0783) from the International Centre for

Diffraction Data (ICDD)]. ................................................................................................ 49

Figure 28. Rocking curve measured at the CoFe(110) reflex for the Co40Fe40B20 (a.) and

Co60Fe20B20 (b.) films. The inset in (b.) shows the FWHM of particular Gaussian fits of the

obtained peaks. .............................................................................................................. 50

Figure 29. The CoFe crystallite sizes (vertical coherence lengths) calculated using the

Scherrer equation for Co40Fe40B20 and Co60Fe20B20 (a) determined from XRD shown in

Figure 27. SEM micrograph collage of Si / SiO2(1.8 nm) / Co40Fe40B20(100 nm) / Pt(5 nm)

before (above) and after annealing (below) recorded in the FIB trench at 36° stage tilt (b.).

....................................................................................................................................... 51

Figure 30. The grazing incidence diffractogram recorded for CoFeB thick films before and

after annealing at various temperature for two stoichiometries: Co40Fe40B20 (a.) and

Co60Fe20B20 (b.). The respective reflexes of the constituent materials are marked by dotted

lines along with the Miller indices. .................................................................................. 52

Figure 31. The evolution of and (inset) spectra recorded for the Co40Fe40B20 (a.) and

Co60Fe20B20 (b.) thick films before and after annealing at various temperatures. For

comparison, the and spectrum of Co50Fe50 is plotted along with the Co40Fe40B20

spectra in (a.). ................................................................................................................ 53

List of figures

114

Figure 32. The complex dielectric function (ε1xx & ε2xx) spectra of the Co50Fe50 (red),

Co40Fe40B20 (blue) and B (grey) 103, together with the simulated ε1xx & ε2xx of (Co50Fe50)+B

with 15 % B inclusion(yellow). More detailed information about the B inclusion is given in

the text. .......................................................................................................................... 54

Figure 33. The annealing temperature dependent evolution of ε1xx (a.) and ε2xx (b.) spectra

for Co40Fe40B20 (solid line) and Co60Fe20B20 (dashed line), and CoFe (black). ............... 55

Figure 34. Drude parameters resistivity (ρ) in black and scattering time (τs) in red colour

as a function of annealing temperature for Co40Fe40B20 (solid symbol) and Co60Fe20B20

(empty symbol). The lines in the figure are drawn to guide the eye (a.). Sheet resistance

of the Co40Fe40B20 (filled circles in red) and Co60Fe20B20 (unfilled circles in blue) layers

passivated with a Pt thin film as a function of the annealing temperature (b.). ............... 57

Figure 35. Measured polar Kerr effect rotation (θK) and ellipticity (ηK) spectra of Co50Fe50

in comparison to the previously published reference107,108 (a.) and calculated (h)2×xy as

the function of photon energy (b.). ................................................................................. 58

Figure 36. The polar Kerr effect measured rotation (θK) and ellipticity (ηK) spectra for the

100 nm thick CoFeB films before and after annealing at the indicated temperatures for

two stoichiometry Co40Fe40B20 (a. & c.) and Co60Fe20B20 (b. & d.). ................................ 59

Figure 37. The calculated xy(h)2 as a function of the photon energy for two

stoichiometries, Co40Fe40B20 and Co60Fe20B20, for Si / SiO2(1.8 nm) / CoXFe(80-

X)B20(100 nm) / Pt(5 nm) layer stacks annealed at 450°C and 600°C. ........................... 60

Figure 38. X-ray diffractogram recorded for Si / SiO2(100 nm) / Co60Fe20B20(20 nm) /

Au(3 nm), in the as-deposited state (squares) and after annealing at 350°C (circles) and

400°C (triangles). ........................................................................................................... 61

Figure 39. Experimental SE ψ and ∆ spectra recorded at 65° AOI for Si / SiO2(100 nm) /

Co60Fe20B20(20 nm) / Au(3 nm) before and after vacuum oven annealing at 350°C and

400°C. ............................................................................................................................ 62

Figure 40.The ε1xx (black) and ε2xx (red) components of the complex of the dielectric

function of Co60Fe20B20 (10 nm to 20 nm) before (continuous line) and after annealing at

350°C (dashed line) and 400°C (dotted line)(a.). Comparison of the complex dielectric

function of as-deposited 20 nm and 100 nm thick Co60Fe20B20 films (b.). ...................... 63

Figure 41. Comparison of Kerr rotation θK (black) and ellipticity ηK (red) for Si /

SiO2(100 nm) / Co60Fe20B20(20 nm) / Au(3 nm) before (continuous line) and after

annealing at 350°C (dashed line) and 400°C (dotted line). ............................................ 65

List of figures

115

Figure 42. A schematic diagram showing a typical layer stack of the MTJ sensor and the

laser-induced temperature profile for setting the exchange bias in the FM/AFM layers. The

CoFeB (FM) layer next to IrMn (AFM) is the pinned layer and the upper CoFeB layer is

the free layer, both sandwich the MgO tunnel barrier. The Ni81Fe19 and CuN serve as the

seed and buffer layer, respectively. The Ta layers work as the passivation (on the top)

and adhesion layer (at the bottom). ................................................................................ 69

Figure 43. Magnetisation as a function of the in-plane applied external magnetic field on

an as-grown sample (a.) and measured along the easy-axis of the bottom pinned layer

after 120 mT field cooling from 227°C (b.). The arrow in green and red represent the

magnetisation of the free layer and pinned layer, respectively. ...................................... 70

Figure 44. SQUID-VSM measured M(H) hysteresis loops recorded at RT for an as-

deposited sample and after consecutive field cooling process from different temperatures

as indicated in the legend. .............................................................................................. 71

Figure 45. Temperature dependence of the coercivity HC (a.)and the exchange bias field

HEB for the determination of the TB. The sample was initially cooled in 120 mT field in-

plane applied magnetic from 327°C to RT before the measurements at the increasing

temperature. ................................................................................................................... 72

Figure 46. Optical micrograph of an exemplary sample annealed with CW irradiation with

different scanning speed, 50 mm·s-1 (a.), 500 mm·s-1 (b.), 2000 mm·s-1 (c.), and

4000 mm·s-1 (d.). In each square, the different laser power was used as indicated, the

patterns, as well as the labels, were created by laser ablation. ...................................... 73

Figure 47. The coercive (a.) and exchange bias field (b.) of the PL as a function of the

applied laser intensity with scanning speed of 50 mm·s-1, 500 mm·s-1, 2000 mm·s-1, and

4000 mm·s-1 for CW laser annealing. For comparison, the ranges of the maximum values

of HC and HEB in the case of oven annealing are shown as hashed areas, including the

error bars. ....................................................................................................................... 74

Figure 48. Coercive field (black symbols, left axis) and exchange bias field strength (red

symbols, right axis) as a function of the laser intensity for the samples annealed using PW

laser. For comparison, the ranges of the maximum values of HC and HEB in the case of

oven annealing (cf.Figure 47) are shown as hashed areas with included error bars. ..... 75

Figure 49. The exchange bias field strength as a function of the CW laser intensity at

different scanning speeds applied on a (0.5 × 0.5) mm2 sample areas for Ru (a.) and

Ni81Fe19 (b.) seed layer. The horizontal dashed bar indicates the maximum exchange bias

field strength (including the error bars of the exchange bias value) achieved for vacuum

oven annealing at 280°C. ............................................................................................... 76

List of figures

116

Figure 50. The difference in the exchange bias field strength determined from the

measurements shown in Figure 49. In addition, results of oven annealing at different

temperatures are shown in the inset. The samples with a Ru seed layer present a larger

exchange bias, both for the oven and laser annealing, with more pronounced differences

arising in the latter. ......................................................................................................... 77

Figure 51. XRD θ–2θ scans of the samples with (a.) Ru and (b.) NiFe seed layer as-

deposited and annealed in an oven at 320°C and 400°C. The IrMn(111) orientation (2θ =

41.36°) is represented by the dashed lines, highlighting a small peak shift in (a.) due to

the presence of the (002)-textured Ru seed layer (2θ = 42.15°). The given crystallite size

was determined from the pronounced Kiessig fringes. ................................................... 78

Figure 52. Change in the laser intensity I = I(0.5 × 0.5) mm2 - I(6 × 6) mm

2 needed to induce the

same exchange bias field strength in (0.5 × 0.5) mm2 square patterns (Figure 55) and

(6 × 6) mm2 square patterns as a function of the laser scanning speed. ........................ 79

Figure 53. The exchange bias field strength as the function of the number of CW laser

annealing repetitions at 5000 mm·s-1 scan speed and 240 kW·cm-2 intensity. ............... 80

Figure 54. MH hysteresis loops of the sample with layer stack Si / SiO2(100 nm) / Ta(5 nm)

/ Ru(5 nm) / IrMn(8 nm) / Co40Fe40B20(2.3 nm) / MgO(1.8 nm) / Co40Fe40B20(2.3 nm) /

Ta(5 nm) as-deposited (black), after one annealing cycle at 250°C (red), followed by nine

cycles of annealing at 300°C (showing only iterations #2 and #9, since no significant

changes were found) ...................................................................................................... 80

Figure 55. Magnetisation loop (a.) and contour plot presenting the FORC distribution (b.)

of the sample with Si / SiO2(1.8 nm) / Co60Fe20B20(100 nm) / Pt(5 nm) in the as-deposited

state. Both measurements were performed by SQUID-VSM. ........................................ 82

Figure 56. FORC distribution of a Si / SiO2(100 nm) / Ta(3 nm) / Ru(3 nm) / IrMn(8 nm) /

CoFeB(5 nm) / Ta(3 nm) layer stack in the as-deposited state (a.) and after setting the

exchange bias by cooling from 200°C in the presence of 120 mT magnetic field(b.). .... 83

Figure 57. MH loop (a.) and FORC distribution (b.) recorded at RT for a Si /

SiO2(100 nm) / Ta(5 nm) / CuN(30 nm) / Ta(3 nm) / Ni81Fe19(2 nm) / IrMn(8 nm) /

Co40Fe40B20(2 nm) / MgO(1.8 nm) / Co40Fe40B20(2.2 nm) / Ta(10 nm)” MTJ layer stack

after setting the exchange bias. ..................................................................................... 84

Figure 58. MOKE microscopy image of the text “HLPH” (“Halbleiterphysik”) inscribed with

a CW laser beam on the Si /SiO2(100 nm) / Ta(3 nm) / Ru(3 nm) / IrMn(8 nm) /

CoFeB(5 nm) / Ta(3 nm) layer stack. The grey contrasts in the image have the pure

magneto-optical origin and correspond to different magnetisation orientations in the FM

layer. The hysteresis loop measured on the laser exposed (red) and unexposed region

(green); the corresponding area-of-interest is marked in the MOKE image. .................. 86

List of figures

117

Figure 59. Exchange bias (red) and coercive field (black) strength determined by MOKE

magnetometry as a function of laser intensity and laser scanning speed for

Si / SiO2(100 nm) / Ta(3 nm) / Ru(3 nm) / IrMn(8 nm) / CoFeB(5 nm) / Ta(3 nm) sample

(a.). The hashed areas in red and black represent the values of exchange bias and

coercive field (with the error bars) obtained for a sample annealed in vacuum at 200°C.

The coercivity (HC) and exchange bias field (HEB) temperature dependence for the oven

annealed sample is shown in (b.). .................................................................................. 89

Figure 60. XPS depth profile of an as-deposited layer Si / SiO2(100 nm) / Ta(3 nm) /

Ru(3 nm) / IrMn(8 nm) / CoFeB(5 nm) / Ta(3 nm) sample. The thickness values provided

in the layer stack and the upper part of the graph are the nominal values used for the

deposition process. ........................................................................................................ 90

Figure 61. Distribution profiles of Co, Mn, Ru, and O for the layer stack Si / SiO2(100 nm) /

Ta(3 nm) / Ru(3 nm) / IrMn(8 nm) / CoFeB(5 nm) / Ta(3 nm) laser annealed at different

laser scanning speeds as indicated in the legend for (a.) 120 kW·cm-2, (b.) 380 kW·cm-2,

and (c) 900 kW·cm-2 laser intensities. ............................................................................. 91

Figure 62. Normalised B distribution profile, for all laser annealed samples of the Si /

SiO2(100 nm) / Ta(3 nm) / Ru(3 nm) / IrMn(8 nm) / CoFeB(5 nm) / Ta(3 nm) layer stack,

indicating a strong migration of B towards the Ta capping layer and the Ru / Ta seed layer.

....................................................................................................................................... 92

Figure 63. XPS B spectra recorded after a sputtering time of the 20s for 900 kW·cm-2 laser

annealed samples with Si / SiO2(100 nm) / Ta(3 nm) / Ru(3 nm) / IrMn(8 nm) /

Co40Fe40B20(5 nm) / Ta(3 nm) layer stack (a.). A clear boron oxide peak was detected at

192.4 eV (dashed line) for the lowest scanning speed only. The B-O peak vanishes after

a sputtering time of 100s (b.). ......................................................................................... 93

Figure 64. Depth profile of the atomic concentration of Co, Mn, and Ru calculated from

XPS measurements of Si / Si(100 nm) / Ta(3 nm) / Ru(3 nm) / IrMn(8 nm) / CoFeB(5 nm)

/ Ta(3 nm) after vacuum field annealing at 200°C (a.). The B diffusion profile for as-

deposited samples and samples annealed at 200°C, and 500°C (b.). ........................... 94

Figure 65. θ-2θ scans for an as-deposited (yellow) and laser annealed with 380 kW·cm-2

(red), and 900 kW·cm-2 (blue) Si / SiO2(100 nm) / Ta(3 nm) / Ru(3 nm) / IrMn(8 nm) /

CoFeB(5 nm) / Ta(3 nm) layer stack for scanning speeds of 50 mm·s-1, 500 mm·s-1, and

5000 mm·s-1. Expected positions for oriented crystalline structures are marked by vertical

dashed lines. .................................................................................................................. 95

Figure 66. Average surface roughness (σavg) of the Si / SiO2(100 nm) / Ta(3 nm) /

Ru(3 nm) / IrMn(8 nm) / CoFeB(5 nm) / Ta(3 nm) layer stack irradiated with 380 kW·cm-2

(red) and 900 kW·cm-2 (yellow) laser intensity for the investigated scanning speeds. The

average surface roughness for the 200°C oven annealed samples are shown as dashed

blue line. Insets are the 3D topography images of the selected samples (in green: laser

annealed samples; in grey: oven annealed samples)..................................................... 96

List of tables

118

List of tables

Table 1. The selection rules of (magneto-) optical transitions ........................................22

Table 2. Nominal and XRR determined thicknesses of the CoXFe(80-X)B20, Co50Fe50, and

Pt layer for as the as-deposited samples .......................................................................48

Table 3. Nominal and XRR determined thicknesses of the Co60Fe20B20 and Au layer for

as the as-deposited samples ..........................................................................................61

Table 4. The optimal laser intensity parameter for the various laser scanning speed. ...74

Abbreviations

119

Abbreviations

AFM Antiferromagnet

FL Free Layer

FM Ferromagnetic

GMR Giant Magnetoresistance

HV High Vacuum (0.1 𝑚𝑃𝑎)

IR Infrared

LASER laser

Light Amplification by Stimulated Emission of Radiation

MTJ Magnetic Tunnel Junction

MO Magneto-optical

MOKE Magneto-optical Kerr Effect

MR Magnetoresistance

MTJ Magnetic tunnel junction

NM Non-magnetic material

PL Pinned Layer

RT Room Temperature

SE Spectroscopic Ellipsometry

SEM Scanning Electron Microscope

SQUID-VSM Superconducting Quantum Interference Device – Vibrating Sample Magnetometry

TEM Transmission Electron Microscope

TMR Tunnel Magnetoresistance

UHV Ultra-High Vacuum (0.1 𝑛𝑃𝑎)

UV Ultraviolet

XPS X-ray Photon Spectroscopy

XRD X-ray Diffraction

XRR X-ray Refractometry

Erklärung

120

Erklärung

Ich erkläre, dass ich die vorliegende Arbeit selbständig und nur unter Verwendung der

angegebenen Literatur und Hilfsmittel angefertigt habe.

Januar, 2020

Apoorva Sharma

curriculum vitae

121

curriculum vitae

Apoorva Sharma

Date of birth 02 May 1987

Place of birth Bikaner, India

Nationality Indian

Gender Male

Marital status Married

Foreign language English, German

Academic

2005-2009

Bachelor’s Electronics and Communication Engineering, University of Rajasthan, Jaipur India.

“Microcontroller based automation system”.

First-class

honours

2010-2012

CSIR-Central Electronics Engineering Research Institute, Pilani-333031, India.

“Magnetic Nanocomposite materials and their application in V-Groove inductor”.

2013-2016

Master of Science at Faculty of Electrical Engineering and Information Technology, University of Technology Chemnitz, Chemnitz Germany.

“Magnetic Characterisation of [Co/Ni] multilayers for magnetoresistive applications”.

Sehr gut

Scientific contributions

122


List of publications

A. Sharma, M.A. Hoffmann, P. Matthes, S. Busse, O. Selyshchev, P. Mack, H. Exner, A. Horn, S.E. Schulz, D.R.T. Zahn, and G. Salvan.

Exchange bias and diffusion processes in laser annealed CoFeB/IrMn thin films.

J. Magn. Magn. Mater. 489, 165390 (2019).

A. Sharma, M.A. Hoffmann, P. Matthes, N. Kohler, S. Busse, M. Muller, H. Exner, S.E. Schulz, D.R.T. Zahn, and G. Salvan.

Magnetic tunnel junctions: Laser annealing versus oven annealing.

IEEE Trans. Magn. 55, 1 (2019).

A. Sharma, M.A. Hoffmann, P. Matthes, O. Hellwig, C. Kowol, D. R.T Zahn, S. E. Schulz, and G. Salvan.

Crystallisation of optically thick films of CoxFe(80-x)B20: evolution of the (magneto-) optical and structural properties

Phys. Rev. B 101, 054438 (2020)

A. Sharma, P. Matthes, I. Soldatov, S. S. P. K. Arekapudi, B Böhm, M. Lindner, O. Selyshchev, N.T.N. Ha, M. Mehring, C. Tegenkamp, S. E Schulz, Dietrich R T Zahn, Y. Paltiel, O. Hellwig and G. Salvan.

Control of magneto-optical properties of cobalt-layers by adsorption of a-helical polyalanine self-assembled monolayers.

J. Mater. Chem. C, 2050-7526 (2020).

M.A. Hoffmann, A. Sharma, P. Matthes, S. Okano, O. Hellwig, R. Ecke, D.R.T. Zahn, G. Salvan, and S.E. Schulz.

Spectroscopic ellipsometry and magneto-optical Kerr effect spectroscopy study of thermally treated Co60Fe20B20 thin films.

J. Phys.: Condens. Matter (2019) (in press).


123

S. Nikam, A. Sharma, M. Rahman, A.Teli, S. Mujawar, D.R.T. Zahn, G. Salvan, P. S. Patil, S.C. Sahoo, and P.B. Patil.

Pulsed Laser Deposited CoFe2O4 Supercapacitor Electrode

RSC Adv., 10, 19353-19359, (2020).

N.T.N. Ha, A. Sharma, D. Slawig, S. Yochelis, Y. Paltiel, D.R.T. Zahn, G. Salvan, and C. Tegenkamp.

Charge-ordered -helical polypeptide monolayers on Au(111)

J. Phys. Chem. C 124, 10, 5734–5739, (2020).

S. Okano, A. Sharma, A. Nishimura, C. Günther, O. D. Gordan, K. Ikushima, G. Salvan, V. Dzhagan and D.R.T. Zahn.

Voltage controlled dielectric function of bilayer graphene.

Adv. Optical Mater., 8, 2000861, (2020).

C. Saengruengrita, A. Sharma, D. Solonenkob, P. Thamyongkita, S. Wacharasindhua, S. Sattayapornc, G. Salvanb, D. R. T. Zahn, and N. Insin.

Iron oxide nanospheres and nanocubes modified with carboxyphenyl porphyrin and their magnetic, optical properties and photocatalytic activities in room temperature amide synthesis.

J. Magn. Magn. Mater. (in press)

C. Saengruengrit, P. Ritprajak, S. Wanichwecharungruang, A. Sharma, G. Salvan, D.R.T. Zahn, and N. Insin.

The combined magnetic field and iron oxide-PLGA composite particles: Effective protein antigen delivery and immune stimulation in dendritic cells J. Colloid Interface Sci. 520, 101 (2018).

J.D. John, S. Okano, A. Sharma, O. Selyshchev, M. Rahaman, N. Miyachi, K. Enomoto, J. Ochiai, I. Saito, T. Masuzawa, T. Yamada, D.H.C. Chua, D.R.T. Zahn, and K. Okano

Transport properties of Se/As2Se3 nanolayer superlattice fabricated using rotational evaporation

Adv. Funct. Mater. 29, 1904758 (2019).


124

R.P. Dhavale, P.P. Waifalkar, A. Sharma, R.P. Dhavale, S.C. Sahoo,

P. Kollu, A.D. Chougale, D.R.T. Zahn, G. Salvan, P.S. Patil, and

P.B. Patil.

Monolayer grafting of aminosilane on magnetic nanoparticles: An efficient approach for targeted drug delivery system

J. Colloid Interface Sci. 529, 415 (2018).

J. D. John, S. Okano, A. Sharma, M. Rahaman, O. Selyshchev, N. Miyachi, K. Enomoto, J. Ochiai, I. Saito, G. Salvan, T. Masuzawa, T. Yamada, D. H. C. Chua, D. R. T. Zahn, and K. Okano.

Observation of two-level defect system in amorphous Se superlattices.

Appl. Phys. Lett., 116, 19, (2020).

M. Rahaman, O. Selyshchev, Y. Pan, I. Milekhin, A. Sharma, G. Salvan, S. Gemming, and D. R. T. Zahn.

Radiative decay of dark exciton related emission in a sandwiched monolayer WSe2 revealed by room temperature micro and nano photoluminescence.

Submitted.

R. Patra, H. Stöcker, A. Sharma, M. Monecke, G. Salvan, R. Mattheis,

S. Pofahl, R. Schäfer, O.G. Schmidt, and H. Schmidt.

Magneto-optical response of multilayer structures with ferromagnetic NiFe, CoFe, or CoFeB thin films.

Submitted.

V.C. Karade, A. Sharma, R.P. Dhavale, R.P. Dhavale, P.S. Patil, J. H. Kim, D.R.T. Zahn, A.D. Chougale, G. Salvan, and P.B. Patil.

APTES monolayer grafting on magnetic nanospheres for controlled release of anticancer drug Nintedanib.

Submitted.


125

List of conferences and workshops

Magnetic Properties of Porphyrin Magnetite Nanocomposites

March 2017, DPG-Frühjahrstagung, Dresden, Germany (Poster).

Magneto-optical Spectroscopy and Spectroscopic Ellipsometry of Co60Fe20B20

April 2017, IEEE Intermag 2017, Dublin, Ireland (Poster).

Imaging Ellipsometer Nanofilm-EP4

June 2017, Accurion summer school 2017, Göttingen, Germany.

Condensed Matter Magnetism : Bulk Meets Nano

October 2017, European School on Magnetism, Halbleiterphysik, Cargèse, France.

Spintronics (won the third prize)

November 2017, Science Slam, Zentrum für den wissenschaftlichen Nachwuchs-TU

Chemnitz, Chemnitz, Germany. (Talk).

Magneto-optical Spectroscopy and Spectroscopic Ellipsometry of Co60Fe20B20

Thin Films

March 2018, 10th Workshop Ellipsometry (WSE 10), Chemnitz, Germany (Talk).

Novel Method of Setting Exchange Bias in Tunnel Magnetoresistance Devices

with Laser Annealing

March 2018, DPG-Frühjahrstagung, Berlin, Germany (Poster).

Spectroscopic Ellipsometry of Diffusion in Magnetic Multilayer Stacks

September 2018, Joint European Magnetic Symposia (JEMS), Mainz, Germany

(Poster).

Magnetite Nanoparticles and Their Biomedical Applications (won first prize)

April 2019, Nanobio19, online poster presentation, (Poster).

126

Acknowledgements

This research work is the result of teamwork, and nothing would have been possible

without the collaboration and the patience of many people. In this context, I thank all the

project partners involved in this work for a very fruitful scientific collaboration.

I would like to thank Prof. Dr. Georgeta Salvan and Prof. Dr. Dr. Dietrich R.T. Zahn for

giving me this opportunity to take up the challenges of this project. I further extend my

gratitude to Prof. Dr. Georgeta Salvan for being a constant source of motivation, for her

patience and for her deep human understanding. She is simply the best forewoman to

work under.

All my thanks to my supervisors Dr. Patrick Matthes and Ms Maria Almeida, for their

continuous support. In particular, I want to thank Dr. Patrick Matthes for the never-ending

discussions in which I always discovered that physics is just simple solutions for difficult

problems. I am really grateful to Ing. Axel Fechner (“The engineer that can fix anything

except broken hearts”) for his technical support and coffee corner discussions. My special

thanks belong to Mrs Sybille Raschke and Mrs Jane Eisentraut, who always took care of

all my administrative work and really pushed me to speak German with them. I want to

thank my fellow group members for the support for providing a very enjoyable scientific

ambience. During my PhD, I met some wonderful people at various conferences who

either directly or indirectly contributed to this thesis. To this, a special thanks to Dr. Ivan

Soldatov and Dr. Prashant Patil for efficient and productive collaborations. I also thank Mr

Fabian Ganss for maintaining the SQUID-VSM facility at its best throughout this work. I

am grateful to Dr. Volodymyr Dzhagan and Dr. Paul Mack for the XPS depth profile

measurements and the valuable discussion for data analysis.

Finally, no thanks will ever be enough for my family, my mother Mrs Rashmi Sharma, my

father Mr Mahesh Kumar Sharma,and my sister Mrs Pallvi Sharma because their trust

and pride in me never rest. My friends Himani, Shun, Ilya, Sasha (Oleksandr) and Josh,

who never fails to remind me that there is more to life than physics alone.

correlation between the structural, optical, and magnetic

Documents