Origin of Ferromagnetism in Oxide-Based Diluted Magnetic Semiconductors

Transcription

1 Origin of Ferromagnetism in Oxide-Based Diluted Magnetic Semiconductors by Numan Akdoğan A dissertation submitted to The Faculty of Physics and Astronomy for the degree of Doctor of Philosophy in Physics Ruhr-Universität Bochum Bochum, Germany February 2008

2 Mit Genehmigung des Dekanats vom wurde die Dissertation in englischer Sprache verfaßt. Mit Genehmigung des Dekanats vom wurden Teile dieser Arbeit vorab veröffentlicht. Eine Zusammenstellung befindet sich am Ende der Dissertation. Dissertation eingereicht am Erstgutachter: Prof. Dr. Dr. h.c. H. Zabel Zweitgutachter: Prof. Dr. K. Westerholt Disputation:

3 Aileme ve bugünlere gelmemde emeği olan herkese.

4

5 Contents I Introduction 1 1 Introduction 3 2 Potential Spintronic Devices Based on DMS Materials Spin-FET Spin-LED Transparent ferromagnet Selective Kerr rotator Theoretical Models for Explaining Ferromagnetism in DMS The Zener model The mean-field model The RKKY model The mean-field Zener model The double exchange model The bound magnetic polaron model Transition metal clustering model II Methods 21 4 Studying DMS Materials with Synchrotron Radiation Synchrotron radiation Production of synchrotron radiation Quantum mechanical description of the interaction between light and matter Absorption Resonant scattering The polarization dependence of the resonant scattering The relation between scattering and absorption phenomena X-ray Magnetic Circular Dichroism The origin of the XMCD effect Detection methods Transmission Total electron yield Fluorescence yield i

6 Contents 6 X-ray resonant magnetic scattering The Alice diffractometer for XRMS and XMCD experiments III Results and Discussion 43 7 Dose Dependence of Magnetism in Co-implanted TiO 2 Rutile Overview Sample Preparation Structural Properties Magnetic Properties Room temperature measurements Low temperature measurements Post-annealing experiments Anomalous Hall effect measurements Discussion Summary and conclusions Room Temperature Ferromagnetism in Co-implanted n-type ZnO Thin Films Overview Sample Preparation Structural Properties Magnetic Properties Room temperature magnetization measurements XRMS and XAS measurements Temperature dependent magnetization measurements Anomalous Hall effect measurements Discussion Summary and conclusions Conclusions 95 Bibliography 97 Publications 105 Acknowledgments 107 Curriculum vitae 109 ii

7 Part I Introduction 1

8

9 1 Introduction Nowadays, the most successful technologies are led by the integrated circuits (ICs) and the data storage industries. The integrated circuits utilize the charge of electrons and holes in order to perform their specific functionalities and ignore the spin of electrons. High speed signal processing and good reliability are the main advantages of ICs. However, their memory elements are volatile which means that when the power is switched-off the stored information is lost. In the case of magnetic data storage technology, the spin of electron is the key parameter. On the other hand, magnetic data storage devices are non-volatile because of the natural remanence characteristic of ferromagnetic materials. Spintronics, short notation for spin-based electronics, is a new research area which seeks to exploit the spin of electrons in addition to their charge in semiconductors. The basic idea is to combine the characteristics of existing magnetic devices with semiconductor devices in order to realize a new generation of devices that are much smaller, more energy efficient, non-volatile, and much faster than presently available. [1, 2, 3, 4, 5, 6, 7, 8, 9]. The key requirement in the development of such devices is an efficient injection, transfer and detection of spin-polarized current from a ferromagnetic material into a semiconductor. Due to the well known problem of a resistance mismatch at metal/semiconductor interfaces, hindering an effective spin injection [10], much interest is now concentrated on the development of room-temperature ferromagnetic semiconductors. Diluted magnetic semiconductors (DMS) refer to the fact that some fraction of atoms in a non-magnetic semiconductor is replaced by magnetic ions (Fig. 1.1). DMS are promising candidates for spintronic applications at practical temperatures, provided that their Curie temperature (T c ) is above room temperature. Therefore, a number of different semiconductor hosts have been investigated to test their magnetic properties. In the past most attention has been paid to (Ga, Mn)As [11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34] and (In, Mn)As [35, 36, 37, 38, 39, 40, 41, 42] systems. However, due to their reported highest Curie temperatures which are around 170 K for (Ga, Mn)As [43, 44, 45] and 35 K for (In, Mn)As [45, 46], they are disqualified for most practical applications. Therefore, there is a large incentive for developing new DMS materials which are ferromagnetic above room temperature. In particular, the calculations of Dietl et al. [47] showed that Mn-doped ZnO would exhibit ferromagnetism above room temperature. Sato et al. have also investigated ferromagnetism of ZnO-based DMS by ab initio electronic structure calculations based on the local spin density approximation and they reported ferromagnetic ordering of 3d transition metal (TM) ions in ZnO [48, 49]. These theoretical predictions initiated a number of experimental studies of TM-doped ZnO [50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61]. In addition, the discovery of room temperature ferromagnetism in Co-doped anatase TiO 2 [62] has also generated much interest in the 3

10 1 Introduction Figure 1.1: Schematic representation of a non-magnetic semiconductor (left), and a diluted magnetic semiconductor (right). In the style of Ref. [43]. Co-doped TiO 2 system as a potential oxide-based DMS [63, 64, 65, 66, 67, 68, 69, 70, 71]. While some of these works reported the observation of ferromagnetism above room temperature, the origin of ferromagnetism in this systems is not well understood yet. The main unresolved question is whether the observed ferromagnetism originates from uniformly distributed TM ions in the host matrix or whether it is due to the precipitation of secondary phases such as metallic clusters. If a DMS contains TM ions below their equilibrium solubility limit no secondary phases are expected. In this case, since the strength of magnetism is proportional to the number of TM ions substituted on the cation cites in a DMS, the realization of high T c ferromagnetism is difficult. On the other hand, at higher TM dopant concentrations doped transition metals start to form unwanted metallic clusters. For this reason, to achieve a high T c ferromagnetism in a single phase DMS, non-equilibrium sample preparation techniques such as low temperature MBE and ion implantation are required. Ion implantation is widely used in silicon technology for integrated circuits due to its reliability, precision and reproducibility [72]. It has also proven to be a reliable method for injecting transition metals into a host semiconductor material beyond their solubility limits [73]. In this thesis ion implantation technique is used to implant cobalt ions into TiO 2 and ZnO. There are two main purposes of this thesis. One is to produce TiO 2 and ZnObased ferromagnetic semiconductors which have T c values higher than room temperature. The second aim of the thesis is to shed some light on the origin of room temperature ferromagnetism whether it originates from clusters or from uniformly distributed magnetic cobalt atoms. In an attempt to fabricate ferromagnetic semiconductors, TiO 2 rutile plates and ZnO films are exposed to magnetic cobalt ions for varying implantation doses. Rutherford backscattering spectrometry (RBS), x-ray diffraction (XRD), atomic force microscopy (AFM), and transmission electron microscopy (TEM) were used to obtain Co distribution profiles and to identify possible second phases in these systems. The magnetic properties of the implanted materials have been investigated using the magneto-optical Kerr effect (MOKE), superconducting quantum interference device (SQUID) magnetometry, and x-ray resonant magnetic scattering (XRMS) as well as x-ray magnetic circular dichroism (XMCD). Hall effect measurements were also performed to determine the type of carriers in this materials. 4

11 The thesis is organized in three parts. The first part gives an introduction to the subject which is followed by some applications of spintronic devices. The theoretical models that are most commonly used to describe the magnetic interactions in diluted magnetic semiconductors are also explained in this part (Chap. 3). The second part deals with the theoretical and experimental background to the magneto-optical effects in soft x- ray energy range. After an introduction to the properties of synchrotron radiation, the XMCD and XRMS techniques are discussed in Chap. 5 and Chap. 6, respectively. The experimental setup which is used for soft x-ray measurements is also described at the end of this part in Sec The experimental results of Co-implanted n-type TiO 2 and n-type ZnO samples are reported in the third part of this thesis. First, the dose dependence of magnetism in Co-implanted n-type TiO 2 rutile plates is presented in Chap. 7. In Chap. 8 the structural and magnetic properties of Co-implanted n-type ZnO films are reported for different implantation doses. The third part ends with conclusions. 5

12 1 Introduction 6

13 2 Potential Spintronic Devices Based on DMS Materials There are two different approaches for realizing spintronic devices. One is metal-based spintronics which uses ferromagnetic metals and the second one is semiconductor-based spintronics consisting of ferromagnetic semiconductors. Metal-based spintronics has been the basis of information storage devices such as read heads for magnetic hard disk drives, since the discovery of the giant magneto-resistance (GMR) effect. The GMR effect 1 was reported by P. Grünberg [74] and A. Fert [75] in 1988 for layered magnetic thin film structures that consist of alternating layers of ferromagnetic and nonmagnetic layers. When the magnetization of ferromagnetic layers are parallel, the material show very low resistance. When the magnetization of ferromagnetic layers are antiparallel, the spin-dependent scattering of carriers becomes maximum and the material exhibits higher resistance. In the metal-based spintronics, researchers and developers now seek to improve the existing GMR devices by developing new materials with enhanced spin-polarization and better spin filtering. Similar rapid development is expected from tunneling magnetoresistance (TMR) devices that are composed of two ferromagnetic layers separated by an insulating metal-oxide layer with a thickness of a few nanometers. In these devices, electrons can easily tunnel through the insulating barrier depends on the relative magnetization of the two ferromagnetic layers, and on the fact that the spin of electrons are preserved as they pass through the barrier. The TMR effect leads to a more pronounced resistance change in small applied fields than is observed in GMR devices. In 1995 Moodera et al. [76] demonstrated TMR at room temperatures with very thin oxide layers. Less than a decade later, Motorola, IBM and Infineon are manufacturing fast magnetic storage devices that incorporates dense arrays of TMR elements, known as magnetic random access memory (MRAM). In the case of the semiconductor-based spintronics, a lot of efforts focus on producing ferromagnetism in semiconductors above room temperature. If successful, this new class of spintronic devices could much easer be integrated with conventional semiconductor technology. In this chapter, some potential semiconductor-based spintronic devices such as spin field-effect transistor (Spin-FET), spin light-emitting diode (Spin-LED), transparent ferromagnet and selective Kerr rotator are briefly summarized. 1 In 2007, P. Grünberg and A. Fert were awarded by the Nobel Prize in Physics for the discovery of this effect. 7

14 2 Potential Spintronic Devices Based on DMS Materials 2.1 Spin-FET In the spin field-effect transistor proposed by Datta and Das [2], carriers are injected with a definite spin orientation into the channel from a spin-polarizing electrode (source), which can be either a ferromagnetic metal or a diluted magnetic semiconductor, and collected by another magnetically polarized electrode (drain) (see Fig. 2.1). The source-drain current depends on relative orientation of the carrier spins and the magnetization directions of the source and drain. If the magnetization directions of the source and drain are parallel, the current becomes maximum. If they are anti-parallel, the transistor will be shut off. By applying a gate voltage, the spin alignment of the carriers in the channel can be controlled, hence, the current can be modulated. If the applied gate voltage is sufficient to align the spin orientation of the carriers anti-parallel to the drain magnetization, then the current in the spin-fet is shut off and it is expected to occur at lower voltages than a conventional charge-controlled transistor. This leads to potential applications of spin-fets in very low-power microprocessors. Figure 2.1: Datta-Das spin-fet. Spin-polarized electrons emitted from a source pass through a narrow channel and are collected by a drain, if the source and drain have parallel alignment of their magnetization. When the gate voltage is on, the electric field causes the precession of spins. With sufficient gate voltage, the electron spins are aligned anti-parallel with the magnetization direction of drain and no current can pass. Thus, the source-drain current can be modulated by the gate voltage in this way. After Ref. [2]. Figure 2.2: A schematic representation of a spin-fet based on TM-doped ZnO. Source and drain are made from Co-doped ZnO. When the gate voltage is on, Mndoped ZnO is injected by holes and becomes a half metallic ferromagnet which allows a spin-polarized current to flow from source to drain. After Ref. [77]. 8

15 2.2 Spin-LED Sato and Katayama-Yoshida [77] also proposed a ZnO-based spin-fet which is represented in Fig Their calculations showed that Mn-doped ZnO can be ferromagnetic by hole doping. In their proposed spin-fet, by applying a gate voltage Mn-doped ZnO is injected by holes and it becomes half metallic ferromagnet. Co-doped ZnO is used as ferromagnetic source and drain. When the gate voltage is on, the spin-polarized electrons can travel from source to drain. When the gate voltage is off, there is no source-drain current. 2.2 Spin-LED In spin light emitting diodes, spin-polarized holes from a DMS and unpolarized electrons from a nonmagnetic semiconductor are injected from either side and recombine in a quantum well. The polarization of the injected holes can be measured by comparing the intensity of the right- and left-circularly polarized light of the emitted electroluminescence (EL). Efficient spin injection has been successfully demonstrated in spin-leds by using a spin-polarized DMS as the injector [6, 78]. In Fig. 2.3 the sample structure of a GaAs-based spin-led proposed by Y. Ohno et al. [6] is represented. Figure 2.3: Injection of spin-polarized holes into a light emitting p n diode using a DMS (Ga, Mn)As. Spin-polarized holes h + travel through the nonmagnetic GaAs and recombine with spin-unpolarized electrons in the (In, Ga)As quantum well. I and σ + represent the current and circularly polarized light emitted from the edge of the quantum well, respectively. Taken from Ref. [56]. The structure of GaAs-based spin-led consists of a p-type ferromagnetic semiconductor (Ga, Mn)As and n-type non-magnetic semiconductor GaAs, which were epitaxially grown by MBE. Since a spontaneous magnetization occurs below T c in the p-type (Ga, Mn)As DMS, this gives possibility to inject spins without requiring a magnetic field. When a forward bias is applied, spin polarized holes from the (Ga, Mn)As are injected through a spacer layer with thickness d into the nonmagnetic region and recombine with 9

16 2 Potential Spintronic Devices Based on DMS Materials spin unpolarized electrons injected from the n-type GaAs in a nonmagnetic (In, Ga)As quantum well (hatched region), producing polarized electroluminescence (EL). Injected spin-polarized electrons can be detected in the form of circularly polarized light emission from the quantum well, which corresponds to the magneto-optical Kerr effect loops. The presence of spin-polarization has been confirmed by observation of hysteresis in the polarization of the emitted light as a function of the magnetic field. Since the T c of (Ga, Mn)As is much below room temperature, the main issue now is to find a proper DMS material that can inject a spin-polarized current efficiently into a semiconductor at room temperature. Realization of such spin-leds would eliminate the need for the polarizing filters that are currently inserted into conventional devices. 2.3 Transparent ferromagnet To realize optically controlled ferromagnetism, Sato and Katayama-Yoshida [77] proposed a photomagnet structure based on TM-doped ZnO grown on GaAs as shown in Fig The concept is based on the theory that ZnO:MnCr and ZnO:FeMn are half-metallic ferromagnets upon hole and electron doping, respectively. For photons of appropriate energy (E g (GaAs) < hν < E g (ZnO)), electrons and holes can be generated in the GaAs substrate near the interface with the ZnO:MnCr or ZnO:FeMn and then injected into ZnO:MnCr or ZnO:FeMn by using positive or negative bias voltage, causing them to become half-metallic ferromagnets. The presence of these ordered states can be detected using the magneto-optical effect via other photons with energy lower than the ZnO bandgap (hν < E g (ZnO)). The potential use of this device has been described in detail in Ref. [77]. Figure 2.4: Schematic of ZnO-based transparent photomagnet. After Ref. [77]. 2.4 Selective Kerr rotator Another potential DMS-based optical device is the selective Kerr rotator [79] which is schematically illustrated in Fig In this structure, the Kerr rotation is controlled by 10

17 2.4 Selective Kerr rotator an electric field applied between two p-n junctions where p-material is DMS (Ga, Mn)N and n-material is GaN. At zero bias voltage the surface p-region is ferromagnetic causing an incident light to be reflected with large Kerr rotation angle. When the bias voltage is increased, the surface depletion of holes would remove the ferromagnetism and thus, no Kerr rotation is observed. Figure 2.5: Working principle of a GaN-based selective Kerr rotator. Taken from Ref. [79]. 11

18 2 Potential Spintronic Devices Based on DMS Materials 12

19 3 Theoretical Models for Explaining Ferromagnetism in DMS Although a number of different theories have been proposed, the mechanisms responsible for the origin of ferromagnetism in diluted magnetic semiconductor is not fully understood and explained yet. In this chapter, the theoretical models that have been most commonly used to explain ferromagnetism in diluted magnetic semiconductors are summarized. 3.1 The Zener model The theory proposed by Zener [80, 81] indicates that the direct superexchange interaction [82] between half filled d-shell electrons of TM cations and completely filled p-orbitals of anion is antiferromagnetic (Fig. 3.1). Since the d-shell electrons from both adjacent TM atoms occupy the same p-level, their spins must be opposite according to Pauli exclusion principle. This leads to an antiferromagnetic coupling of nearest-neighbor TM cations through a shared anion. Figure 3.1: Direct superexchange interaction: Antiferromagnetic coupling of adjacent TM cations through a shared anion. On the other hand, the indirect superexchange interaction between localized d-shell electrons of TM cations mediated by the delocalized band carriers tends to align the spin of the partially filled d-shells in a ferromagnetic manner (Fig. 3.2). In the Zener model, ferromagnetism is only possible when the indirect exchange interaction dominates over the direct superexchange interaction. 13

20 3 Theoretical Models for Explaining Ferromagnetism in DMS Figure 3.2: Indirect superexchange interaction: Ferromagnetic coupling of localized spins through the conduction electrons. 3.2 The mean-field model In the mean-field model, the interaction between the localized Mn atoms through free holes in the material leads to a ferromagnetic alignment of Mn atoms (Fig. 3.3). Due to the possibility of direct superexchange Mn-Mn interactions, there is always a competition between the ferromagnetic and antiferromagnetic interactions. Figure 3.3: The mean-field model: Ferromagnetic coupling of localized Mn spins through the free holes. 3.3 The RKKY model The RKKY (Ruderman-Kittel-Kasuya-Yosida) model [83] describes the magnetic interaction between a single localized magnetic ion and the delocalized conduction band electrons. Due to this interaction the conduction electron close to the magnetic ion is magnetized and act as an effective field to influence the polarization of nearby magnetic ions with the polarization decaying in an oscillatory manner. This oscillation mediates either ferromagnetic or antiferromagnetic exchange coupling depending upon the separation between magnetic ions. This model is efficient when a high concentration of delocalized carriers is present in the host material. 14

21 3.4 The mean-field Zener model 3.4 The mean-field Zener model The mean-field Zener model proposed by Dietl et al. [47] is based on Zener-like indirect superexchange interaction mediated by free holes. In this theory, the delocalized hole carriers mediate a RKKY-like interaction between the localized TM ions resulting in ferromagnetism. As compared to the RKKY interaction, the mean-field Zener model takes into account the complex valence-band structure of zinc-blende ferromagnetic semiconductors. Using this model, Dietl et al. have successfully explained the experimentally observed T c values for (Ga,Mn)As and (Zn,Mn)Te. Most importantly, it was predicted by this theory that Mn-doped p-type GaN and ZnO are ferromagnetic with T c values higher than room temperature as shown in Fig Even though this model indicates hole exchange to be dominant, the theoretical prediction of Sato et al. [49] shows that the high Curie temperature ferromagnetism could also be stabilized in n-type ZnO. Figure 3.4: Computed Curie temperatures for various p-type semiconductors containing 5 % of Mn and hole cm 3 (From Ref. [47] ). 3.5 The double exchange model Sato and Katayama-Yoshida [48, 49] performed first principles ab initio electronic structure calculations and have suggested that the n-type doping in ZnO can increase the Curie temperature of Fe-, Co- and Ni-doped samples. In this model, the ferromagnetic state in TM-doped ZnO is stabilized based on a double exchange interaction, which was first proposed by Zener [84] to explain the experimentally observed ferromagnetism in manganite materials. 15

22 3 Theoretical Models for Explaining Ferromagnetism in DMS The double exchange interaction is explained by hopping of electrons between two neighbor TM ions. In the DMS materials, the 3d levels of TM ions are split by the crystal field into lower energy doublet e g and higher energy triplet t 2g levels. The spin-up ( ) and spin-down ( ) states are also split by the exchange splitting. In Co-doped ZnO, the spin-up Co 3d-states are lower in energy than the spin-down states and are strongly hybridized with the oxygen 2p states [85, 86, 87]. On the other hand, the spin-down states of Co ions are located close to the conduction band of host ZnO. If neighboring Co ions have parallel alignment of their magnetic moments, the electrons in the partially filled 3d-orbitals of the Co ions are allowed to hop from one ion to the other and stabilize the ferromagnetic ground state (Fig. 3.5). Figure 3.5: Ferromagnetic coupling of nearby Co 2+ (3d 7 ) ions via double exchange interaction. 3.6 The bound magnetic polaron model The ferromagnetic ordering of TM metal ions in DMS materials is also described by the bound magnetic polaron (BMP) model [88, 89, 90, 91]. In this model, the formation of bound magnetic polaron is explained by the exchange interaction between many localized TM ion spins via much lower number of weakly localized charge carriers. The overlap of neighboring magnetic polarons leads to a long-range ferromagnetic state and the ferromagnetic transition occurs when the polaron size is equal to the sample size. Recently, Coey et al. [92, 93] applied this model to diluted magnetic oxides. They proposed that the ferromagnetic exchange coupling of TM ions in n-type diluted magnetic oxides is mediated by the shallow donor electrons trapped by the oxygen vacancy that tend to form bound magnetic polarons within their hydrogenic orbits (Fig. 3.6). 16

23 3.6 The bound magnetic polaron model Figure 3.6: Representation of bound magnetic polarons. Cation sites are shown by small circles. Oxygen is not shown, the unoccupied oxygen sites are represented by squares (From Ref. [92]). The general formula for the diluted magnetic oxides is [92] (A 1 x M x )(O δ ) n (3.1) where A is a nonmagnetic cation, M is a magnetic TM cation. O is the oxygen, represents a donor defect and n=1 or 2. An electron associated with a donor defect will be confined in a hydrogenic orbital [94, 92]. The exchange interaction E ex between the hydrogenic electron and TM cation can be written in terms of the s-d exchange parameter J sd as [92] E ex = J sd S s Ψ(r) 2 Ω (3.2) 17

24 3 Theoretical Models for Explaining Ferromagnetism in DMS where S is the spin of the TM cation with a volume Ω = (4/3)πr 3 c, r c is the TM cation radius, and s is the spin of donor electron. The wave function of the donor electron in a 1s orbital is given by Ψ(r) = (πr 3 H ) 1/2 exp( r/r H ), where r H and r are the hydrogenic orbital radius of the donor electron and the distance from the nucleus, respectively. The only way for boosting the E ex, and hence T c, is to increase the donor electron density Ψ(r) 2 in the vicinity of the TM cation [92]. As the donor defect concentration δ increases, the hydrogenic orbitals associated with the randomly positioned defects overlap to form a spin-split donor impurity band [92]. In Fig. 3.7 the electronic band structure of an oxide with 3d TM ions and a spin-split donor impurity band is presented. On passing 3d TM series from Ti to Cu, there are two regions where a high T c is expected. One is near the beginning of the 3d TM series where 3d spin-up states (3d ) cross the Fermi level in the impurity band (Fig. 3.7(c)), and the second one towards the end where 3d spin-down states (3d ) cross the Fermi level (Fig. 3.7(b)) [92]. Figure 3.7: Schematic band structure of an oxide with different 3d TM impurities. (a) The position of the 3d level for low T c, when the splitting of the impurity band is small. (b) and (c) show positions of the spin-down (b) and spin-up (c) 3d levels, respectively, lead to high T c (From Ref. [92]) After all this discussion of the BMP model, the knowledge on defects in diluted magnetic oxides becomes very important. The oxygen defects in BMP model are responsible for the shallow donors and strongly stabilize the ferromagnetic ground state. The oxygen vacancies at the percent level are documented for TiO 2 [95] and ZnO [96]. For ZnO, Zn interstitials (Zn i ) are also reported as a shallow donor [97, 98, 99]. The ab initio calculations of Errico et al. [100] also predict that Co-doped TiO 2 has more oxygen vacancies than the undoped ones. They also suggest that oxygen vacancies in a adequate portion can provide the necessary carriers to stabilize the ferromagnetic state. Recently, the electronic structure calculations of Patterson et al. [101] reported that the oxygen vacancy with a spin half strongly promote ferromagnetic coupling of Co ions in Codoped ZnO, as schematically presented in Fig A ferromagnetic ground state occurs between two Co ions through the oxygen vacancy spin, in agreement with the impurity band model of Coey et al.. 18

25 3.7 Transition metal clustering model Figure 3.8: A schematic energy level diagram for a ferromagnetic coupling between two Co ions via a donor electron trapped in the oxygen vacancy. Reproduced after Ref. [101] 3.7 Transition metal clustering model Another theoretical model to explain the origin of observed ferromagnetism in a DMS is proposed by Schilfgaarde and Mryasov [102] using local density functional calculations. They reported that the TM impurities in a DMS may form very small ferromagnetic clusters (just a few atoms in dimension) which are difficult to detect by most structural characterization techniques. The detection of small magnetic clusters in the oxide matrix is one important experimental challenge and will be discussed in more detail in the third part of this thesis. 19

26 3 Theoretical Models for Explaining Ferromagnetism in DMS 20

27 Part II Methods 21

28

29 4 Studying DMS Materials with Synchrotron Radiation In this chapter the basic properties of synchrotron radiation and the quantum mechanical description of the interaction between light and matter are briefly discussed. In the subsequent chapters XMCD and XRMS techniques as well as the Alice diffractometer which used to carry out XRMS and XMCD experiments at BESSY are explained. 4.1 Synchrotron radiation Synchrotron radiation is emitted when charged particles are accelerated close to the speed of light and are deflected by a magnet. The availability of high-brilliance and its other superior properties compared to conventional light sources such as x-ray tubes, make synchrotron radiation a most versatile tool. Radiation emission from accelerated charged particles was first postulated by Iwanenko and Pomeranschuk in 1944 [103]. For scientific investigations synchrotron radiation has been used for the first time in 1956 by Tomboulin and Hartman [103]. In the following years some research centers started to use synchrotron radiation for research activities. In the 1970s and 1980s, the storage ring facilities explicitly dedicated to the production of synchrotron radiation were constructed. These facilities are called as Second generation sources which produce synchrotron radiation in the dipole bending magnets of the storage ring. The Third generation synchrotron facilities appeared in the 1980s and 1990s. At these new synchrotron facilities synchrotron radiation is mainly produced by insertion devices (wigglers and undulators) and the brilliance of the synchrotron radiation is enhanced by several orders of magnitude in comparison to dipole sources. In future completely coherent x-ray sources might be available with free-electron lasers (FEL) with a brilliance many orders of magnitude higher than the synchrotron radiation from third generation storage rings Production of synchrotron radiation Production of synchrotron radiation in a typical synchrotron radiation facility requires several components [103]: a primary electron accelerating system, a synchrotron accelerator for further acceleration of electrons, a storage ring to store the accelerated electrons for several hours, insertion devices in the storage ring to produce the radiation and beamline systems with monochromators to tune and direct the radiation to an experimental end-station (Fig. 4.1). 23

30 4 Studying DMS Materials with Synchrotron Radiation Figure 4.1: Schematic representation of a typical synchrotron radiation facility. The initial acceleration of electrons In the initial acceleration stage, electrons are emitted by a hot cathode in high vacuum and accelerated by an anode voltage. As a second accelerating stage a microtron is used. In this linear accelerator the electrons pass through a strong high frequency electrical field ten times and they are redirected on larger growing trajectories by a strong constant magnetic field until they reach to high energies (about 50 MeV for BESSY II in Germany) [103]. Then the electrons are passed into the vacuum chamber of the synchrotron accelerator. The synchrotron accelerator In the synchrotron accelerator the electrons are focused on the correct trajectory by strong magnets and are accelerated to the final energy (about 1700 MeV for BESSY II) by the alternating field of cavity resonator [103]. The strength of the magnetic fields is increased as the kinetic energy of electrons increases to keep the electrons on the correct trajectory as they accelerated close to the speed of light. Then the electrons are injected into the storage ring through a transfer channel. The storage ring The electrons travel in the storage ring at nearly the speed of light for several hours. The stable trajectory of electrons are kept by strong electromagnets. In the storage ring, the direction and kinetic energy of the electrons are changed due to collisions between electrons and residual gas particles. This leads to loss of electrons. For this reason, the total number of electrons in the storage ring (or storage ring current) decreases as a function of time. The refilling of process so-called injection takes a few minutes due 24

31 4.1 Synchrotron radiation to high current provided by electron transfer from the synchrotron accelerator into the storage ring [103]. Synchrotron radiation can be produced by so-called insertion devices installed in the straight sections of a storage ring between the dipole bending magnets. Insertion devices The dipole bending magnets associated with second generation storage rings are the simplest sources for circular polarization. Although the synchrotron radiation emitted from bending magnets is highly polarized ranging from linear to nearly circular polarization, at certain desired degree of circular polarization the intensity falls dramatically. However, more intensity is very important for synchrotron radiation experiments. Because, it reduces the measurement time, increases the resolution and enables to investigate very small samples. To increase the intensity, periodic arrays of magnets so-called wigglers and undulators are inserted into straight sections of a storage ring. In a wiggler, a strong magnetic field force the electrons to move on a periodically wiggled trajectory over a length of several meters. In each curve radiation is emitted in the same forward direction. This leads to a strong circular polarization at high intensity. The undulator is one of the the most successful devices to produce circularly polarized synchrotron radiation. In an undulator, the electron beam is periodically deflected by weak magnetic fields. The electrons emit radiation at the wavelength of their periodic motion in the undulator. The spectral resolution of the emitted radiation is proportional to the number of undulator periods and the radiation wavelength can be changed by varying the magnetic field. Beamline and end-station A beamline consists of a monochromator, grazing incidence mirrors, pumps, valves, slits and beamshutter. When synchrotron radiation passes through the beam line optics the source polarization of radiation is almost completely preserved. A monochromator in an individual beamline allows the user to set a certain wavelengths of the light. Finally, the synchrotron radiation is directed by optical elements through a monochromator and comes to the experimental end-station. 25

32 4 Studying DMS Materials with Synchrotron Radiation 4.2 Quantum mechanical description of the interaction between light and matter In this section the quantum mechanical description of the interaction between light and matter is briefly introduced. The Hamiltonian of an atomic electron in a quantized electromagnetic field is given by the following equation [104] H = 1 ( p e A 2m c ) 2 + V 0 ( r) e }{{}}{{} 2mc s A }{{} potential energy kinetic energy Zeemann energy e 2(mc) 2 s E ( p e A c ) + ω k c + ( kσ)c( kσ) kσ } {{ } spin-orbit interaction } {{ } free radiation field, (4.1) where A defines the vector potential. This Hamiltonian can be split into contributions of the undisturbed electron system (H 0 ), free radiation field (H rad ) and interaction Hamiltonian (H int ) [104]: H = H 0 + H rad + H int, (4.2) where H int = e mc A p + e2 A 2mc 2 e ( 2 mc s A ) e e 2 ( ) A 2(mc) 2 c 2 s A, (4.3) H int H 1 + H 2 + H 3 + H 4 (4.4) Each term in the interaction Hamiltonian has different A-dependence. H 1 is linear in A and in the first order it determines the photoelectric absorption processes (Fig. 4.2(a)). In the second order, H 1 term gives rise to resonant scattering processes (Fig. 4.2(c)). The non-resonant Thomson scattering is described by H 2 (Fig. 4.2(b)). The spin-dependent terms H 3 and H 4 are only responsible for non-resonant magnetic scattering which is beyond the scope of this thesis. Next, the absorption and resonant scattering are described in more detail. 26

33 4.2 Quantum mechanical description of the interaction between light and matter Figure 4.2: Quantum mechanical description of the interaction between a photon and an atomic electron. Photoelectric absorption (a) and Thomson scattering (b) can be explained by applying first order perturbation theory. Resonant scattering (c) is a second order process and occurs via an intermediate state. In the style of Ref. [105] 27

34 4 Studying DMS Materials with Synchrotron Radiation Absorption The absorption cross section σ a is given by the transition rate probability divided by the incident photon flux (Φ 0 = c/v ) [106]; σ a = ω i f Φ 0. (4.5) Since the leading contribution of the interaction Hamiltonian to photoelectric absorption comes from the Hamiltonian H 1, the transition rate probability ω i f to first order perturbation theory is given by Fermi s golden rule [106], ω i f = 2π de f f H 1 i 2 ρ(e f )δ(e f E i ω), (4.6) where ρ(e f ) defines the density of final states. For the absorption the initial and final states are given by i = i el k i, σ ph, f = f el 0 ph. (4.7) Here the two different polarization states are indicated by the index σ = 1, 2 and k is the wave vector. For 2p 3d excitation processes of 3d transition metals, it is sufficient to consider dipole transitions only [106]. Then the absorption cross section is given by σ a = 4π2 e 2 c ω f ɛ r i 2 ρ(e f = E i + ω), (4.8) where ɛ is the polarization vector of the electromagnetic wave. Since the density of final states ρ is not the same for different spin directions in a magnetic material, the absorption of left and right circularly polarized light will be different, which is the origin of XMCD effect. A more detailed explanation of the XMCD effect is given in Chap

35 4.2 Quantum mechanical description of the interaction between light and matter Resonant scattering The interaction between x-ray and matter can be written in terms of the scattering amplitude [107] f( q, ω) = f 0 ( q) + f (ω) + if (ω), (4.9) where f and f are the real and imaginary parts of the dispersion corrections. If the photon energy is close to the absorption edges, they take their extremal values. The scattering amplitude f( q, ω) is related to the differential cross-section by the definition [106] dσ dω = f( q, ω) 2. (4.10) The main contributions of the interaction Hamiltonian to two photon processes are H 1 in the second order and H 2 in the first order perturbation theory, yielding [104] f( q, ω) = ( V ω f H 2π c 2 2 i + n ) f H 1 n n H 1 i E i E n + ω + iγ/2, (4.11) where initial and final states are given by i = i el k i, σ i ph, f = i el k f, σ f ph. (4.12) The first term in Eq describes the non-resonant Thomson scattering from all electrons in the atom. The non-resonant charge scattering f 0 ( q) is given by the Fourier transform of the atomic charge distribution. However, at 2p edges of the 3d transition metals the scattering vector is small compared to the atomic radius. Then the scattering amplitude can be approximated by f 0 ( q = 0) = r e Z, where r e is the electron radius and Z is the atomic charge number. The second term in Eq is responsible for resonant scattering. When the incident photon energy is close to an absorption edge ( ω = E n E i ), this second term in the scattering amplitude might be the dominant contribution. The process can be interpreted as an absorption of the incident photon exciting an electron to an intermediate state n and the falling of this electron back to the ground state i leads to a photon emission (Fig. 4.2(c)). If only dipole transitions are considered, the scattering amplitude can be written as [107] 29

36 4 Studying DMS Materials with Synchrotron Radiation f = ( ɛ f ɛ i )( r e Z + F (0) ) + i( ɛ f ɛ i ) mf (1) + ( ɛ f m)( ɛ i m)f (2), (4.13) with F (0) = 3λ [ ] F 1 8π 1 + F1 1, (4.14) F (1) = 3λ [ ] F 1 8π 1 F1 1, (4.15) F (2) = 3λ [ ] 2F 1 8π 0 F 1 1 F1 1. (4.16) m is the unit vector pointing along the magnetization direction and defines the quantization axis of the system. The first term proportional to ɛ f ɛ i refers to non-resonant and resonant charge scattering. The second term including ( ɛ f ɛ i) m is first order in the magnetization and results in circular dichroism and Kerr effects. The last term involving ( ɛ f m)( ɛ i m) is second order in the magnetization and leads to linear dichroism. The functions F (0,1,2) contain the energy-dependent material parameters [107] The polarization dependence of the resonant scattering In this section the polarization dependence of the resonant scattering amplitude f is briefly discussed. For a linearly polarized light, the polarization states are defined either perpendicular ( ɛ σ ) or parallel ( ɛ π ) to the scattering plane (Fig. 4.3). Figure 4.3: Definition of the polarization vectors for the incident and scattered photons. The first term in Eq. 4.13, describing charge scattering, does not change the initial polarization of the photon. While σ σ scattering is independent of the incident angle, π π scattering is dependent. The second term in Eq is proportional to the cross product of the polarization states and resonant magnetic scattering allows to change the initial polarization states, i.e. σ π and π σ scattering. 30

37 4.2 Quantum mechanical description of the interaction between light and matter Since the differential cross section in Eq is proportional to the absolute square of the scattering amplitude, the intensity of σ π and π σ scattering is quadratic in m. Therefore the only contribution linear in m to the scattering comes from the chargemagnetic interference term. This contribution linear in m leads to the transverse-moke effect which is sensitive to in-plane magnetization perpendicular to the scattering plane. The contribution of the third term in Eq to XRMS is generally supposed to be small at the L 2,3 edges of 3d transition metals [107]. For this reason, the discussion on this term is neglected here. In case of the circularly polarized light, the polarization vectors can be written as a linear combination of ɛ σ and ɛ π [106], ɛ ± = 1 2 ( ɛ σ ± i ɛ π ), (4.17) where the positive and negative signs show the right and left circular polarizations, respectively. The polarization state of the circularly polarized light is not changed by neither charge nor magnetic scattering. Within the framework of this thesis, the XRMS experiments using the circularly polarized light are performed in the longitudinal-moke geometry where the magnetization is parallel to the sample surface and the scattering plane. In this geometry, the dominant magnetic contribution to scattering again is caused by the charge-magnetic interference term. Since a circularly polarized light can be thought of to be a superposition of two linear polarization states and therefore includes ɛ π polarization state, transverse-moke can also be measured using circular polarization The relation between scattering and absorption phenomena The relation between absorption and scattering is given by two fundamental theorems; the optical theorem and the Kramers-Kronig relations. According to the optical theorem the absorption cross section is proportional to the imaginary part of the scattering amplitude in the forward direction [106], σ a = 4πr e k f ( q = 0). (4.18) The second step is to give a relation between the real and imaginary parts of the scattering amplitude or the refractive index. The refractive index is subdivided into real and imaginary parts according to following equation [108] n ± = 1 δ ± + iβ ± = 1 (δ ± δ/2) + i(β ± β/2). (4.19) 31

38 4 Studying DMS Materials with Synchrotron Radiation Here, δ and β are the dispersive and absorptive contributions, respectively. δ and β denote the magnetic contributions to the refractive index. If the energy dependence of β is known, the modified Kramers-Kronig relations in case of a magnetic medium result in [108] δ + (E) + δ (E) = 2 π P 0 E β +(E ) + β (E ) E 2 E 2 de, (4.20) δ + (E) δ (E) = 2E π P 0 β + (E ) β (E ) E 2 E 2 de, (4.21) where P defines the principal value [105]. The back transformations are given by β + (E) + β (E) = 2E π P 0 δ + (E ) + δ (E ) E 2 E 2 de, (4.22) β + (E) β (E) = 2 π P 0 E δ +(E ) δ (E ) E 2 E 2 de. (4.23) Using these relations both imaginary and real part of the refractive index can be calculated from a experimentally observed magnetic circular dichroism. 32

39 5 X-ray Magnetic Circular Dichroism X-ray magnetic circular dichroism is a well established technique that measures the difference in absorption of left and right circularly polarized x-rays by a magnetic material. XMCD has the advantage of element specificity that by tuning the photon energy close to a specific absorption edge, the magnetism of different elements can be studied. Using simple sum rules [109, 110], XMCD can also provide quantitative information about the distribution of spin and orbital magnetic moments separately. Furthermore, the new synchrotron radiation sources and improved end-stations make the XMCD experiments more routine on dilute samples. In this chapter the microscopic origin of the XMCD effect and commonly used XMCD detection methods are briefly discussed. 5.1 The origin of the XMCD effect The microscopic origin of the XMCD effect is explained by electronic transitions from spin-orbit split initial 2p states to exchange-split 3d final states (Fig. 5.1). The two-step model proposed by Stöhr and Wu [111] explains the origin of XMCD at the L 2,3 edges of 3d transition metals. In the first step, the interaction of circularly polarized photons with the p shell electrons leads to the excitation of spin-polarized electrons (photoelectrons). The spin-polarization depends on both the edge and the polarization of incoming photon: Left circularly polarized (LCP) photon excites 25% spin-up and 75% spin-down electrons at the L 2 edge. Right circularly polarized (RCP) photon does the opposite. At the L 3 edge, 62.5% spin-up and 37.5% spin-down electrons are excited by LCP photon. For RCP photon the situation is vice versa [112]. In the second step the spin-polarized electrons are analyzed by a spin-resolving detector consisting of 3d final states. In the case of ferromagnetic transition metals, the 3d final states are exchange split and there is an imbalance in the number of available unoccupied 3d spin-up and spin-down states. For this reason, the absorption of the LCP and RCP photons will be different. This difference is opposite at the L 2 and L 3 edges. If there are only unoccupied spin-down states, the detector is only sensitive to spin-down electrons and the XMCD effect is maximized. 5.2 Detection methods Transmission Transmission method is the most direct and accurate method to measure the x-ray absorption spectra [113]. However, since it is difficult to prepare samples for transmission 33

40 5 X-ray Magnetic Circular Dichroism Figure 5.1: A sketch of the spin-dependent absorption process in the two-step model. geometry, this method is not used very often. In transmission measurement, the beam intensity before the sample (I 0 ) is usually measured using the electron yield or photocurrent from a partially transmitting metal grid. The intensity after the sample (I) can be measured using a second grid or a Si photodiode (Fig. 5.2). The absorption coefficient µ ± is obtained by the equation where d is the thickness of sample. I = I 0 exp( µ ± d) (5.1) Figure 5.2: Transmission geometry for XMCD measurements. 34

41 5.2 Detection methods Total electron yield In the the total electron yield (TEY) method, the absorption spectra is obtained by monitoring all excited photoelectrons created during the absorption process in the sample. The electron yield can be measured directly by using a channeltron electron multiplier, or indirectly as the photocurrent of electrons flowing to the sample from ground (Fig. 5.3). Since electrons are only emitted from the surface of the sample (about 20 Å for 3d transition metals [114]), this method is very sensitive to oxidation or other surface reactivity. For very high cross sections, this method may also suffer from the saturation effects [115]. Figure 5.3: Schematic representation of TEY and fluorescence yield methods for measuring the XMCD signal. If magnetic field switching is used to measure the XMCD effect, the data taken with the electron yield method can also be effected. In this case, the trajectories of emitted photoelectrons will also depend on the applied magnetic field. For this reason, electron yield measurements can best be done with changing incoming photon polarization. It should also be noted that even polarization-switched measurements are effected by potential artifacts. If the effective source point for left and right circularly polarized beams is slightly different, this leads to an energy difference between the two beams at a given monochromator position. The slight difference will result in a contribution to the spectrum. Since this effect is independent of the applied field, it is necessary to check that there is no XMCD effect in the absence of sample magnetization [112] Fluorescence yield When the photoelectrons leave the sample, positively charged holes remain. These core holes in an inner shell are subsequently filled by Auger electrons from outer shells (99.2%) and x-ray fluorescence decay (0.8%) [116]. In fluorescence yield method, the fluorescence signal is detected by using a photodiode detector which is located close to the sample surface (Fig. 5.3). Since the magnetic field will not affect the emitted photons, in contrast to TEY method, fluorescence yield method might seem independent of magnetic 35

42 5 X-ray Magnetic Circular Dichroism field effects. However, since most detectors convert photons into electrons, the detector resolution can also be influenced by a strong field. In this case, the detector sensitivity must be controlled before by measuring a XMCD signal for varying the magnetic field. 36

43 6 X-ray resonant magnetic scattering X-ray resonant magnetic scattering (XRMS) combines the advantages of conventional x-ray scattering and XMCD techniques. After the first pioneering experiments [117, 118], XRMS has proven to be a highly effective method for the analysis of the magnetic properties of buried layers and interfaces, including their depth dependence [119, 120]. Moreover, if the photon energy is fixed close to the energy of the corresponding absorption edges, element specific hysteresis loops can be measured by varying the external magnetic field. The most common geometry for XRMS experiments is the longitudinal-moke (L-MOKE) geometry, where the sample magnetization is parallel to the plane of incidence and parallel to the sample surface (Fig. 6.1). For ferromagnetic materials, to obtain the magnetic contribution to the resonant scattering, the asymmetry ratio (I + I )/(I + + I ) is used. Here I ± indicates either the two photon helicities or two antiparallel orientations of the external magnetic field. Since it is easier to realize experimentally, a switching of the external magnetic field is preferred in most experiments. Figure 6.1: L-MOKE geometry for the XRMS measurements. In order to evaluate the magnetization depth profile of multilayers from the magnetic part of the XRMS spectra, a theoretical formulation is needed. A matrix based formalism for magneto-optics with arbitrary magnetization direction has been developed by Zak et al. [121, 122, 123] which offers the possibility to simulate the XRMS spectra in the specular condition for an electromagnetic radiation with an arbitrary incidence angle and polarization. In this formalism, it is assumed that two media are separated by a single boundary (Fig. 6.2). The incoming and reflected electromagnetic waves are described in a basis of polarization states perpendicular (E σ ) and parallel (E π ) to the plane of incidence. These electric field components of electromagnetic waves can be collected in a four-component vector, 37

44 6 X-ray resonant magnetic scattering Figure 6.2: Refraction by a single boundary of two media. E π and E σ are the electric field components parallel and perpendicular to the plane of incidence. The arbitrary direction of the magnetization is given by φ and γ angles in the xyz system. In the style of Ref. [122]. P = E (i) σ E (i) π E (r) σ E (r) π, (6.1) where i and r indicate the incident and reflected waves, respectively. When an electromagnetic wave travels from medium 1 into medium 2, the tangential components of its electric and magnetic fields are conserved. Thus it is more useful to change the basis to F = E x E y H x H y. (6.2) Now a medium boundary matrix A that connects F with P can be defined by the expression F = A P. (6.3) Having the medium boundary matrix A, the boundary matching condition ( F 1 = F 2 ) can be written down as 38

45 A 1 P1 = A 2 P2. (6.4) The arbitrary direction of the magnetization M is specified by means of the polar coordinates φ and γ in the xyz system (Fig. 6.2): M x = M sin φ cos γ (6.5) M y = M sin φ sin γ (6.6) M z = M cos φ. (6.7) With these definitions for an arbitrary direction of M, the dielectric tensor ɛ can be written as ɛ(ω) = N 2 1 i cos φq i sin γ sin φq i cos φq 1 i cos γ sin φq i sin γ sin φq i cos γ sin φq 1, (6.8) where Q is the Voigt parameter. Using Snell s law and Maxwell s equations, the relations between the components of the electric field vector E in the magnetic medium and the expression for the medium boundary matrix A can be calculated. Details of the calculation can be found in Refs. [121, 122, 124]. 39

46 6 X-ray resonant magnetic scattering 6.1 The Alice diffractometer for XRMS and XMCD experiments The XRMS and XMCD experiments within this thesis were carried out using the ALICE diffractometer [125] at the undulator beam lines UE56/1-PGM and UE52-SGM at BESSY II (Berlin, Germany). Due to strong air absorption of x-rays in the considered energy range (450 ev and 1100 ev), the XRMS and XMCD experiments must be performed under vacuum conditions. For this reason, the diffractometer comprises a two-circle goniometer installed in a cylindrical vacuum chamber and works in horizontal scattering geometry (see Fig. 6.3). The chamber is pumped by a turbo-molecular pump which is supported by a scroll type pump for initial rough pumping. Although the chamber is compatible with UHV conditions, high vacuum conditions (< mbar) are enough [126] to carry out the XRMS and XMCD experiments within this study. To ensure the UHV environment of the beam line optics and the storage ring, there is a differential pumping state consisting of a pinhole P1 and a turbo-molecular pump between the beam line and the chamber (Fig. 6.3). Figure 6.3: Schematic topview of the Alice diffractometer. Two linear feedthroughs F1 and F2 which consist of a pentaprism and a pinhole (P2), respectively, are used in the chamber. The pentaprism in front of the pinhole can reflect a laser beam along the path of the x-ray beam which allows the user to align the sample before the vacuum system is pumped [127]. In addition to the pentaprism, a slit (S1) is also installed on the linear feedthrough F1. In order to define the angular resolution during the experiments, S1 is used together with S2 which is the second slit in the chamber located in front of the detector. Setting these slits to 300µm leads an angular resolution of The sample and the detector are moved by two differentially pumped rotating platforms 40

47 6.1 The Alice diffractometer for XRMS and XMCD experiments (RP) which are mounted on top of the chamber with a common center. A manipulator consisting of a XY table and a linear Z translator allows to move the sample linearly in all directions relative to the x-ray beam. The sample can be changed through a load-lock window. A He closed-cycle cryostat is used to cool the sample. Using a heater the sample temperature can be varied between 30 K and 380 K. The cryostat is controlled by a Lakeshore 340 temperature controller. For temperatures higher than 320 K, the cold head and the heater can be thermally decoupled by evacuating a intermediate cavity, which is filled with 800 mbar of He for low temperature measurements. The lowest-possible temperature with closed-cycle He cryostat is 30 K. Other disadvantages of the closed-cycle He cryostat are the unavoidable sample vibrations caused by the expander module of the cold head and the long cooling time until the lowest temperature is reached. In order to solve the mentioned problems a continuous-flow liquid He cryostat was installed. With this new cryostat the sample temperature can be varied between 4.2 K and 475 K by using a heater. A rotatable electromagnet is used to apply a magnetic field in the scattering plane along the sample surface (L-MOKE geometry). With the gap size of 15 mm between magnet poles, a maximum field of ±0.27 Tesla can be reached. The magnetic contribution to the scattered intensity was always measured by switching the magnetic field at fixed photon helicity. The incident beam intensity I 0 is monitored by using the photoelectric current of the refocusing mirror in the beamline optics. The scattered beam (XRMS) is detected by using a photodiode detector as shown in Fig The second photodiode detector also located close to the sample above the scattering plane allows to measure absorption (XMCD) via the fluorescence yield. The absorption spectra can also be taken by the total electron yield (TEY) method by measuring the sample drain current. The corresponding photocurrent from the detectors and the sample drain current are measured by Keithley electrometers. Stepper motors (for the sample, detector and magnet) and magnetic field are controlled via GPIB interface from a Linux PC. Data acquisition is performed using standard SPEC software [128]. 41

48 6 X-ray resonant magnetic scattering 42

49 Part III Results and Discussion 43

50

51 7 Dose Dependence of Magnetism in Co-implanted TiO 2 Rutile 7.1 Overview Titanium dioxide TiO 2 is an interesting material due to many properties. TiO 2 occurs in several crystal structures, such as rutile, anatase and brookite. Rutile is tetragonal (a = 4.59 Å and c = 2.96 Å) and thought to be the most stable configuration with a band gap of 3 ev, while the anatase phase also crystallizes in a tetragonal structure (a = 3.78 Å and c = 9.52 Å) with a band gap of 3.2 ev [129]. Optical transparency to visible light and very high refractive index make TiO 2 a superior candidate for magneto-optic applications. The discovery of room temperature ferromagnetism in Co-doped TiO 2 [62] has attracted considerable attention to titanium dioxide as a host material for magnetic doping. Co-doped TiO 2 has been grown by using a wide variety of growth methods, including pulsed laser deposition (PLD) [63, 130, 131, 132], laser molecular beam epitaxy (LMBE) [133, 134, 135], combinatorial LMBE [62, 136], reactive co-sputtering [137], magnetron sputtering [138, 139], metal organic chemical-vapor deposition (MOCVD) [140], oxygen plasma assisted molecular beam epitaxy (OPA-MBE) [141, 142, 143] and as well as the sol-gel method [64]. The single-crystalline TiO 2 rutile has also been doped by using ion implantation technique [65, 66, 144, 67, 69, 70, 71]. In addition to different growth techniques, different substrates such as Al 2 O 3 [145, 146], SrTiO 3 [141, 63, 142, 131, 130, 143, 139], LaAlO 3 [62, 142, 133, 131, 130], Si [137] and SiO 2 /Si [140] have been used to synthesize Co-doped TiO 2 films. Chambers et al. [141] reported that the solution of Co in TiO 2 is possible at least up to 10% when the TiO 2 is deposited on SrTiO 3 substrate. However, when the TiO 2 films grown on LaAlO 3 substrate the solid solution is about 2-7% [62, 131]. Co metal clusters were observed in the as-grown Co-doped TiO 2 films with a cobalt concentration of 2%. Post-annealing of the samples leads to dissolving of clusters in the TiO 2 matrix [131]. For higher cobalt concentrations, bigger cobalt cluster were reported with a cluster size which is about 150 nm [140]. Moreover, the charge state of substitutional Co in TiO 2 matrix has been determined to be +2 by different groups using x-ray absorption spectroscopy and x-ray photoelectron spectroscopy [141, 142, 64, 133, 135]. Many groups have observed room temperature ferromagnetism in Co-doped TiO 2 for both anatase and rutile phases [62, 137, 131, 65, 66, 67, 68, 69, 70, 71, 147]. A Curie temperature of about 650 K [131] and 700 K [66] was reported by different groups. For further detailed information on magnetic properties of Co-doped TiO 2, it is referred to recently published reviews [52, 54, 57]. Although a lot of experimental reports appeared 45

52 7 Dose Dependence of Magnetism in Co-implanted TiO 2 Rutile on this system, the origin of observed room temperature ferromagnetism in Co-doped TiO 2 is not clear yet. In this chapter 1 the structural, magnetic and electronic properties of Co-implanted n- type TiO 2 rutile single crystals are reported for different implantation doses and for a wide temperature range. The resulting TiO 2 samples have been characterized by RBS to obtain the Co distribution profile and by AFM to check the surface properties after implantation, as well as by XRD and by high resolution (HR) TEM to reveal the presence of precipitates and metallic Co particles. X-ray absorption spectroscopy (XAS) has also been used to determine whether the implanted cobalt ions are in the Co 2+ oxidation state or exhibit pure metallic behavior. The magnetic properties of TiO 2 rutile samples have been investigated using MOKE, SQUID and XRMS techniques. In addition, anomalous Hall effect (AHE) measurements were carried out to verify the occurrence of intrinsic ferromagnetism and relate it to the carrier type in the samples. Strong ferromagnetism is observed in TiO 2 rutile plates after cobalt ion implantation, with magnetic parameters depending on cobalt implantation dose. While the structural data indicate the presence of metallic cobalt clusters, the multiplet structure of Co L 3 edge in the XAS spectra gives a clear evidence for the substituted Co 2+ state. The detailed analysis of the structural and magnetic properties indicates that there are two magnetic phases in Co-implanted TiO 2 plates. One is the ferromagnetic phase due to the the formation of a long range ferromagnetic ordering between implanted magnetic cobalt ions and the second one is a superparamagnetic phase which occurs by the formation of metallic cobalt clusters in the implanted region. Using x-ray resonant magnetic scattering, the element specific magnetic hysteresis curves of cobalt and oxygen in Co-implanted TiO 2 plates are also investigated. Magnetic dichroism was observed at the Co L 2,3 edges as well as at the O K edge. The magnetic hysteresis determined for both the Co L 3 and the O K edges agree with each other concerning their shape and their coercive fields which indicates that the oxygen atoms close to the cobalt atoms are magnetically polarized. The interaction mechanism which leads to ferromagnetic ordering of substituted cobalt ions in the host matrix is also discussed. 7.2 Sample Preparation 40 kev Co + ions implanted into (100)-oriented mm 3 single-crystalline TiO 2 rutile substrates (from Moscow Power Engineering Institute in Russia) by using the ILU- 3 ion accelerator (Kazan Physical-Technical Institute of Russian Academy of Science) with an ion current density of 8µA cm 2. The implantation dose varied in the range of ions cm 2. The sample holder was cooled by flowing water during the implantation to prevent the samples from overheating. The (100)-face rutile plates were irradiated in a single run. The implanted plates were cut by a diamond cutter into smaller pieces for structural, magnetic and electronic studies. As a last step, four gold contacts were evaporated on the corners of the samples for AHE measurements (Fig. 7.1). The list of the Co-implanted TiO 2 samples is given in Table Some parts of this chapter are published in the articles Spin polarization of oxygen atoms in ferromagnetic Co-doped rutile TiO 2 [70] and Dose dependence of magnetism in Co-doped TiO 2 [71]. 46

53 7.2 Sample Preparation Figure 7.1: Sample preparation stages for Co-implanted TiO 2 rutile plates. Table 7.1: List of the T io 2 samples implanted with 40 kev Co + for varying Co ion dose. Sample Substrate Dose ( ion cm 2 ) 1 (100)-TiO 2 rutile (100)-TiO 2 rutile (100)-TiO 2 rutile (100)-TiO 2 rutile (100)-TiO 2 rutile (100)-TiO 2 rutile

54 7 Dose Dependence of Magnetism in Co-implanted TiO 2 Rutile 7.3 Structural Properties In this section, the structural properties of non-implanted and Co-implanted TiO 2 rutile plates are presented. The depth distribution of implanted cobalt ions in the rutile samples as well as the cobalt concentration for each doses are determined by using the RBS technique. The RBS measurements were carried out at the Dynamic Tandem Laboratory (DTL) at Ruhr-Universität Bochum. Fig. 7.2 shows the comparison of the nominal and measured implantation doses determined by the RBS technique. Due to the difficulties for setting the exact ion dose values during implantation, the measured implantation dose always differs from the nominal doses with the exception of the lowest dose doped sample. This deviation increases with increasing of the ion implantation dose and it is more pronounced for the highest doses. Figure 7.2: The comparison of measured and nominal implantation doses. The depth dependence of the cobalt concentration in Co-implanted TiO 2 plates was also investigated using RBS technique for varying Co ion implantation dose (Fig. 7.3). The RBS data show a maximum of cobalt concentration is about 25 at. % for the highest Co doses and it decreases to about 5 at. % for the lowest dose doped samples. Since the penetration depth of Co ions in TiO 2 rutile plates depends only on their energies, independent of the dose, the maximum always occurs at the same depth. An extended inward tail due to cobalt diffusion into the volume of the rutile substrates is also present for each implantation doses. Fig. 7.4 shows the small-angle x-ray reflectivity data taken with synchrotron radiation at the Hamburg Synchrotron Radiation Laboratory (HASYLAB) with an energy of E=8048 ev. The solid line in Fig. 7.4 is a fit to the data points for sample 6 ( ions cm 2 ) which is obtained with the commercial software WinGIXA based on the 48

55 7.3 Structural Properties Figure 7.3: The cobalt concentration as a function of depth measured using RBS technique. Parratt formalism [148]. Since the cobalt concentration in TiO 2 changes depending on depth, for fitting of the reflectivity data the implanted area is sliced into five layers. The roughness and electron density values obtained from the fit are listed in Table 7.2 for each layer. The model used for fitting perfectly matches the RBS data and Fig. 7.5 shows the depth dependence of the cobalt concentration and normalized electron density obtained from the reflectivity fit. The solid line in Fig. 7.5 presents the calculated SRIM 2 profile without taking into account ion sputtering effects. Table 7.2: Fitting parameters of reflectivity curve for sample 6 ( ions cm 2 ). Layer Thickness (nm) Roughness ( Å) Electron density (g cm 3 ) 1. layer layer layer layer layer Substrate The high angle XRD measurements were also carried out at HASYLAB to detect possible additional phases in the samples after implantation. The Bragg scans before and after implantation with different doses are shown in Fig. 7.6 for (100)-oriented TiO 2 rutile samples. Increasing of the implantation dose up to ions cm 2 results in two 2 The Stopping and Range of Ions in Matter (SRIM) [149]. SRIM-2006 software at 49

56 7 Dose Dependence of Magnetism in Co-implanted TiO 2 Rutile Figure 7.4: Small angle x-ray reflectivity data and fit for sample 6 ( ions cm 2 ). Figure 7.5: The cobalt concentration (RBS data) and the simulated electron density as a function of depth for sample 6 ( ions cm 2 ). Solid line represents the calculated SRIM profile. additional peaks which correspond to (1010) and (0002) reflections of the Co hcp structure. Below the implantation dose of ions cm 2 Co nanoclusters can not be detected. 50

57 7.3 Structural Properties Figure 7.6: High-angle Bragg scattering data for non-implanted (solid line) and different dose implanted (100)-TiO 2 samples. The presence of Co clusters is clearly seen for the highest dose (Sample 6). For every implanted sample a tail, shown by an arrow in Fig. 7.6, is present around the main peak of TiO 2 (200) reflection. This tail results from the expansion of the TiO 2 lattice upon cobalt implantation and is not observed before implantation. In addition a new peak is present on the low angle side which corresponds to a cobalt oxide (Co 3 O 4 ) phase and it is also reported by Khaibullin et al. [147]. Due to the difference in sputtering rates of the metal and the host material [150, 151] during high dose ion implantation of rutile films, the cobalt nanoparticles form at the surface and become oxidized forming antiferromagnetic Co 3 O 4 with low Neel temperature of about 40 K. Fig. 7.7 presents the surface morphology of sample 5 ( ions cm 2 ) taken by using AFM technique 3 (Digital Instruments NanoScope MultiMode AFM). The AFM image clearly shows a network of cobalt oxide islands on surface with a roughness of about 2.14 ± 0.25 nm. For further investigations on the effects of ion implantation into TiO 2, high resolution cross sectional TEM measurements were performed. For the preparation of TEM samples, focused ion beam (FIB) technique is used. First, the sample surface is covered by tungsten (W) film to prevent charging effects. Then very small cross sectional piece of implanted sample was cut by FIB 4. Fig. 7.8 presents TEM images of sample 6 ( ions cm 2 ) 3 The AFM measurements are carried out in the laboratory of Prof. C. Wöll (Institut für Physikalische Chemie I, Ruhr-Universität Bochum) within the IMPRS-SurMat scientific collaboration. 4 The sample preparation using FIB and the TEM measurements were performed in the laboratory of Prof. G. Eggeler (Institut für Werkstoffe, Ruhr-Universität Bochum) within the scientific collaboration of IMPRS-SurMat. 51

58 7 Dose Dependence of Magnetism in Co-implanted TiO 2 Rutile Figure 7.7: AFM surface image of (100)-TiO 2 rutile after Co ion implantation with a dose of ions cm 2 (Sample 5). Figure 7.8: Cross-sectional TEM images of sample 6 ( ions cm 2 ). 52

59 7.3 Structural Properties with an increasing magnification from 50 nm to 2 nm. In Figs. 7.8(a) and (b), a general overview of the sample is shown. It is clearly seen that about 40 nm thick surface layer is highly destroyed after ion bombarding. There are many defects and differently sized cobalt clusters in this region. However, in Figs. 7.8(c) and (d) it can be recognized that the structure of TiO 2 is preserved after implantation. Beneath the surface layer there is another cobalt rich layer of about 40 nm thick. Element specific TEM measurements indicate that the cobalt concentration in this layer is much smaller than in the surface layer in agreement with the RBS data (Fig. 7.3). 53

60 7 Dose Dependence of Magnetism in Co-implanted TiO 2 Rutile 7.4 Magnetic Properties The magnetic properties of Co-implanted rutile TiO 2 single crystals have been investigated in detail for different implantation doses and for a wide temperature range from 4.2 K to 390 K using a combination of different experimental techniques including MOKE, SQUID, XRMS and XAS Room temperature measurements Figure 7.9: Longitudinal MOKE hysteresis loops of the (100)-oriented TiO 2 samples for different Co ion doses taken parallel to the easy (closed symbols) and hard axis (open symbols). Fig. 7.9 shows the room temperature hysteresis loops of the TiO 2 samples for different implantation doses measured using a high-resolution MOKE setup [152, 153, 154] in the longitudinal configuration with s-polarized light. Hysteresis loops taken parallel to the easy and to the hard axis orientations are shown in Fig. 7.9 by closed and open symbols, 54

61 7.4 Magnetic Properties respectively. For the easy axis orientation three types of hysteresis loops can be distinguished. For the highest dose ( ions cm 2 ) a square-like hysteresis curve is observed with a large coercive field of H C =0.95 koe. A rather sharp magnetization reversal takes place for this sample with a small step at 0.2 koe. For TiO 2 samples implanted with intermediate ion doses ( ions cm 2 ), the recorded M H loops also demonstrate hysteretic behavior, but the coercive fields decrease significantly. The remanent Kerr signal normalized to the Kerr signal at saturation is 0.94 for the highest dose and decreases for the intermediate doses. The low dose ( ions cm 2 ) implanted sample demonstrates typical superparamagnetic behavior at room temperature. It should be mentioned that the samples with an implantation dose below ions cm 2 provide poor MOKE signals and are therefore omitted from the present discussion. As for the easy axis orientation, the hard axis hysteresis loops measured at intermediate implantation doses are quite similar. However, the hard axis hysteresis loop for the highest dose ( ions cm 2 ) differs drastically, indicating that the in-plane magnetic anisotropy is much stronger for this sample. Figure 7.10: The ratio of remanence Kerr signal to the saturation signal (left) and the coercive field (right) as a function of in-plane rotation angle ϕ for different Co ion doses. The MOKE setup also allows for a rotation of the sample around its surface normal (by the angle ϕ) in order to apply a magnetic field in various in-plane directions and provide information about the in-plane magnetic anisotropy. The in-plane magnetic anisotropies of the samples doped with different doses as determined by the MOKE measurements are shown in Fig For the sample with highest dose, both the remanent Kerr signal normalized to the Kerr signal at saturation (θk rem/θsat K ) and coercive field (H C) are reduced to almost zero value near the hard axis orientations (ϕ = ), while for the magnetic field applied along the easy axis directions (ϕ = ) they are close to unity. It is evident from Fig that both θk rem/θsat K and H C exhibit a strong two-fold symmetry for the highest dose which decreases with decreasing implantation dose. To shed more light on the origin of room temperature ferromagnetism in Co-doped TiO 2, the magnetic properties of Co-implanted TiO 2 rutile films have also been investigated 55

62 7 Dose Dependence of Magnetism in Co-implanted TiO 2 Rutile using the XRMS and XAS techniques. Both the XRMS and XAS experiments were carried out at the undulator beam lines UE56/1-PGM and UE52-SGM at BESSY II (Berlin, Germany) using the ALICE diffractometer [125]. The diffractometer comprises a two circle goniometer and works in horizontal scattering geometry. A maximum magnetic field of ±0.27 Tesla can be applied in the scattering plane along the sample surface either parallel or antiparallel to the photon helicity, which corresponds to the longitudinal magnetooptical Kerr effect (L-MOKE) geometry. For the details of the diffractometer please see Sec The magnetic contribution to the scattered intensity (XRMS) was always measured by reversing the magnetic field direction while keeping the photon helicity fixed. Due to the observation of additional features in the room temperature MOKE hysteresis of sample 6 ( ions cm 2 ), this sample was investigated in more detail than the other samples. In Fig the reflectivity data of sample 6 is shown. These data were taken at room temperature and with a magnetic field applied in the sample plane parallel (I +, solid line) and antiparallel (I, dash line) to the photon helicity. The photon energy of the circular polarized light was tuned close to the L 3 absorption edge of Co (780 ev). The weak oscillations in the range of 2θ = 4 16 are due to the density difference between implanted and non-implanted regions of the TiO 2 plate. The splitting of the two curves, which is proportional to the sample magnetization, is clearly seen. Therefore, as a compromise between high scattering intensity and high magnetic sensitivity for the investigation of the magnetic properties at the Co L edges, the scattering angle was fixed at the position of 2θ = 8.2. For measurements at the O K edge (E=535 ev) the scattering angle was fixed at 2θ = 12, which corresponds to the same scattering vector in the reciprocal space. The energy dependence of the intensity around the Co L 2,3 edges measured in saturation at positive (closed symbols) and negative (open symbols) fields are shown in Fig for sample 6. The magnetic contribution to the resonant scattering can best be visualized by plotting the asymmetry ratio (R = (I + I )/(I + + I )). In Fig the asymmetry ratios for samples with different dose measured at saturation field are plotted against the photon energy. The asymmetry of the samples decreases with decreasing implantation dose in agreement with the MOKE measurements. It is important to note that the fine structure which is clearly seen in Figs and 7.13 around the Co L 3 edge is not typical for metallic cobalt. It is well known that in the case of metallic films the absorption spectra around the L 3 peak of Co consists of a single component [155]. This fine structure of the Co L 3 peak is similar to that observed before for CoO films [155] (see inset in Fig. 7.14, and it is indicative of a Co 2+ state). To make this situation more clear and to record comparable results with previous reports [133], XAS experiments were also performed. The absorption data were taken by the total electron yield (TEY) method (see Sec. 5.2), i.e. by measuring the sample drain current. Since the external magnetic field changes the excited electron trajectories, the XAS spectra were measured with fixed photon helicity at remanence. The angle of incidence was chosen again to be 4.1 with respect to the surface. The absorption spectra were normalized to the incoming photon flux measured from the beam line mirror. The averaged x-ray absorption spectra (σ + +σ )/2 at the Co L 2,3 edges is shown in Fig The XAS spectrum clearly shows a multiplet structure at the L 3 edge. This multiplet structure is a clear sign of the presence of the Co 2+ state in this sample. 56

63 7.4 Magnetic Properties Figure 7.11: Reflectivity scans of sample 6 ( ions cm 2 ) taken with a magnetic field applied in the sample plane parallel (I +, solid line) and antiparallel (I, dash line with open circles) to the photon helicity In Fig the magnetic hysteresis loops of the sample 6 ( ions cm 2 ) are shown which have been recorded at different photon energies of 776 ev and 780 ev. At the energy of 780 ev a square like shape of hysteresis is observed. However, at the energy of 776 ev, the shape of hysteresis curve changes drastically. The coercive field remains at the same value, but jumps appear around remanence. To understand this behavior some data treatment was made. The hysteresis taken at 780 ev was subtracted from the hysteresis measured at 776 ev. The resulting curve has two components: The first one (solid line in Fig. 7.15) has the same shape as the hysteresis curves measured by MOKE for the intermediated doses. The second curve (open triangles in Fig. 7.15) is proportional to the square of the M y component of magnetization. After this data treatment it becomes clear that there are two cobalt phases in this sample and the XRMS gives possibility to separate the hysteresis curves of these different cobalt phases. The discussion for the origin of these two Co phases will be given later, but it is already evident now that they refer to the intrinsic oxide and extrinsic cobalt cluster phase. The magnetic signal at the Ti L 2,3 (Fig. 7.16) and the O K (Fig. 7.17) edges was also investigated for sample 6. Within the sensitivity limit no magnetic signal could be recorded for Ti and O at room temperature. However, at the O K edge, a small but clearly visible magnetic signal was observed at T=30 K which will be discussed next in the section on the low temperature experiments. 57

64 7 Dose Dependence of Magnetism in Co-implanted TiO 2 Rutile Figure 7.12: Energy dependence of scattering intensities at the Co L 2,3 edges for sample 6 ( ions cm 2 ). Figure 7.13: Asymmetry ratios of (100)-TiO 2 rutile for different Co implantation doses measured at saturation field. 58

65 7.4 Magnetic Properties Figure 7.14: X-ray absorption spectra of sample 6 measured at the Co L 2,3 edges by using TEY method. σ + and σ denote the right and left circularly polarized light, respectively. The inset shows the normalized experimental spectrum for CoO/Fe sandwich (solid black) and, for comparison, the normalized TEY spectra of pure cobalt metal (gray) and cobalt oxide (dotted black) taken from Ref. [155]. Figure 7.15: Room temperature hysteresis curves for sample 6 measured at the energies of 776 ev and 780 ev. The solid line and open triangles present the M x and M 2 y components of magnetization after the data treatment. 59

66 7 Dose Dependence of Magnetism in Co-implanted TiO 2 Rutile Figure 7.16: XRMS data measured at the Ti L 2,3 edges for sample 6. Figure 7.17: XRMS data taken at the O K absorption edge for sample 6. 60

67 7.4 Magnetic Properties Low temperature measurements Fig shows the dose dependence of the asymmetry ratios at the Co L edges measured at T=30 K in reflection geometry and at remanence. It can be clearly seen from Fig that the magnetic signal decreases by decreasing the Co ion dose and it goes to zero for the lowest dose at remanence. Figure 7.18: Dose dependence of the asymmetry ratios at the Co L edges measured at remanence. Fig presents the asymmetry ratio of sample 6 ( ions cm 2 ) taken at the oxygen K edge and at T=30 K. Note that the maximum in the asymmetry ratio of oxygen is by roughly a factor of 20 smaller than the asymmetry ratio of cobalt measured at 30 K (closed squares in Fig. 7.18). It should be mentioned that the oxygen polarization has also been observed for samples with lower dose (dose levels of ions cm 2 ). For samples implanted with doses below ions cm 2, the magnetic signal at the O K edge is below the sensitivity limit of experimental setup. Fig shows the element specific magnetic hysteresis curves of sample 6 recorded at the Co L 3 (closed symbols) and O K (open symbols) edges at T=30 K. The shape of the hysteresis curve measured at E=780 ev (Co L 3 edge) is essentially the same as the one recorded previously at room temperature (Fig. 7.15), but the coercive field is increased by a factor of 2 from H C =955 Oe at room temperature to H C =1900 Oe at 30 K. The hysteresis curve measured at E=533 ev (O K edge) is shown by open symbols in Fig The shape of this hysteresis curve is the same as the one recorded at the Co L edge (closed symbols), including the coercive field, but the intensity is much lower. This implies that the oxygen atoms in close proximity to Co atoms are polarized. 61

68 7 Dose Dependence of Magnetism in Co-implanted TiO 2 Rutile Figure 7.19: XMCD signal (asymmetry ratio) at the oxygen K edge measured at T=30 K. Figure 7.20: Hysteresis curves of sample 6 measured at Co L 3 (closed symbols) and O K (open symbols) edges at T=30 K. To study the effects of cobalt clusters on the magnetic properties of Co:TiO 2, temperature dependent magnetization (M T ) measurements are performed using a Quantum Design 62

69 7.4 Magnetic Properties MPMS XL SQUID magnetometer. Fig presents field cooled (FC) and zero field cooled (ZFC) plots for each sample. For ZFC measurements, the samples are cooled in zero field to 5 K and the magnetization is recorded during warming up to 390 K with an applied field of 0.1 koe, parallel to the film surface. For FC measurements the applied field of 0.1 koe is kept during cooling to 5 K and the magnetization is recorded during field warming. The FC and ZFC curves diverge substantially for all doses and the peak in the ZFC curve progressively shifts to higher temperatures with increasing cobalt concentration. These behaviors are not expected for a ferromagnet and suggest the presence of magnetic cobalt nanoparticles in the films or a spin-glass like nature of the system [156, 157]. The trend in the M T curve of sample 1 ( ions cm 2 ) can be attributed to the coexistence of ferromagnetic and superparamagnetic phases with a transition temperature of about 30 K. The XRMS hysteresis of this sample measured at 4.2 K (Fig. 7.22) supports this conclusion. The M T curves for films with higher cobalt concentrations ( ions cm 2 ) indicate the occurrence of superparamagnetism with a blocking temperature of about 100 K and 250 K for sample 2 ( ions cm 2 ) and sample 3 ( ions cm 2 ), respectively. The XRMS hysteresis loops of these films (Fig and Fig. 7.24) show further evidence for the presence of single domain particles as well as superparamagnetism. A large coercive field at 30 K and rapid decrease of H C and remanent magnetization (M R ) observed for these samples. It can be seen from Fig and Fig that both H C and M R goes to zero above 100 K and 250 K for sample 2 and sample 3, respectively. The temperature dependence of sample 4 (Fig. 7.21) is similar to that of sample 2 and sample 3 except the blocking temperature is much higher and above 390 K. It is also important to note that the FC curve of this sample shows a more or less continues behavior versus temperature which is typical for ferromagnets. The reported room temperature ferromagnetism with a two fold in-plane magnetic anisotropy in this sample (Figs. 7.9 and 7.10) indicates that at this dose substituted cobalt ions start to interact ferromagnetically. The FC and ZFC curves of sample 5 ( ions cm 2 ) and sample 6 ( ions cm 2 ) are much closer to each other. This behavior shows that the ferromagnetic phase becomes dominant in these samples. Observation of a two component hysteresis (Fig. 7.15) and the magnetic signal at the O K edge (Fig. 7.19) for sample 6 support this argument. But small peaks at ZFC curves of these samples indicate the existence of superparamagnetic cobalt clusters. These clusters can also be clearly seen in the TEM images of sample 6 (Fig. 7.8). 63

70 7 Dose Dependence of Magnetism in Co-implanted TiO 2 Rutile Figure 7.21: FC (closed squares) and ZFC (open circles) magnetization curves of Coimplanted (100)-TiO 2 rutile samples taken using SQUID magnetometry. 64

71 7.4 Magnetic Properties Figure 7.22: Hysteresis curves of sample 1 ( ions cm 2 ) measured by XRMS at different temperatures with a fixed energy of 776 ev. Figure 7.23: Hysteresis curves of sample 2 ( ions cm 2 ) obtained by XRMS at different temperatures with a fixed photon energy, E=774.5 ev. 65

72 7 Dose Dependence of Magnetism in Co-implanted TiO 2 Rutile Figure 7.24: Hysteresis curves of sample 3 ( ions cm 2 ) taken at different temperatures using XRMS at fixed photon energy of 774 ev. 66

73 7.4 Magnetic Properties Post-annealing experiments Recently, Khaibullin et al. [147] reported that the annealing of Co-implanted TiO 2 plates up to 970 K destroys the magnetism in these samples. These samples were also prepared using ion implantation technique with the same parameters for the samples used in this study. The only difference is the orientation of the TiO 2 rutile plates. In this thesis, (100)-oriented TiO 2 rutile plates were used, whereas they did annealing experiments with the samples have (001)-orientation. Since the magnetic properties for both orientations are more or less the same [71], the results of Khaibullin et al. can also be used to understand the annealing effects on the samples used in this work. Fig shows the thermo-magnetic curves for TiO 2 rutile plates implanted with different doses. The sample with a dose of ions cm 2 shows the magnetic/non-magnetic transition temperature of about 850 K (curve 1). Two characteristic temperatures of magnetic ordering, T C1 700 K and T C2 850 K, were observed for the samples with intermediate doses, ( ions cm 2 (curve 2) and ions cm 2 (curve 3)). From these curves they concluded that two ferromagnetic phases, the low temperature and the high-temperature ones, coexist in these samples. The contribution to the magnetization from the high-temperature phase decreases gradually with the increasing cobalt implantation dose. Finally, for the sample with the highest dose of ions cm 2, the high-temperature phase practically disappears while the low-temperature one dominates (curve 5). Figure 7.25: Thermo-magnetic curves measured in field of 200 mt for TiO 2 rutile plates implanted by cobalt ions with different doses. Taken from Ref. [147]. They also studied the temperature stability of two magnetic phases in the sample implanted with the dose of ions cm 2. Fig shows the thermo-magnetic curves for this sample investigated by iteration of the thermal scanning with a gradual 67

74 7 Dose Dependence of Magnetism in Co-implanted TiO 2 Rutile Figure 7.26: Cyclic thermo-magnetic curves for TiO 2 rutile plate implanted with Co dose of ions cm 2. For details please see the text. After Ref. [147]. increase of the upper temperature limit. The temperature behavior of magnetization does not change if the temperature does not exceed 520 K during the first heating (solid circles). The low-temperature transition still exists during the second heating up to 720 K (open circles). This second heating probably destroys the low-temperature transition because it disappears during the third heating up to 970 K (solid squares) which shows only the contribution of the high-temperature magnetic phase. Finally, after repeating the heating in air up to 970 K (open squares) the magnetization completely vanishes. After 6 months, they annealed the same sample for 1 h in high vacuum of 10 6 Torr at temperature of 950 K and they observed that the low-temperature ferromagnetic phase in the Co-implanted rutile was partially recovered (solid triangles in Fig. 7.26). With these results, Khaibullin et al. [147] concluded that after annealing the sample up to 720 K in air, the oxygen vacancies disappear due to the oxygen diffusion into the sample from the atmosphere. Since these vacancies are thought to strongly stabilize ferromagnetism in this material [92], decreasing of these vacancies destroys the low-temperature ferromagnetic phase with T C1 700 K. When the sample were annealed again in high vacuum, new oxygen vacancies were formed and the low-temperature ferromagnetism was partially recovered in Co-implanted rutile. 68

75 7.5 Anomalous Hall effect measurements 7.5 Anomalous Hall effect measurements In magnetic materials, in addition to the ordinary Hall effect (OHE), there is an additional voltage proportional to the sample magnetization [158], the so-called anomalous Hall effect (AHE). Hence, the Hall voltage can be written as follows, V H = ( R0 I ) Hcosα + t ( RA µ 0 I t ) Mcosθ, (7.1) where t is the film thickness and I is the current. R 0 and R A are the ordinary and anomalous Hall effect coefficients, respectively. µ 0 is the permeability of free space. α is the angle between the applied magnetic field (H) and the normal to the sample. θ is the angle between the sample magnetization (M) and the normal to the sample. The first term in Eq. 7.1 is the ordinary Hall effect and arises from the Lorentz force acting on conduction electrons. This establishes an electric field perpendicular to the applied magnetic field and to the current. The second term is the anomalous Hall effect and it is conventionally attributed to spin dependent scattering mechanism involving a spin-orbit interaction between the conduction electrons and the magnetic moments of the material. At low applied magnetic fields, the Hall voltage (V H ) is dominated by the magnetic field dependence of the sample magnetization M. When the applied magnetic field is high enough to saturate the sample magnetization, the magnetic field dependence of the Hall voltage becomes linear due to the ordinary Hall effect. The observation of the anomalous Hall effect is suggested to be one of the important criteria for DMS materials to be intrinsic [11, 18]. In the past, several groups reported the AHE in highly reduced TiO 2 films doped with either Co or Fe, with an expectation about the possibility of intrinsic ferromagnetism therein [159, 134, 132]. However, recently, Shinde et al. [160] reported the co-occurrence of superparamagnetism and AHE in highly reduced Co-doped TiO 2 rutile films, raising questions about the use of the AHE as a test of the intrinsic nature of ferromagnetism in DMS materials without a detailed characterization of the sample. The AHE measurements 5 in this study are carried out at 4.2 K using a van der Pauw configuration presented in Fig In spite of the fact that the structural and magnetization measurements indicate the presence of magnetic nanoparticles in the Co-implanted TiO 2 films, the anomalous Hall effect is observed for these samples. The AHE data of sample 6 is shown in Fig A rapid increase in the Hall voltage at low field can be interpreted as an AHE which is followed by a slow decrease corresponding to the ordinary Hall effect. It is important to note that the negative slope of the high field data indicates n-type carriers in Co-implanted TiO 2 rutile. 5 These measurements were recorded in the laboratory of Prof. A. Wieck (Institut für Angewandte Festkörperphysik, Ruhr-Universität Bochum) within the scientific collaboration of IMPRS-SurMat. 69

76 7 Dose Dependence of Magnetism in Co-implanted TiO 2 Rutile Figure 7.27: Geometry of the AHE measurements. H is the external magnetic field applied perpendicular to the film surface. Figure 7.28: AHE data of sample 6 taken at 4.2 K. 70

77 7.6 Discussion 7.6 Discussion The origin of the observed two magnetic phases (ferromagnetism and superparamagnetism) in Co-implanted TiO 2 rutile films is attributed to the formation of two cobalt enriched layers with different cobalt concentrations and the valence states of the cobalt dopant. The TEM images (Fig. 7.8) clearly show that nanosize magnetic particles of cobalt metal nucleate in the surface region of the implanted rutile where the cobalt concentration is maximal (see RBS data in Fig. 7.3). Mostly beneath this layer, in the tail part of the depth profile, the implanted cobalt can exist in an ionic state substituting the Ti 4+ ions in the matrix in a form of Co 2+ ions. Figure 7.29: The magnetic phases at room temperature in Co-implanted TiO 2 rutile samples as a function of cobalt ion implantation dose. Fig presents the dose dependence of magnetic phases in Co-implanted TiO 2 rutile plates. At the lowest dose ( ions cm 2 ) the magnetic contribution from the metallic cobalt clusters and the substituted cobalt ions is very small. Hence, this sample gives very weak MOKE signal at room temperature. On the other hand, increasing of cobalt implantation dose leads to both increasing the size of clusters and the accumulation substituted cobalt in the state of paramagnetic Co 2+ ions. For this reason at the cobalt doses of ions cm 2 a superparamagnetic behavior is observed with a blocking temperature increases with increasing implantation dose. At certain concentration the substituted cobalt ions start to interact and these magnetically ordered regions produce ferromagnetism at room temperature in sample 4 ( ions cm 2 ) and sample 5 ( ions cm 2 ). At doses higher than ions cm 2 a strong 71

78 7 Dose Dependence of Magnetism in Co-implanted TiO 2 Rutile ferromagnetic order is formed due to the ion accumulation and indirect exchange interaction between the Co 2+ ions. However, small peaking at ZFC curves of these samples ( ions cm 2 ) and two component hysteresis of sample 6 (Fig. 7.15) show that the superparamagnetic phase is still present in these samples. The interaction mechanism of substituted cobalt ions which leads to ferromagnetism is also very important argument of this study. Since the XAS spectra clearly show the multiplet structure of the Co L 3 peak (see Fig. 7.14), it is for sure that some portion of the implanted cobalt ions in TiO 2 rutile are in Co 2+ oxidation state. When the TiO 2 is doped with cobalt ions, simultaneously many oxygen vacancies are also expected to be produced [100]. The observation of the AHE in the Co-implanted rutile samples give clear evidence for oxygen vacancies which contribute to shallow donor states in TiO 2 and increase the carrier density [95]. As it is discussed in Section 3.6 these oxygen vacancies strongly promote ferromagnetism in Co-implanted TiO 2 films by an indirect exchange of substituted cobalt ions through electrons trapped by neighboring oxygen vacancies. The origin of oxygen spin polarization in Co-implanted TiO 2 rutile samples can be due to the polarization of oxygen atoms close to the cobalt atoms in the host material. The shape of the hysteresis curve measured at the O K edge is the same as the one recorded at the Co L 3 edge supports this explanation. Another important result of this study is the observation of the AHE effect. The AHE is often taken as evidence that the carriers are polarized and that the material exhibiting it is a true DMS. However, after observation of co-occurrence of superparamagnetism and AHE in Co-doped TiO 2 films by Shinde et al. [160] and also in this study, the AHE can be thought as a necessary measurement but it is not sufficient by itself to claim the intrinsic nature of ferromagnetism in a DMS material. 7.7 Summary and conclusions In conclusion, the structural, magnetic and electronic properties of Co-implanted TiO 2 rutile films have been investigated in detail for different implantation doses. The structural data clearly show that the cobalt clusters are present in the samples after high dose cobalt ion implantation. In addition to the cluster formation, substitution of cobalt ions into the rutile lattice is also confirmed by XAS experiments. The origin of observed magnetism in the samples is explained by the coexistence of two different magnetic phases. Cobalt nanoparticles in the surface layer form the superparamagnetic phase in the samples implanted with low and intermediate doses. The Co 2+ ions substituting the host Ti 4+ ions leads to ferromagnetism as a second magnetic phase. The oxygen vacancies formed by ion implantation provide charge compensation and serve as mediators for the exchange interaction between the Co 2+ ions in high dose doped samples. The observation of the AHE effect in Co-implanted TiO 2 rutile can also be thought as an important indicator for the observed long range ordered ferromagnetism. At the highest dose, a strong ferromagnetic order is developed by ion accumulation and indirect exchange interaction between the Co 2+ ions. Thus, an infinite ferromagnetic layer is formed with a strong in-plane magnetic anisotropy following the in-plane anisotropy of the rutile host material. 72

79 8 Room Temperature Ferromagnetism in Co-implanted n-type ZnO Thin Films 8.1 Overview ZnO is a II-VI semiconductor with a wide band gap of about 3.4 ev. The stable crystal structure of ZnO is the wurtzite structure (hexagonal, with a = 3.25 Å and c = 5.12 Å) [161], in which each atom of zinc is surrounded by four oxygen atoms in tetrahedral coordination. The transition metal doped ZnO is interesting from the view point of forming a transparent ferromagnetic material and it has the potential to be a highly multifunctional material with coexistence of magnetic, semiconducting and optical properties. The theoretical predictions of room temperature ferromagnetism in TM-doped ZnO [47, 48, 49] have initiated a number of experimental works on TM-doped ZnO as a potential oxidebased DMS material. The first observation of ferromagnetism in Co-doped ZnO was reported by Ueda et al. [50]. They prepared Zn 1 x Co x O thin films on sapphire substrates using PLD technique with x varies between 0.05 and Following these initial theoretical and experimental reports, different growth methods have been used to deposit Co:ZnO films, including radio-frequency (RF) magnetron co-sputtering [162], pulsed laser deposition (PLD) using a KrF laser [163, 164, 165, 166, 167, 168, 169], combinatorial laser molecular beam epitaxy (LMBE) [170, 171], sol-gel method [99], and as well as ion implantation [61]. The sapphire was widely used as a substrate due to the lower mismatch (2%) between the film and the substrate. In addition to the cobalt, different 3d transition elements have also been used for doping, including Mn [172, 50, 170, 173], Ni [50, 170, 167], V [170, 167, 174], Cr [50, 167], and also Fe [50, 170, 175, 167]. Various solubility limit for Co in ZnO was reported by different groups. Prellier et al. [166] have found that the solubility limit is about 10% in PLD-grown films. Park et al. [176] reported that the cobalt nanoclusters start to form for x 12 at.% in samples grown by sol-gel and RF sputtering techniques. Lee et al. [99] observed some undefined peaks for a cobalt content higher than 25%. Kim et al. [164] showed that the solubility limit to be less than 40% in PLD-grown films. Ueda et al. [50] claimed that the solubility limit is lower than 50% and they clearly observed phase separations into the ZnO and CoO in the film prepared using Zn 0.5 Co 0.5 O targets. This controversy between different research groups seems to result from the growth technique used and/or from the growth conditions such as oxygen pressure and deposition temperature. 73

80 8 Room Temperature Ferromagnetism in Co-implanted n-type ZnO Thin Films Regarding the magnetic properties of the Co:ZnO films, while some groups observed room temperature ferromagnetism for 25% [99] and lower [166, 167, 169, 177] Co concentrations, others reported the absence of ferromagnetism [170, 164, 176]. The review articles on TMdoped ZnO [52, 51, 54, 57, 56] can be seen for the detailed information. In spite of some research groups reported ferromagnetism above room temperature, the nature of magnetic interactions in Co- and other TM-doped ZnO film remains unanswered. In this chapter 1 the structural, magnetic and electronic properties of Co-implanted n-type ZnO films grown on sapphire substrates have been studied in detail by various experimental techniques. RBS, XRD and high resolution TEM techniques were used to determine the depth distribution of implanted cobalt ions and to detect the formation of possible secondary phases such as metallic cobalt clusters in the implanted samples. To determine whether the implanted cobalt ions are in the Co 2+ oxidation state or exhibit pure metallic behavior, XAS experiments were also performed. The magnetic properties of the films were characterized by using MOKE, SQUID as well as XRMS techniques. To determine the carrier type in Co-implanted ZnO films, anomalous Hall effect measurements are also performed. The x-ray diffraction and transmission electron microscopy results show the presence of a (1010)-oriented hexagonal Co phase in the sapphire substrate, but not in the ZnO film. The diameter of the Co clusters are about 5-6 nm, forming a Co rich layer in the substrate close to the ZnO/Al 2 O 3 interface. However, the multiplet structure of the XAS spectra around the Co L 3 edge indicates that the implanted cobalt ions are in the Co 2+ state in the ZnO film. Magnetization measurements show that there are two magnetic phases in the implanted region. One is the intrinsic room temperature ferromagnetism due to the Co substitution on Zn sites in the ZnO layer and the second magnetic phase originates from Co clusters in the sapphire substrate. Magnetic dichroism at the O K edge and the AHE are also observed in Co-implanted ZnO films, supporting the intrinsic nature of the observed ferromagnetism. 8.2 Sample Preparation About 350 Å thick ZnO films were grown on mm 2 epi-polished single-crystalline Al 2 O 3 (1120) substrates by RF (13.56 MHz) sputtering of a ZnO target [178]. The sputtering was carried out in an atmosphere of mbar pure Ar (99.999%) with a substrate temperature of 500 C. In order to increase the quality of ZnO films, we have carried out post-growth annealing in an oxygen atmosphere with a partial pressure of up to 2000 mbar and a temperature of 800 C. After annealing, the ZnO samples were implanted in the ILU-3 ion accelerator (Kazan Physical-Technical Institute of Russian Academy of Science) by using 40 kev Co + ions with an ion current density of 8µA cm 2. The sample holder was cooled by flowing water during the implantation to prevent the samples from overheating. The implantation dose varied in the range of ions cm 2. After implantation, the samples are cut into the square pieces and gold contacts are evaporated 1 Some parts of this chapter are submitted for publication in the article Room temperature ferromagnetism and n-type conductivity in Co-implanted ZnO film [61]. 74

81 8.2 Sample Preparation on the corners of the samples for AHE studies (Fig. 8.1). The list of the Co-implanted ZnO films is given in Table 8.1. Figure 8.1: Sample preparation stages for Co-implanted ZnO/Al 2 O 3 films. Table 8.1: List of the ZnO films implanted with 40 kev Co + for varying Co ion dose. Sample Dose ( ion cm 2 )

82 8 Room Temperature Ferromagnetism in Co-implanted n-type ZnO Thin Films 8.3 Structural Properties The depth dependence of the cobalt concentration in Co-implanted ZnO/Al 2 O 3 films was investigated using the RBS technique at the Dynamic Tandem Laboratory (DTL) at Ruhr-Universität Bochum. The RBS data shows both a maximum of cobalt concentration (about 50 at.%) located close to the ZnO/Al 2 O 3 interface and an extended inward tail due to cobalt diffusion into the volume of the Al 2 O 3 substrate (Fig. 8.2). It is also observed that after ion implantation the thickness of the ZnO layer has decreased from originally 35 nm to 28 nm. According to the SRIM algorithm 2, the average implanted depth of 40 kev Co ions in ZnO/Al 2 O 3 is about 20.4 nm with a straggling of 9.6 nm in the Gaussian-like depth distribution (Inset in Fig. 8.2). However, because of the surface sputtering, ion mixing and heating of the implanted region by the ion beam, there is a redistribution of the implanted cobalt compared to the calculated profile. Figure 8.2: Depth dependence of the cobalt concentration in ZnO/Al 2 O 3 implanted with Co ions with a dose of ions cm 2 (open circles) and ions cm 2 (closed squares), respectively. The inset shows the calculated SRIM profile without taking into account ion sputtering effects. The high-angle XRD experiments provide information on the structural coherence of the films and in our case also give us a chance to detect possible additional phases in the sample after ion implantation. Fig. 8.3 shows a high angle Bragg scans of the ZnO films before (a) and after (b) cobalt implantation. The data were taken using synchrotron radiation at the Hamburg Synchrotron Radiation Laboratory (HASYLAB) (Fig. 8.3(a)) and at the Dortmund Electron Accelerator (DELTA) (Fig. 8.3(b)) with an energy of E=11000 ev. 2 The Stopping and Range of Ions in Matter (SRIM) [149]. SRIM-2006 software at 76

83 8.3 Structural Properties Figure 8.3: High angle Bragg scans of the ZnO films before (a) and after (b) cobalt ion implantation. X-ray diffraction measurements yielded evidence for the (1010) reflection of the Co hcp structure as is clearly seen on the right side of the sapphire substrate peak (Fig. 8.3(b)). The heavy ion bombarding also causes a reduction of intensity of the ZnO (0001) peak and this intensity reduction increases by increasing cobalt concentration. After implantation a tail (shown by an arrow in Fig. 8.3) appears around the main peak of Al 2 O 3 (1120) reflection which is not observed before implantation. This tail likely reflects the lattice expansion of the sapphire substrate upon Co implantation. In order to further investigate both the presence of metallic cobalt clusters and the dam- 77

84 8 Room Temperature Ferromagnetism in Co-implanted n-type ZnO Thin Films Figure 8.4: TEM images of Co-implanted ZnO/Al 2 O 3 film (Sample 6). Cobalt clusters are clearly seen in Al 2 O 3 substrate. age of the sapphire substrate, high resolution cross sectional TEM measurements were performed. The sample preparation for TEM measurements is done by using focused ion beam (FIB) technique. To prevent charging effects, the sample surface was covered by tungsten (W) film and then very small cross sectional piece of implanted sample was cut by using FIB 3. Fig. 8.4 presents TEM images of sample 6 ( ions cm 2 ) with an increasing magnification from 20 nm to 2 nm. In the first image (Fig. 8.4(a)), a general overview of the ZnO/Al 2 O 3 sample is shown. The cobalt clusters can be seen in the sapphire substrate located close to the ZnO/Al 2 O 3 interface. The clusters most likely form because of a decomposition of Co in Al 2 O 3 substrate. Clustering occurs in a Al 2 O 3 at an annealing temperature of 900 C [179]. Obviously ion bombardment heats up the sample locally to this temperature. Fig. 8.4(b) and (c) focuses on the ZnO/Al 2 O 3 interface. These images reveal that the cobalt clusters have a size of about 5-6 nm and that they nearly touch each other. Further information from these images is the deformation of the Al 2 O 3 crystal structure close to the ZnO/Al 2 O 3 interface. This results in a lattice expansion of the substrate which is in agreement with the XRD results shown in Fig However, far from the interface the structure of Al 2 O 3 is preserved and one can see nicely the atomic rows of Al 2 O 3 presented in Fig. 8.4(c). Fig. 8.4(d) shows the ZnO layer with a magnification of 2 nm. Even after heavily ion bombarding, ZnO still has a good arrangement of atomic rows. Moreover, any distinct clusters can not be observed in this region. 3 The sample preparation using FIB and the TEM measurements were performed in the laboratory of Prof. G. Eggeler (Institut für Werkstoffe, Ruhr-Universität Bochum) within the scientific collaboration of IMPRS-SurMat. 78

85 8.4 Magnetic Properties 8.4 Magnetic Properties Room temperature magnetization measurements In this section the magnetization behavior of the Co-implanted ZnO films for different implantation doses are discussed. Fig. 8.5 shows the hysteresis loops of Co-implanted ZnO films which were recorded at room temperature using a high-resolution MOKE setup [152, 153, 154] in the longitudinal configuration with s-polarized light. The MOKE data in Fig. 8.5 clearly indicate that after cobalt implantation, non-magnetic ZnO becomes ferromagnetic at room temperature with a large remanent magnetization. With increasing cobalt concentration the implanted ZnO films exhibit sequentially paramagnetic, weak ferromagnetic and, finally, ferromagnetic response with a square-like hysteresis at room temperature for the dose of ions cm 2. For the highest dose ( ions cm 2 ) the square-like shape of the hysteresis disappears and the coercive field increases drastically. This indicate that at this dose cobalt atoms start to form clusters in the ZnO film. Moreover, no in-plane magnetic anisotropy is observed in Co-implanted ZnO films. Figure 8.5: Room temperature MOKE hysteresis curves of Co-implanted ZnO films measured for varying implantation dose. Since the MOKE technique is only sensitive to the magnetization of thin layers close to the surface (20-30 nm penetration depth), M H measurements have also been carried out 79

86 8 Room Temperature Ferromagnetism in Co-implanted n-type ZnO Thin Films Figure 8.6: The MOKE (a) and SQUID (b) hysteresis curves of sample 6 ( ions cm 2 ). using a Quantum Design MPMS XL SQUID magnetometer. In Fig. 8.6 the MOKE and SQUID hysteresis loops of sample 6 ( ions cm 2 ) are compared. Fig. 8.6(b) presents the SQUID hysteresis loop of this sample after subtraction of a diamagnetic contribution from the sapphire substrate. The coercive field of this hysteresis is more or less the same as measured by MOKE technique. However, some additional contributions appear and the magnetization saturates at considerably higher fields XRMS and XAS measurements To study in detail the observed ferromagnetic behavior, the magnetic properties of Coimplanted ZnO films were investigated using the XRMS and XAS techniques. XRMS has proven to be a highly effective method for the analysis of the magnetic properties of buried layers and interfaces, including their depth dependence [117, 180]. Moreover, if the photon energy is fixed close to the energy of the corresponding absorption edges, element specific hysteresis loops can be measured by varying the external magnetic field [181]. Since there are three elements in the Co-doped ZnO film, the analysis can be carried out separately for Co, O and Zn. The XRMS experiments were performed with the ALICE diffractometer [125] at the undulator beamline UE56/1-PGM at BESSY II (Berlin, Germany). The diffractometer comprises a two-circle goniometer and works in horizontal scattering geometry. A magnetic field can be applied in the scattering plane along the sample surface either parallel or antiparallel to the photon helicity, which corresponds to the longitudinal magneto-optical Kerr effect (L-MOKE) geometry. The maximum field of ±0.27 Tesla was high enough to fully saturate the sample. The magnetic contribution to the scattered intensity was always measured by switching the magnetic field at fixed photon helicity. Fig. 8.7 shows the specular reflectivities of sample 6 measured at the Co L 3 edge (E=780 ev) in magnetic saturation. The measurements were taken at room temperature and with 80

87 8.4 Magnetic Properties Figure 8.7: Reflectivity scans of sample 6 taken at the Co L 3 edge (E=780 ev) with a magnetic field applied in the sample plane parallel (I +, solid line) and antiparallel (I, open circles) to the photon helicity. Inset shows the asymmetry ratio (R) as a function of angle. a magnetic field applied in the sample plane parallel (I +, solid line) and antiparallel (I, open circles) to the photon helicity. Due to the high surface roughness no Kiessing fringes are observed in the reflectivity curves. In addition the splitting of the two curves is clearly seen in Fig The inset in Fig. 8.7 presents the angular dependence of the asymmetry ratio to show how the magnetic signal varies. As a compromise between high scattering intensity and high magnetic sensitivity for the investigation of the magnetic properties at the Co L edges, the scattering angle was fixed at the position of 2θ = 8.2 (shown by a black circle in the inset in Fig. 8.7). For measurements at the O K edge (E 530 ev) the scattering angle was fixed at 2θ = 12, which corresponds to the same scattering vector in the reciprocal space. The energy dependence of the intensity (XRMS) around the Co L 3,2 edges measured in positive (solid line) and negative (open circles) saturation fields is shown in Fig. 8.8 for sample 6. Since the magnetic contribution to the resonant scattering can best be visualized by plotting the asymmetry ratio, R = (I + I )/(I + + I ), in Fig. 8.9 the asymmetry ratios for samples doped with different dose are presented. The asymmetry ratio shows a strong ferromagnetic signal for sample 6 (up to 30 %) and it decreases by decreasing the cobalt implantation dose. The fine structure of the Co L 3 peak in Fig. 8.8 is typical for oxidized cobalt which is observed before for CoO by Regan et al. [155]. They also showed that in the case of metallic cobalt the Co L 3 peak consists of mainly a single component. 81

88 8 Room Temperature Ferromagnetism in Co-implanted n-type ZnO Thin Films Figure 8.8: Energy dependence of scattering intensities (XRMS) at the Co L edges for sample 6 measured at room temperature. Recently, Kobayashi et al. [182] reported that in the case of presence of Co 2+ state in Co-doped ZnO, XAS spectra exhibits a multiplet structure around the Co L 3 edge (see inset in Fig. 8.10). To check whether this behavior is also present in our samples, XAS experiments were carried out at the undulator beamline UE52-SGM at BESSY II using the ALICE diffractometer. The absorption data were taken by the total electron yield (TEY) method (see Sec. 5.2), i.e. by measuring the sample drain current. Since the excited electron trajectories are effected by the external magnetic field, the XAS spectra were taken with fixed photon helicity at remanence. The angle of incidence was chosen to be 4.1 with respect to the surface. The spectra were normalized to the incoming photon flux. Fig shows the averaged x-ray absorption spectra (σ + +σ )/2 at the Co L 2,3 edges. The XAS spectrum clearly shows a multiplet structure at the L 3 edge which is similar to that observed before for Co-doped ZnO by Kobayashi et al.. This multiplet structure is a clear sign of the presence of oxidized cobalt in this sample. The magnetic signal at the Zn L 3 (E 1020 ev) and the O K edges were also investigated. Within the sensitivity limit no magnetic signal could be recorded for Zn (Fig. 8.11). However, a clear magnetic signal was observed at the O K edge. The asymmetry ratio of sample 6 measured at the O K edge is presented in Fig Note that the maximum in the asymmetry ratio of oxygen is much smaller (by roughly a factor of twenty) than the asymmetry ratio of cobalt shown by closed symbols in Fig In addition to sample 6, magnetic signal at the O K edge could also be observed for the samples 4, 5 and 7. In Fig the magnetic hysteresis curves recorded at the Co L 3 (773.4 ev) and O K (526.8 ev) edges are compared. The shape and the coercive field of the hysteresis curves 82

89 8.4 Magnetic Properties Figure 8.9: The asymmetry ratios taken at the Co L edges for sample 6 ( ions cm 2 ) and sample 4 ( ions cm 2 ) shown by closed and open symbols, respectively. Inset presents the asymmetry ratio of sample 2 ( ions cm 2 ) measured at 4.2 K. are the same, but the intensity is much lower for the O K edge. This is a clear indication for a spin polarization of oxygen atoms in the ZnO host matrix. Since an additional magnetic contribution is observed in the SQUID hysteresis curve of sample 6 (Fig. 8.6(b)), it is checked by using XRMS technique whether this feature becomes also visible. For this reason, several hysteresis loops are recorded at different photon energies and a systematic change of the hysteresis loop shape is observed with changing photon energy. Here, the only three hysteresis curves shown by open red circles in Fig. 8.9 are presented. The shape of the hysteresis curve taken at ev (closed symbols in Fig. 8.13) is practically the same as the one measured by MOKE (Fig. 8.6(a)). However, when the incoming photon energy is increased to 781 ev and 783 ev, two different hysteresis loops are observed. Fig. 8.14(a) presents the hysteresis curve recorded at 781 ev. At this energy the hysteresis curve has two components with a small and a large coercive field and it is similar to the SQUID hysteresis (Fig. 8.6(b)). At higher energies the low coercive field component vanishes and at the energy of 783 ev (Fig. 8.14(b)) the hysteresis curve consists of practically only one component with the large coercive field. This large coercive field component originates from the strong interaction between cobalt clusters in the sapphire substrate and it is present even at room temperature. The reason for the observation of different hysteresis curves using XRMS technique can be interpreted by changing of the optical parameters as a function of incident photon energy. Depending on the energy deviation from the L 2,3 resonance condition, both types of magnetic hysteresis can be detected in sample 6. For an energy of ev, which is 83

90 8 Room Temperature Ferromagnetism in Co-implanted n-type ZnO Thin Films Figure 8.10: X-ray absorption spectra of sample 6 measured at the Co L 2,3 edges by using TEY method. σ + and σ denote the right and left circularly polarized light, respectively. Inset shows the experimental Co L 2,3 edge XAS spectrum (top) compared with atomic multiplet calculations done by Kobayashi et al. [182]. very close to the L 3 resonance energy, the contribution to the scattering intensity from the phase with the large coercive field vanishes, resulting in a hysteresis loop (closed symbols in Fig. 8.13) which is similar to the one measured by MOKE (Fig. 8.6(a)). However, at the photon energy of 781 ev, at the resonance condition, both phases contribute to the scattering intensity. The hysteresis taken at this energy (Fig. 8.14(a)) presents the superposition of two phases of cobalt in the host material and it is similar to the hysteresis measured with the SQUID magnetometer (Fig. 8.6(b)). The hysteresis loop recorded at the energy of 783 ev (Fig. 8.14(b)) is representative for the metallic phase of cobalt in Al 2 O 3 with a large coercive field, whereas at this energy the contribution to scattering intensity from the small coercive field component is nearly zero Temperature dependent magnetization measurements To check the temperature dependence of magnetization for ZnO films doped with different doses, field cooled (FC) and zero field cooled (ZFC) M T measurements are carried out using SQUID magnetometry. The measurement procedure for FC and ZFC curves is described in previous chapter in Sec Due to the clustering of cobalt in Al 2 O 3 substrate (see Fig. 8.4) the FC and ZFC curves presented in Fig always show the evidence for the presence of superparamagnetic phase. There is a small peak at about 20 K in ZFC curve of sample 1 ( ions cm 2 ) and this peak shifts to a higher 84

91 8.4 Magnetic Properties Figure 8.11: The XRMS data for sample 6 taken at the Zn L 3 edge. The inset shows the asymmetry ratio Figure 8.12: The asymmetry ratio for sample 6 taken at the O K edge. The inset shows the XRMS data at the O K edge. 85

92 8 Room Temperature Ferromagnetism in Co-implanted n-type ZnO Thin Films Figure 8.13: Hysteresis curves of sample 6 measured at the Co L 3 (closed symbols) and O K (open symbols) edges. Figure 8.14: (a) Hysteresis loop measured at the energy of 781 ev shows the superposition of two different phases of cobalt in the host material. (b) Hysteresis loop taken at the energy of 783 ev. temperatures with increasing cobalt concentration. The trend in the M T curve of sample 1 ( ions cm 2 ) can be attributed to the coexistence of the ferromagnetic phase originating from substituted Co 2+ ions in ZnO and the superparamagnetic phase due to the cluster formation in Al 2 O 3. The hysteresis curve of this sample measured at 5 K (Fig. 8.16) indicates that the superparamagnetic phase in this sample is much 86

93 8.4 Magnetic Properties more dominant. The M T measurements for the samples implanted with higher doses ( ions cm 2 ) shows the occurrence of superparamagnetism with a blocking temperature of about 100 K and 250 K for sample 2 ( ions cm 2 ) and sample 3 ( ions cm 2 ), respectively. The hysteresis curves of these films (Fig. 8.16) also show that the superparamagnetic phase is still dominating over the ferromagnetic phase. But the hysteresis curve of the sample 3 ( ions cm 2 ) indicates that the ferromagnetic phase starts to appear at this dose. The temperature dependent magnetization curves of sample 4 ( ions cm 2 ), sample 5 ( ions cm 2 ) and sample 6 ( ions cm 2 ) show that these samples have a blocking temperature of about 390 K or even more. In addition to this, the ferromagnetic phase is clearly seen in the SQUID hysteresis loops (Fig. 8.16) of these samples and it is present even above room temperature (see Fig. 8.5). Since the MOKE is only sensitive to the surface layers, the superparamagnetic phase in these sample can not be seen in MOKE results. Moreover, the SQUID hysteresis curves of these samples show a systematic decrease in H C of the ferromagnetic phase with increasing cobalt dose. This behavior can be explained as follows: As it is discussed in Chap. 3 the oxygen vacancies strongly promote the ferromagnetism in Co-doped ZnO. With increasing Co dose more oxygen vacancies are filled by Co atoms in the ZnO film. This decreases the concentration of pinning centers and reduces H C. Thus the ferromagnetic component decreases at same rate as the coercive field. The M T data (Fig. 8.15) and the room (Fig. 8.5) and low (Fig. 8.16) temperature hysteresis measurements of sample 7 ( ions cm 2 ) clearly indicate that the cobalt atoms start to cluster also within the ZnO layer at the highest dose. 87

94 8 Room Temperature Ferromagnetism in Co-implanted n-type ZnO Thin Films Figure 8.15: 88 Temperature dependent magnetization curves of Co-implanted ZnO films taken using SQUID magnetometry for varying implantation dose.

95 8.4 Magnetic Properties Figure 8.16: SQUID M H loops of Co-implanted ZnO films measured for different implantation doses at 5 K. 89

96 8 Room Temperature Ferromagnetism in Co-implanted n-type ZnO Thin Films 8.5 Anomalous Hall effect measurements As it is discussed in previous chapter, an additional term also contributes to Hall voltage in a ferromagnetic material due to its spontaneous magnetization. To check this behavior and to determine the majority carriers in Co-implanted ZnO films, AHE experiments were carried out 4. The AHE measurements are performed at 4.2 K using a van der Pauw configuration presented in Fig Figure 8.17: Geometry of the AHE measurements. H is the external magnetic field applied perpendicular to the film surface. Figure 8.18: AHE data of sample 6 ( ions cm 2 ) taken at 4.2 K. The AHE data of sample 6 ( ions cm 2 ) is shown in Fig A sharp rise in the Hall voltage at low field, i.e., AHE, is followed by a slow decrease corresponding 4 These measurements were recorded in the laboratory of Prof. A. Wieck (Institut für Angewandte Festkörperphysik, Ruhr-Universität Bochum) within the scientific collaboration of IMPRS-SurMat. 90

97 8.5 Anomalous Hall effect measurements to the ordinary Hall effect. It is important to note that the negative slope at high fields indicates n-type carriers in Co-implanted ZnO films. 91

98 8 Room Temperature Ferromagnetism in Co-implanted n-type ZnO Thin Films 8.6 Discussion In the literature, the reported highest solubility limit of cobalt ions in ZnO is less than 50 % using a pulsed-laser deposition (PLD) technique [50]. The measured cobalt concentration of at.% in ZnO in this study is rather high, such that the formation of cobalt clusters in ZnO should be expected. However, no clusters could be observed within the ZnO layer for sample 6. It seems that this is due to the properties of ion implantation, which allows doping of transition metals beyond their solubility limits [73]. Fig shows a possible scenario of the dose dependence of magnetic phases in ZnO films at room temperature. At low doses ( ions cm 2 ) the number of substituted cobalt ions in ZnO layer is very few which results in a paramagnetic signal at room temperature. On the other hand, increasing of cobalt implantation dose leads to increasing of the substituted cobalt ions and after certain cobalt concentration they start to interact ferromagnetically. For this reason at the cobalt dose of ions cm 2 a weak ferromagnetic behavior is observed with a T c below room temperature. At higher cobalt concentrations ( ions cm 2 ) the substituted cobalt ions in ZnO interact strongly and stabilize the room temperature ferromagnetism in Co-doped ZnO. At the highest dose of ions cm 2, in addition to the substituted cobalt ions, metallic cobalt clusters are also present in ZnO layer. Figure 8.19: The magnetic phases at room temperature in Co-implanted ZnO films as a function of cobalt ion implantation dose. The difference in the shape of the hysteresis curves obtained by MOKE and SQUID is attributed to the surface sensitivity of the MOKE technique with a maximum penetration 92

99 8.6 Discussion depth of about nm. On the other hand, the SQUID technique probes the whole volume of a sample. The ZnO films has a thickness of 35 nm before implantation. Because of the surface sputtering, the ZnO thickness decreases to about 28 nm after implantation. Thus MOKE provides information only from the ZnO layer, not from the sapphire substrate, i.e. MOKE is only sensitive to the ferromagnetic contribution from the ZnO layer. In this layer a small fraction of nonmagnetic ZnO atoms are replaced by magnetic Co ions, giving raise to the MOKE hysteresis. However, SQUID collects magnetic contributions from both the Co-implanted ZnO film and from the cobalt clusters in Al 2 O 3 (Fig. 8.20). Therefore, the difference between the MOKE and SQUID data appear as a result of the depth-dependent Co content in the implanted layer. Figure 8.20: The cluster formation in Al 2 O 3 substrate after cobalt ion implantation. Another important result of this study is the observation of oxygen spin polarization in Co-implanted ZnO films. Since the shape of the hysteresis curve measured at the O K edge (Fig. 8.13) is the same as the one recorded by MOKE (Fig. 8.6(a)), spin polarization of oxygen atoms in this sample can not be due to the cobalt clusters in the sapphire substrate. Otherwise, the hysteretic shape of the polarized oxygen should be similar to the hysteresis of metallic cobalt clusters in sapphire with a large coercive field. This shows that the oxygen atoms are polarized due to the spontaneous ferromagnetic order in the ZnO film. The main question that arises here is the mechanism which leads to the observed long range ferromagnetic ordering in Co-doped ZnO. Recently, Patterson [101] calculated the electronic structures of Co substituted for Zn in ZnO, for Zn and O vacancies, and for interstitial Zn in ZnO using the B3LYP hybrid density functional theory. He reported that the singly-positively charged O vacancy is the only defect in Co-doped ZnO which can mediate ferromagnetic exchange coupling of Co ions at intermediate range (just beyond near neighbor distances). The ground state configuration is one where the majority Co spins are parallel whereas the minority spins are parallel to each other and to the oxygen vacancy spin, so that there are exchange couplings between these three spins which lead to an overall ferromagnetic ground state of the Co ions (see Fig. 3.8 in Sec. 3.6). No substantial exchange coupling was found for the positively charged interstitial Zn defect which has also spin half. The exchange coupling mechanism explained by Patterson is essentially the same as the impurity band model of Coey et al. (see Sec. 3.6), in which the polarons bound to the oxygen vacancies mediate ferromagnetic coupling between Co ions. In order to have the magnetic moments of the Co ions align ferromagnetically, one mediating electron is required with an oppositely directed spin. The oxygen spin polarization has not explicitly been considered in the aforementioned band structure calculations and may be due to ferromagnetic splitting of nearest neighbor oxygen p-levels. 93