Metallic Amorphous Thin Films and Heterostructures with Tunable Magnetic Properties

Transcription

1 Digital Comprehensive Summaries of Uppsala Dissertations from the Faculty of Science and Technology 1218 Metallic Amorphous Thin Films and Heterostructures with Tunable Magnetic Properties ATIEH ZAMANI ACTA UNIVERSITATIS UPSALIENSIS UPPSALA 2015 ISSN ISBN urn:nbn:se:uu:diva

2 Dissertation presented at Uppsala University to be publicly examined in 10134, Ångströmlaboratoriet, Polacksbacken, Uppsala, Friday, 27 February 2015 at 10:15 for the degree of Doctor of Philosophy. The examination will be conducted in English. Faculty examiner: Prof. Yves Idzerda (Montana state university). Abstract Zamani, A Metallic Amorphous Thin Films and Heterostructures with Tunable Magnetic Properties. Digital Comprehensive Summaries of Uppsala Dissertations from the Faculty of Science and Technology pp. Uppsala: Acta Universitatis Upsaliensis. ISBN The primary focus of this thesis is to study the effect of doping on magnetic properties in amorphous Fe 100 x Zr x alloys. Samples with compositions of x = 7,11.6 and 12 at.% were implanted with different concentrations of H. Moreover, the samples with a composition of x = 7 at.% were also implanted with He, B, C and N. Magnetic measurements were performed, using SQUID magnetometry and MOKE, in order to compare the as-grown and the implanted films. The Curie temperature (T c ) increases and the coercivity (H c ) decreases, with increasing dopant volume. We also found that H c increases with temperature for B and C doped samples. Magnetization curves at low temperature validate the presence of non-collinear spin configurations in the as-grown films, which is suppressed after doping, resulting in films with tunable soft magnetic properties. We have also studied the effect of interlayer mixing and finite size effects on FeZr in Fe 92 Zr 8 /AlZr multilayer films, and found an anomalous increase of T c with decreasing thickness. Strain induced changes in the magnetization of an amorphous Co 95 Zr 5 film at the orthorhombic phase transition of the BaTiO 3 substrate, was also studied. The results show that structural modifications of the substrate increases the stress and hence changes the magnetic anisotropy in the amorphous Co 95 Zr 5 layer. Finally, the magnetization reversal of Co and CoX heterostructures, with X being Cr, Fe, Ni, Pd, Pt and Ru, has been studied. For this purpose a synthetic antiferromagnet structure, FM/ NM/FM, was used, where FM is a ferromagnetic Co or CoX layer and NM is a nonmagnetic Ru spacer layer. The FM layers are coupled antiferromagnetically across the NM layer. For a range of FM layer thicknesses, the exchange stiffness parameter A ex and the interlayer coupling (J RKKY ) of the Co or CoX layers were obtained. This is done by fitting M(H) curves, measured by SQUID magnetometry, to a micromagnetic model. The alloying in CoX resulted in a decreasing A ex and also a reduced M S. The experimental results are in a good agreement with DFT calculations. Keywords: Amorphous, Thin Film, Magnetic properties, FeZr alloys, Ion implantation Atieh Zamani, Department of Physics and Astronomy, Materials Physics, 516, Uppsala University, SE Uppsala, Sweden. Atieh Zamani 2015 ISSN ISBN urn:nbn:se:uu:diva (

3

4

5 List of papers This thesis is based on the following papers, which are referred to in the text by their Roman numerals. I Tuning magnetic properties by hydrogen implantation in amorphous Fe 100 x Zr x thin films Atieh Zamani, Anders Hallén, Per Nordblad, Gabriella Andersson, Björgvin Hjörvarsson and Petra E. Jönsson Journal of Magnetism and Magnetic Materials, 346, (2013) II Magnetic properties of light ion implanted amorphous Fe 93 Zr 7 films Atieh Zamani, Reda Moubah, Martina Ahlberg, Henry Stopfel, Unnar Arnalds, Anders Hallén, Gabriella Andersson, Björgvin Hjörvarsson and Petra E. Jönsson (Submitted to Journal of Applied Physics) III IV V Soft Room-Temperature Ferromagnetism of Carbon-Implanted Amorphous Fe 93 Zr 7 Films Reda Moubah, Atieh Zamani, Anders Olsson, Shengwei Shi, Anders Hallén, Stefan Carlson, Dimitri Arvanitis, Per Nordblad, Björgvin Hjörvarsson, and Petra E. Jönsson Applied Physics Express, 6, (2013) Origin of the anomalous temperature dependence of coercivity in soft ferromagnets Reda Moubah, Martina Ahlberg, Atieh Zamani, Anders Olsson, S. Shi, Z. Sun, Stefan Carlson, Anders Hallén, Björgvin Hjörvarsson, and Petra E. Jönsson J. Appl. Phys. 116, (2014) Reversed interface effects in amorphous FeZr/AlZr multilayers Martina Ahlberg, Atieh Zamani, Erik Östman, Hossein Fashandi, Björgvin Hjörvarsson and Petra E. Jönsson Phys. Rev. B. 90, (2014) VI Strain induced changes in magnetization of amorphous Co 95 Zr 5 based multiferroic heterostructures

6 Reda Moubah, Fridrik Magnus, Atieh Zamani, Vassilios Kapaklis, Per Nordblad, and Björgvin Hjörvarsson AIP Advances, 3, (2013) VII Effects of substitution on the exchange stiffness and magnetization of Co films C. Eyrich, A. Zamani, W. Huttema, M. Arora, D. Harrison, F. Rashidi, D. Broun, B. Heinrich, O. Mryasov, M. Ahlberg, O. Karis, P. E. Jönsson, M. From, X. Zhu and E. Girt Phys. Rev. B. 90, (2014) Reprints were made with permission from the publishers. Papers not included: I coauthored the following articles that are not included in this thesis: Intense Quantum Confinement Effects in Cu 2 O Thin Films Panagiotis Poulopoulos, Sotirios Baskoutas, Spiridon D. Pappas, Christos S. Garoufalis, Sotirios A. Droulias, Atieh Zamani, and Vassilios Kapaklis Journal of Physical Chemistry C, 115, (2011) Using light transmission to watch hydrogen diffuse Gunnar K. Pálsson, Andreas Bliersbach, Max Wolff, Atieh Zamani, and Björgvin Hjörvarsson Nature Communications 3, 892 (2012) Thermally driven redistribution of phases and components in Cu/Sn thin films Anna Oleshkevych, Atieh Zamani, Igor Kotenkob, Svitlana Voloshkob, Sergey Sidorenkob, and Adrian R. Rennie Journal of Alloys and Compounds, 535, (2012) Broadband ferromagnetic resonance system suitable for a wide range of ultrathin films Eric Montoya, Tommy McKinnon, Atieh Zamani, Erol Girt, and Bret Heinrich Journal of Magnetism and Magnetic Materials 356, (2014) Comments on my own participation I Responsible for the sample preparation, structural and magnetic measurements and manuscripts preparation.

7 II Responsible for the sample preparation and magnetic measurements and manuscripts preparation. III Responsible for the sample preparation and magnetic measurements. IV Responsible for the sample preparation and took part in structural and magnetic measurements. V Responsible for sample preparation and took part in structural and magnetic characterizations. VI Took part in sample preparation. VII Took part in sample preparation, structural and magnetic characterizations and data analysis.

8

9 Contents 1 Introduction Magnetic Properties Exchange Interaction and magnetic ordering Interlayer Exchange Coupling Exchange Stiffness Constant Magnetism of 3d metals and alloys Anisotropy and Demagnetization Energies A Micromagnetic Model Critical exponents and thin films Sample Preparation and Material Composition Analysis Sample deposition by magnetron sputtering Rutherford Backscattering Spectroscopy Ion Implantation Sample Structural Characterization Thin film thickness from X-ray reflectivity (XRR) Magnetic Characterization SQUID Magnetometry MOKE Magnetometry Magnetic Measurements Summary of Results Paper I, II, III, IV and V Paper VI Paper VII Summary and Conclusion Outlook Summary in Swedish Acknowledgement References... 56

10

11 1. Introduction Magnetism has been known since ancient times, when the property of magnetism in lodestone mineral was discovered in Greece. To reveal and use the hidden forces of magnetism has since been an obsession to scientists. The early studies of magnetism gave a basis for understanding the origin of ferromagnetism in metals such as Fe, Co and Ni. In general, pure single element metals are rarely used in applications. This is because any specific property of a metal can usually be tailored by alloying one or several other elements. Sometimes, these added elements not only affect the physical properties but can also change the crystal structure, and amorphous alloys are a typical example of that. In amorphous (non-crystalline) solids, the positions of the atoms do not form a well defined pattern, therefore, the amorphous solid lacks long range order and has order only within a short-range (atomic-scale). [1 3] Thus their internal random structure resembles a fluid but their external structure exhibits the rigidity of a solid. [4] Metallic amorphous alloys were first synthesized in the 1960 s. The process required extremely fast cooling of the alloy melt (10 5 K/s) in such a way that during the solidification there is not enough time for ordering of the atoms into a crystalline form. [4 6] The first amorphous and ferromagnetic alloy with a substantial magnetization, Fe 75 P 15 C 10, was prepared and reported by Duwez and Lin in [7] There are several techniques to produce amorphous materials, such as condensation of metal vapors and sputtering techniques. Recently, materials engineering has produced alloys which can be cooled by a few K/s and still exhibit an amorphous structure and therefore allows for the production of much larger amorphous structures. [8] Alloys that can be formed into an amorphous structure at slow cooling rates are often called bulk metallic glasses (BMG). [4 6,9] Bulk metallic glasses contain two or more elements that show geometric size differences between the constituent elements, where this size difference facilitates the formation of the amorphous phase. Ferromagnetic amorphous alloys are formed by combining transition metals (Fe, Co, Ni) with good glassformers such as Zr and Pd. [10, 11] Systematic studies have been performed and multicomponent Fe [12]-, Ni [13]-, Al [14]-, Mg [15]- based bulk metallic glasses have been discovered and extensively investigated. The exact chemical compositions of these compounds strongly influence the amorphization. In the present work, the formation of metallic amorphous iron is stabilized due to the addition of Zr atoms in the growth process. Already a small amount of Zr addition (about 7 at. % [16,17]), is enough to form amorphous Fe 100 x Zr x due 11

12 Figure 1.1. Two-dimensional illustration of (a) substitutional disorder in a crystalline lattice, (b) atomic structure of a binary amorphous metallic alloy. to the size-mismatch between these two elements, where the Zr atomic size is considerably larger than Fe. Disorder is one of the main characteristics of amorphous alloys. An example of disorder within crystalline materials is chemical or substitutional disorder, illustrated in Fig. 1.1(a). Here the crystal structure is still preserved despite the presence of the solute atoms. The atomic arrangements in amorphous materials, illustrated in Fig. 1.1(b), have no long-range order, but still shows short-range ordering between the near neighbors. Since amorphous solids lack the basic unit cell found in crystalline structures, they cannot be defined by the use of primitive vectors and a basis set. They can however be described by the average number of atoms that are distributed within a shell of radius r around a central atom, as indicated in Fig. 1.1(b). Experimentally, determining atomic-level structure of amorphous materials is not an easy task. One may look at how the probability of finding atoms varies as a function of distance from a given reference atom, which defines the radial distribution. Specifically, it provides information on bond distances and the coordination number of atoms located at various distances from a central atom. For structural characterization in amorphous FeZr, this type of analysis has mainly been used to obtain structural insight on the overall amorphous nature of our samples. Due to the spatial homogeneity of the disordered atomic structure, the amorphous alloys ideally should exhibit isotropic magnetic properties. [18 21] On the other hand, in crystalline and nano-crystalline materials, the presence of crystalline grains strongly affect the magnetic properties. [22] This leads to a grain size effect on the magnetic hardness (the coercivity, H c ), where, in the range nm, the coercivity rapidly decreases with decreasing grain size. [23 26] Both amorphous and nanocrystalline magnetic metallic materials are thus useful for applications that require soft magnetic properties. For instance, bulk Fe-based amorphous alloys are of potential use in inductors [8], in high power transformers [27], and magnetocaloric applications [28]. Since 12

13 such alloys can exhibit varying degrees of short-range order, from purely amorphous to nanocrystalline, their properties are adjustable in a way that is impossible for a crystalline bulk material. [18 21] The amorphous alloys also exhibit properties which are very interesting for micro- and nano-scale patterning. This is due to their lack of crystalline order and associated defects such as dislocations, hence there are no grain boundaries or lateral imperfections that limit the structuring. Also, amorphous metals can be found in various applications such as sport equipment, and scalpels due to their mechanical properties. [8, 29] In the field of material science, the ability to affect material properties by decreasing the material thickness has had immense importance. [30 32] Thin films provide a good platform to study the properties of matter and has been the geometry of many samples studied in this thesis. Thin film technology is today used in many applications and is especially interesting for microelectronic and magnetic recording-devices, having properties that are very sensitive to the thickness. [33, 34] In this thesis I present studies of thin magnetic films and amorphous materials, which have been subjected to doping and to thickness variations. Bulk like amorphous FeZr implanted with different light dopants FeZr/AlZr multilayers to study finite size effects CoZr to study substrate induced strain in the magnetic layer Crystalline Co(X)/Ru/Co(X) trilayers to study exchange stiffness and magnetization The results are described in detail in Chapter 6. All the thin films and multilayer structures studied in this thesis were deposited using magnetron sputtering. The thickness of the amorphous samples and their textured structure characteristics were investigated by using X-ray reflectivity. Additional structure characterization using Extended X-Ray Absorption Fine Structure (EXAFS) measurements was used to study the Fe local environment. The magnetic properties were investigated using MOKE and SQUID magnetometry. 13

14 2. Magnetic Properties Magnetism in solids can largely be reduced to atomic scale properties. The early description of magnetism was based on the magnetic field created by charge traveling in a circular orbit around the nucleus of an atom. This magnetic moment is associated with the orbital motion of electrons, called orbital magnetic moment. Moreover, electrons possess an intrinsic magnetic moment, called spin. The total magnetization of a material is therefore provided by the sum of these two principal sources, as illustrated in Fig Figure 2.1. Picture of magnetism in the atomic level, illustrating the orbital and spin contributions to the magnetic moment. The total magnetic moment of an atom is determined by the sum of magnetic moments from all of its electrons. However, the inner electron shells will be fully occupied. Since there is an equal amount of spin up and spin down states in every shell, the magnetic moments in the filled shells are cancelled out. Therefore, only the outermost not fully occupied shell (or shells) is of interest for magnetic properties. For the 3d transition ferromagnetic metals (Fe, Co and Ni), there is an unfilled d-shell that has five orbitals, each can be 14

15 filled by two electrons. According to the Pauli exclusion principle, two electrons within one orbital must have opposite spin directions. The occupation of these orbitals for single atoms is determined by Hund s rules. [35] Based on Hund s first rule, parallel spins will try to singly occupy different orbitals and remain unpaired before any orbital becomes doubly occupied. This is no longer valid once the single atom is placed within a bulk solid since the magnetic moments are not isolated from one another and the interaction between them is considerable. Considering a crystal lattice, the surrounding electric field of the crystal, i.e. the electric field that one atom feels in the vicinity of other atoms, needs to be taken into account. A very important consequence of this is a quenched orbital magnetic moment. For example in 3d metals, the orbital magnetic moment is small and the magnetization is mainly carried by spin moments. In a simplistic picture the quenched orbital magnetic moment is associated with a new set of orbitals, due to breaking of spherical symmetry inside the crystal. Importantly, these new orbitals have zero orbital angular momentum and therefore no orbital magnetic moment. However, spin-orbit coupling will generate the small orbital magnetic moment found in 3d metals. Spin-orbit coupling will also couple the spin magnetic moment to the lattice. The symmetry of the lattice then determines the preferred magnetization direction (the easy axis directions), known as the magnetocrystalline anisotropy. [36 38] There are two especially important models when describing the magnetic behavior of materials: the Heisenberg model and the Stoner model. The Heisenberg model describes the exchange interactions between localized magnetic moments and Stoner model describes the interaction between itinerant moments (delocalized moments). The two different models are successful in describing different aspects of magnetic properties. For instance Stoner model can describe the formation of magnetic moments however it does not describe the temperature effects very well, while the Heisenberg model can capture local variations in magnetic interactions and can estimate the ferromagnetic ordering temperature (T C ). We begin with a discussion on Heisenberg model and itinerant behavior will be discussed in section Exchange Interaction and magnetic ordering The exchange interaction energy determines the relative spin orientation between two localized magnetic moments at different atoms. [36, 37, 39] The exchange Heisenberg Hamiltonian based on spin-spin interactions is expressed in a vector model as: E ex = J ij Ŝ i Sˆ j, (2.1) <ij> where the sum is over all spin pairs < ij > of i-th and j-th atoms, J ij is the exchange coupling parameter between magnetic sites, Ŝ i and Sˆ j are the spin 15

16 unit vectors of atoms at sites i and j respectively. [36] If J ij has a positive value, the lowest energy state results from a parallel alignment of the spin moments which gives rise to ferromagnetic (FM) order. If J ij has a negative value, the spins of adjacent atoms align antiparallel, i.e., giving rise to antiferromagnetic (AF) order. The Heisenberg model is strictly speaking valid for magnetic insulators where the magnetic moments are localized at atomic sites and where these local magnetic moments are coupled by a pair exchange interaction. However, it can also be applied to 3d transition metals. The Bethe-Slater curve (Fig. 2.2) qualitatively describes the variations in strength of the direct exchange as a function of the ratio of the interatomic separation (D) to the radius of the 3d shell (d). [40] The main trend can be explained by using a simple pair interaction of two atoms sharing two electrons. For the parameter D/d <1.5 the electrons from two neighboring atoms are forced to stay close to each other and the Pauli exclusion principle requires the spins of electrons to be antiparallel, which will result in an antiferromagnetic interaction between the atoms. Once the D/d ratio increases, 3d electrons move away from each other, occupying two different orbital states. Therefore, a ferromagnetic state with the exchange coupling (J) becoming increasingly positive, is allowed. After reaching a maximum, the exchange coupling starts to decrease due to a decreasing spatial overlap of the wave functions of the electrons. At large interatomic distances, the atoms cannot feel each others presence, hence the exchange energy goes to zero. As an example, the exchange interaction between Fe atoms is very sensitive to their interatomic distance and generally shows an increased ferromagnetic (FM) interaction with increasing atomic separation. The Curie temperature is proportional to the pair exchange interaction energy J,[41] T C = J Z S2 3k B, (2.2) where J = J ij (r ij ) is the mean value of the exchange coupling parameter, Z is the coordination number, S is the magnitude of the magnetic moment and k B is the Boltzmann constant. The ferromagnetic transition temperature is therefore determined by the exchange interactions, where a stronger exchange interaction leads to a higher T C.[39] According to the Bethe-Slater curve, the exchange interaction for Fe is around the zero-crossing point of the curve and increases almost linearly with increased interatomic separation and also changes the sign from negative to positive. E.g., fcc-fe is AF with a Néel temperature of 70K, while bcc-fe is FM with a Curie temperature of 1043K. [42] The amorphous Fe 100 x Zr x system has been considered as a model system to study the magnetic properties of Fe in an amorphous environment. [43 55] It behaves as a ferromagnet, due to a dominance of FM direct exchange cou- 16

17 Figure 2.2. The Bethe-Slater curve, shows the dependence of the exchange interaction on the ratio of interatomic separation to the diameter of the 3d shell. [40] pling, even though the atomic arrangement allows a co-existence of both FM and AF exchange interactions due to variations of the Fe-Fe interatomic distances. The atoms at the shortest distance are antiferromagnetically coupled and the more distant ones are ferromagnetically coupled. This inhomogeneity of magnetic interactions creates an exchange frustration, which leads to noncollinear ferromagnetism in amorphous FeZr. Figures 2.3 a and b summarizes the values of T C and M S, respectively, for several studies of the amorphous Fe 100 x Zr x system. The magnetic ordering temperature and the magnetic moment of Fe 100 x Zr x are both dependent on the alloy composition. The Curie temperature and saturation moment increases with Fe concentration, up to a maximum value at about x 13. For low Zr concentrations (x < 13), the antiferromagnetic interactions rises with increased Fe concentration and makes it harder to reach the saturation magnetization. Also, since the average J decreases, the T C goes down as well. Furthermore, there are varying results regarding the evolution of the magnetic moment per Fe atom, as indicated by measurements using very high magnetic fields (up to 11T), showing an increase of the as determined moment above the composition of x 13 (see Fig. 2.3 b). [45, 48 55] Interlayer Exchange Coupling Layered structures consisting of two magnetic layers separated by a metallic non-magnetic spacer layer, are found to experience an interlayer exchange 17

18 T C (K) 280 our group buschow shirakawa83 fukamichi castano97 coey84 heiman unruh84 ryan ma Fe (at%) a Mag. mom. (µ B /Fe) heiman79 heiman79 fukamichi81 shirakawa83 coey84 unruh84 ryan87 castano97 buschow Fe (at%) Figure 2.3. Literature results on a) the Curie temperature and b) the magnetic moment in amorphous Fe 100 x Zr x.[43 55] b coupling of Ruderman-Kittel-Kasuya-Yosida (RKKY) type. The interlayer exchange coupling energy [56, 57] E RKKY = J RKKY ˆM i ˆ M j, (2.3) depends on the exchange coupling constant J RKKY and the angle between the directions of the magnetic moments in the two magnetic layers at their respective spacer layer interface. ˆM i and Mˆ j are the unit vectors for the magnetization at layer i and layer j, respectively. RKKY interaction between the magnetic layers is mediated by spin polarization of conduction electrons in the metallic spacer layer. The coupling strength is given by J RKKY, which oscillates in sign as a function of the non-magnetic spacer thickness, i.e. leading to either a parallel or antiparallel magnetization alignment of the magnetic layers with respect to each other. [56, 57] 2.2 Exchange Stiffness Constant The coupling term J from Equation 2.1 is a measure of the interaction strength between the two neighboring atomic spins, which does not consider the number of nearest neighbor pairs. To describe how variation in magnetization directions between atoms, for example in domain walls, evolve spatially there exists a parameter called exchange stiffness constant (A ex ). [39] The exchange stiffness serves as a material specific parameter of a ferromagnet and it scales with the exchange energy (E ex ). It can be derived from the Heisenberg model by an appropriate averaging process using a magnetization density which is slowly varying in space 18 A ex = nj S2, (2.4) a

19 where n is the number of nearest neighbors, and a is the interatomic lattice parameter. [39] This is for instance important when describing the exchange energy in spin waves, which have spatially varying directions of the magnetization. Moreover, A ex has important role in the magnetization reversal process and can be applied to spintronic devices, where the switching current density is very sensitive to the exchange stiffness constant. Typically in a bulk sample, A ex can be quantified using experimental techniques like ferromagnetic resonance (FMR), Brillouin Light Scattering (BLS), or neutron scattering (NS) by magnon excitations. The energy for such excitations is inversely proportional to the film thickness. As a result, A ex is difficult to measure in thin films, since the mentioned techniques cannot probe the required energy range. Therefore a method has been developed [58] for estimating the exchange stiffness in thin films, which has been used in paper VII, by measuring the magnetization in antiferromagnetically coupled FM Co layers. The basis for this method is the micromagnetic model, described in detail in section Magnetism of 3d metals and alloys Generally, the magnetic behavior of an atom in solids and compounds is different from that of a single atom. This is because the exchange interaction will correlate the magnetic moments between neighboring atoms, so the magnetism becomes a collective effect. The only pure metals that are ferromagnetic at room temperature are Fe, Co & Ni, which exhibit strong permanent (spontaneous) magnetization even in the absence of any applied magnetic field. The magnetic moment of these 3d-metals can only be explained by nonlocalized 3d electrons, as shown by Stoner. [35] These itinerant electrons can be described by band structures as shown for bcc Fe in Fig The partially filled d-bands are split into two sub-sets, which contains the majority (spin-up) and minority (spin-down) states. Stoner first proposed a basic model for describing the itinerant ferromagnetism based on the partial occupation of these states. The occupation number (N) of electrons for each spin direction can be expressed by: N = E F D (E)dE, where D(E) is either the majority ( ) or minority ( ) electron density of states (DOS). Finally the magnetization (M) arises from the fact that there are more majority than minority electrons below the Fermi level (M = µ B (N N )). The Stoner criterion (N(ε F ) I ex 1), where I ex is the Stoner exchange integral, gives the necessary condition for ferromagnetism. [35] The d-bands of transition metals have a large density of states at the Fermi level N(ε F ), which is necessary for stabilizing a ferromagnetic order. However, a high DOS at the Fermi level is not enough to establish ferromagnetism, there also needs to exist a sufficient exchange interaction (I ex ). For Fe, Co and Ni the Stoner criterion is fulfilled, which gives an exchange splitting of about 1-2 ev between 19

20 Figure 2.4. Density of states (DOS) of bcc Fe, for both spin up (majority) and spin down (minority) states. The non-magnetic state and ferromagnetic state correspond to the dashed and solid lines, respectively. The Fermi level is indicated by ε F. The density of states at the Fermi level, in the non-magnetic state, is large enough to fulfill the Stoner criterion. The figure is adopted from [35]. the spin sub-bands. This results in a large magnetic moment per atom and consequently an appreciable magnetization. The magnetization of these 3d metals can be controlled by alloying. The 3d transition metal alloys (combining different transition metals) follow the Slater-Pauling curve, shown in Fig According to the Slater-Pauling curve, there is a relation between the magnetization (average magnetic moment) and the number of valence electrons per atom. For the left side of the curve one can find the weak ferromagnetic Fe based alloys, while the right side of the curve is based on the strong Co and Ni based alloys. Strong ferromagnet refers to the case where the majority band (spin up) is fully occupied, while the weak ferromagnet has empty states in the majority band which can be further filled. This is illustrated for bcc Fe (weak ferromagnet) in Fig The Fe magnetic moment for some selected Fe based compounds found in the Slater-Pauling curve are shown in Table 2.1, where one finds that the Fe magnetic moment has variation depending on the crystalline structure, concentration and the alloying element. Since Fe generally has the highest magnetic moment, these values are higher than the average magnetic moment per atom found in the Slater-Pauling curve. As an example, for the case of Fe in Ni 80 Fe 20, there is a significant enhanced magnetic moment compared to both fcc and bcc- Fe. There is some degree of charge-transfer between the alloying element and the host metal which can affect the magnetization. Similarly, it has been proposed that charge transfer from Zr to Fe in amorphous FeZr alloys has an 20

21 Figure 2.5. The Slater-Pauling curve shows how the net magnetic moment per atom varies as a function of the number of valence electrons per atom. The figure is adopted from [35]. Element Structure M s (µ B /Fe atom) Fe fcc 1.5 Fe bcc 2.23 Fe 50 Co 50 bcc 2.7 Fe 90 Ni 10 bcc 2.41 Ni 80 Fe 20 fcc 2.7 Table 2.1. The magnetic moment per Fe atom for different Fe alloys. The results are for bulk samples [59 64], except for Fe which is obtained from Fe grown on Ni [60, 61]. impact on the magnetization. [65 67] As illustrated in Fig. 2.6, neither the majority or minority bands of Fe are completely filled, and the presence of Zr partially fills these bands due to charge transfer. Therefore, the magnetic moments of the amorphous FeZr alloys become smaller than those of the Slater- Pauling curve. After doping with C or B, these dopant atoms can act as acceptors for electrons, as illustrated in Fig This can decrease the number of electrons in the minority band, resulting in increased magnetic moment. Figure 2.6 illustrates the above descriptions, however it may not capture the full complexity. Regarding the influence of other dopants in the amorphous 21

22 Figure 2.6. Density of states (DOS) of Fe to describe the magnetic properties of amorphous Fe(Zr) and FeZr-B or -C alloys, where alloying will affect the occupation of spin up and down states. FeZr system, it is not yet known if charge transfer effects has any influence on magnetic properties. 2.4 Anisotropy and Demagnetization Energies The magnetic anisotropy energy defines the directional dependence of magnetic properties in a ferromagnet, and arises from two different effects: magnetocrystalline anisotropy (MCA) and shape anisotropy. [39] MCA is an intrinsic property of a ferromagnet, that arises from the crystalline structure of the material. The energetically preferred direction for the magnetization is called the easy axis. The easy and hard axis arise from the interaction of the spin magnetic moment with the crystal lattice (through spin-orbit coupling). In Fig. 2.7, we show how the easy axis directions (red/grey arrows) direction depends on the crystal structure and also how the magnetization generally depends on the magnetic field direction relative to the easy axis. Figure 2.7. The easy axes and hard axes denotes the preferred and unfavorable directions of the magnetization respectively. Magnetization is in the direction which corresponds to an energy minimum. For single-crystal hcp Co, the easy axis is parallel to the hexagonal c-axis of the crystal. In bulk single-crystal bcc Fe (cubic anisotropy), the easy directions are <100> and the hard directions are <111>. [38] 22

23 When a material has only one easy axis, it is said to have uniaxial magnetic anisotropy. According to eq. 2.5, the uniaxial anisotropy energy (E An ) can be identified in terms of the anisotropy constants (K ui ). E An = K u1 ( ˆn Ŝ) 2 K u2 ( ˆn Ŝ) 4... (2.5) where ˆn is the direction of the easy axis and Ŝ the magnetization direction. E An governs the energy cost of rotating the magnetization away from its easy axis, which is consequently an influencing factor for the size and shape of the hysteresis curve in ferromagnetic materials. [39] As an example, for the single-crystal hcp Co, the easy axis is parallel to the hexagonal c-axis of the crystal. At room temperature, K u1 for Co is erg/cm 3 and M S is 1247 emu/cm 3 (paper VII), the anisotropy field (H K = 2K u1 M s ) is then 3.4 koe in the direction of the c-axis (K u2 and higher order terms are small and can be neglected). The anisotropy field determines the effective field which is trying to line up the magnetic moment along the easy axis. Therefore one needs an appreciable field in the bulk Co to rotate the magnetic moment away from the c-axis. The other effect one has to consider is shape anisotropy, which depends on the shape of the sample. In a magnetic sample, the magnetic field produced by the free poles at the exterior surface of the sample opposes the magnetization in the sample, creating a demagnetizing (opposing) field of H dem = N M, where N is a tensor that depends on the sample shape. In the case of thin films, when the thickness of the film is very small compared to the lateral size, the demagnetizing field becomes negligible for in-plane magnetization (N = 0). When the magnetization is oriented perpendicular to the film surface, by an external field, then the magnetic surface charges create a strong demagnetizing field (N = 4π) in the out-of-plane direction. For Co films fully saturated along the surface normal the demagnetization field is approximately H dem = NM S = 4π 1247 emu/cm 3 = 15.7 koe. (2.6) As an example, for the thin Co films presented in paper VII, since the demagnetization field is four times larger than the anisotropy field perpendicular to the film plane, the orientation of the magnetic moment will be in-plane. The shape of a ferromagnetic material can change when the magnetization is aligned along an external magnetic field. This phenomenon is called magnetostriction. A related effect, called magnetoelastic anisotropy, occurs when magnetic materials are subjected to stress that can change the magnetization direction. [39] The magnetoelastic energy per unit volume in an isotropic elastic material with uniaxial stress σ, is given by [68] E me = K me ( ˆσ Ŝ) 2, where K me = 3 λσ, (2.7) 2 23

24 where λ is the magnetostriction coefficient, ˆσ is the unit vector along the stress and Ŝ is the unit vector along the magnetization. Magnetoelastic anisotropy plays an important role in thin magnetic films, since stress can be induced by the substrate. [39] Magnetoelastic anisotropy in CoZr grown on BaTiO 3 is studied in paper VI. 2.5 A Micromagnetic Model We have used a micromagnetic model [58] for obtaining the exchange stiffness of Co thin films studied in Paper VII. This model is based on the formation of spin spirals within two ferromagnetic layers that are antiferromagnetically coupled across a non-magnetic spacer layer. The magnetization is assumed to be in-plane, which is reasonable since the shape anisotropy in the Co layers is higher than the magnetocrystalline and surface anisotropies. In this model, each Co layer consists of N atomic planes, where each plane is acting as a macro spin that interacts only with its nearest neighbors through the direct exchange interaction shown in Fig. 2.8.!!!!!!!!!!!!!!!!! "#!!! $%!!! "#!!! $%!!! "#! Figure 2.8. This figure qualitatively demonstrates the formation of exchange-spring in CoX(FM)/Ru(NM)/CoX(FM) trilayer, when subjected to an external magnetic field. An index i, is used for labeling the atomic planes in FM layers. As assumed by the micromagnetic model, the spins within each atomic plane rotate coherently. This figure is re-printed from [69]. 24

25 One should note that the micromagnetic model does not consider the magnetocrystalline anisotropy (MCA) contribution. All the layers are textured along the [0001] direction of the hcp crystal structure, i.e. the [0001] easy axis is oriented perpendicular to the film surface. Therefore, in-plane rotation of magnetic moment is not affected by the presence of MCA in CoX films, with X being Cr, Fe, Ni, Pd, Pt and Ru. Moreover it is assumed that the magnetic moments in each atomic plane rotate coherently. By applying the field in-plane, the total energy per unit area is E Mag = E RKKY E ex E z E RKKY = J RKKY cos(θ N θ N+1 ) E ex = 2A N 1 ex d cos(θ i θ i+1 )+ i=1 E z = M s Hd 2N i=1 cos(θ i ), 2N 1 i=n+1 cos(θ i θ i+1 ) (2.8) where E RKKY is the (here antiferromagnetic) coupling between the ferromagnetic sub-layers separated by non-magnetic spacer layers. E ex refers to the direct exchange interaction between nearest neighbor atomic planes. θ i is the magnetization angle of plane i with respect to the applied magnetic field. E z is the Zeeman energy acting on all magnetic atomic planes per unit area. H is the applied magnetic field and d is the thickness of each magnetic layer. In order to calculate M(H), first the magnetization angles of each sublayer within the Co layers as a function of the external field is determined. This can E be done by minimizing eq. 2.8 with respect to θ i, i.e. mag θ i = 0. Since each atomic layer forms its own angle with respect to the direction of the external field, one can calculate the total magnetic moment M(H) according to M(H)= M 2N s 2N i=1 cos(θ i ), (2.9) which can be directly compared to magnetization measurements. When H = 0, then θ i = π/2 for i = 1 to N and θ i = π/2 for i = N +1 to 2N, and M(H)=0 due to the antiferromagnetic RKKY coupling across the non-magnetic spacer. The saturation magnetization M S was obtained by measuring M(H) in very high fields using a sensitive magnetometer. The only remaining and unknown parameters in eq. 2.8 are A ex and J RKKY, which are used as fitting parameters. The fitted values of A ex, from paper VII are presented in section

26 2.6 Critical exponents and thin films Close to the magnetic phase transition, ferromagnetic systems exhibit critical behavior that can be characterized by a set of critical exponents. The magnetization (M), susceptibility (χ), and correlation length (ξ ), vary as power laws, in the vicinity of T C [38]: M(T ) (T C T ) β T < T C χ(t ) (T T C ) γ T > T C (2.10) ξ T C T ν T = T C The spin correlation length (ξ ), is the maximum distance over which the magnetic moments (or spins) are correlated. The critical exponents can be obtained, for example, from mean field calculations, which gives β = 1/2, γ = 1 and ν = 1/2. [38] Thin films consist of a single or a few layers of atoms, where the spatial dimension in the out-of-plane direction is much smaller than in the in-plane direction. The out-of-plane dimensional reduction leads to a reduced spin correlation in that direction. The magnetic ordering temperature decreases due to the geometric confinement and this is known as a finite size effect, which is related to the reduced spin-spin correlation length (ξ ). [30] The ferromagnet transition temperature then has a power law behavior T C (d)/t C ( )=1 c 0 d λ, (2.11) where T C ( ) is the bulk Curie temperature and T C (d) is the transition temperature of the material with thickness d, c 0 is a constant and λ = 1/ν is a shift exponent. [70] 26

27 3. Sample Preparation and Material Composition Analysis 3.1 Sample deposition by magnetron sputtering For the purpose of thin film sample preparation, sputtering is a standard tool and the most commercially practiced coating method, resulting in films with the same stoichiometry as the source. [71, 72] The deposition process takes place under vacuum condition, which reduces possible impurities in the film. Originally sputtering was performed by the diode sputtering method. However it was found that a magnetic field can improve the deposition process, providing a new technique called magnetron sputtering. A magnetron consists of a target to be deposited on a substrate, with permanent magnets located behind the target, as illustrated in Fig The target atoms are sputtered by gas ions, and deposited onto the substrate. Generally a chemically inert (noble) gas is used, e.g. Ar, to avoid unwanted chemical reactions during the growth process. The target material is held at a negative potential up to a couple of hundred volts, which constitutes a cathode. Positively charged Ar ions are accelerated toward the negatively biased target (cathode), creating secondary electrons when Ar + ions bombard the target. These secondary electrons gain energy from the applied potential and will continue to further ionize the Ar gas by inelastic scattering. As illustrated in Fig. 3.1, the magnetic field confines these electrons in a close vicinity of the target, leading to creation of a dense plasma in front of the sputtering target. Consequently the Ar + ion bombardment of the target is increased. When collisions between target atoms and the incident Ar + ions takes place, the atoms are knocked out from the target if the kinetic energy is large enough. Since these sputtered atoms are not charged, they travel straight out of the magnetic trap to coat the substrate and form the film. The target composition is maintained during the process and the deposition rate scales with the Ar pressure and the electrical power used to sustain the plasma. The planar rotation of the substrate during growth creates films of uniform coating thickness. The main advantage of magnetron sputtering is that the magnets, which are located behind the target plate, cause the electrons to be trapped in a helical path (see Fig. 3.1), until striking the Ar atoms and 27

28 ,. / $ + $ +,-,. / + $!" # + $ $ +!"!" # $ + $%&'(")(*!! Figure 3.1. A cross-sectional schematic of two planar circular magnetrons during the co-sputtering of an alloy. Both the target material and the substrate are placed inside an ultra-high vacuum chamber. The vacuum chamber used for the deposition is filled with Ar gas. The Ar + positive gas ions are created by collisions of Ar atoms with electrons and accelerated towards a negatively charged target, bombarding the target. Magnetic fields, close to the target, trap electrons which increases the ionization of Ar atoms. The target atoms that are sputtered away condense on the rotating substrate after a series of collisions with inert gas atoms. Some of the power is dissipated as heat in the target, so cooling is essential. This figure is a re-print from [73]. creating Ar + ions. This leads to a higher probability of collisions between the electrons and Ar atoms, therefore more target atoms are sputtered and deposition rate is increased. Also, due to the presence of the magnetic field, a lower gas pressure can be used in the chamber and the system can be operated at lower voltages. 3.2 Rutherford Backscattering Spectroscopy Rutherford Backscattering Spectroscopy (RBS) is a characterization technique used for quantitatively identifying elemental concentrations within a solid material. RBS characterization is based on binary Coulombic collision between target (sample) atoms and high kinetic energy ( MeV), typically 4 He + particles. These particles (incident ions) are positively charged, impinging on the sample with energy E i, as illustrated in Fig The energy of the backscattered particles (E s ) depends on the mass of the atom in the sample (M 2 ) from which they scatter. Moreover, the ions will lose energy in electronic interactions as they pass through the sample. The energy of the backscattered particles will therefore depend on the depth of the collision, in addition to the 28

29 Figure 3.2. An incident ion with mass M 1 and initial kinetic energy E i impinging the sample is scattered out of the sample (backscattered) with energy E s and is registered by a detector. The detector records their energy and the number of backscattered particles in a certain solid angle. Due to conservation of momentum and energy, the mass of the target atom can be extracted. Furthermore, since the Rutherford cross section is well known and the number of incident particles can be determined, the depth profile (concentration) of mass M 2 can be determined from the yield measured in the detector. mass M 2. Since the energy of the scattered ion is measured, the electronic loss of the incident ion is obtained and can be used to calculate the depth where the collision took place, if the target material is known. Thick layers will exhibit a broad peak in the RBS spectrum, where the low energy edge provides information on the depth of the inner interface and the high energy edge corresponds to the scattering from the interface closer to the surface. The energy of the backscattered particles from a light element is lower than particles scattered from a heavy element, which makes it possible to separate various masses in composite targets. The probability that the incident particle backscatter by the sample atoms can be determined by the interaction cross section, which is known for all elements in the periodic table. The number of backscattered particles is proportional to the concentration of atoms in the sample and by comparing to computer simulations, quantitative depth profiles can be derived from experimental data. RBS is ideally suited for analyzing films containing heavy elements on a substrate of light elements, for instance metal layers on a Si substrate. For multi-elemental samples and films of lower mass elements on a heavier substrate it can be difficult to resolve the individual elements. As an example, the RBS spectra of FeZr-film on a Si substrate, with a 4 nm buffer and a 5 nm capping layers of AlZr is illustrated in Figure 3.3. The yield 29

30 Figure 3.3. The figure shows an RBS spectrum (red/grey dotted line) from a sample consisting of a FeZr (40 nm) layer sandwiched between AlZr capping (5 nm) and buffer (4 nm) layers on a Si substrate. The analysis was done with a normal incidence 2.0 MeV 4 He ion beam and the detector at a backscattering angle of 170. By fitting the experimental data using SIMNRA software [74] (black thicker line), it can be concluded that the film composition is Fe 89 Zr 11. of backscattered particles is seen as a function of their energy, expressed in Channel no.. By fitting the spectrum using the SIMNRA software [74], the Fe and Zr concentration and the film thickness can be extracted. 3.3 Ion Implantation The ion implantation technique has been developed for introducing atomic dopant elements embedded in the near-surface region of a material. The implantation process is based on accelerating ions and injecting them into a target (the sample). The most widespread use of ion implantation can be found in the semiconductor field, but the technology enables all elements in the periodic table to be implanted into any crystalline or amorphous host material. [75,76] The ion implantation process starts by electrostatic extraction of ions from an ion source, which typically is a plasma containing the species in ionized form, illustrated in Figure 3.4. After pre-acceleration of ions from the plasma, these ions are analyzed in a magnet to select the correct ion mass, energy and charge state. Various magnetic and electrostatic lenses and steering de- 30

31 vices guide the ions in the form of a mono-energetic, isotope-pure beam, down along the beam line to the target. Before hitting the target the beam is typically raster scanned in horizontal and vertical directions to obtain a homogeneous areal coverage of the sample surface. In the sample, the ions gradually lose their kinetic energy in random elastic collisions with the host matrix atoms and inelastic interactions with the target electrons. Interactions with target atoms results in large angular deviations from the ion trajectory and also to displacements of target atoms from their positions, so called damage. If sufficient momentum is transferred in the elastic collisions, the displaced target atoms can also create new recoils and increase the damage. Eventually the ions comes to rest and most of the deposited kinetic energy of the ions has been transferred to the target phonon system (heat). The ion penetration depth and the width of the distribution depend on the ion species, the ion energy and the target material. The two stopping mechanisms mentioned above are termed nuclear collisions (elastic scattering) and electronic collisions (inelastic scattering), and these depend on the energy, or the velocity of the ion. The total energy loss during the ion trajectory is given by the sum of nuclear and electronic losses. Figure 3.4. A simplified figure of the basic constituents of a typical ion implanter. Positive ions are extracted from the ion source and mass analyzed on the high voltage platform and then accelerated through the acceleration tube towards the ground potential at the target. The beam is raster scanned to obtain a homogeneous coverage across the sample surface. The range distribution of implanted ions and the damage caused by the collisions can readily be estimated by simulations, using for instance the SRIM software [76]. This is a Monte Carlo program based on the binary collision approximation for amorphous targets. Figure 3.5a demonstrates the concentration profile of the implanted ions along the sample thickness. The main disadvantages with this technique is structural damage due to the displacement collisions. Figure 3.5b demonstrates the damage profile of the implanted ions along the sample thickness. This is particularly problematic 31

32 Figure 3.5. a) SRIM simulation [76] of ion implantation profiles in amorphous FeZr targets, where the film is implanted with 11 at.% of N, C, B, He and H. b) Vacancy (damage) profile within the amorphous FeZr target and the implanted ions distribution for a dose of ions/cm 2. The implantation dose (fluence) is controlled by the ion current and the implantation time. in crystalline materials and, for high enough doses, amorphisation may occur. Therefore, thermal annealing is often needed to return the damaged material to a crystalline phase. For the case of amorphous material, the arrangement of atoms is not ordered and therefore small changes of structural disorder upon implantation is less relevant. Other potential problems with ion implantation is beam heating, which may occur at a combination of high energy and high flux and/or fluence. In such cases, self-heating of the sample during the implantation may induce diffusion, or other unwanted effects. Sputtering of the surface can also be substantial for large fluence of heavier ions. In spite of these drawbacks, ion implantation is today widely used for introducing dopant elements and optimizing alloying composition with excellent control of depth distribution and concentration of the implanted species. In this thesis ion implantation has been used for doping amorphous FeZr films in Papers I-IV with the aim to investigate the effect of dopant elements on the structural, electronic, and magnetic properties of the amorphous FeZr system. Our results show this method to be specially useful because it allows precise local tuning of the magnetic properties. 32

33 4. Sample Structural Characterization 4.1 Thin film thickness from X-ray reflectivity (XRR) Crystalline solids are formed when atoms are stacked together in a fixed geometric pattern. There is a periodicity and long-range order in a crystalline solid, i.e. a unit cell exists that repeats itself and fills the space. It is possible to structurally characterize crystalline materials to describe the arrangement of atoms with X-Ray Diffraction (XRD), with suitably chosen X-ray sources. The wavelength of these x-rays are about the atomic size (λ 1 Å) that makes them suitable for structural characterization. For details on the structural characterization by XRD and analysis the reader is referred to the literature [77, 78]. The X-Ray Reflectivity (XRR) can be used for the film thickness measurement of crystalline and amorphous films. The XRR measurements are based on the reflected intensity of incident X-rays by a sample. During X-ray reflectivity measurement, the sample is rotated at a grazing angle θ (to increase pathway of the the X-rays through the sample) in a range of zero to a few degrees while the detector rotates 2θ. From each interface of the multilayer film, a fraction of the X-rays are reflected. The reflection at the surface and interfaces corresponds to electron density variations between the different film layers. Ultimately, the interference of the reflected X-rays from all interfaces creates an oscillating pattern, where the peaks correspond to the total film thickness are called Kiessig fringes. [79] The thickness of the film is then obtained from the distance between these fringes, where the thickness is inversely proportional to this distance. Moreover, from the amplitude of the fringes, the film density and roughness (originating from uneven surface or interface) can be calculated. To extract the above mentioned quantities, the data are compared to a model structure for which the reflectivity is obtained by solving the Fresnel equations. These equations describe the behavior of electromagnetic waves when traveling from one medium to another medium of a different density, hence with differing indices of refraction. We have used the XRR followed by simulations, for determination of the film thickness and interface roughness of multilayer structures. As an example, Figure 4.1 demonstrates the use of a structural model to obtain the thickness of each layer in the multilayer structure used in paper V. 33

34 The multilayers consist of bilayers repeated in a sequence. The spacing of the multilayers Λ, correspond to the respective layers thicknesses (Λ = L FeZr + L AlZr ), as for the structure shown in Figure 4.1. Due to the periodicity of multilayer structures it is possible to study their layer thickness by using XRR measurements. Figure 4.1 shows representative examples of data together with the fits for multilayer structure of [FeZr (d)/alzr (30 Å)] 10 investigated by XRR. The notation of the samples is for FeZr/AlZr nominal thicknesses. The samples were grown on native oxide Si substrates, using 100 Å and 40 Å AlZr as buffer and capping layers, respectively. The nominal values and the values obtained from XRR are presented in paper V. Figure 4.1. X-ray reflectivity from [FeZr (d)/alzr (30 Å)] 10 multilayer films illustrated on the right side of this figure, and a single 250 Å thick FeZr layer film. Experimental data are presented as black lines and the fits as cyan lines. The data were fitted using the GenX software [80]. Kiessig fringes come from the interference between film surface and the substrate. 34

35 5. Magnetic Characterization Two types of magnetometers have been used for the magnetic measurements: SQUID (Superconducting Quantum Interference Device) and MOKE (Magneto- Optic Kerr Effect). All investigations are performed on ferromagnetic thin films which require specific protocols and precautions to derive the characterizing parameters: Curie temperature, T c, saturation magnetization, M s (T ), remanence, M r (T ) and coercivity, H c (T ). Strictly, T c is only defined in zero magnetic fields for ferromagnets, it is thus necessary to use low field experiments to derive the Curie temperature. Suitable protocols and methods are: M(T ), M r (T ) or χ(t ), where M labels magnetization and χ susceptibility. In-plane measurements of M(H) at different temperatures provide measures of M s (T ), H c (T ) and M r (T ). Demagnetizing effects are negligible for magnetic field in-plane experiments. The following sections are divided into three categories. First, a brief description of the experimental setups, i.e., SQUID and MOKE magnetometers is given. Secondly, hysteresis loop (M(H)) experiments for determination of remanent magnetization (M r (T )), coercivity (H c (T )) and saturation magnetization (M s (T )) are discussed. Finally, for Curie temperature (T c ) determinations, results of magnetization as a function of temperature (M(T )), remanent magnetization (M r (T )) or magnetic susceptibility (χ(t )) measurements are presented. SQUID Magnetometry SQUID magnetometers provide a sensitive method to obtain magnetization data over a wide temperature range and with applied magnetic fields up to several Tesla. The magnetic moment from the sample couples inductively to a superconducting pick up coil. The magnetic flux in the pick up coil is transferred to the SQUID detector. The SQUID detector is an extremely sensitive flux-to-voltage converter, allowing very small magnetic moments ( 10 8 electromagnetic unit (emu)) to be measured. SQUID magnetometers are often used for obtaining the saturation magnetizations (M s ) of thin magnetic films, since they often provide both high external fields and give absolute magnetic moments. Details of SQUID magnetometers are given in Ref. [81]. MOKE Magnetometry The rotation of the polarization plane of light at the surface of a metal, the Magneto-optic Kerr effect, is employed in (MOKE) magnetometry. An 35

36 experimental MOKE setup is illustrated in Fig. 5.1, where linearly polarized laser light is reflected on a magnetic sample and then goes through the analyzer before arriving at the detector. A magnetized sample induces rotation of the initial polarization of the light. The rotation angle, is proportional to the magnetization of the film. [82] MOKE experiment can be performed in three Figure 5.1. A laboratory MOKE setup configuration is shown. The magnetic field is provided either by a pair of Helmholtz coils or conventional electromagnets. The angle of the analyzer is 90 to that of the polarizer. different geometries: transverse, polar and longitudinal MOKE, illustrated in Fig [82] To probe the magnetization of the samples, we have used the longitudinal geometry because the magnetization of the thin films is oriented in the film plane. In the longitudinal geometry, the induced rotation is only sensitive to the magnetization parallel to the plane of incidence. The polar Figure 5.2. Upon reflection from the surface of a magnetized medium, the rotation of the incident polarized light depends on the direction of magnetization. The figure illustrates the different MOKE geometries provided. and transverse configurations are used when the magnetization is perpendicular to the sample plane and the magnetization is perpendicular to the plane of 36

37 incidence and parallel to the surface, respectively. MOKE is a surface sensitive method (penetration depth about 100 nm). The method does not provide absolute values of the magnetization. Magnetic Measurements A ferromagnetic material shows a hysteretic M(H) curve, where magnetization traces out a loop as a function of applied field. MOKE or SQUID measurement can both be used to record M(H) loops, for determining characteristics of ferromagnetic materials such as coercivity (H c ), remanence (M r ), and saturation magnetization (M S ). Since the SQUID magnetometer probes the entire volume of a sample, the initial experimental value of the saturation magnetic moment from M(H) loop, provides the total magnetic moment from both magnetic film and the substrate. After approaching the saturation field, the contributions from the para- or diamagnetic substrate gives a positive or negative slope of the M(H) curve. As an example of this behavior, an M vs. Figure 5.3. M vs. H measured at 10 K on a Co/Si sample, where a) shows the raw data and b) the M-H loop after the diamagnetic contribution has been subtracted. H curve of a Co (8.14 Å) film grown on Si substrate is shown in Fig. 5.3 a. In this case, due to the diamagnetic Si substrate contribution, the M dia (H) is subtracted in order to obtain the magnetic moment of the ferromagnetic film, see Fig. 5.3b. Finally, to obtain the absolute magnetization (magnetic moment per volume), the magnetic moment of the sample is divided by its volume. The maximum field that can be applied by a specific magnetometer can limit the possibility of reaching the saturation magnetization of the sample, as shown in Fig. 5.4, for different Zr concentrations in Fe 100 x Zr x. The magnetization does not always saturate at the knee of the M vs. H, but they show a steady increase up to the highest applied field, which indicates magnetization in a non-collinear phase (for the 7.6 and 9.7 at.%, Zr samples). 37

38 Figure 5.4. M vs. H measured at 5 K on Fe 100 x Zr x samples with different Zr content (x). The magnetic saturation is not reached at the maximum applied field of (0.5 T) for the samples with 7.6 and 9.7 at.% Zr. The Curie temperatures of the investigated films have been determined using different methods: M(T ), M r (T ) and χ ac (T ). M (emu/cm 3 ) x=9.7at.% x=7.6at.% T (K) x=12.0at.% x=12.9at.% x=11.6at.% Figure 5.5. Field cooled (FC), magnetization vs. T measured on Fe 100 x Zr x films with different Zr content. The applied magnetic field is 10 Oe for the x = 7.6 at.% sample and 5 Oe for the other samples. In paper I, by using SQUID magnetometry, the T c of Fe 100 x Zr x films were obtained from the field-cooled (FC) magnetization versus temperature shown in Fig

39 Figure 5.6. Temperature dependence of the (normalized) remanent magnetization M r (T )/M r (80K) (from paper II). In the FC measurements the sample is cooled from 300 K to 10 K in a constant small magnetic field and then evolution of the sample magnetization is measured during warming. The Curie temperature is derived at the point, where the spontaneous magnetization disappears (i.e. where a rapid increase of M(T ) first appears.) Figure 5.7. In- and out-of-phase components (χ, χ ) of the ac-susceptibility versus temperature for the as-grown Fe 93 Zr 7 sample. The onset of the out-of-phase component (χ ) corresponds to T c = 122 ± 2 K. The amplitude of the ac field is 70 µt and the frequency 215 Hz (from paper II). In paper II, by using a MOKE setup, the Curie temperature of FeZr films were determined by measuring the remanent magnetization (M r (T )). The re- 39

40 manence M r (T ) values were extracted from M(H) loops measured at different temperatures. The value of the Curie temperature is obtained at the point where M r (T ) reaches zero. Figure 5.6 shows the temperature dependence of the normalized remanence M r (T )/M r (80K) for the as-grown FeZr and some of the implanted films. Along with the two methods presented above to determine the Curie temperature, a third method is by alternating current (ac)-susceptibility measurements. The ac-susceptibility has been measured on an as-grown FeZr sample. The temperature dependence of in-phase ( χ ) and out-of-phase ( χ ) components of the ac-susceptibility were measured over the temperature range of K. Fig. 5.7 shows low field ac-susceptibility vs. temperature curves for this sample, where the applied ac field is 70 µt and the frequency is 215 Hz. The Curie temperature is estimated from the onset of the out-of-phase component ( χ ) which corresponds to T c = 122 K. 40

41 6. Summary of Results 6.1 Paper I, II, III, IV and V Amorphous FeZr (Fe 100 x Zr x ) films were synthesized by dc-magnetron sputtering. A single amorphous layer was used for the films investigated in papers I, II, III, IV and a multilayer structure was used in paper V. All FeZr films are sandwiched between a buffer and a cap layer of AlZr. Amorphous Fe 100 x Zr x exhibits Curie temperatures (T c ) that changes with the Zr concentration but are below room temperature (see Figure 6.1, paper I and references therein), therefore they can not be adopted for room-temperature applications. In order to use these materials for room temperature applications, increasing the Curie temperature is crucial. We report on two different approaches to manipulate the magnetic behavior of these amorphous Fe 100 x Zr x alloys. For this purpose, we first study the dependency of the Curie temperature upon variation in chemical composition through dopant incorporation (paper I, II, III, IV), where the process of introducing dopant atoms into these films was accomplished by ion implantation technique. In the next step we look into the Figure 6.1. The Curie temperature as a function of Zr concentration for amorphous Fe 100 x Zr x films. 41

42 dependency of the critical temperature upon FeZr thickness variations (paper V). Composition and thickness variations, such as those investigated here, show that a significant increase in Curie temperature (T c ) is obtained in both cases, resulting in Curie temperatures above room temperature. For paper I, the Fe 100 x Zr x samples were deposited by co-sputtering of Fe and Zr targets, to make films with the compositions of x = at.%. The Curie temperature changes with varying Zr contents and has a maximum at about 12.0 at.% Zr, as illustrated in Figure 6.1. Afterwards, the films with compositions of x = 11.6 and 12.0 at.%, were doped with H, using ion implantation. Figure 6.2. a) M vs. H measured at 5 K and b) FC magnetization vs. T recorded in a field of 5 Oe for Fe 88.4 Zr 11.6 samples with different hydrogen content. These figures are adopted from paper I. Figure 6.2 a, shows the hysteresis curves measured at 5 K for the as-grown and hydrogenated Fe 88.4 Zr 11.6 samples. The coercivity reduces drastically upon hydrogenation down to the order of 1 Oe for 16.0 at.% H. Figure 6.2 b, shows the magnetization as a function of temperature for the same samples, where T C increases rapidly with H doping. For paper II, a compound target of Fe 93 Zr 7 was used for depositing the alloys, which were then doped with H, He, B, C and N at the same average dopant concentration of about 11 at.%. For these samples, T c is found to increase with increasing atomic radius (r a ) of the doping element. Furthermore, we compare these results with another set of samples, which are implanted with different concentrations of H and He, where the Curie temperature shows an increase with concentration. We combine both the concentration (c) and the size of the dopant and define a parameter V c = cra. 3 Also, to describe the impact on the Fe-Fe distance by doping, we define a length scale r c = 3 cr a (or r c = 3 V c ). r c can be considered an indirect measure of the Fe-Fe interatomic distances. Figure 6.3 shows the increase of T c for the implanted samples, compared to the as-grown FeZr film, called T c, as a function of r c. The data shows a close to linear relation between the Curie temperature and r c. The magnetic properties of doped FeZr can be qualitatively explained by the Bethe-Slater 42

43 Figure 6.3. The increase in Curie temperature T c for the implanted samples (relative to the as-grown FeZr film) versus r c = 3 V c, where V c = cr 3 a (paper II). curve [40], described earlier in section 2.1. The exchange interaction in the FeZr system is close to the zero-crossing point of the curve. The direct exchange depends on the Fe-Fe interatomic distance and T c is proportional to the exchange energy. The presence of dopants, drives the exchange interaction towards stronger FM coupling as Fe-Fe distance increases, hence greatly enhancing T c. In Paper III, the Fe 93 Zr 7 films were doped with C at 5.5 and 11 at.%. The enhanced Curie temperature and decrease of the coercivity are very similar to the effects of H doping (as in Paper I and II). The Curie temperature reaches well above room temperature and the films become magnetically soft. The chemical bonding of Fe for as-grown and implanted films were characterized by X-ray photoelectron spectroscopy (XPS) that reveals the electronic effects (electron transfer from Fe to C). This is due to the presence of Fe-C bonds that are formed upon the implantation. Paper IV reports an additional electronic effect in C and B doped Fe 93 Zr 7 films, where the coercivity enhances with increasing temperature, as shown in Fig This is interpreted as the result of charge-transfer effects, where due to the large electron affinity of C and B, electron transfer from Fe to C and B occurs. As a result of this formation of interatomic covalent bonds, the local anisotropies increase, as evidenced by an increase of coercivity with temperature. XPS measurements were carried out, which show that the binding energy peaks shift to higher energies. 43

44 Figure 6.4. The coercivity as a function of temperature for the as-grown FeZr films and implanted by H, He, B, C and N. Inset shows a zoom of the curve for B-implanted sample, showing an anomalous temperature dependence of the coercivity. This figure is a adopted from paper IV. Paper V investigates the dependence of the critical temperature on the thickness of the FeZr layer, and reveals an anomalous increase of T C when the thickness of the magnetic layer (FeZr) is decreased from 60 Å to 20 Å, as shown in Fig Amorphous Fe 92 Zr 8 /AlZr multilayer structures were deposited by co-sputtering of Fe and Zr targets. The critical temperature enhancement is linked to a reversed interface effect, wherein changes in the Fe-Fe distance and coordination number favors a higher effective magnetic coupling at the interfaces compared to the interior of the bulk layer. This indicates the inevitable intermixing of the layers, i.e., interfacial mixing of AlZr and FeZr layers, that occurs during the sample deposition, resulting in Al entering the FeZr layer. The results are explained with a model where such interface effects are combined with finite size scaling. In order to explain why in this model T c increases as the thickness decrease, we assume that the FeZr layer consists of two regions: the interior FeZr layer and the interface region between the FeZr and AlZr. XRR results provide the thickness of the interface region (intermixed region). The average exchange interactions in the interior layer (J 1 ) is smaller than that at the interface (J i ) layer. As the result, the average effective exchange interactions (J ef f ) increases as the FeZr layer thickness decreases. Along with a decreasing layer thickness, the enhancement of J ef f results in a rise in T c. At the interface, Fe 92 Zr 8 becomes doped with small amounts of Al and Zr. By comparing to Fig. 2.3 a and the results in paper I-IV, this implies that for small Al/Zr doping levels, T c increases. However, when the FeZr layer becomes thin, T c decreases again due to the finite size effects. 44

45 T C (K) bulk FeZr thickness (Å) Figure 6.5. The Curie temperature (solid squares) and magnetic moment (open circles) as a function of the FeZr layer thickness. The bulk (250 Å) values of both properties are represented by the dotted horizontal line. The solid line and the dash-dotted line represent fits. This figure is a adopted from paper V Mag. mom. (μ B /Fe) 6.2 Paper VI In paper VI, a different approach to control the magnetization is investigated, where the structural transitions of a ferroelectric substrate govern the magnetic properties of a deposited ferromagnetic thin film. We have studied ferromagnetic amorphous CoZr grown on a ferroelectric BaTiO 3 substrate, to understand strain induced coupling between these materials by making use of temperature dependent structural transitions of the substrate. During the deposition process of the structure BaTiO 3 (001)/ Al 70 Zr 30 (3 nm)/co 95 Zr 5 (15 nm) / Al 70 Zr 30 (3 nm), an external magnetic field of 0.01 T was applied parallel to the [010] BaTiO 3 (001) substrate direction. This results in an in-plane magnetic anisotropy in the amorphous Co 95 Zr 5 layer. To assist the growth of an amorphous Co 95 Zr 5 (15 nm) layer, an amorphous Al 70 Zr 30 (3 nm) seed layer was used. Another functionality of the seed layer is the suppression of direct interface effects between the BaTiO 3 substrate and the Co 95 Zr 5 film on top. The results in paper VI, show changes in the magnetic anisotropy of the amorphous Co 95 Zr 5 layer at the structural phase transitions of the BaTiO 3 substrate. Once the orthorhombic phase transition of the BaTiO 3 substrate occurs (220 K), this will induce an enhancement of the uniaxial magnetic anisotropy in the amorphous Co 95 Zr 5 film. Therefore, the magnetic moment orients towards the easy uniaxial axis. The strain induced in the Co 95 Zr 5 thin film, when entering any of BaTiO 3 structural phases, is derived from the change of the lattice parameters of BaTiO 3. Figure 6.6 (a) indicates that in the orthorhombic BaTiO 3 phase, the strain induced hard axis is along the [100] direction while 45

46 a!!!!"!!"#$%%&" H//[010] M (emu.cm -3 ) Rhombohedral Orthorhombic Tetragonal M (emu.cm -3! ) b! H//[100]!"!!"#%$%&" T (K) Figure 6.6. Changes of the magnetization with temperature at an applied magnetic field of 5 Oe. The measurements were performed after cooling the sample in zero field. The field was applied in the film plane parallel (a), and perpendicular (b), to the [100] BaTiO 3 (001) substrate direction. Sudden jumps emerge near 190 and 278 K that are related to the structural phase transitions of the BaTiO 3 substrate. the tetragonal and rhombohedral phases do not show this feature. Figure 6.6 (b) indicates that the strain from the BaTiO 3 substrate will influence the magnetic response of the Co 95 Zr 5 by aligning magnetic domains in the [010] direction when entering a new phase. Since there is an AlZr buffer layer, our results support the view that the anisotropy is governed by bulk-strain rather than interfacial lattice strain. Heterostructures of ferromagnetic and ferroelectric materials offer interfacemediated magnetoelectric coupling. The ability to externally control the properties of magnetic films via modulating the surface/interface magnetism, would potentially provide a composite system where the magnetization can be controlled by applying an external electric field. This can be interesting for the spintronic data storage devices, since controlling the magnetization by electric field is a promising strategy to reduce the energy consumption. 46

47 6.3 Paper VII The aim of paper VII is to determine the exchange stiffness (A ex ) in pure and alloyed Co layers by fitting the M(H) dependence of a FM/NM/FM film structure to the one dimensional micromagnetic model described in section 2.5. For this study, the synthetic antiferromagnetic structure (SAF) of Co ( nm)/ru (0.5 nm)/co ( nm) trilayers was selected as shown in Fig. 6.7 a. A Ru layer is sandwiched between two Co layers, which are antiferro- Figure 6.7. a) The cross section of the multilayer system prepared for the measurements of A ex in Co and CoX alloys. The layer structure consists of: a seed layer of Cu followed by Ru, two ferromagnetic CoX layers (antiferromagnetically coupled by a Ru spacer layer) and a Ru capping layer. b) Experimental data (SQUID measurement) from measured M(H) curves of Co (10 nm) / Ru / Co (10nm). Fits to the experimental data points were carried out using the micromagnetic model outlined in text. The filled red/grey circles are the measured values while the line is the fitted curve. Both figures are adopted from [69]. magnetically coupled. The films are textured with the [0001] oriented perpendicular to the surface. Consequently, the uniaxial anisotropy has no effect on magnetization reversal in these films. Moreover, by using the micromagnetic model, A ex of the alloyed CoX thin films in CoX (10 nm) / Ru (0.38 nm) / CoX (10 nm) were determined, where X refers to the alloying elements of Cr, Fe, Ni, Pd, Pt and Ru. Figure 6.7 b shows the calculated fit and the measured M(H) curve of Co (10 nm) / Ru / Co (10 nm). The data is normalized to the saturation magnetization, M S. The fit was performed by using Eq. 2.8 and Eq. 2.9, using J RKKY and A ex as fitting parameters. The results show that the thickness dependence of A ex for the Co layers which are thicker than 80 Å is very weak. In the case of thinner films, the exchange stiffness drops. This accordingly makes the interface contribution more prominent. The M S values of these Co films do not show the same sharp 47