See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/30017194

Magnetic and structural properties of the Gd/Ni-Bilayer-System

Thesis · September 2015

Source: OAI

CITATION READS

1 2,143

1 author:

Alexander Barth
Oerlikon Metco AG
11 PUBLICATIONS   111 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Alexander Barth on 09 September 2015.

The user has requested enhancement of the downloaded file.

 Magnetic and structural
properties of the
Gd/Ni-Bilayer-System

DISSERTATION

zur Erlangung des akademischen Grades

eines Doktors der Naturwissenschaften (Dr. rer. nat.)
an der Universität Konstanz
Fachbereich Physik

vorgelegt von
Alexander Barth
Tag der mündlichen Prüfung: 26. Juli 2007
Referenten:

Prof. Dr. Günter Schatz

Prof. PhD. William E. Evenson

Konstanzer Online-Publikations-System (KOPS)

URL: http://www.ub.uni-konstanz.de/kops/volltexte/2007/3637/
URN: http://nbn-resolving.de/urn:nbn:de:bsz:352-opus-36373
Contents

Contents iv

1 Motivation and Introduction 1

2 Magnetism and magnetic materials: Fundamental principles 3

2.1 The magnetic field . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Matter in external magnetic fields . . . . . . . . . . . . . . . . . . 4
2.2.1 Dia- and paramagnetism . . . . . . . . . . . . . . . . . . . 7
2.2.2 Ferromagnetism . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2.3 Ferri- and Antiferromagnetism . . . . . . . . . . . . . . . . 15
2.3 Finite size effects in thin magnetic films . . . . . . . . . . . . . . . 16
2.3.1 Interlayer exchange coupling . . . . . . . . . . . . . . . . . 17
2.4 Film growth and epitaxy . . . . . . . . . . . . . . . . . . . . . . . 18

3 Experimental methods and set-ups 21

3.1 The UHV set-up MEDUSA . . . . . . . . . . . . . . . . . . . . . 21
3.2 Magnetic characterization . . . . . . . . . . . . . . . . . . . . . . 24
3.2.1 Superconducting QUantum Interference Device:
SQUID . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.2.2 X-ray Magnetic Circular Dichroism: XMCD . . . . . . . . 29
3.3 Structural investigations . . . . . . . . . . . . . . . . . . . . . . . 36
3.3.1 Scanning Tunneling Microscopy (STM) . . . . . . . . . . . 36
3.3.2 X-ray diffraction and reflectometry . . . . . . . . . . . . . 39
3.3.3 Electron Diffraction: MEED . . . . . . . . . . . . . . . . . 40
3.3.4 Chemical composition by AES . . . . . . . . . . . . . . . . 41
3.3.5 Rutherford backscattering . . . . . . . . . . . . . . . . . . 42
3.3.6 Transmission electron microscopy . . . . . . . . . . . . . . 44
Contents iv

4 Experimental Results 47
4.1 Sample description and preparation . . . . . . . . . . . . . . . . . 47
4.2 Thin Gadolinium layers . . . . . . . . . . . . . . . . . . . . . . . . 49
4.3 Structure and morphology of the Gd/Ni-bilayer system . . . . . . 54
4.3.1 X-ray analysis of Gd/Ni/Au thin films . . . . . . . . . . . 55
4.3.2 In-situ Auger spectroscopy on Gd/Ni bilayers . . . . . . . 58
4.3.3 Characterization of the layer structure and chemical com-
position by RBS and TEM . . . . . . . . . . . . . . . . . . 59
4.3.4 Investigating the growth of Gd on Al2 O3 and Si3 N4 by STM 62
4.4 Magnetic characterization of Gd/Ni-bilayers . . . . . . . . . . . . 67
4.4.1 In-plane anisotropy in Gd/Ni bilayers . . . . . . . . . . . . 68
4.4.2 Field dependent magnetization measurements on the exam-
ple bilayer system: Gd(50Å)/Ni(75Å)/Au(20Å) . . . . . . 69
4.4.3 Temperature dependence of the magnetic properties of the
example system Gd(50Å)/Ni(75Å)/Au(20Å) . . . . . . . . 72
4.4.4 Element-specific hysteresis loops of an example Gd/Ni bi-
layer system . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.4.5 Dependence on the nickel layer thickness . . . . . . . . . . 83
4.4.6 Influence of the substrate temperature . . . . . . . . . . . 89
4.5 Proposed model of the bilayered system . . . . . . . . . . . . . . . 95
4.6 Preliminary results on Ni/Gd/Ni-trilayers . . . . . . . . . . . . . 96

5 Outlook and conclusions 99

6 Summary 101

7 Zusammenfassung 103

List of figures 109

Bibliography 117

Danksagung 119

A Properties of the GdNi-system 121

B Depth profile by RBS 123

Chapter 1

Motivation and Introduction

Rare earth elements have been the scope of intensive research for many of their
different properties over the past decades. The high magnetic permanent mo-
ment of these elements aroused a lot of interest and research on their magnetic
behavior or structure, which was already conducted in the early 1970s. But due
to the challenging problems associated with the handling of rare earths like their
strong reactivity with water and oxygen, strong alloying with other metals or
their high content of impurities was keeping research only to a small number of
involved groups and publications per year. Another issue is the non-epitaxial
or even amorphous growth of nearly all rare earth elements if evaporated under
UHV conditions. As will be shown later in section 4.2 there are only a few sub-
strates which allow an epitaxial growth of gadolinium for example. But still the
search for materials with high magnetization or either high or low coercivity for
magnetic data storage technology or magnetic sensing elements attract notice to
this class of elements especially gadolinium.
Gadolinium can be seen as the only room temperature ferromagnet among the
rare earth materials due to the highest Curie temperature, which is 16°C. The
low value of the coercivity limits its possible applications for data storage. Its
permanent magnetic moment amounts to 7.98 µB , what exceeds for example the
value of nickel by a factor of 12. Its magnetic moment is generated completely by
the spin moment µs of the 4f-shell whose complicated RKKY interaction to its
neighbors will be described in sections 2.2 and 2.2.2.2. Besides these properties
gadolinium possesses other outstanding characteristics that are worth mentioning
like the highest cross section for capturing thermal neutrons of 49000 barn or the
high magnetocaloric effect. Its physical properties are a density of 7.895 cmg 3 , a
melting point at 1313°C and an atomic weight of 157.25 u. There are 17 isotopes
known but only 7 are abundant in natural gadolinium. The hcp structure with
the lattice parameters a=3.63 Å and c=5.78 Å is stable up 1235°C (αGd) above
it transforms to its bcc phase (βGd). It was discovered in 1880 by Jean Charles
Galissard de Marignac and its main applications are as a contrast agent in med-
ical NMR, material for nuclear reactor construction and in microwave devices.
The other magnetic material used in this work, nickel, becomes ferromagnetic
Chapter 1. Motivation and Introduction 2

below TC =358°C. Its magnetic moment is 0.606 µB (at T=0 K). The physi-
cal properties are: Density 8.908 cmg 3 , atomic weight 58.69 u and melting point
Tm =1455°C and its crystal structure is fcc.
A large number of the publications on gadolinium and its magnetic behavior in
the recent decade was on the topic of exchange interactions, either as a pure
metal or in alloys with different transition metals, especially iron, cobalt CoNi
and Permalloy. Systems containing pure nickel layers are hardly discussed in lit-
erature. Possible reasons are the difficulties to determine the interface structure
[13], the tendency of alloying [14, 15] and the comparable low Curie temperature
of the formed GdNi alloys [16].
Interlayer exchange coupling attracted a lot of interest since the discovery of the
giant magnetoresistance effect GMR in 1986 [17, 18]. In this field of application
gadolinium can play an important role due to its high magnetic moment, low
coercivity and its intrinsically low magnetic anisotropy. The coupling of Gd with
ferromagnetic materials is one of the most interesting topics that was investigated
in the recent years [19, 20, 21] as well as the coupling of magnetic layers through
a gadolinium spacer [14, 15, 22]. The interaction is still only partially understood
because of the complex interaction of the 3d and 4f magnetic moments [23], lead-
ing to twisted [24] or even helic spin configurations as in the case of Dy and Ho
[25, 26].
In this study the magnetic interplay and interactions between thin gadolinium
and nickel layers were investigated and their connection to the structural prop-
erties of the bilayers. For this, several thin films of gadolinium were studied to
investigate the growth on different substrates and two series of Gd/Ni bilayers
with varying nickel layer thickness, on two different substrates and two different
deposition temperatures. Electron diffraction (MEED, LEED), Auger electron
spectroscopy, X-ray diffraction (XRD), transmission electron microscopy (TEM),
Rutherford backscattering (RBS) and scanning tunneling microscopy (STM) were
applied to examine structural properties on an atomic and microscopic scale. The
overall magnetic properties of the whole samples were studied by a superconduct-
ing quantum interference device (SQUID) and element-specific by X-ray circular
dichroism (XMCD). It will be shown that morphology and magnetic behavior are
closely connected in this system and that different coupling schemes can evolve
from this.
Chapter 2

Magnetism and magnetic

materials: Fundamental
principles

A short introduction will be given initially to the subject of magnetism and mag-
netic materials. Generally three different types of magnetic response of materials
are distinguished, namely dia-, para- and ferromagnetism. The basic concepts
that lead to these different types will be discussed and the further subtypes ferri-
and antiferromagnetism will be introduced to the reader. Special emphasis will be
given to the different nature of the ferromagnetic behavior of the so called band-
or 3d-transition metal magnet nickel and the 4f- or rare earth magnet gadolinium.
Theoretical predictions at the end of this chapter about the interplay of these two
materials will lead to later discussion of the observed experimental results.

2.1 The magnetic field

In the literature one can find two complementary ways of describing magnetic
phenomena. One is in terms of circulating currents evolving from Maxwell’s
equations and the other in terms of magnetic poles, which are more suitable for
describing experimental observations. Even the unit system that is used depends
on the choice of the description and development of the theory and definitions
of magnetism. Most text books about electrodynamics work with the cgs system
(centimeter-gram-second), like [8, 27]. So this is usually used for formulations
from Maxwell’s equations. The more phenomenological description in terms of
poles is in most cases carried out in the SI-unit system. Since most measurement
devices ,that were used in this work like the SQUID-magnetometer, deliver results
in cgs units, these units will be used in the following description of magnetism.
A short conversion table is given in table 1.1.
Additional information about the conversion between the two unit systems is
comprehensively summed up in [28], where it is written: ”Conversions between
Chapter 2. Magnetism and magnetic materials: Fundamental principles 4

cgs units and mks (SI) units seem to have been designed to torment both the
novice and the seasoned professional alike”.
These two formulations can be connected by the definition of the magnetic dipole
moment m ~ caused by an electrical current density ~j(~r):
Z
1
m
~ = [~r × ~j(~r)]d3 r (2.1)
2c

The magnetic induction B ~ is the response of a material if a magnetic field

H~ is applied to it. In vacuum, B
~ and H~ are equal in cgs units (in SI B
~ = µ0 H).
~
In para- and diamagnetic materials, their relation is linear. The magnetization
M~ of the medium has to be taken into account, leading to the following relation
between B ~ and H:~

~ =H
B ~ + 4π M
~ (2.2)

The magnetization M~ can have its origin in induced and/or permanent mag-
netic moments of the atoms or molecules and is usually expressed as the magnetic
moments per unit volume (see table 2.1).

property cgs SI
p
H = 4πε0 r2 Ampere
1 p
 
dipole field H = [oersted]
r2 m
magnetic induction ~ =H
B ~ + 4π M ~ [Gauss] B ~ = µ0 (H~ +M ~ ) [Tesla]
M~  emu
 M~
susceptibility χ H~ cm3 oersted H~ [dimensionless]
-23 erg
Bohr’s magneton µB e~
2me c
= 0.927*10 Oe e~
2me
= 0.927*10-20 TJ

Table 2.1: Conversion table of characteristic magnetic variables cgs↔SI

2.2 Matter in external magnetic fields

If a material is exposed to an external magnetic field its response can be expressed
by the ratio of the magnetization that results and the applied field strength at a
fixed temperature. This quantity is called the susceptibility χ:

M h emu i
χ= ; . (2.3)
H cm3 Oe
But it’s the magnetic induction B~ what is directly measured. The ratio of
~ and the magnetic field strength H
B ~ is called the permeability µ of the material.
5 2.2. Matter in external magnetic fields

According to equation equation (2.2), µ is linearly related to χ:

µ = 1 + 4πχ (2.4)

The value and sign of χ is taken to divide materials according to their mag-
netic properties into dia-, para- or ferromagnetic as discussed in the following
sections.
The basis of all magnetic phenomena of matter is the atomic magnetic mo-
ment caused by the electrons. The nucleus carries a magnetic moment, too, but
it is much weaker and gives rise only to small changes in the electronic structure
that can be observed by hyperfine interaction experiments. The origin of the
atomic magnetic moment is the orbital moment ~l and the total spin moment ~s
of the electrons. For light to medium elements the individual angular momenta
~li and spins ~si of electrons in open shells add up separately to L~ and S, ~ which
couple to the total angular momentum J~ = L ~ + S.
~ This mechanism is called
Russell-Saunders- or L-S-coupling.
In heavy elements, the so called jj-coupling appears, where ~l and ~s of a single
electron first ~
~ P couple to j, and then these add up to the total angular momen-
~
tum J = i ji of the atom. Both of the materials under investigation, nickel
and gadolinium, build up their magnetic moment according to the L-S-coupling.
Although gadolinium is a considerable heavy element this coupling mechanism is
valid due to the the fact that the open shells are in low lying 4f orbital.

The magnetic moment arising from the total angular momentum of the elec-
tron cloud is given by:

µB ~
~ = −gJ
m J. (2.5)
~
With the number gJ being the Landé factor and µB the so called Bohr
magneton (see table 1.2). The g-factor can be computed from the quantum
numbers J, S and L by

J(J + 1) + S(S + 1) + L(L + 1)

gJ = 1 + . (2.6)
2J(J + 1)
In order to obtain the values for J, S and L one has to derive the quantum
states of all electrons in shells that are not completely filled according to Hund’s
rules. But this formalism only applies correctly to free atoms. In a solid, and
especially in the magnetic 3d-transition metals, the formation of electronic band
structures from delocalized electrons and interactions between the orbitals of
neighboring atoms lead to strong deviations from the calculated values, as shown
later in the sections sections 2.2.2.1 and 2.2.2.2.
Chapter 2. Magnetism and magnetic materials: Fundamental principles 6

Figure 2.1: Spin and effective magnetic moment: The spin and magnetic
moment of the rare earth metals from lanthanum to lutetium [29].

p
Ion Configuration g J(J + 1) [µB ] µexp [µB ]
Ce3+ 4f1 5s2 5p6 2.54 2.4
Pr3+ 4f2 5s2 5p6 3.58 3.5
Nd3+ 3 2
4f 5s 5p 6
3.62 3.5
Pm3+ 4 2
4f 5s 5p 6
2.68 -
Sm3+ 4f5 5s2 5p6 0.84 1.5
Eu3+ 6 2
4f 5s 5p 6
0 3.4
Gd3+ 7 2
4f 5s 5p 6
7.94 7.98
Tb3+ 4f8 5s2 5p6 9.72 9.77
Dy3+ 9 2
4f 5s 5p 6
10.63 10.6
Ho3+ 10 2
4f 5s 5p 6
10.6 10.4
Er3+ 4f11 5s2 5p6 9.59 9.5
Tm3+ 12 2
4f 5s 5p 6
7.57 7.3
Yb3+ 13 2
4f 5s 5p 6
4.54 4.5

Table 2.2: Comparison of calculated and measured magnetic moment from the
paramagnetic state per ion of rare earth metals [25]. Eu3+ should have no mag-
netic moment, but low lying excited states are occupied, producing a nonzero
magnetic moment.
7 2.2. Matter in external magnetic fields

The magnetic moments of rare earth materials, however, can be computed

very accurately by equation (2.5) because the partially filled 4f and 5f shells are
well shielded by the outer orbitals 5s, 5d, 5p and 6s, as shown in figure 2.6. The
good accordance to values of the 3+ ions shows that the electrons in the 5d and
6s shells are all delocalized in the conduction band.

2.2.1 Dia- and paramagnetism

Diamagnetic materials possess a negative value for the susceptibility χ, with
|χ|  1. The sign indicates that the magnetization points opposite to an external
magnetic field. Diamagnetism occurs in all materials but is usually dominated by
para- and ferromagnetic phenomena. It can be explained by loop currents that are
induced in the atoms by the external field, corresponding to spin precession. This
current, in turn, produces an induced magnetic moment, which points opposite
to the external field according to Lenz’s law. The following expression for the
diamagnetic susceptibility can be derived using quantum mechanical perturbation
theory [2, 30]:

N Ze2
2 
χdia = − r . (2.7)
6me c2

N : Number of atoms/molecules per unit volume

Z : Atomic number
2
hr i : average quadratic radii of the occupied orbitals in partially filled shells

Typical values for χdia are in the range of 10−6 emu

cm3 Oe
.

In a paramagnetic material the atoms (or molecules) possess a permanent

magnetic moment arising from their electronic configuration. These permanent
magnetic moments do not interact with each other and can be freely oriented.
This leads to a small but positive value for χpara .
Langevin developed an expression for freely orientable magnetic moments m ~ in
an external field. Due to thermal excitation, they will not align perfectly to an
applied magnetic field. This classical picture leads to the Curie law:

N m2 C
χpara = = . (2.8)
3kB T T
C is called the Curie constant.

But due to the quantum mechanical nature of the atomic magnetic moment,
the alignment with respect to the external field is quantisized. If this is taken
into account the magnetization is given by
Chapter 2. Magnetism and magnetic materials: Fundamental principles 8

Mpara = N gJµB BJ (x). (2.9)

BJ (x) is called the Brillouin function. Its argument is x=JgµB H/kB T. It

converges in the limit of J → ∞ to the Langevin function, and the result equa-
tion (2.8) is obtained. Equation (2.9) gives us an expression for the susceptibility,
incorporating the total angular moment J of the atoms:

N g 2 J(J + 1)µ2B C
χpara = = . (2.10)
3kB T T
But there are additional mechanisms that cause a paramagnetic response to
external magnetic fields. The weakly-bound electrons in the conduction band
carry a magnetic spin moment m ~ S = -2 µ~B and cause a temperature-independent
paramagnetic behavior called Pauli spin paramagnetism. The energy levels of
spins parallel and antiparallel to the external field are shifted by ∆E = 2µB B0 .
This gives more conduction electrons with m ~ S parallel than antiparallel to the
~ as shown in figure 2.2. One thereby finds a net magnetization
external field B,

3N µ2B
 
N − N↑↓
MP auli = µB = . (2.11)
V 2kB TF
N , N↑↓ are the numbers of electrons with magnetic moments parallel and
antiparallel to the external field. It should be mentioned that due to the negative
g factor of electrons the spin ~s and the associated magnetic moment m ~ s point in
opposite directions.

Figure 2.2: Density of states: Without an external field the density of states
of the free electron gas is unperturbed and the number of spin-up and spin-down
electrons are equal (left graph). An external magnetic field causes a shift of the
distribution and there are more electrons with spin antiparallel to the field.

The so called Van Vleck paramagnetism will only be briefly mentioned. It is

caused by excited states in atoms that don’t exhibit a magnetic moment in the
9 2.2. Matter in external magnetic fields

ground state. It’s also not temperature dependent and gives small positive values
of χ.

2.2.2 Ferromagnetism
The forms of magnetism presented so far have their origin in the properties of
the atoms or molecules. With ferromagnetism, the interaction between atomic
magnetic moments comes into play. Ferromagnetic materials are characterized
by high values of χ and a spontaneous magnetization.
In order to describe ferromagnetism, we require additional physics. The depen-
dence of M on H for a typical ferromagnetic material is called hysteresis loop.
Starting with a demagnetized sample, the magnetization rises with increasing
magnetic field until it reaches a maximum value called the saturation magneti-
zation MS . In this state all atomic magnetic moments are aligned. Decreasing
the field again to zero, a remaining magnetization, called the remanence MR ,
can be observed. One has to apply an opposite field HC , the coercivity, in order
to obtain a net magnetization of zero again. With increasing field strength the
magnetization rises again until all moments are aligned opposite to the saturation
before, and the process can be continued to form a loop.

In 1907 Weiss presented a phenomenological model of ferromagnetism, in-

troducing an interaction between the permanent magnetic moments of the atoms
called ”the molecular field”, which he added to the applied field:

~ tot = H
H ~ ext + H
~W = H
~ ext + γ M
~ (2.12)
γ: molecular field constant.

Introducing this expression into equation (2.10) leads to the Curie-Weiss law,
describing the behavior of magnetic materials above their Curie temperature TC ,
where they are paramagnetic.

CH
ext
M C M T −Cγ C
= ⇒ χ= = = (2.13)
Hext + γM T Hext Hext T − TC
This can be expanded to the temperature regime below TC by applying the
same replacement to the calculations of Langevin for paramagnetic materials. So
without an external field there is still the molecular field HW and the spontaneous
magnetization is given by the intersection of the following two expressions for M.

M = N mL(α) (2.14)
HW
M = (2.15)
γ
Chapter 2. Magnetism and magnetic materials: Fundamental principles 10

The argument of the Langevin function L is α=mH/kB T.

But even the normalized magnetization M/M0 over T/TC is only approxi-
mately represented by equation (2.14) (see figure 2.3).

Figure 2.3: The relative spontaneous magnetization of Fe, Co and Ni can be

well described by Weiss’s theory with (a) the classical Langevin function or the
Brillouin function for (b) J=1 and (c) J=1/2 [6].

This model takes no account of the quantum mechanical nature of the elec-
trons which cause the magnetic behavior and gives no explanation for the origin
of the postulated molecular field. This major drawback of the theory of Weiss
was solved by Heisenberg in 1928 by introducing an exchange interaction between
the spins of different atoms.
X
Ĥex = ~i S
Jij S ~j (2.16)
ij

Jij is called the exchange integral. The major result of this idea is that it
can lead to a parallel alignment of neighboring spins due to this interaction of
electrostatic origin. It is a direct result of Pauli’s exclusion principle since elec-
trons with parallel spins can’t be at the same place what reduces the electrostatic
repulsion. If only the number z of nearest neighbors are taken into account within
the molecular field theory, one obtains an expression for Jij :

3kB TC
Jij = . (2.17)
2zS(S + 1)
11 2.2. Matter in external magnetic fields

And the magnetization follows from equation (2.9) with J=1/2:

 
mγM
M = N µ tanh . (2.18)
kB T

This gives a very accurate temperature dependence for the normalized mag-
netization (see figure 2.3).

2.2.2.1 Band magnetism in 3d-transition metals

In the case of the three ”classical” ferromagnets Fe, Co and Ni the open shells
from which a permanent magnetic moment evolves are the 3d orbitals. But these
electrons are in the conduction band and therefore delocalized, which contradicts
the idea of localized interacting magnetic moments as in Weiss’s and Heisenberg’s
theories. Although the temperature dependence and phase transition are well pre-
dicted (see figure 2.3), the magnetic moment per atom and its change from the
ferro- to the paramagnetic phase couldn’t be explained.

Stoner introduced his model of band magnetism in 1938 in analogy to Pauli

spin paramagnetism of the conduction band electrons. In Fe, Co and Ni the
Fermi energy is within the 3d band, making it possible to shift electrons to ex-
cited states. The energy needed is taken from minimizing electrostatic repulsion
as mentioned in the above paragraph. This is represented by the exchange inter-
action in equation (2.16) plus an exchange term that shifts the energy levels for
spin parallel or antiparallel. This shift is not equal for both spin states, and by
favoring one spin orientation, it causes a spontaneous magnetization:

E↑ (~k) = E(~k) − µB B + JN↓ (2.19)

E↓ (~k) = E(~k) − µB B + JN↑ .

This splitting is shown for nickel in figure 2.4 [30]. Here the exchange energy
leads to a complete filling of the majority band with 5 electrons and 0.54 holes in
the minority band (the missing 0.46 holes are in the 4s-band and do not contribute
to the magnetic moment).
Chapter 2. Magnetism and magnetic materials: Fundamental principles 12

Figure 2.4: Energy bands of nickel: Above TC there is an equal number of

spin-up and spin-down electrons (left side). At T=0K there is an excess of 0.54
electrons in one of the 3d-subbands [30].

The susceptibility follows from equation (2.3) using D(EF ), the density of
states at the Fermi level:

µ2B D(EF )
χStoner = . (2.20)
JD(EF )V
1−
| {z 2 }
≥0 Stoner-criterion

The Stoner criterion is a good measure of wether a material will be ferro-

magnetic. The computed values are shown in figure 2.5

Figure 2.5: Values of the Stoner criterion: Elements with D(EF )*J ≥ 1 are
ferromagnatic [2].
13 2.2. Matter in external magnetic fields

2.2.2.2 Rare Earth Magnetism

The open shells of the lanthanides and actinides are 4f- and 5f orbitals. The
focus here will be on the row of rare earth metals that includes gadolinium.
The 4f orbital is well shielded as can be seen in figure 2.6. So the exchange
interaction, which produces a spontaneous magnetization, can not be caused by
direct overlapping of orbitals. The diameter of the 4f orbital is only 0.4 Å (peak
of distribution) but the interatomic distance is 3.634 Å in the c-plane. So the
interaction must be mediated indirectly.

Figure 2.6: The open 4f shell is well shielded from the influence of neighboring
atoms in Gadolinium. For comparison the size of the Wigner-Seitz-Cell RW S is
inserted. Taken from [26].

This mediation is performed by delocalized electrons in the conduction band

and is called Rudermann-Kittel-Kasuya-Yosida coupling, abbreviated RKKY, af-
ter the scientists who contributed to its description. First, Rudermann and Kittel
developed the idea of an indirect interaction of nuclear magnetic moments via con-
duction electrons [31]. Kasuya and de Gennes expanded the model to rare earths
[32], and Yosida applied it to transition metal alloys [33].
The basic idea of this coupling is that the spin of the 4f orbital polarizes the
conduction electrons, which, in turn transports this polarization to neighboring
spins. In the following only a short summary of the RKKY formalism will be
given.
Chapter 2. Magnetism and magnetic materials: Fundamental principles 14

The interaction of two localized f-spins l and l’ can be described in a similar

manner as in equation (2.16) by:
X
Ĥf f 0 = ~l − R
j(R ~ l0 )S
~l · S
~ l0 . (2.21)
l,l0

~ l and R
The sum is over all pairs of spins at the sites R ~ l0 in the system. j
describes the indirect exchange coupling, depending on the density of states in
the conduction band at the two positions. It follows:

~ ~0
~ l0 ) = 1 ~k − ~k 0 )|2 D(k)(1 − D(k )) ei(~k−~k0 )·(R~ l −R~ l0 )
X
~l − R
j(R |j sf ( (2.22)
N 2 k,k0 Ek0 − Ek

jsf : Exchange interaction between localized f- and delocalized s-spins

~k, ~k 0 : Wave vectors of the conduction electrons at the two different sites
Ek , Ek0 : Their respective energies
D : Density of states.

In order to solve this eingenstate equation analytically, several assumptions

have to be made. The direct interactions f↔f and s↔s are neglected; jsf = j = con-
stant and the density of conduction electrons is taken to be isotropic. Then
equation (2.22) becomes:

2
j(R ~ l0 ) = 9π j F (2kF |R
~l − R ~ l − R~l0 ). (2.23)
EF
The function F, given below, shows an oscillatory behavior:

x cos x − sin x
F (x) = . (2.24)
x4
It describes well the long range RKKY interaction, as it decreases by R−3 . In
figure 2.7 the exchange interaction from equation (2.22) is plotted for gadolinium.
Positive values express ferromagnetic and negative antiferromagnetic coupling.
Gadolinium behaves like a ”normal” ferromagnet, i.e. the magnetic moments
within a domain are parallel. In most other rare earth materials the indirect
coupling forms chiral or cone-helical structures of the magnetic moments, and
parts of the magnetic moments cancel each other [34, 35].
15 2.2. Matter in external magnetic fields

Figure 2.7: RKKY ex-

change interaction of
Gd: The coupling oscil-
lates from ferro- to anti-
ferromagnetic. Taken from
[26].

2.2.3 Ferri- and Antiferromagnetism

In antiferromagnetic materials the coupling between neighboring magnetic mo-
ments leads to antiparallel alignment. Thus, there is no net magnetization. Ex-
amples are chromium and manganese, that even have a very high magnetic mo-
ment per atom. Above the ordering temperature TN , the Néel temperature, the
materials turn paramagnetic. But compounds like Fe2 O3 (hematite) can also be
antiferromagnetic (like most oxides with Fe, Co, Ni).
In contrast, ferrimagnetic materials have oppositely oriented magnetic moments
but they do not cancel each other completely. Instead a net magnetization is
left. Ferrimagnetism can occur in oxides like FeO·Fe2 O3 (magnetite) or alloys,
e.g. CoPt3 , CrPt3 and some RE-TM alloys.
The expressions for antiferro- or ferrimagnetic coupling can be extended to multi-
layered structures, as has been investigated e.g in [36, 37, 38]. Here the magnetic
coupling between neighboring layers of rare earths and transition metals was stud-
ied. The interaction of magnetic layers separated by a nonmagnetic spacer can
lead to an oscillating coupling depending on the spacer thickness [18, 39, 40].
Examples of this phenomenon are multilayers of Co/Cu or Fe/Ru and Fe/Cr.
Chapter 2. Magnetism and magnetic materials: Fundamental principles 16

2.3 Finite size effects in thin magnetic films

In thin films the dimensionality is reduced to a 2D system. On the other hand,
magnetism is an ensemble property, especially ferromagnetic coupling. Thus,
strong influence of the macroscopic structure on the magnetic behavior can be
expected.
The magnetic anisotropy is one of the properties influenced by the characteristics
of a thin layered film. There are three major contributions to magnetic anisotropy:

ˆ Magnetocrystalline anisotropy has its origin in the crystal structure

and is thereby an intrinsic property of the material. It is caused by different
distances to the next neighbors of an atom along different crystal axes. Its
contribution is comparatively small, such as Ka =4.2*105 erg/cm3 for iron.

ˆ Shape anisotropy dominates in many thin film systems. Large dipole-

dipole interactions cause it. If the easy axis lies within the film, the aligned
domains can interact throughout the whole film, and a large demagnetizing
field can be established . The Gd, Ni and Gd/Ni layered films investi-
gated here all show an in-plane magnetization due to shape anisotropy (see
figure 4.20 in section 4.4.1).

ˆ Surface and boundary anisotropy: Atoms at surfaces and boundaries

can interact only on one side with neighboring atoms of the same species.
This asymmetry also leads to a magnetically anisotropic behavior. This
effect can be very strong in multilayered systems with very thin magnetic
layers like Co/Pd with tCo ≤ 13Å, as shown in figure 2.8. Above the critical
thickness of the Co-layer the easy axis is driven by the shape anisotropy.

All these different contributions are usually summarized by a measurable

value of an effective anisotropy Kef f that consists of a volume part KV and a
surface/boundary part KS :

KS
Kef f = KV + 2 t = film thickness. (2.25)
t
Reduced dimensionality influences other properties, too, e.g. decrease of
the Curie temperature [42], saturation magnetization MS or critical temperature
TC in superconductors [41]. Systems of islands or small grains (see section 2.4)
undergo a transition to a superparamagnetic behavior [43].
17 2.3. Finite size effects in thin magnetic films

Figure 2.8: The effective anisotropy Kef f ·t changes its sign at tc , so that the
easy axis goes from perpendicular to the film surface to parallel [41].

2.3.1 Interlayer exchange coupling

The exchange coupling between magnetic layers has attracted a lot of interest
in the past 20 years. Most research has been done on systems of two, in most
cases, identical magnetic layers separated by a non-magnetic spacer. These lay-
ered structures can exhibit the so called giant magnetoresistance effect GMR,
describing the change in conductivity across the layers, depending on their mag-
netic alignment. Exchange coupling of magnetic layers was first investigated in
the 1970s and especially from the mid-1980s on in the course of magneto-optical
devices, in order to stabilize the magnetization of layers containing rare earth
metals. They exhibit a high Kerr rotation but are thermally very unstable.
In the following a very brief description of the exchange coupling between two
neighboring magnetic layers will be given, introduced by Esho [44] and Kobaya-
shi [40]. Only magnetostatic interactions are considered, what is sufficient for the
investigated system Gd/Ni. The coupling phenomena in GMR-systems involve a
mediation of the magnetizations through the conduction electrons of the spacer
layer, according to the RKKY coupling as first described by Grünberg [17] and
Parkin in 1986 [18].
The magnetization of each layer is considered to rotate coherently and the easy
axis lies in the film plane. The external field H and the field produced by the
adjoint layer act on it. The external field is also applied parallel to the film plane.
Chapter 2. Magnetism and magnetic materials: Fundamental principles 18

The total free energy per unit area can then be written as:

E = − Ms1 t1 H cos θ1 − Ms2 t2 H cos θ2 (2.26)

| {z }
Zeeman energies
1
+ σw [1 ± cos(θ1 − θ2 )]
|2 {z }
coupling energy
2
− (Ku1 − 2πMs1 )t1 cos2 θ1 − (Ku2 − 2πMs2
2
)t2 cos2 θ2
| {z }
effective anisotropy energies

t1 , t2 : thicknesses of the layers; Ku1 , Ku2 : uniaxial anisotropy constants;

θ1 , θ2 : angles between the applied field H and the magnetization of the single
layers.

The stable configurations can then be computed from:

∂E ∂2E
=0 and > 0. (2.27)
∂θi ∂θi2

From these equations follows that the stable solutions are:

θ1 , θ2 = 0, π. (2.28)

So the magnetic moments align either parallel or antiparallel to each other

in this simple model, depending on the sign of the term of the coupling energy in
equation (2.26). Helical structures, as they appear in e.g. Dy or Tb can lead to
different angles.

2.4 Film growth and epitaxy

In the previous section 2.3 the influence of the limited dimensionality on mag-
netism was discussed. But there are more implications for the properties of the
material, like altered crystal structures, alloys that do not exist in the bulk mate-
rial, or conductivity, superconductivity or optical properties changes that can be
tailored. Due to the rising demand for the precise production of thin films sev-
eral techniques have been developed. Those used in this work will be presented
in the next chapter. Epitaxy is usually performed under at least high vacuum
conditions (i.e. p < 10−6 mbar) in order to supply a continuous and homogenous
flow of atoms (sometimes molecules or small clusters) onto the substrate without
deflections or reactions with the residual gas.
In the following, the interactions between the material and the substrate will be
19 2.4. Film growth and epitaxy

discussed. Their origin lies in surface and boundary tension energies that are
minimized by growing in a distinct way. Three major modes of epitaxial growth
are distinguished (see figure 2.9) but mixed forms of them or transitions from one
to another are also known.

Figure 2.9: Epitaxial growth: Illustration of the three major modes of epitaxial
growth, beginning on the left from submonolayers to several monolayers on the
right [3].

An effective tension energy can be defined from the following boundary ten-
sions: γSV between substrate and vacuum, γAV between adsorbate and vacuum
and γSA between substrate and adsorbate. Its sign and value are good indices to
predict the growth mode:

∆γ = −γSV + γAV + γSA (2.29)

ˆ Volmer-Weber growth: A negative value of ∆γ expresses a stronger

interaction between the adsorbed atoms than with the substrate. The ad-
sorbate tends to agglomerate and form islands. That’s why this growth
mode is called island growth. Nevertheless, films growing in this manner
can form very smooth granular structures (e.g. Pt on WSe2 [45]). An ex-
ample of the build-up of very rough surfaces is Pb on glass (especially at
elevated temperatures).
Chapter 2. Magnetism and magnetic materials: Fundamental principles 20

ˆ Frank-van-der-Merve growth: Also known as layer-by-layer growth.

∆γ > 0, i.e. the material tends to accumulate at the substrate. Thus,
a first monolayer is established followed by additional layers. Sites at kinks
and holes are preferred by the adsorbed atoms, which favors the 2D-growth
pattern.

ˆ Stranski-Krastanow growth: After forming one or several smooth mono-

layers, further deposited atoms form islands. ∆γ ≈ 0 characterizes this
behavior.
Chapter 3

Experimental methods and

set-ups

3.1 The UHV set-up MEDUSA

The samples were all produced in the UHV-chamber MEDUSA, and acronym for
Maschinerie zur Erforschung DUenner Schichten and Atomagglomerationen, at
a base pressure of pbase =1*10−10 mbar. During evaporation from the compact
e-beam evaporators, models EFM3 and EFM4 from Focus GmbH, the pressure
did not rise above 1*10−9 mbar. EFM stands for Evaporator with integrated Flux
Monitor. A schematic layout of the chamber is shown in figure 3.1.
The samples were introduced into the UHV system through a load-lock that
could be evacuated down to 1*10−7 mbar within a few hours. With a substrate
heating it is possible to remove gaseous adsorbates from the samples without
degrading the vacuum in the main chamber. A manipulator system, manufac-
tured by Vacuum Generators, was adapted to the sample transfer system of the
MEDUSA. It is used to bring the samples to the evaporation position and is in-
stalled in all the measurement devices. The big advantage of this manipulator
system is that samples can be heated up to 800 K and cooled by liquid nitrogen
down to 100 K. The main chamber is equipped with a set of two EFM evap-
orators (see figure 3.2 a). They are positioned in a way to aim at the same
spot, where the sample is located. These sources were used for evaporating nickel
and gadolinium. In addition, a home-built boat evaporator was installed for the
preparation of gold capping layers. But due to poor reliability and the small
amount of material that could be stored even if all four available boats were
used only for capping material, it was replaced by a water-cooled crucible evapo-
rator system, which can be equipped with up to four different materials at a time.

The EFM evaporators are a very compact form of a e-beam evaporation

source and it fits to a CF40-flange. Electrons are emitted from a tungsten wire
that has a thorium content of 0.5 %. A high voltage is supplied between the
Chapter 3. Experimental methods and set-ups 22

Figure 3.1: Schematic layout of MEDUSA UHV-chamber (top view).

filament and the crucible, usually 800-1000 V, in order to accelerate the elec-
trons towards the crucible. The current between the filament and the crucible
is monitored to check the applied heating power. The crucible and the filament
are suited in a copper cooling shroud to minimize the rise in pressure because of
heating the surrounding. On top of the cooling shroud, where the evaporated ma-
terial comes through, a chimney is mounted. It collects a part of the electrically
charged atoms, and so a current, proportional to the evaporated flux can be mea-
sured. Due to the low ionization energies of gadolinium, these atoms are usually
3+ ionized, what gives a much higher current than for nickel. The evaporation of
both materials posed great difficulties. The filament is exposed to the evaporated
material unlike in most other e-beam evaporators. Gadolinium lowers the work
function of the filament dramatically, that’s why the heating up of the crucible
has to be monitored very carefully, and the filament current has to be reduced im-
mediately as soon as the material is beginning to be evaporated. Nickel tends to
alloy with the crucible materials molybdenum and tungsten. This can lead to the
destruction of the crucibles. The boat evaporator used was suffering the known
drawbacks like poor reliability and that only small amounts of material could be
stored. Aside form that materials like Co, Pt or Pd can not be evaporated from
this kind of source, due to the high temperatures needed. On the other hand a
stable and high flux of gold could be reached within a few minutes and the system
needed no cooling. Still it was replaced later by a home-built thermal evaporator,
consisting of boron-nitride crucibles surrounded by a tantalum filament. But this
23 3.1. The UHV set-up MEDUSA

Figure 3.2: Schematics of an EFM evaporator and source alignment in

MEDUSA: In part a) the major components of the used EFM evaporators are
shown. On the right side in b) the arrangement of all three sources with respect
to the sample is sketched.

source needs water cooling and has to be completely unmounted for every refill
of materials.
Figure 3.2 b) shows the arrangement of the evaporation sources around the sample
position. During deposition of all materials the sample did not need to be moved
giving a very good reproducibility of the samples. The sample was mounted on
a OMNIAX manipulator manufactured by Vacuum Generators. It can be moved
in all directions and is rotatable around the z- and y-direction. As can be seen
in figure 3.1 to the left of it (looking through the viewport at the sample stor-
age), an AUGER spectrometer from Omicron is mounted in order to investigate
the chemical composition of the samples. Its e-gun was also used for electron
diffraction under a small angle, called MEED. To the right a LEED system from
Omicron is installed to investigate crystal structures with a higher precision.
By a transfer mechanism, samples can be brought into a separate STM chamber
in-situ. The STM was manufactured by RHK, but the pictures presented later
were all taken with a system by Omicron, situated in another chamber. For this
a special transfer adapter was developed for bringing the samples under UHV to
the other chamber.

Beside these devices the MEDUSA chamber is also equipped with a sputter
gun to clean and smoothen metal substrates and a mass spectrometer for inves-
tigating residual gases. A sample storage could hold up to four samples and a
Chapter 3. Experimental methods and set-ups 24

smaller one two samples. By that a whole series of samples could be produced
with using the load-lock only once. This assured a very low pressure and espe-
cially a very low oxygen content in the chamber. The content of oxygen and water
in the residual gas was minimized by using a titanium sublimation pump. The
main and STM chamber were pumped by iongetter pumps supplied by Varian and
Leybold. The load-lock was equipped with a turbo molecular pump from Pfeiffer
which was pre-pumped by a oil-free plunger pump manufactured by Leybold.

3.2 Magnetic characterization

The magnetic behavior of the bilayers produced is the central issue of this work.
The different nature and the very different temperature dependence of gadolinium
and nickel is the reason of the very interesting switching behavior. Rare earth
and transition metals align antiferromagnatically in many systems investigated
in the recent years [46, 47]. Possible applications of these systems range from
sensing devices to components in data storage devices.
All samples were thoroughly characterized by a SQUID-magnetometer from Quan-
tum Design. Field and temperature dependence of the whole bilayer system could
be investigated and so the interplay of the two species was accessible. Experiments
with XMCD, X-ray Magnetic Circular Dichroism, were performed to observe the
magnetization of each material separately.
In the following the basic concepts and physics, involved in these experimental
methods, will be introduced and discussed. Citations for further going literature
will be given to the reader, in order to keep this section to a reasonable extent.

3.2.1 Superconducting QUantum Interference Device:

SQUID
SQUID magnetometers represent the most sensitive devices for measuring mag-
netic moments and magnetic fields respectively. Moments down to 10−4 ∗ φ0 can
be detected. φ0 = h/2e is called the flux quantum (see below). The set-up used
was manufactured by Quantum Design (San Jose, CA) and carries the model
name MPMS-5S XL. The samples can be magnetized by fields of up to ±50 kOe
and measurements can be carried out in a temperature range from 1.7 K to 400 K.
Nowadays many different measurement methods for magnetic moments and fields
are available like MOKE (Magneto-Optical Kerr Effect), VSM (Vibrating Sample
Magnetometry) etc. A brief overview and comparison is given in [48]. The major
advantages of SQUIDs are the temperature variability, which was used inten-
sively in this work, the high accuracy and good reproducibility. Its disadvantages
are long measuring times, high costs for purchase and helium consumption and
background signals from substrates and mounting straws. In the following basic
25 3.2. Magnetic characterization

concepts and physics involved in this technique will be introduced. A more de-
tailed description can be found in [49, 50, 51].

Figure 3.3: Two Josephson contacts forming a SQUID loop.

Figure 3.3 shows a schematic view of a dc SQUID loop consisting of two

Josephson junctions arranged to a ring structure. A Josephson junction is an
insulating layer between two superconductors, coupling them weakly. From the
Ginzburg-Landau theory it can be derived, that a dc current flows through such
a junction in absence of neither an external voltage nor magnetic field, the so
called dc Josephson effect:

j = jm sin δ0 (3.1)

jm : maximum current density, δ0 : phase difference between both sides of the

junction.

So without an external voltage applied, a current is flowing across the junc-

~ and U in the describing
tion. If an external voltage is applied, the potentials A
Schrödinger equation have to be gauge transformed, leading to the ac Josephson
effect.
 
2eU21
j = jm sin δ0 − t (3.2)
~

δ0 : Phase difference of the field free junction from equation (3.1), U21 : Ap-
plied external voltage.
Chapter 3. Experimental methods and set-ups 26

An additional oscillating current with the frequency ω = 2eU21 /~ arises over

the junction.
In the presence of a weak magnetic field, i.e. the BCS theory is still valid and
~ is a position-dependent phase shift can
the only effect of the vector potential A
be computed by:

Z2
2π ~ · d~l hc h
δ = δ0 + A φ0 ≡ (cgs) (SI: ). (3.3)
φ0 2e 2e
1

φ0 is the so called flux quantum. With these three equations (3.1), (3.2)
and (3.3) also complex arrangements of Josephson junctions can be described,
like a SQUID loop as shown in figure 3.3, applying simple Kirchhoff’s rules. A
magnetic flux φ is flowing perpendicular through the ring. A current going from
position 1 to 2 splits up onto the two branches a and b.

I = Ia + Ib = Ima sin δa + Imb sin δb (3.4)

   
δb + δa δb − δa
= 2Im sin cos
2 2

The second line evolves from assuming the same maximum current in both
branches (Ima =Imb ). Following the closed path around the ring (i.e. 1→2→1) the
phase can only change by discrete values 2πn, thus the magnetic flux is quantized
accordingly.
I
∇Φ · ~l = δtotal = 2πn (3.5)
C

The phase differences in the upper and the lower branch are connected to the
enclosed magnetic flux by the following equation, which can be computed from
~ · d~l over a half loop.
the integration of A

I 2 Z
2π ~ · d~l + 2π 4πλL ~j · d~l = 2πn + 2πφ .
δa − δb = 2πn + A (3.6)
φ0 φ0 c φ0
C C0

C’: Path from the right side of the upper junction to the right side of the
lower junction (so without passing an insulator layer), λL : London penetration
depth.
27 3.2. Magnetic characterization

Applying this result for the phase difference to equation (3.4) and taking into
account that the phase difference on a single branch, e.g. δb is connected to the
external magnetic flux by:
 
πφ 1
δb + = n+ π. (3.7)
φ0 2
By that equation (3.4) can be simplified to:
  
 πφ 
I = 2Im cos . (3.8)
φ0 
One can see from equation (3.8) that the detected current oscillates with
changing magnetic flux penetrating the SQUID loop by multiples of the flux
quantum φ0 . From this it is clear, that a device as shown in figure 3.3 can be used
to measure very small changes in magnetic fields. But in order to detect fluxes
greater than φ0 the SQUID loop is used in the MPMS system in a flux-locked-
loop feedback. A magnetic flux, opposite to the one from the sample, is applied
to the SQUID with an oscillating component (usually some 100 kHz). If the two
constant fluxes cancel each other the frequency of the oscillating component is
doubled, due to |cos(x)|. This is detected by a lock-in amplifier. This degenerative
feedback makes it possible to detect magnetic fluxes up to the maximum flux, that
can be generated by the magnet in the feedback loop.
In the following a short description of the set-up used and the special tech-
niques to achieve a resolution down to 10-4 ∗ φ0 will be given. All measurements
are controlled by a computer, which is following a previously generated measuring
sequence. In figure 3.4 a schematic overview of the used SQUID system is shown.
This inner part is suited in a helium dewar which is insulated by a vacuum gap
and a surrounding liquid nitrogen shield.
Straws, made of delrin, which is a purely diamagnetic plastic, are used to mount
the samples. The straw, holding the sample, is attached to a rod, which can be
moved up and down through the sensing coils. The detected signal is then trans-
ferred to the actual SQUID loop, that is positioned at the bottom of the helium
dewar. The set-up was equipped with the optional RSO mode (Reciprocating
Sample Option).
In the left part of figure 3.5 two types of pick-up coils and and the detected
signal, that is recorded if a sample is moved through it, are shown. The used
set-up holds a second-order gradiometer, labeled b) in the drawing. The use
of such a pick-up coil brings many benefits. Into a simple coil a change in the
homogenous external field would already induce a voltage. For comparison a first-
order gradiometer is in the picture, too. The first-order coils are insensitive to a
change of the homogenous external field, whereas field gradients induce a voltage.
The second-order coil rejects both homogenous fields and linear gradients, but is
more complicated in production, since the distances and diameters have to obeyed
accurately. Furthermore Vandervoort [52, 53] proved that the noise due to the
Chapter 3. Experimental methods and set-ups 28

Figure 3.4: Schematic overview of the SQUID magnetometer MPMS

XL-5S: Right: Probe components; Left: Zoom of the solenoid [52].

resistance of the solenoid is reduced by an order of magnitude.

The right part of figure 3.5 illustrates the RSO mode, which was used for all
samples, because the less precise DC mode has a too low resolution. The sample
is moved through the pickup coil with a frequency of up to 4 Hz and amplitudes
between 0.5 and 4 cm. This oscillation of the detected SQUID voltage is than
fed into a second lock-in, improving the resolution by one order of magnitude
in comparison with the DC mode. Two positions for adjusting the sample with
respect to the coils can be chosen. The center position, where the sample is
in the middle of the gradiometer, and the position, where the maximum slope
in the detected SQUID voltage. Temperature dependent measurements were all
performed at the center position, because it is less sensitive to displacements of
the sample. In hysteresis, the sample doesn’t drift due to thermal expansion of
the mounting straw, and the more precise maximum slope position can be used.
29 3.2. Magnetic characterization

Figure 3.5: Gradiometer pick-up coils: Left: Gradiometer designs and de-
tected voltage over displacement. Right: Illustration of the RSO mode; The
sample is moved at up to 4 Hz though the pick-up coils [52].

3.2.2 X-ray Magnetic Circular Dichroism: XMCD

Magneto-optical effects are known since the discovery of the Faraday effect in
1845. John Kerr discovered in 1876 the rotation of the polarization of light,
that is reflected at a ferromagnetic surface. This magneto-optical Kerr effect
is the basis of nowadays MOKE magnetometers and is driven by the different
absorption of left and right circular polarized light. The induced transitions,
using optical wavelengths (a few eV), are within the valence band. In 1975 Erskine
and Stern proposed the idea of a magneto-optical effect in the x-ray regime of
circular polarized light, called XMCD [54]. The first XMCD experiment could
be performed in 1987 by Schütz on a thin Fe foil [55]. This was made possible
after a tunable x-ray source with a sufficiently high flux was available. Therefore
synchrotron radiation is needed for these experiments.
In contrast to experiments using visible light in XMCD electrons in low lying
core levels are excited to states in the valence band or the continuum. With the
energy levels of the core shells being characteristic for the different elements, this
method is element-specific [56]. This was used within this work, in order to record
hysteresis loops of nickel and gadolinium, separately.
Chapter 3. Experimental methods and set-ups 30

3.2.2.1 From X-ray absorption XAS to XMCD

X-rays are strongly attenuated in matter according to the Beer-Lambert law

I(E, d) = I0 e−µd . (3.9)
In the energy range of x-rays photon absorption is the dominant contribution
to the attenuation, i.e. that electrons are exited from an initial state |ii to a
final state |f i and the photon energy is transferred onto them. Contributions by
Compton effect and photo effect can be neglected. So there appear characteristic
lines in the absorption spectrum that are defined and labeled K, L1 , L2 , L3 , M1 ...
after the initial states |1s1/2 i, |2s1/2 i, |2p1/2 i, |2p3/2 i, |3s1/2 i...

The transition probability is described by Fermi’s golden rule:

2π
Γi→f = |hf |Ĥ|ii|2 ρ(Ef ). (3.10)
~
From this equation one can see that two quantities determine the transition rate:
The interaction strength, described by the first term, which leads to the selection
rules (equation (3.11)) and the density of states near the Fermi level as final
states. Here we only consider electric dipole transitions (E1):

∆j = 0;
∆l = ±1;

+1 left circular

∆m = 0 linear (3.11)

−1 right circular

∆s = 0.

In XMCD the dependence of the absorption on the magnetization is used.

This effect is caused by the different number of final states that are available for
electrons with ”spin up” or ”spin down”, due to the different shift in energy of the
two subbands (see figure 2.4). Figure 3.6 shows schematically the absorption of
a photon with the energy E=~ω at the L2,3 -edge in a 3d-transition metal. The
possible final states are 4s and 3p.
The definition of the actual XMCD-signal is the difference in the absorption
coefficients µ+ (E) and µ− (E):
∆µ(E) = µ+ (E) − µ− (E) (3.12)
µ+ (E) : P~ ↑↑ M
~
µ− (E) : P~ ↑↓ M
~

P~ : Polarization/helicity of the photons; M

~ : Magnetization of the sample.
31 3.2. Magnetic characterization

Figure 3.6: Dipole transi-

tions in a 3d-metal: Accord-
ing to the selection rules tran-
sitions to the 4s and 3d or-
bitals are possible. The 4s-
band is very narrow and by
that can not take up many
electrons, but is not split due
to exchange interactions. The
3d-band splits up into a spin-
up and a spin-down branch,
making transitions with ∆m=-
1 more likely. Taken from [57].

In the experiments carried out at BESSY, the Berliner Elektronenspeicherring-

Gesellschaft für Synchrotronstrahlung, in the course of this work, the magnetiza-
tion of the samples was changed with an external magnetic field and the polar-
ization of the incoming photons was kept constant. For simplicity the measured
intensity with respect to the ingoing intensity I0 was recorded.

This sensitivity in the number of available states in the valence band is used
in NEXAFS experiments, with which the density of states above the Fermi-level
can directly be observed.
This difference in the cross-section of the absorption of x-rays between P~ ↑↑ M~
and P~ ↑↓ M ~ can lead to a significant difference in the absorption spectrum as
shown in figure 3.7.
From the definition equation (3.12) and figure 3.6 one sees, that a difference
in absorption is only visible if the magnetization is parallel/antiparallel to the
beam direction. Measurements in transmission on the in-plane magnetized system
Gd/Ni were carried out under an angle of 45°. So only the projection of M ~ on
the incident beam ~k is detected, lowering the XMCD-signal. Figure 3.8 shows a
schematic of the used UHV set-up ALICE at the BESSY.
The synchrotron BESSY, situated at Berlin, offers a large variety of x-ray
radiation for different investigation methods. The radiation is generated by elec-
Chapter 3. Experimental methods and set-ups 32

Figure 3.7: XAS and XMCD spectra of Nickel: The XMCD spectrum
(bottom) is the normalized difference between the two XAS spectra at the top.
The data was recorded from sample A4, consisting of Gd(50Å)/Ni(75Å)/Au(20Å)
on Si3 N4 .

trons that circulate in packages in the storage ring at an energy of 1.7 GeV and a
beam current of up to 400 mA. At 16 straight sections so called wigglers or undu-
lators consisting of superconducting magnets are installed that force the electrons
on a wiggled trajectory, see figure 3.9. The electrons are accelerated towards the
center of each half circle, what causes them to emit the x-ray radiation, which
has wide spread spectrum. The arrangement of the magnets determines, wether
linear, circular or elliptical radiation is emitted. At each site several experiments
are installed in order to optimize the use of the synchrotron. The energy loss
33 3.2. Magnetic characterization

Figure 3.8: UHV-chamber ALICE at BESSY: The sample is rotatable and

can be moved in all directions for alignment. The sample holder can hold up to
four samples.

Figure 3.9: Layout of an undulator: By changing the arrangement of the

for segments the type of the emitted radiation can be changed. Taken from
www.bessy.de.
Chapter 3. Experimental methods and set-ups 34

of the electrons, because of the emission of radiation, is compensated by four

cavity-resonators along the storage ring. Between the straight sections specially
designed permanent magnets focus the e-beam and force it on its circular path
through the storage ring, which is kept at a pressure below 1·10-11 mbar.

3.2.2.2 Sum rules

The magnetic moment of an atom can be divided into two contributions, originat-
ing from its spin polarization hSz i and its orbital momentum polarization hLz i,
according to:

orbital momentum : µL = −µB hLz i

spin momentum : µS = −µB hSz i.
(3.13)

With the transition probability and the number of electrons in the initial
states being different XMCD can be used to determine these two contributions.
In the following a short introduction to the so called sum rules will be given.
More detailed information is given in [57, 58, 59, 60]. The description will be
limited to the L2 - and L3 -edge as they are important for the magnetic transition
metals.

Figure 3.10 shows all possible transitions from 2p- to 3d-orbitals. The pro-
portionate contribution of the single transitions is indicated by the thickness of
the arrows and the percentage value. These differences in the transition proba-
bility leads to a spin polarization of the final state, called Fando effect.
The same is true for the orbital momentum l. The selective transition produces
a polarization of the orbital momentum of the final state. Both polarizations,
derived directly from the calculated transition probabilities, are also given in fig-
ure 3.10 as hli and hσi for the case of right circular polarized light.
Starting from the different transition probabilities, one can derive the XMCD
spectrum of a sample that has a magnetic moment, which is completely caused
by spin polarization and µl = 0. Then the L2 and L3 peak would have different
sign, but the same height. Whereas a hypothetical sample with µs = 0 and µl > 0
would show two peaks with the same sign but the L2 peak would be twice as high.

In an actual XMCD measurement one would of course observe a superposi-

tion of these two cases, leading to a spectrum like in figure 3.7. But the consid-
erations above can be used to evaluate the contributions of µl and µs from the
ratio of the areas under the peaks.
35 3.2. Magnetic characterization

Figure 3.10: Transition probabilities for 2p→3d: σ + -polarized light excites

the initial states with a different probability, indicated by the thickness of the
arrows and the percentage of the overall transition strength.

1
µs = − (A3 − 2A2 )µB
C
2
µl = − (A3 + A2 )µB (3.14)
3C
A2 and A3 are the areas under the L2 and L3 peaks. The integration bound-
aries are given by the limits of the absorption band of the respective transition.

Z
A3 = µ+ (E) − µ− (E)dE
L3
Z
A2 = µ+ (E) − µ− (E)dE (3.15)
L2

The constant C cannot be derived from the spectra, but must be evaluated
from reference measurements by e.g. SQUID that give a number of the total
atomic momentum of the sample.
Chapter 3. Experimental methods and set-ups 36

3.3 Structural investigations

The morphology and crystal structure of the samples influence strongly the mag-
netic behavior of the Gd/Ni-bilayer system. In order to resolve the great differ-
ences between the samples deposited on Al2 O3 and Si3 N4 several methods were
deployed using diffraction (XRD, XRR, MEED), energy transfer (AES, RBS) or
scanning techiques (STM, TEM).
In this section only a brief introduction of these investigation techniques will be
given, as far as it is necessary for the understanding of the later presented results.

3.3.1 Scanning Tunneling Microscopy (STM)

The first set-up of an scanning tunneling microscope was developed and pub-
lished by G. Binning and H. Rohrer in 1982 [61, 62]. For this they were awarded
the nobel prize in 1986. It allows to picture surfaces directly in real space down
to atomic resolution. Later they extended the idea of a scanning probe to the
atomic force microscope (AFM), that allows to investigate non-conductive sur-
faces. Nowadays many subtypes are available to investigate magnetic structures,
adhesion forces, etc. or manipulate surfaces on the nanoscale.

Figure 3.11: Illustration of a SPM: Left: conceptional idea; Right: Realization

in the case of a STM, using a tunneling current between probe and sample.

The basic idea is to scan the surface with a probe, that has only a few
nanometers in diameter, line by line and then bring these lines together to form
a picture. Figure 3.11 illustrates this technique. In order to do that, a very
accurate positioning of the probe is necessary. This can be achieved by attaching
the probe to a piezo tube, as shown in figure 3.12. Applying voltages up to 400 V
in case of the used set-up from Omicron the tube bends and can be moved over
the surface. The positioning in z-direction is done at some models by the same
tube, applying off-set voltages to all segments at a time. The UHV-STM from
Omicron uses a separate one. Additionally a coarse positioning system is needed,
usually operating with piezo drives, too.
In the case of a STM the measured feedback property is a tunneling current
between the tip and the surface, what limits this technique to conductive samples.
37 3.3. Structural investigations

Figure 3.12: Piezo tube: By applying voltages up to 400 V to the different

parts of the tube, the tip can be moved in all directions.

The tip in the used set-up consisted of 0.38 mm thick tungsten wire, which was
etched in 3m-NaOH to a tip with a curvature below 20 nm. In an ideal picture,
as in figure 3.11, the current flows through a single atom at the end of the tip.
SEM pictures of a STM tip are shown in figure 3.13. The curvature at the end is
below 10 nm. Although most tips produced were of similar shape, their imaging
qualities could be very different. Some problems like a multitip or oxidization
could be removed by electric field emission.

Figure 3.13: SEM picture of a STM-tip: The tips produced were of a high
constant quality.

A short description of the theoretical background will be given, following

the description of a tunneling hamiltonian after Bardeen [63]. If the overlap of
the wave functions of the surface atoms and the tip and the applied voltage U
are small, perturbation theory can be used, what leads to an expression for the
tunneling current.

I ∼ U ρtip (EF )ρsurf ace (~r0 , EF ) (3.16)

ρ are the density of states at the Fermi-level in the tip and the surface. So
the detected distances, later combined to a picture, is not directly the surface,
but the electronic density of states near the Fermi-level.
Chapter 3. Experimental methods and set-ups 38

Figure 3.14: Feedback control in a typical STM: a) Conceptional set-up of

a STM, b) constant current mode for feedback, c) constant height mode. Taken
from [45].

The density of the surface states decreases exponentially in first order with
an effective decreasing length κeff in vacuum.

r
2me B
κef f = + |~k|| |2 (3.17)
~2
Wtip + Wsample |eU |
B = −
2 2

With B being the barrier height, ~k|| the electron wave vector parallel to the cur-
rent. By that the tunneling current drops exponentially with increasing distance
z to the surface and is therefore a very sensitive measure for the surface structure.

I ∼ exp[−2κef f z] (3.18)
39 3.3. Structural investigations

In order to scan a sample the movement of the tip over the surface has to be
controlled by a feedback loop. Two major modes of operation are distinguished,
according to the parameter, which is kept constant, see also figure 3.14.

ˆ constant current mode: The feedback adjusts the z-Piezo in such a way,
that the tunneling current remains constant. The recorded z component is a
real space picture of the surface. This mode is mostly used on larger scales,
in order to prevent the tip from touching the surface and to investigate
larger height differences.

ˆ constant height mode: In this mode the scanning piezo is only moved in
x- and y-direction. The z value is kept constant and so the current changes
with changing distance between the tip and the surface. A picture of the
surface is obtained by plotting the current over the x/y-position. This mode
is usually applied to investigate surfaces on an atomic scale.

Besides the imaging of the surface, information about band structure, work
function etc. can be obtained by I-U- or I-z-spectroscopy [64, 65].

3.3.2 X-ray diffraction and reflectometry

X-rays are a powerful tool to investigate the crystal structure and morphology
of thin films and bulk materials. It is a well established technique and many
commercial systems are available. Most of the measurements within the scope
this work were performed at the ESRF in Grenoble by Mireille Maret.
Two different investigations have been undertaken: Large angle diffraction (XRD)
and small angle/gracing incident refractometry (XRR). Detailed descriptions can
be found in [2, 30, 66]. XRD relies on the Bragg condition, that states that under
certain angles the scattered waves from the atoms interfere constructively.

2dhkl θhkl = nλ (3.19)

dhkl : Distance between two adjacent lattice planes, indexed by hkl, the Miller
indices.

With the distances in a solid being in the range of several Å only, the wave
length λ must be on the same scale. That is why x-rays are needed. They are pro-
duced by electrons accelerated onto a copper target, producing bremsstrahlung
and characteristic radiation. The device used, manufactured by Phillips, operated
on the Cu-α line (λ = 1.5418 Å). The radiation is further filtered by a graphite
monochromator. In the so called θ-2θ geometry, the incident beam and the di-
rection of detection are under the same angle with respect to the surface. The
source is fixed and with the sample being rotated by θ, the detector is moved
by the angle of 2θ. Under large angles (≥20°) the resolved distances are in the
Chapter 3. Experimental methods and set-ups 40

range of the interatomic distances in a solid, making it possible to investigate the

crystalline structure of a sample.
If one goes to very small angles (2θ ≤∼6°) the beam is partially reflected at
the boundaries of the different layers in a thin film, due to a change in density
and thereby a change in refractive index. The resulting graph shows an oscilla-
tory behavior, the so called kissing fringes. From their distances and heights the
thickness of the single layers can be computed. This method is very sensitive to
the surface and interface roughness since it averages over a large area. Because
of the small incident angles additional slits have to be introduced, in order to
keep the illuminated area within the sample size. But this even further reduces
the luminescence and longer measuring times are needed (several hours). This
method is called X-ray reflectrometry (XRR) and was used to calibrate the layer
thicknesses and to investigate the interface roughness.

3.3.3 Electron Diffraction: MEED

This diffraction technique is a simple but effective way to investigate the crystal
structure of surfaces. Detailed descriptions can be found in [2, 30] . MEED is an
acronym for Medium Energy Electron Diffraction. It was deployed in-situ and
can be used even during evaporation. Electrons are emitted from an e-gun, in
the used set-up with energies up to 3.5 keV, and hit the sample under a glancing
angle below 8°. The reflected electrons hit a fluorescent screen, from which the
diffraction pattern can be recorded by a CCD camera. The de-Broglie wave length
of the electrons is:

h
λ= √ ≈ 0.12Å. (3.20)
2me E
One drawback of the used set-up in the MEDUSA is that the available en-
ergies are not higher. The e-gun is originally part of the Auger spectrometer.
Another disadvantage is that the screen is part of the LEED system and by that
very close to the sample and it is spherical. The latter leads to that from the
obtained pictures it is very difficult to evaluate absolute numbers for lattice pa-
rameters, but qualitative investigations are possible.

Reflexes can be detected at positions, where the Laue condition ∆~k = G ~

~
is fulfilled, with ∆k being the change in the electron wave vector and G an ~
arbitrary vector between two lattice points in the reciprocal space. Constructive
interference occurs if the Ewald-sphere, representing all points accessible to the
scattered electrons in the reciprocal space, and lattice points of the investigated
crystal intersect. In the special case of glancing incident the diffraction takes place
on a 2 dimensional surface, so there is no condition for constructive interference
in the 3rd dimension and the lattice points in the z-direction get so close that
they form rods. Figure 3.15 shows a schematic of the used set-up. In [66] further
41 3.3. Structural investigations

Figure 3.15: Schematic layout of the MEED set-up in the MEDUSA UHV-
chamber.

descriptions of electron diffraction are given, showing the impact of crystallinity,

chemical order and also morphology of surfaces to the diffraction patterns.

3.3.4 Chemical composition by AES

In Auger spectroscopy core electrons are excited to the continuum, leaving a
vacancy in its low lying initial orbital. An electron from a energetically higher
orbital takes its place and the energy difference Ei →Ef can either be emitted as
a X-Ray photon or transferred to another electron, which leaves the atom. The
emission of an Auger electron is more likely in light atoms, whereas the X-ray
emission appears more often in heavy atoms (Z > 60). In figure 3.16 a simple
comparitive illustration of these processes is given.
The emitted electrons have distinct kinetic energies that are characteristic
for the emitting atom. This is used in AES in order to determine the chemical
composition of the investigated surface. Electrons are accelerated towards the
surface by 3 keV in the used set-up. The secondary electrons are collected and
their kinetic energy analyzed in a cylindrical mirror analyzer. It consists of a
grounded tube, having a diameter of 200 mm, and several insulated metal rings,
that build up an electrical field gradient. By changing the slope and height of the
applied electrical potential only electrons with a certain energy are focused into
the detector, a so called channeltron. The CMA can filter electrons with energies
from 0 eV to 1250 eV with a resolution of about ± 1 eV.
The set-up, used in the MEDUSA chamber, is equipped with a lock-in amplifier.
It applies an additional small ac voltage to the potential in the CMA and that’s
why the detected signal is not N(E) but dN(E)/dE.
Chapter 3. Experimental methods and set-ups 42

Figure 3.16: Auger effect and X-ray emission: The illustration shows
the two competing processes that can occur, if a core electron is removed. The
example atom is titanium. The energy of the Auger electron, leaving the atom, is
EL2 - EM 4 - EM 3 = 423 eV, whereas the emitted x-ray photon possess the energy
EL2 - EM 4 = 458 eV. Taken from [67].

3.3.5 Rutherford backscattering

RBS is a scattering method employing ion beams. Like in the classic Rutherford
experiment the Coulomb force between the projectiles and the nuclei in a solid is
the basic interaction of the scattering process.

Figure 3.17: RBS Schematic layout of a RBS-experiment. The reflected ions are
detected under the angle θ.

RBS is a very powerful tool in order to investigate the structure and in-
terfaces of thin films, since it makes it possible to determine the depth profiles
of the different materials in a sample. But for that one needs a source of high
energetic ions with a well defined energy and a very sensitive detection systems.
43 3.3. Structural investigations

The experiments were carried out in the group of Professor Hans Hofsäss at the
university of Göttingen. In this work only a brief description of this method will
be introduced. For further reading there is lots of literature available, especially
[2, 7].
Treating the target and projectile as hard spheres, that interact via Coulomb
forces, one can derive the so called kinematic factor K from energy and momen-
tum conservation. It describes the energy transfer.
"p #
1 − [(mp /Mt ) sin θ]2 + (mp /Mt ) cos θ
K= (3.21)
1 + (mp /Mt )

mp : mass of the projectile ion

Mt : mass of the target nucleus
θ : scattering angle

Figure 3.18: RBS: The kinematic factor K for different projectiles and scattering
angles. Taken from [7].

In figure 3.18 the dependence of the kinematic factor on the target mass is
shown for different projectile masses on the left and for different scattering angles
on the right. In the set-up used 4 He ions were used and the detector was placed
under an angle of θ = 127°. The three elements in the investigated samples differ
much in their atomic weights (Ni: 58 u, Gd: 158 u, Au: 196 u), but as can be
seen in figure 3.18 (left) the slope of the curve is already above 50 u small.
Chapter 3. Experimental methods and set-ups 44

In thin films the energy loss of the particles passing through the material is
exploited. The further the ions penetrate the sample until they are scattered the
more energy they lose. So the energy difference between an ion scattered at the
surface and one at a depth x inside the thin film is given by:
"   #
K dE  1 dE 
∆E = + · x. (3.22)
sin α1 dx E0 sin α2 dx KE0
α1 , α2 : Angle between the incident/outgoing ion beam and the surface.

3.3.6 Transmission electron microscopy

High Resolution Transmission Electron Microscopy was used to get cross-sectional
images of the samples. The used set-up was a FEI Tecnai F20 microscope.
The first TEM was built in 1931 by Knoll and Ruska, only six years after de
Broglie predicted the wave-like nature of electrons in 1925 (see equation (3.20)).
In a classical TEM the ray path of the electrons is like in an optical microscope
and one obtains directly an image of the specimen. The glass lenses are replaced
by electromagnetic leses. The electrons are set free from a nearly punctual source
and accelerated by voltages in the range of 100 kV to 300 kV. The resulting wave
length is in the sub Angstroem regime, but lens defects and charging effects limit
the resolution it to atomic distances.

Figure 3.19: Different modes of TEM: Positioning the detector at large angles
gives a chemical contrast proportional to I ∼ Z2 .
45 3.3. Structural investigations

There are many different types of TEMs and operation modes for them, so
only the techniques, that were actually used, will be presented. All measurements
were carried out by Richard Vanfleet at Brigham Young University in Utah. In
order to obtain a chemical contrast between the different materials in the samples
at a very high resolution, so called HAADF (high angular annualar dark field)
imaging was performed. The detector collects only electrons that are scattered
over large angles. The intensity depends on Z2 . The spatial resolution is obtained
in combination with scanning the sample by a highly focused electron beam.
In order to get a cross-sectional view on the sample it was glued by epoxy
onto a piece of silicon. The metal film is then in the middle of the sandwich,
shown schematically in figure 3.20. Pieces of about 1 by 2 mm were cut out from
the sandwich. One side is polished flat onto which the sample is fixed to a holder,
consisting of a glass rod, by wax. The other side standing out is polished at a
small angle (1° to 2°). That puts a wedge shape to the specimen and the thin
edge of the wedge is positioned so that the area of interest is thinned in the point.

Figure 3.20: Sample preparation for TEM: Left: The region of interest is
placed in the middle of the sandwich. Right: Positioning on the sample holder.
Chapter 3. Experimental methods and set-ups 46
Chapter 4

Experimental Results

4.1 Sample description and preparation

A schematic of the samples is shown in figure 4.1. The substrates used were
sapphire and silicon nitride. The Al2 O3 substrates have a polished [0001] oriented
surface and are single crystalline. The sample dimensions were 3.5 x 3.5 x 0.3 mm3
and 10 x 5 x 0.5 mm3 for SQUID and X-ray measurements. They were supplied by
Crystek. The silicon nitride substrates were produced by Silson by passivation of
the surface of a Si-wafer with nitrogen. After polishing one side a quadratic cavity
was etched into the backside leaving a 150 - 180 nm thick membrane. The outer
dimensions were the same as for the Al2 O3 substrates, large ones for performing
XMCD measurements and small ones for SQUID measurements. The membranes
are very fragile, so a few of them broke while mounting the sample into a straw
before investigating it by SQUID. The Si3 N4 surface is amorphous although the
pure Si inside is single crystalline. Their quality varied considerably from one
batch to another. But even within one simultaneously processed shipment from
Silson large differences in the cleanliness and surface roughness were observed.
Figure 4.2 shows an AFM picture of a Si3 N4 membrane of poor quality. There
is a varying number of unknown particles on the surface that can significantly
influence the growth conditions and thereby the magnetic properties of the Gd/Ni
bilayers. In most cases the surface of the Si3 N4 substrates was cleaner. The best
XMCD results presented in section 4.4.4 were obtained from samples grown on
Si3 N4 substrates that were supplied by our collaborator Olav Hellwig.
First, a Gadolinium layer of a constant thickness of 50 Å was deposited and
on top of it a layer of nickel with a varying thickness between 15 Å up to 100 Å.
A capping layer of gold, 20 Å thick, prevented the bilayer system from oxidation.
Special care has to be taken for the deposition of gadolinium. Because of its
high reactivity, the evaporation must be done at very low residual pressure and
especially at very low oxygen content in the residual gas. During evaporation
the pressure could be kept below 1·10−9 mbar and a titanium sublimation pump
minimized the oxygen and water concentration. In order to improve the vacuum
Chapter 4. Experimental Results 48

Figure 4.1: Sample composition: The thickness of the Gd layer was kept
constant to 50Å followed by xÅ (x=15, 25, 50, 75 and 100) of Ni. 20 Å Au was
deposited as a capping layer.

conditions, a cooling trap was filled with liquid nitrogen, lowering additionally
the pressure during the time of evaporation. Two different series of samples of
the same composition were produced, one at Tdep = 100 K by cooling the sample
stage with liquid nitrogen and one at room temperature. The liquid nitrogen
pipeline was acting as an additional cooling trap for the residual gas due to its
large surface.

Figure 4.2: AFM picture of a Si3 N4 substrate: Some substrates were

contaminated with a large number of particles, left over from the membrane
processing.
49 4.2. Thin Gadolinium layers

As a first step a series of calibration samples were produced, since the amount
of evaporated material can only be monitored indirectly if using an EFM e-beam
evaporators(see figure 3.2). This has to be done every time new material is
introduced, because slight differences in the alignment of the crucible with respect
to the filament and the chimney can cause different evaporation rates, although
the same current IF lux is observed. It turned out that for gadolinium the flux can
deviate by up to a factor of 3 from the previous calibration.

4.2 Thin Gadolinium layers

As mentioned before, as a first step, samples of a pure gadolinium layer covered
by gold were produced for calibrating the real flux of evaporated material at
a certain ion current IF lux , measured with the e-beam evaporator control unit.
This was done by small angle x-ray reflectrometry at the ESRF, the European
Synchrotron Radiation Facility in Grenoble. In this paragraph only data from
diffraction methods will be presented. Measurements by SQUID and STM were
performed, too. They will be shown in sections 4.3 and 4.4.
In figure 4.3 a large angle θ − 2θ XRD scan is shown. There is a strong reflex
from the inner crystalline silicon and a broader peak which is attributed to the
above Si3 N4 layer with a thickness of about 180nm. If Gd would grow crystalline
on this substrate a peak at 2θ = 28° should be visible, coming from a reflex of
the [10-10]-plane. There is no indication of crystalline growth on Al2 O3 neither
[47]. In literature there are reports of crystalline growth of Gd only on tungsten
[68] and niobium [69].
The amorphous growth of gadolinium was also observed by electron diffraction,
as shown in the two following figures. The first series of pictures were obtained by
low energy electron diffraction, LEED (see figure 4.4). For this purpose the sample
has to be rotated by 90° around the z-axis of the manipulator and the evaporation
has to be interrupted. On the other hand, LEED is independent of the azimuthal
orientation of the sample and information obtained from the dependence on the
energy of the incoming electrons is the advantage of this technique. In the top
row, the images a) and b) show diffraction patterns of a clean Al2 O3 -[0001] surface
at 249.5 eV (a) and 432 eV (b). Sapphire is an insulating material, so in order to
avoid a too strong charging effect of the sample, conducting silver glue was used
for mounting it onto the sample holder. The diffraction patterns indicate the high
crystalline order of the substrate and a well oriented flat surface. At the bottom
row images are shown after deposition of 50 Å gadolinium (c) and 50 Å nickel (d),
both taken at 400 eV electron energy. The deposition temperatures were 100 K for
gadolinium and room temperature for nickel. Only a blurry intensity distribution
was observed in both cases in the whole accessible energy range up to 1000 eV,
i. e. that the material grows amorphous on Al2 O3 . No LEED spots could be
observed neither from a 50 Å nickel layer deposited onto a 50 Å gadolinium layer.
Chapter 4. Experimental Results 50

Figure 4.3: XRD-scan of a 80Å Gd film: Gd deposited on a Si3 N4 substrate

and covered by 30Å Au.

Figure 4.4: LEED images: a) Al2 O3 at E = 249.5 eV, b) l2 O3 at E = 432 eV,

c) 50 Å Gd Tdep = 100 K E = 400 eV, d) 50 Å Ni Tdep = 300 K E = 400 eV.
51 4.2. Thin Gadolinium layers

The same was observed by Treubel in [47] during evaporation by MEED,

where the electrons arrive under a small incident angle, typically 8° - 10°. The
used screen was the one from the LEED set-up, which is spherical, so the picture
can only taken qualitatively. The distances between the reflexes do not give the
correct lattice parameter. But it can be clearly seen that the reflexes of the plain
substrate vanish already after the deposition of 2.33 Å gadolinium at 100 K.
This proofs the amorphous growth. But all diffraction methods are insensitive
to a nanocrystalline structure with grains smaller than the coherence length. As
shown later in this work, STM reveals grains typically below 3 nm in diameter.
Attempts to grow a textured Gd film on other substrates like WSe2 , Pt- or Cu
seed layers were unsuccessful, too. Similar measurement on Si3 N4 would be in vain
since the substrate itself is amorphous and no constructive interference appears.

Figure 4.5: MEED images of Gd on Al2 O3 : MEED at 3 keV of the initial

growth of Gd at 100 K deposition temperature. The clear reflex pattern of the
Al2 O3 substrate disappear after depositing one monolayer of Gd [47].

The calibration of the flux of evaporated material was done by small angle
x-ray reflectrometry. These measurements also reveal the quality respectively the
smoothness of the interface and surface of the sample. All calibration samples
include the layer of interest of unknown thickness and a capping layer of gold,
usually 25 to 40 Å. In the case of gadolinium the interface is very rough, so the
minima in the interference pattern are only poorly developed. In order to sim-
ulate the data in more detail a system of two layers of pure elements and an
intermediate alloy layer was assumed. The good miscibility of Gd and Au even
Chapter 4. Experimental Results 52

Figure 4.6: XRR scan of a Gd/Au bilayer: The as-grown sample shows
only a poorly developed interference pattern. Three months later no pattern was
observable anymore.

leads to a slow intermixing on a long term. Figure 4.6 shows a XRR-scan between
2θ = 1° and 10° of a Gd(80Å)/Au(25Å) bilayer. Al2 O3 [0001] was used as substrate
and the deposition temperature was kept to 100 K. The same XRR-measurement
was repeated three months later and the interference pattern vanished completely,
what gives the indication of a strong intermixing caused by a solid state reaction
between Gd and Au. In contrast in the case of Gd/Ni bilayers such aging was
not observed. Even one year after production the magnetic behavior remained
the same (see section 4.4). The intermixing of Gd with transition metals is not
observed for all 3d-elements and differs drastically from one material combination
to another. Strong intermixing in multilayer systems is reported for e.g. Gd/Co
[70] and Gd/Cu [71], whereas for Gd/Fe [24] or Gd/Ti [72] no or only a small
tendency to interdiffusion is known.
For comparison a XRR-measurement of a Al2 O3 [0001]/Ni(114Å)/Au(25Å) bilayer
is shown in figure 4.7. It matches very good with simulated data for this bilayer
system. It must be mentioned at this point that the thickness of the single layers
and of the whole system is actually derived from comparing the measured interfer-
ence pattern with a calculated one, that reproduces the measurement the closest.
The minima, called Kissing fringes, can be clearly seen, due to the smooth and
sharp interface between the two layers. The result can be accurately simulated
by simple bilayer system.
The morphology and structure of the Gd film are crucial properties for its
magnetic behavior. In order to connect the observed magnetic and structural
properties of the produced Gd/Ni bilayers the following two aspects are necessary
to know.
53 4.2. Thin Gadolinium layers

Figure 4.7: XRR scan of a Ni/Au bilayer: The interference minima are
distinctive and the observed pattern can be simulated by two thin layers with a
sharp interface.

Figure 4.8: Dependencies of TC : The average grain size <D> has a strong
influence on TC [73] (left); For a 50 Åthick Gd film the Curie temperature is
already significantly lower as for bulk material [68] (right).

Michels et al, published in [73] data about the grain size dependence of
the Curie temperature in Gd. It turned out that for grains of less than 5 nm
diameter the Curie temperature is reduced considerably. As can be seen in sec-
tion 4.3.4 there are significant differences in the growth of Gd on the two used
substrates. But the values of TC measured in this work are still smaller than
the ones published in [73] for the corresponding grain size. This might be caused
Chapter 4. Experimental Results 54

by the reduction in TC in thin films, as described in [68]. If the film thickness

is decreased below 30 monolayers TC drops drastically. Both finite size effects
contribute to the obtained magnetic transition temperatures and the different
switching behavior of the bilayers on different substrates. Figure 4.8 shows the
dependency of the Curie temperature on the grain size (left graph) and on the
film thickness (right graph).

4.3 Structure and morphology of the Gd/Ni-

bilayer system
The Gd/Ni bilayers were characterized by many different methods in order to
characterize the structure of the materials, the layer structure, the chemical com-
position and the interface condition. In literature the reported growth modes
and patterns of bi- and multilayered films of Gd together with transition metals
are not consistent. Some account for the Gd/Co system strong or even complete
intermixing [14, 22, 74], whereas in [70] sharp interfaces are reported. Gd/Cu
multilayers are similar in structure due to the diffusion of Cu into the Gd layer
[71, 75]. There is hardly any data about the Gd/Ni system. In [13] interdiffusion
is observed if Gd is evaporated onto a Ni single crystal. Whereas in [76, 77] only
a slight intermixing around the interface is reported in Gd/NiFe (80 at. % Ni
content) multilayers. Smooth layers with sharp interfaces were obtained in mul-
tilayered thin films of Gd together with the transition metals of the first period
to the left of Co, like Gd/Fe [19, 24]. Very good structural quality exhibit layered
thin films of Gd/V [78], Gd/Cr [70] and Gd/Mn [37, 79].
The tendency of intermixing seems to be similar for systems of Gd in contact
with transition metals of the lower periods, like interdiffusion with Ag and Pd
and not together with Y [80], Nb [69, 81] and Mo [82]. Complete intermixing of
even thick layers appears in Gd multilayers containing noble metals like Pt, Au
and Ag [47, 83]. Therefore it was decided to deposit first the Gd layer, then the
Ni and at last the capping layer of Au. From reference samples like in figure 4.6
it was also observed that the magnetic moment of the Gd layer is decreased ap-
proximately by a factor of 3, due to the disturbed exchange interaction among
the Gd atoms. A similar decrease. is reported in [84] for Gd-Au bulk alloys.
In order to obtain a definite picture of the structure and morphology of the sam-
ples many different investigation methods were deployed. First x-ray diffraction
and reflectrometry measurements are presented. The chemical composition could
be investigated in-situ during evaporating material by AES, what is discussed
in the section afterwards. Then RBS and TEM measurements are shown in the
following subsection. The last part of this section deals with results from STM
pictures of pure Gd layers on the different substrates and of complete samples
and the dependence of the material growth on the deposition temperature.
55 4.3. Structure and morphology of the Gd/Ni-bilayer system

4.3.1 X-ray analysis of Gd/Ni/Au thin films

In the previous section in figures 4.6 and 4.7 small angle X-ray reflectometry
measurements were presented, showing that thin gadolinium grows on Al2 O3 and
Si3 N4 with a large surface roughness. In addition the intermixing with the Au
capping layer of pure Gd films could be observed. Similar reflection experiments
on bilayers from the series containing a 50 Å thick gadolinium layer were in vain
and the layered structure could not be resolved due to the low coverage and high
interface roughness.

Figure 4.9: Large angle X-ray diffraction of a Gd/Ni bilayer: This bilayer
contains a 600 Å thick Gd layer. A peak attributed to the Gd(002) around 30.5°
is visible (calculated position 30.92°).

In the following X-ray diffraction measurements of Gd/Ni and Ni/Gd bilay-

ers capped by a gold film, in addition to the one in figure 4.3 of a pure thin Gd
film, will be discussed.
The first figure 4.9 shows the XRD scan under large angles of a bilayer with
the composition Al2 O3 /Gd(600Å)/Ni(100Å)/Au(40Å). If the film thickness is in-
creased to this high value a peak attributed to the Gd(002) reflex can be observed
in contrast to the samples with only 50 Å Gd layer and as others described in [47].
This gives rise to the interpretation that the observed vanishing of the MEED
and LEED diffraction patterns is due to the very small size of the formed grains.
Chapter 4. Experimental Results 56

In section 4.3.4 it will be shown that Gd grows in a granular structure but

with a very small average grain size around 2 nm. Nanocrystals of this size
can possess a crystalline ordering without delivering discrete diffraction patterns,
because the formed structures are smaller or of similar size as the coherence length
of the used radiation or particles. But the Gd(002) peak is very broad, like it was
also for Gd films on Si and Nb and Pt seed layers, published in [83].
Figure 4.10 shows the results of XRD measurements on two bilayers. The
graph at the top, labeled a), is taken from a Al2 O3 /Gd(200Å)/Ni(75Å)/Au(50Å)
sample.
There is a very sharp but small peak at 20.5° that can be ascribed to the
GdNi(101) reflex, which is expected at 20.77°. The fine structure of the peak
suggests that at the interface the alloyed material forms large grains or even a
continuous film. The low signal height in contrast shows that not much material
is involved in the alloy.
In addition a small peak is observable at the position of the Gd(002) reflex. The
small grain size and low crystalline quality of Gd grown on Al2 O3 are mainly
responsible for the broadening and lowering of the signal, and of course the fact,
that the layer has partially alloyed. There are no reflexes from the Ni layer. In-
dependent measurements on pure Ni films grown on Al2 O3 , Si3 N4 or glass have
shown that under the given conditions of deposition rate and temperatures nickel
always grows amorphous. The inset on the right shows a magnification of a peak
that is identified as the Au(022) reflex, so the capping material grows at least
partially polycrystalline on the smooth but amorphous Ni surface.

In the graph b) below the results of the same measurements on a sample

of the following composition, Al2 O3 /Ni(75Å)/Gd(200Å)/Au(50Å) are presented.
There are two major differences:
The Gd(200) peaks appears much stronger but still very broad and the Au(022)
reflex vanishes completely. Compared to the one in figure 4.9 the Gd(200) peak
is still much smaller because of the thinner Gd film thickness. The GdNi(101)
peak is only slightly smaller and narrower compared to the graph above. This
shows that nearly as much alloy was formed as in the film with inversed stacking
order. The vanishing of the Au(022) reflex can be explained either by a strong
formation of a GdAu alloy, what is supported by the measured reduction in Msat
at 4 K, or the poor growth conditions on the rough Gd surface.
57 4.3. Structure and morphology of the Gd/Ni-bilayer system

Figure 4.10: XRD scans of Gd/Ni and Ni/Gd bilayers:

a) Al2 O3 /Gd(200Å)/Ni(75Å)/Au(50Å): There is a small peak observable that
comes from the Gd(002) reflex, a very sharp but still small reflex is visible that
can be attributed to the alloy GdNi(101) peak. The Au capping layer also shows
crystalline quality, giving rise to the Au(022) reflex
b) Al2 O3 /Ni(75Å)/Gd(200Å)/Au(50Å): In reversed order the Gd(002) reflex is
much stronger developed, but the Au peak around 64° vanishes.
Chapter 4. Experimental Results 58

4.3.2 In-situ Auger spectroscopy on Gd/Ni bilayers

The element specific detection of an Auger spectrometer was used to investigate
how fast Ni can cover a thin Gd film. For this, a 100 Å thick Gd film was
prepared at 100 K onto a Al2 O3 [0001] substrate. While depositing Ni at a constant
rate onto this Gd film a so called peak-to-peak measurement was performed.
The energy analyzer of the Auger electrons is switched during this between two
energies that are characteristic for the different elements. The relative signal
height is then a gauge for the abundance of each material. The recorded data are
shown in figure 4.11.

Figure 4.11: Auger analysis: Simultaneous measurement of the relative signal

strength, originating at the characteristic energies 61 eV (Ni) and 138 eV (Gd)
as afunction of thickness [47].

Generally, Auger spectroscopy is a very surface sensitive method, because

the generated Auger electrons can penetrate only a few monolayers of material.
If one assumes a flat film of Gd at the beginning of the measurement shown in
figure 4.11, the slow saturation of the Ni signal suggests a strong intermixing. It
takes 50 Å of nickel to dispose of the the gadolinium signal, i.e. to completely
cover it. But in the context of the observations in the previous sections, that
the deposited gadolinium forms a very rough film, this gradual vanishing may be
attributed to the morphology. The deposited nickel slowly fills up the ”valleys”,
leaving blank spots of gadolinium at the surface. D. LaGraffe interpreted in
[13] similar AES plots by intermixing, but did not perform any investigations
on the morphology of the films. After the deposition of 100 Å Ni, Gd was
evaporated onto the sample. In this case the signals saturate much faster. It only
takes 10 Å of gadolinium to get a constant ratio of Auger electrons from the two
59 4.3. Structure and morphology of the Gd/Ni-bilayer system

species. This indicates that less intermixing takes place. But in this succession
the incoming Gd impinges a flat surface of Ni, as can be concluded from STM
pictures of Gd/Ni/Au samples in figure 4.19. Section 4.5 presents a model of
the Gd/Ni bilayer system that exhibits all the observed structural and magnetic
properties.
On the other hand AES has shown strong intermixing at higher temperatures. In
[47] peak-to-peak measurements of a Gd/Ni bilayer are shown during annealing
the sample. A strong increase in the Gd signal was observed after the sample
temperature has reached 250°C. This alloy formation at elevated temperatures
is reported in literature for many rare earth/transition metal layered thin films
[13, 70, 71].

4.3.3 Characterization of the layer structure and chemical

composition by RBS and TEM
RBS experiments were carried out at the University of Göttingen in the group of
Prof. Hans Hofsäss. The used setup possesses a spatial depth resolution down
to 1 nm, which is at the cutting edge of this technique. A sketch of it is shown
in figure 4.12. Samples of the stacking composition Gd(50Å)/Ni(100Å)/Au(20Å)
deposited on Al2 O3 and Si3 N4 at substrate temperatures of 100 K (samples A5)
and room temperature (samples A9), were examined.

Figure 4.12: RBS measurements: The incident He+ -ions had 450 keV kinetic
energy. The recoiled He-ions were detected under an angle of 127° with respect
to the incoming beam.

From the results the layer thicknesses were calculated. For reference the Au
layer was assumed to be exactly 20 Å. This presumption is justified by the RBS
measurements since the Au peak is the same for all samples. In the following
table 4.3.3 the measured values of the layer thickness are listed. The Gd and Ni
layer appear thicker than expected. There are two possible reasons for this: First
the calibration via the Au layer thickness is not exact, i. e. that if it is only
15 Å the derived values for the other layers decrease, too. Secondly the rough
interface between the two materials leads to an apparent extension, respectively,
broadening of the RBS signal.
Chapter 4. Experimental Results 60

Figure 4.13: Comparison of RBS spectra: a) Deposition at T = 300 K,

dependence on substrate. b) Deposition at T = 100 K dependence on substrate.

In figure 4.13 the results of the RBS experiments on four Gd/Ni-bilayers are
shown. In the graph on the top, labeled a) the bilayers deposited at room tem-
perature are compared for differences between the two substrates and in b) the
ones that were produced at 100 K. What all the scans have in common is the
asymmetric shape of the peaks. The peak coming from the Gd-layer is slightly
flatter on its side to higher energies, whereas the one originating from the Ni-layer
is spread out more to the edge towards lower energies. This means that the inter-
face between the to materials is broader than the interfaces between the substrate
and the Gd-layer and the Ni-layer and the Au capping layer. This means that
the Gd/Ni interface is not sharp but spread. Just as in the case of the results
from AES in figure 4.11 in the previous section two explanations are possible:
nickel enters the Gd layer and intermixing to a GdNi alloy takes place or the
interface is very rough and the later deposited nickel fills up the deepenings first,
so that there is no intermixing on an atomic scale, forming any GdNi alloy. Link-
ing these results to the later presented STM images favors the second explanation.
61 4.3. Structure and morphology of the Gd/Ni-bilayer system

In addition, the measurements by SQUID on samples with a varying Gd layer

thickness also support the idea of an ”apparent” intermixing by the large surface
roughness. Although the fact, that the asymmetry of the RBS-signal of the Ni
layer is larger than the one of the Gd layer, indicates that some nickel entered
the gadolinium film.

Sample Substrate Tdep Gd [Å] Ni [Å] Au [Å]

A5 Al2 O3 100 K 76 114 20
A5 Si3 N4 100 K 62 101 20
A9 Al2 O3 300 K 53 113 20
A9 Si3 N4 300 K 62 115 20

Table 4.1: Calculated layer thicknesses from RBS spectra.

Several TEM experiments were carried out by Richard VanFleet at Brigham

Young University in Provo, Utah. He investigated two bilayers consisting both of
Si3 N4 /Gd(50Å)/Ni(50Å)/Au(20Å). The one labeled A1 was deposited at 100 K,
the other, labeled A7, at 300 K substrate temperature. Because this technique
requires to cut a cross-sectional thin slice out of the sample, only measurements on
system on Si3 N4 substrates were performed. Figure 4.14 shows in the left part a
cross-sectional view through the samples. One can only see the boundary between
the Ni and the Au layer. The brightness changes too little going from the Gd to
the Ni layer. In the right part of the figures the chemical composition is displayed
on the scanning depth following the indicated path in the TEM image, using an
EDX (Energy Dispersive X-ray Analysis) detection system simultaneously.
Again in the case of low deposition temperatures the different elements seem to
be distributed over a broader area in the EDX graphs. This is in contradiction
with the expected lower mobility of the arriving atoms during evaporation, as
predicted by the nucleation theory. A possible explanation for the obtained data
is again that due to higher roughness of the Gd layer the measurement averages
over an area much larger than the typical distances in the height profile of the
Gd film. Like in the Auger experiments the overlap can be an artefact of the very
rough interface and that the compositional analysis of the bilayer is done on a
larger area, broadening the signals.
Additional TEM experiments would be very helpful, especially if it would be
possible to get images of a cross section at higher resolution, resolving the layered
structure. Thicker layers like depositing e. g. 200 Å Gd should show the same
compositional overlap. If not this would indicate an intermixing process.
Chapter 4. Experimental Results 62

Figure 4.14: TEM cross-sections: Comparing two samples

Si3 N4 /Gd(50Å)/Ni(50Å)/Au(20Å), deposited at two different temperatures
(A1: 100 K; A7: 300 K).

4.3.4 Investigating the growth of Gd on Al2 O3 and Si3 N4

by STM
All described results about the structural properties so far were by methods that
average over large areas, like XRD, AES and RBS, i.e. from several µm to mm.
This is either due to a minimum beam diameter of the incident particles or caused
by the incident angle. The images taken by TEM show structures on a smaller
length scale, but the interface still appears as ambiguous area, that can be gener-
ated by intermixing but also by interface roughness. Especially the results from
AES measurements during deposition would support the idea of a chemically
mixed GdNi layer.
63 4.3. Structure and morphology of the Gd/Ni-bilayer system

In the following paragraph, a series of STM images are shown that were taken
from several pure Gd films without any capping layer. The samples were trans-
ported under UHV conditions to another UHV chamber within a few minutes.
A special transport manipulator with a new sample adapter was constructed in
order to fit the two different holder systems of the UHV set-ups. During the
transport the pressure didn’t rise above 1·10−7 mbar, what is believed to be suf-
ficient, in order not to degenerate the gadolinium due to oxidation. In air even
much thicker layers would oxidize completely within a very short time.

Figure 4.15: A44: 50 Å Gd on Al2 O3 , Tdep = 100 K:

a) 1µm x 1µm STM image and height scan below, along the indicated line; b)
50nm x 50nm STM image and the corresponding height profile.

The first sample investigated by STM, shown in figure 4.15 and labeled A44,
is a 50 Å thick Gd film on a Al2 O3 substrate. The deposition temperature was
100 K. Picture a) is a scan of an area of 1 by 1 µm. Below the height profile along
the indicated path is displayed. As one can see on this image, the thin film exhibits
a very rough surface. The line scan reveals height differences around 30 Å on
Chapter 4. Experimental Results 64

distances of several 100 nm. In part b) on the right an image of 50 by 50 nm2 is

shown. On this scale the nanocrystalline structure of the thin film can be seen.
In the height profile below several grains are mapped. They have a diameter in
the range of 1 nm, but there is a large variety of grain sizes. As investigated by
[73], summarized in figure 4.8, this leads to a significant decrease in the Curie
temperature. It is believed that the average grain size is the leading effect for the
different magnetic behavior between the bilayers on the two different substrates.
For comparison in figure 4.16 STM images from film of 50 Å Gd on Si3 N4 , that
was also deposited at Tdep = 100 K, are shown.

Figure 4.16: A46: 50 Å Gd on Si3 N4 , Tdep = 100 K:

a) 500 nm x 500 nm STM image and height scan below, along the indicated line;
b) 200nm x 200nm STM image.

The left image a) in figure 4.16 is taken on a 500 by 500 nm2 square. Again
as above there are very large height differences. But it has to be kept in mind
that the substrate Si3 N4 itself has a higher surface roughness compared to Al2 O3 .
This is due to the contamination with large particles that are left from the mem-
brane processing, as it is described in section 4.1 and especially figure 4.2. But
the irregular structure is comparable to the observations before on Al2 O3 in fig-
ure 4.15. But on shorter length scales significant differences can be seen. The
65 4.3. Structure and morphology of the Gd/Ni-bilayer system

film is grown in a granular structure, comparable to the sample on Al2 O3 but

the grains are larger. The line scan below the picture, labeled b), shall give an
estimate of the average grain size of approximately 10 nm. For this diameter the
expected Curie temperature of 250 K, predicted from figure 4.8, gets closer to the
observed one of 150 K in figure 4.24. The remaining difference is attributed to
the fact that the film is only 50 Å thick. Higher mobility of the incoming atoms
and a different ratio of the interaction energies between substrate, adatoms and
among the film material, as described in section 2.4, leads to this changed growth
pattern.

Figure 4.17: A47: 50 Å Gd on Al2 O3 , Tdep = 300 K: The imgage taken

has shows an area of 200 nm by 200 nm.

The influence of the substrate temperature was examined, too and will be
connected to the differences in the magnetic behavior in section 4.4.6. Two exam-
ple images of a 50 Å thick Gd film on Al2 O3 (figure 4.17) and on Si3 N4 (figure 4.18)
are listed in the following. The growth of the material differs again in the aver-
age grain size and in the surface roughness between the two substrates, but the
differences to the samples deposited at a lower temperature of 100 K are not as
significant. For Al2 O3 the granular structure is nearly the same, what leads to
nearly the same TC of the Gd layer in both series, compared in section 4.4.6. But
larger height differences, i.e. larger surface roughness can be seen. This fits nicely
to the model in section 4.5 of a bilayer with rough interface but low or even no
chemical intermixing of the two materials.
In figure 4.18 a scan on 3 µm by 3 µm large area on a Gd film on Si3 N4 is shown.
The film was deposited at room temperature. Again a granular film is detected
with a slightly increased surface roughness, compared the low temperature sam-
ple. In the middle part one can see a somehow blurry area. This is a charging
Chapter 4. Experimental Results 66

Figure 4.18: A50: 50 Å Gd on Si3 N4 , Tdep = 300 K: The scanned section

is a 3 µm large square. In the inner part a slightly blurry area can be seen, which
is an artefact of a charging effect by the STM, maybe caused by partial oxidation.
This area was scanned previously.

effect. This area was scanned before the scanning scale was increased. This charg-
ing effect denotes that a part of the tunneling current remained. A reason for this
might be a partial oxidation of the surface or the resistance of the sample was too
high, because of the insulating substrate. The film had to be connected by silver
glue to the substrate holder. Therefore taking images at a much smaller scale like
100 nm failed. SQUID measurement later will show that the Curie temperature
is nearly the same for the samples, produced at 300 K, than for the ones at 100 K.
So the grain size distribution is expected to be equivalent. The higher surface
roughness leads to the observed reduction of the saturation magnetization.
For comparison at the end of this section STM images of a complete bilayer
system capped with a 20 Å Au film are presented in figure 4.19. The film com-
position was Gd(50Å)/Ni(100Å)/Au(20Å), grown on Al2 O3 at Tdep = 100 K. In
this case the observed surface morphology is completely different. The surface
consists of much larger grains of some 10 nm in diameter. Even on a large scan
range as in the left picture (500 nm x 500 nm) the film is very smooth and only
small height differences of less than 20 Å are observed. The pure Gd film, which
was produced under the same conditions, shown in figure 4.15, however has a
much rougher surface. This supports the idea that the later deposited nickel and
gold first fills up the holes and cracks and by that smoothes the surface. Espe-
cially gold is known to do so, but since the deposited Au film is much thinner
than the one of Ni the flattening must be caused by the thicker Ni layer.
67 4.4. Magnetic characterization of Gd/Ni-bilayers

Figure 4.19: Al2 O3 /Gd(50Å)/Ni(100Å)/Au(20Å): The surface of a com-

plete bilayer sample is very smooth and the observed grains have a size of about
10 nm.

4.4 Magnetic characterization of Gd/Ni-bilayers

The magnetic behavior of the produced Gd/Ni-bilayers was investigated by SQUID
and XMCD. SQUID measurements could be done on samples on both sub-
strates, whereas XMCD is limited to the Si3 N4 membranes. With both meth-
ods the field and temperature dependence was examined thoroughly to get a
complete picture of the magnetism in the single layers and their interaction.
In the first part, field dependent measurements on a selected bilayer system
Gd(50Å)/Ni(75Å)/Au(20Å) will be presented. The second part deals with the
temperature dependence of the magnetic behavior and the interlayer exchange
coupling. The way the magnetic characterization and the insights on the mor-
phology and structure of the bilayers are connected, is debated in the next sec-
tion 4.5.
SQUID is the most sensitive device for measuring magnetic moments (see sec-
tion 3.2.1). The advantageous features were namely the accurate temperature
control and the large field range. On the other hand the detected signal included
contributions from the substrate and the mounting straw. They can be sub-
tracted easily by assuming them to be linearly proportional to the external field.
But still the obtained data is the net magnetization of the whole bilayer system.
The behavior of the single layer has to be deduced from the observed SQUID
data, starting from a model of the system. With the element specific XMCD
technique the different contributions can be separated.
Chapter 4. Experimental Results 68

4.4.1 In-plane anisotropy in Gd/Ni bilayers

The bilayer systems, consisting of gadolinium and nickel have their easy-axis of
magnetization within the film plane. It is driven by the shape anisotropy. At
temperatures above the Curie point of the Gd layer one observes an easy- and
hard-axis loop of the Ni layer, if its thickness is sufficiently large, i.e. dNi > 50
Å, to build up a closed film (see figure 4.20). The saturation magnetization of
both orientations match very good. For lower Ni film thicknesses the anisotropy
decreases, what can be explained by the fact that the material gets distributed
into smaller portions.

Figure 4.20: Magnetic anisotropy: With no magnetic moment coming from

the Gd layer at T = 300 K. The thin Ni film shows the expected in-plane
anisotropy.

For all temperatures below the Curie temperature of the gadolinium layer,
the anisotropy becomes stronger, so that fields of up to 10 kOe are by far too
small to saturate the magnetization perpendicular to the sample surface.
This strong in-plane anisotropy of thin Gd films is reported e.g. in [85]. Mea-
surements with a maximum field of 50 kOe achieve the same value for Msat for
both orientations. These M(H) measurements in figure 4.21 show the strongly
pronounced in-plane easy axis of magnetization.
69 4.4. Magnetic characterization of Gd/Ni-bilayers

Figure 4.21: At 4 K: In the Gd aligned the system has its easy-axis in the film
plane, too. The significantly lower MS for the hard-axis loop is due to the large
anisotropy.

4.4.2 Field dependent magnetization measurements on the

example bilayer system: Gd(50Å)/Ni(75Å)/Au(20Å)
The field dependence of the magnetization was determined at several distinct tem-
peratures. The Gd/Ni bilayers show a very interesting coupling phenomena, that
strongly depends on the measuring temperature. The Ni layer possesses a rather
constant saturation magnetization over the whole temperature range of interest
between 4 K and 300 K, what is far below its Curie temperature of 631 K. Its coer-
civity varies only slightly as shown in section 4.4.4. The magnetic response of the
underlying gadolinium layer in contrast varies strongly on the temperature, grain
size and film thickness. It possesses a high magnetic moment, which rises fast
with decreasing temperature. But due to the small coercivity the magnetostatic
energy MS ·HC is smaller than the one of the Ni layer. This situation is called the
”Nickel aligned state”. It expresses the fact, that in this intermediate temperature
range the magnetic moment of the Ni layer aligns parallel to an external field Hext
and the one of the Gd layer is oriented antiparallel to it for small fields. These
different properties lead to interesting hysteresis loops, that are characterized by
a negative remanence, i.e. the remanent magnetization points opposite to the pre-
viously applied external field and Mrem,Gd > Mrem,N i , but HC,Gd MGd < HC,Gd MGd .
A comprehensive analytical model for such an exchange coupled bilayer system
is described in [86] for a DyFe2 /YFe2 superlattice. Figure 4.22 shows a typi-
cal hysteresis loop from the bilayer sample Al2 O3 /Gd(50Å)/Ni(75Å)/Au(20Å),
deposited at T = 100 K and labeled A4, which exhibits the described negative
remanence. The hysteresis loop was recorded at T = 45 K and the external mag-
netic field pointed along the film plane, called in-plane loop. A linear off-set was
Chapter 4. Experimental Results 70

subtracted as background signal. Arrows were added to indicate the evolution of

the measured magnetization.
Starting at high positive fields both layers are aligned in field direction. If the
field is reduced the Gd begins to reverse its magnetic moment due to the an-
tiferromagnetic exchange coupling with the Ni layer at about 150 Oe. The net
magnetization decreases with decreasing external field and reaches a negative
value of 15 % of MS . In order to saturate the moment of the Gd layer completely,
a negative field of -50 Oe has to be applied. Raising the external field to higher
negative values, switches the nickel layer at 250 Oe. Now the magnetic moment
of both layers point again parallel to the external field. The same occurs if one
starts at high negative fields and a closed hysteresis loop is formed.

Figure 4.22: Al2 O3 /Gd(50Å)/Ni(75Å)/Au(20Å) deposited at 100 K: The

high magnetic moment of the Gd layer aligns opposite to the external field in
remanence, overruling the one from the Ni layer. The effective magnetization is
antiparallel to the magnetizing field.

The Ni aligned state exists in the temperature range where T < TC (Gd) and
the nickel layer takes up more magnetostatic energy, due to its higher coercivity.
The shape of the hysteresis changes completely for temperatures below the so
called compensation temperature, which is around 50 K (see figure 4.25). But
first the hysteresis at three characteristic temperatures above Tcomp are presented
in figure 4.23 of the bilayer system, already introduced in figure 4.22. In the left
71 4.4. Magnetic characterization of Gd/Ni-bilayers

Figure 4.23: In-plane hysteresis loops in the Ni aligned state: The mag-
netization at different temperatures of A4: Gd(50Å)/Ni(75Å)/Au(20Å) grown on
Al2 O3 (left) and on Si3 N4 (right).
Chapter 4. Experimental Results 72

column measurements on a sample deposited on Al2 O3 are shown. The right

column displays the results, if Si3 N4 was used as a substrate.
At 300 K the Gd is nonmagnetic and the observed hysteresis is a simple square
loop, that shows the magnetization of the Ni layer. The coercivity of 20 Oe is
very small, as expected for a thin film with the easy-axis parallel to the external
field. There are hardly any differences between the two substrates. Measure-
ments at 65 K differ significantly. While on Al2 O3 the Gd is still nonmagnetic
and a usual easy-axis loop is recorded with an increased coercivity compared to
at 300 K, the bilayer on Si3 N4 goes through a overcompensated hysteresis. Here,
65 K is already below the Curie temperature of the Gd layer, although it possesses
the same thickness in both samples. The gadolinium magnetization is switched
sharply antiparallel to the external field and by that to the moment of the nickel
layer at 10 Oe and a negative remanent moment of M(0)/MS ≈ -0.2 is observed.
With increasing negative field the magnetic moment does not change much until
H = -60 Oe, when the Ni layer is switched by the external field. This means that
the Gd layer is switched completely, despite the reorienting at zero field shown
in figure 4.22. The switching of the nickel moment takes place in a narrow field
range between 60 to 100 Oe, too.
At 45 K the situation in both samples is similar. In both cases the gadolin-
ium is ferromagnetic and its moment is reoriented antiparallel to the magnetizing
external field in remanence due to the exchange coupling with the nickel layer.
The magnetic properties of the nickel layer seem to be similar in both cases.
Still, there are significant differences between the two hysteresis loops. Again the
switching is much sharper for both layers and takes place at lower field values if
Si3 N4 is used as substrate. On Al2 O3 the nickel layer reorients into the external
field direction at -300 Oe, starting from high positive field values, and on Si3 N4
at -200 Oe. The negative remanence is M(0)/MS ≈ -0.35, compared to -0.12 on
Al2 O3 . But the gadolinium is not completely switched here, in contrast to the
system deposited on Si3 N4 .

4.4.3 Temperature dependence of the magnetic properties

of the example system Gd(50Å)/Ni(75Å)/Au(20Å)
The hysteresis loops already have shown that the Gd/Ni bilayer system exhibits
some peculiar magnetic properties and that there are significant differences be-
tween the samples grown on Al2 O3 and on Si3 N4 .
In figure 4.24 measurements of the remanence MR and of the saturation mag-
netization MS are shown. For obtaining the remanence the sample was saturated
in a +10 kOe field after adjusting the temperature. Then, the external field was
shut off and the remanent magnetization could be measured. It turned out, that
special care had to be taken, in order not to switch the Gd layer, which possesses
only a very small coercivity in the Ni aligned state, by setting the field to zero in
the so called oscillate mode of the SQUID. Small inserts, showing the magnetic
73 4.4. Magnetic characterization of Gd/Ni-bilayers

Figure 4.24: MR (T) and MS (T): In both graphs it can be seen, that the Gd
layer deposited on Si3 N4 becomes ferromagnetic at higher temperatures.

moments of the layers in the remanent state were added to visualize the interlayer
exchange coupling.
The remanence starts to decrease for the sample on Si3 N4 already at 100 K and
goes through zero at approximately 77 K. This point is called the compensation
Chapter 4. Experimental Results 74

temperature Tcomp . It reaches a minimum at the so called overcompensation tem-

perature Toc = 45 K. If the temperature is decreased even further the remanence
jumps directly to positive values. In this temperature regime the Gd layer is
aligned in field direction and the magnetic moment of the Ni layer couples anti-
ferromagnetically to it. This is called the Gd aligned state, which will be further
discussed in the next paragraph. The bilayer, deposited on Al2 O3 , has a nearly
constant remanence, generated by the Ni layer, down to 60 K. Then there is a
sharp drop to negative values, because the Gd layer becomes ferromagnetic. The
minimum is at 50 K, but at 45 K the remanence is still negative, so this should
be interpreted as the compensation temperature, because the transition from the
Ni to the Gd aligned state is not gradual, but switches theoretically at a distinct
temperature. The smaller value at 45 K can be explained by the incomplete
switching of the Gd moment as shown in figure 4.23.
The ferromagnetic ordering in the Gd layer can be seen at best in measuring the
saturation magnetization. For this, the magnetic moment in a constant external
field of 10 kOe was measured. Both curves were shifted to match the value of MS
at T = 300 K, which was deduced from the hysteresis loops at 300 K. By that
the background contribution, which was assumed to be constant over the investi-
gated temperature range, can be subtracted. The magnetic moment of the Si3 N4
sample starts to rise slowly already at 150 K and increases strongly with decreas-
ing temperature below 100 K. As observed in the remanence measurements, the
magnetic moment of the Gd layer builds up at much lower temperatures, if Al2 O3
is used as a substrate. MS starts to increase fast at 75 K and reaches the values
of the bilayer on Si3 N4 at the overcompensation temperature of 45 K. This shows
that in both cases the total magnetic moment of the Gd layer in both samples
is the same and there is no reduction of it due to intermixing, which may cause
quenching of the Ni moments or reduce the coupling strength between the single
Gd atoms, since the RKKY ordering mechanism is so sensitive on the interatomic
distance (see section 2.2.2.2). The differences in the magnetic properties must be
evoked by diverse conditions in the Gd layer, that have their origin in the different
epitaxial growth on the two substrates, that were investigated in section 4.3.
Figure 4.25 displays the hysteresis loops of both samples at 4 K, which is
in the Gd aligned state. In the top row there is a zoom of the inner part from
-400 Oe to +400 Oe. The graphs below show the full measured field range from
-10 kOe to +10 kOe. Reducing the external field from high positive values a kink
is observed in both cases, which is attributed to the switching of the Ni moment.
The step height matches very well the saturation magnetization of the Ni layer,
which was obtained from the measurement at 300 K. On Al2 O3 the Ni moment
is switched between 5 kOe and 2 kOe, compared to the range between 4 kOe to
1 kOe in the case of Si3 N4 . Since the total magnetic moments of the Ni and Gd
layers seem to be the same for both samples, as can be concluded from the mea-
surement of MS in figure 4.23, the different coupling strength can be attributed to
the different morphology of the Gd-Ni interface. The larger surface area, if Al2 O3
is used as a substrate, couples the Ni moment stronger to the leading moment of
the Gd layer (see section 4.3). In the magnetization reversal at low fields, shown
75 4.4. Magnetic characterization of Gd/Ni-bilayers

Figure 4.25: Hysteresis loops in the Gd aligned state: For temperatures

below the overcompensation temperature Tov = 45 K the Gd moment points
parallel to the external field and the magnetic moment of the Ni layer is coupled
antiferromagnatically to it.

in the top row, only a square loop is observed. This leads to the assumption that
the moment of the Ni layer reverses coherently and antiferromagnetically coupled
with the Gd moment. By that the Ni layer undergoes a very peculiar hysteresis
loop with three switching events in the full field range. The coercivity of the
bilayer on Si3 N4 is slightly bigger than for the Al2 O3 sample, but still amounts
to only 25 Oe (Si3 N4 ) and 15 Oe (Al2 O3 ).

S. Esho measured the switching of the transition metal Co in bilayer systems

of Gd/Co by MOKE [44]. In order to separate the Kerr rotations of the two mate-
rials, the laser beam was coming in at a small angle, what reduces the penetration
depth. This was done for the front (Gd on top) and the back (Ni layer) sides.
The results are shown in figure 4.26 together with calculated loops from [40]. Un-
fortunately such measurements could not be performed on the produced samples,
because the used XMCD set-up could neither generate fields above 1000 Oe nor
could the sample be kept at a temperature below the compensation point. And
no MOKE system, which can operate at temperatures below 45 K, was available
to investigate the magnetization of the single layers in the Gd-aligned state as in
Chapter 4. Experimental Results 76

[44].

Figure 4.26: MOKE measurements on the TM layer in Gd/TM bilayer

systems: a) Gd/Co bilayer [44]. b) T. Kobayashi calculated the magnetiza-
tion over H for Fe, but the contribution of the Gd layer below to the MOKE
signal could not be ruled out completely and only loops of the whole system were
recorded like in figure 4.25.

As mentioned in section 4.2, the Gd/Au samples degenerated due to fast

intermixing. Even a strong reduction in MS (T) was observed for these bilay-
ers. Investigations on the Gd/Ni/Au layered systems did not show any changes
in time concerning their magnetic behavior. In figure 4.27 a selection of four
hysteresis loops of the Al2 O3 /Gd(50Å)/Ni(75Å)/Au(20Å) bilayer system is pre-
sented. There are only very small changes between the measurements, that were
done right after the preparation of the samples and one year later. It seems the
total magnetic moment is slightly reduced during that time, but the decrease is
only a few percent.

4.4.4 Element-specific hysteresis loops of an example Gd/Ni

bilayer system
XMCD measurements made it possible to observe the field dependency of the
magnetization for each element in the Gd/Ni bilayers separately. These exper-
iments, carried out at BESSY in Berlin, were only done on Si3 N4 membranes,
because it is more easy to measure the transmitted beam going through a thin
membrane, than detecting the field dependent absorption of reflected X-rays, as
it would be necessary for the Al2 O3 samples.
There are only few data, which were obtained in a measurement performed on
the reflected beam under a small incident angle of the X-rays. Nominally the
sample holder could be cooled down to liquid helium temperatures, but the long
distance from the cooled part of the manipulator to the actual sample holder
and the heating of the membrane by the high intensity X-ray beam, made it
77 4.4. Magnetic characterization of Gd/Ni-bilayers

Figure 4.27: Comparison of hysteresis loops of sample A4

Al2 O3 /Gd(50Å)/Ni(75Å)/Au(20Å): The magnetic behavior does not
change due to intermixing after storing the samples for one year. a) and b)
field ranging ±10 kOe; c) Zoom of central part.

impossible to investigate hysteresis loops in the Gd aligned state, i.e. that the
actual sample temperature was always above 45 K. The maximum magnetic field,
that could be applied, was 1000 Oe and would not be sufficient to orient both
layers into the external field direction (see figure 4.23) below the compensation
temperature. But the expected antiferromagnetic coupling of the Gd moment to
the leading magnetic moment of the Ni layer was affirmed. The hysteresis loops
of the two layers at the minimum achievable temperature are shown in figure 4.28.
Chapter 4. Experimental Results 78

Usually before recording a hysteresis loop by varying the external magnetic

field, the maximum in the XMCD spectrum was determined. An example for
such a scan of the X-ray absorption on the photon energy is shown in figure 3.7.
For each energy the absorption was measured for the maximum positive field (µ+ ,
graph at the top) and maximum negative field (µ− , graph in the middle). The
normalized difference is the XMCD signal, ∆µ, shown in the plot in the bottom
row of figure 3.7.

The actual temperature of the illuminated and probed part of the sample can
be concluded from comparison of the loops in figure 4.28 and SQUID measure-
ments at different temperatures. Especially the switching field of the Ni moment
is a suitable property for this. By that the temperature range can be narrowed
down to 60-70 K.
In the top row of figure 4.28 the assumed composition is sketched. In order to
model the obtained hysteresis loops in the bottom row the thin film was divided
in three parts: A pure Gd film at the bottom, a pure Ni film on top, and an inter-
mediate Gd monolayer, that is polarized by the adjoint magnetic moment of the
nickel. In the middle schematic M(H) graphs are plotted to illustrate the antifer-
romagnetic coupling and the polarization of the Gd boundary layer. Arrows were
added to the curves, illustrating the devolution of the magnetizations. The M(H)
measurements were performed sequentially at the two characteristic energies of
the L3 -edge of Ni and the M5 -edge of Gd. The signal height at each energy is
a real measure of the magnetic moment, but it can not be taken for comparison
between the two loops, since it depends on the X-ray absorption properties of the
different specimen.
The Ni layer undergoes a typical ferromagnetic hysteresis with a coercivity of
approximately 100 Oe. It switches completely in a narrow field range and the
magnetization then saturates.
The XMCD signal of the Gd layer is completely different. Decreasing the exter-
nal field from a start at a high positive field, the magnetization switches sharply
at +10 Oe. This reorientation before zero field is due to the antiferromagnetic
coupling to the Ni layer, as described in sections 4.4.2 and 4.4.3. But in this
measurement an additional feature, which is not accessible by SQUID, could be
observed. If the external field is risen again to larger negative values, the XMCD
signal of Gd decreases about 10% just at the point where the Ni is switched. This
corresponds in a 50 Å film to approximately one monolayer of gadolinium. The
interface layer in the Gd film is polarized by the neighboring Ni. This effect was
reported in other rare earth/transition metal systems [23, 87, 88]. It is magne-
tized antiparallel to the Ni layer at all times, whereas the remaining part of the
Gd layer couples antiparallel only for small fields. If the external field is large
enough, its magnetic moment points in the field direction. By that, in the case of
large fields Ni and Gd are magnetized by the external field and the the interface
layer is polarized in the opposite direction. The observed total magnetic Gd mo-
ment is smaller as in the remanent state. In remanence the magnetic moment of
the two thick layers are antiparallel and so the magnetic moment of the interface
79 4.4. Magnetic characterization of Gd/Ni-bilayers

Figure 4.28: Hysteresis loops for Gd and Ni separated: Sketches of the

alignment of the magnetic moments of each layer (top); Schematic of the hystere-
sis loops (middle row); measured loops of a Si3 N4 /Gd(50Å)/Ni(75Å)/Au(20Å)
bilayer.
Chapter 4. Experimental Results 80

layer and the Gd layer are parallel and the maximum XMCD signal is obtained
at this point.

Figure 4.29: XMCD at T = 25 K: If measured in reflection geometry the

switching of the interface layer is strongly pronounced in the XMCD signal. The
measurement at the Ni-edge at Eγ = 853 eV shows a ferromagnetic hysteresis
with nearly 100 % remanence.

In addition to measurements of the XMCD signal in transmission, experi-

ments under a small incident angle (between 10° to 15°) were performed. Ab-
sorption in the reflected beam is thereby detected. In this set-up the observed
X-ray intensity is influenced by different contribution, not only the absorption as
before, if the transmitted intensity is explored. A complicated interplay between
81 4.4. Magnetic characterization of Gd/Ni-bilayers

a strong depth sensitivity, structural properties, Bragg reflections and the XMCD
absorption give hysteresis loops as shown at the bottom in figure 4.29. Careful
alignment leads to a pronounced contribution of the interface layer of gadolinium,
which is polarized by the Ni layer above. The course of the hysteresis loops is
illustrated by arrows. The sample was cooled down to nominally 25 K, but again
the high intensity of the x-ray beam and thermal radiation from the surround-
ing falsified the measured value. In this beam alignment the deviation seems to
be smaller, because one is not limited to the thin membrane with a poor heat
conductance and small heat capacity. So the transferred heat dissipates faster.
The shape of all loops is still as expected for the Ni aligned state, that lies in
the temperature regime above 45 K up to the Curie point of thegadolinium layer
below.
The experiment on the Ni layer gives a usual ferromagnetic square loop and is
presented in the upper part of figure 4.29. The remanence is nearly 100 % and
the coercivity 100 Oe. The hysteresis loop at the Gd-edge, using again a incident
X-ray energy of Eγ = 1153 eV, is of a completely different shape. Starting from
a high negative magnetic field, the XMCD signal remains constant just before
Hext reaches zero, it increases at 10 Oe. This change is caused by the switching
of the magnetic moment of the Gd film. It aligns antiparallel to the magnetic
moment of the Ni layer, which is still oriented in field direction, due to the ex-
change coupling between the layers. In addition there is a larger change of the
reflected X-ray intensity at ±100 Oe. It is contributed to the magnetic reversal of
the interface Gd layer, which is coupled antiferromagnetically to the Ni moment,
since this change in the XMCD signal coincides with the switching field of Ni
layer.
The development of the field dependence of the magnetization is decipted in
figure 4.30. The temperatures given in the plot are the nominal values. At 36 K
the contribution of the Gd film is clearly reduced compared to the measurement
at 25 K, shown left of it in the first row. This is evident to the rapid reduction in
MS of the gadolinium with increasing temperature, see figure 4.24. At 63 K there
is hardly any Gd film contributions visible to the x-ray absorption. The jumps
attributed to the polarized interface layer are still visible. Even at nominally
100 K this behavior can be observed. The magnetic reversal takes place at an
external magnetic field of 60 Oe, what corresponds to the switching field of the Ni
layer at the same temperature. For comparison of the switching fields, the hys-
teresis loops at 63 K and 100 K are displayed in figure 4.31. They were recorded
immediately after or before the measurements at the Gd edge in figure 4.30. The
coercivity decreases slightly as expected for a ferromagnetic material [30, 89] and
the values match well with the observed Gd switching events in figure 4.30. The
deviation of the saturation magnetization at the starting point and at the ending
point of the measurement at -200 Oe is probably caused by the decrease in beam
intensity during the measurement, what becomes more important in the case of
very small magnetic signals.
Chapter 4. Experimental Results 82

Figure 4.30: XMCD at the Gd-M5 -edge: The polarization of the Gd inter-
face layer appears even at temperatures above the Curie point. The indicated
temperatures are only the nominal values. The XMCD signal was normalized to
a saturation value of 1.

Figure 4.31: XMCD at the Ni L3 edge in reflection: With the coercivity

of the Ni layer decreasing, the switching of the Gd interface (see figure 4.30 layer
shifts to lower fields).
83 4.4. Magnetic characterization of Gd/Ni-bilayers

4.4.5 Dependence on the nickel layer thickness

To study the magnetic interaction between the two materials in the Gd/Ni bi-
layer system the thickness of the nickel layer was varied between 15 Å to 100 Å,
whereas the gadolinium layer was kept to 50Å. In addition a few samples with
thicker Gd layers were produced, that will be presented at the end of this section.

The first figure in this section, figure 4.32, shows a series of hysteresis loops
at 45 K, which is the compensation point of the bilayered samples of Gd/Ni with
a gadolinium layer of 50 Å, so there is a minimum in the remanent magnetization.
Al2 O3 was used as a substrate and the deposition temperature was 100 K. In the
right column measurements over the full investigated field range from -10 kOe to
+10 kOe are listed. The left column shows the inner part of the hysteresis loops
in a range from -400 Oe to +400 Oe. As one can see overcompensated loops
with negative remanent magnetization only appear in samples with a nickel layer
thickness of 75 Å and 100 Å. Another feature is that the shape changes strongly
with increasing nickel film thickness. At low coverages, especially 15 Å and 25 Å,
the bilayer saturates at much higher fields. A possible explantation for this is,
that the gadolinium layer is not fully covered by nickel and that due to alloying
with the Au capping layer magnetic moment, coming from the gadolinium is lost.
This goes along with the observed decrease in saturation magnetization, that is
beyond the decrease caused by a smaller nickel content. It must be mentioned
at this point that the scale of the magnetization axis in figure 4.32 is always the
same in the left column but differs from one row to the other in the right column.
The bilayers containing nickel layers of 75 Å and 100 Å saturate in an external
field of 400 Oe already completely. In the case of 100 Å of nickel the magnetic
moment of the Gd switches nearly completely at remanence, leading to a higher
negative remanent magnetization compared to the bilayer with a 75 Å thick Ni
layer.
In figure 4.33 the remanent magnetization of GdNi bilayers grown on Al2 O3
are shown for different nickel layer thicknesses. The Gd layer is kept constant to
50 Å. The two samples containing a Ni layer of 15 Å and 25 Å possess nearly
no remanent magnetization above 50 K. Below that temperature the Gd layer
becomes ferromagnetic. The low remanence can be seen in the hysteresis loops
in figure 4.32 for these two samples. The large difference below 50 K is another
indication that due to incomplete covering of the Gd by the Ni, intermixing of Gd
with the capping material Au takes place reducing the ferromagnetic portion. In
the measurements on the three other bilayers the magnetostatic interaction be-
tween the two layers can be observed. The small rise in Mrem (T) directly above
50 K in the graph of the 50 Å sample is an artefact of the measurement. When
reducing the external field from 50 kOe to 0 Oe in the so called oscillate mode a
small negative is applied in the field adjustment, which may switch the magneti-
zation if the coercivity is very small. In the case of 75 Å and 100 Å compensation
is observed with a minimum in the remanence at 50 K and 45 K, respectively.
Chapter 4. Experimental Results 84

Figure 4.32: Hysteresis loops for different Ni film thickness on Al2 O3 :

The series is taken on samples grown on Al2 O3 at a measuring temperature of
45 K. The left column is a zoom of the inner part of the full loops, which are
shown in the right column. Notice the different scale of the y-axis, displaying the
magnetic moment.
85 4.4. Magnetic characterization of Gd/Ni-bilayers

Figure 4.33: Mrem (T) for different Ni content on Al2 O3 : Compensation

can only be observed in the systems containing at least 50 Å of nickel. Negative
remanence appears even only in the two samples with 75 Å and 100 Å thick Ni
layers.

In the Gd-aligned state the remanent magnetization rises strongly. The antipar-
allel alignment of the two magnetic moments causes a smaller Mrem (T) of the
sample with 100 Å than of the one with 75 Å.
In figure 4.34 the same series of bilayers deposited also at 100 K but on Si3 N4
is presented. The results are similar to the ones obtained for the Al2 O3 -series.
But there are a few differences, that should be mentioned. The samples with the
thinnest nickel layers, 15 Å and 25 Å, show a much sharper switching behavior,
i.e. the magnetization saturates at much lower fields, compared to the graphs in
figure 4.32. The sample containing a 50 Å nickel layer shows a very particular
switching behavior. It seems as if this configuration is already in the Gd aligned
state. The inset in the left graph shows the hysteresis of the Ni layer at fields
between 300 Oe and 900 Oe. The origin of this switching behavior is still unclear,
since it is not present in the two following graphs of the samples with 75 Å and
100 Å thick nickel layers, neither in the M(H) measurements on the same bilayer
on Al2 O3 . The most obvious explanation would be that the Gd layer has more
than 50 Å, but there is no indication for that in the measurements on magnetism
on the sample on Al2 O3 . The Ni content is very accurate as can be seen in the
two graphs at the end of this section figures 4.36 and 4.36.
The two last rows show hysteresis loops with a negative remanence, due to the in-
Chapter 4. Experimental Results 86

Figure 4.34: Hysteresis loops for different Ni content: Si3 N4 was used as
a substrate at a deposition temperature of 100 K. The two columns are like in
figure 4.32.
87 4.4. Magnetic characterization of Gd/Ni-bilayers

teraction of the two adjoint magnetic layers. As already discussed in section 4.4.2
the switching is much sharper compared to Al2 O3 as a substrate and the bilayer
was produced under the same conditions. Interestingly the 100 Å thick nickel
layer saturates earlier (bottom row) than the one in the row above, the hysteresis
of the bilayer with 75 Å of Ni. This means that the magnetic moment of the Ni
layer has a lower coercivity in the 100 Å sample. The switching of the Gd layer
in contrast is the same for both samples, showing a reorientation opposite to the
external field at ±15 Oe. This magentizing process is nearly complete, i.e. nearly
the whole layer is switched.

Figure 4.35: Mrem (T) for different Ni content on Si3 N4 : Despite the
higher TC of the Gd layer compered to the Al2 O3 samples, the compensation
temperature is the same.

In the next figure 4.35 the remanent magnetization over temperature is shown
for the series of GdNi bilayers on Si3 N4 . Again the sample with a 50 Å nickel
layer stands out. There is only an increase in Mrem for temperatures below 100 K,
where the Gd layer becomes ferromagnetic (compare to figure 4.24). Only a small
dip can be seen around 50 K. The sample containing 15 Å of nickel only shows
an increase in Mrem , too, below the Curie temperature of the Gd layer. But
the bilayer with 25 Å of nickel already shows a compensation behavior around
TC,Gd , but its minimum is at 55 K. The bilayer systems with 75 Å and 100 Å of
nickel are again very similar. In the Ni aligned state between below 100 K the
remanence decreases gradually until reaching its minimum at 45 K and 40 K re-
spectively. Going to lower temperatures Mrem increases rapidly due to changing
Chapter 4. Experimental Results 88

to the gadolinium aligned state and the strongly increasing magnetic moment of
the Gd layer. Due to the antiparallel alignment of Ni and Gd Mrem is smaller for
the 100 Å sample than for the 75 Å one, as expected. Another interesting feature
is that the compensation temperature is nearly the same as for the bilayers on
Al2 O3 , although the Curie temperature is much higher if Si3 N4 is used as a sub-
strate. This means that the overall amount of the magnetic moments determines
the temperature where the compensation takes place.

Figure 4.36: Msat over Ni thickness for Al2 O3 samples: The saturation
magnetization at 300 K is plotted over the Ni layer thickness.

The last two figures 4.36 and 4.37 of this section show the saturation magne-
tization of the bilayers at 300 K over the nickel layer thickness. At this temper-
ature there is no ferromagnetic contribution from the Gd. But intermixing and
polarization of Gd moments at the interface reduce Msat at 300 K. If deposited
on Al2 O3 the extrapolation of the series intersects the x-axis at 8.7 Å. Whereas
for the series deposited on Si3 N4 it seems as if no magnetic moment of the Ni
layer is vanishing. But it must be stated at this point, that for a more reliable
extrapolation more data, i.e. more samples are needed.
The difference can originate from several effects, derived from the results of the
structural investigations in section 4.3 and especially the differences revealed by
STM in section 4.3.4.
First, the rougher interface between Gd and Ni on Al2 O3 leads to a larger num-
ber of antiparallel polarized gadolinium moments. Additionally the Gd grains are
smaller on Al2 O3 , what may promote intermixing of the two materials. The mag-
89 4.4. Magnetic characterization of Gd/Ni-bilayers

Figure 4.37: Msat over Ni thickness for Si3 N4 samples: There is nearly no
magnetic moment vanishing, if the bilayer is grown on Si3 N4 .

netic moment of the nickel atoms is nearly completely suppressed in Gdx Ni1−x
alloys [90], what may be an explanation of the reduced saturation magnetiza-
tion of the bilayers. This interconnects the results from STM with the magnetic
behavior of the bilayer systems on the two different substrates.

4.4.6 Influence of the substrate temperature

As described in section 4.1, a series of bilayers was produced on both substrates
at room temperature, but apart from that under the same conditions, like de-
position rate, base pressure and pretreatment of the substrates. In section 4.3.4
differences in the growth and film morphology observed by STM are described. In
the following these findings and results from SQUID measurements are brought
together. Again the observations will be presented for the example system of
Gd(50Å)/Ni(75Å)/Au(20Å) deposited on Al2 O3 and Si3 N4 . Deviations in other
samples are mentioned at the corresponding passage.

In figure 4.38 the temperature dependence of the remanent magnetization

Mrem (top row) and the saturation magnetization Msat (bottom row) are plot-
ted for four different samples, all consisting of the same layer composition. The
legend indicates the used substrate and the deposition temperature. From the
Msat -graph one can see that the Curie temperature of the gadolinium layer re-
mains nearly the same for both substrates. But in both cases of samples, produced
Chapter 4. Experimental Results 90

Figure 4.38: Comparison of M(T) for different deposition temperatures:

The remanence in the top graph shows a shift of the compensation temperature
for both used substrates of about -5 K. The lower graph shows that TC of Gd
seems unchanged but Msat is much lower if deposited at 300 K.

at room temperature, Msat does not rise as fast as for the ones at 100 K depo-
sition temperature. If Al2 O3 is used as a substrate the deviation of the final
values at 4 K is much larger, compared to the two samples on Si3 N4 . But also
above the Curie temperature of the Gd layer Mrem and especially Msat of the
room temperature samples are smaller than for the samples that were cooled to
100 K during deposition. This means that not only magnetic moment from the
Gd layer is missing but from the Ni layer, too. The fact that the differences
between Tdep 100 K and 300 K are much larger for Al2 O3 samples than for Si3 N4
weakens the explanation of the reduction due to stronger oxidation. Besides, the
pressure during deposition was only slightly higher if the samples were not cooled,
3·10−9 mbar instead of 1·10−9 mbar. In addition this could hardly be responsible
for the lower Msat at high temperatures. In section 4.3.4 the STM pictures re-
91 4.4. Magnetic characterization of Gd/Ni-bilayers

vealed an increased roughness of the Gd film of the room temperature samples,

but stronger in the case of Al2 O3 substrates. This higher interface roughness can
lead to a reduced magnetic moment of the bilayer by two processes that probably
both play a role. First it increases the overall surface area giving more interface
Gd atoms the chance to couple directly with the magnetic moment of the adjoint
Ni layer by the interface polarization, described in section 4.4.4. Secondly, this
very large interface roughness may cause that the gadolinium is not completely
covered by nickel and that nonmagnetic GdAu alloys may partially form, reduc-
ing Msat below TC (Gd).
Samples with thinner nickel layers exhibit a similar reduction in Msat through-
out the whole investigated temperature range. Overcompensation, i.e. negative
remanence, only appears in the two samples with 75 Å and 100 Å on Si3 N4 , just
as for the series deposited at 100 K, described in section 4.4.5. Whereas the
100 Å nickel system on Al2 O3 does not show any compensation, but a simple
reduction of Mrem . Possibly the gadolinium layer does not have the desired thick-
ness or contaminations of the substrate surface have lead to additional oxidation.
Besides higher interface roughness stronger interdiffusion must be concerned as a
reason for the smaller magnetic moment of the bilayers. This reduces selectively
the magnetic moment of the Ni layer.

In figure 4.39 the saturation magnetization over the nickel layer thickness
shown for the samples deposited at room temperature. The graph at the top dis-
plays the data of the Al2 O3 samples. Compared to the ones deposited at 100 K,
that are shown in figure 4.36, the vanishing nickel moment is much larger. This
indicates a strongly increased alloy formation at the interface. But again, in or-
der to have a more reliable number for the exact portion of nickel that looses its
magnetic moment due to alloying with gadolinium, many more samples should be
investigated. The value obtained from the sample with 50 Å of Ni appears faulty.
The graph at the bottom shows the same dependence for the Si3 N4 series. Msat
at 300 K of the samples produced at 100 K was hardly reduced. In contrast in the
series that was deposited at room temperature Msat (300 K) is largely reduced,
nearly to the value of the Al2 O3 samples.

In figure 4.40 hysteresis loops of room temperature samples are compared

with ones of samples that were produced at 100 K. The left column presents
data from bilayers deposited on Al2 O3 the right column of ones on Si3 N4 . In the
top row hysteresis loops at 300 K are shown. At this temperature only the Ni
layer is ferromagnetic and contributes to the saturation magnetization, but Msat
is reduced by the antiparallel polarized gadolinium atoms at the interface. One
can see that Msat (300 K) is strongly reduced by about 30 % if the substrates are
not cooled. The reduction is larger for the Si3 N4 samples at this temperature.
This is in accordance with the stronger increase in surface roughness described
in section 4.3.4, leading to a larger portion of the Gd moments to be polarized
antiparallel to the Ni. At room temperature the stronger interdiffusion raises the
Chapter 4. Experimental Results 92

Figure 4.39: Msat at 300 K of bilayers produced at room temperature:

On both substrates Msat is significantly reduced due to alloying and interface
polarization.

number of Ni atoms in the Gd layer, whose magnetic moment is annihilated by

the Gd surrounding.
The four graphs in the two lower rows of figure 4.40 display the M(H)-dependence
at the temperature of maximum overcompensation. For Si3 N4 this is 45 K and for
Al2 O3 40 K in the case of the room temperature samples. The ones produces at
100 K all measurements were performed at 45 K. The full field range is shown in
the middle and the section around 0 Oe at the bottom. Msat is again reduced for
the bilayers on both substrates by the higher deposition temperature. But now the
decrease is nearly the same for both substrates, i.e. that more magnetic moment
from the Gd layer is missing, what effects the lower compensation temperatures.
There are several possible causes for this: Stronger oxidation during deposition,
more Gd polarized by the adjoint Ni, stronger intermixing and no complete cover
by nickel of the gadolinium, leading to alloying with the capping Au layer. The
93 4.4. Magnetic characterization of Gd/Ni-bilayers

Figure 4.40: Comparison of hysteresis loops for different deposition tem-

peratures: The left column shows data of Al2 O3 samples, the right one of Si3 N4
samples. Top graph M(H) at 300 K; in the middle hysteresis loops at 45 K over
the full applied field range; the bottom row is a zoom of the inner part of the
M(H)-scans right above.
Chapter 4. Experimental Results 94

first one is believed to play a subordinate role since the pressure is only slightly
higher during deposition. The second argument should induce a similar decrease
as at 300 K. So the two last processes seem to cause this effect. With Msat (300 K)
being more reduced on Si3 N4 , the additional decrease due to missing Gd moment
must be emphasized for the Al2 O3 samples. So it can be assumed that stronger
alloying of Gd with Ni but also of Gd with the capping material Au is responsible
for this effect. The zoom of the area around the origin of the graphs show the
decrease in magnetic moment of both layers since they switch at different external
fields, showing clearly that the total magnetic moment of the Gd layer is strongly
reduced, especially for the Al2 O3 sample. This favors the argument that Gd
moment is vanishing due to the formation of nonmagnetic GdAu alloys.

Figure 4.41: Hysteresis loops at 4 K : Top row shows the full magnetic field
range; the bottom row a zoom of the inner section between -400 Oe to 400 Oe.

The figure 4.41 compares the hysteresis loops at 4 K. The observed changes
in Msat are equal to the ones in figure 4.40. The plots in the bottom row reveal an
interesting feature. The coercivity of the Gd layer does not change, although its
total moment decreases. This indicates that the Gd grains being ferromagnetic
are similar as also observed by STM and that differences in the interface structure
cause the changed magnetic behavior.
95 4.5. Proposed model of the bilayered system

4.5 Proposed model of the bilayered system

In this section a model of the bilayer system Gd/Ni is proposed that reproduces
the observed structural and magnetic properties of the investigated samples. The
experimental methods resolving the structure of the samples have shown an amor-
phous growth of the materials, especially of the initial Gd layer. Its surface pos-
sesses a roughness in the range of the overall film thickness, dependent on the
chosen substrate and deposition temperature. The ordering on an atomic scale,
revealed by MEED and XRD seems in contrast to be unchanged. Only for a very
thick coverage of gadolinium, here 600 Å crystalline ordering on a very low level
could be observed. But still the small grain size leads to a strong broadening of
the peak. There was no indication from these structural investigations for a real
intermixing on a atomic scale. The overlap of the AES signals can be interpreted
as a result of the incoming Nickel filling up the cracks and slots first and leveling
out the surface at the end. Figure 4.42 gives a sketch of the proposed model for
the structure and morphology of the Gd/Ni bilayer system. The Gd layer consists
of nanocrystallites of a few nm in diameter. The nickel on top forms a smooth
flat film. Intermixing is limited to a small region at the interface.

Figure 4.42: Illustration of the model for GdNi-bilayers: The granular

Gd film is very rough. The Ni on top preferably accumulates in the slots of the
Gd layer.

The magnetic data shows the interplay of a Ni layer possessing an approxi-

mately constant magnetic moment over the investigated temperature range from
4 K to 300 K and a Gd layer, whose magnetic moment varies strongly with
changing temperature. This temperature dependence changes with the choice of
substrate and slightly with the deposition temperature. The transition tempera-
ture to ferromagnetic coupling of the gadolinium layer is mainly governed by the
grain size and the overall film thickness. But even in the case of thick Gd films in
the bilayers, there is a small additional increase in magnetization at 50 K, what
corresponds to the Curie temperature of the GdNi alloy, shown in figure A.2. But
intermixing is limited in this picture to a thin band at the interface. The lack
of magnetic moment at room temperature can be attributed to the polarized Gd
interface layer according to [91], which could be observed in the reflection XMCD
measurements (see figures 4.30 and 4.31).
Chapter 4. Experimental Results 96

4.6 Preliminary results on Ni/Gd/Ni-trilayers

A set of samples of a Ni/Gd/Ni trilayer system was produced. All of them
covered by 20 Å Au film as a protective capping layer. The Ni thickness was
kept to 40 Å and the one of the Gd layer was varied between 2.5 Å and 20 Å. It
turned out that in this series the Ni content of the samples was not constant, due
to a technical problem of the EFM evaporator. But nevertheless, very interesting
coupling phenomena between the two ferromagnetic Ni layers over the spacer was
observed. A short summation of the obtained data will be presented in this last
section, since in order to obtain a more complete picture many more experiments
should be undertaken, including investigations of the structural properties again.

Figure 4.43: Remanence of a Ni/Gd/Ni trilayer: The sample composition

was Al2 O3 /Ni(40Å)/Gd(5Å)/Ni(40Å)/Au(20Å) deposited at Tdep = 100 K.

In the following a brief selection of SQUID measurements on the trilayer

system Al2 O3 /Ni(40Å)/Gd(5Å)/Ni(40Å)/Au(20Å) is presented. So far investi-
gations with XRD and XRR to resolve structural features were unsuccessful as
well as XMCD measurements. The reasons for this are, that the intermediate Gd
layer is too thin to be visible in investigations of the film structure. Why there
was no XMCD signal observable in any sample, is still unclear. The measured
saturation magnetization at 300 K, taken from figure 4.44 is about a factor of 2.5
smaller than expected for a 80 Å thick Ni film. Two possible explanations may
be given for this. First the beam of nickel atoms was misaligned and the layer
thickness was not correct on the substrates. Second the moments of the Ni layers
97 4.6. Preliminary results on Ni/Gd/Ni-trilayers

align antiparallel and therefore most of the XMCD signal is canceled out and the
effective magnetic moment is below the resolution of this method in transmission.
Figure 4.43 shows a measurement of the remanent moment of Ni/Gd/Ni trilayer.
The sample Al2 O3 /Ni(40Å)/Gd(5Å)/Ni(40Å)/Au(20Å) was magnetized by an
external field of 10 kOe at each temperature and then the magnetic moment was
measured after switching off the magnetic field like it was done with the bilayered
systems in section 4.4. The remanence possesses a strange temperature depen-
dence. At temperatures above 200 K it is negative, so oppositely orientated to the
magnetizing field, what is affirmed by the hysteresis loop at 300 K in figure 4.44.
It changes to positive values between 220 K and 200 K. Then a temperature
regime begins with a positive remanence of approximately the same magnitude.
At 50 K the Gd or maybe a GdNi alloy becomes ferromagnetic and the remanent
moment increases strongly with decreasing temperature, overruling the magnetic
moment of the pure Ni layers.
To find a model that reproduces this behavior is a challenging task. Especially
the negative remanence at higher temperatures is not understood so far. The
magnetic moments of the two nickel layers seem to align antiparallel since the
saturation magnetization is reduced significantly. Now two possible configura-
tions can be named. First, the two layers have different magnetic properties due
to the different growth conditions and the intermediate Gd has no influence at
higher temperatures since it consists of only 2 monolayers nominally. So temper-
ature dependence of the saturation magnetization and the coercivity may differ
between the two layers of nickel and so it is either the lower or the upper one,
which remains aligned in the field direction, while the other is reoriented antifer-
romagnetically to it. The excess in Msat of one layer over the other is then visible
as the remanent moment.
Another possible explanation includes the interface layer of gadolinium. From the
measurements by XMCD on Gd/Ni bilayers it is known that the interface layer
is polarized by the adjoint nickel moment in section 4.4.4. Changes in either the
nickel layers with decreasing temperature or the gadolinium itself may lead to
a different polarization of the interface moments. This idea is supported by the
observation that Mrem changes from negative to positive to approximately the
same value. In order to support this assumption it would be necessary to know
the morphology of this thin Gd film, wether it grows as a closed film or in islands
or if intermixing occurs. But the more or less constant moment of this polarized
layer is unclear in this model.
In figure 4.44 hysteresis loops in the three temperature ranges are shown. The
left side gives the magnetization over the whole measured field range of -10 kOe
to +10 kOe. The right column contains zooms of the inner part of the hysteresis
loops. An interesting feature that appears here is visible in the large range loop
at 45 K. Between 3 and 6 kOe there is a steplike rise in the magnetization as it
was observed for the bilayer systems in the hysteresis loops at 4 K. But the step
height is far too small to originate from the switching of the complete nickel layers.
In [92] a similar steplike rise is reported in Gd/Co multilayers. The additional
measurements on the field dependence of the magnetoresistance revealed in this
Chapter 4. Experimental Results 98

Figure 4.44: Hysteresis loops at three chosen temperatures: In the field

dependent measurement at 300 K the negative remanence is really observed. The
arrows in the plot shall indicate the devolution of the magnetization. At 45 K
and 4 K the remanet moment points in the previously applied field direction.

work a gradual reorientation of the spins at the interface. At low fields the spins
are only partially aligned, so that there is an angle between the ones of the two
different materials, called a twisted-spin state, as reported for Fe/Gd multilayers
[93]. The high fields finally achieve to force both to be parallel. The same seems
to appear here. In contrast to the bilayers the step represents not a magnetic
moment equivalent to the complete Ni layer. So this increase may be attributed
to a parallel alignment of spins that are twisted at lower magnetic fields.
Chapter 5

Outlook and conclusions

The results have shown a strong influence of the film growth and thereby structure
and morphology on the magnetic properties. This correlation should be further
investigated by improving the quality of the Gd layer. Possible ways to do that
would be to use other substrates like tungsten or niobium single crystals or to use
seed layers of these materials. It was also shown in literature that seed layers of
yttrium or chromium can support epitaxial growth of gadolinium. The next in-
teresting point would be to produce and study bilayers of the same compositions
on Al2 O3 and Si3 N4 like in this work but in opposite stacking order. As can be
seen in section 4.3.1 the crystalline quality of the gadolinium layer improves and
a sharper interface is formed, if nickel is deposited first. In addition intermixing
of Gd and Ni is reduced. The problem that gadolinium would alloy readily with
any noble metal, that can be considered as a capping material, could be reduced
by introducing an additional layer of titanium, molybdenum or chromium acting
as an interdiffusion barrier. These materials are immiscible with gadolinium, but
titanium fulfills best the needs for adequate evaporation conditions, no miscibil-
ity and no magnetic moment. SQUID measurements on a few of such bilayers
already showed the the expected behavior of increased Msat at room temperature
due to the sharper interface and thereby smaller number of antiparallel aligned
Gd moments at the interface. But so far no diffusion barrier was introduced to
the capping layer. In addition to a magnetic characterization by SQUID mea-
surements using MOKE at low temperatures, e.g. in the Gd aligned state, should
give hysteresis loops like in figure 4.26.
The preliminary results on the trilayered systems Ni/Gd/Ni in section 4.6 already
indicate an interesting exchange coupling of the nickel layers through the non-
magnetic Gd, respectively GdNi, layer above its Curie temperature. Especially
the change in sign of the remanent magnetization with decreasing temperature is
not understood so far. Again investigations on structure and magnetism of such
trilayers are considered to be a fruitful topic for further experiments, especially
to record magnetization by MOKE for each layer separately. For that of course
a transparent substrate has to be used.
But also on the systems presented in this work additional measurements should
Chapter 5. Outlook and conclusions 100

be carried out to clarify the shape of the interface between the two materials
using HRTEM and in-situ STM. More information about the magnetism could
be obtained by further XMCD experiments, MOKE measurements at low tem-
peratures (similar to figure 4.26) and PEEM revealing the domain structure. The
alloy formation at the interfaces is still not clear and additional information on
the processes contributing to this would be desirable. One of the major problems
here will be to divide real chemical intermixing on an atomic level and apparent
intermixing as a consequence of the large surface roughness. Here the preparation
of smooth epitaxial gadolinium films would help, that could be studied by in-situ
Auger spectroscopy.
All measurements have shown that the strong temperature dependence of the
magnetism in gadolinium produces a variety of interesting, temperature depen-
dent coupling phenomena. In this final section a few ideas were mapped out, that
are considered to be worthwhile to be pursued.
Chapter 6

Summary

The bilayer system of gadolinium and nickel exhibits due to the magnetostatic
exchange coupling between the two materials and the strong temperature depen-
dence of the ferromagnetic ordering of gadolinium interesting magnetic properties.
It was shown, that this response to external magnetic fields is directly connected
to the structural properties and morphology of the thin films.
Changes in the deposition conditions have lead to differences in the growth of the
Gd/Ni bilayers, resulting in different magnetic characteristics of the samples. It
was shown, that if the ratio of layer thickness and by that of the overall magnetic
moment of each species is in a certain range, overcompensation can be observed,
i.e. negative remanence occurs at a characteristic temperature. This overcom-
pensation temperature depends on the condition, when the products of the total
magnetic moments and their coercivities (Mi ·HC , i) of both layers change their
order in magnitude. So this characteristic was nearly the same for all systems,
neither deposited on Al2 O3 or Si3 N4 nor at room temperature or at 100 K. The
small differences like a shift of about 5 K to lower temperatures in the case of
the room temperature samples can be attributed to additional alloying of the
materials and enhanced polarization of interface Gd moments (see section 4.4.6).
One of the key results of this work is the antiparallel alignment of the magnetic
moments in remanence, which can be deduces from SQUID measurements but
was clearly visible in the XMCD measurements in section 4.4.4. Especially the
simultaneous switching of the layers due to their exchange coupling could be elab-
orated. Another surprising feature, that was observed, was the polarization of a
small portion of the gadolinium. The gadolinium interface layer aligns antiparal-
lel to the magnetic moment of the adjoint nickel layer even at temperatures far
above the Curie point of the thin gadolinium film (see figures 4.28 and 4.30).
The change from the Ni-aligned to the Gd-aligned state, taking place at the point
of maximum overcompensation, was investigated in a series of SQUID measure-
ments, presented in section 4.4.3. Similar behavior and transitions have been
reported for similar systems of Gd together with a transition metal, like Gd/Fe,
Gd/Co and Gd/CoNi to name some of them. One major result of this research
was, that this point of change mainly depends on the total amount of material
Chapter 6. Summary 102

deposited. For example on Si3 N4 ferromagnetic ordering begins at much higher

temperatures as on Al2 O3 , but the maximum negative remanence appears around
50 K for both systems. Whereas samples produced at room temperature all show
a small shift to lower temperatures due to a reduction of magnetic moment in
consequence of enhanced alloying.
Previously to this work it was believed that gadolinium grows completely amor-
phous on the used substrates, because neither MEED nor LEED or XRD have
shown reflexes due to constructive interference, but with increasing film thick-
ness reflexes attributed to crystalline gadolinium and GdNi alloy were observed
in section 4.3.1. A major field of the research in this work was on the struc-
tural properties of the bilayers. For this HRTEM and RBS were employed to
directly image the cross-section of the bilayer (see section 4.3.3). But so far the
obtained results of both methods could not be driven to a sufficiently high level
of resolution, where as HRTEM is expected to be the most promising. With this
technique it should also be possible to investigate the crystallinity of the single
layers, like it was done in [70]. STM revealed the morphology of thin gadolinium
films and how the choice of substrate and deposition temperature affect it (see
section 4.3.4).
All the made observations could be summarized in a model, described in sec-
tion 4.5 of a thin nanocrystalline gadolinium film with a very high surface rough-
ness. The later deposited nickel preferentially accumulates in the deepenings,
evening the sample surface. Intermixing is believed to be limited to a narrow re-
gion around the interface but must be considered, since on one hand a reduction
of the saturation magnetization of the Ni layer is observed and on the other hand
GdNi XRD reflexes, too.
Preliminary results on Ni/Gd/Ni trilayers are shown in section 4.6, but only a
small amount of the obtained data is presented, because it is understood to a very
small degree only and the measurements need to be verified, too. The found mag-
netic coupling exhibits an interesting temperature dependence with a transition
from antiferromagnetic to ferromagnetic around 200 K.
Chapter 7

Zusammenfassung

Das Bilagensystem aus Gadolinium und Nickel zeigt interessante magnetische Ei-
genschaften, hervorgerufen durch die magnetostatische Austauschkopplung zwi-
schen den beiden Materialien und die starke Temperaturabhänigkeit des Ferro-
magnetismus von Gadolinium. Es konnte gezeigt werden, dass dieses Ansprechen
auf externe magnetische Felder direkt mit den strukturellen Eigenschaften und
Morphologie der dünnen Filme verknüpft ist.
Änderungen der Aufdampfbedingungen führten zu Unterschieden im Filmwachs-
tum der Gd/Ni Bilagen und damit zu unterschiedlichen magnetischen Eigenschaf-
ten der Proben. Wenn das Verhältnis der Schichtdicken und damit das magne-
tische Gesamtmoment der einzelnen Materialien in einem bestimmten Bereich
gewählt wird, so tritt Überkompensation auf, d. h. eine negative Remanenz er-
gibt sich bei einer charakteristischen Temperatur. Die Überkompensationstem-
peratur hängt davon ab, wann die Produkte aus magnetischem Gesamtmoment
und Koerzitivität (Mi ·HC , i) der beiden Schichten ihre Reihenfolge bezüglich der
Größe wechseln. Deshalb war diese charakteristische Größe in etwa gleich für alle
Proben, egal ob sie auf Al2 O3 oder Si3 N4 aufgebracht wurden. Auch die Depositi-
onstemperatur hatte darauf nur einen geringen Einfluss. Die geringe Verschiebung
um 5 K zu niedrigeren Temperaturen im Falle der Proben, die bei Raumtempe-
ratur hergestellt wurden, kann durch zusätzliche Durchmischung und verstärkte
Polarisation der Gd-Momente an der größeren Grenzschicht erklärt werden (s.
section 4.4.6).
Eines der grundlegenden Ergebnisse dieser Arbeit ist die antiparallele Ausrich-
tung der magnetischen Momente in Remanenz, die aus den SQUID-Daten indi-
rekt geschlossen werden kann und durch die XMCD-Messungen in section 4.4.4
direkt beobachtet werden konnten. Speziell das simultane Ummagnetisieren der
Schichten aufgrund der Austauschkopllung konnte herausgearbeitet werden. Ein
weiteres überraschendes Detail, das beobachtet wurde, war die Polarisierung ei-
nes Teils des Gadoliniums. Die Gadoliniummomente an der Grenzschicht richten
sich antiparallel zum magnetischen Moment der angrenzenden Nickelschicht aus,
selbst bei Raumtemperatur, was weit über der Curietemperatur des dünnen Ga-
doliniumfilms liegt (s. figures 4.28 and 4.30).
Chapter 7. Zusammenfassung 104

Der Wechsel vom so genannten Ni-dominierten zum Gd-dominierten Zustand, der

bei der Temperatur stattfindet, wo die Überkompensation maximal ist, wurde in
einer Reihe von SQUID-Messungen untersucht, die in section 4.4.3 aufgeführt
sind. Ähnliches Verhalten und ähnliche Übergänge sind für andere Systeme aus
Gd in Verbindung mit einem Übergangsmetall bereits in der Literatur bekannt,
wie Gd/Fe, Gd/Co und Gd/CoNi, um ein paar Beispiele zu nennen. Hervorzuhe-
ben ist hierbei, dass dieser Wechsel hauptsächlich von der Menge der deponierten
Materialien und damit von den Gesamtmomenten abhängt. So setzt die ferroma-
gnetische Ordnung im Gadolinium auf Si3 N4 bereits bei deutlich höheren Tempe-
raturen ein als auf Al2 O3 und trotzdem tritt die maximale negative Remanenz bei
etwa 50 K in beiden System auf. Wohingegen Proben, die bei Raumtemperatur
hergestellt wurden, alle eine geringe Reduzierung der Überkompensationstempe-
ratur aufweisen. Diese wird dadurch verursacht, dass ein Teil des magnetischen
Moments verloren geht durch die verstärkte Interdiffusion und durch die größe-
re Fläche der Grenzschicht wegen der höheren Rauhigkeit, was dazu führt, dass
mehr Gd-Momente antiparallel zum Nickel polarisiert werden. Die Überkompen-
sationstemperatur wird also nicht durch eine vollständige Durchmischung der
Gadoliniumschicht mit Nickel zur GdNi-Legierung, deren TC = 50 K beträgt,
festgelegt, sondern durch die magnetischen Gesamtmomente. TC der Gadolini-
umschicht hängt stark von der Korngröße beim Wachstum und der Schichtdicke
ab. So kann die Überkompensationstemperatur durch eine Vergrößerung der Di-
cke des Gadoliniumfilmes zu höheren Temperaturen getrieben werden.
Vor dieser Arbeit wurden die vorausgegangenen Daten so gedeutet, dass Gado-
linium vollständig amorph auf den verwendeten Substraten aufwächst, da weder
MEED, LEED noch XRD Reflexe durch konstruktive Interferenz gezeigt haben.
Geht man aber zu größeren Schichtdicken können Reflexe beobachtet werden,
die kristallinem Gadolinium und der GdNi-Legierung zugeschrieben werden, wie
in section 4.3.1 aufgeführt. Ein Schwerpunkt in dieser Untersuchung lag auf den
strukturellen Eigenschaften der Bilagen. HRTEM und RBS wurden eingesetzt,
um direkt den Querschnitt durch die Bilagen abzubilden (s. section 4.3.3). Je-
doch konnten die Ergebnisse mit beiden Methoden noch nicht eine ausreichende
Auflösung erreichen. Man erwartet, dass HRTEM die dafür vielversprechndste
Technik ist. Damit sollte es möglich sein, die Kristallinität der einzelnen Schich-
ten direkt untersuchen zu können, wie in [70]. Mit STM konnte die Morphologie
dünner Gadoliniumfilme abgebildet werden und wie diese von der Wahl des Sub-
strates und Depositionstemperatur abhängt (s. section 4.3.4).
Alle Beobachtungen, die gemacht wurden, sind in einem vorgeschlagenem Modell
aus einem nanokristallinem Gadoliniumfilm mit einer sehr hohen Oberflächenrau-
higkeit zusammengefasst, das in section 4.5 beschrieben wird. Das anschließend
deponierte Nickel lagert sich bevorzugt in den Einschnitten und Vertiefungen an,
was die Oberfläche glättet. Durchmischung ist wahrscheinlich auf einen engen Be-
reich um die Grenzschicht beschränkt, muss aber in Betracht gezogen werden, da
einerseits eine Reduzierung der Sättigungsmagnetisierung der Nickelschicht beob-
achtet wird und andererseits bei der Röntgenbeugung GdNi-Reflexe auftauchen.
Vorläufige Ergenisse an Ni/Gd/Ni Trilagen sind in section 4.6 aufgeführt, wo-
105

bei nur ein kleiner Teil, der erhaltenen Daten präsentiert wurden, weil sie im-
mer noch nur unzureichend erklärt werden können und noch kein schlüssiges
Modell dafür gefunden wurde. Eine Verifizierung der Messungen muss ebenfalls
noch durchgeführt werden. Es zeigte sich aber ein sehr interessantes magneti-
sches Kopplungsverhalten zwischen den Nickelschichten mit einem Übergang von
antiferromagnetisch zu ferromagnetisch bei etwa 200 K.
Chapter 7. Zusammenfassung 106
List of Figures

2.1 Spin and effective magnetic moment of the rare earth metals . . . 6
2.2 Density of states in an external field . . . . . . . . . . . . . . . . . 8
2.3 Normalized temperature dependence of M/M0 . . . . . . . . . . . 10
2.4 Ni 3d band . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.5 Values of the Stoner criterion . . . . . . . . . . . . . . . . . . . . 12
2.6 Radial distribution of the electrons in Gd . . . . . . . . . . . . . . 13
2.7 RKKY exchange interaction of Gd . . . . . . . . . . . . . . . . . 15
2.8 Effective anisotropy Kef f ·d . . . . . . . . . . . . . . . . . . . . . . 17
2.9 Growth modi of epitxial thin films . . . . . . . . . . . . . . . . . . 19

3.1 Schematic layout of MEDUSA UHV-chamber . . . . . . . . . . . 22

3.2 Schematics of an EFM evaporator and source alignment in MEDUSA 23
3.3 DC-SQUID loop . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.4 Schematic overview of the MPMS XL-5S . . . . . . . . . . . . . . 28
3.5 Illustration of the gradiometer and the RSO mode . . . . . . . . . 29
3.6 Schematic illustration of dipole transitions . . . . . . . . . . . . . 31
3.7 Example of a XMCD spectrum . . . . . . . . . . . . . . . . . . . 32
3.8 XMCD set-up ”ALICE” at BESSY . . . . . . . . . . . . . . . . . 33
3.9 Layout of an undulator . . . . . . . . . . . . . . . . . . . . . . . . 33
3.10 Transition probabilities for 2p→3d . . . . . . . . . . . . . . . . . 35
3.11 Illustration of a SPM . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.12 Piezo tube in a SPM . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.13 SEM picture of a STM-tip . . . . . . . . . . . . . . . . . . . . . . 37
3.14 Conceptional set-up of a STM . . . . . . . . . . . . . . . . . . . . 38
3.15 Schematic layout of the MEED set-up in the MEDUSA UHV-
chamber . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
List of Figures 108

3.16 Auger effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

3.17 Illustration of RBS technique . . . . . . . . . . . . . . . . . . . . 42
3.18 Kinematic factor K for RBS . . . . . . . . . . . . . . . . . . . . . 43
3.19 Different modes of TEM . . . . . . . . . . . . . . . . . . . . . . . 44
3.20 Sample preparation for TEM . . . . . . . . . . . . . . . . . . . . . 45

4.1 Sample composition . . . . . . . . . . . . . . . . . . . . . . . . . . 48

4.2 AFM picture of a Si3 N4 substrate . . . . . . . . . . . . . . . . . . 48
4.3 XRD-scan of a thin Gd film . . . . . . . . . . . . . . . . . . . . . 50
4.4 LEED images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.5 MEED images of Gd on Al2 O3 . . . . . . . . . . . . . . . . . . . . 51
4.6 XRR scan of a Gd/Au bilayer . . . . . . . . . . . . . . . . . . . . 52
4.7 XRR scan of a Ni/Au bilayer . . . . . . . . . . . . . . . . . . . . 53
4.8 TC of thin Gd films: . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.9 Large angle X-ray diffraction of a Gd/Ni bilayer . . . . . . . . . . 55
4.10 XRD scans of Gd/Ni and Ni/Gd bilayers . . . . . . . . . . . . . . 57
4.11 Element analysis by AUGER . . . . . . . . . . . . . . . . . . . . 58
4.12 RBS setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.13 RBS spectra of Gd(50Å)/Ni(100Å)/Au(20Å) samples . . . . . . . 60
4.14 TEM images of Gd/Ni bilayers . . . . . . . . . . . . . . . . . . . 62
4.15 STM images of Gd on Al2 O3 . . . . . . . . . . . . . . . . . . . . 63
4.16 STM images of Gd on Si3 N4 . . . . . . . . . . . . . . . . . . . . . 64
4.17 STM images of Gd on Al2 O3 . . . . . . . . . . . . . . . . . . . . 65
4.18 STM images of Gd on Si3 N4 . . . . . . . . . . . . . . . . . . . . . 66
4.19 Surface of a complete bilayer sample . . . . . . . . . . . . . . . . 67
4.20 Magnetic anisotropy of a Gd/Ni bilayer at 300K . . . . . . . . . . 68
4.21 Magnetic anisotropy of a Gd/Ni bilayer at 4K . . . . . . . . . . . 69
4.22 Typical hysteresis loop of a Gd/Ni-bilayer with negative remanence 70
4.23 Hysteresis loops at different temperatures . . . . . . . . . . . . . . 71
4.24 MR and MS over temperature . . . . . . . . . . . . . . . . . . . . 73
4.25 Hysteresis loops in the Gd aligned state . . . . . . . . . . . . . . . 75
4.26 Hysteresis of the TM in the RE aligned state . . . . . . . . . . . . 76
4.27 Aging of Gd/Ni bilayers . . . . . . . . . . . . . . . . . . . . . . . 77
4.28 XMCD measurements in the Ni aligned state . . . . . . . . . . . . 79
109 List of Figures

4.29 XMCD in reflection geometry at T = 25 K . . . . . . . . . . . . . 80

4.30 XMCD in reflection geometry . . . . . . . . . . . . . . . . . . . . 82
4.31 XMCD at the Ni L3 edge in reflection . . . . . . . . . . . . . . . . 82
4.32 Hysteresis loops for different Ni film thickness on Al2 O3 . . . . . . 84
4.33 Mrem (T) for different Ni content on Al2 O3 . . . . . . . . . . . . . 85
4.34 Hysteresis loops for different Ni content on Si3 N4 . . . . . . . . . . 86
4.35 Mrem (T) for different Ni content on Si3 N4 . . . . . . . . . . . . . 87
4.36 Msat over Ni thickness for Al2 O3 samples . . . . . . . . . . . . . . 88
4.37 Msat over Ni thickness for Si3 N4 samples . . . . . . . . . . . . . . 89
4.38 Comparison of M(T) for different deposition temperatures . . . . 90
4.39 Msat at 300 K of bilayers produced at room temperature . . . . . 92
4.40 Comparison of hysteresis loops for different deposition temperatures 93
4.41 Comparison of hysteresis loops for different deposition temperatures 94
4.42 Illustration of the model for GdNi-bilayers . . . . . . . . . . . . . 95
4.43 Remanence of a Ni/Gd/Ni trilayer . . . . . . . . . . . . . . . . . 96
4.44 Hysteresis loops of a Ni/Gd/Ni trilayer . . . . . . . . . . . . . . . 98

A.1 Phase diagram of the binary alloy system Gd-Ni . . . . . . . . . . 121

A.2 TC of Gdx Ni1−x alloys . . . . . . . . . . . . . . . . . . . . . . . . 122

B.1 RBS depth profile . . . . . . . . . . . . . . . . . . . . . . . . . . . 123

List of Figures 110
Bibliography

[1] G. Schatz, “Surfaces and thin film physics,” 2000. Vorlesungsskript zur
Wahlpflichtfachvorlesung.

[2] H. Ibach and H. Lüth, Festkörperphysik. Berlin, Deutschland: Springer,

1999.

[3] A. Barth, “Laser-Modifikation von dünnen, magnetischen Filmen,” Master’s

thesis, Universität Konstanz, 2002.

[4] M. Dippel, Charakterisierung des Schmelzverhaltens selbstorganisierter

Indium–Nanostrukturen auf WSe2 –Substraten. PhD thesis, Universität Kon-
stanz, 2000.

[5] I. Guhr, “CrPt3 nanostrukturen auf WSe2 ,” Master’s thesis, Universität Kon-
stanz, 2003.

[6] N. Spaldin, Magnetic Materials. Cambridge, UK: Cambridge University

Press, first edition ed., 2003.

[7] G. Schatz, Nukleare Festkörperphysik. Stuttgart, Deutschland: Teubner-

Studienbücher: Physik, third edition ed., 1997.

[8] T. Fließbach, Elektrodynamik. Deutschland: Spektrum Akademischer Ver-

lag, second ed., 1997.

[9] W. Demtröder, Experimentalphysik 2. Berlin, Deutschland: Springer, 1995.

[10] W. Demtröder, Experimentalphysik 3. Berlin, Deutschland: Springer, 1996.

[11] H. Lüth, Solid Surfaces, Interfaces and Thin Films. Berlin: Springer, 2001.

[12] T. Ulbrich, “Aufbau eines MOKE-Systems zur Untersuchung magnetischer

Nanostrukturen,” Master’s thesis, Universität Konstanz, 2003.

[13] D. LaGraffe, P. A. Dowben, and M. Onellion, “The chemistry of the

gadolinium-nickel interface,” J. of Vacuum Science and Technology, vol. 8,
pp. 2738–2742, 1990.

[14] B. Altuncevahir and A. R. Koymen, “Deposition-order-dependent coercivity

of CoNi/Gd bilayers,” J. of Appl. Phys., vol. 90, no. 6, pp. 2939–2942, 2001.
Bibliography 112

[15] B. Altuncevahir and A. R. Koymen, “Positive and negative exchange bias in

CoNi/Gd/CoNi trilayers and CoNi/Gd bilayers,” JMMM, vol. 261, pp. 424–
432, 2003.
[16] T. R. McGuire and R. J. Gambino, “Magnetic and transport properties of
Gd-Ni amorphous alloys,” IEEE Transactions on Magnetics, vol. 14, no. 5,
pp. 838–840, 1978.
[17] P. Grünberg, R. Schreiber, Y. Pang, M. B. Brodsky, and H. Sowers, “Layered
magnetic structures: Evidence for antiferromagnetic coupling of Fe layers
across Cr interlayers,” Phys. Rev. Lett., vol. 57, pp. 2442–2445, Nov 1986.
[18] S. S. P. Parkin, N. More, and K. P. Roche, “Oscillations in exchange coupling
and magnetoresistance in metallic superlattice structures: Co/Ru, Co/Cr,
and Fe/Cr,” Phys. Rev. Lett., vol. 64, pp. 2304–2307, May 1990.
[19] L. T. Baczewski, R. Kalinowski, and A. Wawro, “Magnetization and
anisotropy in Fe/Gd multilayers,” JMMM, vol. 177-181, pp. 1305–1307, 1998.
[20] A. Koizumi, M. Takagaki, M. Suzuki, N. Kawamura, and N. Sakai, “Anoma-
lous magnetic hysteresis of Gd and Fe moments in a Gd/Fe multilayer mea-
sured by hard x-ray magnetic circular dichroism,” Phys. Rev. B, vol. 61,
pp. R14909–R14912, June 2000.
[21] A. Pogorily, E. Shypila, and C. Alexander, “A study of magnetization in
exchange-coupled FM/Gd bilayers,” JMMM, vol. 286, pp. 493–496, 2005.
[22] J. Colino, J. P. Andres, J. M. Riveiro, J. L. Martýnez, C. Prieto, and J. L.
Sacedon, “Spin-flop magnetoresistance in Gd/Co multilayers,” Phys. Rev. B,
vol. 60, no. 9, pp. 6678–6684, 1999.
[23] R. E. Camley, “Surface spin reorientation in thin Gd films on Fe in an applied
magnetic field,” Phys. Rev. B, vol. 37, no. 7, pp. 3608–3611, 1987.
[24] D. Haskel, Y. Choi, D. R. Lee, J. C. Lang, G. Srajer, J. S. Jiang, and
S. D. Bader, “Hard x-ray magnetic circular dichroism study of a surface-
driven twisted state in Gd/Fe multilayers,” J. of Appl. Phys., vol. 93, no. 10,
pp. 6507–6509, 2003.
[25] J. Jensen and A. R. Mackintosh, Rare Earth Magnetism. Oxford, UK: Oxford
University Press, 1991.
[26] R. J. Elliot, Magnetic properties of rare earth materials. London, UK:
Plenum Press, 1972.
[27] J. D. Jackson, Classical Electrodynamics. New York, USA: Wiley, third ed.,
1999.
[28] L. J. Swartzendruber, “Properties, units and constants in magnetism,”
JMMM, vol. 100, pp. 573–575, 1991.
113 Bibliography

[29] B. T. Matthias, H. Suhl, and E. Corenzwit, “Spin exchange in superconduc-

tors,” Phys. Rev. Lett., vol. 1, no. 3, pp. 92–94, 1958.

[30] C. Kittel, Einführung in die Festkörperphysik. 12th edition ed., 1996.

[31] M. A. Rudermann and C. Kittel, “Indirect exchange coupling of nuclear

magnetic moments by conduction electrons,” Phys. Rev. Lett., vol. 96, no. 1,
pp. 99–93, 1954.

[32] T. Kasuya, “A theory of metallic ferro- and antiferromagnetism on Zener’s

model,” Progress of Theoretical Physics, vol. 16, no. 1, pp. 45–57, 1956.

[33] K. Yosida, “Magnetic properties of Cu-Mn alloys,” Phys. Rev., vol. 106, no. 5,
pp. 893–898, 1957.

[34] M. B. Salomon, “Long-range incommensurate magnetic order in Dy-Y mul-

tilayer,” Phys. Rev. Lett., vol. 56, p. 259, 1986.

[35] C. F. Makrzak, D. Gibbs, P. Böni, and A. I. Goldman, “Magnetic rare-earth

superlattices,” J. Appl. Phys., vol. 63, no. 8, pp. 3447–3452, 1988.

[36] S. Parkin, “Systematic variation of the strength and oscillation period of

indirect magnetic exchange coupling through the 3d, 4d and 5d transition
metals,” Phys. Rev. Lett., vol. 67, no. 25, pp. 3598–3601, 1991.

[37] P. Pankowski, S. Pizzini, J. Pelka, A. Wawro, and L. Baczewski, “Growth

mode and structural characterization of epitaxial TM/RE thin films,” Jour-
nal of alloys and compounds, vol. 362, pp. 56–60, 2004.

[38] Z. S. Shan, D. J. Sellmyer, S. S. Jaswal, Y. J. Wang, and J. X. Shen, “Mag-

netism of rare-earth-transition-metal nanoscale multilayers. ii: Theoretical
analysis of magnetization and perpendicular magnetic anisotropy,” Phys.
Rev. B, vol. 42, no. 16, pp. 10446–10459, 1990.

[39] Z. S. Shan, “Magnetism of rare-earth-transition-metal multilayers,” Phys.

Rev. Lett., vol. 63, pp. 10446–10459, 1989.

[40] T. Kobayashi, H. Tsuji, S. Tsunashima, and S. Uchiyama, “Magnetization

process of exchange-coupled ferrimagnetic double-layered films,” Jpn. Journ.
of Appl. Phys., vol. 20, no. 11, 1981.

[41] K. L. Chopra, Thin film phenomena. McGraw-Hill Book Company, 1969.

[42] S. Jaakkola, S. Parviainen, and S. Penttilä, “Volume dependence of the

Curie temperature of rare-earth-3d transition metal compounds,” J. Phys.
F, vol. 13, pp. 491–502, 1983.

[43] A. Maier, Strukturelle und magnetische Eigenschaften von CoPt3 –

Nanostrukturen auf WSe2 . PhD thesis, Universität Konstanz, 2001.
Bibliography 114

[44] S. Esho, “Anomalous magneto-optical hysteresis loops of sputtered Gd-Co

films,” Japn. Journ. of Appl. Phys. Supplement, vol. 15, pp. 93–98, 1976.

[45] R. Pallesche, “Überstrukturen beim Wachsum von Pt auf WSe2 ,” Master’s

thesis, Universität Konstanz, 2006.

[46] R. E. Camley, “Properties of magnetic superlattices with antiferromag-

netic interfacial coupling: Magnetization, susceptibility, and compensation
points,” Phys. Rev. B, vol. 39, pp. 12316–12319, Jun 1989.

[47] F. Treubel, Ferrimagnetismus im Zweischichtsystem Gd/Ni. PhD thesis,

Universität Konstanz, 2005.

[48] R. Zorn, “Magnetometrie,” 30. Ferienkurs des Institutes für Festkörper-

forschung 1999, vol. 30, pp. A6.1–22, 1999.

[49] W. Buckel and R. Kleiner, Supraleitung. Weinheim: Wiley-VCH, sixth edi-

tion ed., 2004.

[50] J. B. Ketterson and S. N. Song, Superconductivity. Cambridge, UK: Cam-

bridge University Press, first edition ed., 1999.

[51] J. C. Gallop, SQUIDs, the Josephson effects and superconducting electronics.

Bristol, UK: Adam Hilger Series, first edition ed., 1990.

[52] Q. Design, “QDMPS 5XL manual,” 1998.

[53] A. Zieba and S. Forner, “Detection coil, sensitivity function, and sample
geometry effects for vibrating sample magnetometers,” Rev. Sci. Instrum.,
vol. 53, no. 9, 1982.

[54] J. L. Erskine and E. Stern, “Calculation of the M23 magneto-optical absorp-

tion spectrum of ferromagnetic nickel,” Phys. Rev. B, vol. 12, pp. 5016–5024,
1975.

[55] G. Schütz, W. Wagner, W. Wilhelm, P. Kienle, R. Zeller, R. Frahm, and

G. Materlik, “Absorption of circularly polarized X-rays in iron,” Phys. Rev.
Lett., vol. 58, no. 7, pp. 737–740, 1987.

[56] E. Meltchakov, H. C. Mertins, W. Jark, and F. Schäfers, “Magnetic circualr

dichroism of gd/transition metal multilayer structures around the gd m5,4 -
edges,” Nuclear Instruments and Methods in Physics, vol. 467–468, pp. 1411–
1414, 2001.

[57] S. Eisebitt, “Zirkulardichroismus in der Rumpfabsorption,” Ferienkurs des

IFF: Magnetische Schichtsysteme, vol. 30, pp. C6.1–C6.28, 1999.

[58] W. Kuch, Abbildende magnetische Mikrospektreskopie (Habitilationss-

chrift). PhD thesis, Universität Halle-Wittenberg, 2002.
115 Bibliography

[59] C. Sorg, Magnetic properties of 3d and 4f ferromagnets studied by X-ray

absorption spectroscopy. PhD thesis, Freie Universität Berlin, 2005.

[60] B. T. Thole, P. Carra, F. Sette, and G. van der Laan, “X-ray Circular Dichro-
ism as a probe of orbital magnetization,” Phys. Rev. Lett., vol. 68, no. 12,
pp. 1943–1947, 1992.

[61] G. Binnig, H. Rohrer, C. Gerber, and E. Weibel, “Tunneling through a con-

trollable vakuum gap,” Appl. Phys. Lett., vol. 40, no. 2, pp. 178–180, 1982.

[62] G. Binnig, H. Rohrer, C. Gerber, and E. Weibel, “Surface studies by scanning

tunneling microscopy,” Phys. Rev. Lett., vol. 49, pp. 57–60, 1982.

[63] J. Bardeen, “Tunneling from a many-particle point of view,” Phys. Rev. Lett.,
vol. 6, p. 57, 1961.

[64] R. Wiesendanger, Scanning Tunneling Microscopy. Berlin, Deutschland:

Springer, 1999.

[65] WSxM© . 4.0 Develop 9.0. Nanotec Electronica S.L., 2006. Freies Software-
programm − Download unter http://www.nanotec.es.

[66] M. Henzler and W. Göpel, Oberflächenphysik des Festkörpers. Stuttgart:

B. G. Teubner, 1994.

[67] E. Labs. homepage, 2006. http://www.eaglabs.com/en-

US/references/tutorial/augtheo/caiatheo.html.

[68] M. Farle, K. Baberschke, U. Stetter, A. Aspelmeier, and F. Gerhardter,

“Thickness-dependent Curie temperature of Gd(0001)/W(110) and its de-
pendence on the growth conditions,” Phys. Rev. B, vol. 47, no. 17, pp. 11571–
11574, 1993.

[69] J. S. Jiang, D. Davidovic, D. H. Reich, and C. L. Chien, “Oscillatory super-

conducting transition temperature in Nb/Gd multilayers,” Phys. Rev. Lett.,
vol. 74, 1995.

[70] G. Bertero, T. C. Hufnagel, B. M. Clemens, and R. Sinclair, “TEM analysis

of Co-Gd and Cr-Gd multilayer strutures,” Journal of Materials Research,
vol. 8, p. 771, 1992.

[71] D. Raiser and J. C. Sens, “Study of the diffusion at the Gd/Cu interface,”
Applied Surface Science, vol. 55, p. 277, 1992.

[72] A. V. Svalov, A. Fernandez, V. Vaskovskiy, M. Tejedor, R. L. Anton,

J. Barandiarán, and G. Kurlyandskaya, “Interlayer coupling and magnetic
compensation in Co/Ti/(Gd-Co)/Ti multilayers,” Proceedings of the 3rd
Moscow intnational Symposium on Magnatism, 2005.
Bibliography 116

[73] D. Michels, C. Krill, and R. Birringer, “Grain-size-dependent Curie transition

in nanocrystalline Gd: the influence of interface stress,” JMMM, vol. 250,
pp. 203–211, 2002.

[74] J. A. Gonzales, J. P. Andres, M. A. Arranz, M. A. L. de la Torre, and J. M.

Riveiro, “Interdiffusion and magnetic properties of Gd1−x Cox /Co multilay-
ers,” J. of Phys.: Condensed Matter, vol. 14, pp. 5061–5066, 2002.

[75] T. Ohkochi, N. Hosoito, and K. Mibu, “Depth-selective measurements of

induced magnetic polarization in Cu layers of Gd/Cu multilayers by 119 Sn
Mössbauer spectroscopy,” J. of Phys.: Condensed Matter, vol. 17, pp. 4023–
4033, 2005.

[76] R. Ranchal, C. Aroca, M. S. P. S´anchez, and E. Lopez, “Improvement of

the structural and magnetic properties of Permalloy/Gadolinium multilayers
with Mo spacers,” Appl. Phys. A, vol. 82, pp. 697–701, 2006.

[77] Y.-H. Fan and H. Brückl, “Magnetic moment compensation in exchange-

biased trilayers with antiparallel spin alignment,” Appl. Phys. Lett., vol. 83,
no. 15, pp. 3138–3140, 2003.

[78] L. T. Baczewski, “Induced magnetic moment of V atoms in ultra-thin epi-

taxial V/Gd bilayers.”

[79] K. Mergia, L. Baczewski, S. Messoloras, S. Hamada, T. Shinjo, H. Gamari-

Seale, and J. Hauschild, “Polarized neutron reflectivity study of a Gd/Cr
multilayer,” Appl. Phys. A, vol. 74, no. Suppl., 2002.

[80] T. Trappmann, M. Gajdzik, C. Sürgers, and H. v. Löhneysen, “Scanning tun-

neling spectroscopy of rare-earth metals: Gadolinium on yttrium,” Europhys.
Lett., vol. 39, no. 2, pp. 159–164, 1997.

[81] C. Surgers and H. V. Löhneysen, “Growth and characterization of Nb/Gd

multilayers for different substrate temperatures,” Thin Solid Films, vol. 218,
no. 1–2, pp. 69–79, 1992.

[82] C. Waldfried, D. N. McIlroyy, and P. A. Dowben, “The electronic struc-

ture of gadolinium grown on Mo(112),” Journ. Phys.: Cond. Matt., vol. 9,
pp. 10615–10638, 1997.

[83] I. Zoto, Magnetic Properties of Coupled Bilayers and Trilayers Thin Films.
PhD thesis, University of Alabma, 2006.

[84] S. Poon and J. Durand, “Critical phenomena and magnetic properties of

an amorphous ferromagnet: Gadolinium-gold,” Phys. Rev. B, vol. 16, no. 1,
1977.

[85] D. LaGraffe and P. A. Dowben, “Magnetic ordering of thin Gd overlayers,”

Phys. Rev. B, vol. 40, no. 2, 1989.
117 Bibliography

[86] G. J. Bowden, J.-M. L. Beaujour, A. A. Zhukov, B. D. Rainford, and P. A. J.

de Groot, “Modeling the magnetic properties of DyFe2 /YFe2 superlattices,”
Jour. of Appl. Phys., vol. 93, no. 10, pp. 6480–6482, 2002.

[87] D. Weller, S. F. Alvarado, W. Gudat, K. Schröder, and M. Campaga, “Ob-

servation of suface-enhanced magnetic order and surface reconstruction on
Gd(0001),” Phys. Rev. Lett., vol. 54, no. 14, pp. 1555–1558, 1985.

[88] K. Takanashi, H. Kurokawa, and H. Fujimori, “A novel hysteresis loop and

indirect exchange coupling in Co/Pt/Gd/Pt multilayer films,” Appl. Phys.
Lett., vol. 63, pp. 1585–1587, 1993.

[89] T. Pan, G. W. D. Spratt, L. Tang, L.-L. Lee, Y. Feng, and D. E. Laugh-

lin, “Temperature dependence of coercivity in co-based longitudinal thin-film
recording media,” J. Appl. Phys., vol. 81, no. 8, pp. 3952–3954, 1997.

[90] M. Mizumaki, K. Yano, I. Umehara, F. Ishikawa, K. Sato, A. Koizumi,

N. Sakai, and T. Muro, “Verification of Ni magnetic moment in GdNi2 laves
phase by magnetic circular dichroism measurement,” Phys. Rev. B, vol. 67,
pp. 132404–132407, 2003.

[91] N. Ishimatsu, H. Hashizmue, S. Hamada, N. Hosoito, C. S. Nelson, C. T.

Venkataraman, G. Srajer, and J. C. Lang, “Magnetic structure of Fe/Gd
multilayers determined by resonant X-ray magnetic scattering,” Phys. Rev.
B, vol. 60, no. 13, pp. 9596–9606, 1999.

[92] H. Nagura, K. Tkanashi, S. Mitani, K. Saito, and T. Shima, “Current-

perpendicular-to-plane magnetoresistance in Gd/Co multilayers with twisted
spin structure,” JMMM, vol. 240, pp. 183–185, 2002.

[93] K. Cherifi, C. Dufour, P. Bauer, G. Marchal, and P. Margin, “Experimental

magnetic phase diagram of a gd/fe multilayered ferrimagnet,” Phys. Rev. B,
vol. 44, no. 14, pp. 7733–7736, 1991.

[94] T. B. Massalski, Binary Alloy Phase Diagrams, vol. 3. USA: ASM

International, second edition ed., 1990.
Bibliography 118
Danksagung

Zum Abschluß dieser Arbeit möchte ich allen danken, die zu ihrem Gelingen
beitrugen und mich durch ihre tatkräftige Mitarbeit oder durch interessierte Dis-
kussion unterstützten. Im Folgenden möchte ich einige davon speziell erwähnen.

Herrn Prof. Günter Schatz gilt mein besonderer Dank. Die kollegiale Atmo-
sphäre in der Arbeitsgruppe wird nicht zuletzt von den allmorgendlichen Kaffee-
runden getragen, sondern auch von seinem guten Zuspruch in Zeiten, in denen es
gar nicht gut läuft. Und davon kann es im Laufe einer Doktorarbeit einige haben.
Die auf der einen Seite lockeren aber dafür stetigen Diskussionen über Ergebnisse
und Fortschritte im Labor sind eine große Hilfe gewesen und liessen viele neue
Ideen entstehen.

Prof. Dr. Manfred Albrecht möchte ich für die zahlreichen Diskussionen und
Anregungen danken, die dieser Arbeit eintscheidende Impulse gegeben haben und
natürlich für das Korrektur lesen. Für seine Zukunft in Chemnitz als frischgeba-
ckener Professor wünsche ich ihm weiterhin viel Erfolg.

Meinem Vorgänger an der MEDUSA, Frank Treubel, danke ich für die gu-
te Einarbeitung in die Bedienung dieser manchmal sehr zickigen Dame, doch zu
Stein musste zum Glück noch niemand erstarrren. Außerdem bedanke ich mich
für die Einarbeitung bei der Betreuung des SQUIDs, aber auch außerhalb der
Universität für die leider einseitigen Tennis-Matches, die gemeinsamen Besuche
von Tanzbällen und -kursen.

Dr. Marta Marszalek und Prof. Bill Evenson gilt mein Dank für die gemein-
same Vorbereitung einer Veröffentlichung über das Gd/Ni-Bilagensystem. Dabei
sind die grundlegenden Ideen zur Interpretation vieler Daten entstanden. Au-
ßerdem möchte ich Bill Evenson im Besonderen für den Feinschliff an meinem
Englisch danken.

Dr. Olav Hellwig muss hier ausdrücklich erwähnt werden für die Unterstüt-
zung und Organisation der Beamtime am BESSY. Seine Expertise erst führten
zu den XMCD-Messungen, die in dieser Arbeit präsentiert wurden.
Danksagung 120

Für Messungen mittels RBS danke ich Prof. Dr. Hans Hofsäss, für die TEM-
Bilder Richard Vanfleet und Dr. Mireille Maret für die Röntgen-Kleinwinkelmessungen.

Meinen weiteren Bürokollegen und Mitgliedern der AG Schatz möchte ich für
die tolle Zusammenarbeit danken. In einer so langen Zeit ist ein gutes, kollegiales
Arbeitsklima, wie ich es hier vorgefunden habe, unerlässlich.

Ein ganz spezieller Dank geht an den Hochschulsport. Nicht nur , dass er
dafür sorgte, dass ich das Mensaessen wieder abstrampeln konnte, sondern auch
für die Organisation von folgenschweren Skifreizeiten.

Zum Schluss noch ein Dankeschön an meine Familie, die mich während mei-
ner Promotion immer unterstützte und mir erst das Studium der Physik ermög-
lichten.
Appendix A

Properties of the GdNi-system

Figure A.1: Phase diagram of the binary alloy system Gd-Ni [94].
Appendix A. Properties of the GdNi-system 122

Figure A.2: Dependence of the

Curie temperature of GdNi al-
loys on the composition [16].
Appendix B

Depth profile by RBS

Shortly after the submission of this thesis it was possible to transform the RBS
energy spectrum in figure 4.13 into a depth profile. The following figure shows
the chemical composition of a cross-section for the sample A5 on Al2 O3 . The
nominal layer structure is Gd(50Å)/Ni(100Å)/Au(20Å) and it was deposited at
100 K.

Figure B.1: RBS depth profile: Converted spectra from figure 4.13 into a
position dependent chemical composition of a cross-sectional view.
View publication stats