Exploring Spin in Novel Materials and Systems

Exploring spin in novel materials and systems

DISSERTATION

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University

Lei Fang

Graduate Program in Physics

The Ohio State University

2011

Dissertation Committee:

Professor Ezekiel Johnston-Halperin, Advisor

Professor Fengyuan Yang

Professor William Putikka

Professor Louis DiMauro

Lei Fang

2011

Abstract

The field of spintronics has attracted tremendous attention from researchers because of its overwhelming advantages over the traditional electronic devices such as: higher speed, higher integration density, less power consumption, true nonvolatility, tunable optoelectronics, potentials for quantum information technology, etc. The metal based spintronics has brought huge success in the commercial market from the application of GMR effect, such as the hard drive read head and MRAM. However, coupling the spin degree of freedom to the semiconductor material gives more opportunities for new performance and functionality, due to its particular properties of the bandgap engineering, the easy control of the carrier concentrations and transport properties through doping, applying gate voltage and band offsets, as well as the potential in optoelectronic applications and the already existing overwhelming dominance of the semiconductor industry.

There are active researches in all types of semiconductor materials for spintronics.

However, recent work has focused on structures which rely on either metallic spin injectors that are not well suited to multifunctional operation or magnetic semiconductors that do not function under ambient conditions. This places significant limitations on the potential utility of devices based on these principles. To realize multifunctionality in spintronics, one part of my work is focused on using an organic-based room temperature

ii magnetic semiconductor to enable multifunctional operation and fully functional semiconductor spintronic devices. The other approach to realize the multifuctionality is to develop a strong intrinsic multiferroic which could be potentially used in spintronics applications. By using magneto-optical kerr effect, the ferromagnetism is confirmed in the world’s strongest engineered multiferroic – strained EuTiO3 thin film. This is second part of my thesis work. Moreover, theory prediction points out that dimensionality of a system plays an important role in the mechanism of spin relaxation, for example, longer spin relaxation time is expected in quasi-1d structure. However, traditional optical probing of spin is not available in the 1d structure due to its polarization anisotropy. The third part of my work is controlling the polarization anisotropy of quasi-1d system through oxide coating it with media which has close dielectric constant. Consequently this approach promises optical pumping and probing spin dynamics in quasi-1d systems.

Exploring spin in these novel materials and systems gives potential for new applications in spintronics field as well as realizes the desirable multifunctionalities.

iii

Dedication

This document is dedicated to my husband Taeyoung Choi and my parents.

Acknowledgments

I would like to thank my advisor Ezekiel Johnston-Halperin for guidance and support during the time I have worked for him. His broad and deep understanding of physics, as well as enthusiasm, has inspired me to grow as a young scientist.

I would also like to recognize all of my colleagues in my lab that have helped me do my research. Yi-Hsin Chiu, Dongkyun Ko, Howard Yu, Yong Pu, Ke Li and Patrick Truitt for the support in research. I would also like to thank the undergraduates that have helped on the project over the years, Bryce Smith, Scott Stowe, Cole Robinette your willingness to work hard was a great help. Thanks to all the collaborators in our department from different research groups, Prof. Yang, Xianwei Zhao, Prof. Epstein,

Deniz, K. Bozdag, Chia-Yi Chen, Prof. Hammel and Jay Jung. It was great pleasure to work with you for all my three projects. Thanks to all the others in my lab who gave me the support to make it through it all.

Thank you to Prof. Darrell Schlom, June Lee from Cornell. The cross school collaborating has brought great outcome in terms of the science discovery and publication.

I would like to thank the physics department’s support staff especially Ellen Copeland,

Kris Dunlap, Melanie Holbert and JD Wear for all their help. I would also like to thank everyone in the machine shop for their help in building and training me to build the physical setup for the projects.

To my good friends, Sarah Parks, Mark Murphuy, Charles Ruggiero, Jaclyn Kurash, Jing

Wang, Donghui Quan, thank you for all the support you have given me through the years and help through the tough decisions I needed make in grad school. Your support is part of why I made it here.

To my parents, you have always provided me with the support, insight and unconditional love I have needed. Thank you for your understanding and encouragement.

Finally, to my husband, Taeyoung Choi, thank you for the support you have given me.

You have helped me through the difficult periods here and I am forever grateful. I love you.

Vita

June 2000 ...... Tongling No.1 High School, Tongling,

China

June 2004 ...... B.S. Physics University of Science and

Technology of China.

2004 to present ...... Graduate Teaching and Research Associate,

Department of Physics, The Ohio State

University

Publications

3. “A strong ferroelectric ferromagnet created by means of spin-lattice coupling”

June Hyuk Lee, Lei Fang, Eftihia Vlahos, Xianglin Ke et al. Nature, 466, 954-959, 2010

2. “Process-dependent defects in Si/HfO2/Mo gate oxide heterostructures”

Shawn Walsh, Lei Fang, J. K. Schaeffer, E. Weisbrod and L. J. Brillson, Appl. Phys.

Lett. 90, 052901 (2007)

vii

1. “Process-dependent electronic states at Mo/hafnium oxide/Si interface”

Shawn Walsh, Lei Fang, J. K. Schaeffer and L. J. Brillson, J. Vac. Sci. Technol., A. 25,

1261 (2007)

Fields of Study

Major Field: Physics

viii

Table of Contents

Abstract ...... ii

Dedication ...... iv

Acknowledgments...... v

Vita ...... vii

List of Tables ...... xiv

List of Figures ...... xv

Chapter 1 Introduction-Spintronics...... 1

1.1 Overview-Introduction to spintronics ...... 1

1.2 Spin injection...... 4

1.2.1 Electrical spin injection ...... 4

1.2.2 Optical spin injection ...... 6

1.3 Spin relaxation...... 6

1.4 Spin detection ...... 9

1.4.1 Electrical spin detection ...... 9

1.4.2 Optical spin detection ...... 11 ix

1.5 Spintronics in novel material and systems-focus of my research ...... 11

1.6 Reference ...... 14

Chapter 2: Electrical spin injection from an organic-based magnet in a hybrid organic/inorganic heterostructure ...... 16

2.1 Introduction ...... 16

2.2 Theory and background ...... 17

2.2.1 Magnetic ordering of V[TCNE]x ...... 17

2.2.2 Spin-LED detection methods-optical transition ...... 20

2.2.3 Previous work on spin injection from ferromagnets into a semiconductor ...... 24

2.3 Fabrication of devices and principle of operation ...... 30

2.3.1 Spin-LED device fabrication with V[TCNE]x~2 ...... 30

2.3.2 Principle of spin-LED device ...... 33

2.4 Measurement setup ...... 34

2.4.1 Custom designed sample mount ...... 34

2.4.2 Transport measurements ...... 36

2.4.3 Polarization resolved electroluminescence measurement ...... 36

2.4.4 Magnetic circular dichroism measurement ...... 38

2.5 Experiment results ...... 39

2.5.1 Spin-LED function: I-V and MR results ...... 39

2.5.2 Electroluminescence from full-spin-LED device and control measurements .. 42

2.5.3 Polarization analysis ...... 45

2.6 Conclusions and outlook...... 51

2.7 Reference ...... 53

Chapter 3: Polarization control of semiconductor nanowires ...... 55

3.1 Introduction ...... 55

3.2 Theory and background ...... 56

3.2.1 Spin relaxation in quasi 1-D semiconductors ...... 56

3.2.2 Polarization anisotropy in quasi 1-D system ...... 60

3.2.3 Optical pumping in quasi 1-D system ...... 64

3.3 Growth of semiconductor nanowires ...... 70

3.3.1 Growth mechanism – The vapor-liquid-solid method ...... 70

3.3.2 Pulsed laser deposition of InP nanowires ...... 72

3.3.3 Chemical vapor deposition of ZnO nanowires ...... 73

3.3.4 Index matching with oxide sputtering ...... 74

3.3.5 Structure synthesis of InP and ZnO nanowires ...... 75

3.4 Polarization resolved photoluminescence measurement for NWs ...... 77

3.4.1 Measurement set up ...... 77

3.4.2 PL results for bare NWs and oxide coated NWs ...... 78

3.4.3 Polarization analysis ...... 81

3.5 Hanle measurements for nanowires ...... 86

3.5.1 Measurement set up and theory ...... 87

3.5.2 Calibration Measurement on GaAs epilayer ...... 89

3.5.3 Results on nanowire and outlook ...... 91

3.6 Conclusions and outlook ...... 93

3.7 Reference:...... 95

Chapter 4: MOKE measurement of ferromagnetism in a strong multiferroic created by means of spin-lattice coupling ...... 98

4.1 Introduction ...... 98

4.2 Theory and background ...... 100

4.2.1 Magnetic and electrical phase control in epitaxial EuTiO3 film ...... 100

4.2.2 From compressive to biaxial strain ...... 102

4.3 Initial results from collaborators ...... 103

4.3.1 Growth of epitaxial EuTiO3 in MBE ...... 103

4.3.2 Structure characterization of thin films ...... 105

4.3.3 SHG measurement for ferroelectricity ...... 107

4.4 Measurement of ferromagnetism ...... 109

4.4.1 Motivation for using MOKE ...... 109

xii

4.4.2 MOKE measurement set up ...... 110

4.4.3 Ferromagnetism at 2 K ...... 115

4.4.4 Temperature dependence of ferromagnetism ...... 116

4.4.5 Quantifying MOKE with SQUID ...... 118

4.4.6 MOKE with Faraday set up ...... 120

4.5 Conclusions ...... 122

4.6 Reference:...... 123

Appendix A: Two important aspects in MOKE measurement for EuTiO3 ...... 125

A.1 Design sample mount for in-plane magnetization ...... 125

A.2 Controlling temperature in spectromag with superfluid He ...... 126

All Reference ...... 129

xiii

List of Tables

Table 2.1. Wavefunction at k=0 of conduction band and HH and LH of valence band .. 21

xiv

List of Figures

Figure 1.1. Image of read head and MRAM devices ...... 2

Figure 1.2 Scheme of Datta-Das spin FET ...... 3

Figure 1.3 Schematic illustrating T1 and T2 ...... 7

Figure 1.4 Electrical spin injection and detection in Fe/GaAs ...... 10

Figure 2.1. Chemical structure and spin density distribution of TCNE molecule ...... 18

Figure 2.2 Schematic energy diagram for V[TCNE]x~2 ...... 19

Figure 2.3 E-k diagram of bulk GaAs near k=0...... 20

Figure 2.4 Allowed optical transitions for bulk GaAs and QW with probalities ...... 23

Figure 2.5 GaMnAs/InGaAs QW Spin-LED device and its EL spectrum ...... 25

Figure 2.6 Hysteretic EL polarization at different temperatures up to Tc ...... 26

Figure 2.7 Control measurements of GaMnAs/InGaAs QW spin injection ...... 27

Figure 2.8 Circular polarization as a function of field for Fe-LED ...... 29

Figure 2.9 HH and LH circular polarization for Fe-LED ...... 30

Figure 2.10 Schematics of cross section of V[TCNE]x~2 spin-LED device ...... 32

Figure 2.11 Band diagram of V[TCNE]x~2 spin-LED device ...... 33

Figure 2.12 Schematic showing opposite signal from HH and LH EL polarization ...... 34

Figure 2.13 Picture of custom-designed sample mount with device on it ...... 35

Figure 2.14 Measurement set up for EL polarization ...... 38

Figure 2.15 I-V characteristics for V[TCNE]x~2 and bare LED device ...... 40

Figure 2.16 Magntoresistance of V[TCNE]x~2 spin-LED and bare devices at T = 100K . 41

Figure 2.17 Temprature dependence of electrical characterization of spin-LED device 42

Figure 2.18 Optical characterisics of spin-LED device ...... 43

Figure 2.19 Raw polarization for spin-LED and control devices ...... 45

Figure 2.20 EL polarization of HH and LH for spin-LED device comparing to SQUID . 46

Figure 2.21 Schematic showing polarization analysis ...... 47

Figure 2.22 EL polarization of HH and LH at lower bias and higher temperature ...... 49

Figure 2.23 Pump power dependence of CP for photoluminescence ...... 50

Figure 3.1 Spin relaxation time as a function of channel width ...... 59

Figure 3.2 Photoluminescence of a single InP nanowire parallel and perpendicular to the

nanowire axis ...... 61

Figure 3.3 Normalized PL intensity as a function of polarization angle for a single WZ

and ZB InP nanowire ...... 63

Figure 3.4 Band structure and selection rules for ZB and WZ InP nanowire ...... 64

Figure 3.5 Diagram showing circular polarization as two linear polarization combination

...... 65

Figure 3.6 Diagram of light absorption inside a nanowire ...... 66

Figure 3.7 Diagram of index matching nanowire ensemble with oxide material ...... 68

Figure 3.8 Schematic showing GaAs nanowire growth and phase diagram...... 70

Figure 3.9 Schematic of pulsed laser deposition system ...... 73

xvi

Figure 3.10 Setup of UHV sputtering system ...... 74

Figure 3.11 SEM images of as-grwon InP nanowire ensembles ...... 75

Figure 3.12 TEM images of as-grown InP nanowire ...... 76

Figure 3.13 SEM images of as-grown ZnO nanowire ...... 77

Figure 3.14 Measurement set up for linear polarization resolved photoluminescence..... 78

Figure 3.15 PL comparison of as-grown and oxide coated InP nanowire ...... 79

Figure 3.16 PL comparison of as-grown and oxide coated ZnO nanowire ...... 81

Figure 3.17 Schematic showing polarizaiton analysis of nanowire ensembles ...... 82

Figure 3.18 Polarizaiton analysis for as-grown and oxide coated nanowires ...... 85

Figure 3.19 Measurement set up for photoluminescence Hanle measurement ...... 87

Figure 3.20 PL spectrum and circular polarization of GaAs epilayer ...... 90

Figure 3.21 Hanle curve of GaAs epilayer ...... 91

Figure 3.22 PL spectrum and circular polarization of as-grown InP nanowire ...... 92

Figure 3.23 PL spectrum and circular polarization of oxide coated InP nanowire ...... 93

Figure 4.1 Calculation of ε33 and polarization as a function of strain ...... 101

Figure 4.2 Phase diagram of EuTiO3 as a function of strain ...... 102

Figure 4.3 Phase diagram of EuTiO3 - bulk and film structure ...... 103

Figure 4.4 Schematic of a MBE system...... 104

Figure 4.5 θ-2θ XRD and rocking curve of EuTiO3 films grown on different substrates

...... 106

Figure 4.6 STEM image of EuTiO3 grown on DyScO3 ...... 107

Figure 4.7 SHG measurement of EuTiO3 films ...... 108

xvii

Figure 4.8 Temperature dependence of dc suscepbility of DyScO3 and its transmission

curve ...... 110

Figure 4.9 Different geomery of MOKE set up ...... 111

Figure 4.10 Sample mount of MOKE and schematics of measurement ...... 112

Figure 4.11 MOKE measurement setup outside of crystat ...... 114

Figure 4.12 Analysis process of MOKE data ...... 115

Figure 4.13 MOKE results of EuTiO3 on different substrates at T = 2K ...... 116

Figure 4.14 Temperature dependent hysterisis loop of EuTiO3/DyScO3 ...... 117

Figure 4.15 Temperature dependence of the remanent kerr angle and coercivity for

EuTiO3/DyScO3 ...... 118

Figure 4.16 SQUID measurements of EuTiO3/DyScO3 ...... 120

Figure 4.17 MOKE measurements in Faraday geometry ...... 121

Figure A.1 Schematic showing different geometry of the actual MOKE measurement 126

Figure A.2 Picture of the sample mount for MOKE ...... 128

xviii

Chapter 1 Introduction-Spintronics

1.1 Overview-Introduction to spintronics

Ever since 1988 when the giant magnetoresistence effect (GMR) was discovered[1-2], the field of spintronics has attracted tremendous attention from researchers and has brought revolution in both the technique aspect and fundamental understandings of nature. Spintronics is short for spin electronics, it is using the carriers’ spin degree of freedom in electronic devices to realize their performance and functionality. Spintronics has many advantages over the traditional electronic devices such as: higher speed, higher integration density, less power consumption, true nonvolatility, tunable optoelectronics, potentials for quantum information technology, etc.

So far, all metal spintronic structures has gained lots of commercial success. For example, the magnetic read heads in hard drives[3] and nonvolatile magnetic random access memory (MRAM) are realized based on GMR effect. Figure 1.1 lists the commercialized device applications. Meanwhile, semiconductor based spintronics offers benefits which are not included in all metal structures and therefore it is a very active research topic in multidisciplinary fields. The benefits come from the special characteristics only in semiconductors: its bandgap which can be tuned over a significant range in ternary compounds; the easy control of the carrier concentrations and transport

1 properties through doping, applying gate voltage and band offsets; the optical properties which could be used in optoelectronic applications; and the already existing overwhelming dominance of the semiconductor industry. Coupling the spin degree of freedom to the traditional semiconductor electronics gives opportunities for its new performance and functionality.

Fig. 1.1 (a) A single channel GMR isolator composed of a driver chip and a receiver chip in an 8 pin.

[4] (b) Schematic diagram of MRAM produced by freescale [5].

There are number of spin dependent device structures proposed by researchers.

One of the earliest and generic devices is the Datta-Das spin field effect transistor (spin-

FET)[6], as Fig. 1.2 is showing. Different from a usual FET, where there is a source, a narrow channel, a drain and a gate for controlling to flow (ON) or not (OFF), in a spin-

FET device, ferromagnetic material is used as the source and drain contact. The electrons with same polarization are injected at the source and are transported ballistically through the channel. When they arrive at the drain, they can either enter the drain (ON) if their spin polarization is same as the drain or they are scattered away (OFF) from the drain if their spin polarization are different. The use of the gate is to generate an effective

2 magnetic field through the spin-orbit coupling in the substrate material to control the spin polarization as they are transporting in the channel. In this way, by changing the gate, the spin transport is controlled and their spin polarization at the drain is determined. As a result, the conductance of the device is controlled. The concept of this device has triggered tremendous research both in theory and experiment.

Fig 1.2 Scheme of the Datta-Das spin FET. The source and drain are ferromagnetic, which controls the magnetization of source and drain. The gate voltage controls the spin precession in the narrow channel to determine spin polarization at the drain[7].

Other than the Datta-Das spin-FET, varieties of other semiconductor spintronic devices have been developed over last two decades. All these devices have to satisfy four essential requirements: 1) Efficient generation of spin (or spin injection) into the semiconductor material. 2) The spin polarization has to transport efficiently and have long spin lifetime so they still survive when they reach the detection point. 3) Effective manipulation of the spin carriers to provide the desired functionality and 4) Effective way to detect spin polarization. These aspects will be discussed in the chapter.

1.2 Spin injection

In a non-magnetic semiconductor, there are equal numbers of spin-up and spin- down carriers when it is in equilibrium and it consists of two independent spin channels with the same conductivity. Since it is impossible to separately contact these two channels electrically, the only way to generate spin polarized current is to inject spin polarized carriers inside a semiconductor, which could be realized either electrically or optically.

1.2.1 Electrical spin injection

In electrical spin injection a magnetic electrode is connected to the sample. When the current drives spin-polarized electron from the electrode to the sample, a net spin polarization accumulates at the sample. In the pursuit of realizing spin injection from ferromagnetic metal to semiconductors, ohmic contact was first tried as it was believed that with a small interface resistance, the spin polarization can be transferred from a ferromagnetic metal into a semiconductor. However, typical metal-semiconductor ohmic contacts result from heavily doping the semiconductor surface, leading to spin-flip scattering and loss of the spin polarization[4]. Despite much effort, no effective spin injection was demonstrated in the 1990’s [4]. Later, Schmidt et al, pointed out that the conductivity mismatch between the metal and semiconductor was an intrinsic problem for injecting spins from metal into semiconductors [8]. Schmidt et al. derived that the spin injection efficiency is proportional to the ratio between the conductivity of the semiconductors and FM σsc / σfm [8]. While this is not a problem for nonmagnetic metal

4 as σfm ~ σnm, for semiconductors since σfm >> σsc, the spin injection efficiency is very low.

To solve the conductivity mismatch problem in spin injection between metal and semiconductors, injection through a spin-dependent tunnel barrier between a FM and a semiconductor was proposed by Rashba [9]. When the impedance of the barrier at the interface is sufficiently high, most of the voltage drop happens at the interface, which is determined by the (spin dependent) density of the electronic states of the two electrodes that are involved in the tunneling process. The current through the barrier is then sufficiently small so that the two electrodes remain in equilibrium and the relative (spin- dependent) conductivities of the electrodes play no substantial role in spin-dependent transport across the interface [4]. After this discovery, higher and higher efficient electrical spin injection from metal to semiconductor has been realized in different systems. Either a metal-semiconductor schottky tunnel barrier [10-12] or a metal- insulator-semiconductor could be used [13-14].

Besides the use of tunnel barriers, there are several other ways to realize the spin injection in the FM/SM system. The most straightforward way is to use a ferromagnetic semiconductor as there is no conductivity mismatch problem between two semiconductors. In fact, spin injection from a magnetic semiconductor into a semiconductor heterostructure had already been demonstrated successfully before the conductivity mismatch problem was found [15-16]. Half-metallic ferromagnet whose spin polarization is 1 is actually ideal as the most efficient spin injector. However in reality obtaining half-metallic spintronic behavior is fraught with different problems [4].

For the use of ferromagnetic semiconductors, since their Curie temperatures (Tc) are much lower than room temperature (highest reported is below ~170K), which is not practical for spintronic device. Researchers in many different areas are trying to find materials which are suitable for ambient conditions. The use of an organic ferrimagnet with Tc higher than room temperature as a spin injector is an alternate way for realizing it and it will introduced in details in Chapter 2.

1.2.2 Optical spin injection

A non-equilibrium spin polarization could also be realized through optical pumping [17]. The angular momentum of absorbed circularly polarized light is transferred to the orbital angular momentum of the electrons. The electron spins are then polarized through the spin orbit coupling. The details of the optical pumping and detection will be introduced later in Chapter 2 and 3.

1.3 Spin relaxation

After spin is injected into the material, spin relaxation and dephasing will happen.

Constants T1 and T2 are used to describe the process, as illustrated in Fig. 1.3. T1 is the spin lattice relaxation time. It is the typical time in which the longitudinal magnetization reaches the equilibrium. Equivalently it is the time of spin population to reach thermal equilibrium with the lattice. In T1 process the energy that is reduced from the transition for the spin system needs to go somewhere, namely other repositories for thermal energy, such as translations, rotations and vibrations, collectively called the lattice.

Fig. 1.3 (a) The decay of a nonequilibrium occupation (spin relaxation), characterized by the time T1.

Spin up and down eigenstates are shown split by a magnetic field. (b) Phase coherence between two states in indicated by the shaded circle. The decay of phase coherence, characterized by T2, does not involve a change in occupation or a change in the energy of the population. [19]

The other time scale T2 or T2* is the spin-spin relaxation time, or transverse relaxation time. The origin of T2 comes from different mechanisms by which spins interact, for example the local or internal magnetic field at a dipole produced by neighboring dipoles or the spread in precession rates produced by the magnetic field that one nucleus produces at another. Classically it is the time that it takes for an ensemble of transverse spins initially precessing in phase to lose their phase with each other. The loss of the ensemble spin phase is irreversible. If the external field is inhomogenus, one can consider the macroscopic magnetization of the sample in the x-y plane as the sum of smaller magnetization vectors each arising from a small volume experiencing a homogeneous field. Within the small volume with a homogeneous magnetic field, the intrinsic relaxation time which is characteristic of the magnetization decay is T2 and it is an irreversible process. Furthermore, due to the inhomogeneity of the magnetic field, each of the small volume will precess with its own characteristic Larmor frequency. As a result, the magnetization from each small volume of spin will get out of phase with each 7 other. The time constant which describes this spin dephasing of the ensemble spins is

T2*. The phase loss due to the inhomogeneity of the magnetic field is reversible. The

1 1 relation between them is *   B0 , where B0 is the field inhomogeneity. Due to T2 T2 the inhomogeneity of the magnetic field in the experiment, the ensemble effect of T2* is usually dominating [18].

When the system is anisotropic, T1 = T2. The equality of the two times is very convenient for comparing experiment and theory, since measurements usually yield T2, while theoretically it is often more convenient to calculate T1. In many cases a single symbol τs is used for spin relaxation and dephasing (and called indiscriminately either of these terms)[18].

There are mainly four types of spin relaxation mechanisms, the D’yakonov-

Perel’s mechanism, the Elliott-Yafet mechanism, Vir-Aronov-Pikus mechanism and

Hyperfine interaction mechanim. They will be introduced briefly in Chapter 3 section

3.2.1.

For spintronics applications, longer spin relaxation time τs is desired as most times when the spin lives longer, it implies better performance and functionality of the device. There have been numerous researches to explore the spin relaxation time of semiconductors. The longest spin relaxation time found in bulk GaAs is about hundreds of ns. Longer spin relaxation time is predicted in quasi-1d system and the effort toward experimentally confirming it will be introduced in Chapter 3.

1.4 Spin detection

After spin is injected into a semiconductor and relaxed in the transporting channel, there should be ways to detect those spin polarization. In experiment it could also be detected electrically or optically.

1.4.1 Electrical spin detection

Even though efficient spin injection into semiconductor was reported shortly after the proposal of tunnel barrier, successful electrical detection was not realized until 2007 in Fe/GaAs system with a schottky tunnel barrier[20]. The typical device structure is called “non-local” measurement, and it is shown in Fig. 1.4(a). In this method, ferromagnet contacts are used both as the source and detector electrodes. The voltage at the detector is measured relatively to a reference electrode, which corresponds to the chemical potential of the spin states. It is argued that both the spin-valve effect and Hanle effect should be observed to demonstrate electrical detection of spin transport conclusively[20]. The spin valve effect is the measurement of a non-equilibrium spin population using a non-local ferromagnetic detector through which no charge current flows. The potential at the detection electrode should be sensitive to the relative magnetizations of the detector and the source electrodes. The existence of the Hanle effect is more rigorous, which is the modulation and suppression of the spin-valve signal due to precession and dephasing in a transverse magnetic field[20]. Both effects are observed in the device and electrical spin detection is confirmed, as shown in Fig. 1.4(b) and (c). Afterward, same measurement geometry is applied to different materials, for

9 example, electrical spin injection and detection is realized in graphene[21] and silicon[22].

Fig. 1.4 (a) A schematic diagram of the non-local experiment (not to scale). The large arrows indicate the magnetizations of the source and detector. Electrons are injected along the path shown in red. The injected spins (purple) diffuse in either direction from contact 3. The non-local voltage is detected at contact 4. Other choices of source and detector among contacts 2, 3 and 4 are also possible. (b) Non- local voltage versus in-plane magnetic field, By (swept in both directions) at a current I1,3 =1.0mA at

T=50 K. The raw data are shown in the upper panel is fitted by a second-order polynomial. The lower panel shows the data with this background subtracted. (c) Non-local voltage, V4,5, versus perpendicular magnetic field, Bz , for the same contacts and bias conditions as in (b) The data in the lower panel have the background subtracted. The data shown in black are obtained with the magnetizations of contacts 3 and 4 parallel, and the data shown in red are obtained in the antiparallel configuration.[20]

Beside nonlocal geometry, other measurement schemes are also applied for electrical spin detection. Appelbaum et al. used a hot electron injection and demonstrate 10 electrical detection of spin transport in Si [23]. Dash et al. used three channel “local” measurement and obtained a spin lifetime of 140ps for electrons and 270ps for holes in silicon at room temperature [24].

1.4.2 Optical spin detection

There also exist different types of optically detection methods. One of the most widely used is applying the opposite process of optical pumping and look at the circular polarization of the luminescence, which is called as a spin-LED detection [10-11, 15-16].

Not only for direct band gap material such as GaAs, this method is also used to detect spin injection and transport in indirect bandgap material such as silicon[25]. Details of the physics underlying this method are introduced in 2.2.2.

Other optical detection methods including Faraday or Kerr rotation[26-27], or optical hanle effect have also been used to study spin dynamics in direct band gap semiconductors[28].

1.5 Spintronics in novel material and systems-focus of my research

Spintronics is developing very fast towards its goal to contribute both to fundamental science and technology applications. In the mean time, it is very important to explore suitable materials and systems that give better performance and functionality to spintronic devices, which is the focus of my research topics.

My work consists of three independent projects. The first two projects are both material oriented. Recent work has focused on structures which rely on either metallic

11 spin injectors that are not well suited to multifunctional operation or magnetic semiconductors that do not function under ambient conditions. This places significant limitations on the potential utility of devices based on these principles. Therefore, it is very important to develop a material which meets the requirement as a room temperature magnets as well as exhibiting multifunctionality property. An organic-based magnet is a good candidate for the purposes. Therefore, one of the two material oriented projects is to realize spin injection from an organic-based magnet into an inorganic heterostructure.

Success of this project will open the door to a new class of active, hybrid spintronic devices with multifunctional behavior, which comes from the optical, electronic and chemical sensitivity of the organic layer.

Other than using organic materials, multiferroics is also a very appealing material for researchers in spintronics field. In multiferroic material, the strain, electric dipole and magnetic moment could be coupled together. The interactions between the magnetic and electric polarization leads to additional functionalities, for example, the magnetic properties could potentially be controlled by electric fields [29]. However, so far the existing multiferroic materials are exhibiting properties a factor of 1000 smaller than useful ferroelectrics or ferromagnets. To solve this problem, engineering an intrinsic multiferric material by applying strain on it was proposed. The proposed strained films-

EuTiO3 are grown by molecular beam epitaxy (MBE) and my second project is to test the magnetic properties of this multiferroic material by magneto-optical Kerr effect. It is found that this material is indeed multiferroic, with ferroelectricity and magnetization orders of magnitude higher than any existing multiferroic materials. This discovery

12 confirms that one could engineer a strong multiferroic which could potentially be used in real spintronics applications.

Not only could new materials bring more functionality for spintronics field, existing material with different dimensions could also exhibit different properties in terms of spins. For example, for III-V semiconductor, typical spin relaxation time is on the order of hundreds of ns, it is proposed that if the dimension is reduced to 1d, the spin relaxation time could even get longer therefore better for applications. Meanwhile, optical pump and probing is a very powerful technique because it allows probing spin dynamically and it has been used in studying spin dynamics in 3d, 2d and 0d systems.

However, there is a fundamental problem for using this technique for 1d system – polarization anisotropy. The third project that I have been working on is to realize optical studies of the spin relaxation time in semiconductor quasi-1d system by oxide coating the structure to eliminate or reduce the polarization anisotropy in the system.

These three projects will be introduced in Chapter 2, 3 and 4 in details, respectively. The results of my research validate the use of new materials and systems in spintronics for better performance and functionality. Hopefully it also triggers development of other novel materials and systems which have potentials for the field.

1.6 Reference

1. M. N. Baibich et al., Phys. Rev. Lett. 61, 2472 (1988).

2. G. Binasch et al., Phys. Rev. B 39, 4828 (1989).

3. C. Tsang et al., IEEE Trans. Magn. 30, 3801 (1994).

4. S. A. Wolf et al., Science 294, 1488 (2001).

5. D. D. Awschalom and M. E. Flatte, Nat. Phys. 3, 153 (2007).

6. S. Datta and B. Das, Appl. Phys. Lett. 56, 665 (1990).

7. B. T. Jonker, Proc. IEEE. 91, 727 (2003).

8. G. Schmidt et al., Phys. Rev. B 62, R4790 (2000).

9. E. I . Rashba, Phys. Rev. B. 62, R16267 (2000).

10. H. Zhu et al., Phys. Rev. Lett. 87, 016601 (2001).

11. A. T. Hanbicki et al., Appl. Phys. Lett. 80, 1240 (2002).

12. C. Adelmann et al., Phys. Rev. B. 71, 121301(R) (2005).

13. X. Jiang et al., Phys. Rev. Lett. 94, 056601 (2005).

14. W. Han et al., Phys. Rev. Lett. 105, 167202 (2010).

15. R. Fiederling et al., Nature 402, 787 (1999).

16. Y. Ohno et al., Nature 402, 790 (1999).

17. F. Meier and B. P. Zakharchenya, Optical Orientation Amsterdam, The Netherlands:

North-Holland, vol. 8. (1984).

18. I. Zutic, J. Fabian and S. Das Sarma, Rev. Mod. Phys. 76, 323 (2004).

19. M. E. Flatte, J. M. Byers and W. H. lau. Chap: Spin dynamics in semiconductors in

Semiconductor spintronics and quantum computation. edited by D. D. Awschalom, D.

Loss and N. Samarth Springer: New York (2002).

20. X. Lou et al., Nat. Phys. 3, 197 (2007).

21. N. Tombros et al., Nature 448, 571 (2007).

22. O. M. J. Van'tErve et al., Appl. Phys. Lett. 91, 212109 (2007).

23. I. Appelbaum et al., Nature 447, 295 (2007).

24. S. P. Dash et al., Nature 462, 491 (2009).

25. B. T. Jonker et al., Nat. Phys. 3, 542 (2007).

26. J. M. Kikkawa and D. D. Awschalom, Phys. Rev. Lett. 80, 4313 (1998).

27. S. A. Crooker et al., Science 309, 2191 (2005).

28. J. S. Colton et al., Phys. Stat. Sol. (B) 233, 445 (2002).

29. N. A. Spaldin and M. Fiebig, Science 309, 391 (2006).

Chapter 2: Electrical spin injection from an organic-based magnet in a hybrid

organic/inorganic heterostructure

2.1 Introduction

The field of semiconductor spintronics promises the extension of spin-based electronics beyond memory and magnetic sensing into active electronic components with implications for next-generation computing[1] and quantum information[2]. The development of organic-based magnets with room temperature magnetic ordering[3] and semiconducting functionality[4] promises to further broaden this impact by providing a route to all-organic spintronic devices[5-6] and hybrid organic/inorganic structures capable of exploiting the multifunctionality[7] and ease of production of organic systems[8] as well as the well established spintronic functionality of inorganic materials.

In this project we successfully extract of spin-polarized current from a thin film of the organic-based room temperature ferrimagnetic semiconductor V[TCNE]x (x~2; TCNE:

-2 tetracyanoethylene; TC ~ 400 K, EG ~ 0.5 eV,  ~ 10 S/cm) and its subsequent injection into a GaAs/AlGaAs light-emitting diode (LED). The orientation of this spin current is determined by polarization analysis of the electroluminescence from the LED and is found to be parallel to the magnetization of the V[TCNE]x~2 layer, in agreement with theoretical predictions[9-10]. Detailed analysis of the optical selection rules in the LED, coupled with control measurements of magnetic circular dichroism in the V[TCNE]x~2

16 layer, reveals the magnitude of the electron spin polarization to be largely insensitive to both electrical bias and temperature. This successful demonstration of spin injection in a hybrid organic/inorganic structure opens the door to a new class of active, hybrid spintronic devices with multifunctional behavior defined by the optical, electronic and chemical sensitivity[11] of the organic layer.

This chapter mainly contains four parts. The first part is the theory and background of the project which includes background knowledge of V[TCNE]x~2, theory and experimental background of the spin LED detection method. The second part is the device fabrication of the organic/inorganic hybrid structure. The third part is the measurement set-up. Finally the experimental results for spin injection will be introduced in the fourth part.

2.2 Theory and background

2.2.1 Magnetic ordering of V[TCNE]x

The chemical structure of the TCNE (tetracyanoethylene) molecule is shown in

Fig. 2.1(a); the molecular unit TCNE is a very good electron acceptor and accepts electron with opposite spin in its π* antibonding orbital, forming charge transfer salts in which it is present as the [TCNE]- ion. This unpaired π* electron produces a net spin of

- - 1/2 on [TCNE] . A single-crystal polarized neutron diffraction study of [Bu4N]+[TCNE]

(Bu4N =tetra-n-butylammonium) paramagnetic charge transfer salt determined the spin density distribution in [TCNE]- [12], as shown in Fig. 2.1(b). The spin of 1/2 is delocalized over the entire [TCNE]- molecular ion, with 33 % of the total spin on each of

17 the sp2 hybridized carbon atoms (central C atoms in Fig. 2.1(a)), 13 % of the spin on each of the four nitrogen atoms, and a negative corresponding to about 5 % of the total spin on each of the four sp hybridized carbon atoms[12].

Fig. 2.1 (a) Molecular compound and structure for Tetracyanoethylene (TCNE) molecule. [13] (b)The unique electronic structure of V[TCNE]x~2 leads to semiconductor charge transport with fully spin- polarized conduction and valance bands. [12]

When the [TCNE]- molecule forms a compound with metal, for example V, direct bonding of the TCNE N atoms to V creates a three dimensional network. This makes

V[TCNE]x~2 a ferrimagnetic material with large magnetic momentum per formula unit. There is direct antiferromagnetic exchange coupling between the unpaired spins of V2+

- * (t2g , S = 3/2) and TCNE ( , S = 1/2). As shown in Fig. 2.2, the unpaired spin in the TCNE- anion is distributed over the entire molecule[12] and occupies the * antibonding level. The * orbital can accept another electron with opposite spin, which costs an

* additional Coulomb energy [4, 14] Uc ~ 2 eV. Therefore, the  band is split into two

* * oppositely spin-polarized subbands: occupied  and unoccupied  +Uc. The t2g levels lie

* within the Coulomb gap[15] and define the valence band while the  + Uc levels define the conduction band with a 0.5 eV bandgap. This leads to fully spin-polarized valance

* (t2g) and conduction bands ( + Uc) with parallel spin alignment. This proposed electronic structure with ferromagnetically aligned conduction and valence bands is consistent with theoretical calculations [9-10, 16] and experimental studies[6, 15, 17].

Fig. 2.2 Schematic energy diagram for V[TCNE]x~2. The unique electronic structure of V[TCNE]x~2 leads to semiconductor charge transport with fully spin-polarized conduction and valance bands.

The family of molecule-based magnets M[TCNE]x • y(solvent)(M = transition metal) was introduced by the synthesis of V[TCNE]x • y(CH2Cl2), the first room- temperature molecule-based magnet, with magnetic ordering temperature of ~ 400K [3].

The high magnetic ordering temperature was proposed to result from the direct bonding of the TCNE N atoms to V and a three-dimensional network structure in the solid [3].

This compound is amorphous, chemically unstable, and extremely air-sensitive [3].

2.2.2 Spin-LED detection methods-optical transition

The injected spins could be detected either optically or electrically. As mentioned in Chapter 1, one of the optically detection methods is to use a spin-light emitting diode

(spin-LED) [18-19]. The principle for optical detection is based on quantum selection rule of the radiative recombination of carriers and the consequent emission of light. In a spin-LED device, when a forward bias is applied, oriented electron spins are injected into an LED structure that usually contains a GaAs/AlGaAs quantum well. The spin-polarized electrons recombine with unpolarized holes in the quantum well and radiatively emit light. The remaining net spin polarization of the electrons that are injected into quantum well could be detected quantitatively by analysis of the right (σ-) and left (σ+) circular polarization of the emitted light through the optical selection rule [20].

Fig. 2.3 E-k diagram of bulk GaAs near k=0. The heavy hole and light hole bands are degenerate at k=0

The band diagram is illustrated in Fig. 2.3 for cubic semiconductor. For a bulk zincblende semiconductor, in this case GaAs, the wavefunction at the conduction band minimum has s symmetry and has a j=1/2 with mj=±1/2 degeneracy; while the valence band maximum has p symmetry and the spin-orbit interaction splits it into a four-fold degenerate P3/2 level and a two-fold degenerate P1/2 level which is 0.34eV lower. The P1/2 level (j = 1/2) is called split-off band. The P3/2 band (j = 3/2) contains two sub-bands, the heavy hole band (HH) with mj = ±3/2 and the light hole band (LH) with mj = ±1/2. The optical transitions between the conduction band and the valence band are equivalent to a classical dipole and the selection rule only allows Δmj = ±1 happen, which corresponding to right (angular momentum l = -1; σ-) or left (angular momentum l = +1; σ+) circularly polarized light. The transition probabilities can be obtained by calculating the matrix

± element of the transition <Ψf |Hint|Ψi>, where Hint = X ± iY for σ light. Table 1 is showing the wavefuntion of all the states that involve in the ground state transitions, from which all the transition probabilities could be calculated.

Table 2.1. Angular and spin part of the wave function at k=0 of the conduction band and HH and LH of the valence band.

Fig. 2.4 (a) is a showing all the possible transitions that follow selection rules.

+ The red lines correspond to Δmj = -1 transition, which gives σ light (LCP) and the blue

- lines correspond to Δmj = +1 transition, which gives σ light (RCP). The relative probabilities are listed next to the lines. Due to the fast spin relaxation of holes in the valence band, the circular polarization of the luminescence along z direction is directly related to the electron spin polarization at the conduction band. For bulk case,

n  n considering the net spin polarization at the CB is Pspin  (n↑ and n↓ are the n  n number of spin up and spin down electrons, respectively) the detected circular

I RCP  I LCP polarization defined as P  could be written in terms of n↑ and n↓ circ I RCP  I LCP

RCP LCP I  I (n  3n )  (3n  n ) 1 n  n as Pcirc  RCP LCP    0.5Pspin . Therefore for bulk I  I (n  3n )  (3n  n ) 2 n  n

GaAs, due to the degeneracy of the HH and LH states, the maximum detected circular polarization is 0.5.

Fig 2.4. Allowed transitions for bulk GaAs and typical GaAs/AlGaAs quantum well. Red arrows indicate LCP and blue arrows indicate RCP. (a) For bulk GaAs, HH and LH states degenerate. The transition to HH and LH give same energy of optical spectrum. (b) For quantum well, HH and LH states separate and give different peaks for optical spectrum, therefore allows analysis for the circular polarization of HH and LH individually.

In a QW, as shown in Fig. 2.4 (b), due to quantum confinement the HH and LH bands are separated in energy. This separation in energy allows spectrally distinguish the

HH and LH transitions as well as analysis of the circular polarization of HH and LH separately. For HH case, the circular polarization becomes

RCP LCP I  I n  n PHH  RCP LCP   Pspin . As for the LH case, the circular polarization is I  I n  n

RCP LCP I  I n  n PLH  RCP LCP   Pspin . Therefore, if distinct HH and LH transitions are I  I n  n

23 detected, the polarization signal from HH and LH should be equal in magnitude and opposite in sign.

2.2.3 Previous work on spin injection from ferromagnets into a semiconductor

There have been many reports using a spin LED structure to detect electrical spin injection from either a ferromagnetic metal [21-23] or a ferromagnetic semiconductor

[18-19, 24-26] into LED. Two of those reports will be introduced in details in this section as background of spin LED measurements.

The first work showing here used a InGaAs/GaAs quantum well LED to detect spin injection from a ferromagnetic semiconductor (Ga,Mn)As[19]. As shown in Fig. 2.5, the field was applied parallel to the sample surface and the light was emitted from the side of sample. With this structure the in-plane magnetization of the spin injection should be detected.

Fig. 2.5 Electrical spin injection in an epitaxially grown ferromagnetic semiconductor heterostructure, based on GaAs. a, Spontaneous magnetization develops below the Curie temperature TC in the ferromagnetic p-type semiconductor (Ga,Mn)As, depicted by the black arrows in the green layer.

Under forward bias, spin-polarized holes from (Ga,Mn)As and unpolarized electrons from the n-type

GaAs substrate are injected into the (In,Ga)As quantum well (QW, hatched region), through a spacer layer with thickness d, producing polarized EL. b, Total electroluminescence (EL) intensity of the device (d=20 nm) under forward bias at temperature T=6 K and magnetic field H =1000 Oe is shown

(black curve) with its corresponding polarization (red curve). Current I=1.43 mA. Note that the polarization is largest at the QW ground state (E=1.34 eV). The EL and polarization are plotted on semi-log and linear scales, respectively. Inset, a current-voltage plot characteristic of a 20-nm spacer layer device. [19]

The spontaneous magnetization below Tc of the (Ga,Mn)As creates spin-polarized holes, under forward bias, the spin-polarized holes were injected into the quantum well and recombined with unpolarized electrons and emit light. Fig. 2.5(b) is showing the EL spectrum of the device (black curve) and polarization (red curve) at temperature T = 6K 25 and magnetic field H = 1000 Oe. When plotting the relative changes in EL polarization as a function of the in-plane magnetic field for various temperatures above and below Tc of the spin injector, as shown in Fig. 2.6, the EL polarization was clearly showing a hysteresis loop with bigger saturation magnetization and larger coercive field at lower temperatures. No polarization hysteresis was observed for T > 52K, when (Ga,Mn)As becomes paramagnetic. In the inset figure, the temperature dependence of the relative remanent polarization followed the sample’s magnetization moment M (solid black curve). Both facts are direct evidence that spins are injected from (Ga,Mn)As into the quantum well structure.

Fig. 2.6 Hysteretic electroluminescence polarization is a direct result of spin injection from the ferromagnetic (Ga,Mn)As layer. Shown are relative changes in the energy integrated (shaded grey area in Fig. 2.5b) polarization ΔP, at temperatures T=6-52 K, as a function of in-plane field from a device with d =140 nm. E=1.34 eV, I=2.8 mA. Triangles indicate points taken when the field is swept up or down. Inset, the relative remanent polarization (ΔP at H=0 Oe) shown in solid squares at T=6-94

K, and the temperature dependence of the (Ga,Mn)As magnetic moment, measured by a SQUID magnetometer (solid black curve), demonstrating that polarization is proportional to magnetic moment. [19]

Control measurements are very important in spin LED measurements, since there are various measurement artifacts that look like spin injection signal. First measurement artifact is that there could be an intrinsic magnetic response from the quantum well LED structure. To exclude it, a non-magnetic material was used instead of (Ga,Mn)As and same measurement and analysis were taken for comparison. The second measurement artifact is that there could be magnetic circular dichroism from the magnetic layer, which is an optical filtering effect. To exclude it circular polarization of the photoluminescence was measured.. As shown in Fig. 2.7, in both control measurements, the polarization exhibit no hysteresis loop as observed in the spin injection device, therefore, spin injection was confirmed from the (Ga,Mn)As into quantum well.

Fig. 2.7 The absence of hysteretic polarization. Data shown from a forward biased nonmagnetic device at T=6 K (a) and from a magnetic structure under optical excitation (b) verify that electrical spin injection occurs within the device depicted in Fig. 2.5a. a, Magnetic field dependence of ΔP from a non-magnetic device is shown in contrast to a magnetic structure, with identical layers (d=20 nm). b,

No hysteresis is observed from the photoluminescence (PL) polarization of a magnetic sample excited with E =1.398 eV (QW PL ground state, E=1.345 eV). [19] 27

Spin injection from ferromagnetic metal, such as Fe, into quantum well structure was also detected by spin-LED method, as shown in one of the earliest papers[22]. Here a

20-nm thick Fe was grown directly on top of the III-V semiconductor layers in MBE chamber without exposing to air. The EL measurements were performed with Faraday geometry, where the magnetic field is parallel to the light propagation. In this case, EL polarization should follow the out of plane magnetization of Fe. In Fig. 2.8, the solid black curve is a typical out of plane magnetization curve of Fe measured by SQUID. EL polarization of the Fe-LED (black dots) follows well the magnetization of Fe, which indicates successful spin injection. The spin injection efficiency is about 2%.

Furthermore, the deviation of the polarization degree from the magnetization curve for

B>4T is explained by the fact that spin injection is a superposition of spin signal and

Zeeman splitting in GaAs. For a nonmagnetic contact shown with the open symbols, no signature of spin injection is observed and it also shows a linear dependence on the magnetic field, which is explained by the Zeeman splitting induced spin alignment in

GaAs.

Fig. 2.8 Circular polarization degree P as a function of external magnetic field measured at 25 K from

LEDs with (full squares) and without (open triangles) Fe cap layer. The out-of-plane magnetization curve of a thin Fe layer is shown for comparison in arbitrary units (solid line). The dotted lines are guides to the eye and indicate the contribution of the Zeeman splitting induced spin alignment in

GaAs. [22]

The spin injection at room temperature was explored in this paper, as Fe has very high Tc. For the room temperature signal, the circular polarization of both HH and LH luminescence were studied and the schematic for this measurement is shown in Fig.

2.9(a). Due to different peak position of HH and LH, the EL polarization could be plotted separately, as seen in Fig. 2.9(b). The EL polarization for HH and LH here show same magnitude and opposite sign, just as predicted by in 2.2.1. The spin injection efficiency in this Fe-LED device is 2%. As mentioned in Chapter 1, it was discovered that the major difficulty for spin injection from metal to semiconductor is due to the conductivity mismatch between the metal injector and the semiconductors[27-28]. It is proposed to use a Schottky tunneling barrier between the metal and semiconductor to solve the resistance mismatch problem[28-30]. The result showing here is the first paper proving possible 29 spin injection through a Schotty barrier from ferromagnetic metal to semiconductor. With improved Schottky barrier over the last decade, the spin injection efficiency has increased significantly to 30-40% for Fe spin-LEDs [21, 23].

Fig. 2.9 (a) Electroluminescence spectrum of LED with Fe cap layer recorded at 300 K. The shaded areas indicate the integrated intensities used to determine the circular polarization degree P for heavy- hole (HH) and light-hole (LH) transitions. The schemes indicate the circular polarization of the EL light from recombination of electrons with spin +1/2 for heavy-hole and light-hole transitions. (b)

Circular polarization degree P for heavy-hole (full squares) and light-hole (open squares) transitions as a function of external magnetic field measured at 300 K from LED with Fe cap layer. The magnetization curve of a thin Fe layer is shown for comparison in arbitrary units with two opposite signs (solid lines) [22].

2.3 Fabrication of devices and principle of operation

2.3.1 Spin-LED device fabrication with V[TCNE]x~2

The MBE grown LED structure is purchased from the Electronic materials and devices laboratory of the Ohio State University. It contains III-V material which is grown on a p-doped (p = 1×1018 cm-3) GaAs (100) substrate with layer structure as follows: 300

17 -3 16 -3 nm p-GaAs (1×10 cm ) buffer layer/ 200 nm p-Al0.1Ga0.9As (p = 1×10 cm )/ p- contact/ 25 nm i-Al0.1Ga0.9As barrier/ 10 nm i-GaAs QW/ 25 nm i-Al0.1Ga0.9As barrier/

16 -3 + 100 nm n-Al0.1Ga0.9As (n = 1×10 cm ) drift layer/ 15 nm n  n Al0.1Ga0.9As grading

16 -3 18 -3 + + 18 - layer going from n = 1×10 cm to 5×10 cm / 15 nm n - Al0.1Ga0.9As (n = 5×10 cm

3) Schottky contact.

For device fabrication, Al (7 nm) and Au (23 nm) are thermally deposited on the back of the p-GaAs substrate to form an Ohmic p-contact. The wafer is then patterned using standard photolithography and wet etching to define a mesa of 200 m by 1.5 mm.

In order to have electrical contact only to the top of the mesa, the wafer is coated with an insulating photoresist and patterned to have a window of 100 m by 1 mm directly over the mesa. For full spin-LED devices, the samples are transferred to a series of interconnected argon glove boxes (O2, H2O < 0.5 ppm) for further processing, beginning with the deposition of V[TCNE]x~2 uniformly over the photoresist (including the window). The V(TCNE)x~2 films (~ 400 nm) are grown via chemical vapour deposition

(CVD) with a deposition rate of 5.5 nm/min. The samples are then transferred to an integrated vacuum chamber for thermal deposition of the top metal contacts. An optically transparent aluminium layer (7 nm) is deposited over the mesa and a narrow gold stripe

(23 nm) is subsequently deposited partially on top of the mesa in order to ensure good electrical contact to the aluminium. For electrical connection from the device to the sample puck, gold wires are connected using pressed indium contacts for the top contacts and copper tape is used for the bottom contact. For the type 1 control samples using non- ferromagnet electrodes, the deposition of V[TCNE]x~2 is omitted and Al and Au are

31 directly deposited on top of the mesa. For type 2 control measurements measuring magnetic circular dichroism, the same thickness of V[TCNE]x~2 is directly deposited on same size of unprocessed LED wafer. The whole device after the fabrication is shown in

Fig. 2.10.

Fig. 2.10 Cross section of the whole device after fabrication

2.3.2 Principle of spin-LED device

Fig. 2.11 Band diagram of full spin-LED device under forward bias (+V). Spin-polarized electrons are injected from V[TCNE]x~2 into n-AlGaAs and recombine with unpolarized holes at the GaAs QW.

Dashed box represents V[TCNE]x~2 / n-AlGaAs interface. Inset: A schematic cross section of full spin-LED device.

Figure 2.11 is showing the band diagram of full device when a forward bias is applied. The forward bias flattens the energy levels of LED structure and electrons are injected from V[TCNE]x~2 to LED structure. Since the π*+Uc level (conduction band of

V[TCNE]x~2) is spin oriented (Psource), electrons are polarized and injected into the conduction band of the n-AlGaAs layer of LED structure(Pinj). Since the work function of

V[TCNE]x~2 unknown, the interface between V[TCNE]x~2 and n-AlGaAs is not clear and it is represented by the dashed box. The spin oriented electrons pass through the n-type

AlGaAs and recombine with unpolarized holes at quantum well (PQW) and emit circular polarized light. The emitted light is due to electrically injected electrons and therefore it is called electroluminescence (EL). In this process, some spin information gets lost due to

33 limited spin injection efficiency and spin relaxation. However, as given in 2.2.2, by analyzing the circular polarization of EL signal, the spin polarization of electrons at the quantum well could be traced and it should follow the magnetization of the spin injection source. The HH and LH peaks should show same magnitude and opposite sign, as illustrated in Fig.2.12.

Fig. 2.12 .The expected circular polarization signal of HH/LH transitions (right) for a given magnetization of the spin injector (left). Due to quantum selection rules, HH and LH show opposite polarization behavior.

2.4 Measurement setup

2.4.1 Custom designed sample mount

Due to the oxygen and moisture sensitivity of V[TCNE]x~2 film, an oxygen and moisture free environment is required for the device transferring from the argon glove box for sample processing to the measurement chambers. This environment is realized by a custom designed sample mount, a picture of which is shown as Fig. 2.13(a).

Fig. 2.13 Pictures of the (a), custom-design sample holder. (b), mounted V[TCNE]x~2 spin-LED device and (c) zoom-in picture showing the mesa and Au, Al contact

The sample mount contains three separate parts, part 1 is where the sample is mounted on and it is made of copper. Four pins are set up at the top of the mount with three of them connected to the three electrically isolated pads, which are also electrically isolated from the mount itself. One of those three pads is at the bottom of the mount and two of them are in the middle and they are all for top contact connections. The other pin is connected to the mounts itself for bottom contact. The whole piece is inserted to part 2, which is filled with argon before loading the sample. After sample is loaded, part 3 seals part 2 with soft Indium seal. The device is loaded into the sample mount in argon glove box therefore transferring with this mount protects the device from oxygen and moisture.

A picture with device on the mount is shown in Fig. 2.13(b). Fig 2.13(c) shows a zoomed-in image of the sample. Au and Al contacts are clearly seen on top of the mesa.

2.4.2 Transport measurements

For pure transport studies devices are transferred to a Quantum Design Physical

Property Measurement System (QD PPMS; 300 - 2 K, up to 9 T) after a brief exposure to air (~ 1 min). A Keithley 2400 sourcemeter is employed for electrical measurements. IV scans are measured as a function of temperature while limiting the current in order to prevent damaging of the p-i-n diode and V[TCNE]x~2 layer. For V[TCNE]x~2 spin-LED, low temperature (T < 60 K) IV behavior could not be investigated as it requires excessive bias (V > 20 V) and results in the damaging of the device.

2.4.3 Polarization resolved electroluminescence measurement

For electro-optical measurements, the devices in the mount are then loaded into a magneto-optical cryostat with a split coil superconducting magnet capable of producing magnetic fields of up to 8 T (with windows both parallel and perpendicular to the field) and temperatures down to 1.2 K. A magnetic field is applied perpendicular to the sample surface, the total spin of the electron and hole along the axis of quantization can be mapped to the circular polarization of the emitted photon propagating along the same axis. For quantum well luminescence, the quantization axis is determined by the growth direction, which is z direction. To detect the luminescence in z direction, the devices are loaded with sample surface perpendicular to both the direction of the magnetic field and the collection optics. For electroluminescence (EL) measurements, a Keithley 2400 source meter is used to apply constant current between the top (lower bias) and bottom

(higher bias) contacts.

As discussed before, to study the spin injection in this V[TCNE]x~2 spin-LED device, the circular polarization of the emitted electroluminescence is measured. A higher current density is required due to attenuation of the luminescence from the V[TCNE]x~2 layer. However, too high current should also be avoided to prevent irreversible damage to the p-i-n structure. After studying several devices, a compliant applied current of I = 1 mA and corresponding voltage of V = 20 V is set.

For the optical measurements, a sketch of the set up is shown in Fig. 2.14. Lens 1 collimates the emitted light to Lens 2, which acts as a beam reducer together with Lens 3 so that the beam after Lens 3 is also collimated with smaller diameter and collected by an objective lens. After the objective lens, the luminescence is collected by an optical fiber and finally into a 0.3m PI-Acton spectrometer (SpectraPro-2300i) and a liquid N2 cooled

CCD camera (Spec-10). A variable waveplate and a linear polarizer are placed in the collection path between Lens 3 and the objective lens to resolve circular polarization of the luminescence. At each field, the retardance of the variable waveplate is first set to λ/4 retardance so RCP is converted to s-polarized linearization and LCP is converted to p- polarization. The linear polarizer after is set to pass only s-polarization so the intensity of

RCP (IRCP) is detected. Then 3λ/4 retardance is set for the variable waveplate in which

RCP is converted to p-polarized and LCP is set to s-polarized, while the setting for linear polarizer remains the same so the intensity of LCP light (ILCP) is detected. Alternating variable waveplate setting allows differential measurement of the intensity of the RCP and LCP components of the luminescence separately. The net circular polarization PEL is

RCP LCP RCP LCP defined as PEL = (I – I )/(I + I ). In this way PEL is collected and averaged 10 37 times for each field point. The averaged PEL is the PEL value in each field point. The full field dependence is determined by sweeping the field from -1000 Oe to +1000 Oe and back to -1000 Oe for every 100 Oe.

Fig. 2.14 Measurement set up for the polarization resolved electroluminescence. The device emits both left and right circularly polarized light. The variable waveplate changes the circularly polarized light to linearly polarized light and the following linear polarizer only passed one component of linearly polarized light, which corresponds to the intensity of one component of circularly polarized light emitted by the device. Switching the variable waveplate from λ/4 to 3λ/4 allows the detector to record the intensity of the right and left circularly polarized light emitted by the device, separately.

2.4.4 Magnetic circular dichroism measurement

For measurement of the magnetic circular dichroism (MCD), a linearly polarized optical excitation with a wavelength of 700nm and spot size on the order of 100μm is 38 focused onto a sample, which is comprised of V[TCNE]x~2 deposited on an unprocessed

AlGaAs/GaAs wafer. The incident angle is approximately 15°. The detection method is same as the electroluminescence measurement in Fig.2.14.

2.5 Experiment results

2.5.1 Spin-LED function: I-V and MR results

The electrical properties of both V[TCNE]x~2 spin-LEDs and bare LEDs are studied and compared, as shown in Fig. 2.15. Both devices show modified p-i-n diode behavior. However, the V[TCNE]x~2 spin-LED turns on at a much higher voltage than the bare LED (15V vs. 2.5V), indicating an additional series impedance from the

V[TCNE]x~2 layer in the V[TCNE]x~2 spin-LED device, consistent with thermally activated transport in an intrinsic semiconductor with a bandgap of 0.5 eV.

Fig. 2.15 Electrical characteristics of the spin-LED device at T = 60 K. I–V curve of bare spin-LED

(diamond symbols) and V[TCNE]x~2 spin-LED (triangle symbols). Both spin-LED devices show a standard p-i-n diode behavior. V[TCNE]x~2 spin-LED exhibits larger turn on voltage due to intrinsic high resistivity of V[TCNE]x~2 at low temperature.

Figure 2.16 shows the magnetoresistance measurement for device with and without V[TCNE]x~2, a positive linear magnetoresistance is observed in V[TCNE]x~2 spin-LED devices with the same sign and magnitude as isolated V[TCNE]x~2 films, confirming charge flow through the V[TCNE]x~2 layer; while the bare LED device shows the expected weak negative magnetoresistance for a non-magnetic III-V semiconductor.

Those electrical measurements set up the base for the spin injection measurement, which is that charge flow is going through the V[TCNE]x~2 layer and electrons’ spin are polarized.

Fig. 2.16 Magnetoresistance of LED devices. V[TCNE]x~2 spin-LED exhibit a linear positive MR at

100 K as expected from highly resistive V[TCNE]x~2 layer in the hybrid LED structure. At same temperature bare LED (doped-AlGaAs) show negative MR that is consistent with previously reported magneto-transport behavior in III-V systems.

The temperature dependence of the current-voltage curves for both the full hybrid spin-LED and the bare LED structure are shown in Fig. 2.17 on a semi-log scale. The bare LED shows linear behavior on the semi-log plot down to a temperature of 30 K, indicating that the p-i-n diode follows the expected exponential dependence on source- drain bias. In contrast, the V[TCNE]x~2 spin-LED appears nearly linear at high temperature but deviates strongly from simple exponential behaviour as the temperature is reduced. This suggests an increasingly dominant contribution from transport within the

V[TCNE]x~2 layer at low temperature.

Fig. 2.17 Electrical characterization of LED devices. (a) Semi-log IV curves of a bare LED device from 300 K to 30 K. IV curves exhibit weak temperature dependence and demonstrate typical p-i-n diode charge transport behavior. (b) Semi-log IV curves of a V[TCNE]x~2 spin-LED from 300 K to 60

K. IV curves show typical p-i-n diode charge transport behavior with a strong temperature dependence. Bending in the high voltage regime indicate an additional series resistance to the diode structure.

2.5.2 Electroluminescence from full-spin-LED device and control measurements

Figure 2.18(a) shows an EL spectrum from a V[TCNE]x~2 spin-LED at a temperature of T = 60 K and with an applied current of I = 0.5 mA and a bias of V =

+18.5 V. There are three distinct peaks. The peak at 1.533 eV represents the transition

42 from the conduction band of the quantum well to the HH states of the valence band. The higher energy peak at 1.541 eV is attributed to the corresponding LH transition, consistent with theoretical predictions for the HH/LH splitting [31]. The broad peak centered at 1.471 eV is due to recombination in the p-doped GaAs substrate. More details of the HH and LH peaks are shown in Fig. 2.18(b), red curve is for RCP and black curve

RCP LCP RCP LCP is for LCP. The circular polarization is determined as PEL = (I – I )/(I + I ), where IRCP and ILCP are determined individually by integrating over the appropriate spectral peaks (shaded regions in Fig. 2.18(b))) and PEL is calculated for the HH and LH, respectively. As mentioned earlier, ten of those scans are taken and averaged to get a value of PEL at every field point.

Fig. 2.18 (a) Optical characteristics of the spin-LED device at T = 60 K. Typical EL spectra of

V[TCNE]x~2 spin-LED device.(I = 0.5 mA,V = +18.5 V). (b) Zoomed in spectrum for the electroluminescence for Fig 2.17. The shaded areas in the spectra indicate the region of polarization integration over the QW HH and LH peaks, respectively.

Figure 2.19(a) is showing the circular polarization for both HH and LH for full spin-LED device as a function of magnetic field. The HH polarization exhibits strong

43 magnetic field dependence while the LH polarization shows no measureable variation.

While the full inversion between HH and LH predicted in Fig. 2.12 is not observed, the

HH/LH asymmetry strongly suggests that the signal is due to electrical spin injection. As discussed above, this discrepancy can be further quantified by exploring the typical sources of extraneous field response in a spin-LED [19]: magnetic circular dichroism

(MCD) in the V[TCNE]x~2 and the intrinsic magnetic field response of the AlGaAs/GaAs quantum well.

The MCD results are shown in Fig. 2.19(b). In contrast to the EL signal, the HH and LH PL shows identical response to the magnetic field and the amplitude of the MCD is significantly weaker than the spin injection signal in Fig. 2.19(a). The intrinsic magnetic field response of the quantum well is measured on a bare LED (no V[TCNE]x~2 layer) using the same EL measurement as described for the full spin-LED. The polarization for the HH and LH again show identical field response, in this case a simple linear field dependence (Fig. 2.19(c)). The likely source of this signal is the combination of Zeeman splitting in the quantum well and dichroism from the cryostat windows.

Fig. 2.19 Circular Polarization measurement for V[TCNE]x~2 spin-LED device and the control devices at T = 60 K. Filled symbols are for field sweeping up and open symbols are for field sweeping down.

(a) EL polarization of V[TCNE]x~2 spin-LED device at I = 0.5 mA, V = +18.5 V. HH and LH show different magnetic field dependence. (b). PL polarization (MCD) measurement of a V[TCNE]x~2 coated QW sample. Symmetric polarization response from HH and LH is observed. (c). EL polarization from a bare spin-LED device. Both HH and LH show a linear dependence on magnetic field. Note that a field-independent background polarization is subtracted for all scans.

2.5.3 Polarization analysis

The measured PEL is a superposition of the spin injection signal (Pspin), the MCD signal (PMCD) and the bare EL signal (Pbare) for both HH and LH transitions. Pbare can be removed by a linear fit to the calibration data in Fig. 3(c), leaving the corrected

HH / LH HH / LH HH / LH polarization as PEL  PMCD  Pspin . The background corrected polarization signals are plotted in Fig. 2.20 where it can be seen that PEL of the HH peak reaches

45 saturation at ~ 200 Oe, closely tracking the out of plane magnetization of the V[TCNE]x~2

(Fig. 2.20 green line) and providing further support for the presence of electrical spin injection.

Fig. 2.20 Circular polarization for HH and LH transitions in a V[TCNE]x~2 spin-LED device at T = 60

K, I = 0.5 mA and V = +18.5 V. after polarization from EL control is subtracted(see text). HH polarization follows out of plane magnetization curve (green line, measured using SQUID magnetometry) of V[TCNE]x~2 and reaches saturation at ±200 Oe. LH polarization does not show significant field dependence due to the serendipitous cancellation of spin injection and MCD signals at this temperature.

Using this simple model the lack of full inversion between HH and LH in Fig.

HH LH HH LH 2.20 can be explained by noting that theory predicts PMCD  PMCD and Pspin  Pspin . As shown in Fig. 2.21.

Fig. 2.21 A cartoon showing the polarization relationship, PEL=Pspin+PMCD When the MCD signal and spin signal are close in value, the measured signal is twice for HH and cancels out for LH, which is observed in Fig 2.20

LH HH When PMCD and Pspin are comparable, PEL is zero and PEL is twice of the actual

LH HH spin signal. Moreover, subtracting PEL from PEL will result in a cancellation of the

MCD terms and give twice the optical polarization ( 2 Pspin ). This analysis yields a

sat saturated spin polarization of Pspin =0.098 ± 0.007% (the error is one standard deviation at saturation). The saturated spin polarization is related to the injected electron spin

sat inj  r polarization at the quantum well (inj) by Pspin  , where   (1 ) and τr and τs   s are the recombination time and spin relaxation time of the injected electrons, respectively[20]. Since η is always greater than one, the measured is always

47 attenuated from inj. Independent photoluminescence studies give a η of 8.48, and thus a

inj of 0.83 ± 0.07% at 60 K. This number is comparable to the initial results from Fe and

GaMnAs based spin-LED structures[19, 22].

The dependence of PEL on bias and temperature is explored in Figs. 2.22(a) and

(b), respectively. The large impedance of the V[TCNE]x~2 layer, and the resulting high turn on voltage for the spin-LED, limits the range of applied bias over which measureable

EL can be obtained. Figure 2.22(a) shows a measurement of PEL at a bias of +15.2 V and a current of 0.08 mA (the minimum values that give measureable luminescence). The signal is noisier due to the low light level but shows no significant change from Fig. 2.20, indicating an insensitivity to bias within the measurement window. Figure 2.22(b) shows

PEL at a temperature of 140 K, the highest temperature at which EL from the quantum well can be clearly resolved. This data provides further support for the simple two- component model for the polarization presented above. At 140K the magnitude for

HH LH both Pspin and Pspin are reduced from 60K, consistent with an increased η of 18.09, while the MCD signal does not depend significantly on temperature (data not shown). As a result, at 140K the MCD dominates the polarization, and the EL approaches the

sat polarization signal shown in Fig. 3(b). When this η is combined with Pspin = 0.036 ±

0.006%, a value for inj(140 K) of 0.66 ± 0.16% is determined. This relatively modest dependence on temperature (inj(60 K) = 0.83 ± 0.07%) is consistent with spin injection from other room temperature ferromagnets such as Fe[21], where the spin injection efficiency tracks with the bulk magnetization, and is distinct from materials such as

LSMO where surface relaxation leads to a rapid depolarization of the spin current at elevated temperature [32].

Fig. 2.22 Circular polarization for HH and LH transitions in a V[TCNE]x~2 spin-LED device after polarization from EL control is subtracted for all measurements (see text). (a). EL polarization at T =

60 K, I = 0.08 mA and V = +15.2 V. Injection with lower applied bias gives similar though noisier HH and LH polarization. (b), EL polarization at T = 140 K, I = 1.5 mA and V = +10.1 V. Circular polarization from EL persists up to 140 K, the highest temperature at which EL from the quantum well can be clearly resolved.

 The value of   (1 r ) is estimated by measuring pump power dependence of  s the spin polarization of the quantum well from photoluminescence (PL) using the following approach[20]. Right circularly polarized light at 1.790 eV is used to excite the sample, creating free carriers in both the AlGaAs barriers and the GaAs quantum well that subsequently relax to the ground state of the quantum well. The polarization of the

RCP LCP RCP LCP subsequent luminescence is given by PPL = (I – I )/(I + I ) and is measured as a

49 function of laser intensity (Fig. 2.23). In an intrinsic semiconductor this polarization will be independent of excitation intensity and is determined by the optical selection rules of the system under study. Under the assumption that the majority of the carriers are excited in the barriers due to our non-resonant excitation, we expect to create a spin polarization of 50% in the conduction band [20]. Ignoring relaxation effects and the finite spin lifetime, this polarization is transferred to the ground state of the quantum well. Due to the spectrally distinct nature of the heavy- and light-hole ground states this polarization is transferred with perfect fidelity to the optical polarization, predicting PPL = 50%. As discussed in the main text, including relaxation effects and recombination time reduces

PPL from this ideal value by η.

Fig. 2.23 The dependence of PPL on pump power at T = 60K and T = 140K. With increasing pump power, a saturation of polarization is observed for both temperatures and gives an estimate for η.

For an intrinsic semiconductor this value of  can be determined directly from the

relation P0  Pspin . However, for n-doped material there is an additional complication to

50 consider: the 50% free carrier polarization introduced by the pump only accounts for the optically excited carriers, while the full electron spin polarization must include both the photocarriers and the equilibrium carriers introduced by the donors. As a result, the net spin polarization will be reduced by a factor of noptical /(noptical + ndonor). As the intensity of the optical excitation increases this ratio approaches unity and the simple relationship

P0  Pspin is recovered. This can be seen directly in Fig. 2.23 for temperatures of 60 K and 140 K as the observed polarization saturates with increasing laser power, yielding saturation polarizations of 5.89±0.03% and 2.76±0.02%, respectively. These polarizations can in turn be used to calculate values of 8.48±0.05 and 18.09±0.15.

2.6 Conclusions and outlook.

In conclusion, optical detection of electrical spin injection across a

V[TCNE]x~2/AlGaAs interface has been demonstrated in an active hybrid organic/inorganic spin-LED device. Circular polarization of the electroluminescence that tracks the magnetization of the V[TCNE]x~2 layer consistent with the expected HH/LH asymmetry. These studies validate the spintronic functionality of organic-based magnets and lay the foundation for a new class of multifunctional hybrid spintronic structures and represent the first all-semiconductor spintronic device with the potential for room temperature operation. Potential benefits of this new paradigm include low cost and substrate-generic fabrication [33-35] as well as multifunctional operation exploiting the chemical, electronic and photonic functionality of the organic layers. Moreover, the technique of using extensively studied inorganic heterostructures as a sensitive probe of

51 free carrier spin physics enables the fundamental study of spin physics in organic and molecular systems.

However, the spin injection efficiency is still low (0.8% at T = 60 K) and the highest temperature observing the spin injection signal is not room temperature yet. The low spin injection efficiency is likely due to the poor interface between the V[TCNE]x~2 and inorganic LED structure. There is uncontrolled oxidization on the surface and no chemical treatment on the surface is performed prior deposition of V[TCNE]x~2 layer. On the other hand, for the well developed all inorganic spin-LED devices such as Fe-spin-

LED, Fe is deposited in-situ in MBE chamber on the AlGaAs/GaAs heterostructure, which guarantees a good interface to maintain spin polarization during injection process.

Therefore, chemical surface treatment before depositing V[TCNE]x~2 is proposed to increase the spin injection efficiency. Moreover, the limit of temperature for detecting the spin signal is due to the fast decay of spin polarization as increasing temperature in GaAs system[36-37]. The use of other semiconductor material, such as GaN, whose spin relaxation has much less dependence on the temperature comparing to GaAs [38], could be used as spin detector to realize room temperature operation.

2.7 Reference

1. International technology roadmap for semiconductors. Executive summary for 2009.

(2009).

2. D. P. Divincenzo, Science 270, 255 (1995).

3. J. M. Manriquez et al., Science 252, 1415 (1991).

4. V. N. Prigodin et al., Adv. Mater. 14, 1230 (2002).

5. J.-W. Yoo et al., Phys. Rev. B. 80, 205207 (2009).

6. J.-W. Yoo et al., Nat. Mater. 9, 638 (2010).

7. J.-W. Yoo et al., Phys. Rev. Lett. 97, 247205 (2006).

8. K. I. Pokhodnya, A. J. Epstein, and J. S. Miller, Adv. Mater. 12, 410 (2000).

9. A. L. Tchougreeff and R. A Dronskowski, J. Comput. Chem. 29, 2220 (2008).

10. H. Matsuura, K. Miyake, and H. Fukuyama, J. Phys. Soc. Jpn. 79, 034712 (2010).

11. J. S. Miller and A. J. Epstein, Angew. Chem. Int. Edit. 33, 385 (1994).

12. A. Zheludev et al., J. Am. Chem. Soc.116, 7243 (1994).

13. D. A. Dixon and J. S. Miller, J. Am. Chem. Soc. 109, 3656 (1987)

14. C. Tengstedt et al., Phys. Rev. B., 69,165208 (2004).

15. C. Tengstedt et al., Phys. Rev. Lett. 96, 057209 (2006).

16. G. C. De Fusco, et al., Phys. Rev. B. 79, 085201 (2009).

17. J. B. Kortright, et al., Phys. Rev. Lett. 100, 257204 (2008).

18. R. Fiederling et al., Nature 402, 787 (1999).

19. Y. Ohno et al., Nature 402, 790 (1999).

20. F. Meier and B. P. Zakharchenya, Optical Orientation Amsterdam, The Netherlands:

North-Holland, vol. 8. (1984).

21. A. T. Hanbicki et al., Appl. Phys. Lett. 80, 1240 (2002).

22. H. Zhu et al., Phys. Rev. Lett. 87, 016601 (2001).

23. C. Adelmann et al., Phys. Rev. B. 71, 121301(R) (2005).

24. G. Kioseoglou et al., Nat. Mater. 3, 799 (2004).

25. B. T. Jonker et al., Nat. Phys. 3, 542 (2007).

26. Y. Chye Phys. Rev. B. 66, 201301(R) (2002).

27. G. Schmit et al., Phys. Rev. B. 62, 4790(R) (2000).

28. I. Zutic, J. Fabian and S. Das Sarma, Rev. Mod. Phys. 76, 323 (2004)

29. E. I. Rashba Phys. Rev. B. 62, 16267(R) (2000).

30. A. Fert and H. Jaffres Phys. Rev. B. 64, 184420 (2001).

31. B. T. Jonker, Proc. IEEE. 91, 727 (2003).

32. J.-H. park et al., Phys. Rev. Lett. 81, 1953 (1998)

33. Editorial. Sumitomo’s PLED. III-Vs Rev.16, 27 (2003)

34. Q. Wang & D. Ma, Chem. Soc. Rev. 39, 2387 (2010)

35. H. Klauk, Chem. Soc. Rev. 39, 2643 (2010)

36. J. M. Kikkawa and D. D. Awschalom, Phys. Rev. Lett. 80, 4313 (1998).

37. W. H. Lau , J.T. Olesberg, and M.E. Flatte, Phys. Rev. B. 64, 161301 (1998).

38. J. H. Buss et al., Phys. Rev. B. 81, 155216 (2010).

Chapter 3: Polarization control of semiconductor nanowires

3.1 Introduction

Studying the spin relaxation time τs is very important for spintronics and quantum computing[1]. Although comprehensive studies of spin relaxation time in 3d, 2d and 0d have been done[2-4], to implement spintronics into real application, understanding the spin relaxation time in the quasi 1D system is crucial.

Semiconductor nanowires have been a hot topic among researchers in physics, chemistry, material science and engineering for decades. The attraction of nanowire comes from the interesting fundamental properties in this low-dimensional system as well as its exciting potentials in nanoscale electronic and photonic applications [5]. The study of electrical properties of nanowires enables defining fundamental limits in these one- dimensional systems that are relevant to possible applications such as high-performance field-effect transistors and quantum based electronics. Study of optical properties of nanowires provides the unique and highly-flexible way for creating multicolor nanophotonic structures and integrating photonics with a number of existing technologies, such as nanowire LED and laser, as well as highly sensitive nanoscale photon detectors.

It is well known that nanowire is special for its low dimensionality. Since a complete one-dimension system is impossible to achieve due to finite size of the system,

55 a “quasi” 1D system is usually used to describe it. Due to its dimensionality, there are lots of interesting electronic and optical properties. Of these properties, the intrinsic polarization anisotropy of nanowires has attracted significant attention due to the potential for creating polarization sensitive devices. After the first report discovering the polarization anisotropy of a single InP nanowire[6], there are a number of following up studies exploring its impact on the fundamental properties [7-8] as well as applications for photodetectors [9]. In terms of studying the spin relaxation in this quasi 1-D system, it is found that due to this property, traditional optical pumping method is no longer valid. This part of thesis is trying to realize optical study of spin properties of nanowire by controlling the polarization anisotropy of the system.

In this chapter, first the theory and background of the project will be introduced, including polarization anisotropy of nanowire system and underlying physics why optical pumping is forbidden in the system. Following that is the growth techniques for various types of III-V and II-VI semiconductor nanowires. Finally the measurement scheme, the experimental results and details of the analysis will be shown.

3.2 Theory and background

3.2.1 Spin relaxation in quasi 1-D semiconductors

In semiconductor spintronics, a net spin polarization could be induced either optically or electrically[10-11]. Afterwards the spin relaxation will happen and there are mainly four types of mechanism for spin relaxation: (i) The D’yakonov and Perel’ mechanism [12], a creation of an effective magnetic field Beff, which causes spin relaxes

(ii) the Elliot-Yafet mechanism[13-14], where spin relaxation is due to momentum 56 relaxation directly through spin-orbit coupling, (iii) Bir-Aronov-Pikus (BAP) processes which involve a simultaneous ﬂip of electron and hole spins due to electron-hole exchange coupling [15] and (iv) hyperfine interaction, where electron spins interact with magnetic momenta of lattice nuclei with the nuclear spins. Among those mechanisms, the

DP mechanism is the dominant factor in semiconductor spin relaxation[16].

The DP mechanism is due to the spin splitting of conduction band, which is equivalent to a wave-vector dependent effective magnetic field Beff which cause the electron spins to process and thus relax. In 3d zinc-blende semiconductors, the orientation of the Beff is random with respect to the trajectory of the electrons, which results in the spin precession along random directions and loss of spin coherence. In 2d case, such as a quantum well, Beff is in the plane of quantum well but still random because of the spread of k in the x-y plane, leading to similar spin relaxation as the 3d case. However, if we look at 1d case, where the crystal structure of the semiconductor and the axial direction of the nanowire define a single axis direction, therefore, Beff is restricted to this direction. As a result, all spins precess along the same direction and the spin coherence is much better maintained in 1-d case.

In DP mechanism, There are two types of spin splitting of conduction band: in a material with bulk inversion asymmetry (BIA) and in a heterostructure the structural inversion asymmetry (SIA)[17].

The SIA form is called Bychkov-Rashba term and it has the form

H ( xk y  y kx ) (Eq (3.1)). where σi (i=x,y,z) are the Pauli matrices and the constant

η reflects the strength of the spin splitting in the conduction band and its value is defined

57 by the details of the semiconductor band structure[16]. This form of the spin Hamiltonian is equivalent to the interaction of the spin with the effective magnetic field H = (ћ/2)

σ•Beff , where Beff = ηDP v×ẑ and v = ћ k/m*. The quantum-mechanical description of the evolution of the spin 1/2 is equivalent to the evolution of the classic momentum S with the equation of motion dS/dt = Beff × S. Using the Bychkov-Rashba term for 2d electron gas channels with finite width, the theory predict (i) for very large spin splitting (η lp ≥1), where lp is the electron mean free path, we violate a general condition for the motional narrowing regime (defined in (ii)) for the DP spin relaxation. Each elementary rotation is not small and the information about the spin polarization is lost after the first random scattering event. For this regime, τs is on the order of τp, which is the momentum relaxation time, and is the shortest of all regimes. (ii)When η is small (η lp < 1), it is the

2D 2 regime of motional narrowing and a well-known equation  s ~ p (lp ) defines the time of spin relaxation in the 2D system. As the strip of the 2D electron gas L is narrowed, the behavior is unchanged until ηL~1. (iii) For smaller channel widths (ηL<1), where the 2D channel with L decreases below the mean free path lp, the channel width L acts as a new mean free path in the system, substituting lp in equations, DP spin

2D 2 relaxation is suppressed very effectively with  s ~ s (L) . This means that when the dimension of the 2d channel is reduced to the order of electron mean free path, a much longer spin relaxation time is expected (shown in Fig. 3.1(a) and (b)) [16].

Fig. 3.1 (a) Different regimes of the spin relaxation on the plane (ηDP ,L) of model parameters: ηDP

Lp≥1 (elementary rotations during free flights are not small), s ~ p ; ηDPLp<1 , ηDPL ≥ 1 (2D spin

2D -2 2D relaxation), s ~ p (ηDP Lp) ; ηDPL <1 (supression of spin relaxation, quasi-1D regime), s ~ s

-2 -4 -2 (ηDP L) ~ p ηDP L ; L≤Lp (L substitutes Lp, quantum-mechanical quantization in the channel) (b) s as a function of the channel width L [16]

In addition to the Bychkov-Rashba coupling caused by SIA, the BIA induces a

Dresselhaus type of coupling. As a result of the relatively low crystal symmetry, such as zinc-blende structure, the effective 2 × 2 electron Hamiltonian for the conduction electrons contains spin-dependent terms that are cubic in the electron wave vector k. For semiconductor nanowires without structural asymmetry, the effect of SIA is small and effect of BIA dominates spin relaxation process[16, 18]. The BIA induced Dresselhaus term is written as[16]:

2 2 2 2 2 2 H BIA [ x kx ( k y    kz )  y k y ( kz  kx )  z kz (kx   k y )] (Eq(3.2))

2 2 where , are the average wave vectors along the confinement directions y and z, respectively. This leads to a linear and a cubic k term due to BIA for spin splitting.

The Dresselhaus linear term is similar to the Rashba term in Eq. (1). As a result, the

Dresselhaus linear term should give similar dependence of s on the nanowire diameter

2  s  (d) . The Dresselhaus cubic term due to BIA gives additional spin relaxation which is independent of the nanowire diameter. Hence the cubic term imposes a limit in the increase of s. The EY mechanism may also limit the spin relaxation time at very small d due to enhanced momentum scattering from the nanowire surface. The goal of the project is to characterize the spin relaxation time in semiconductor nanowires at various nanowire diameters to verify the theory prediction and distinguish relevant spin relaxation mechanisms in 1D semiconductors through the saturation of s and its dependence on temperature and magnetic field.

3.2.2 Polarization anisotropy in quasi 1-D system

The first report on the polarization anisotropy of semiconductor nanowires is from the Lieber’s group in 2001 published in Science [6]. In this report, photoluminescence of free standing, single indium phosphide (InP) nanowires are studied. Fig. 3.2 shows that for a single InP nanowire, both the excitation and emission spectra show high anisotropy.

In both case, the ratio of parallel to perpendicular emission is greater than an order of magnitude. The reason for the large polarization response is attributed to the large dielectric contrast between nanowire and its environment. When the nanowire diameter is much smaller than the wavelength of the exciting light, the classical electrostatic theory is valid. The nanowire could be treated as an infinite dielectric cylinder in a vacuum and the excitation light could be treated as electromagnetic fields. When the incident field is 60 polarized parallel to the cylinder, the electric field inside the cylinder is still Ei//=Ee//. But when polarized perpendicular to the cylinder, the electric field amplitude is attenuated

2 e according to Ei  Ee , where Ei is the electric field inside the cylinder, Ee the    e excitation field, and ε(εe) is the dielectric constant of the nanowire cylinder

(vacuum)[19].

Fig. 3.2 Polarized excitation and emission spectra of nanowires. (A) Excitation spectra of a 15-nm- diameter InP nanowire. These spectra were recorded with the polarization of the exciting laser aligned parallel (solid line) and perpendicular (dashed line) to the wire axis. The polarization ratio, ρ=(I// -

I┴)/(I// + I┴ ), is 0.96. Inset, plot of the polarization ratio as a function of energy. (B) Emission spectra of the same wire as in (A).These spectra were taken with the excitation parallel to the wire, while a polarizer was placed in the detection optics. The polarization ratio of the parallel (solid line) to perpendicular (dashed line) emission is 0.92. The spectra were taken with integration times of 10 s.

Inset, plot of the polarization ratio as a function of energy. (C) Dielectric contrast model of polarization anisotropy. The nanowire is treated as an infinite dielectric cylinder in a vacuum while the laser polarizations are considered as electrostatic fields oriented as depicted. [6] 61

After this finding, there has been very active research exploring what the polarization anisotropy brings both for the fundamental studies and the application for photonics. Mishra and coauthors [7] report polarization dependent photoluminescence study from both zinc-blende (ZB, cubic) and wurtzite (WZ, hexagonal) structure InP nanowires. In their study, they found that WZ structure show a PL position at a higher energy than ZB structure (1.49eV vs. 1.418eV) at the same temperature. Moreover, the polarization of the PL emission from ZB and WZ structure display very distinctive differences. In their measurement, circularly polarized 780 nm laser excitation was used to excite all the linearly polarized electronic states with equal probability, and the emitted

PL was analyzed for linear polarization. In Fig. 3.3, the circles show the variation of the

PL integrated intensity from ZB NW as a function of the polarization angle, with the solid line fit to cos2θ. The ZB InP NW emission is strongly polarized PL (82%) parallel to the nanowire axis. It is pointed out that in ZB InP, similar to the band diagram of GaAs in Fig 2.3 of last chapter, the recombination InP is between the s-like conduction band electrons with Г6 symmetry and the p-like doubly degenerate light and heavy hole velence bands with Г8 symmetry. As a result, the emitted light is completely unpolarized.

Therefore, the anisotropic polarization response is purely because of the dielectric contrast between the NW and the environment [7].

Fig. 3.3 Integrated intensity of PL emission at 15 K from single ZB NW and WZ NW as a function of linear polarization analyzer angle. [7]

In the paper, as comparison, the polarization response from WZ InP nanowire show opposite sign to the ZB structure. The reason is attributed to the special of WZ structure (shown in Fig. 3.4 [20]). A hexagonal wurtzite crystal exhibits a completely different symmetry. The s-like conduction band exhibit Г7 symmetry, while the p-like hole bands split into three separate hole bands due to a combination of the spin-orbit interaction and crystal field splitting[20-21]. The lowest energy hole states has a Г9 symmetry, while the two higher energy hole bands have Г7 symmetry. The lowest energy exciton state is between a Г7 electron and a Г9 hole, which means a dipole allowed only for E perpendicular to the c axis of the nanowire, and forbidden for E parallel to the c axis[20-21]. Thus the lowest energy exciton transition in WZ InP should be strongly polarized perpendicular to the wurtzite c axis, which is the nanowire axis. That’s why the polarization responses from ZB and WZ nanowires have opposite sign. Moreover, the weaker polarization anisotropy in WZ nanowire (49%) is attributed to a combination of

63 the dipole selection rules of the WZ crystal and the dielectric contrast inherent for all semiconductor NWs[7]. Because of this reason, it is very important to make sure the structure of the nanowire that we are studying.

Fig. 3.4 Band structure and selection rules for ZB and WZ structures. Crystal splitting and spin-orbit splittings are indicated schematically. Transitions which are allowed for various polarizations of photon electric vector with respect to crystal “c” axis are indicated[20].

3.2.3 Optical pumping in quasi 1-D system

In terms of the spin relaxation study in nanowires, the inherent polarization anisotropy is actually an important fundamental problem that forbids people from using the optical techniques to explore the spin in this quasi 1d system. As introduced in section

2.2.2, for recombination process the optical selection rules in zincblende semiconductors determines that only left and right circularly polarized light are emitted from the optical transitions. The optical pumping process is an opposite process of the recombination, where the optical selection also determines that only left or right polarized light with angular momentum l = ±1 could excite the transitions in Fig. 2.4. For a bulk material, 64 optical pumping usually generates 50% of up or down spin polarization with a right or left circularly polarized light, respectively[22].

Fig. 3.5 Diagram showing a circular polarization (red) as combination of two perpendicular linear polarization (blue and green)

A circular polarization could be written as a combination of two linear polarizations which are perpendicular to each other with (2n+1)π/2 phase difference between them, as shown in Fig. 3.5. For a bulk material, when the incident light reaches the surface, the circular polarization is still circularly polarized inside the material. In a quasi 1d system shown in Fig. 3.6, the circular polarization, could be considered as decomposition of two linear polarized electrostatic fields: one parallel to the nanowire axis Ee// and one perpendicular to the nanowire axis Ee┴ with │Ee//│=│Ee┴│. When the light enters the nanowire, in the parallel direction the intensity is not reduced (Ei//=Ee//) due to the continuity of E// at the surface however, for the perpendicular direction, the

2 e field is attenuated to Ei  Ee [19]. Even though it is derived for static electric    e field, since the nanowire diameter is much smaller than the wavelength of the light, it is 65 still valid inside the nanowire[8]. The total electrostatic field inside the wire is the superposition of Ei// and Ei┴. Since │Ei//│≠│Ei┴│, the circular polarization of light inside the nanowire is strongly suppressed and no longer a circularly polarized light, but becomes an ecliptical case. Therefore, optical pumping inside the quasi 1d structure is forbidden and will not generate a net spin polarization inside the nanowires.

Fig. 3.6 Diagram showing the absorption of light electrostatic field inside a nanowire when decomposing the circular polarization into two linear polarization parallel and perpendicular to the nanowire axis. The parallel direction is continuous while the perpendicular direction is strongly attenuated due to the dielectric mismatch of the nanowire and its environment.

More specifically, the large difference between Ei// and Ei results in a dramatic anisotropy in the optical absorption coefficients // and  for directions parallel and perpendicular to the nanowire axis, respectively [23], and it is given as

2 2 c Ei // d v c Ei d v  //  2 and    2 ((Eq (3.3)), where d is the Ee // Ee electron-hole dipole moment inside the semiconductor giving the optical transitions

66 between the conduction band and the valence band. Usually for semiconductor nanowires whose diameters are significantly larger than the Bohr radius (for InP ~9 nm), the quantum confinement plays no role and the dipole matrix elements are isotropic [23].

Thus the anisotropy of optical absorption becomes

2 2   E    2     i    e  , ((Eq (3.4))  //     //  Ei      e  which leads to a polarization in absorption

2 2  //  ( e  )  4 e pa   2 2 . ((Eq (3.5))  //  ( e  )  4 e

The light emission, such as photoluminescence and electroluminescence, from semiconductor nanowires in air also exhibits strong polarization due to similar mechanism [24]. The anisotropy in the intensity (I) of light emission is [8]

2 I 6 e  2 2 , ((Eq (3.6)) I // ( e  )  2 e which gives the polarization in light emission,

2 2 I //  I ( e  )  4 pe   2 2 . ((Eq (3.7)) I //  I ( e  )  8

From Eq 3-7, we can see that for ZB nanowires the polarization anisotropy in both absorption and luminescence are purely from the dielectric mismatch between the nanowires and the environment they are in. When the nanowires are embedded in a dielectric material instead of the air (diagram shown in Fig.3.7), for example, for InP nanowires (dielectric constant  ~ 9.6 [25])embedded in a material such as TiO2 (with e

~ 7[25]), the absorption polarization should decrease significantly from 92% in air (e ~ 67

1) to 13% and the luminescence polarization decreases from 89% in air to 9%. The absence of optical anisotropy will allow the excitation of spins in semiconductor nanowires through the absorption of circularly polarized light and subsequent characterization of spin relaxation process. For example, with index matching a GaAs nanowire ( ~ 10.9) with a dielectric material with e ~ 7, circularly polarized light with a small component of linear polarization (13%) will be absorbed and a 44% of spin polarization will be generated. It is even possible to realize 50% spin polarization with index matching nanowire with same dielectric constant material.

Fig. 3.7 Diagram showing index matching randomly oriented nanowires with a material which has a close dielectric constant.

In our experiment, randomly oriented nanowire ensembles are studied. In the ensemble system, the polarization anisotropy does not disappear, but is only reduced comparing to a single nanowire, as predicted by theories [8, 26] and also observed experimentally [9, 27]. In nanowire ensembles, the incident polarized light preferentially excites a group of nanowires which oriented nearly parallel to the excitation polarization within the ensemble. They absorb most intensity of the excitation and emit linearly

68 polarized light in the same direction. As a result, a suppression of the polarization anisotropy is observed. This effect is referred as polarization memory effect [8]. A more detailed model predicting the polarization anisotropy of nanowire ensembles will be explained later in the chapter.

Since the polarization anisotropy exists in the ensemble of nanowires, we propose controlling the polarization anisotropy through oxide coating these nanowire ensembles with oxide media which has a close dielectric constant with the nanowire material. When the light interacts with the oxide coated nanowires, there are two interfaces need to be considered. First, the oxide coating should be thick enough so that light sees it as a bulk material. As a result the decomposed electrical fields in both directions (E// and E┴) have same magnitude inside the media as their values in air, respectively. Therefore, it is still circularly polarized in the media. Second, since the dielectric constant of the media is close or even same as the nanowire material, when the excitation light propagates to their interface, the dielectric mismatch effect vanishes so that E// and E┴ inside the nanowire again remains the same magnitude. Therefore, the polarization anisotropy inside the nanowire is largely reduced by the oxide coating and optical pumping becomes valid in the nanowire system. Testing of the oxide coating is realized through measuring the linear polarization dependent photoluminescence of the oxide coated nanowires.

3.3 Growth of semiconductor nanowires

3.3.1 Growth mechanism – The vapor-liquid-solid method

The growth of semiconductor is via the vapor-liquid-solid mechanism (VLS) by employing metal nanoparticles as catalysts [28], usually Au (Fig. 3.8(a), GaAs as an example). In the process, the metal nanoparticles should be heated above the eutectic temperature for the metal–semiconductor system with the existence of a vapor-phase source of the semiconductor, therefore creating a liquid droplet of the metal- semiconductor alloy. A continued supplying of the semiconductor vapor is also required so that it continuously feeds into the droplet and eventually supersaturates the eutectic, leading to nucleation of the solid semiconductor. As shown in Fig. 3.8(b), the coexistence of the liquid phase of the metal-semicondutor alloy and the solid semiconductor in the equilibrium phase diagram is required; therefore only in the shaded region the growth process could happen. The solid–liquid interface is the growth interface, where the continued semiconductor vapor enters the lattice and resulting in the growth of the nanowire with the alloy droplet riding on the top.

Fig. 3.8. Schematics of (a) GaAs nanowire growth catalyzed by Au nanoparticles. (b) Au/GaAs phase diagram.

There are two competing interfaces during nanowire growth, 1. the liquid/solid interface between the droplet and the grown nanowire and 2. the vapor/solid interface between the semiconductor vapor supply and the exposed surface of the growing nanowire. Growth through the first interface results in the axial VLS growth of the nanowire, while the second interface results in vapor–solid growth and thickening in the radial direction. Depending on the detailed growth conditions such as the material, pressure, flow rate, temperature, background gases which are the by-products of growth reactions, either mechanism can dominate in an actual growth process. Uniform nanowires with negligible diameter variation can be achieved through careful control of the growth conditions. In the VLS growth process the diameter of the nanowire is determined by the diameter of the nanoparticles that are used as catalysts and the length of the nanowires are usually controlled by the growth time. Moreover, the crystal orientation of the nanowire during VLS growth is chosen to minimize the total free energy, as the process is thermodynamically driven.

The vapor phase semiconductor reactants usually are generated through momentum and energy transfer methods such as pulsed laser deposition (PLD) [29] or molecular beam epitaxy (MBE) [30] from solid targets or through decomposition of precursors in a chemical vapor deposition (CVD) process. For the compound semiconductor case, metal-organic chemical vapor deposition (MOCVD) [31] or PLD

[32] are generally used to provide the reactants. For the nanowires used in this project, the InP and ZnO nanowires, PLD and CVD are used as growth methods, respectively and will be introduced in next two sessions.

3.3.2 Pulsed laser deposition of InP nanowires

In our study, InP nanowires are grown by a pulsed laser deposition (PLD) system with a three-zone tube furnace as the growth chamber. Silicon wafer with 500nm thermal oxide were cleaned with standard solvent cleaning procedure (15 mins sonication in

Aceton, 15 mins sonication in methanol) and treated of 1hr UV-Ozone to minimum the organic residues on the surface, then it is immersed in a solution of 0.1% poly-L-lysine

(Ted Pella) for 15mins, then rinsed with DI water and blown dry with N2. Diluted 50nm

Au colloid solution (0.01% concentration, BBI International) was dropped on the wafer to distribute the catalyst for 15mins and then blown dried. The density of nanowire could be controlled by the dilution of the Au colloid catalyst and the nanowire diameter can also be controlled by choosing different sizes of commercially purchased Au colloid.

While Au colloids are left and attached to the substrate, 2g InP polycrystalline powder

(99.999%, Alfa Aesar) is grounded for 2 hrs and pressed for 12 hrs to form an InP target.

Fig. 3.9 is showing a Schematic of the laser ablation system. The InP target was placed just outside the opening of the tube furnace at the upstream end. The Au colloid covered substrate is placed at downstream side of the tube furnace, typically 9 inches from the center of the furnace on the downstream side. The tube was evacuated to below 10 mTorr by a mechanical pump before the heating started. The center zone of the tube furnace was heated to 850°C and the substrate was heated to between 450°C and 550°C for the nanowire growth, which could be controlled by a left zone heater. When the growth temperature was reached, argon gas with a total pressure between 60 and 100 Torr and a flow rate of 50 – 100 Standard Cubic Centimeters per Minute (sccm) was flown from the

72 target side (upstream end) to the substrate side (downstream end). Then, an excimer pulsed laser beam of a wavelength of 248 nm and duration of 10 ns is focused onto a 2.4

× 3.9 mm2 spot on the InP target to start the ablation which generates In and P vapors.

The laser energy density is maintained at 1 - 2 J/cm2 per pulse and a frequency of 2 - 10

Hz. The argon flow carries the In and P vapor to the preheated substrate near the downstream end of the furnace. The Au nanoparticles on the substrate form alloy with the

InP vapors and become liquid nanodroplets, from which InP nanowires nucleate and grow. A typically growth duration is 10 – 30 minutes.

Fig. 3.9 Schematic of the laser ablation system for the synthesis of semiconductor nanowires.

In addition, by mixing InP powder with Se doping or Zn doping to form the target, either n-type or p-type InP nanowire could be grown, respectively. Different doping of the nanowire enables studying transport properties of the nanowire as well as p-n junction nanowire for nanowire LEDs and photodiodes.

3.3.3 Chemical vapor deposition of ZnO nanowires

ZnO nanowires were grown using the same tube furnace (without the excimer laser) by heating a 1:1 molar mixture of ZnO and graphite powders at the center of the 73 furnace to 900 °C to generate Zn vapor, which was carried by a gas flow of 2% O2 in Ar to a Si/SiOx substrate positioned at a few cm downstream from the source powders[33].

3.3.4 Index matching with oxide sputtering

The dielectric environment change of the nanowires was realized by coating the as-grown nanowires with a thick Ta2O5 layer which has a dielectric constant of ~4 [25] in order to reduce the dielectric mismatch between the semiconductor nanowires and their environment. The oxide coating was grown in an ultrahigh vacuum (UHV) sputtering system using an off-axis geometry to minimize the damage to the nanowires. As shown in Fig. 3.10, the UHV magnetron sputter system is equipped with a load lock chamber and two substrate heaters capable of controlling the substrate temperature up to 850 °C with the fluctuation of ±1 °C. The main chamber has a base pressure of 5×10−10 Torr. We used RF sputtering in a sputter gas of 5% O2 in Ar with a total pressure of 18 mTorr. The deposition rate is 0.8 nm/min.

Fig. 3.10 set up of the UHV sputtering system.

3.3.5 Structure synthesis of InP and ZnO nanowires

The as grown nanowires are synthesized by scanning electron microscopy (SEM) and tunneling electron microscopy (TEM). Fig. 3.11 is SEM of the InP nanowire (a) is showing that the nanowires are uniform and straight with random orientation. More zoomed in image ((b)) is showing that the diameters of nanowire are about 60-80nms.

Fig. 3.11 SEM image of as grown InP from PLD. (b) is a zoom in view of (a) showing the dimension of nanowire.

In Fig. 3.12, TEM images are showing that the each nanowire has a single crystal core and a polycrystalline shell structure (a). The diffraction pattern shown in (b) indicates that the core has cubic structure, which corresponds to the zincblende crystal structure. Therefore, in the nanowire we grow, the polarization anisotropy only comes from the dielectric contrast inherent for all semiconductor NWs. The polarization anisotropy from the dipole selection rules of the WZ crystal (discussed in section 3.2.2) is not considered here.

Fig. 3.12 TEM images of as grown InP from PLD. (a) core-shell structure with single crystal in the core and polycrystalline for the shell (b) diffraction pattern indicating zincblende structure for the core

InP. (c) Zoomed in image from (a) showing a <111> growth direction of the core.

Figure 3.13 is the SEM image of the ZnO nanowires. Figure 3.13(a) is showing that the ZnO nanowires are less dense than InP nanowires and there are large number of pollycrytalline ZnO formed from Au colloid as background. The wires still have random orientation, with some of them straight and some are twisted. More zoomed in image

((b)) is showing that the diameters of nanowire are about 20-40nms.

Fig. 3.13 SEM image of as grown ZnO from CVD. (b) is a zoom in view of (a) showing the dimension of nanowire.

3.4 Polarization resolved photoluminescence measurement for NWs

3.4.1 Measurement set up

The samples are loaded to the sample mount of a superconducting spectromag system and cooled down to 5K. A 2mW, 690nm CW Ti-sapphire laser was focused on the sample with a spot size of ~100um. Within this spot size, an ensemble of nanowires’ luminescence is detected. The pump wavelength is chosen because of the large laser scattering signal from the nanowire surface. To avoid the laser scattering dominating the

NW luminescence, a much lower wavelength 690nm is used. The schematic for measurement set up is shown in Fig. 3.14. The collection path is very similar to the EL measurement in Chap 2. To measure the linear polarization anisotropy, a Glan-Laser linear polarizer is placed at the pump beam to fix the excitation beam polarization and another Glan-Laser linear polarizer and an achromatic zero-order half waveplate are placed at the collection path. When rotating the angle of the half waveplate, NW luminescence with different linear polarization is detected.

Fig. 3.14 Measurement set up for linear polarization resolved photoluminescence.

For the measurement of the ZnO nanowires, all the optics are replaced with blue optics with 300-500nm range since ZnO is a wide band gap material with a band gap of

3.3 eV. By doubling a 690nm laser frequency with a BBO crystal, a 345nm excitation wavelength is used to pump the ZnO nanowire system.

3.4.2 PL results for bare NWs and oxide coated NWs

Fig. 3.15 shows the comparison of as-grown InP NWs ((a) and (b)) and oxide coated NWs ((c) and (d)). From the SEM images, the oxide is conformally coated outside the bare wires and reaches diameter between 500 to 700 nm. Typical photoluminescence spectra from InP nanowires with excitation and detection parallel and perpendicular to each other are plotted. The gray shaded area centered ~849nm is where the nanowire luminescence mainly is. The broadness of the peak is due to inhomogeneity of nanowire diameter. The spectrum is asymmetric and has a tail at lower energy end, which is due to surface defects and nucleation in bulk/ film form during growth. There is a satellite peak

78 around 890 nm, which may be contributed to polycrystalline compounds formed on wire surface. The ratio of the intensity of the NW luminescence when the excitation and

I detection are perpendicular and parallel is   0.59 , as indicated by the red and black I // spectrum in (b). The large deviation from single InP nanowire polarization anisotropy of

I   0.02 [6] is because of the ensemble effect, as reduced polarization anisotropy is I // expected in ensemble of NWs than in a single NW [9]. The satellite peak at 890nm which is not from nanowires has much less polarization dependence. The PL intensity is directly related to the polarization difference between the incoming beam and photoluminescence.

Fig. 3.15 (a) SEM image of as-grown InP nanowire. The nanowires are randomly oriented with an average diameter 50-80nm. (b) Photoluminescence for as-grown InP nanowire with pump and collection polarization parallel (black) and perpendicular (red) to each other. The parallel intensity is much stronger than perpendicular one. (c) SEM image of Ta2O5 coated InP nanowire. The oxide is conformally coated and the diameter reaches 500-700nm. (d) Photoluminescence for oxide coated InP nanowire with pump and collection polarization parallel (black) and perpendicular (red) to each other.

As a comparison, the typical spectrum of nanowire after oxide coating is shown in (d). The gray region is again to highlight the luminescence of the nanowire from luminescence of defect. The luminescence peak is around 867nm. Notice that there is a shift of the PL peak from bare wire, which is due to the shift in population of nanowires which contribute more strongly to luminescence. Since the nanowire luminescence is quenched by the oxide coating, the shoulder around 890nm is more pronounced.

Moreover, the PL intensity are almost same for the parallel and perpendicular

I polarization with a polarization ratio   0.98 . We conclude that the angle dependence I // gets largely reduced in the oxide coated nanowires. The 890nm satellite peak still has no polarization dependence.

Similarly for ZnO nanowires shown in Fig. 3.16 (a), the as-grown ZnO nanwire has an average diameter of 20-40 nm and a length of several microns. The wires have very sharp and strong luminescence as seen in (b), center ~368.5nm with polarization

I ratio   0.68 , as indicated by the red and black spectrum. The larger observed I //

I polarization ratio  in ZnO nanowires than in InP nanowires is because the dielectric I // constant of ZnO is smaller than InP Fig. 3.16 (c) shows that the Ta2O5 coating outside

ZnO naowires increases the diameter to 200-250nm. After oxide coating with Ta2O5, the

I polarization ratio is largely increased to   0.94 , as seen in (d), therefore, the I // polarization anisotropy in ZnO nanowires is also reduced by oxide coating.

Fig. 3.16 (a) SEM image of as-grown ZnO nanowire. The nanowires are randomly oriented with an average diameter 20-40nm. (b) Photoluminescence for as-grown ZnO nanowire with pump and collection polarization parallel (black) and perpendicular (red) to each other. The parallel intensity is much stronger than perpendicular one. (c) SEM image of Ta2O5 coated ZnO nanowire. The oxide is conformally coated and the diameter reaches 200-250nm. (d) Photoluminescence for oxide coated

ZnO nanowire with pump and collection polarization parallel (black) and perpendicular (red) to each other. The difference in parallel and perpendicular intensity for the oxide coated wires is significantly reduced.

In both InP and ZnO nanowires, It is confirmed that with oxide coating, the polarization anisotropy is reduced significantly. More details of the angle dependence are introduced in the following section.

3.4.3 Polarization analysis

A more careful angle dependence of PL intensity is done by rotating the half waveplate with small angle step. PL spectrum is taken at each angle of half waveplate and an integration of the shaded area is treated as intensity of the luminescence. After 81 normalizing with highest intensity the luminescence is plotted as a function of angle. PL intensity oscillates with the angle, reaching maximum and minimum every 90 degree. A model to explore the relation between the intensity and angle between the excitation and detection polarization is developped. This model is based on the experiment scheme, for randomly oriented ensemble nanowires, which is an extension of previous models developed to describe single wire measurement [6]. A schematic of the polarization anisotropy analysis for an ensemble of nanowires is shown in the Fig. 3.17.

Fig. 3.17 Schematics for the polarization analysis, dashed lines showing randomly oriented nanowires.

For a single randomly oriented nanowire, the angle between the nanowire axis and the excitation field E0 is β, the angle between the excitation and detection directions is α.

With this geometry, the detected luminescence intensity from this single wire could be written as

2 2 2 6 e 2 I  (E0 cos ) [cos (  )  2 2 sin (  )] (   e )  2 e

2 6 2  ( e E sin )2 [cos2 (   )  e sin2 (   )]   0 (  )2  2 2 e e e 82

 (E cos)2 cos2 (  ) 0 // absorption + // emission

6 2  (E cos  )2  e sin2 (   ) 0 (   )2  2 2 e e // absorption +  emission

2  ( e E sin  )2 cos2 (   )   0 e  absorption + // emission

2 6 2  ( e E sin  )2  e sin2 (   )    0 (   )2  2 2 e e e  absorption +  emission

(Eq(3.8))

2 2   E    2  where the anisotropy of the optical absorption is    i    e  (Eq (3.4)) and  //     //  Ei      e 

2 I 6 e the anisotropy in the intensity of light emission is  2 2 (Eq (3.6)). ε and I // ( e  )  2 e

εe are the dielectric constants for the nanowire and its environment, respectively. For randomly oriented nanowire ensembles, the total intensity is the sum of luminescence from all the nanowires, which is the integration of the angle β from 0 to 180 degree, This integration brings the relation between the luminescence intensity and the angle α between the detection and excitation polarization optics:

2 2 2 2 2 2 2  e  (10 e  4 e  2 )(9 e  2 e   )  ( e ) (3 e  ) cos2 8(  )2 (3 2  2    2 ) e e e (Eq(3.9))

As Fig. 3.18 top panel showing, for as-grown InP nanowire, experimental value is plotted as red dots. The corresponding calculation value is using the dielectric constant for InP nanowire (ε=9.6), and the dielectric constant for the environment of the bare

83 nanowires, which is air with εe=1. With the two numbers Eq (9) is plotted on top of the experimental value as the blue line. The calculated blue curve follows the same period as

I the experimental values, with the normalized ratio   0.41 for ideal case. This I // parameters-free model is only a function of dielectric constants. And it correlates well with the experimental result.

The same measurements and analysis are performed with Ti2O5 coated nanowires, the experiment value is plotted as the black symbol top panel in 3.18 and the theory curve is plotted as the green curve with replacing εe=1 (air) to εe=4.41 in Eq (9), which is the dielectric constant of bulk Ti2O5 [25]. The intensity vs angle still oscillates with a much

I smaller amplitude and the polarization ratio is increased to   0.94 . The calculation I //

I value for oxide coated wires (green line) is   0.895 . The increasing in the I //

I experimental ratio  from 0.6 to 0.94 indicates reducing of the polarization anisotropy I // through oxide coating.

Fig. 3.18 Polarization analysis for bare and oxide coated nanowires. Top panel: comparison between experimental (red and black) and model (blue and green) for bare and Ta2O5 coated InP nanowries.

Same period is shown in both with higher polarization anisotropy for the calculation. Bottom panel:

Same comparison for ZnO nanowires.

Similarly, bottom panel of Fig. 3.18 shows the same measurements for as-grown and oxide coated ZnO nanowires. In the blue spectrum range, the PL intensity also oscillates as a function of angle. For as-grown ZnO nanowires, the experimental value

I I  0.69 matches well with the parameter-free calculation value  0.63 (εZnO = 4 and I // I //

εe = 1). In oxide coated case, the calculation predicts almost complete vanish in the

I polarization anisotropy, shown in the green curve with   0.99 , as the dielectric I // constant of Ta2O5 (εTa2O5 = 4.41) is very close to that of ZnO (εZnO = 4).However, in

I experiment,some polarization anisotropy is still observed with   0.94 . This is I // attributed to the finite diameter of the oxide coated wires. Due to technical reason, oxide 85 coating for ZnO nanowire only reaches diameters between 200 and 250nm. With this diameter the oxide coating is not thick enough to totally eliminate the polarization anisotropy when the light excitation enters the Ta2O5 media, as some reports showing that in this diameter regime the polarization anisotropy still exists in nanowires.[26, 34]

Therefore, even though the interface between the Ta2O5 media and ZnO nanowire gives no polarization anisotropy, there is some polarization anisotropy contribution from the interface between air and Ta2O5 media. The calculation considers infinite surrounding of oxide media and assumes the polarization anisotropy only comes from the inner interface.

This assumption gives rise to the mismatch of experimental and calculation for ZnO nanowires. Nevertheless, oxide coating ZnO nanowires largely reduces the polarization anisotropy in the system, as predicted.

3.5 Hanle measurements for nanowires

After proving that with changing the environment of semiconductor nanowires the linear polarization is significantly reduced, next is to apply this to the spin measurement in the quasi-1D system. As introduced in section 3.2.3, with the oxide coating, circular polarization in quasi-1D system is possible and optical pumping could be realized. One of the most direct ways to measure the spin lifetime is to use an optical Hanle measurement

[35-37].

3.5.1 Measurement set up and theory

Fig. 3.19 Measurement set up for Hanle measurement in photoluminescence.

The Hanle measurement set up is shown in Fig. 3.19. Instead of linear pumping, a quarter waveplate is placed in the pump path to generate circularly polarized light. As discussed before, with oxide coating, the circularly polarized light is still circularly polarized in the quasi-1D system and the optical pumping creates a net spin polarization in the system in the z direction (perpendicular to the sample surface). The sample is placed parallel to the direction of the magnetic field, resulting in a depolarization to the Sz from this transverse magnetic field. This depolarization is realized through precession of spin to the applied field with the Larmor frequency Ω = μBgB/ћ where μB is the Bohr magneton and g is the g factor [22]. The depolarized Sz is detected by measuring the polarization of the photoluminescence same way as the electroluminescence introduced in last chapter.

1 1 1 The “spin lifetime” Ts is defined as   (Eq (3.10)), where τr is the Ts  r  s electron recombination time (lifetime) and τs is the spin relaxation time. The electron- hole recombination and the spin relaxation are the two main reasons for the loss of the average electron spin information. At time t after excitation, the electron spin in z direction Sz is S0cos(Ωt)exp(-t/ τs). Its average value is obtained by averaging this with the distribution of lifetime W(t):

 S  S dtW (t)exp(t / )cost (Eq (3.11)) Z 0  s 0

Under usual condition for thermalized electrons W(t) = τ-1exp(-t/τ). Eq (3.11) gives the expression for the Hanle curve [22]:

Sz (0) S0 Sz (B)  2 , Sz (0)  (Eq (3.12)) 1 (Ts ) 1 r / s

From Eq (12), by measuring the degree of polarization in zero field S0 and half

 width of the magnetic depolarization curve (Hanle curve), B1/ 2  , τr and τs can be gBTs determined individually if the g factor is known.

It is very common to use Hanle measurement to find the spin lifetime in various systems. Epstein et al. [35] found that Hanle measurements in InAs quantum dots exhibit a complex structure, suggesting multiple spin lifetime components. gTs is on the order of hundreds of ps and the dot dimensions influence the spin lifetime and its dependence on temperature. Colton et al. [38] measured Hanle in MBE grown GaAs epilayers with different doping concentration. The lifetimes obtained were 14 and 26 ns for samples

88 nominally doped at 1x1015 and 3x1015 cm-3, respectively. Moreover, the relation between the pump power and the Hanle curve width is also explored. The steady-state electron lifetime (τr) in Eq (3.12) depends on the electron concentration (n) and the electron–hole pair generation rate (G, number of pairs generated per unit volume per unit time): τr =

1 G 1 n/G. Therefore, Eq(3.10) could be written as   . Since G is directly Ts n  s proportional to the pump power, spin lifetime Ts is expected to increase as the pumping power is reduced, resulting in a narrower Hanle curve (smaller B1/2 for Hanle curve from

Eq (3.12)). Hanle effect is a simple method to extract carrier spin relaxation times from semiconductor systems.

3.5.2 Calibration Measurement on GaAs epilayer

Since the spin lifetime in 1d system is relatively complicated, it is important to make sure the validity of the measurement setup. As a calibration measurement, an epilayer of GaAs sample is tested for Hanle curve. The sample is grown on GaAs substrate, with 200nm of undoped GaAs buffer layer. On top of it is 300nm of undoped

16 Al0.9Ga0.1As with the top layer 2μm of 3x10 n-doped GaAs layer. A 780nm CW Ti- sapphire laser was focused on the sample with a spot size of ~100um. Different laser power was used to explore the power dependence of the Hanle curve. Fig. 3.20 (a) is showing a PL spectrum of the LCP and RCP luminescence for the GaAs epilayer sample at T = 5K with circular pumping excitation at 19.5 mW power. The main peak is around

I RCP  I LCP 819 nm. The circular polarization P  is plotted as a function of circ I RCP  I LCP 89 wavelength(blue curve in (b)). There is a peak around 813nm, which is not the same as the PL peak. It is probably due to the maximum spin polarization is happening not in the same energy with the recombination. Nevertheless, a well defined circular polarization is measured, indicting spin generation and detection in the GaAs epilayer sample.

Fig. 3.20 (a) PL spectrum for GaAs epilayer at 5K (b) circular polarization as a function of wavelength at zero field. Maximum CP appears ~813nm.

A Hanle curve at this pump power is also measured and shown in Fig. 3.21(a).

After fitting the typical Hanle curve to the Lorentz form given by Eq (12), a spin lifetime

Ts of 18.4ns is obtained (assuming g factor for GaAs -0.44). In (b), the dependence of

Hanle curve on the pumping power is also studied. As the pump power decreasing, narrower Hanle curves are observed and spin lifetime of 20.7ns and 28.7ns are obtained for 14.5mW and 7.2mW pumping power, respectively. The lower pump power gives longer spin lifetime here, which is consistent with the other literature reports [38-39].

Fig. 3.21 (a) Hanle curve for GaAs epilayer at 5K with 19.5mW pump power. Typical Hanle curve is observed (b) Pump power dependent Hanle curve for GaAs epilayer at 5K. Less power gives narrower curve, thus, longer spin lifetime.

3.5.3 Results on nanowire and outlook

Both bare and oxide coated InP nanowires are studied. A 20mW, 690nm CW Ti- sapphire laser was focused on the sample with a spot size of ~100um. The high power is used to excite more spin polarized carriers at zero field. For as-grown InP nanowires,

I RCP  I LCP circular polarization calculated from P  is plotted on top of the PL circ I RCP  I LCP spectrum in Fig. 3.22(a). NW peak is around 852nm and circular polarization shows a dip around the nanowire peak. The reason for this feature is not clear. However, the large field scan for Hanle curve indicates that this dip does not depend on the magnetic field, as shown in (b). This is what we expect from a bare nanowire sample, in which the optical pumping is forbidden by the index mismatch of the nanowire and its environment and therefore no spin signal should be observed.

Fig. 3.22 (a) PL specturm for an InP nanowire sample detecting RCP and LCP. Black curve is the circular polarization value vs. wavelength. A dip is observed near NW peak, reason for which is not very clear. (b) Hanle measurement for this sample. CP value is from the integration indicated in (a), which is the main NW peak. No Hanle curve is observed, as expected.

However, after nanowire is coated with oxide, Hanle curve is still not obtained for this specific sample. Fig. 3.23(a) is showing that after oxide coating, the nanowire peak at

852 nm is significantly quenched and 870 and 890nm peak become dominant. The circular polarization (black curve) shows a much more complicated feature comparing with the bare wires, reason of which is also not very clear. Three regions of CP are plotted in (b) and no peak is showing any Hanle curve. In principle only the 845nm peak is supposed to give Hanle signal as it is closest to the nanowire peak. However, since the intensity of the peak is very low (~700 counts), the resolution is not enough for observing a Hanle curve.

Fig. 3.23 (a) PL specturm for InP nanowire coated with Ta2O5. Nanowire peak at 845nm is significantly quenched and 870nm and 890nm peak become more pronounced. Black curve is the circular polarization value vs. wavelength. More complicated feature is observed and reason for which is not very clear. Three shaded areas are the integration range taken for CP value plotted in (b) .

Colors for different peaks are matching in two figures. (b) Hanle measurement for oxide coated sample. No Hanle curve is observed for all peaks, which is not expected for the nanowire peak at

845nm.

Exploring of the circular polarization in quasi-1d system is paused by the difficulties in growing the required material with PLD system. However, GaAs nanowire with a MOCVD system and research with the GaAs nanowire is taking place now in the group.

3.6 Conclusions and outlook

We discussed the limitation for using optical techniques probing the spin dynamics in quasi-1d systems, which is the polarization anisotropy induced by the index mismatch between the nanowire material and its environment. To solve this problem,

93 index matching the nanowire by changing its environment to a material whose dielectric constant is close to that of the nanowire is proposed. Two categories of nanowire, InP and

ZnO are successfully grown by PLD and CVD methods, respectively. Polarization anisotropy for photoluminescence in an ensemble of these nanowires is observed and quantitatively matches with the calculations based on the model for nanowire ensembles.

Tuning PL anisotropy of nanowires is realized by coating nanowires with Ta2O5. By choosing environmental media whose dielectric constant is close to that of the nanowires, polarization anisotropy is significantly reduced and optical spin injection into nanowires becomes possible. Although an obvious Hanle signal has not been observed in the oxide coated systems, it is attributed to the low PL intensity of the oxide coated nanowires. The research is continued with developing GaAs/AlGaAs core-shell nanowires grown by

MOCVD system.

3.7 Reference:

1. V. A. Sih, E. Johnston-Halperin, and D.D. Awschalom, Proc. IEEE. 91, 752

(2003).

2. J. M. Kikkawa and D.D. Awschalom, Nature 397, 139 (1999).

3. J. M. Kikkawa et al., Science 277, 1284 (1997).

4. J. Berezovsky et al., Science 320, 349 (2008).

5. W. Lu and C. M. Lieber, J. of Phys. D- Appl. Phys. 39, R387 (2006).

6. J. F. Wang et al., Science 293, 1455 (2001).

7. A. Mishra et al. Appl. Phys. Lett. 91, 263104 (2007).

8. H. E. Ruda and A. Shik, Phys. Rev. B 72, 115308 (2005).

9. Y. H. Yu et al., Nano Lett. 8, 1352 (2008).

10. X. H. Lou et al., Nat. Phys. 3, 197 (2007).

11. S. A. Crooker et al., Science 309, 2191 (2005).

12. D'yakonov and Perel'. Sov. Phy. Solid State, 13, 3023 (1972).

13. R. J. Elliot, Phys. Rev. 96, 266 (1955).

14. Y. Yafet, Solid State Physics. edited by F. Seitz and D. Turnbull. Vol. 14.

(Acedemic, New York, 1963).

15. G. L. Bir, Sov. Phy. JETP 42, 1382 (1976).

16. A. A. Kiselev and K.W. Kim, Phys. Rev. B 61, 13115 (2000).

17. S. D. Ganichev et al., Phys. Rev. Lett. 92, 256601 (2004).

18. R. Winkler, Phys. Rev. B 69, 045317 (2004).

19. L. D. Landau and E. M. Lifshits, Electrodynamics of Continuous Media.

(Pergamon, New York, 1984).

20. J. L. Birman, Phys. Rev. Letts. 2, 157 (1959).

21. J. L. Birman, Phys. Rev. 114, 1490 (1959).

22. F. Meier and B. P. Zakharchenya, Optical Orientation Amsterdam, The

Netherlands: North-Holland, vol. 8. (1984).

23. P. Ils et al., Phys. Rev. B 51, 4272 (1995).

24. X. F. Duan et al., Nature 409, 66 (2001).

25. Handbook of Chemistry and Physics. 86th ed., edited by D. R. Lide, (CRC Press,

Taylor & Francis Group, New York, 2005).

26. H. E. Ruda and A. Shik, J. Appl. Phys. 100, 024314 (2006).

27. Y. Kravtsova et al., Appl. Phys. Lett. 90, 083120 (2007).

28. R. S. Wagner, Whisker Technology, edited by A. P. Levitt. (Wiley, New York,

1970).

29. A. M. Morales and C. M. Lieber, Science 279, 208 (1998).

30. J. L. Liu et al., J. Cryst. Growth 200, 106 (1999).

31. K. Haraguchi et al., Appl. Phys. Lett. 60, 745 (1992).

32. X. F. Duan and C. M. Lieber, Adv. Mater. 12, 298 (2000).

33. P. C. Chang et al., Chemistry of Mater. 16, 5133 (2004).

34. H. -Y. Chen et al., Optics Express 16, 13465 (2008).

35. R. J. Epstein et al., Appl. Phys. Lett. 78, 733 (2001).

36. J. Furst et al., Semicond. Sci. Technol. 20, 209 (2005).

37. Y. Masumoto et al., Phys. Rev. B 74, 205332 (2006).

38. J. S. Colton et al., Phys. Stat. Sol. (B) 233, 445 (2002).

39. O. Maksimov et al., Physica E 32, 399 (2006).

Chapter 4: MOKE measurement of ferromagnetism in a strong multiferroic created

by means of spin-lattice coupling

4.1 Introduction

Ferroelectric ferromagnets are exceedingly rare, fundamentally interesting materials that could give rise to new technologies in which the low power and high speed of field-effect electronics is combined with the permanence and routability of voltage- controlled ferromagnetism[1-2]. Furthermore, the properties of the few existing compounds that simultaneously exhibit these phenomena[1-5] are insignificant in comparison with those of the useful ferroelectrics or ferromagnets: their spontaneous polarizations (Ps) or magnetizations (Ms) are smaller by a factor of 1000 or more. The same holds for magnetic- or electric-field-induced multiferroics[6-8]. Due to the weak properties of single-phase multiferroics, composite and multilayer approaches involving strain-coupled piezoelectric and magnetostrictive components are the closest to application today[1-2]. Recently, however, a new route to ferroelectric and ferromagnets was proposed[9] by which magnetically ordered insulators that are neither ferroelectric nor ferromagnetic are transformed into ferroelectric ferromagnets using a single control parameter: strain. The system targeted, EuTiO3, was predicted to simultaneously exhibit strong ferromagnetism (spontaneous magnetization Ms ~ 7 µB/Eu) and strong

2 ferroelectricity (spontaneous polarization Ps ~ 10 µC/cm ) simultaneously under large

98 biaxial compressive strain[9]. These values are orders of magnitude higher than any known ferroelectric ferromagnet and rival the best materials that are solely ferroelectric or ferromagnetic. However, it is hard to realize the material that could provide the large compressive strain that is required to bring the simultaneous multiferroic phase. In this project, we first show theoretically the emergence of a multiferroic state could be realized under biaxial tension strain with even lower strain size required. In experiment, the lower tension strain requirement enables higher quality crystalline films. Furthermore, ferroelectric behavior is detected in the strained EuTiO3 film by second harmonic generation (SHG) and ferromagnetism is observed with magnetic-optical Kerr effect

(MOKE) and SQUID. The resulting genesis of a strong ferromagnetic ferroelectric points the way to high temperature manifestations of this spin-phonon coupling mechanism[10]. The work demonstrates that a single experimental parameter, strain, simultaneously controls multiple order parameters and is a viable alternative tuning parameter to composition[11] for creating multiferroics.

In this chapter, first the theoretical background of the strain induced multiferroic phase will be introduced, followed by the calculation from compressive to tensile strain.

The initial results from collaborators give the structural characterization of the strain and unstrained films as well as the SHG measurement for ferroelectric. Then I will focus on explaining the use of MOKE to measure the ferromagnetism. Finally SQUID results, which quantify the magnetization measured by MOKE will be shown.

4.2 Theory and background

4.2.1 Magnetic and electrical phase control in epitaxial EuTiO3 film

In Ref 9, Fennie and Rabe proposed a new approach for designing a strongly coupled multiferroic in which the interplay of spins, optical phonons and strain leads to different ordered states. They pointed out the criteria a bulk system must satisfy for this mechanism: (1) It must be an AFM-PE insulator in which at least one infrared-active (ir) phonon is coupled to the magnetic order, (2) the spins in the AFM ground state should align with the application of a magnetic field of modest strength, (3) this alignment should decrease the frequency of the spin-coupled ir-active phonon, and, (4) the key to the approach, the ir-active mode of interest must be strongly coupled to strain. [9] From literatures[12] in which they found in bulk EuTiO3, at the onset of AFM order, the static dielectric constant ε0 undergoes a sharp reduction of about 10%, indicating a hardening of the lattice. Further, it is demonstrated that in an increasing magnetic field ε0 increases, saturating at a field large enough to fully align the spins (approximately 1.5 T). This behavior is attributed to a coupling between the spins and an ir-active phonon, given by

2 2   0    Si  S j , where ω is the frequency of an ir-active phonon (lattice) mode,

ω0 is the bare phonon frequency, λ is the macroscopic spin-phonon coupling constant and

is the nearest-neighbor spin-spin correlation function. It turns out such spin- lattice coupling, in conjunction with spin, could tune multiple ferroic order parameters simultaneously.

100

Fig. 4.1 (a) Static dielectric constant ε33 and (b) spontaneous polarization Ps as a function of epitaxial

-1 compressive strain η. For ε33, the solid lines are fits proportional to (η-ηc) , while the dashed vertical lines indicate ηc [9].

By first principle density functional calculations, they calculated the lowest-lying ir-active A2u phonon frequency, the static dielectric tensor along c axis ε33 as well as the spontaneous polarization Ps as a function of biaxial compressive strain η for FM and

AFM EuTiO3 (as seen in Fig. 4.1), which leads to the phase diagram as a function of compressive epitaxial strain for EuTiO3 system, as shown in Fig. 4.2. Here the compressive strain is given negative sign. When the applied strain is between -0.9% and -

1.2%, the competition between an AFM-PE phase and a FM-FE phase allows magnetic phase control with an applied electric field, and electric phase control with an applied magnetic field, with modest critical fields. When the strain is larger than -1.2, the (001)

EuTiO3 would transform from its paraelectric and antiferromagnetic unstrained ground state to a simultaneously ferromagnetic and ferroelectric ground state. 101

Fig. 4.2 EuTiO3: Compressive epitaxial strain (η) phase diagram.[9]

4.2.2 From compressive to biaxial strain

Due to the high compressive strain that is required to reach the multiferroic state, it is hard to find an appropriate substrate to provide the strain. Even though the substrate does exist, for example, a commensurate EuTiO3 film on a LaAlO3 substrate would have a biaxial strain of εs < -2.9%, experimentally achieving the biaxial compression required while maintaining a high quality crystalline film is not realized at this time. This motivates the extension of the calculated strain phase diagram of (001) EuTiO3 from biaxial compressive strain to biaxial tension strain. The calculated[13] full phase diagram is shown in Fig. 4.3 (a), it is found that the theoretical critical strain needed for multiferroic phase is lower in biaxial tension strain (+0.75%) than the compressive one (-

1.2%). Lower strains are easier to be realized and thicker films could be grown without being cracked. In (b) and (c), crystal structure for a bulk EuTiO3 is shown, with a lattice constant of 3.905 Å. When it is grown on a (110) DyScO3 substrate, whose lattice constant is 3.944 Å[14], the EuTiO3 film experience a strain of 1.1% (as seen in (c)) and therefore a suppression of the dimension in c axis.

102

Fig. 4.3 Predicted effect of biaxial strain on EuTiO3 and our approach to imparting such strain in

EuTiO3 films via commensurate epitaxy on (110) DyScO3, (001) SrTiO3, and (001) LSAT substrates.

(a), First principles epitaxial phase diagram of strained EuTiO3 from –2% (biaxial compression) to

+2% (biaxial tension) calculated in 0.1% steps. Regions with paraelectric (PE), ferroelectric (FE), antiferromagnetic (AFM), and ferromagnetic (FM) behaviour are shown. (b)(c), Schematic view of bulk unstrained EuTiO3 and epitaxially strained EuTiO3 thin film on the DyScO3 substrate, showing the in-plane expansion due to biaxial tension [13].

4.3 Initial results from collaborators

4.3.1 Growth of epitaxial EuTiO3 in MBE

EuTiO3 thin films were grown by reactive molecular-beam epitaxy (MBE) system

[15]. MBE is an advanced technique for the growth of thin epitaxial layers and nanostructures of compound semiconductors [16-17]. For MBE system, the growth is

−8 −9 conducted in an ultra-high vacuum of 10 ~ 10 Pa, under which O2, CO2, H2O, and N2 contamination on the growing surface can be neglected. The high vacuum makes it possible to reach a growth rate down to nm/min, so that epitaxial growth and precise control of the growth thickness is possible. MBE produces high-quality single crystal 103 layers with very abrupt interface, monolayer control of thickness, precise doping and compositional accuracy. Because of the high degree of control and flexibility using MBE, it is a valuable tool in the development of sophisticated, magnetic and photonic devices.

Fig. 4.4 Schematics showing set up of a MBE system. [18]

Fig. 4.4 is showing a schematic of a MBE set up [18]. The elemental europium and titanium sources are used and the growth substrate temperature is 650 °C in a background partial pressure of molecular oxygen of 310-8 Torr. To assess the theoretical phase diagram, commensurate EuTiO3 films are prepared on substrates in the same structural family (perovskites) with lattice constants suitably close to that of EuTiO3 so that films thick enough to test for ferroelectricity and ferromagnetism can be grown. The

22-nm-thick commensurate (001)-oriented EuTiO3 films are grown on (001)

(LaAlO3)0.29–(SrAl1/2Ta1/2O3)0.71 (LSAT), (001) SrTiO3 and (110) DyScO3 substrates.

These substrates impart biaxial strains of -0.9%, 0.0% and about +1.1%, respectively.

The structural, ferroelectric and ferromagnetic properties of these films (in addition to

104 multiple control samples including bare substrates) are measured to assess the predictions of Fig. 4.3(a).

4.3.2 Structure characterization of thin films

The films are characterized by X-ray diffraction (XRD, Fig. 4.5). θ-2θ scans in

Fig. 4.5(a) show 22 nm thick phase-pure epitaxial (001)-oriented EuTiO3 films with

Kiessig fringes. The presence of these fringes indicates that each film has a smooth film surface. (b), (c) and (d) shows the rocking curve of the 002 peak for each sample overlaid on the rocking curve of the nearest substrate peak. The full width at half maximum (FWHM) of the rocking curve of the EuTiO3 film grown on (001) LSAT is equal to 11 arc sec (0.003°), comparable to that of the underlying LSAT substrate. For

EuTiO3 on (110) DyScO3, the FWHM of the rocking curve of the EuTiO3 along the [1-

10] and [001] in-plane directions of the DyScO3 substrate is equal to 9 arc sec (0.0025°), comparable to that of the DyScO3 substrate. This is the narrowest FWHM ever reported for EuTiO3 films and demonstrates its high crystalline quality. XRD result proves all

EuTiO3 films are phase perfect and epitaxially grown on the substrates [13].

105

Fig. 4.5 (a) θ-2θ XRD scans of EuTiO3 films grown on (001) LSAT (green), (001) SrTiO3 (blue) and

(110) DyScO3 (red), Clear thickness fringes are seen. The substrate peaks are denoted with asterisks

(*). (b) XRD rocking curves from the 002 EuTiO3 film (green) and 002 LSAT substrate (black) peaks.

(c) and (d) XRD rocking curves from 002 EuTiO3 film (red) and 220 DyScO3 substrate (gold) peaks along the [1-10] and [001] in plane directions, respectively. [13]

Cross-sectional scanning transmission electron microscopy (STEM, Fig. 4.6) is also performed to characterize sample. STEM image is down to atomic scale chemical imaging of composition and bonding. It reveals that the films are commensurate, smooth, and of high structural perfection. Cross-sectional chemical mapping of the interface between the EuTiO3 and DyScO3 substrate demonstrates that the film/substrate interface is abrupt so that the strain is properly introduced to EuTiO3 [13].

106

Fig. 4.6 Structural characterization by STEM of 22 nm thick commensurate EuTiO3 films grown on

(110) DyScO3, (001) SrTiO3, and (001) LSAT substrates. Top: Annular dark field (ADF) and spectroscopic imaging of the same EuTiO3/DyScO3 heterostructure characterized. Bottom: ADF-

STEM images of the structure showing a coherent interface and a low density of defects in the EuTiO3 film. A-site and B-site elemental maps of the interface obtained by combining the Eu-M4,5 (green) and

Dy-M4,5 (red) electron energy loss spectroscopy (EELS) edges, and the Ti-L2,3 (yellow) and Sc-

L2,3(blue) edges extracted from two separate 256256 pixel spectrum image acquisitions (one for the

Eu/Dy edges, the other for Sc/Ti). Intermixing is limited to 1-2 atomic layers at the interface. [13]

4.3.3 SHG measurement for ferroelectricity

Temperature-dependent second harmonic generation (SHG) measurements

(Fig. 4.7 (a)) indicate the paraelectric-to-ferroelectric transition temperature (Tc) of the films [13]. Materials lacking inversion symmetry exhibit SHG and ferroelectrics lack inversion symmetry. Therefore SHG activity is a necessary, but insufficient condition for ferroelectricity. Only the EuTiO3/DyScO3 film exhibits an SHG response with Tc ~ 250 K.

The absence of SHG response from the EuTiO3 strained at εs = –0.9% (EuTiO3/LSAT)

107 and εs = 0.0% (EuTiO3/SrTiO3) indicates that EuTiO3 in these strain states is not ferroelectric.

Fig. 4.7 (a) Temperature dependence of SHG signal of EuTiO3/DyScO3 (red), EuTiO3/SrTiO3 (blue), and EuTiO3/LSAT (green). The result shows that the commensurate EuTiO3 strained in biaxial tension by +1.1% on DyScO3 is ferroelectric below Tc ~ 250 K, while other two are not ferroelectric at all temperatures. (b) SHG hysteresis loop (top) and corresponding polarization loop (bottom) for

EuTiO3/DyScO3 at 5 K. [13]

Fig. 4.7 (b) shows the domain dynamics probed by monitoring the SHG intensity as a function of applied in-plane electric field. After removing the background, the lower panel shows a clear hysteresis loop of the SHG signal as a function of E field, indicating ferroelectricity of the film. At this point, only the ferromagnetism needs to be tested to complete the map of the world’s strongest multiferroic material.

108

4.4 Measurement of ferromagnetism

4.4.1 Motivation for using MOKE

Testing for ferromagnetism in strained EuTiO3/DyScO3 films is complicated by the large paramagnetic response of the DyScO3 substrate above its Neel temperature and large antiferromagnetic response below its Neel temperature[19] (seen in Fig. 4.8 (a)). If a superconducting quantum interference device (SQUID) magnetometer is used under typical measurement magnetic fields, the background paramagnetic response of the thick substrate swamps the signal from the strained EuTiO3 film. Given this problem, it was proposed to use the magneto-optic Kerr effect (MOKE) to test the magnetic property of the samples. Fig. 4.8 (b) is the transmission curve of DyScO3 at 25 °C (black) and

1400°C (red)[20]. The transitivity reaches maximum ~700nm so probing around this wavelength of the laser the signal is mostly sensitive to the EuTiO3 film while picking up least signal from the DyScO3 substrate. In MOKE measurement, when the beam is incident on the sample, the magnetic optical interaction between the magnetization of the sample and the linear polarization of light rotates its linear polarization and therefore the reflected beam has a different linear polarization. The difference in polarization angle between the incident and reflected probe is the Kerr angle, θKerr. The Kerr angle is directly proportional to the magnetization of the sample (it is a product of the magnetization of sample and an optical matrix that is determined by the sample’s lattice structure), so measuring the Kerr angle gives qualitative probe to the sample’s magnetization. Since MOKE is a technique which is spectrally sensitive, this renders the measurement insensitive to the magnetization of the substrates at all temperatures.

109

Fig. 4.8 (a) The temperature dependence of the dc susceptibility of DyScO3 with a 100 Oe field applied along [110] direction. The inverse susceptibility curves are shown in the inset figures. The pink lines represent the Curie–Weiss fit over the temperature range T=200–320 K [19] (b)

Transmission curve of DyScO3 at 25 °C (black) and 1400°C (red) [20]

4.4.2 MOKE measurement set up

There are three different types of measurement geometries for Magneto-Optic

Kerr effect (MOKE) measurements, as shown in Fig. 4.9, depending on the easy axis of the magnetization for different samples. (1) Polar mode, the easy axis is perpendicular to the sample surface. (2) Longitudinal mode, the easy axis is parallel to the sample surface and parallel to the plane of incidence. (3) Transverse mode, the easy axis is parallel to the sample surface and perpendicular to the plane of incidence. In this sample, the easy axis is parallel to the sample surface so that longitudinal mode is used here.

110

Fig. 4.9 Different geomeries of the magneto-optical Kerr effect (MOKE)(a) Polar mode, sample magnetization perpendicular to the sample surface. (b) Longitudinal mode, sample magnetization parallel to the sample surface, also parallel to the plane of incidence. (c) Transverse mode, sample magnetization parallel to the sample surface, but perpendicular to the plane of incidence.

The MOKE measurements are made in a magneto-optical cryostat with a split coil superconducting magnet capable of producing magnetic fields of up to 8 T (with windows both parallel and perpendicular to the field) and temperatures down to 1.2 K.

For this measurement, the samples are mounted so that the field is applied in the plane of the sample with the face of the sample positioned at the eucentric point of the cryostat.

Due to the large magnetostrictive nature of DyScO3, bare DyScO3 and EuTiO3/DyScO samples tend to fracture in high magnetic fields if they are mounted directly to the copper sample stage. As a result, as can be seen Fig. 4.10 (a), the samples are mounted using spring clamps that allow for modest distortion of the sample while holding them securely in place for optical measurements. This mounting geometry allows the probe light to pass directly through the cryostat by reflecting at a glancing angle from the sample surface

111

(see Fig. 4.10 (b)). This geometry maximizes the sensitivity of the probe light to the in- plane component of the sample magnetization during the application of the in-plane magnetic field.

Fig. 4.10 (a) Sample mount for MOKE measurements, spring clips secure samples without strain due to thermal contraction and magnetostriction while preserving optical access. (b) Schematic of MOKE measurement, probe light is incident at a glancing angle to maximize sensitivity to the in-plane magnetization.

Exploiting the optical nature of the MOKE technique, the probe is tuned to a wavelength between 690 nm and 750 nm to maximize the absorption in the EuTiO3 epilayer while minimizing interactions with the substrate. This renders the measurement insensitive to the paramagnetic substrates as well as to the antiferromagnetic state of

DyScO3 below its Neel temperature of 3.1 K. The polarization of the probe beam is prepared in a linear polarization perpendicular to the plane of incidence (parallel to the sample surface) and is analyzed outside the cryostat after it has reflected from the sample under study. 112

The measurement geometry outside the cryostat is shown in Fig. 4.11. The incoming beam is turned into an AC signal by a mechanical chopper MC1000A from

Thorlabs which can produce a frequency of 25Hz to 1kHz. In this measurement 1 kHz is used. To detect this change in linear polarization between the incident and reflected probe, which is the Kerr angle, θKerr, the reflected beam is split into s and p linearly polarized beam by a linear polarizer (A and B signal) and the two beams are focused separately into the two photodiodes in a diode bridge. The intensity of A and B is directly transformed to electrical signal through the two photodiodes and the diode bridge is designed to have an output of A-B signal. This signal is amplified by a voltage amplifier and finally fed into a lock-in amplifier. The frequency of the chopper is fed into the lock- in amplifier so that only the chopped beam is detected. The gain and bandwidth of the amplifier, as well as the sensitivity, time constant and waiting time of the lock-in are set for best signal to noise ratio to detect A-B signal (VA-B). For the measurements, initially

VA-B is set to zero to reach a balance of s and p wave by rotating the ½ waveplate. When the magnetic field is applied, the change in magnetization changes the relative composition of s and p wave and the corresponding change in VA-B is recorded, which directly reflects the magnetization of the samples.

113

Fig. 4.11 MOKE set up outside of cyrostat, linear polarized beam interacts with sample and the resulting linear polarization of the reflected beam is measured by changing the optical signal to electrical signal through a diode bridge and then detected by lock-in amplifier.

The directly measured signal is voltage. To transfer it into angle, a calibration procedure needs to be performed to get the relationship between the real angle and measured voltage. For calibration,the ½ waveplate is rotated independently in the set up and the corresponding voltage change could be read from the lock-in. Since the change in angle is readable from the waveplate, a linear fit can be obtained to get the relation between the voltage change and the real angle. This linear coefficient is applied to get the

θKerr from the voltage signal.

The data is processed as follows: Fig 4.12. (a) is raw data from a single field sweep at a temperature of 2 K. The field is swept from –50 Oe to +50 Oe and back to -50

Oe. When the field comes back to -50 Oe, the signal does not exactly reproduce due to a linear drift in time. This linear drift is subtracted to generate the data shown in (b).

114

Finally, the offset is set to zero and five full field scans are averaged to improve the signal to noise, as shown in (c). Note that despite the averaging, the magnetic field reversal does not show noticeable broadening, indicating excellent reproducibility in Hc.

Fig. 4.12 Analysis of the MOKE data. (a) Raw data from a single MOKE scan (–50 Oe to +50 Oe to –

50 Oe). (b) A linear drift in time is subtracted so that the first and last data points coincide. (c) The offset in θKerr is subtracted so that the loop is symmetric and five individual scans are averaged (this data is the same as in Fig. 4.13. Excellent reproducibilty in Hc prevents significant broadening of the magnetization reversal.

4.4.3 Ferromagnetism at 2 K

The MOKE response at a temperature of 2 K from all three EuTiO3 films and a bare DyScO3 substrate is shown in Fig. 4.13. All the θKerr curves are obtained after data processing described above. The EuTiO3 strained by εs = +1.1% (EuTiO3/DyScO3) exhibits a clear ferromagnetic hysteresis loop, with sharp switching to full saturation, signifying that it is ferromagnetic in contrast to the EuTiO3 with εs = –0.9%

(EuTiO3/LSAT) or εs = 0.0% (EuTiO3/SrTiO3).

115

Fig. 4.13 MOKE measurements showing that commensurate EuTiO3 strained in biaxial tension by

+1.1% on DyScO3 is ferromagnetic at 2 K (red). The control samples are not ferromagnetic.

EuTiO3/SrTiO3 (blue), EuTiO3/LSAT (green), and bare DyScO3 substrate (gold).

4.4.4 Temperature dependence of ferromagnetism

To explore the Curie temperature of the ferromagnetism, its temperature dependence is explored. Due to linear drifts in time (See Fig. 4.12 (a)), simply monitoring

θKerr as a function of temperature is not practical. As a result, the temperature dependence is determined by taking a series of hysteresis curves from low to high temperature.

MOKE signal at different temperatures from 1.64K to 5K are shown in Fig. 4.14.The magnetization shows sharp switching to full saturation up to Tc, even for the small signal at T = 4.3 K, indicating high quality of the film. As the temperature increases, the hysteresis curve is shrinking down. Both the full saturation θKerr and the coercivity are decreasing, indicating the magnetization is reducing as temperature increases.

116

Fig. 4.14 Temperature dependent hysteresis loop obtained from EuTiO3/DyScO3 sample with MOKE measurements. At all temperatures, sharp switching with full saturation is observed. As temperature is increasing, the Kerr angle and coercivity are reducing. No hysteresis loop is observed at 5K

From Fig. 4.14, remanent Kerr angle could be extracted at each temperature. The remanent Kerr angle is written as θremanent and it is calculated from the form

saturation(H) saturation(H)   (Eq 1.). The temperature dependence of θremanent remanent 2 extracted is plotted with black solid points in Fig. 4.15. Error bars’ values are given by the standard deviation of the five averages inidicated in Fig. 4.12. The remanent Kerr angle monotonically decreases as temperature is increased until, above a temperature of

~4.3 K, the signal to noise ratio of the remanence drops below one. The data is fit to a

 power law of the form: M(T)  M 0 (Tc  T) (Eq. 2.) and is shown by the dashed line in

Fig. 4.15. The model fits well with the experimental curve and yields a TC of 4.24±0.02 K and a critical exponent β of 0.42±0.02, in reasonable agreement with the theoretical estimate of 0.5 for a three dimensional ferromagnet [21]. The coercivity Hc is also

117 extracted from the magnetization curves by the equation

H  H  H (Eq. 3.) for all temperatures and is displayed as blue solid c (remanent0) (remanent0) points in Fig. 4.15. The error bars represent the standard deviation of five curves. Hc monotonically decreases with increasing temperature, consistent with the TC calculated above. While far from exhaustive, this analysis supports the assertion that EuTiO3 is a well behaved ferromagnet for a biaxial strain of +1.1%.

Fig. 4.15 Temperature dependence of the remanent Kerr angle and coercivity for EuTiO3/DyScO3.

Error bars represent one sigma variation for temperature, θKerr, and Hc. The dashed line is a power law fit to θremanent and yields a TC of 4.24±0.02 K and a critical exponent of 0.42±0.02, consistent with the theoretical prediction of 0.5 for a three dimensional ferromagnet.

4.4.5 Quantifying MOKE with SQUID

Since the MOKE measurements of ferromagnetism are not on an absolute scale,

SQUID measurements are used to quantify the spontaneous magnetization of EuTiO3 strained by εs = +1.1%. These measurements are made in nominally zero residual

118 magnetic field to minimize the paramagnetic response of the substrate. The strained

EuTiO3 is cooled in a 100 Oe field, a field found sufficient to polarize the EuTiO3 film, yet sufficiently small to minimize the residual magnetic field during subsequent SQUID measurements. Clear hysteresis loops are observed below 4 K (see Fig. 4.16(a)), however, the magnetization does not saturate at 1.8K, probably due to some complication from the antiferromagnetic substrate at this temperature. The spontaneous magnetization of the strained EuTiO3 is seen to be large, i.e., several µB/Eu, close to the theoretical prediction of a spontaneous magnetization of 7 μB/Eu. The temperature dependence of the

SQUID measurement is plotted on top of MOKE result as comparison (Fig. 4.16(b)). The observed magnetization is seen to rise at the same temperature (TC) as the MOKE signal and follows it until the antiferromagnetic transition of the DyScO3 substrate at 3.1 K, where the substrate signal masks the magnetization of the strained EuTiO3 film, in agreement with the MOKE results. From the combination of these results, and since the ferromagnetism occurs well above the antiferromagnetic transition of the DyScO3 substrate, we conclude that the observed ferromagnetism is not correlated with the magnetic ordering of the underling substrate.

119

Fig. 4.16 (a) The isothermal SQUID measurement quantifying the magnetization of strained EuTiO3 on DyScO3 at T = 1.8 and 3.8 K. At 1.8 K, the magnetization reaches as high as 3 μB/Eu. (b)

Temperature dependence of the magnetization measured using both MOKE and SQUID. The red data points with error bars (representing one sigma variations for both temperature and θKerr) show the temperature dependence of the remanent value of θKerr for EuTiO3/DyScO3, θremanent. The observed magnetization in SQUID is seen to rise at the same temperature (TC) as the MOKE signal and follows it until the antiferromagnetic transition of the DyScO3 substrate at 3.1 K, where the substrate signal masks the magnetization of the strained EuTiO3 film. [13]

4.4.6 MOKE with Faraday set up

In the pursuit of the magnetization for EuTiO3 samples, a MOKE measurement with Faraday set up is also performed. Instead of detecting the polarization change of the reflected beam, when the sample is transparent enough, the polarization change of the transmitted beam could also be detected, as shown in Fig. 4.17 (a). Since the beam is going through the substrate layer and the substrate is much thicker comparing to the film

(which is tens of nm thick), the transmission geometry enables capturing most signal from the substrate layer. In this geometry, the direction of the magnetic field is perpendicular to the sample surface, as seen in Fig. 4.17(a). We conduct this

120 measurement in the EuTiO3/DyScO3 sample, and Kerr angle as a function of field at different temperatures are shown in Fig. 4.17(b). In this case, the detected Kerr angle is symmetric to the magnetic field. At T = 2K, a hysteresis loop is observed and it becomes very vague at T = 5K and disappears at T =10K. This symmetry to the magnetic field is attributed to the magnetorestriction property of the DyScO3 substrate. When the light passes through the material which has this property, the lattice constant of the structure is changed, therefore causes a change in the dielectric constant of the material and consequently rotates the light polarization. It is not related to the magnetism of the sample so it does not care about the sign of the magnetic field, resulting in a symmetric curve to the magnetic field. This observation may stimulate some research direction which uses MOKE to probe the magnetorestriction of certain materials.

Fig. 4.17 (a) Schematic showing MOKE measurement set up for Faraday geometry, where the polarization of a transmitted light is detected and therefore information on the whole substrate is gathered. (b) θKerr at different temperatures. Symmetric curves are observed and it is attributed to the magnetorestriction of the DyScO3 substrate.

121

4.5 Conclusions

By using magneto-optical Kerr effect, we successfully measured the ferromagnetism from a strained EuTiO3 films and our results confirm the theorized mechanism[9] and open the door to higher-temperature implementations of strong ferromagnetic ferroelectrics[10], which would allow for dramatic improvements in numerous devices and applications: magnetic sensors[22-23], energy harvesting, high- density multistate memory elements[23], wireless powering of miniature systems[24], and tunable microwave filters, delay lines, phase shifters and resonators[25-26]. This strong intrinsic multiferroic also has great potential in the applications for spintronics field. Moreover, the method of using MOKE as a sensitive probe for magnetism in perovskites structure thin films opens a door for a new category of measurement techniques in these systems.

122

4.6 Reference:

1. W. Eerenstein, N. D. Mathur, and J. F. Scott, Nature, 442, 759 (2006.)

2. R. Ramesh and N. A. Spaldin, Nat. Mater. 6, 21 (2007).

3. J. Y. Park et al., Appl. Phys. Lett. 91, 152903 (2007).

4. N. Ikeda et al., Nature 436, 1136 (2005).

5. J. -P. Rivera and H. Schmid, Helvetica Physica Acta, 54, 273 (1981).

6. T. Kimura et al., Nature 426, 55 (2003).

7. T. Lottermoser et al., Nature 430, 541 (2004).

8. T. Kimura et al., Nat. Mater. 7, 291 (2008).

9. C. J. Fennie and K. M. Rabe, Phys. Rev. Lett. 97, 267602 (2006).

10. J. H. Lee and K. M. Rabe, Phys. Rev. Lett. 104, 207204 (2010).

11. T. Goto et al., Phys. Rev. Lett. 92, 257201 (2004).

12. T. Katsufuji and H. Takagi, Phys. Rev. B 64, 054415 (2001).

13. J. H. Lee et al., Nature 466, 954 (2010).

14. J. H. Haeni et al., Nature, 430, 758 (2004).

15. J. H. Lee et al., Appl. Phys. Lett. 94, 212509 (2009).

16. R. C. Jaeger, Introduction to Microelectronic Fabrication. "Film Deposition"

(Upper Saddle River: Prentice Hall. 2002).

17. W. P. McCray, Nat. Nanotech. 2, 259 (2007).

18. T. Steiner, Semiconductor Nanostructures for optoelectronic applications.

(Boston: Artech House. 2004)

19. X. Ke et al., Appl. Phys. Lett. 94, 152503 (2009).

123

20. R. Uecker et al., J. Cryst. Growth 295, 84 (2006).

21. F. Ono et al., J. Alloys Compd. 317, 254 (2001).

22. J. Y. Zhai et al., Appl. Phys. Lett. 88, 062510 (2006).

23. M. Gajek et al., Nat. Mater. 6, 296 (2007).

24. A. Bayrashev, W. P. Robbins and B. Ziaie, Sensor Aatuat A-Phys. 114, 244

(2004).

25. Y. K. Fetisov and G. Srinivasan, Appl. Phys. Lett. 88, 143503 (2006).

26. J. Das et al., Adv. Mater. 21, 2045 (2009).

124

Appendix A: Two important aspects in MOKE measurement for EuTiO3

A.1 Design sample mount for in-plane magnetization

It was known that the easy axis of strained EuTiO3 film is in plane. To use

MOKE to probe the magnetism, longitudinal geometry should be used. For longitudinal geometry, if A.1(a) is used, the component of the linear polarization in the direction of the sample’s magnetization is very small (sinθ is small with small θ) and makes it difficult to detect the signal. Therefore, geometry in A.1 (b) was proposed, the beam is incident on the sample from the front big window and the reflected beam is coming out from the back big window. In this case θ is big and the detected component of the linear polarization in the direction of the sample’s magnetization is much larger. However, if the sample is mounted exactly at the centre of the sample space, due to the limited size of window, it makes it impossible for the optical alignment because the beam cannot hit the sample. To solve this problem, an off-centre sample mount in Fig. A.1(d) was designed and the geometry in Fig. A.1(c) was used for the measurement. The off-centre design of the mount puts the samples at a position where incident beam is able to hit the sample when it is coming from the front window and bigger component of the magnetization could be detected.

125

Fig. A.1 Schematic showing the geometry of the MOKE measurements (a) incident and reflect through the side small window. Small angle θ makes the signal too small to detect. (b) incident and reflect through the front and back big window. Big angle θ gives much bigger signal to be detected (c). sample mount off-centred makes the optical alignment possible while keeping big angle θ. (d) Picture showing the off-centred sample mount.

A.2 Controlling temperature in spectromag with superfluid He

When measuring MOKE of strained EuTiO3 sample, temperatures between 1.7K and 5K were needed. Temperatures in the VTI (variable temperature insert) are monitored through the temperature sensor that came with the cryostat as well as the thermo coupler installed in the sample mount. The temperatures below 2.17K are realized through the superfluid phase of liquid helium. To reach the superfluid phase, the VTI space of the cryostat is pumped with an oil-free rough pump with base pressure 0.05

126

Torr. The needle valve is then all the way opened. Large amount of liquid helium is fed to VTI from the helium space. The level of the liquid helium in the VTI could be directly watched through the optical windows. When the level reaches top of the cryostat window, needle valve is reduced to 10% open. As the pump is pumping to VTI space, the pressure in the VTI space drops and the temperature is decreased. When the temperature is lower than the lambda point of liquid helium, which is 2.17K, superfluid phase of helium is realized and the bubbles of liquid helium disappear, leaving a stable environment for the optical measurements. The exact temperature is determined by the initial amount of helium fed in and the pressure at the gas-liquid interface. Lower pressure gives lower temperature and larger initial amount of helium gives higher temperature. Therefore, these two parameters need to be controlled together to reach the desired temperature.

Temperatures between 2.17K and 5K are realized by controlling the pressure of

VTI space. In this case, a higher pressure in the VTI space is needed to reach the temperatures higher than 2.17K since higher interface pressure gives higher temperatures.

Because the pressure at the pump cannot be changed, the pressure control is then realized through the diaphragm VAT valve that connects the pump and VTI space. Above 2.17K the helium is not in superfluid and it will affect the optical measurement, the helium is only fed to the bottom of the VTI space and the heat transfer is realized through two thick copper wires at the bottom of the sample mount, as shown in Fig. A.2. The two copper wires are immersed into helium and the temperature of helium is transfer to the sample very quickly since the media are all metal. The diaphragm valve cannot give a very

127 precise and quantitative control of the pressure, therefore, one should adjust the valve with super carefulness to get the desired temperature. Even little touch of the valve could change the pressure so much that the temperature will be too much away from what we need. One could watch the temperature on the temperature sensor and adjust the VAT valve at the same time so that the pressure could be tuned accordingly. Again the exact temperature is determined by the initial amount of helium fed in and the pressure at the gas-liquid interface and these two need to be adjusted together.

Fig. A.2 Picture of the sample mount with two thick copper wires at the

bottom to transfer the temperature of helium at the bottom of the VTI

space to the sample.

128

All Reference

1.1. M. N. Baibich et al., Phys. Rev. Lett. 61, 2472 (1988).

1.2. G. Binasch et al., Phys. Rev. B 39, 4828 (1989).

1.3. C. Tsang et al., IEEE Trans. Magn. 30, 3801 (1994).

1.4. S. A. Wolf et al., Science 294, 1488 (2001).

1.5. D. D. Awschalom and M. E. Flatte, Nat. Phys. 3, 153 (2007).

1.6. S. Datta and B. Das, Appl. Phys. Lett. 56, 665 (1990).

1.7. B. T. Jonker, Proc. IEEE. 91, 727 (2003).

1.8. G. Schmidt et al., Phys. Rev. B 62, R4790 (2000).

1.9. E. I . Rashba, Phys. Rev. B. 62, R16267 (2000).

1.10. H. Zhu et al., Phys. Rev. Lett. 87, 016601 (2001).

1.11. A. T. Hanbicki et al., Appl. Phys. Lett. 80, 1240 (2002).

1.12. C. Adelmann et al., Phys. Rev. B. 71, 121301(R) (2005).

1.13. X. Jiang et al., Phys. Rev. Lett. 94, 056601 (2005).

1.14. W. Han et al., Phys. Rev. Lett. 105, 167202 (2010).

1.15. R. Fiederling et al., Nature 402, 787 (1999).

1.16. Y. Ohno et al., Nature 402, 790 (1999).

1.17. F. Meier and B. P. Zakharchenya, Optical Orientation Amsterdam, The

Netherlands: North-Holland, vol. 8. (1984).

1.18. I. Zutic, J. Fabian and S. Das Sarma, Rev. Mod. Phys. 76, 323 (2004).

129

1.19. M. E. Flatte, J. M. Byers and W. H. lau. Chap: Spin dynamics in semiconductors in

Semiconductor spintronics and quantum computation. edited by D. D. Awschalom, D.

Loss and N. Samarth Springer: New York (2002).

1.20. X. Lou et al., Nat. Phys. 3, 197 (2007).

1.21. N. Tombros et al., Nature 448, 571 (2007).

1.22. O. M. J. Van'tErve et al., Appl. Phys. Lett. 91, 212109 (2007).

1.23. I. Appelbaum et al., Nature 447, 295 (2007).

1.24. S. P. Dash et al., Nature 462, 491 (2009).

1.25. B. T. Jonker et al., Nat. Phys. 3, 542 (2007).

1.26. J. M. Kikkawa and D. D. Awschalom, Phys. Rev. Lett. 80, 4313 (1998).

1.27. S. A. Crooker et al., Science 309, 2191 (2005).

1.28. J. S. Colton et al., Phys. Stat. Sol. (B) 233, 445 (2002).

1.29. N. A. Spaldin and M. Fiebig, Science 309, 391 (2006).

2.1. International technology roadmap for semiconductors. Executive summary for 2009.

(2009).

2.2. D. P. Divincenzo, Science 270, 255 (1995).

2.3. J. M. Manriquez et al., Science 252, 1415 (1991).

2.4. V. N. Prigodin et al., Adv. Mater. 14, 1230 (2002).

2.5. J.-W. Yoo et al., Phys. Rev. B. 80, 205207 (2009).

2.6. J.-W. Yoo et al., Nat. Mater. 9, 638 (2010).

2.7. J.-W. Yoo et al., Phys. Rev. Lett. 97, 247205 (2006).

2.8. K. I. Pokhodnya, A. J. Epstein, and J. S. Miller, Adv. Mater. 12, 410 (2000).

130

2.9. A. L. Tchougreeff and R. A Dronskowski, J. Comput. Chem. 29, 2220 (2008).

2.10. H. Matsuura, K. Miyake, and H. Fukuyama, J. Phys. Soc. Jpn. 79, 034712 (2010).

2.11. J. S. Miller and A. J. Epstein, Angew. Chem. Int. Edit. 33, 385 (1994).

2.12. A. Zheludev et al., J. Am. Chem. Soc.116, 7243 (1994).

2.13. D. A. Dixon and J. S. Miller, J. Am. Chem. Soc. 109, 3656 (1987)

2.14. C. Tengstedt et al., Phys. Rev. B., 69,165208 (2004).

2.15. C. Tengstedt et al., Phys. Rev. Lett. 96, 057209 (2006).

2.16. G. C. De Fusco, et al., Phys. Rev. B. 79, 085201 (2009).

2.17. J. B. Kortright, et al., Phys. Rev. Lett. 100, 257204 (2008).

2.18. R. Fiederling et al., Nature 402, 787 (1999).

2.19. Y. Ohno et al., Nature 402, 790 (1999).

2.20. F. Meier and B. P. Zakharchenya, Optical Orientation Amsterdam, The

Netherlands: North-Holland, vol. 8. (1984).

2.21. A. T. Hanbicki et al., Appl. Phys. Lett. 80, 1240 (2002).

2.22. H. Zhu et al., Phys. Rev. Lett. 87, 016601 (2001).

2.23. C. Adelmann et al., Phys. Rev. B. 71, 121301(R) (2005).

2.24. G. Kioseoglou et al., Nat. Mater. 3, 799 (2004).

2.25. B. T. Jonker et al., Nat. Phys. 3, 542 (2007).

2.26. Y. Chye Phys. Rev. B. 66, 201301(R) (2002).

2.27. G. Schmit et al., Phys. Rev. B. 62, 4790(R) (2000).

2.28. I. Zutic, J. Fabian and S. Das Sarma, Rev. Mod. Phys. 76, 323 (2004)

2.29. E. I. Rashba Phys. Rev. B. 62, 16267(R) (2000).

131

2.30. A. Fert and H. Jaffres Phys. Rev. B. 64, 184420 (2001).

2.31. B. T. Jonker, Proc. IEEE. 91, 727 (2003).

2.32. J.-H. park et al., Phys. Rev. Lett. 81, 1953 (1998)

2.33. Editorial. Sumitomo’s PLED. III-Vs Rev.16, 27 (2003)

2.34. Q. Wang & D. Ma, Chem. Soc. Rev. 39, 2387 (2010)

2.35. H. Klauk, Chem. Soc. Rev. 39, 2643 (2010)

2.36. J. M. Kikkawa and D. D. Awschalom, Phys. Rev. Lett. 80, 4313 (1998).

2.37. W. H. Lau , J.T. Olesberg, and M.E. Flatte, Phys. Rev. B. 64, 161301 (1998).

2.38. J. H. Buss et al., Phys. Rev. B. 81, 155216 (2010).

3.1. V. A. Sih, E. Johnston-Halperin, and D.D. Awschalom, Proc. IEEE. 91, 752

(2003).

3.2. J. M. Kikkawa and D.D. Awschalom, Nature 397, 139 (1999).

3.3. J. M. Kikkawa et al., Science 277, 1284 (1997).

3.4. J. Berezovsky et al., Science 320, 349 (2008).

3.5. W. Lu and C. M. Lieber, J. of Phys. D- Appl. Phys. 39, R387 (2006).

3.6. J. F. Wang et al., Science 293, 1455 (2001).

3.7. A. Mishra et al. Appl. Phys. Lett. 91, 263104 (2007).

3.8. H. E. Ruda and A. Shik, Phys. Rev. B 72, 115308 (2005).

3.9. Y. H. Yu et al., Nano Lett. 8, 1352 (2008).

3.10. X. H. Lou et al., Nat. Phys. 3, 197 (2007).

3.11. S. A. Crooker et al., Science 309, 2191 (2005).

3.12. D'yakonov and Perel'. Sov. Phy. Solid State, 13, 3023 (1972).

132

3.13. R. J. Elliot, Phys. Rev. 96, 266 (1955).

3.14. Y. Yafet, Solid State Physics. edited by F. Seitz and D. Turnbull. Vol. 14.

(Acedemic, New York, 1963).

3.15. G. L. Bir, Sov. Phy. JETP 42, 1382 (1976).

3.16. A. A. Kiselev and K.W. Kim, Phys. Rev. B 61, 13115 (2000).

3.17. S. D. Ganichev et al., Phys. Rev. Lett. 92, 256601 (2004).

3.18. R. Winkler, Phys. Rev. B 69, 045317 (2004).

3.19. L. D. Landau and E. M. Lifshits, Electrodynamics of Continuous Media.

(Pergamon, New York, 1984).

3.20. J. L. Birman, Phys. Rev. Letts. 2, 157 (1959).

3.21. J. L. Birman, Phys. Rev. 114, 1490 (1959).

3.22. F. Meier and B. P. Zakharchenya, Optical Orientation Amsterdam, The

Netherlands: North-Holland, vol. 8. (1984).

3.23. P. Ils et al., Phys. Rev. B 51, 4272 (1995).

3.24. X. F. Duan et al., Nature 409, 66 (2001).

3.25. Handbook of Chemistry and Physics. 86th ed., edited by D. R. Lide, (CRC Press,

Taylor & Francis Group, New York, 2005).

3.26. H. E. Ruda and A. Shik, J. Appl. Phys. 100, 024314 (2006).

3.27. Y. Kravtsova et al., Appl. Phys. Lett. 90, 083120 (2007).

3.28. R. S. Wagner, Whisker Technology, edited by A. P. Levitt. (Wiley, New York,

1970).

3.29. A. M. Morales and C. M. Lieber, Science 279, 208 (1998).

133

3.30. J. L. Liu et al., J. Cryst. Growth 200, 106 (1999).

3.31. K. Haraguchi et al., Appl. Phys. Lett. 60, 745 (1992).

3.32. X. F. Duan and C. M. Lieber, Adv. Mater. 12, 298 (2000).

3.33. P. C. Chang et al., Chemistry of Mater. 16, 5133 (2004).

3.34. H. -Y. Chen et al., Optics Express 16, 13465 (2008).

3.35. R. J. Epstein et al., Appl. Phys. Lett. 78, 733 (2001).

3.36. J. Furst et al., Semicond. Sci. Technol. 20, 209 (2005).

3.37. Y. Masumoto et al., Phys. Rev. B 74, 205332 (2006).

3.38. J. S. Colton et al., Phys. Stat. Sol. (B) 233, 445 (2002).

3.39. O. Maksimov et al., Physica E 32, 399 (2006).

4.1. W. Eerenstein, N. D. Mathur, and J. F. Scott, Nature, 442, 759 (2006.)

4.2. R. Ramesh and N. A. Spaldin, Nat. Mater. 6, 21 (2007).

4.3. J. Y. Park et al., Appl. Phys. Lett. 91, 152903 (2007).

4.4. N. Ikeda et al., Nature 436, 1136 (2005).

4.5. J. -P. Rivera and H. Schmid, Helvetica Physica Acta, 54, 273 (1981).

4.6. T. Kimura et al., Nature 426, 55 (2003).

4.7. T. Lottermoser et al., Nature 430, 541 (2004).

4.8. T. Kimura et al., Nat. Mater. 7, 291 (2008).

4.9. C. J. Fennie and K. M. Rabe, Phys. Rev. Lett. 97, 267602 (2006).

4.10. J. H. Lee and K. M. Rabe, Phys. Rev. Lett. 104, 207204 (2010).

4.11. T. Goto et al., Phys. Rev. Lett. 92, 257201 (2004).

4.12. T. Katsufuji and H. Takagi, Phys. Rev. B 64, 054415 (2001).

134

4.13. J. H. Lee et al., Nature 466, 954 (2010).

4.14. J. H. Haeni et al., Nature, 430, 758 (2004).

4.15. J. H. Lee et al., Appl. Phys. Lett. 94, 212509 (2009).

4.16. R. C. Jaeger, Introduction to Microelectronic Fabrication. "Film Deposition"

(Upper Saddle River: Prentice Hall. 2002).

4.17. W. P. McCray, Nat. Nanotech. 2, 259 (2007).

4.18. T. Steiner, Semiconductor Nanostructures for optoelectronic applications.

(Boston: Artech House. 2004)

4.19. X. Ke et al., Appl. Phys. Lett. 94, 152503 (2009).

4.20. R. Uecker et al., J. Cryst. Growth 295, 84 (2006).

4.21. F. Ono et al., J. Alloys Compd. 317, 254 (2001).

4.22. J. Y. Zhai et al., Appl. Phys. Lett. 88, 062510 (2006).

4.23. M. Gajek et al., Nat. Mater. 6, 296 (2007).

4.24. A. Bayrashev, W. P. Robbins and B. Ziaie, Sensor Aatuat A-Phys. 114, 244

(2004).

4.25. Y. K. Fetisov and G. Srinivasan, Appl. Phys. Lett. 88, 143503 (2006).

4.26. J. Das et al., Adv. Mater. 21, 2045 (2009).

135