-
UPPSALA UNIVERSITY
MASTER THESIS
Magnetic Coupling and TransportProperties of Fe/MgO
Superlattices
Author:Tobias Warnatz
Supervisor:Dr. Fridrik Magnus
Materials PhysicsDepartment of Physics and Astronomy
Thesis Number:FYSMAS1036
Series:FYSAST
September 8, 2015
1
-
Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . 7
2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . 92.1 Sample Preparation . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . 92.2 X-Ray
Reflectivity . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . 112.3 Polarized Neutron Reflectivity . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . 142.4 Magneto-Optical Kerr Effect
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . 162.5 Four-Terminal
Sensing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . 17
3 Results and Discussion . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . 183.1 Structural
Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . 183.2 Magnetic Properties . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . 203.3 Magnetic
Ordering . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . 293.4 Magnetotransport . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4 Conclusions and Outlook . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . 42
5 Acknowledgment . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
References . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . 45
3
-
SammanfattningMagnetiska strukturer med multilager är
intressanta både för applikationer (hårddiskar, magnetfältssen-
sorer etc.) och för fundamental forskning (exempelvis koppling
mellan magnetiska lager). Material medpassande gitterparametrar kan
tillväxas epitaxiellt och forma ett supergitter. Den mycket
kristallina kvalitetetenav såna materialprover kan leda till en
bättre förståelse av den magnetiska kopplingsmekanismen och
äventill uppkomsten av helt nya magnetiska fenomen.
I den här avhandlingen används Fe/MgO supergitter som skapats
via en kombination av likströms-och
radiofrekvens-katodförstoffning. Tjockleken hos de omagnetiska
lagren av MgO varierades för attundersöka kopplingsmekanismen. Den
strukturella karaktäriseringen utfördes med
röntgenreflektivitetetoch röntgendiffraktion. Mätningarna med
röntgenreflektivitet anpassades med GenX för att erhålla
exaktavärden på lagrens tjocklek och ojämnhet. En hög
kristallinitet och mycket jämna gränsytor mellan lagrenerhölls för
samtliga tjocklekar på MgO-lagren.
Den longitudinella magnet-optiska Kerreffekten användes för att
studera supergittrens magnetiska egen-skaper. Ovanliga steg i
hystereskurvan erhölls för alla prover. Vi fann att varje enskilt
steg svarade mot att ettindividuellt Fe-lager bytte riktning.
Kopplingen favoriserar en antiparallell linjering i rumstemperatur
ochkopplingens styrka beror kraftigt på MgO-lagrets tjocklek.
Kopplingsmekanismen kommer från utbytetmellan lagren genom
spinnpolariserad kvanttunnling. Dock avslöjade temperaturberoende
mätningar enmagnetisk koppling assisterad av materialorenheter vid
höga temperaturer och även av kopplingsbeteendetav en perfekt
tunnelförbindelse vid låga temperaturer.
Genom mätningar med polariserad neutronreflektivitet var det
möjligt att bekräfta en periodisk, vinkel-rät och antiparallell
linjering i de ferromagnetiska lagren för de tjocka respektive
tunna lagren. Den vinkel-räta kopplingen verkar vara ett resultat
av kampen mellan svagare, antiparallell koppling och den
magne-tokristallina anisotropin.
Mätningar av elektron-transport i planet påvisade steg i
magnetoresistansen som svarade mot stegenuppmätta i hystereskurvan
huvudsakligen genom den anisotropa magnetoresistanseffekten. Dock
upp-mättes ett svagt bidrag från tunnelmagnetoresistanseffekten
vilket gör att dessa prover också är lovandeför framtida mätningar
av elektron-transport ur planet.
AbstractMagnetic multilayer structures are interesting for
applied science (hard-drives, magnetic field sensor
etc.) as well as for fundamental research (coupling mechanism of
the magnetic layers). Materials withsuitable lattice parameters can
be epitaxially grown to form a superlattice. The high crystalline
quality ofthose samples may lead to a better understanding of the
coupling mechanism as well as to the emergenceof novel
phenomena.
In this thesis, Fe/MgO superlattices were grown via a
combination of direct-current and radio-frequencysputtering. The
thickness of the MgO spacer layers was varied to investigate the
coupling mechanism.The structural characterization was done via
x-ray reflectivity and x-ray diffraction. The x-ray
reflectivitymeasurements were fitted via GenX to obtain precise
thickness and roughness values of the layers. A highcrystalline
quality and very smooth interfaces were found for all MgO
thicknesses.
The longitudinal magneto-optical Kerr effect was used to study
the magnetic properties of the super-lattices. Unusual steps in the
hysteresis curve were found in all samples. It was found that each
stepcorresponds to the switching of an individual Fe layer. The
coupling favors an antiparallel alignment atroom temperature and
its strength highly depends on the spacer layer’s thickness. The
coupling mechanismwas assigned to the interlayer exchange by
spin-polarized quantum tunneling. However, temperature depen-dent
measurements revealed an impurity assisted coupling at high
temperatures and the coupling behaviorof a perfect tunnel junction
for low temperatures.
Through polarized neutron reflectivity measurements, it was
possible to confirm a periodic, perpendicu-lar and antiparallel
alignment of the ferromagnetic layer for thick and thin MgO spacer
layers, respectively.The perpendicular coupling seems to be a
result of the competition between weaker, antiparallel couplingand
magnetocrystalline anisotropy.
In-plane transport measurements revealed steps in the
magnetoresistance corresponding to the steps inthe hysteresis curve
mainly due to the anisotropic magnetoresistance effect. However,
also a small con-tribution of the tunnel-magnetoresistance effect
was measured making these samples promising for futureout-of-plane
transport measurements.
5
-
1. Introduction
In 1986, Peter Grünberg demonstrated antiferromagnetic coupling
of Felayers in Fe/Cr/Fe multilayers [1], which led subsequently to
the discoveryof the giant magnetoresistance effect (GMR). Due to
the giant resistance dif-ference observed for parallel and
antiparallel alignment of the ferromagneticlayers, the effect was
quickly adapted in commercial devices for magnetic fieldsensing
(e.g. in hard drives). In 2006, a multi-stepwise reversal of
magneticlayers in Fe/Cr/Fe superlattices was observed leading to a
partitioned GMReffect [2]. Such structures are of particular
interest since not only two binarystates (high and low resistance),
but also intermediate resistance values couldbe used for reading
out digital information leading to a higher storage
density.However, this publication raised little attention mainly
because the resistancechange in this partitioned GMR effect was
rather small (max. 2.4% at 300K)making it impractical for
commercial devices.
Even though it was already proposed in 1975 [3] and
experimentally provenin 1995 [4] that a magnetoresistance effect
can also be achieved with an insu-lating spacer layer it was not
until 2004 [5, 6] that Fe/MgO/Fe junctions raisedmajor attention.
This was because the achieved magnetoresistance effect wasmuch
larger than any other reported effect through metallic or
non-metallicspacer layers. Optimization in the growth led to
astonishing magnetoresistancevalues of 500% at room temperature
[7]. Hence, the tunnel-magentoresistanceeffect (TMR) replaced the
GMR in most devices and even led to other noveldevices (like MRAMs
[8]). The size of the TMR effect in single Fe/MgO/Fejunctions opens
up the question whether a significant partitioned TMR can
beobtained in MgO-based multilayers.
A partitioned TMR effect can only be achieved if an interlayer
coupling ex-ists through the insulating layer. Replacing the
metallic Cr spacer layer withinsulating MgO eliminates the coupling
mechanism discovered by Grünberg.However, a different coupling
mechanism occurs. This coupling mechanismwas first proposed in 1989
by a method based on spin current [9]. A laterapproach described
the coupling based on quantum interference due to spin-dependent
reflections at the interfaces [10, 11]. The experimental proof
ofthis interlayer exchange by spin-polarized quantum tunneling came
in 2002[12]. In contrast to the
ferromagnetic-antiferromagnetic-oscillating couplingof metallic
spacer layers depending on the spacer layer thickness, an
insu-lating layer exhibits a non-oscillating coupling behavior.
However, similarto the metallic spacer layer one observes an
exponential decay in the cou-pling strength with increasing spacer
layer thickness [13]. Recently, we have
7
-
demonstrated the possibility for a partitioned TMR by a
multi-stepwise rever-sal in Fe/MgO/Fe superlattices [14]. However,
even though the main couplingmechanism was attributed to the
aforementioned spin-polarized quantum tun-neling, it was not
possible to be certain about the main coupling mechanism.
In the present work, high quality MgO[Fe/MgO]10Pd superlattices
with avariation in the MgO thickness have been prepared to
distinguish betweendifferent coupling mechanisms. Structural
characterization was done usingx-ray reflectivity (chapter 3.1),
magnetic characterization was done using themagneto-optical Kerr
effect (chapter 3.2) and polarized neutron reflectivity(chapter
3.3) was used to determine the magnetic ordering of
ferromagneticlayers. Finally, in-plane magneto-transport
measurements were performed toinvestigate the magnetoresistance.
Furthermore, samples with a comparablehigh quality grown on SrTiO3
will be characterized to investigate if such sub-strates can be
used for out-of-plane TMR measurements in Fe/MgO/Fe
super-lattices.
8
-
2. Methods
In this chapter basic methods will be explained in order to
understand andinterpret the presented results in chapter 3.
2.1 Sample PreparationA structure consisting of two or more
discrete layers grown with different
materials is called a multilayer. If the materials have a
comparable latticeparameter, a superlattice can be formed. A
superlattice is defined by a longout-of-plane structural coherence
[15] (atomic registry). The lattice mismatchbetween Fe (2.866 Å
[15]) and MgO (4.213 Å [15]) at first sight appears to betoo big
(47.0%) to form a superlattice, but a 45◦ rotation of the Fe layer
on topof MgO (fig. 2.1) leads to a sufficient small lattice
mismatch of only 3.94%.An obvious choice of substrate for such a
superlattice is single crystallineMgO (100). Another suitable
material in terms of lattice matching wouldbe SrTiO3. The 45◦
rotation of Fe on top of this material results in an evensmaller
lattice mismatch (3.65%). Furthermore, this material can be
dopedwith small amounts of Nb to become conductive. By using doped
SrTiO3 as asubstrate for the Fe/MgO superlattices, out-of-plane
transport measurementscan be performed.
Fe [1
00]
MgO [100]
45°
Easy
Axis
Hard Axis
FeMg O
Figure 2.1. Schematic illustration of the tetragonal Fe/MgO
structure formed by a 45◦
rotation of the Fe layer on top of MgO.
9
-
Table 2.1. Spacer layer sputtering times and thicknesses.
Sample MgO Sputtering Time in s MgO Thickness in ÅA 450 19.9B
420 17.4C 350 16.7D 300 14.6
Pd- Capping
MgO
Fe
MgO Substrate
~45 Å {~15-20 Å {~21-23 Å {
~1 mm {} x10
Figure 2.2. Schematic illustration of the superlattice samples
(side view).
The Fe and MgO layers were deposited with direct-current (DC)
and radio-frequency (RF) magnetron sputtering, respectively. In
order to ensure the highquality of the samples, the chamber’s base
pressure was below 2· 10−9 Torrand Ar with a purity of 99.99999%
was used as a sputtering gas. The 1 mmthick MgO (100) and SrTiO3
(100) 10 mm x 10 mm substrates were annealedfor 40 minutes at
550◦C. After that, the temperature was reduced to the
growthtemperature of 165◦C [16]. Once the temperature had
stabilized, Fe was sput-tered (epitaxially grown) for 75 s with a
power of 50 W and an Ar pressureof 2.00 mTorr. MgO was RF-sputtered
with a power of 60 W and the sameAr pressure. The Fe magnetron was
turned off during the MgO sputteringand vice versa, to avoid
cross-contamination. This process was repeated 10times before Pd
was deposited for 30 s with a power of 50 W and the sameAr pressure
as a capping layer (fig. 2.2). In order to study the interlayer
cou-pling of the ferromagnetic (Fe) layers, different MgO layer
thicknesses wereprepared (tab. 2.1). High resolution transmission
electron microscopy images(HRTEM) were taken to investigate the
quality of the samples, as shown infigure 2.3. A nice layering is
observed, indicating very smooth surfaces aswell as an atomic
registry through the whole sample.
10
-
Figure 2.3. HRTEM cross section image of a representative Fe/MgO
(bright /darklayers) superlattice grown on a MgO (100) substrate.
Very smooth interfaces (a) anda structural registry of the
individual layers with respect to the single crystalline sub-strate
lattice (b) can be observed.
2.2 X-Ray ReflectivityX-ray reflectivity is a powerful tool for
investigating thin films as well as
multilayers. The experimental data can be used to obtain
thickness, roughnessand other sample specific values [17]. The
x-ray source emits polychromaticphotons (Bremsstrahlung and
characteristic x-rays). In order to perform x-rayreflectivity (XRR)
measurements, monochromatic x-rays are needed (here:copper Kα ). A
grid can be used to select the desired wavelength. Due
tointerference effects at the grid, only photons with a certain
wavelength are re-flected. In general, it does not matter if the
grid (monochromator) is placedbefore or after the sample. A slit in
front of the x-ray source is used to reducethe beam’s divergence,
but reduces the beam’s intensity simultaneously. Thechosen slit
size usually depends on the probed structure and the desired
reso-lution. A 2θ -ω-scan in a standard Bragg-Brentano-geometry was
used for allthe measurements. The sample is rotated at a rate ω and
the detector is rotatedwith a rate of 2θ , which is twice that of ω
. In this case, it is ensured that theangle α between the source
and the sample and between the sample and thedetector is always the
same (fig. 2.4). XRR measurements start at very smallangles. Thus,
total reflection occurs up to the critical angle αc. After
that,ordinary reflection progresses to lattice planes, at which
Bragg’s law can beused
n ·λ = 2d · sin(θ) (2.1)with n as a positive integer number, λ
as the incident wavelength, d as thelattice plane distance and the
reflection angle θ . X-rays scatter at the inter-faces between the
layers due to the variation of electron densities
(differentmaterials). Reflections from different interfaces will
lead to a path difference
11
-
Substrate
X-RaySource Slit
Sample
Detec
tor
(α( α
ω
2θ
Figure 2.4. Schematic illustration of a XRR setup.
between the reflected waves (fig. 2.4). Constructive
interference occurs only ifthe reflected waves are still in phase.
Hence, Bragg reflections can only be ob-served if the relation in
equation 2.1 is obtained. By knowing the wavelengthof the incoming
photons and the reflection angle, it is possible to calculate
thespacing between different layers.
It is convenient to plot the XRR data over the momentum transfer
Q insteadof the angle θ , since Q includes the used wavelength and
different data sets(e.g. neutron and x-ray reflection) can be
compared. The momentum transferQ can be calculated via the
following equation.
Q =4πλ
· sin(θ). (2.2)
A special case occurs if XRR measurements are performed on
multilayersamples. A multilayer sample consists of two or more
discrete layers period-ically arranged through the whole sample.
Due to the chemical modulationin the sample, the Bragg condition
can be satisfied for the multilayer stack(Bragg peaks) and for the
thickness of the whole sample (Kiessig fringes) [15]as illustrate
in figure 2.5. By measuring the distance between two
neighboringBragg Peaks (or Kiessig oscillations) one can calculate
the bilayer (or total)thickness. If the XRR data is plotted over Q,
the measured distance corre-sponds to 2πd with the bilayer (or
total) thickness d. It is worth mentioning thatonly the bilayer
thickness and not the thickness of the individual layers can
beobtained by measuring the Bragg peaks spacing. The width and
height of theBragg peaks correspond to the thickness of the
individual layers. By using afitting software (e.g. GenX [18]) it
is possible to receive information about theinterface roughness,
the thickness of the individual layers and variations in
thematerials densities (due to stress/strain or vacancies). As
already mentionedabove, x-rays are only sensitive to a variation in
the electronic density. Hence,
12
-
0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 1 . 2 1 . 4 1 . 61 E - 8
1 E - 7
1 E - 6
1 E - 5
1 E - 4
1 E - 3
0 . 0 1
0 . 1
1C r i t i c a l A n g l e
K i e s s i g O s c i l l a t i o n s
Inten
sity (a
rb. un
its)
Q ( 1 / Å )
B r a g g P e a k s
Figure 2.5. GenX simulation of a MgO[Fe(52 Å)/MgO(11 Å)]10Pd(20
Å) superlatticewith an interface roughness of 2 Å (Fe) and 3 Å
(MgO).
it is not possible to distinguish between layers made of
different isotopes sincethe electronic density would be the
same.
If higher angles (Q-values) are chosen to illuminate the sample,
diffrac-tion occurs. The measured peaks correspond then to the
lattice parametersof the materials within the sample. The lattice
parameters (d) can be cal-culated via equation 2.1. A special case
occurs, if multilayers with a highout-of-plane structural coherence
(atomic registry) are measured. In this case,satellite peaks (also
known as superlattice peaks) occur next to the principlepeaks.
13
-
Figure 2.6. Schematic illustration of the PNR geometry [19]. The
angles αi and α findicate the angles of the incoming and reflected
neutron beam, respectively. γ is theangle between the magnetization
and the polarization of the incoming neutrons, whereγ = 0◦ and 90◦
correspond to a sample’s magnetization parallel to the y-axis (up)
andx-axis (right), respectively.
2.3 Polarized Neutron ReflectivityNeutron reflectivity
measurements exploit the wave character of neutrons
and are therefore similar to XRR measurements. However, neutrons
are sen-sitive to variations in the nuclear (not electronic)
densities, which makes thecombination of both techniques a powerful
tool for characterizing samples.One of the biggest advantages of
neutrons for the sample characterization istheir spin. Hence, it is
possible to probe unpaired electron spins (magneticmoments) in the
sample. This technique is called polarized neutron reflec-tivity
(PNR) and can be used to probe the direction of the magnetization
ineach layer as a function of thickness (depth dependent). A
distinction is madebetween spin-flip (SF) and non-spin-flip (NSF)
measurements. NSF measure-ments can probe a variation in nuclear
density and a magnetization parallel tothe incoming neutron’s spin
(fig. 2.6).
The NSF measurement is divided into up-up (UU, spin of incoming
neu-trons up, measured neutrons up) and down-down (DD, spin of
incoming neu-trons down, measured neutrons down). Both measurements
are sensitive to avariation in nuclear density. However, the
contribution of the magnetic partvaries for both channels depending
on the orientation of the magnetization(parallel or antiparallel)
to the incoming neutrons spin as illustrated in figure2.6 and
described in the following equations.(
Vuu VudVdu Vdd
)=
2πh2
mN[(
bn 00 bn
)+
(by bxbx −by
)](2.3)
bx = bm · sinγ (2.4)
by = bm · cosγ. (2.5)
14
-
V are the neutron scattering potentials, bn is the nuclear
scattering length, bmis the magnetic scattering length and γ is the
angle between the magnetizationand the polarization of the incoming
neutrons. A magnetization along the z-axis (along the momentum
transfer Qz) cannot be measured via PNR.
The SF measurement is divided in up-down (UD, incoming neutrons
up,measured neutrons down) and down-up (DU, incoming neutrons down,
mea-sured neutrons up). SF measurements are not sensitive to a
variation in thenuclear density, but it is possible to probe the
sample’s magnetization per-pendicular to the spin of the incoming
neutrons. Hence, the combination ofSF and NSF measurements is a
powerful method to reveal the magnetiza-tion of individual magnetic
layers perpendicular to the scattering plane. BothSF channels are
usually identical and lead to the same information. In somespecial
cases (e.g. magnetic chirality) both SF channels may differ and
leadto additional information, but these cases are beyond the scope
of this thesisand discussed elsewhere [19]. Figure 2.7 shows the
scattering length den-sity (SLD) profile of a Fe/MgO superlattice.
Since Fe and MgO are differentmaterials, one observes the same
periodicity in the nuclear and electron den-sity. Furthermore,
since the magnetization of the layers are aligned parallel
toneutrons guide field, one observes a magnetic periodicity
identical to the nu-clear and electronic one. The biggest
difference (contrast) is exhibited for theelectron density. The
nuclear density difference is low and hardly gives anysignal.
However, the magnetic contrast is very pronounced illustrating that
acombination of both techniques is very powerful.
Figure 2.7. XRR (green), nuclear UU (red) and magnetic UU (blue)
SLD for a[Fe(21.6 Å)/MgO(16.7 Å)]10Pd superlattice with a
magnetization along the neutronsguide field.
15
-
Mα α
Mα α
Mα α
Polar Longitudinal
Transverse
Figure 2.8. Schematic illustration of the three MOKE
geometries.
2.4 Magneto-Optical Kerr EffectThe Magneto-optical Kerr effect
(MOKE) is widely used to measure hys-
teresis curves as well as to image magnetic domains (Kerr
microscopy) ofmetallic surfaces [20]. Magnetic thin films, as well
as particles or other sys-tems can be investigates. It can be
separated into three different geometries.The polar (P-MOKE),
longitudinal (L-MOKE) and transverse (T-MOKE) asillustrated in
figure 2.8, where the magnitude of the effect decreases in thesame
order.
Figure 2.9 illustrates the experimental setup of a L-MOKE
measurement.To exploit the L-MOKE effect, linear polarized light
has to be used. In thiscase, the electric wave vector of the
incident light is normal to the scatter-ing plane (s-polarized).
The electric field of the incident light couples to theunpaired
electron spins in the sample’s magnetic material due to
spin-orbitinteraction [15]. Hence, the reflection from a magnetic
sample will lead toa rotation of the polarization (the electric
wave vector is not anymore per-fectly s-polarized) and ellipticity
corresponding to the magnetization of thesample. By using a second
polarizer in front of the detector, which only al-lows p-polarized
light to pass, the Kerr rotation can be detected. By applyingan
alternating, magnetic field parallel to the surface of the sample
(collinearto the magnetization M in the longitudinal MOKE setup in
figure 2.8) one canmeasure the sample’s hysteresis loop. L-MOKE is
only sensitive to a magneti-zation collinear to the direction of
the incident light (fig. 2.9). A magnetizationperpendicular to its
direction won’t lead to a Kerr rotation and can thereforenot be
measured by a L-MOKE setup. However, by an in-plane rotation of
thesample, it is possible to measure the magnetization collinear to
the directionof the incident light for e.g. magnetic hard and easy
axes.
16
-
H
Electromagnets
Detector Laser
Polarizer 1Polarizer 2
Sample
Figure 2.9. Schematic illustration of the L-MOKE setup.
A
V
Rwire
Rwire
Rwire
Rwire
Rsample
Figure 2.10. Schematic illustration of the four-terminal sensing
setup with the am-peremeter A and the voltmeter V.
2.5 Four-Terminal SensingThe four-terminal sensing method is
used to separate lead and contact re-
sistance from the sample’s resistance. The sample is contacted
by four wiresas shown in figure 2.10. The outer wires are used to
apply a constant current.The inner wires are used to measure the
potential. The internal resistance ofthe voltmeter (around 10 MΩ)
is much higher than the sample’s resistance.Hence, the current
flows mainly through the sample and the measured voltagemostly
depends on the potential drop in the sample and not the potential
dropin the wires and contacts. A precise value of the sample’s
resistance can thenbe obtained by dividing the measured voltage by
the applied current.
17
-
3. Results and Discussion
3.1 Structural PropertiesXRR measurements have been used to
determine the structural quality of
the samples. The XRR measurement shown in figure 3.1 is
representative forall samples and illustrates the high degree of
perfection in the sample’s layer-ing. The Bragg peak position of
the experimental data and fit overlap, whichmakes it possible to
determine the bilayer (one Fe and one MgO layer) thick-ness. Since
the width and height of these peaks are captured by the fit as
well,the bilayer thickness can be divided into individual thickness
values for Feand MgO leading to the values presented in table 2.1.
As expected from theHRTEM images (fig. 2.3), very smooth interfaces
with a roughness of only0.5-1 monolayers (tab. 3.1) were obtained
from the fitting. The atomic densi-ties were first set to
literature values ( formula unitsunit cell volume ) and then varied
by 10% tocompensate for variation of the unit cell volume due to
stress, strain or vacan-cies. The XRD measurements (fig. 3.1 inset)
show satellite peaks around theFe (002) peak. Those peaks are
called superlattice peaks and only appear whenthere is a high
structural coherence normal to the layers, as already
ascertainedfrom the HRTEM images.
The same measurements were performed on a [Fe/MgO]9Fe
superlatticegrown on SrTiO3 with the exact same conditions as the
sample above. Inthis sample, the superlattice was terminated with a
Fe layer instead of a MgOlayer to avoid the possibility of island
growth of the Pd capping layer. Pd isknown to form large islands on
top of MgO which could result in the cappinglayer not fully
covering the superlattice. However, this affects neither the
mag-netic nor the structural properties of the sample. The quality
of the layering(XRR data in fig. 3.2) is comparable to the layering
of the samples grown onMgO substrates shown above (fig. 3.1).
Furthermore, superlattice peaks canbe observed in the XRD spectrum
indicating the formation of a superlattice.However, the Fe (002)
peak is shifted by 0.7◦ in 2θ (further away from thebulk lattice
constant) in the sample grown on SrTiO3. One can conclude, thatthe
different substrate leads to differences in stress/strain causing
this shift.
Table 3.1. Layer roughness from the GenX fit for Sample A.
Layer Roughness in ÅMgO Substrate 1.2
Fe 2.4MgO 1.8
Pd 4.8
18
-
0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 01 01
1 0 2
1 0 3
1 0 4
1 0 5
1 0 6
1 0 7
1 0 8
5 0 5 5 6 0 6 5 7 0 7 51 00
1 0 1
1 0 2
1 0 3
Inten
sity (
arb.
units
)
2 T h e t a ( d e g r e e )
F e ( 0 0 2 )
X R R D a t a G e n X F i t
Inten
sity (a
rb. un
its)
Q ( 1 / Å )
Figure 3.1. Experimental XRR data (black dots) of sample A
[Fe(20.9 Å)/MgO(19.9Å)]10 and GenX fit of the same data set (red
line). The experimental XRD data (inset)exhibits superlattice peaks
indicating a high structural coherence.
0 . 2 0 . 4 0 . 6 0 . 8 1 . 0
1 0 2
1 0 3
1 0 4
1 0 5
1 0 6
1 0 7
1 0 8
5 5 6 0 6 5
1 0 2
1 0 3
Inten
sity (
arb.
units
)
2 T h e t a ( d e g r e e )
F e ( 0 0 2 )
Inten
sity (a
rb. un
its)
Q ( 1 / Å )
Figure 3.2. Experimental XRR data (black dots) of a [Fe/MgO]9Fe
superlattice grownon SrTiO3. The experimental XRD data (inset)
exhibits superlattice peaks indicatinga high structural
coherence.
19
-
Easy Axes Hard Axes
Figure 3.3. Schematic illustration of Fe’s four-fold anisotropy
(topview) with two easyaxes (dashed, orange lines) and two hard
axes (solid, black lines).
3.2 Magnetic PropertiesFe intrinsically has a magnetocrystalline
anisotropy. In the case of bcc Fe,
the easy axis is along the (100) direction and the hard axis is
along the (110)direction. Hence, the magnetocrystalline anisotropy
is four-fold symmetric.Since Fe is rotated by 45◦ (fig. 2.1) on top
of MgO, the easy axes point alongthe diagonals and the hard axes
along the sides as illustrated in figure 3.3. Itis worth to
mention, that Fe’s four-fold anisotropy in thin films can only
beobserved if a sufficiently good crystalline quality is
ensured.
The four-fold anisotropy can be verified by measuring the
sample’s mag-netic response to an applied, alternating, magnetic
field along both axes (fig.3.4). If a strong, external field is
applied along the hard axis (e.g. to the lefthand side of figure
3.4a inset), all magnetic moments are aligned along theexternal,
magnetic field as seen by the saturation of the sample’s
magnetiza-tion (around -90 mT for sample A). When reducing the
external field, a steadydecrease of the magnetization can be
observed. Reducing the field leads to acoherent rotation of the
magnetization towards the easy axes (e.g. upper andlower left hand
side corner in figure 3.3). At zero external field, a
remainingmagnetization (remanence) can still be observed. In this
case, all magneticmoments are aligned along the easy axes (e.g.
upper and lower left hand sidecorner in fig. 3.3) leading to a
component pointing still along the hard axis(left hand side). By
calculating the resulting contribution M = Ms · cos45◦,where Ms is
the saturation magnetization, one gets Mr = 0.7Ms, which equalsto
the observed remanent magnetization. Reversing the field leads to
an abruptjump in the hysteresis curve, which can be attributed to a
flipping of the mag-netization over to the easy axes in the other
direction (e.g. upper and lowerright hand side corner in figure
3.3). By increasing the external field further,a coherent rotation
of the magnetization towards the hard axis occurs until
thesaturation can finally be observed again.
By applying a magnetic field along the easy axis (fig. 3.4b
inset) one ob-serves a much lower saturation field compared to the
hard axis as well as
20
-
- 2 0 0 - 1 0 0 0 1 0 0 2 0 0
- 1 . 0
- 0 . 5
0 . 0
0 . 5
1 . 0
- 2 0 - 1 6 - 1 2 - 8 - 4 0 4 8 1 2 1 6 2 0
- 1 . 0
- 0 . 5
0 . 0
0 . 5
1 . 0
Magn
etiza
tion (
norm
alized
)
B ( m T )
a )
Magn
etiza
tion (
norm
alized
)
B ( m T )
b )
Figure 3.4. Normalized, experimental L-MOKE data of sample A
[Fe(20.9Å)/MgO(19.9 Å)]10 along the hard (a) and easy (b) axis.
clear, symmetric steps between the saturation magnetization and
remanence.The steps occur before zero external field as soon as the
field is reduced from-Hs (red branch) and the total number of steps
is close to the number of ferro-magnetic layers in the whole sample
(10). Hence, one can assume that thewell-defined steps correspond
to the switching of the individual ferromag-netic layers or at
least to sufficiently big domains within the probed samplearea as
already reported for Fe/Cr/Fe superlattices [2]. Due to the
four-foldanisotropy, the layers (or domains) have to rotate by
either 90◦ or 180◦ in orderto be aligned along one easy axis.
Therefore, one can conclude that a couplingof the magnetic layers
is present and that this coupling is not ferromagneticsince the
plateaus can be observed before the external field is reversed.
More-over, one observes a remanence of 0.5Ms. Since L-MOKE cannot
measurea magnetization perpendicular to the plane of the incident
light, it might bethat half the layers point along one easy axis
(e.g. left hand side in fig. 3.4binset) and the other half point
along the other easy axis (e.g. up) resulting in anet magnetization
of 0.5Ms. An antiferromagnetic interlayer coupling shouldresult in
a remanent state with zero net magnetization. However, if the
inter-layer coupling is of similar magnitude as the
magnetocrystalline anisotropy,then it might be that the coupling is
not strong enough to overcome Fe’s mag-netocrystalline anisotropy
and a minimization of the energy is achieved by anangle of 90◦
between the layers. Another possibility is that the coupling
acrossthe MgO in fact is biquadratic, which would lead to the same
result. However,with only the L-MOKE data it is not possible to
give a clear statement aboutthe occurring phenomena. The alignment
of the ferromagnetic layers will bediscussed in more detail in
section 3.3.
To study the coupling mechanism in more detail, different MgO
spacerlayer thicknesses have been compared (fig. 3.5). Two
important changes areobserved with decreasing MgO thickness.
Firstly, the switching fields, includ-ing the field required to
saturate the magnetization, increase with decreasingthickness.
Secondly, the steps in the hysteresis become less pronounced
forthinner MgO spacer layers. Since L-MOKE probes only a small area
(size of
21
-
- 1 0 0 - 8 0 - 6 0 - 4 0 - 2 0 0 2 0 4 0 6 0 8 0 1 0 0
- 1 . 0
- 0 . 5
0 . 0
0 . 5
1 . 0
2 1 . 6 / 1 6 . 7 Å F e / M g O
Magn
etiza
tion (
norm
alized
)
B ( m T )
2 2 . 9 / 1 7 . 4 Å F e / M g O 2 0 . 9 / 1 9 . 9 Å F e / M g
O
2 3 . 4 / 1 4 . 6 Å F e / M g O
Figure 3.5. Normalized, experimental L-MOKE data of sample A-D
along the easyaxis.
the laser spot) it might be that thinner MgO layers lead to the
formation ofmagnetic domains which are smaller than the spot size.
Hence, one cannot as-sume a reversal of the whole magnetic layer,
but rather a successive reversal ofsmall domains, leading to a
successive Kerr rotation and therefore less abruptsteps in the
hysteresis curve.
Via Kerr microscopy, the domain structure of the remanent state
betweensamples with thick (22.2 Å) and thin (16.5 Å) MgO spacer
layers have beencompared (fig. 3.6). Since the Fe layers are only 2
nm thick and the MgOlayers are basically transparent, one has to
take into account, that the observedKerr microscope image is a
superposition of different Fe layers. Differentbrightness values
may correspond to different Fe layers. However, it is ob-vious that
the sample with thicker MgO spacer layers forms huge domains(around
0.5 mm), whereas the other sample consists of 0.1 mm or even
smallerdomains. Hence, the different domain sizes provide a good
explanation for theslope of the different hysteresis curves in
figure 3.5.
The changes in the switching fields can be related to the
strength of the in-terlayer exchange coupling by using the
following equation [21]. The negativebilinear coupling term J1
corresponds to the antiferromagnetic coupling termJAF , where Bs is
the saturation field, which can be obtained from the
L-MOKEmeasurements, Ms is the volume magnetization (literature
value) and dFe is the
22
-
Figure 3.6. Kerr microscopy image of a [Fe(23.0 Å)/MgO(22.2
Å)]10 (a) and a[Fe(25.3 Å)/MgO(16.5 Å)]10 (b) superlattice at
remanence. The thick MgO layer fa-vors the formation of large
domains, up to 0.5 mm across. A thin MgO layer leads tothe
formation of much smaller domains.
thickness of one Fe layer, which can be obtained from the XRR
fit. The re-sult is plotted together with the relative remanence
over the MgO thickness infigure 3.7.
− J1 = JAF =BsMsdFe
4(3.1)
Increasing the thickness of the insulating layer leads to a
decrease in theantiferromagnetic coupling strength (J1 decreasing
in magnitude) and an in-crease of the remanence (fig. 3.7). The
coupling strength follows an expo-nential decrease with an increase
of the spacer layer thickness approaching avalue of 0 (no
coupling), whereas the relative remanence’s behavior may bebest
described by an abrupt jump between a value either close to 50% or
0%.The smallest MgO thickness (indicated by a black arrow) follows
neither thetrend of the saturation field nor the trend of the
relative remanence. Due to thesmall thickness, a higher pinhole
density may occur. Pinholes lead to a fer-romagnetic coupling and
may weaken the antiferromagnetic coupling throughthe MgO. Since the
behavior of this sample is not purely connected to thesmaller MgO
thickness, this sample will be neglected in further analysis.
Previous experimental and theoretical studies have confirmed a
remarkableimpact of defects or vacancies within the MgO layers on
the magnetic cou-pling. Thick (8 Å [22] or 10 Å[23]) or perfect MgO
spacer layers [22] shouldlead to a ferromagnetic coupling. However,
impurities (e.g. oxygen vacancies)enhance the antiferromagnetic
coupling strength [24]. It is worth mentioning,that even with the
impurity-assisted coupling mechanism a weak, ferromag-netic
behavior above 10 Å is predicted [23]. The clear evidence for a
non-ferromagnetic coupling in samples above 15 Å has not been
reported yet. Infact, all recent theories and experiments suggest a
loss of any coupling above15 Å MgO thickness.
23
-
1 4 1 5 1 6 1 7 1 8 1 9 2 0 2 1 2 2 2 3
- 1 . 0 x 1 0 - 4
- 8 . 0 x 1 0 - 5
- 6 . 0 x 1 0 - 5
- 4 . 0 x 1 0 - 5
- 2 . 0 x 1 0 - 5
0 . 0
0
1 0
2 0
3 0
4 0
5 0
J 1
(J/m2
)
M g O T h i c k n e s s ( Å )
J 1
Mr/M
s (%)
M r / M s
Figure 3.7. Antiferromagnetic coupling strength (black dots) and
relative remanence(orange squares) of samples A-D and older samples
over the spacer layer thickness.The dashed line serves as a guide
to the eye.
The 90◦ coupling between the ferromagnetic layers observed for
sampleswith thick MgO spacer layers may have different origins. One
possibility is thecompetition between antiferromagnetic coupling
and Fe’s four-fold anisotropy.The thicker spacer layer weakens the
antiferromagnetic coupling so that thecoupling is not strong enough
to overcome the second hard axis to form anantiferromagnetic
alignment. Hence, a minimization of the energy is achievedwith a
biquadratic-like (90◦) alignment of the magnetization. This theory
isexperimentally confirmed for thinner MgO (6 Å) spacer layers
[25]. Anotherpossibility is the transition from an
antiferromagnetic to a biquadratic cou-pling due to oxidation of
the Fe/MgO interfaces and magnetic impurities inthe spacer layer as
shown for MgO thicknesses between 4.6 and 8.1 Å [26].The
biquadratic coupling can then be described by the loose spin model,
whichis described elsewhere [19].
To distinguish between both possibilities, temperature dependent
L-MOKEmeasurements have been performed. The biquadratic coupling
should becomedominant at low temperatures if an oxidation of the
interfaces is present [26].Hence, samples exhibiting an
antiferromagnetic alignment at room tempera-ture (thin MgO layers)
should exhibit a biquadratic behavior at lower tem-peratures and
samples exhibiting a perpendicular alignment of the magneti-zation
at room temperature should exhibit a higher saturation field
(strongercoupling) at low temperatures. Furthermore, temperature
dependent L-MOKE
24
-
- 1 . 0
- 0 . 5
0 . 0
0 . 5
1 . 0
- 6 0 - 4 0 - 2 0 0 2 0 4 0 6 0
3 0 K
Magn
etiza
tion (
norm
alized
)
B ( m T )
1 5 6 K
a )
2 6 9 K 3 8 2 K
- 4 0 0 - 3 0 0 - 2 0 0 - 1 0 0 0 1 0 0 2 0 0 3 0 0 4 0 0
- 1 . 0
- 0 . 5
0 . 0
0 . 5
1 . 0 3 0 K
Mag
netiza
tion (
norm
alized
)
B ( m T )
b )
1 5 6 K 2 6 9 K 3 8 2 K
Figure 3.8. Hysteresis loop along the easy axis of a [Fe(23.0
Å)/MgO(22.2 Å)]10 (a)and a [Fe(21.6 Å)/MgO(16.7 Å)]10 (b)
superlattice at various temperatures.
measurements are suitable to prove if impurities and defects are
present withinthe MgO layers. The quantum interference model
predicts an increase ofthe coupling strength with higher
temperatures due to the thermal populationof the excited electronic
states [24] (higher tunneling probability) and a de-crease with
lower temperatures (smaller tunneling probability). However,
theimpurity-assisted interlayer exchange coupling across a tunnel
barrier predictsthe inverse behavior [24].
The temperature dependent hysteresis loops are shown in figure
3.8. Thesample with the thick MgO spacer layer (fig. 3.8 a)
exhibits a square hystere-sis loop at low temperatures (typical for
bulk Fe) indicating the absence ofany coupling mechanism. At higher
temperatures, the formation of the char-acteristic hysteresis steps
and a reduced remanence occurs. It seems that thecoupling is
present above 60 K with a maximum (highest saturation field)
be-tween 120 and 160 K. By increasing the temperature further, the
saturationfield is reduced and the steps become less pronounced. It
is worth mentioning,that even above 380 K characteristic hysteresis
steps can be observed.
The sample with the thin MgO follows a similar behavior.
However, thesaturation field is at all temperatures higher than the
saturation field of thesample with the thick spacer layer. Below
200 K, the sample’s hysteresis looplooks similar to a hysteresis
loop of Fe’s hard axis as shown in figure 3.4a. Athigher
temperatures, the characteristic steps occur and the saturation
field isreduced as already reported for the other sample. The steps
occur together withthe sharp drop of the remanence, indicating the
antiferromagnetic coupling.Even at 380 K, the steps can be
observed. The relative remanence as well asthe saturation field of
both samples are plotted over the temperature in figure3.9 a and b,
respectively.
The interpretation of the temperature dependent data is not
trivial. First ofall, the coupling of both samples seems to vanish
below 60 K (thick MgO) and200 K (thin MgO), which is consistent
with the quantum interference model
25
-
0 5 0 1 0 0 1 5 0 2 0 0 2 5 0 3 0 0 3 5 0 4 0 00 . 00 . 10 . 20
. 30 . 40 . 50 . 60 . 70 . 80 . 91 . 0
0 5 0 1 0 0 1 5 0 2 0 0 2 5 0 3 0 0 3 5 0 4 0 00 . 00 . 10 . 20
. 30 . 40 . 50 . 60 . 70 . 80 . 91 . 0
2 2 . 2 Å M g O
Mr/M
s
T e m p e r a t u r e ( K )
a ) 1 6 . 7 Å M g O
Mr/M
s
T e m p e r a t u r e ( K )
0 5 0 1 0 0 1 5 0 2 0 0 2 5 0 3 0 0 3 5 0 4 0 005
1 01 52 02 53 03 54 04 5
0 5 0 1 0 0 1 5 0 2 0 0 2 5 0 3 0 0 3 5 0 4 0 0 0
5 0
1 0 0
1 5 0
2 0 0
2 5 0
3 0 02 2 . 2 Å M g O
Satur
ation
field
(mT)
T e m p e r a t u r e ( K )
b )1 6 . 7 Å M g O
Satur
ation
field
(mT)
T e m p e r a t u r e ( K )
Figure 3.9. Relative remanence (a) and saturation field (b) of a
[Fe(23.0 Å)/MgO(22.2Å)]10 (orange circles) and a [Fe(21.6
Å)/MgO(16.7 Å)]10 (green squares) superlatticeat various
temperatures between 10 K and 390 K.
26
-
for perfect spacer layers. However, before the coupling
vanishes, a steady in-crease of the saturation field (coupling
strength) is observed, which indicatesrather an impurity-assisted
interlayer exchange coupling. It seems that at leastfor the thick
MgO layers a superposition of the impurity-assisted couplingand the
behavior of a perfect tunnel-junction exists, whereby the
impurity-assisted coupling wins at higher temperatures. It is
predicted that an increasein the MgO thickness will lead to a
coupling comparable to the one of an idealtunnel-junction [13].
Hence, the impurity-assisted coupling becomes less im-portant for
thicker MgO spacer layers. This is consistent with the low
tem-perature data of the thick MgO spacer layer. First an absence
of coupling dueto a smaller tunneling probability (below 60 K) and
then an increase of thesaturation field (coupling strength) up to
100 K. After 100 K, the saturationfield is constant up to 170 K. In
this region, the impurity-assisted coupling andthe coupling of a
perfect tunnel barrier seems to be equally strong. For
tem-peratures above 160 K, a steady decrease of the saturation
field indicates thedominance of the impurity assisted coupling. The
relative remanence of thissample exhibits values above 0.5Ms only
during the absence of any coupling(below 60 K) and is apart from
that almost constant. Hence, even at low tem-peratures (60 K) the
90◦ coupling between the ferromagnetic layers is present.It seems
that the coupling is still too weak to overcome the second hard
axisto form the antiferromagnetic alignment of the Fe layers.
The low temperature behavior of the thin MgO spacer layer is
even morepuzzling. For temperatures above 200 K it follows similar
to the thick MgOlayers the behavior of the impurity-assisted
coupling. However, below 200 Kthe hysteresis loops look like loops
along the hard axis. An explanation mightbe, that the different
thermal expansion coefficients between Fe and MgO mayinduce a
lattice distortion at low temperatures swapping the magnetic easy
andhard axes. Due to this exchange, it is impossible to find traces
of a superpo-sition of the impurity-assisted and the coupling of a
perfect MgO layer asinterpreted for the thick MgO spacer. To
confirm this interpretation and tofind traces, one has to rotate
the sample by 45◦ (hard axis at room tempera-ture) and perform the
same L-MOKE measurement, but this has yet to be done.
Finally, L-MOKE measurements along the easy axis (fig. 3.10)
were per-formed for the sample grown on SrTiO3 first presented in
figure 3.2. Oneobserves similar magnetic properties compared to
samples grown on MgO.The remanence is reduced to 0.34Ms indicating
a different layer configurationat zero external field. Furthermore,
one observes a hysteresis step around 40mT, which is far away from
the other steps indicating a stronger coupling ofone of the
layers.
27
-
- 6 0 - 4 0 - 2 0 0 2 0 4 0 6 0
- 1 . 0
- 0 . 5
0 . 0
0 . 5
1 . 0
Magn
etiza
tion (
norm
alized
)
B ( m T )
Figure 3.10. Experimental L-MOKE data (black squares) of a
[Fe/MgO]9Fe superlat-tice grown on SrTiO3.
28
-
Figure 3.11. Descending branch of sample A’s hysteresis loop
along the easy axis toillustrate the chosen field steps for the PNR
measurements.
3.3 Magnetic OrderingTo study the magnetic alignment of
individual layers, PNR measurements at
different external field steps have been performed. Before each
measurement,the sample was saturated along the neutron guide field
(up) to start with a well-defined magnetic history of the sample.
Both NSF channels (UU and DD) aswell as one SF channel (DU) were
measured and fitted together with the XRRdata. Furthermore, both
NSF were measured at saturation to obtain a precisevalue for Fe’s
magnetic moment which may vary with the thickness of theFe layers
[27] and was then used for the fitting of the PNR data at
differentexternal fields. The raw data was reduced by Gunnar K.
Palsson’s Super AdamReduction program (SARED, unpublished). The
reduction process included(i) redefinition of the region of
interest to reduce the noise (ii) dividing bythe monitor to
compensate fluctuations in the neutrons flux (iii) direct
beamnormalization to obtain the reflectivity values and (iv)
background subtractionto further improve the signal to noise
ratio.
For sample A (thickest MgO layer), PNR measurements were carried
outat the field values corresponding to remanence, and the first
and second mag-netization steps, as shown in figure 3.12. The
experimental PNR data as wellas the fit are plotted in figure 3.12.
For the sake of clarity, the fit and data ofthe first and second
step are plotted with an offset of 1E-3 and 1E-6 respec-tively. At
remanence, both data sets exhibit a clear peak at QB = 0.157
1/Åcomparable to the position of the first Bragg peak in the XRR
spectrum (QB =0.162 1/Å). Furthermore, a clear peak at QB/2 =
0.0789 1/Å and another peak
29
-
0 . 0 0 0 . 0 5 0 . 1 0 0 . 1 5 0 . 2 0 0 . 2 51 E - 1 31 E - 1
21 E - 1 11 E - 1 0
1 E - 91 E - 81 E - 71 E - 61 E - 51 E - 41 E - 30 . 0 1
0 . 1
s e c o n d s t e p
f i r s t s t e p
U U D a t a G e n X F i t
Refle
ctivity
a )
Q ( 1 / Å )
r e m a n e n c e
0 . 0 0 0 . 0 5 0 . 1 0 0 . 1 5 0 . 2 0 0 . 2 51 E - 1 31 E - 1
21 E - 1 11 E - 1 0
1 E - 91 E - 81 E - 71 E - 61 E - 51 E - 41 E - 30 . 0 1
0 . 1b ) D U D a t a
G e n X F i t
Refle
ctivity
Q ( 1 / Å )
r e m a n e n c e
f i r s t s t e p
s e c o n d s t e p
Figure 3.12. Experimental (black dots) and fitted (red/ orange
line) UU (a) and DU(b) PNR data of sample A (thick MgO) at
different external fields. For the sake ofclarity, the measurement
of the first and second step are plotted with an offset of 1E-3and
1E-6, respectively.
30
-
Figure 3.13. Schematic illustration of the first three magnetic
layers at remanence forsample A. For the sake of clarity, the
non-magnetic spacer layers are not shown.
at Q = 0.232 1/Å is present in both channels. The latter peak is
the higherorder of the QB/2 peak. The best fit is obtained for the
in-plane direction of themagnetization of the layers alternating
between values close to the in figure2.6 defined angle γ = 0◦ and
90◦ (fig. 3.13) as already reported [14].
The peak at 0.157 1/Å in the NSF channel occurs due to the
variation inthe nuclear density (bilayer thickness). The peak at
QB/2 in the NSF channelis purely due to the magnetic periodicity,
which is therefore twice the struc-tural periodicity (two bilayer
thicknesses). According to equation 2.3, the NSFchannel is not
sensitive to a magnetization perpendicular to the neutrons
spin.Hence, the magnetic periodicity varies between no contribution
(magnetiza-tion perpendicular to the neutrons spin) and high
contribution (magnetizationcollinear to the neutrons spin), which
is exactly twice the nuclear periodicityresulting in the QB/2 peak
as illustrated in figure 3.13. The SF channel is onlysensitive to a
magnetization perpendicular to the neutrons spin. Hence, theQB/2
peak there can be identically explained. The occurring QB peak
seemsto be puzzling, since the variation of the nuclear density
does not contributeto the SF channel. In former studies [14], this
peak was explained by hugedomains. The Kerr microscope images
confirmed the presence of domains.However, they should not result
is such a pronounced peak. A more realisticexplanation is the
constructive interference due to the second harmonics (n=2)in
Bragg’s law (equ. 2.1) leading to a Q2B/2 = QB peak. The same
occursof course also for the NSF channel, but the magnetic and
nuclear contributionoverlap at the same position making the second
harmonics (net magnetization)less obvious.
The increase of the external field leads to a clear reduction
and broadeningof the QB/2 peak in both channels indicating a
reduced magnetic periodicity.In contrast to the NSF channel, the QB
peak in the SF channel becomes muchbroader. By applying an external
field collinear to the incoming neutrons spinsome layers may align
parallel to the external field reducing both, the contribu-tion of
these individual layers as well as the measurable net magnetization
to
31
-
Table 3.2. Azimuthal angles γ of the net magnetization of
individual Fe layers forthree different applied field values (as
defined in figure 3.11).
Layer Remanence First Step Second Step1 82◦ 2◦ 4◦
2 -1◦ 9◦ 2◦
3 84◦ 83◦ 20◦
4 3◦ 4◦ 7◦
5 84◦ 79◦ 64◦
6 0◦ 1◦ 1◦
7 83◦ 79◦ 72◦
8 5◦ 1◦ 0◦
9 85◦ 79◦ 85◦
10 0◦ -1◦ -18◦
the SF channel resulting in a reduction and broadening of all
three SF channelpeaks. The obtained magnetic angles for the best
fit are summarized in table3.2.
One notices the small offset in the γ = 90◦ angle at remanence.
The smallguide field (to define the direction of the neutrons spin
during the measure-ment, 1 mT) is pointing upwards (γ = 0◦). Hence,
the layers forming a γ = 90◦angle are a little bit tilted upwards
due to the weak external field. Furthermore,a small misalignment of
the sample with respect to the guide field may alsofavor an
imperfect configuration of the sample’s magnetization.
By increasing the external field to 4.5 mT, it was possible to
measure thefirst plateau occurring in the hysteresis curve (fig.
3.11). The biggest changeis observed in the top layer, which is now
aligned along the externally appliedfield. This layer has only one
nearest neighbor and experiences therefore theweakest coupling.
Hence, the applied field is strong enough to switch thewhole layer
without having a major impact on the alignment of the other
layersmagnetization.
By increasing the external field up to 7.5 mT, the second
plateau can bemeasured. This plateau is much narrower than the one
measured before mean-ing that there is some uncertainty in the
exact position on the hysteresis curvefor this measurement.
Furthermore, all the unchanged layers have the sameamount of
nearest neighbors. If the coupling exists only between the
nearestneighbors, then every layer has the same probability to
switch its magnetiza-tion. Layer 3 experiences the biggest impact
of the new external field. How-ever, also layer 5 and 7 are further
away from the favored 90◦ alignment, whichmay be a hint that some
domains have switched. It seems that the stepwise re-versal of the
layers starts from the weakest coupled layer and propagates
thenthrough the whole sample. That might imply that the coupling
extends notonly to the nearest neighbors, but also to the next
nearest neighbors or evenfurther.
32
-
Figure 3.14. Descending branch of sample C’s hysteresis loop
along the easy axis toillustrate the chosen field steps for the PNR
measurements.
The presented angles of the net magnetization correspond to the
best ob-tained fit. Varying these angles will always lead to a
worse fit. However, smallchanges may not have a remarkable impact.
Therefore, the uncertainty of thepresented angles is estimated to
be ≤ 10◦.
The same measurements were performed for sample C with a thinner
MgOspacer layer. The measurement protocol was identical to the one
used for theprevious measurements and the data reduction was done
in the exact same way.Again, a precise value for Fe’s magnetization
was obtained by a fit of a NSFmeasurement at saturation. A
measurement at remanence and a measurementwith an external field of
47 mT was performed (fig. 3.14).
One of the biggest differences between the PNR results of both
samples isthe missing QB/2 peak in the remanent state of the NSF
channel. Since the SFchannel exhibits a clear QB/2 peak, one can
conclude that a magnetic period-icity with twice the bilayer
thickness exists perpendicular to the neutron spin.The GenX fit of
the data set confirms the periodic, antiparallel alignment ofthe Fe
layers (fig. 3.16) as already deduced from the L-MOKE
measurements.Even though the sample was saturated along the
neutrons guide field (up), aremanent magnetization perpendicular to
the guide field was observed.
By reducing the external field, the strongest coupled layer will
flip first toform an antiparallel alignment (down) to the
neighboring layers. However,since the external field forces the
layer to be aligned in the opposite direction(up), a 90◦ flip might
be the compromise between the coupling (antiferromag-netic) and the
external field (ferromagnetic). This phenomena is known as
33
-
0 . 0 0 0 . 0 5 0 . 1 0 0 . 1 5 0 . 2 0 0 . 2 51 E - 1 11 E - 1
0
1 E - 91 E - 81 E - 71 E - 61 E - 51 E - 41 E - 30 . 0 1
0 . 1
0 . 0 0 0 . 0 5 0 . 1 0 0 . 1 5 0 . 2 0 0 . 2 51 E - 1 11 E - 1
0
1 E - 91 E - 81 E - 71 E - 61 E - 51 E - 41 E - 30 . 0 1
0 . 1
Refle
ctivity
Q ( 1 / Å )
s t e p
U U D a t a G e n X F i t
Refle
ctivity
Q ( 1 / Å )
r e m a n e n c e
0 . 0 0 0 . 0 5 0 . 1 0 0 . 1 5 0 . 2 0 0 . 2 51 E - 1 11 E - 1
0
1 E - 91 E - 81 E - 71 E - 61 E - 51 E - 41 E - 30 . 0 1
0 . 0 0 0 . 0 5 0 . 1 0 0 . 1 5 0 . 2 0 0 . 2 51 E - 1 11 E - 1
0
1 E - 91 E - 81 E - 71 E - 61 E - 51 E - 41 E - 30 . 0 1
Refle
ctivity
Q ( 1 / Å )
D U D a t a G e n X F i t
Q ( 1 / Å )
s t e p
r e m a n e n c e
Figure 3.15. Experimental (black dots) and fitted (red/ orange
line) UU (a) and DU (b)PNR data of sample C (thin MgO) at different
external fields. For the sake of clarity,the measurement at
remanence has an offset of 1E-3
34
-
Figure 3.16. Schematic illustration of the first three magnetic
layers at remanence forsample C. For the sake of clarity, the
non-magnetic spacer layers are not shown.
spin-flop-transition, mainly investigated in antiferromagnets,
but recently alsoadapted to "artificial antiferromagnets" like
magnetically coupled multilayers[28]. It precisely describes the
sudden 90◦ rotation of the magnetization withrespect to the
external field at a critical magnetic field [29].
Another explanation might be, that an uniaxial anisotropy is
embedded inthe fourfold anisotropy as was already reported for
other Fe/MgO superlattices[25]. In this case, the antiferromagnetic
coupling would be favored along onespecific easy axis. To confirm
this assumption, PNR measurements have to beperformed on the same
sample rotated by 90◦ (neutrons guide field along theother easy
axis).
The enhanced external field gives rise to the QB/2 peak in the
NSF channelindicating a magnetization collinear to the neutrons
spin (guide field). At thesame time, a broadening of the QB and
QB/2 peak in the SF channel can beobserved. That indicates a
reduction in the periodic alignment perpendicularto the neutrons
guide field. The obtained magnetic angles for the best fit
aresummarized in table 3.3.
The remanent state exhibits a periodic, antiparallel alignment
of the Fe lay-ers perpendicular to the guide field. Due to the
guide field, the layers are tilteda little bit upwards. The
magnetization of the first step looks at first randomlydistributed.
However, a detailed analysis may reveal the occurring phenom-ena.
At saturation, all magnetic moments are aligned along the external
field(up). By reducing the field, the strongest coupled layer does
not flip by 180◦,but rather by 90◦ (spin-flop) as already discussed
above. It seems that thestrongest coupled layer is layer number 5
(the middle layer), since this layeris already in equilibrium
position. Furthermore, this layer has the most near-est and
next-nearest neighbors resulting in the strongest coupling. A
largervariation in the magnetic angles as seen in the other sample
(thick MgO) ispresent since a competition between the strong
external field and the antifer-romagnetic coupling occurs. A
coherent rotation from a parallel (γ = 0◦) to anantiparallel
alignment (γ = ± 90◦) due to spin-flop-transition occurs
leading
35
-
Table 3.3. Azimuthal angles γ of the net magnetization of
individual Fe layers for twodifferent applied field values (as
defined in figure 3.14). The numbers in red highlightare doubtful
magnetizations along the hard axis.
Layer Remanence First Step1 81◦ 0◦
2 -94◦ 7◦
3 81◦ 53◦
4 -94◦ -24◦
5 81◦ 81◦
6 -94◦ -17◦
7 81◦ 67◦
8 -94◦ -4◦
9 81◦ 42◦
10 -94◦ -4◦
to a large variation in the magnetic angles. That would also
explain the slopein the hysteresis curve (fig. 3.14). However, the
four-fold magnetocrystallineanisotropy breaks the coherent rotation
and leads to a jump over the γ = ± 45◦hard axes. After this jump, a
coherent rotation of the magnetization to the val-ues γ = ± 90◦
proceeds. The angles of layer 3 and 9 are close to the value ofthe
magnetic hard axis. Since the neutrons measure the whole sample it
mightbe that these observed angles are an average about multiple
domains leadingto the unlikely orientations.
It is not surprising that the outermost layer (1) exhibits an
magnetic angle ofγ = 0◦. Again, this layer is the weakest coupled
layer and will therefore onlyturn at an external field close to
remanence. Hence, one can conclude that thereversal process from
saturation to remanence seems to start with the rotationof the
innermost layer and propagates then coherently to the outermost
layers.
36
-
Figure 3.17. Schematic illustration of the setup for in-plane
transport measurementsof a [Fe(26.0 Å)/MgO(17.0 Å)]10 superlattice
in cross section. The capping layer andsubstrate are
insulating.
3.4 MagnetotransportBy attaching silver contacts to the side of
a sample with relatively thin MgO
spacer layers, in-plane transport measurements with external
field dependencecould be performed in a standard four-terminal
sensing setup. Figure 3.17 and3.18 illustrate the setup in side and
top view, respectively. Such a measurementgeometry is not ideal for
studying TMR effects since most of the current willpass along the
low-resistance Fe layers without tunneling through the
MgO.Nonetheless, if the TMR is large enough a small contribution
could be detecteddue to current spreading throughout the thickness
of the film.
Easy Axes Hard Axes
A
V
Figure 3.18. Top-view of the setup for in-plane transport
measurements with silvercontacts on the sides.
37
-
3 . 8 4 4
3 . 8 4 6
3 . 8 4 8
3 . 8 5 0
3 . 8 5 2
3 . 8 5 4
3 . 8 5 6
3 . 8 5 8
3 . 8 6 0
- 2 5 0 - 2 0 0 - 1 5 0 - 1 0 0 - 5 0 0 5 0 1 0 0 1 5 0 2 0 0 2
5 0
- 1 . 0
- 0 . 5
0 . 0
0 . 5
1 . 0
Resis
tance
(Ohm
)0 . 3 6 %
m a g n e t i c f i e l d
Magn
etiza
tion (
norm
alized
)
B ( m T )
c u r r e n t
Figure 3.19. Magnetoresistance (upper part) with respect to the
hysteresis (lower part)along the horizontal hard axis. The red line
is the ascending and the black line thedescending branch.
First, in-plane transport measurements along the hard axis have
been per-formed as shown in figure 3.19. A decrease in the
resistance is observedwhen approaching the remanent state with a
small peak at zero external field.The main effect seems to come
from the anisotropic magnetoresistance effect(AMR). If the
material’s magnetization is parallel to the applied current, a
highresistance can be observed. A magnetization perpendicular to
the applied cur-rent leads to a smaller scattering cross section of
the 3d-orbitals and thereforeto a smaller resistance (fig. 3.20)
[30]. As already explained in section 3.2,a magnetization parallel
to the hard axis at saturation (parallel to the appliedcurrent) and
therefore a high resistance is observed. By reducing the
externalfield, the sample’s magnetization rotates along the easy
axes, increasing theperpendicular contribution of the magnetization
with respect to the appliedcurrent and therefore reducing also the
measured resistance.
The peak close to zero external field cannot be explained by the
AMR. Sincethe MgO layers are very thin and a very small remanence
can be observed, onecan assume that the individual layers exhibit a
periodic, antiparallel alignmentalong one easy axis (either top
right and bottom left or top left and bottomright corner), as
already shown for slightly thinner MgO spacer layers in sec-tion
3.3. Hence, at zero external field the magnetization component
collinear
38
-
Figure 3.20. Schematic illustration of the AMR effect (left). An
applied external fieldleads to a change in the sample’s
magnetization inducing a change in the scatteringcross section of
the 3d-orbitals (light green ellipses). The real shape of the
3d-orbitalsis illustrated on the right hand side [30].
to the applied current is smallest, which would lead to the
smallest AMR value.However, the TMR exhibits a high resistance for
a periodic, antiparallel align-ment (as assumed for the remanent
state) and a low resistance for a periodic,parallel alignment (as
assumed for the saturated state). Thus, the TMR shouldexhibit a
maximum resistance at zero external field and a minimum
resistanceat saturation (contrary to the AMR). Hence, one can
conclude that at least asmall part of the s-(conductance)-electrons
tunnels through the MgO layers toexhibit the small peak at
remanence. A contribution of the TMR may alsoexist by approaching
the remanent state, however since this effect is so tinyit cannot
be distinguished from the AMR. The maximum magnetoresistance(mainly
due to the AMR) is just 0.36%.
In-plane magnetoresistance measurements along the easy axis have
beenperformed to investigate if the steps in the hysteresis curve
give rise to a changein the in-plane resistance (fig. 3.21). Again,
the main contribution comes fromthe AMR. A high resistance can be
observed if the sample’s magnetization isparallel to the applied
current (along the horizontal easy axis). By reducingthe external
magnetic field, a successive switching of the layers
perpendicularto the applied field (along the vertical easy axis due
to spin-flop) leads to astepwise decrease in the resistance. This
switching has been confirmed for asample with a slightly thinner
MgO spacer in section 3.3. However, at zeroexternal field one
observes a pronounced peak, which cannot be explained bythe AMR.
Similar to the measurement along the hard axis, this peak can
beattributed to the TMR. Since some s-electrons tunnel through the
insulatingspacers, a periodic, antiparallel alignment leads to a
rise in resistance. Again,the effect observed close to remanence is
a superposition of a low resistancedue to the AMR (magnetization
perpendicular to current) and a high resis-tance due to the TMR
(periodic, antiparallel magnetization of the ferromagen-tic
layers). The magnetoresistance effect is only 0.16%, but since the
AMR
39
-
3 . 8 7 4
3 . 8 7 6
3 . 8 7 8
3 . 8 8 0
3 . 8 8 2
3 . 8 8 4
3 . 8 8 6
- 3 0 - 2 0 - 1 0 0 1 0 2 0 3 0
- 1 . 0
- 0 . 5
0 . 0
0 . 5
1 . 0
Resis
tance
(Ohm
)0 . 1 6 %
m a g n e t i c f i e l d
Magn
etiza
tion (
norm
alized
)
B ( m T )
c u r r e n t
Figure 3.21. Magnetoresistance (upper part) with respect to the
hysteresis (lower part)along the horizontal easy axis. The red line
is the ascending and the black line thedescending branch.
weakens the TMR contribution, one can assume that an
out-of-plane measure-ment should lead to a much higher TMR effect.
Firstly, because no AMRwould be present (no weakening of the TMR)
and secondly more s-electronswould tunnel through the MgO barrier
since no alternative path (in-plane tothe ferromagnetic layers) is
available. Even though the observed effects arerelatively weak, it
is remarkable that the steps in the magnetization can alsobe
observed in the in-plane resistance making these structures
promising forfuture out-of-plane measurements.
The analysis of future TMR measurements is not trivial since the
TMRis expected to be weakened, but also enhanced by different
effects. It wasshown that the main coupling mechanism at room
temperature is impurity-assisted spin-polarized tunneling. The
impurities (probably O-vacancies) leadto a sharp decrease in the
theoretically possible TMR. An ideal tunnel junc-tion can exhibit a
TMR of more than 1800%, whereas the TMR of a defectafflicted
junction is reduced to 800% or less [13]. This reduction is
attributedto the incoherent scattering of the tunneling electrons.
On the other hand,the smooth interfaces in the samples (confirmed
by XRR measurements) willhave a positive effect on the TMR, which
highly depends on the interfacequality [15] and structural
relaxation [31]. Furthermore, the TMR increases
40
-
with increasing MgO thickness, where a peak in the TMR can be
observed atdifferent MgO thicknesses ("hot spots"). One of these
"hot spots" is a MgOthickness around 22 Å [32] close to the
thickness of sample A. A 25 Å thickMgO spacer leads to a TMR of 67%
[33]. By increasing the MgO thickness to30 Å a TMR of even 150% at
room temperature can be observed [34]. How-ever, no coupling was
observed in these studies. Therefore, the top Fe layerhad to be
magnetically hardened with a Co layer to perform TMR measure-ments.
Hence, only a trilayer (hard Fe/MgO/Fe) could be measured. All
thehere presented MgO thicknesses exhibit a coupling, which makes
out-of planemagentotranport measurements through several layers
possible.
41
-
4. Conclusions and Outlook
[Fe/MgO]10 superlattices have been studied. The structures were
grown bysputter deposition, where the sputtering time for the MgO
deposition was var-ied for different samples. The different
sputtering times led to different MgOthicknesses between the
samples, which influenced the magnetic interlayerexchange coupling
mechanism. By fitting XRR measurements, it was possi-ble to obtain
precise values for the thickness of the individual layers as wellas
to extract further information like interface roughness or
variations in thedensity. These measurements confirmed the high
degree of perfection in thelayering. XRD measurements were
performed to examine the structural qual-ity. Every sample
exhibited superlattice peaks, illustrating a high
out-of-planecoherence.
L-MOKE measurements were carried out to study the magnetic
propertiesof the samples. It was found that unusual steps occur in
the hysteresis curves.Each step corresponds to the switching of
large domains in individual Fe lay-ers. Kerr microscopy
measurements confirmed that a thick MgO spacer layerleads to the
formation of larger domains. Furthermore, a thicker spacer
layerleads to a weaker magnetic coupling between the Fe layers. The
coupling isfound to be antiferromagnetic, whereby samples with
relatively thick MgOspacer layers exhibit a perpendicular alignment
of the individual layers at re-manence due to a competition between
the antiferromagnetic coupling andFe’s fourfold magnetocrystalline
anisotropy. Temperature dependent measure-ments revealed that the
dominant coupling mechanism at high temperatures isthe
impurity-assisted interlayer exchange coupling due to
spin-polarized tun-neling. However, it seems that samples with a
thick spacer layer follow ratherthe behavior of a perfect (no
defects) tunnel junction at low temperatures.Samples with thin MgO
spacer layers exhibit a typical hysteresis loop of ahard axis for
low temperature measurements along the easy axis. It seems
thatinternal strain swaps the magnetic anisotropy axes at low
temperatures makingit impossible to investigate the coupling at low
temperatures. At higher tem-peratures the coupling could be
assigned to the impurity-assisted interlayer ex-change coupling due
to spin-polarized tunneling. It is worth mentioning, thatneither in
theory nor in experiments has a coupling across such thick
MgObarriers been reported previously.
PNR measurements have been performed to confirm the periodic
couplingof the individual layers. The occurring QB/2 peak confirmed
the periodic mag-netic alignment at remanence, which was found to
be antiparallel and perpen-dicular for thin and thick MgO spacer
layers, respectively. By fitting the PNR
42
-
data together with the XRR data and by performing measurements
at differentexternal fields, it was possible to confirm that the
hysteresis steps correspondto the switching of individual layers
and to get an impression of the switchingsequence. It was found
that the reversal of the layers starts with a weakly cou-pled
outermost layer (only one nearest neighbor) and propagates then
throughthe whole sample.
In-plane transport measurements showed mainly an AMR, however
somefeatures can be assigned to a TMR. Unfortunately, the TMR
contribution wasreduced by the AMR and a quantification of this
effect was not possible. Thisis perhaps not surprising as the
in-plane resistivity is not strongly affected bythe TMR across the
MgO layers. Hence, out-of-plane transport measurementsare of utmost
importance to get solely a TMR contribution. Therefore, thegrowth
on SrTiO3 substrates has been studied, as this material can be
dopedto become conductive. It was found that samples grown on
SrTiO3 exhibitsimilar XRR and XRD data and similar steps in the
hysteresis curve makingit suitable for out-of-plane transport
measurements. However, the growth hasto be optimized and the
magnetic properties have to be studied more carefullyexceeding the
scope of this thesis.
Looking ahead, temperature dependent measurements of the sample
withthe thin MgO spacer layers along the hard axis have to be
performed in or-der to confirm the swapping of the magnetic axes at
low temperatures and tostudy the coupling mechanism at these
temperatures. The next step wouldbe the growth optimization on
doped SrTiO3 substrates. Once a sufficientlyhigh sample quality is
achieved, L-MOKE and PNR measurements similar tothe ones carried
out for samples grown on MgO have to be performed. Out-of-plane
transport measurements can then be performed through the
wholesample. Additionally, a patterning of individual islands can
be performed. Bychoosing the size of the islands according to the
domains size one can ensurethat each hysteresis step corresponds to
the switching of the whole layer. Inthis case, PNR measurements may
be also easier to fit. To exploit the fullpotential of such
samples, one could also combine different MgO but alsoFe
thicknesses to create novel devices (e.g. a shift-register). Such a
devicemay have more stable states than just the antiparallel or
perpendicular coupledstates shown in this thesis.
43
-
5. Acknowledgment
First of all, I would like to thank Dr. Fridrik Magnus for
fulfilling his su-pervision duties flawlessly. Without his
assistance during the measurements,contagious enthusiasm, permanent
availability (particularly during the holi-days) and helpful
explanations as well as critical remarks, this thesis wouldnot
exist in its current state.
Furthermore, Dr. Gunnar Karl Palsson has to be honored for his
supportduring the PNR measurements as well as for his indispensable
assistance dur-ing the following data reduction and analyzes
processes.
Prof. Bengt Lindgren has to be named for teaching me the GenX
basics,helping me tirelessly improving the XRR and PNR fits as well
as explainingme mysterious PNR data nobody else understood.
Moreover, Dr. Spyridon Pappas deserves my gratitude for
sacrificing a lotof his time to help me with the temperature
dependent L-MOKE measure-ments.
Also, Sotirios Droulias sacrificed a lot of his time to teach me
XRR andXRD basics (and advanced stuff). Thank you!
Additionally, I appreciate Emil Melander’s help with the Swedish
abstractas well as his organized after-work group activities.
Besides, I would like to express my deepest gratitude to Dr.
Vassilios Ka-paklis, who introduced me to this group, awakened my
interest in materialsscience and gave me valuable feedback on an
almost final version of this the-sis.
I would like to thank Dr. Reda Moubah for teaching me the
sputteringtechnique as well as providing me with a sputter recipe,
which was used togrow the superb samples.
I am grateful for fruitful fist-and second-hand discussions with
Prof. BjörgvinHjörvarsson, which led to a much deeper understanding
of the occurring cou-pling mechanism and occurring phenomena of the
PNR measurements, re-spectively as well as for his constant support
and motivation.
Finally, I want to thank the materials physics group of Uppsala
Universityfor supporting me whenever I needed help, half-serious,
half-humorous dis-cussions at the coffee table and activities apart
from academia.
Last, but not least I want to thank my family for both, the
financial andmoral support during all the years of study (almost
6), my friends for draggingme away from work and you, dear reader,
for reading my thesis (or at least theacknowledgment).
44
-
References
[1] P. Grünberg, R. Schreiber, Y. Pang, M. Brodsky, and H.
Sowers, “Layeredmagnetic structures: evidence for antiferromagnetic
coupling of Fe layersacross Cr interlayers,” Physical Review
Letters, vol. 57, no. 19, p. 2442, 1986.
[2] V. Ustinov, M. Milyaev, L. Romashev, T. Krinitsina, A.
Burkhanov,V. Lauter-Pasyuk, and H. Lauter, “Multi-stepwise GMR and
layers magneticreversal in uniaxial Fe/Cr superlattices,” Physica
status solidi (c), vol. 3, no. 5,pp. 1249–1256, 2006.
[3] M. Julliere, “Tunneling between ferromagnetic films,”
Physics Letters A,vol. 54, pp. 225–226, Sept. 1975.
[4] C. Chappert, A. Fert, and F. N. Van Dau, “The emergence of
spin electronics indata storage,” Nature materials, vol. 6, no. 11,
pp. 813–823, 2007.
[5] S. S. Parkin, C. Kaiser, A. Panchula, P. M. Rice, B. Hughes,
M. Samant, andS.-H. Yang, “Giant tunnelling magnetoresistance at
room temperature withMgO (100) tunnel barriers,” Nature materials,
vol. 3, no. 12, pp. 862–867, 2004.
[6] S. Yuasa, T. Nagahama, A. Fukushima, Y. Suzuki, and K. Ando,
“Giantroom-temperature magnetoresistance in single-crystal
Fe/MgO/Fe magnetictunnel junctions,” Nature materials, vol. 3, no.
12, pp. 868–871, 2004.
[7] Y. Lee, J. Hayakawa, S. Ikeda, F. Matsukura, and H. Ohno,
“Effect of electrodecomposition on the tunnel magnetoresistance of
pseudo-spin-valve magnetictunnel junction with a MgO tunnel
barrier,” Applied physics letters, vol. 90,no. 21, pp.
212507–212507, 2007.
[8] J. Åkerman, “Toward a universal memory,” Science, vol. 308,
no. 5721,pp. 508–510, 2005.
[9] J. C. Slonczewski, “Conductance and exchange coupling of two
ferromagnetsseparated by a tunneling barrier,” Physical Review B,
vol. 39, no. 10, p. 6995,1989.
[10] P. Bruno, “Theory of interlayer magnetic coupling,”
Physical Review B, vol. 52,no. 1, p. 411, 1995.
[11] M. Stiles, “Exchange coupling in magnetic
heterostructures,” Physical ReviewB, vol. 48, no. 10, p. 7238,
1993.
[12] J. Faure-Vincent, C. Tiusan, C. Bellouard, E. Popova, M.
Hehn, F. Montaigne,and A. Schuhl, “Interlayer magnetic coupling
interactions of two ferromagneticlayers by spin polarized
tunneling,” Physical review letters, vol. 89, no. 10,p. 107206,
2002.
[13] J. P. Velev, M. Y. Zhuravlev, K. D. Belashchenko, S. S.
Jaswal, E. Y. Tsymbal,T. Katayama, and S. Yuasa, “Defect-mediated
properties of magnetic tunneljunctions,” Magnetics, IEEE
Transactions on, vol. 43, no. 6, pp. 2770–2775,2007.
[14] T. Warnatz, “Synthesis and Characterization of Fe/MgO
Superlattices,” tech.rep., Uppsala University, 2015.
45
-
[15] H. Raanaei, Tailoring Properties of Materials at the
Nanoscale. PhD thesis,Uppsala University, 2009.
[16] H. Raanaei, H. Lidbaum, A. Liebig, K. Leifer, and B.
Hjörvarsson, “Structuralcoherence and layer perfection in Fe/MgO
multilayers,” Journal of Physics:Condensed Matter, vol. 20, no. 5,
p. 055212, 2008.
[17] V. Holy, J. Kubena, I. Ohli, K. Lischka, W. Plotz, et al.,
“X-ray reflection fromrough layered systems,” Physical review B,
vol. 47, no. 23, p. 15896, 1993.
[18] M. Björck and G. Andersson, “GenX: an extensible X-ray
reflectivityrefinement program utilizing differential evolution,”
Journal of AppliedCrystallography, vol. 40, no. 6, pp. 1174–1178,
2007.
[19] H. Kronmüller and S. Parkin, Handbook of Magnetism and
Advanced MagneticMaterials. No. Bd. 3 in Handbook of Magnetism and
Advanced MagneticMaterials, Wiley, 2007.
[20] M. Grimsditch and P. Vavassori, “The diffracted
magneto-optic kerr effect: whatdoes it tell you?,” Journal of
Physics: Condensed Matter, vol. 16, no. 9,p. R275, 2004.
[21] P. Poulopoulos, P. Isberg, W. Platow, W. Wisny, M. Farle,
B. Hjörvarsson, andK. Baberschke, “Magnetic anisotropy and exchange
coupling in Fe/V(0 0 1)superlattices on MgO (0 0 1),” Journal of
magnetism and magnetic materials,vol. 170, no. 1, pp. 57–66,
1997.
[22] T. Katayama, S. Yuasa, J. Velev, M. Y. Zhuravlev, S. S.
Jaswal, and E. Y.Tsymbal, “Interlayer exchange coupling in
Fe/MgO/Fe magnetic tunneljunctions,” Evgeny Tsymbal Publications,
p. 3, 2006.
[23] M. Y. Zhuravlev, J. Velev, A. Vedyayev, and E. Y. Tsymbal,
“Effect of oxygenvacancies on interlayer exchange coupling in
Fe/MgO/Fe tunnel junctions,”Journal of Magnetism and Magnetic
Materials, vol. 300, no. 1, pp. e277–e280,2006.
[24] M. Y. Zhuravlev, E. Y. Tsymbal, and A. Vedyayev,
“Impurity-assisted interlayerexchange coupling across a tunnel
barrier,” Physical review letters, vol. 94,no. 2, p. 026806,
2005.
[25] C. Bellouard, J. Faure-Vincent, C. Tiusan, F. Montaigne, M.
Hehn, V. Leiner,H. Fritzsche, and M. Gierlings, “Interlayer
magnetic coupling in Fe/MgOjunctions characterized by vector
magnetization measurements combined withpolarized neutron
reflectometry,” Physical Review B, vol. 78, no. 13, p.
134429,2008.
[26] Y. Chiang, J. J. Wong, X. Tan, Y. Li, K. Pi, W. Wang, H.
Tom, andR. Kawakami, “Oxidation-induced biquadratic coupling in
Co/Fe/MgO/Fe(001),” Physical Review B, vol. 79, no. 18, p. 184410,
2009.
[27] I. M. L. Billas, J. A. Becker, A. Châtelain, and W. A. de
Heer, “Magneticmoments of iron clusters with 25 to 700 atoms and
their dependence ontemperature,” Phys. Rev. Lett., vol. 71, pp.
4067–4070, Dec 1993.
[28] A. Bogdanov, A. Zhuravlev, and U. Rößler, “Spin-flop
transition in uniaxialantiferromagnets: Magnetic phases,
reorientation effects, and multidomainstates,” Physical Review B,
vol. 75, no. 9, p. 094425, 2007.
[29] F. Keffer and H. Chow, “Dynamics of the antiferromagnetic
spin-floptransition,” Physical Review Letters, vol. 31, no. 17, p.
1061, 1973.
[30] R. Gross and A. Marx, “Spinelektronik Vorlesungsskript zur
Vorlesung SS
46
-
2004,” Walther-Meissner-Institut, Garching, 2004.[31] H. Yang,
M. Chshiev, A. Kalitsov, A. Schuhl, and W. Butler, “Effect of
structural relaxation and oxidation conditions on interlayer
exchange couplingin Fe| MgO| Fe tunnel junctions,” Applied Physics
Letters, vol. 96, no. 26,p. 262509, 2010.
[32] R. Matsumoto, A. Fukushima, T. Nagahama, Y. Suzuki, K.
Ando, and S. Yuasa,“Oscillation of giant tunneling
magnetoresistance with respect to tunnelingbarrier thickness in
fully epitaxial Fe/MgO/Fe magnetic tunnel junctions,”Applied
physics letters, vol. 90, no. 25, p. 2506, 2007.
[33] J. Faure-Vincent, C. Tiusan, E. Jouguelet, F. Canet, M.
Sajieddine, C. Bellouard,E. Popova, M. Hehn, F. Montaigne, and A.
Schuhl, “High tunnelmagnetoresistance in epitaxial Fe/MgO/Fe tunnel
junctions,” Applied PhysicsLetters, vol. 82, no. 25, pp. 4507–4509,
2003.
[34] C. Tiusan, J. Faure-Vincent, M. Sicot, M. Hehn, C.
Bellouard, F. Montaigne,S. Andrieu, and A. Schuhl, “Spin filtering
effects in monocristalline Fe/MgO/Femagnetic tunnel junctions,”
Materials Science and Engineering: B, vol. 126,no. 2, pp. 112–119,
2006.
47
1 Introduction2 Methods2.1 Sample Preparation2.2 X-Ray
Reflectivity2.3 Polarized Neutron Reflectivity2.4 Magneto-Optical
Kerr Effect2.5 Four-Terminal Sensing
3 Results and Discussion3.1 Structural Properties3.2 Magnetic
Properties3.3 Magnetic Ordering3.4 Magnetotransport
4 Conclusions and Outlook5 AcknowledgmentReferences