MagneticCouplingandTransportPropertiesofFe/MgOSuperlattices U U

(1)

U

PPSALA

U

NIVERSITY

M

ASTER

T

HESIS

Magnetic Coupling and Transport

Properties of Fe/MgO Superlattices

Author:

Tobias Warnatz

Supervisor:

Dr. Fridrik Magnus

Materials Physics

Department of Physics and Astronomy

Thesis Number:

FYSMAS1036

Series:

FYSAST

(2)

(3)

Magnetiska strukturer med multilager är intressanta både för applikationer (hårddiskar, magnetfältssen-sorer etc.) och för fundamental forskning (exempelvis koppling mellan magnetiska lager). Material med passande gitterparametrar kan tillväxas epitaxiellt och forma ett supergitter. Den mycket kristallina kvaliteteten av såna materialprover kan leda till en bättre förståelse av den magnetiska kopplingsmekanismen och även till 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 att undersöka kopplingsmekanismen. Den strukturella karaktäriseringen utfördes med röntgenreflektivitetet och röntgendiffraktion. Mätningarna med röntgenreflektivitet anpassades med GenX för att erhålla exakta värden på lagrens tjocklek och ojämnhet. En hög kristallinitet och mycket jämna gränsytor mellan lagren erhölls 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 ett individuellt Fe-lager bytte riktning. Kopplingen favoriserar en antiparallell linjering i rumstemperatur och kopplingens styrka beror kraftigt på MgO-lagrets tjocklek. Kopplingsmekanismen kommer från utbytet mellan lagren genom spinnpolariserad kvanttunnling. Dock avslöjade temperaturberoende mätningar en magnetisk koppling assisterad av materialorenheter vid höga temperaturer och även av kopplingsbeteendet av en perfekt tunnelförbindelse vid låga temperaturer.

Genom mätningar med polariserad neutronreflektivitet var det möjligt att bekräfta en periodisk, 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 stegen uppmätta i hystereskurvan huvudsakligen genom den anisotropa magnetoresistanseffekten. Dock upp-mättes ett svagt bidrag från tunnelmagnetoresistanseffekten vilket gör att dessa prover också är lovande för framtida mätningar av elektron-transport ur planet.

Abstract

Magnetic 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 with suitable lattice parameters can be epitaxially grown to form a superlattice. The high crystalline quality of those samples may lead to a better understanding of the coupling mechanism as well as to the emergence of novel phenomena.

In this thesis, Fe/MgO superlattices were grown via a combination of direct-current and radio-frequency sputtering. 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 reflectivity measurements were fitted via GenX to obtain precise thickness and roughness values of the layers. A high crystalline 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 step corresponds to the switching of an individual Fe layer. The coupling favors an antiparallel alignment at room temperature and its strength highly depends on the spacer layer’s thickness. The coupling mechanism was 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 behavior of 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 coupling and magnetocrystalline anisotropy.

(6)

(7)

1. Introduction

In 1986, Peter Grünberg demonstrated antiferromagnetic coupling of Fe layers in Fe/Cr/Fe multilayers [1], which led subsequently to the discovery of the giant magnetoresistance effect (GMR). Due to the giant resistance dif-ference observed for parallel and antiparallel alignment of the ferromagnetic layers, the effect was quickly adapted in commercial devices for magnetic field sensing (e.g. in hard drives). In 2006, a multi-stepwise reversal of magnetic layers in Fe/Cr/Fe superlattices was observed leading to a partitioned GMR effect [2]. Such structures are of particular interest since not only two binary states (high and low resistance), but also intermediate resistance values could be used for reading out digital information leading to a higher storage density. However, this publication raised little attention mainly because the resistance change 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 proven in 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 raised major attention. This was because the achieved magnetoresistance effect was much larger than any other reported effect through metallic or non-metallic spacer layers. Optimization in the growth led to astonishing magnetoresistance values of 500% at room temperature [7]. Hence, the tunnel-magentoresistance effect (TMR) replaced the GMR in most devices and even led to other novel devices (like MRAMs [8]). The size of the TMR effect in single Fe/MgO/Fe junctions opens up the question whether a significant partitioned TMR can be obtained in MgO-based multilayers.

(8)

demonstrated the possibility for a partitioned TMR by a multi-stepwise rever-sal in Fe/MgO/Fe superlattices [14]. However, even though the main coupling mechanism 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 a

variation in the MgO thickness have been prepared to distinguish between different coupling mechanisms. Structural characterization was done using x-ray reflectivity (chapter 3.1), magnetic characterization was done using the magneto-optical Kerr effect (chapter 3.2) and polarized neutron reflectivity (chapter 3.3) was used to determine the magnetic ordering of ferromagnetic layers. Finally, in-plane magneto-transport measurements were performed to investigate the magnetoresistance. Furthermore, samples with a comparable high quality grown on SrTiO3will be characterized to investigate if such

(9)

2. Methods

In this chapter basic methods will be explained in order to understand and interpret the presented results in chapter 3.

2.1 Sample Preparation

A structure consisting of two or more discrete layers grown with different materials is called a multilayer. If the materials have a comparable lattice parameter, a superlattice can be formed. A superlattice is defined by a long out-of-plane structural coherence [15] (atomic registry). The lattice mismatch between Fe (2.866 Å [15]) and MgO (4.213 Å [15]) at first sight appears to be too big (47.0%) to form a superlattice, but a 45◦rotation of the Fe layer on top of 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 crystalline MgO (100). Another suitable material in terms of lattice matching would be SrTiO3. The 45◦ rotation of Fe on top of this material results in an even

smaller lattice mismatch (3.65%). Furthermore, this material can be doped with small amounts of Nb to become conductive. By using doped SrTiO3as a

substrate for the Fe/MgO superlattices, out-of-plane transport measurements can be performed. Fe [100 ] MgO [100] 45° Easy Axis Hard Axis Fe Mg O

Figure 2.1.Schematic illustration of the tetragonal Fe/MgO structure formed by a 45◦

(10)

Table 2.1. Spacer layer sputtering times and thicknesses.

Sample MgO Sputtering Time in s MgO Thickness in Å

A 450 19.9 B 420 17.4 C 350 16.7 D 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 high quality of the samples, the chamber’s base pressure was below 2· 10−9 Torr and Ar with a purity of 99.99999% was used as a sputtering gas. The 1 mm thick MgO (100) and SrTiO3(100) 10 mm x 10 mm substrates were annealed

(11)

Figure 2.3. HRTEM cross section image of a representative Fe/MgO (bright /dark layers) superlattice grown on a MgO (100) substrate. Very smooth interfaces (a) and a structural registry of the individual layers with respect to the single crystalline sub-strate lattice (b) can be observed.

2.2 X-Ray Reflectivity

X-ray reflectivity is a powerful tool for investigating thin films as well as multilayers. The experimental data can be used to obtain thickness, roughness and other sample specific values [17]. The x-ray source emits polychromatic photons (Bremsstrahlung and characteristic x-rays). In order to perform x-ray reflectivity (XRR) measurements, monochromatic x-rays are needed (here: copper Kα). A grid can be used to select the desired wavelength. Due to

interference effects at the grid, only photons with a certain wavelength are re-flected. In general, it does not matter if the grid (monochromator) is placed before or after the sample. A slit in front of the x-ray source is used to reduce the beam’s divergence, but reduces the beam’s intensity simultaneously. The chosen 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 all the measurements. The sample is rotated at a rate ω and the detector is rotated with a rate of 2θ , which is twice that of ω. In this case, it is ensured that the angle α between the source and the sample and between the sample and the detector is always the same (fig. 2.4). XRR measurements start at very small angles. Thus, total reflection occurs up to the critical angle αc. After that,

ordinary reflection progresses to lattice planes, at which Bragg’s law can be used

n· λ = 2d · sin(θ ) (2.1)

(12)

Substrate X-Ray Source Slit Sample Detecto r ( α ( α ω 2θ

Figure 2.4.Schematic illustration of a XRR setup.

between the reflected waves (fig. 2.4). Constructive interference occurs only if the 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 wavelength of the incoming photons and the reflection angle, it is possible to calculate the spacing between different layers.

It is convenient to plot the XRR data over the momentum transfer Q instead of 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 transfer Qcan be calculated via the following equation.

Q=4π

λ · sin(θ ). (2.2)

(13)

0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 1 . 2 1 . 4 1 . 6 1 E - 8 1 E - 7 1 E - 6 1 E - 5 1 E - 4 1 E - 3 0 . 0 1 0 . 1 1 C 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 In te n s it y ( a rb . u n it s ) 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) superlattice

with an interface roughness of 2 Å (Fe) and 3 Å (MgO).

it is not possible to distinguish between layers made of different isotopes since the electronic density would be the same.

(14)

Figure 2.6. Schematic illustration of the PNR geometry [19]. The angles αiand αf

indicate the angles of the incoming and reflected neutron beam, respectively. γ is the angle between the magnetization and the polarization of the incoming neutrons, where

γ = 0◦and 90◦correspond to a sample’s magnetization parallel to the y-axis (up) and

x-axis (right), respectively.

2.3 Polarized Neutron Reflectivity

Neutron 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 the combination of both techniques a powerful tool for characterizing samples. One of the biggest advantages of neutrons for the sample characterization is their spin. Hence, it is possible to probe unpaired electron spins (magnetic moments) in the sample. This technique is called polarized neutron reflec-tivity (PNR) and can be used to probe the direction of the magnetization in each layer as a function of thickness (depth dependent). A distinction is made between spin-flip (SF) and non-spin-flip (NSF) measurements. NSF measure-ments can probe a variation in nuclear density and a magnetization parallel to the incoming neutron’s spin (fig. 2.6).

The NSF measurement is divided into up-up (UU, spin of incoming trons up, measured neutrons up) and down-down (DD, spin of incoming neu-trons down, measured neuneu-trons down). Both measurements are sensitive to a variation in nuclear density. However, the contribution of the magnetic part varies for both channels depending on the orientation of the magnetization (parallel or antiparallel) to the incoming neutrons spin as illustrated in figure 2.6 and described in the following equations.

(15)

V are the neutron scattering potentials, bn is the nuclear scattering length, bm

is the magnetic scattering length and γ is the angle between the magnetization and 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 the nuclear density, but it is possible to probe the sample’s magnetization per-pendicular to the spin of the incoming neutrons. Hence, the combination of SF and NSF measurements is a powerful method to reveal the magnetiza-tion of individual magnetic layers perpendicular to the scattering plane. Both SF channels are usually identical and lead to the same information. In some special cases (e.g. magnetic chirality) both SF channels may differ and lead to additional information, but these cases are beyond the scope of this thesis and discussed elsewhere [19]. Figure 2.7 shows the scattering length den-sity (SLD) profile of a Fe/MgO superlattice. Since Fe and MgO are different materials, one observes the same periodicity in the nuclear and electron den-sity. Furthermore, since the magnetization of the layers are aligned parallel to neutrons guide field, one observes a magnetic periodicity identical to the nu-clear and electronic one. The biggest difference (contrast) is exhibited for the electron density. The nuclear density difference is low and hardly gives any signal. However, the magnetic contrast is very pronounced illustrating that a combination 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 neutrons

(16)

M α α M α α M α α Polar Longitudinal Transverse

Figure 2.8. Schematic illustration of the three MOKE geometries.

2.4 Magneto-Optical Kerr Effect

The Magneto-optical Kerr effect (MOKE) is widely used to measure hys-teresis curves as well as to image magnetic domains (Kerr microscopy) of metallic 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) as illustrated in figure 2.8, where the magnitude of the effect decreases in the same order.

(17)

H

Electromagnets

Detector

Laser

Polarizer 1 Polarizer 2 Sample

Figure 2.9.Schematic illustration of the L-MOKE setup.

A V Rwire Rwire Rwire Rwire R sample

Figure 2.10. Schematic illustration of the four-terminal sensing setup with the

am-peremeter A and the voltmeter V.

2.5 Four-Terminal Sensing

(18)

3. Results and Discussion

3.1 Structural Properties

XRR measurements have been used to determine the structural quality of the samples. The XRR measurement shown in figure 3.1 is representative for all 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, which makes 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 Fe and MgO leading to the values presented in table 2.1. As expected from the HRTEM images (fig. 2.3), very smooth interfaces with a roughness of only 0.5-1 monolayers (tab. 3.1) were obtained from the fitting. The atomic densi-ties were first set to literature values (_{unit cell volume}formula units ) and then varied by 10% to compensate 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 the Fe (002) peak. Those peaks are called superlattice peaks and only appear when there is a high structural coherence normal to the layers, as already ascertained from the HRTEM images.

The same measurements were performed on a [Fe/MgO]9Fe superlattice

grown on SrTiO3 with the exact same conditions as the sample above. In

this sample, the superlattice was terminated with a Fe layer instead of a MgO layer to avoid the possibility of island growth of the Pd capping layer. Pd is known to form large islands on top of MgO which could result in the capping layer 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 on MgO substrates shown above (fig. 3.1). Furthermore, superlattice peaks can be 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 the bulk lattice constant) in the sample grown on SrTiO3. One can conclude, that

the 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 Å

MgO Substrate 1.2

Fe 2.4

MgO 1.8

(19)

0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 1 01 1 02 1 03 1 04 1 05 1 06 1 07 1 08 5 0 5 5 6 0 6 5 7 0 7 5 1 00 1 01 1 02 1 03 In te n s it y ( a rb . u n it s ) 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 In te n s it y ( a rb . u n it s ) Q ( 1 / Å )

Figure 3.1.Experimental XRR data (black dots) of sample A [Fe(20.9 Å)/MgO(19.9

Å)]10and 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 02 1 03 1 04 1 05 1 06 1 07 1 08 5 5 6 0 6 5 1 02 1 03 In te n s it y ( a rb . u n it s ) 2 T h e t a ( d e g r e e ) F e ( 0 0 2 ) In te n s it y ( a rb . u n it s ) Q ( 1 / Å )

Figure 3.2.Experimental XRR data (black dots) of a [Fe/MgO]9Fe superlattice grown

on SrTiO3. The experimental XRD data (inset) exhibits superlattice peaks indicating

(20)

Easy Axes Hard Axes

Figure 3.3.Schematic illustration of Fe’s four-fold anisotropy (topview) with two easy

axes (dashed, orange lines) and two hard axes (solid, black lines).

3.2 Magnetic Properties

Fe 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 along the diagonals and the hard axes along the sides as illustrated in figure 3.3. It is worth to mention, that Fe’s four-fold anisotropy in thin films can only be observed 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, magmag-netic field along both axes (fig. 3.4). If a strong, external field is applied along the hard axis (e.g. to the left hand side of figure 3.4a inset), all magnetic moments are aligned along the external, 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 steady decrease of the magnetization can be observed. Reducing the field leads to a coherent rotation of the magnetization towards the easy axes (e.g. upper and lower left hand side corner in figure 3.3). At zero external field, a remaining magnetization (remanence) can still be observed. In this case, all magnetic moments are aligned along the easy axes (e.g. upper and lower left hand side corner 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 Msis the saturation magnetization, one gets Mr= 0.7Ms, which equals

to the observed remanent magnetization. Reversing the field leads to an abrupt jump 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 lower right 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 the saturation can finally be observed again.

(21)

- 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 M a g n e ti z a ti o n ( n o rm a liz e d ) B ( m T ) a ) M a g n e ti z a ti o n ( n o rm a liz e d ) B ( m T ) b )

Figure 3.4. Normalized, experimental L-MOKE data of sample A [Fe(20.9

Å)/MgO(19.9 Å)]10along 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 the well-defined steps correspond to the switching of the individual ferromag-netic layers or at least to sufficiently big domains within the probed sample area as already reported for Fe/Cr/Fe superlattices [2]. Due to the four-fold anisotropy, the layers (or domains) have to rotate by either 90◦or 180◦in order to be aligned along one easy axis. Therefore, one can conclude that a coupling of the magnetic layers is present and that this coupling is not ferromagnetic since the plateaus can be observed before the external field is reversed. More-over, one observes a remanence of 0.5Ms. Since L-MOKE cannot measure

a magnetization perpendicular to the plane of the incident light, it might be that half the layers point along one easy axis (e.g. left hand side in fig. 3.4b inset) and the other half point along the other easy axis (e.g. up) resulting in a net magnetization of 0.5Ms. An antiferromagnetic interlayer coupling should

result 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 an angle of 90◦between the layers. Another possibility is that the coupling across the 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 about the occurring phenomena. The alignment of the ferromagnetic layers will be discussed in more detail in section 3.3.

(22)

- 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 M a g n e ti z a ti o n ( n o rm a liz e d ) 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 easy

axis.

the laser spot) it might be that thinner MgO layers lead to the formation of magnetic 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 of small domains, leading to a successive Kerr rotation and therefore less abrupt steps in the hysteresis curve.

Via Kerr microscopy, the domain structure of the remanent state between samples with thick (22.2 Å) and thin (16.5 Å) MgO spacer layers have been compared (fig. 3.6). Since the Fe layers are only 2 nm thick and the MgO layers are basically transparent, one has to take into account, that the observed Kerr microscope image is a superposition of different Fe layers. Different brightness 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 smaller domains. Hence, the different domain sizes provide a good explanation for the slope 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 negative bilinear coupling term J1 corresponds to the antiferromagnetic coupling term

JAF, where Bsis the saturation field, which can be obtained from the L-MOKE

(23)

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 to the 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 in figure 3.7.

− J1= JAF=

BsMsdFe

4 (3.1)

Increasing the thickness of the insulating layer leads to a decrease in the antiferromagnetic 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 a value of 0 (no coupling), whereas the relative remanence’s behavior may be best 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 the trend of the saturation field nor the trend of the relative remanence. Due to the small thickness, a higher pinhole density may occur. Pinholes lead to a fer-romagnetic coupling and may weaken the antiferfer-romagnetic coupling through the MgO. Since the behavior of this sample is not purely connected to the smaller MgO thickness, this sample will be neglected in further analysis.

(24)

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 J1 ( J /m 2 ) M g O T h i c k n e s s ( Å ) J ₁ M r/ 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 samples with thick MgO spacer layers may have different origins. One possibility is the competition between antiferromagnetic coupling and Fe’s four-fold anisotropy. The thicker spacer layer weakens the antiferromagnetic coupling so that the coupling is not strong enough to overcome the second hard axis to form an antiferromagnetic alignment. Hence, a minimization of the energy is achieved with a biquadratic-like (90◦) alignment of the magnetization. This theory is experimentally confirmed for thinner MgO (6 Å) spacer layers [25]. Another possibility is the transition from an antiferromagnetic to a biquadratic cou-pling due to oxidation of the Fe/MgO interfaces and magnetic impurities in the 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, which is described elsewhere [19].

(25)

- 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 M a g n e ti z a ti o n ( n o rm a liz e d ) 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 M a g n e ti z a ti o n ( n o rm a liz e d ) 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 within the MgO layers. The quantum interference model predicts an increase of the coupling strength with higher temperatures due to the thermal population of the excited electronic states [24] (higher tunneling probability) and a de-crease with lower temperatures (smaller tunneling probability). However, the impurity-assisted interlayer exchange coupling across a tunnel barrier predicts the inverse behavior [24].

The temperature dependent hysteresis loops are shown in figure 3.8. The sample 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 of any coupling mechanism. At higher temperatures, the formation of the char-acteristic hysteresis steps and a reduced remanence occurs. It seems that the coupling is present above 60 K with a maximum (highest saturation field) be-tween 120 and 160 K. By increasing the temperature further, the saturation field 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, the saturation field is at all temperatures higher than the saturation field of the sample with the thick spacer layer. Below 200 K, the sample’s hysteresis loop looks similar to a hysteresis loop of Fe’s hard axis as shown in figure 3.4a. At higher temperatures, the characteristic steps occur and the saturation field is reduced as already reported for the other sample. The steps occur together with the sharp drop of the remanence, indicating the antiferromagnetic coupling. Even at 380 K, the steps can be observed. The relative remanence as well as the saturation field of both samples are plotted over the temperature in figure 3.9 a and b, respectively.

(26)

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 . 0 0 . 1 0 . 2 0 . 3 0 . 4 0 . 5 0 . 6 0 . 7 0 . 8 0 . 9 1 . 0 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 . 0 0 . 1 0 . 2 0 . 3 0 . 4 0 . 5 0 . 6 0 . 7 0 . 8 0 . 9 1 . 0 2 2 . 2 Å M g O M r/ M s T e m p e r a t u r e ( K ) a ) 1 6 . 7 Å M g O M r/ 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 0 0 5 1 0 1 5 2 0 2 5 3 0 3 5 4 0 4 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 0 2 2 . 2 Å M g O S a tu ra ti o n f ie ld ( m T ) T e m p e r a t u r e ( K ) b ) 1 6 . 7 Å M g O S a tu ra ti o n f ie ld ( m T ) 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(22.2

Å)]10(orange circles) and a [Fe(21.6 Å)/MgO(16.7 Å)]10(green squares) superlattice

(27)

for perfect spacer layers. However, before the coupling vanishes, a steady in-crease of the saturation field (coupling strength) is observed, which indicates rather an impurity-assisted interlayer exchange coupling. It seems that at least for the thick MgO layers a superposition of the impurity-assisted coupling and the behavior of a perfect tunnel-junction exists, whereby the impurity-assisted coupling wins at higher temperatures. It is predicted that an increase in the MgO thickness will lead to a coupling comparable to the one of an ideal tunnel-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 due to a smaller tunneling probability (below 60 K) and then an increase of the saturation field (coupling strength) up to 100 K. After 100 K, the saturation field is constant up to 170 K. In this region, the impurity-assisted coupling and the 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 the dominance of the impurity assisted coupling. The relative remanence of this sample exhibits values above 0.5Msonly 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 axis to form the antiferromagnetic alignment of the Fe layers.

The low temperature behavior of the thin MgO spacer layer is even more puzzling. For temperatures above 200 K it follows similar to the thick MgO layers the behavior of the impurity-assisted coupling. However, below 200 K the hysteresis loops look like loops along the hard axis. An explanation might be, that the different thermal expansion coefficients between Fe and MgO may induce a lattice distortion at low temperatures swapping the magnetic easy and hard 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 as interpreted for the thick MgO spacer. To confirm this interpretation and to find 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. One

observes similar magnetic properties compared to samples grown on MgO. The remanence is reduced to 0.34Msindicating a different layer configuration

(28)

- 6 0 - 4 0 - 2 0 0 2 0 4 0 6 0 - 1 . 0 - 0 . 5 0 . 0 0 . 5 1 . 0 M a g n e ti z a ti o n ( n o rm a liz e d ) B ( m T )

Figure 3.10.Experimental L-MOKE data (black squares) of a [Fe/MgO]9Fe

(29)

Figure 3.11.Descending branch of sample A’s hysteresis loop along the easy axis to illustrate the chosen field steps for the PNR measurements.

3.3 Magnetic Ordering

To 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) as well as one SF channel (DU) were measured and fitted together with the XRR data. Furthermore, both NSF were measured at saturation to obtain a precise value for Fe’s magnetic moment which may vary with the thickness of the Fe layers [27] and was then used for the fitting of the PNR data at different external fields. The raw data was reduced by Gunnar K. Palsson’s Super Adam Reduction program (SARED, unpublished). The reduction process included (i) redefinition of the region of interest to reduce the noise (ii) dividing by the monitor to compensate fluctuations in the neutrons flux (iii) direct beam normalization to obtain the reflectivity values and (iv) background subtraction to further improve the signal to noise ratio.

For sample A (thickest MgO layer), PNR measurements were carried out at 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 well as the fit are plotted in figure 3.12. For the sake of clarity, the fit and data of the 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/Å

comparable to the position of the first Bragg peak in the XRR spectrum (QB =

(30)

0 . 0 0 0 . 0 5 0 . 1 0 0 . 1 5 0 . 2 0 0 . 2 5 1 E - 1 3 1 E - 1 2 1 E - 1 1 1 E - 1 0 1 E - 9 1 E - 8 1 E - 7 1 E - 6 1 E - 5 1 E - 4 1 E - 3 0 . 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 R e fl e c ti v it y 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 5 1 E - 1 3 1 E - 1 2 1 E - 1 1 1 E - 1 0 1 E - 9 1 E - 8 1 E - 7 1 E - 6 1 E - 5 1 E - 4 1 E - 3 0 . 0 1 0 . 1 b ) _{D U D a t a} G e n X F i t R e fl e c ti v it y 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

(31)

Figure 3.13.Schematic illustration of the first three magnetic layers at remanence for sample 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 higher order of the QB/2peak. The best fit is obtained for the in-plane direction of the

magnetization of the layers alternating between values close to the in figure 2.6 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 in the nuclear density (bilayer thickness). The peak at QB/2 in the NSF channel

is purely due to the magnetic periodicity, which is therefore twice the struc-tural periodicity (two bilayer thicknesses). According to equation 2.3, the NSF channel 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 contribu(magnetiza-tion (magnetiza(magnetiza-tion collinear to the neutrons spin), which is exactly twice the nuclear periodicity resulting in the QB/2peak as illustrated in figure 3.13. The SF channel is only

sensitive to a magnetization perpendicular to the neutrons spin. Hence, the QB/2 peak there can be identically explained. The occurring QB peak seems

to be puzzling, since the variation of the nuclear density does not contribute to the SF channel. In former studies [14], this peak was explained by huge domains. The Kerr microscope images confirmed the presence of domains. However, they should not result is such a pronounced peak. A more realistic explanation is the constructive interference due to the second harmonics (n=2) in Bragg’s law (equ. 2.1) leading to a Q_2B/2 = QB peak. The same occurs

of course also for the NSF channel, but the magnetic and nuclear contribution overlap at the same position making the second harmonics (net magnetization) less obvious.

The increase of the external field leads to a clear reduction and broadening of the Q_B/2peak in both channels indicating a reduced magnetic periodicity. In contrast to the NSF channel, the QBpeak in the SF channel becomes much

(32)

Table 3.2. Azimuthal angles γ of the net magnetization of individual Fe layers for three different applied field values (as defined in figure 3.11).

Layer Remanence First Step Second Step

1 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 channel peaks. The obtained magnetic angles for the best fit are summarized in table 3.2.

One notices the small offset in the γ = 90◦ angle at remanence. The small guide 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 also favor an imperfect configuration of the sample’s magnetization.

By increasing the external field to 4.5 mT, it was possible to measure the first plateau occurring in the hysteresis curve (fig. 3.11). The biggest change is observed in the top layer, which is now aligned along the externally applied field. This layer has only one nearest neighbor and experiences therefore the weakest coupling. Hence, the applied field is strong enough to switch the whole layer without having a major impact on the alignment of the other layers magnetization.

(33)

Figure 3.14. Descending branch of sample C’s hysteresis loop along the easy axis to illustrate 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, small changes may not have a remarkable impact. Therefore, the uncertainty of the presented angles is estimated to be ≤ 10◦.

The same measurements were performed for sample C with a thinner MgO spacer layer. The measurement protocol was identical to the one used for the previous 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 NSF measurement at saturation. A measurement at remanence and a measurement with an external field of 47 mT was performed (fig. 3.14).

One of the biggest differences between the PNR results of both samples is the missing Q_B/2peak in the remanent state of the NSF channel. Since the SF channel 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 of the 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), a remanent magnetization perpendicular to the guide field was observed.

(34)

0 . 0 0 0 . 0 5 0 . 1 0 0 . 1 5 0 . 2 0 0 . 2 5 1 E - 1 1 1 E - 1 0 1 E - 9 1 E - 8 1 E - 7 1 E - 6 1 E - 5 1 E - 4 1 E - 3 0 . 0 1 0 . 1 0 . 0 0 0 . 0 5 0 . 1 0 0 . 1 5 0 . 2 0 0 . 2 5 1 E - 1 1 1 E - 1 0 1 E - 9 1 E - 8 1 E - 7 1 E - 6 1 E - 5 1 E - 4 1 E - 3 0 . 0 1 0 . 1 R e fl e c ti v it y Q ( 1 / Å ) s t e p U U D a t a G e n X F i t R e fl e c ti v it y 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 5 1 E - 1 1 1 E - 1 0 1 E - 9 1 E - 8 1 E - 7 1 E - 6 1 E - 5 1 E - 4 1 E - 3 0 . 0 1 0 . 0 0 0 . 0 5 0 . 1 0 0 . 1 5 0 . 2 0 0 . 2 5 1 E - 1 1 1 E - 1 0 1 E - 9 1 E - 8 1 E - 7 1 E - 6 1 E - 5 1 E - 4 1 E - 3 0 . 0 1 R e fl e c ti v it y 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)

(35)

Figure 3.16.Schematic illustration of the first three magnetic layers at remanence for sample C. For the sake of clarity, the non-magnetic spacer layers are not shown.

spin-flop-transition, mainly investigated in antiferromagnets, but recently also adapted to "artificial antiferromagnets" like magnetically coupled multilayers [28]. It precisely describes the sudden 90◦ rotation of the magnetization with respect to the external field at a critical magnetic field [29].

Another explanation might be, that an uniaxial anisotropy is embedded in the fourfold anisotropy as was already reported for other Fe/MgO superlattices [25]. In this case, the antiferromagnetic coupling would be favored along one specific easy axis. To confirm this assumption, PNR measurements have to be performed on the same sample rotated by 90◦(neutrons guide field along the other easy axis).

The enhanced external field gives rise to the QB/2peak in the NSF channel

indicating a magnetization collinear to the neutrons spin (guide field). At the same time, a broadening of the QB and QB/2 peak in the SF channel can be

observed. That indicates a reduction in the periodic alignment perpendicular to the neutrons guide field. The obtained magnetic angles for the best fit are summarized in table 3.3.

(36)

Table 3.3. Azimuthal angles γ of the net magnetization of individual Fe layers for two different applied field values (as defined in figure 3.14). The numbers in red highlight are doubtful magnetizations along the hard axis.

Layer Remanence First Step

1 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 slope in the hysteresis curve (fig. 3.14). However, the four-fold magnetocrystalline anisotropy 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 of the magnetic hard axis. Since the neutrons measure the whole sample it might be that these observed angles are an average about multiple domains leading to the unlikely orientations.

(37)

Figure 3.17. Schematic illustration of the setup for in-plane transport measurements

of a [Fe(26.0 Å)/MgO(17.0 Å)]10superlattice in cross section. The capping layer and

substrate are insulating.

3.4 Magnetotransport

By attaching silver contacts to the side of a sample with relatively thin MgO spacer layers, in-plane transport measurements with external field dependence could be performed in a standard four-terminal sensing setup. Figure 3.17 and 3.18 illustrate the setup in side and top view, respectively. Such a measurement geometry is not ideal for studying TMR effects since most of the current will pass along the low-resistance Fe layers without tunneling through the MgO. Nonetheless, if the TMR is large enough a small contribution could be detected due 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 silver

(38)

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 R e s is ta n c e ( O h m ) 0 . 3 6 % m a g n e t i c f i e l d M a g n e ti z a ti o n ( n o rm a liz e d ) 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 the descending 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 observed when 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 high resistance can be observed. A magnetization perpendicular to the applied cur-rent leads to a smaller scattering cross section of the 3d-orbitals and therefore to 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 applied current) and therefore a high resistance is observed. By reducing the external field, the sample’s magnetization rotates along the easy axes, increasing the perpendicular contribution of the magnetization with respect to the applied current and therefore reducing also the measured resistance.

(39)

Figure 3.20.Schematic illustration of the AMR effect (left). An applied external field leads to a change in the sample’s magnetization inducing a change in the scattering cross section of the 3d-orbitals (light green ellipses). The real shape of the 3d-orbitals is 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 should exhibit a maximum resistance at zero external field and a minimum resistance at saturation (contrary to the AMR). Hence, one can conclude that at least a small part of the s-(conductance)-electrons tunnels through the MgO layers to exhibit the small peak at remanence. A contribution of the TMR may also exist by approaching the remanent state, however since this effect is so tiny it cannot be distinguished from the AMR. The maximum magnetoresistance (mainly due to the AMR) is just 0.36%.

(40)

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 R e s is ta n c e ( O h m ) 0 . 1 6 % m a g n e t i c f i e l d M a g n e ti z a ti o n ( n o rm a liz e d ) 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 the descending 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 AMR would be present (no weakening of the TMR) and secondly more s-electrons would tunnel through the MgO barrier since no alternative path (in-plane to the ferromagnetic layers) is available. Even though the observed effects are relatively weak, it is remarkable that the steps in the magnetization can also be observed in the in-plane resistance making these structures promising for future out-of-plane measurements.

(41)

(42)

4. Conclusions and Outlook

[Fe/MgO]10superlattices have been studied. The structures were grown by

sputter deposition, where the sputtering time for the MgO deposition was var-ied for different samples. The different sputtering times led to different MgO thicknesses between the samples, which influenced the magnetic interlayer exchange coupling mechanism. By fitting XRR measurements, it was possi-ble to obtain precise values for the thickness of the individual layers as well as to extract further information like interface roughness or variations in the density. These measurements confirmed the high degree of perfection in the layering. XRD measurements were performed to examine the structural qual-ity. Every sample exhibited superlattice peaks, illustrating a high out-of-plane coherence.

L-MOKE measurements were carried out to study the magnetic properties of 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 layer leads to the formation of larger domains. Furthermore, a thicker spacer layer leads to a weaker magnetic coupling between the Fe layers. The coupling is found to be antiferromagnetic, whereby samples with relatively thick MgO spacer layers exhibit a perpendicular alignment of the individual layers at re-manence due to a competition between the antiferromagnetic coupling and Fe’s fourfold magnetocrystalline anisotropy. Temperature dependent measure-ments revealed that the dominant coupling mechanism at high temperatures is the impurity-assisted interlayer exchange coupling due to spin-polarized tun-neling. However, it seems that samples with a thick spacer layer follow rather the behavior of a perfect (no defects) tunnel junction at low temperatures. Samples with thin MgO spacer layers exhibit a typical hysteresis loop of a hard axis for low temperature measurements along the easy axis. It seems that internal strain swaps the magnetic anisotropy axes at low temperatures making it 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, that neither in theory nor in experiments has a coupling across such thick MgO barriers been reported previously.

PNR measurements have been performed to confirm the periodic coupling of the individual layers. The occurring QB/2peak confirmed the periodic

(43)

data together with the XRR data and by performing measurements at different external fields, it was possible to confirm that the hysteresis steps correspond to the switching of individual layers and to get an impression of the switching sequence. It was found that the reversal of the layers starts with a weakly cou-pled outermost layer (only one nearest neighbor) and propagates then through the whole sample.

In-plane transport measurements showed mainly an AMR, however some features can be assigned to a TMR. Unfortunately, the TMR contribution was reduced by the AMR and a quantification of this effect was not possible. This is perhaps not surprising as the in-plane resistivity is not strongly affected by the TMR across the MgO layers. Hence, out-of-plane transport measurements are of utmost importance to get solely a TMR contribution. Therefore, the growth on SrTiO3 substrates has been studied, as this material can be doped

to become conductive. It was found that samples grown on SrTiO3 exhibit

similar XRR and XRD data and similar steps in the hysteresis curve making it suitable for out-of-plane transport measurements. However, the growth has to be optimized and the magnetic properties have to be studied more carefully exceeding the scope of this thesis.

Looking ahead, temperature dependent measurements of the sample with the 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 to study the coupling mechanism at these temperatures. The next step would be the growth optimization on doped SrTiO3 substrates. Once a sufficiently

(44)

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 would not exist in its current state.

Furthermore, Dr. Gunnar Karl Palsson has to be honored for his support during the PNR measurements as well as for his indispensable assistance dur-ing the followdur-ing 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 explaining me mysterious PNR data nobody else understood.

Moreover, Dr. Spyridon Pappas deserves my gratitude for sacrificing a lot of 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 and XRD basics (and advanced stuff). Thank you!

Additionally, I appreciate Emil Melander’s help with the Swedish abstract as 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 materials science 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 sputtering technique as well as providing me with a sputter recipe, which was used to grow the superb samples.

I am grateful for fruitful fist-and second-hand discussions with Prof. Björgvin Hjörvarsson, 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 University for supporting me whenever I needed help, half-serious, half-humorous dis-cussions at the coffee table and activities apart from academia.

(45)

References

[1] P. Grünberg, R. Schreiber, Y. Pang, M. Brodsky, and H. Sowers, “Layered magnetic structures: evidence for antiferromagnetic coupling of Fe layers across 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 magnetic reversal 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 in data 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, and S.-H. Yang, “Giant tunnelling magnetoresistance at room temperature with MgO (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, “Giant

room-temperature magnetoresistance in single-crystal Fe/MgO/Fe magnetic tunnel 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 electrode composition on the tunnel magnetoresistance of pseudo-spin-valve magnetic tunnel 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 ferromagnets separated 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 Review B, 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 ferromagnetic layers 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 tunnel junctions,” Magnetics, IEEE Transactions on, vol. 43, no. 6, pp. 2770–2775, 2007.

(46)

[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, “Structural coherence 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 from rough layered systems,” Physical review B, vol. 47, no. 23, p. 15896, 1993. [18] M. Björck and G. Andersson, “GenX: an extensible X-ray reflectivity

refinement program utilizing differential evolution,” Journal of Applied Crystallography, vol. 40, no. 6, pp. 1174–1178, 2007.

[19] H. Kronmüller and S. Parkin, Handbook of Magnetism and Advanced Magnetic Materials. No. Bd. 3 in Handbook of Magnetism and Advanced Magnetic Materials, Wiley, 2007.

[20] M. Grimsditch and P. Vavassori, “The diffracted magneto-optic kerr effect: what does 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, and K. 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 tunnel junctions,” Evgeny Tsymbal Publications, p. 3, 2006.

[23] M. Y. Zhuravlev, J. Velev, A. Vedyayev, and E. Y. Tsymbal, “Effect of oxygen vacancies 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 interlayer exchange 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/MgO junctions characterized by vector magnetization measurements combined with polarized 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, and R. 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, “Magnetic moments of iron clusters with 25 to 700 atoms and their dependence on temperature,” Phys. Rev. Lett., vol. 71, pp. 4067–4070, Dec 1993.

[28] A. Bogdanov, A. Zhuravlev, and U. Rößler, “Spin-flop transition in uniaxial antiferromagnets: Magnetic phases, reorientation effects, and multidomain states,” Physical Review B, vol. 75, no. 9, p. 094425, 2007.

(47)

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 coupling in 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 tunneling barrier 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 tunnel

magnetoresistance in epitaxial Fe/MgO/Fe tunnel junctions,” Applied Physics Letters, vol. 82, no. 25, pp. 4507–4509, 2003.