ABSTRACT
STUART, SEAN CLAYTON. Oxide-Graphene Interfaces for Graphene Spintronics. (Under the
direction of Dan Dougherty and Marc Ulrich.)
Graphene’s high carrier mobility and low spin-orbit scattering allow for efficient spin
transport, which has been demonstrated by several publications over useful length scales.
Spintronic devices require an oxide tunneling barrier to allow for efficient spin injection
from a magnetic contact and can employ magnetic oxide gates for spin manipulation. This
thesis concerns the production and characterization of oxide films for graphene based
spintronics.
Pulsed laser deposition (PLD) was used to grow thin, uniform MgO films on graphene of
suitable quality for tunneling barriers. This was an important result, improving on previous
deposition techniques significantly.
Progress toward more sophisticated spintronic devices requires controllable
manip-ulation of spin polarized charge carriers. We have identified Cr2O3as a material whose
magnetoelectric properties would enable voltage controlled switching of the exchange
interaction. Magnetoelectric Cr2O3films were produced by PLD. These films were
character-ized by x-ray diffraction, photoelectron spectroscopy and atomic force microscopy (AFM).
The magnetoelectric properties of Cr2O3were characterized by a novel combination of
electrostatic (EFM) and magnetic force microscopy (MFM). Magnetoelectric annealing was
used to produce varying sized magnetoelectric domains imaged by MFM. A local electric
field was applied with a conducting AFM tip, and the local switching of the polarization
© Copyright 2015 by Sean Clayton Stuart
Oxide-Graphene Interfaces for Graphene Spintronics
by
Sean Clayton Stuart
A dissertation submitted to the Graduate Faculty of North Carolina State University
in partial fulfillment of the requirements for the Degree of
Doctor of Philosophy
Physics
Raleigh, North Carolina
2015
APPROVED BY:
J.E. Rowe J.P Maria
Dan Dougherty
Co-chair of Advisory Committee
Marc Ulrich
DEDICATION
To my family, for their support, encouragement and motivation.
To my grandfather who suggested I become a Doctor. To my mother who explained
that this did not necessarily mean a medical doctor, and spent years teaching me math and
research skills. To my father for the wealth of technical and troubleshooting knowledge. To
my siblings for keeping me grounded.
To my in-laws for their support, childcare and countless meals which helped me
gradu-ate and allowed me to finish this thesis.
To my wife for the love, support, and active encouragement provided throughout my
graduate and undergraduate education. She pushed me to be the best version of myself.
Finally to my son who gives me incredible joy and who provided much needed distractions
BIOGRAPHY
The author developed an interest in physics while attending Bellport Sr. High School in
Bellport NY. There he attended an authentic research class which allowed him to read about
quantum computing and related physics topics. Following this interest he attended the
University of Maryland at Baltimore County for a bachelors in physics. While attending
UMBC and for a year afterwards he had the opportunity to work at the Naval Research
Lab where he developed skills in machining, solidworks modeling and vacuum equipment
ACKNOWLEDGEMENTS
I would like to thank my advisers for their years of support and enormously helpful
TABLE OF CONTENTS
LIST OF TABLES . . . viii
LIST OF FIGURES. . . ix
Chapter 1 Introduction. . . 1
1.1 Graphene . . . 1
1.2 Graphene Band Structure . . . 3
1.3 Graphene Production . . . 10
1.3.1 Graphene Production by Chemical Vapor Deposition . . . 11
1.3.2 Graphene Production From Silicon Carbide . . . 12
1.4 Spin Relaxation in Graphene . . . 14
1.4.1 Elliott-Yafet Spin Relaxation . . . 16
1.4.2 Dyakonov Perel Spin Relaxation . . . 17
1.4.3 Spin Relaxation Experiments . . . 19
1.5 Graphene Spintronics . . . 20
1.5.1 Tunnel Junctions . . . 20
1.5.2 Producing Tunnel Barriers on Graphene . . . 22
1.5.3 Non Local Spin Valves . . . 23
1.5.4 Hanle measurment . . . 25
1.5.5 Datta Das Device and the Rashba Effect . . . 26
1.5.6 Graphene Spin Field Effect Transistor . . . 27
1.6 Magnetoelectric Chromia . . . 29
1.6.1 Magnetoelectrics . . . 29
1.6.2 Chromium Oxide . . . 31
Chapter 2 Methods . . . 36
2.1 Atomic Force Microscopy . . . 36
2.1.1 Magnetic Force Microscopy . . . 43
2.1.2 Electrostatic Force Microscopy . . . 46
2.1.3 Asylum AFM MFP-3D . . . 47
2.2 X-Ray Photoelectron Spectroscopy . . . 49
2.2.1 X-Ray Generation . . . 51
2.2.2 Electron Energy Detection . . . 52
2.2.3 Thickness Estimation by Photoelectron Spectroscopy . . . 54
2.2.4 Riber XPS system . . . 55
2.3 X-Ray Diffraction . . . 56
2.4 Pulsed Laser Deposition . . . 57
2.4.2 Tuning the Deposition . . . 66
Chapter 3 Magnesium Oxide Deposition . . . 68
3.1 Abstract . . . 69
3.2 Introduction . . . 69
3.3 Methods . . . 71
3.4 Results . . . 72
3.4.1 AFM study of film morphology . . . 73
3.4.2 XRD study of film crystalinity. . . 75
3.4.3 XPS study of film chemistry . . . 75
3.5 Discussion . . . 79
3.5.1 Film morphology and Crystallinity . . . 79
3.5.2 Film Composition and Defects . . . 80
3.6 Conclusion . . . 81
Chapter 4 Chromium Oxide Deposition . . . 83
4.1 Abstract . . . 83
4.2 Introduction . . . 84
4.3 Methods . . . 85
4.4 Results . . . 86
4.4.1 AFM studies of Film Morphology . . . 86
4.4.2 XRD studies of Film Crystallinity . . . 92
4.4.3 XPS studies of Film Composition and Interface Chemistry . . . 92
4.5 Discussion . . . 96
4.6 Conclusion . . . 97
Chapter 5 Magnetoelectric Films. . . 98
5.1 Abstract . . . 99
5.2 Introduction . . . 99
5.3 Results & Discussion . . . 101
5.4 Conclusion . . . 107
5.5 Methods . . . 108
Chapter 6 The ME-Hanle Effect . . . 110
6.1 Introduction . . . 111
6.2 Weiss Molecular Field . . . 112
6.3 Effective Field . . . 113
6.4 ME Hanle Simulation . . . 115
6.4.1 Modeling a Hanle Measurement of Spin Transport in Graphene . . . 115
6.4.2 Modeling a Graphene Spin-FET with ME Gate . . . 118
BIBLIOGRAPHY . . . 124
APPENDICES . . . 135
Appendix A Riber Manual . . . 136
A.1 Procedure Outline . . . 137
A.2 Sample Introduction/Exchange . . . 138
A.3 X-Ray Operation . . . 139
A.3.1 X-Ray Interlocks . . . 139
A.3.2 X-Ray Startup Procedure . . . 140
A.3.3 X Ray Shutdown Procedure . . . 141
A.4 Data Acquisition . . . 141
A.5 Heater Operation . . . 142
A.6 Instrument Specifications RIBER LAS 3000 . . . 143
A.7 Adjustment and Troubleshooting . . . 144
A.7.1 Discriminator Adjustment . . . 144
A.7.2 Adjusting the Position . . . 144
A.7.3 X Ray Source Troubleshooting . . . 145
Appendix B Supplemental: ME Films . . . 148
Appendix C ME-Hanle Simulation Programs . . . 153
C.1 Hanle Simulation . . . 153
C.2 Spin-FET Simulation . . . 157
LIST OF TABLES
Table 2.1 List of important X-Ray lines produced by magnesium94. . . . 52
LIST OF FIGURES
Figure 1.1 3D band structure of graphene (left) and 2D band structure around the K point (right) based on the results of the preceding calculations 2
Figure 1.2 2D hexagonal structure of graphene. Carbon atoms shown in green, with dashed lines outlining a primitive unit cell containing 2 carbon atoms. Thea1anda2vectors define the unit cell of graphene. . . 4
Figure 1.3 AFM of graphene grown on copper foil. The right image shows topog-raphy which is dominated by the copper foil morphology. The left image shows the phase contrast indicating regions with graphene in red and without graphene in blue. The graphene is slightly (< 1nm) higher in the topography image however the phase contrast is necessary to locate the boundaries. . . 12
Figure 1.4 AFM image of a commercial graphene/Si sample. The roughess shown is attributed to PMMA residue on the graphene. . . 13
Figure 1.5 AFM image of graphene grown on silicon carbide (SiC) by evaporation in ultra high vacuum. Left shows a large scale image. The major features are SiC steps. The image at right is a zoom in showing the single layer and brighter bilayer graphene. . . 14
Figure 1.6 Diagram of the Elliott-Yafet (EY) and Dyakonov-Perel (DP) effect. The left side shows the EY effect where the spin ocassionally flips along the conduction path. The green line shows the linear dependence of spin relaxation on number of scattering events characteristic of EY effect. The right side shows the DY effect where the spin gradually precesses along the conduction path. The Red line shows the inverse dependence of spin relaxation on number of scattering events charac-teristic of DP effect. The spin relaxation in materials with both effect is still under debate as illustrated by the question mark. Diagram from Boross et. al33. . . . 16
Figure 1.7 Spectral gap via spin-orbit interactions (2λl) induced by a∆shift of a
carbon atom out of the plane. The red and blue lines show the results with and without including d-orbitals. Figure from Gmitra et al41. . . 19
Figure 1.8 Mott model of charge transport across an interface between a fer-romagnet and a semiconductor. The spin transport is modeled by separating the current into a spin up (I↑) and a spin down(I↓) channel. The effective resistance (R↑,R↓) in the magnet is spin dependent. The large difference between (R↑- R↓) and RS C makes the spin injection
Figure 1.9 A. SEM image of a non local spin valve on graphene. B. Side View diagram showing the device geometry. C and D. the voltage of spin up vs spin down electrons with the film magnetizations shown by black arrows. The figure is from Tombros et. al.44 . . . . 24
Figure 1.10 Datta-Das device, an electron wave analog of the electro-optic mod-ulator52 . . . . 27
Figure 1.11 Graphene Spin-FET geometry from Semenov et. al.55. A Ferromag-netic Dielectric (FMD) is used to alter the polarized electron spins, injected at the source S, in the graphene channel. The effect is tuned through the gate voltage Vg and detected at the source D. The arrows
in the graphene channel illustrate the spin precession induced by exchange interactions with the FMD. . . 28
Figure 1.12 Diagram of the time reversal and spatial inversion symmetries of different magnetic and multiferroic orders. When both time reversal and spacial inversion symmetries are present multiferroic proper-ties (including magnetoelectric properproper-ties) are present. Figure from Eerenstein et. al.56 . . . . 31
Figure 1.13 Chromium (III) oxide structure. The chromium atoms are shown in green and oxygen shown in grey. The chromium atoms are arranged in two sub-lattices, one with blue and one with red arrows corre-sponding to up and down magnetic polarization. The terminating chromium atoms are are aligned with the same magnetization. . . 32
Figure 1.14 Spin polarized UV photoelectron spectroscopy of a Cr2O3film on sapphire. The red and blue data show alternate spin polarization of collected electrons. The green data are the difference spectrum, revealing the existence of a spin polarized state near the Fermi level. Spectrum reproduced from He et. al.60 . . . . 33
Figure 2.1 Schematic of the inter-atomic forces involved in atomic force mi-croscopy. The tip - sample interaction is dominated by the atomic forces between a terminating atom (or group of atoms) of a tip and the atoms at the surface of a sample. The tip is moved along the contour B by tracking these forces and adusting the tip (or sample) position. The atom A is drawn to show how atomic forces can be probed in an STM measurment when a atom transduces a atomic force to a tunneling current signal. Illustration from Binnig et. al64. . 37
Figure 2.2 An AFM tip that is large (radius R2) compared to the surface features (radius R1) will not resolve the topography correctly. In this case the tip geometry is a dominant feature in the measured topography (dashed line). Diagram from Westra et. al73. . . . 39
Figure 2.3 Schematic of atomic force microscopy with optical deflection mea-surment. Left shows the tip deflected up where the laser reflection is measured by the blue detector. Right shows a less deflected tip where the reflection moves to the red detector. . . 39
Figure 2.4 The choice driving frequency relative to the harmoic frequency of a AFM tip determines wether the measured forces are attractive or repluslive. Figure from Seo et. al67. . . . 43
Figure 2.5 An example of magnetic force microscopy. The red and blue features are magnetic bits encoded into a hard drive platter. . . 44
Figure 2.6 Magnetic Force variant of Atomic Force Microscopy. The red is a magnetic film on the tip and sample. The magnetic force between the tip and sample are probed by offsetting the tip from the surface, decreasing the effects of shorter range inter-atomic forces. The phase shift of the tip oscillation is related to the magnetic force. . . 45
Figure 2.7 The Asylum AFM showing the scan head and the sample stage. Pic-ture from Asylum Research86 . . . . 48
Figure 2.8 The Asylum AFM showing the optical path and z- piezoelectric inside the scan head. Picture from Asylum Research86. . . . 49
Figure 2.9 Hemispherical electron analyzer geometry. Electrons with selected energy (green) will pass through the electric field of the Hemispheres and be detected. Electrons with energy too high (red) or too low (blue) will be discarded. . . 53
Figure 2.10 Photo of the Riber XPS system. . . 55
Figure 2.12 Time evolution of the plasma produced in a pulsed laser deposition system. The top-left image shows the location of the target, laser and the heater (where the sample is placed). The plasma expands over 15
µs (top panels) untill it hits the heater and rebounds to fill the target -heater area. Image from Lowndes et. al.98 . . . . 58
Figure 2.13 Plot of velocity of ablated particles as a function of laser fluence. Plot from Tasaka et. al100. . . . 59
Figure 2.14 Plot of velocity of ablated particles as a function of chamber pressure and distance from target. Plots from Riabinina et. al101. . . . 60
Figure 2.15 Image of pulsed laser deposition system used in this research. . . 62
Figure 2.16 Diagram of pump system used in PLD chamber. The bottom port of the chamber is connected to a turbo pump which is separated by a gate valve. Two smaller metal valves allow for the indtroduction of nitrogen for venting and for use of a secondary pump used for the initial chamber evacuation. A O2leak valve allows for controlled introduction of oxygen. . . 63
Figure 3.1 AFM images showing the initial stages of MgO growth on HOPG. All three depositions grown at 10−5Torr of O
2at room temperature. a) 10 shots, 50 Hz, at 380 mJ laser energy. Small∼5Ahigh islands are uniformly spread across the surface rms∼0.07 nm b) 50 shots, 50 Hz, 400 mJ 50% covered surface with the MgO islands<1 nm in height. rms 0.4 nm c) 75 shots, 50 Hz 400 mJ. A completely coverage surface, the pinholes seen in b) are filled in by the additional MgO rms∼0.1 nm. . . 72
Figure 3.2 RMS roughness of MgO on HOPG as a function of oxygen pressure during deposition. All films were deposited with 100 shots at 400 mJ and 50 Hz at room temperature. . . 74
Figure 3.4 XRD spectrum and AFM image from∼100 nm MgO films deposited on HOPG at elevated substrate temperatures. A) XRD of films grown at room temperature, 150◦C, and 200◦C showing emergence of MgO (200) peak at elevated temperatures. The unlabeled peaks are present in the XRD spectrum of the HOPG substrate. B) AFM image of MgO features. C) AFM line profile of 200nm high MgO features observed at elevated temperatures. . . 76
Figure 3.5 XPS spectra of a MgO film grown on HOPG with 100 shots at 400 mJ and 50 Hz at room temperature and 10−5Torr of oxygen before and after a 300◦C anneal for 30 min in UHV. The shaded region marks the hydroxide peak and the dashed line the O-Mg peak. The larger hydroxide peak at 30◦photoelectron emission angle as compared to normal emission indicates that the hydroxide is predominantly on the surface. Final state effects are anticipated as the cause for the shift in the hydroxide peak after annealing. . . 77
Figure 3.6 XPS (Mg 1s) spectra of a MgO Film grown on HOPG with 100 shots at 400 mJ and 50 Hz at room temperature and 10−5Torr of oxygen before and after a 300◦C anneal for 30 min in UHV. . . . 78
Figure 4.1 AFM images illustrating the growth of Cr2O3on HOPG at 100◦C with a roughness of 0.8 nm RMS. Topography is shown at left and associated line profile is on the right . . . 86
Figure 4.2 AFM images illustrating the topography of Cr2O3on a HOPG sample at 400◦C. The large scale features are due to the HOPG topography. . 87
Figure 4.3 AFM images illustrating the growth of Cr2O3on a SiC - Graphene sam-ple at 400◦C. A. The topography after deposition. B. The topography of the initial SiC- graphene sample . . . 88
Figure 4.4 AFM images illustrating the growth of Cr2O3on HOPG at 600◦C with increasing film thicknesses. A, and associated line profile B, show the initial growth is seen to nucleate in 10-20 nm high islands. In C and D a nominally 100 nm thick film is shown, with a roughness of 0.7 nm RMS. A nominally 300 nm film is shown in E and F, with a mainly smooth surface is interrupted by deep gaps. . . 89
Figure 4.5 AFM images illustrating the growth of Cr2O3on HOPG at 600◦C with a post anneal at 1000◦C. Topography A, and associated line profile B. 90
Figure 4.7 XRD of a film of Cr2O3grown on HOPG at 400◦C. The peak shown at 39.75◦is consistent with Cr2O3(0001) crystalinity. The other XRD peak is from HOPG via a tungsten X-ray line . . . 93
Figure 4.8 XPS of a film of Cr2O3grown on HOPG at 400◦C. The peak shape is characteristic of Cr2O3 . . . 94
Figure 4.9 XPS performed on a partial film of Cr2O3films on HOPG deposited at various temperatures. The top spectrum is a large area scan showing a single C 1s peak considtent with sp2bonds. Detailed view (bottom) show a small attenuation in theπ∗shakeup peak after deposition. . . 95
Figure 5.1 Structural and chemical properties of Cr2O3films grown on HOPG. Figure 5.1A shows the topography of a chromium-oxide films grown on hopg with major features representative of the HOPG surface, and particulates common to the PLD process. Figure 5.1B shows a Cr2O3 <0006>peak with a FWHM of 0.32◦for a film grown at 400◦C. Figure
5.1C shows Cr 2p XPS with a structure characteristic of Cr2O3. . . 102
Figure 5.2 Magnetoelectric domains produced by magnetoelectric annealing in
∼1T applied magnetic field and variable electric fields. Figure 5.2A shows small domains (∼2µm) produced by a 1.75 kV/cm electric field. Figure 5.2B shows larger domains (∼5µm) produced by 1.875 kV/cm electric field. Figure 5.2C shows a much larger (∼ 15µm) domain produced by a 2kV/cm field. . . 103
Figure 5.3 Results of applying a local electric field varying from 2.5V to -2.5V with a conducting AFM tip in contact with the surface. Figure 5.3A, B, and C shows the topography, EFM in red and MFM in blue. Figure 5.3D shows line profiles from the MFM and EFM in Figures 5.3B and C. 104
Figure 6.1 Chromium - graphene lattice mismatch. The graphene is shown in blue and the chromium atoms are in green. . . 114
Figure 6.2 A simulation of a Hanle measurment of spin transport in graphene. The governing equation (Equation 6.3) is used with the values from Han et. al30. in Equation 6.5. A scale factor is applied to obtain numerical agreement with previously published values of∆RN L30. . 117
Figure 6.4 Calculation of a graphene - Cr2O3Spin-FET device performance. The non-local resistance is plotted as a function of applied voltage. A change in resistance of severalΩis easily measurable for a reason-able electric field. The device modeled assumes a 100nm Cr2O3 thick-ness, a 5µm spin transport length and tranport properties matching
published values30. . . 119
Figure 6.5 Part (a) The Hanle measurment device geometry where spin current is injected at E2and a change in voltage is measured between E3and E4. By adding a resistor Rs e n s e the Vs e n s e is translated into a current (Io u t). The measured Vs e n s e is shown in part (b) as a function of the Rs e n s e and the applied magnetic field. The measured Io u t is shown in part (c), showing a very similar signal to Vs e n s e. Figure from Wen et. al.140. . . 121
Figure A.1 Vacuum diagram of the Riber XPS . . . 137
Figure A.2 Peak resolution at different focusing settings . . . 144
Figure A.3 Pins on filament connection to X-Ray Gun . . . 145
Figure A.4 X Ray interlock flowchart . . . 147
Figure B.1 AFM, EFM and MFM of a pattern prepared on 40±10nm Cr2O3-HOPG sample. Vertical bars were patterned magnetoelectric writing using biases of 2, 1, -1 and -2 V. Image A is an AFM image showing the film topography in the area used for all images shown. B and C are EFM images taken with a non-magnetic tip charged to+0.5V and -0.5V respectively. D,E and F are MFM images taken with a magnetic tip charged to 0,+0.5 and -0.5V respectively. The lack of change in the relative contrast for MFM images D,E and F shows that contrast is primarily magnetic as opposed to electrostatic. . . 149
Figure B.2 AFM, EFM and MFM of two Cr2O3films deposited on a graphene -SiC substrate. A NSCU logo is used to define a magnetic pattern in the film shown at right. Image A shows the topography in gold, image B shows the EFM in red and image C shows the MFM in Blue. A linear ramp in voltage is used to form a pattern in a second sample at left. Image D shows the topography, image E shows the EFM and image F shows the MFM. . . 150
Figure B.4 XPS-based thickness calibration for Cr2O3films grown on HOPG by PLD. The calibration curve features 4 films grown under 400◦C in 100 mTorr of O2. Each of the points represents the ratio of the carbon 1s peak area to the chromium 2p32peak area . The decay was fitted toAe−tx +y
0. The calculated decay constant (t) of 4900±1300 was determined to be the number of shots corresponding to the mean free path of the electrons within the film (corrected for the 15◦take-off angle of phototelectron leaving the film). The constanty0accounts for trace carbon contamination of the samples and the constant A accounts for the sensitivity and density factors. The mean free path of the electrons in the Cr2O3films was taken to be 2.0 nm in the NIST database95. The thickness of the film corresponding to 4900±1300 shots is 2.0±0.5 nm. . . 152
Figure C.1 A simulation of a Hanle measurment of spin transport in graphene. The governing equation (Equation 6.3) is used with the values from Han et. al30. shown in Equation 6.5. A scale factor is applied to the calculate results correspond to the measured values of RNL previously published30. . . 154
Figure C.2 Calculation of a graphene - Cr2O3Spin-FET device performance. The non-local resistance is is plotted as a function of applied voltage. A change in resistance of severalΩis easily measurable for a reasonable electric field. The device modeled assumes a 100nm Cr2O3thickness, a 5µm spin transport length and tranport properties matching pub-lished values30. . . 157
Figure D.1 The effect of capactive forces on the average phase in EFM images. The phase shift parabolically dependent on the tip bias as expected. 161
CHAPTER
1
INTRODUCTION
1.1
Graphene
The isolation of Graphene by Gaim and Novoselov1in 2004 spawned a remarkable amount
of research into the electrical and physical properties of two dimensional (2D) materials2,3.
Graphene has a 2D Electron Gas (2DEG) because it is a truly 2D material and, under ideal
conditions, a semimetal. The Graphene band structure can be see in Figure 1.1. Near
1.1. GRAPHENE CHAPTER 1. INTRODUCTION
the dispersiond E/d K is linear near the fermi level the effective mass of the electrons is
zero. This property of graphene makes electrons behave as massless dirac fermions4. This
is a rare example of quantum relativistic effects observable in a solid state experimental
system4.
Figure 1.1.3D band structure of graphene (left) and 2D band structure around the K point (right) based on the results of the preceding calculations
The high mobility in graphene makes it appealing for a range of electronic applications.
The highest measured electron mobility in graphene is 200,000 cm2V−1s−1in a suspended
graphene sample5. However, for most methods of graphene production the effects of
substrates, defects, and impurities limit the mobility significantly6. Several methods of
pro-ducing graphene have been developed to obtain high quality graphene including solution
processing7, and CVD for on various materials8,9. There is usually a trade-off between high
1.2. GRAPHENE BAND STRUCTURE CHAPTER 1. INTRODUCTION
exfoliation, including stacks of different 2-D materials10, however this method has a very
low yield and is not suitable for large scale production.
In addition to graphene’s high mobility, the nature of the band gap in graphene is of
interest for applications replacing a traditional semiconductor11. Graphene has no intrinsic
band gap, since the occupied and unoccupied band structure meet at or near the fermi
energy. This fermi energy changes depending on the level of electrostatic or chemical
doping. This is an issue for devices such as a transistor, where a band gap determines the
"on-off" threshold voltage. There has been significant work towards inducing a band gap12,
however this necessarily reduces mobility. Instead of modifying graphene to replace a
traditional semiconductor, applications that do not require a band gap are immediately
accessible with graphene and are able to use its full potential. For example, graphene’s
high mobility combined with its low spin-orbit interaction make it an ideal material for
spintronics.
1.2
Graphene Band Structure
To understand the electrical properties of graphene, it is necessary to calculate it’s band
structure13,14. The crystal structure of graphene is a hexagonal lattice with a carbon-carbon
distance of 1.42Å. This structure is shown in Figure 1.2, the dashed lines represent the
diamond shaped primitive lattice which contains 2 carbon atoms. Carbon has a 1s22s22p2
electron structure. In graphene each carbon atom forms threes p2 bonds, leaving one
conduction electron per atom. In this section we calculate and plot the band structure
1.2. GRAPHENE BAND STRUCTURE CHAPTER 1. INTRODUCTION
1
2
a
1a
23
4
5
6
1.2. GRAPHENE BAND STRUCTURE CHAPTER 1. INTRODUCTION
band structure shown in Figure 1.1 is calculated using a linear combination of atomic
orbitals (LCAO) approximation, which begins with the physical structure of the material.
To calculate the band structure a Bloch function is used:
ψk~=
X
~
R∈G
eik~·R~φ(x~−R~) (1.1)
WhereG denotes the set of all lattice vectors,R~is a specific lattice vector,k~is the
momen-tum, andφis an atomic wave function andφ(x~)is a linear combination of thepz orbitals
of both carbon atoms in the primitive cell. We can label these asφ1andφ2so thatφ(x~)is
defined as:
φ(x~) =b1φ1(x~) +b2φ2(x~) (1.2)
where b1and b2are the respective amplitudes.
The Hamiltonian for an electron in graphene (at positionx~) is then given by:
H = p~
2
2m +
X
~
R∈G
Va t(x~−x~1−R~) +Va t(x~−x~2−R~)
(1.3)
where thex~1andx~2are the positions of the two carbon atoms in the unit cell. TheR~term
describes the set of vectors that translate between equivalent lattice sites.
1.2. GRAPHENE BAND STRUCTURE CHAPTER 1. INTRODUCTION φ1 H φ1
=ε1
φ1 φ1 + φ1 X ~
R6=0
Va t(x~−x~1−R~) +Va t(x~−x~2−R~)
+Va t(x~−x~2)
! φ1
(1.4)
Whereε1is the energy ofφ1,Va t is the potential between the location atx~and a specific
carbon atom. Theφ1atomic wavefunction is located at first atom in the unit cell, so the sum
does not includeR~=0 to prevent double counting that term. Equation 1.4 can be simplified
using gauge invariance to setε1,2=0 and defining the sum (in the large parentheses) as
U1,2corresponding toφ1,2.
To calculate the band structure we need to solve the Schrodinger equation:
Hψk~
=E(k~)ψ~k
(1.5)
We can do this by writing theψwave function in the basisφas shown:
ψk~
=X i φi ψk~
(1.6)
Using this basis the Schrodinger equation (Equation 1.5 changes to:
HψK~
=Hφi
ψk~
= φi Ui ψk~
(1.7)
The first term to calculate isφ1
ψk~
:
φ1
ψ~k
= φ1
eik~·~0+ei~k·a~1+eik~·a~2. . .+eik~·a~n b 1
φ1
+b2
φ2
(1.8)
1.2. GRAPHENE BAND STRUCTURE CHAPTER 1. INTRODUCTION
nearest neighbor interactions. This allows us to set theeik~·a~n>2 terms in Equation 1.8 to zero.
We will consider the interaction between the wavefunctionφ1and each ofφ2,φ2(x+a1),
andφ2(x+a2). These interactions take into account the nearest neighbors from the number
1 lattice site shown in Figure 1.2. The termsφ∗
1andb1φ1are evaluated only atR =0 to
avoid including second order interactions. The results of this approximation are:
φ1
ψ
=b1+b2
Z
φ∗ 1φ2
1+e−ik~·a~1+e−i~k·a~2=b
1+b2γα(k~) (1.9)
φ2
ψ
=b2+b1
Z
φ∗ 2φ1
1+eik~·a~1+eik~·a~2=b
2+b1γα∗(k~) (1.10)
Here the integrals are equivalent due to symmetry and renamedγ. These integrals
calcu-late the overlap between nearest neighborpx orbitals. The sum of exponentials in both
equations are namedα(k~). The last difference between these equations is the sign of the
exponential. This can be explained by the difference in sign of the basis vectors needed to
go from the first atom in the unit cell. From Figure 1.2 one can see that the first atom in the
unit cell, labeled 1, has atoms 3 and 4 for nearest neighbors. Atoms 3 and 4 are equivalent
to atom 2, with the subtraction of vector a1and a2.
The right hand side of the Schrodinger equation (Equation 1.7) is:
φ1 U1 ψ = φ1 U1
1+e−ik~·a~1+e−ik~·a~2 b 1
φ1
+b2
φ2
(1.11)
1.2. GRAPHENE BAND STRUCTURE CHAPTER 1. INTRODUCTION φ1 U1 ψ =
(b1
Z
φ∗ 1U1φ1
+b2
Z
φ∗ 1U1φ2
1+e−ik~·a~1+e−ik~·a~2. (1.12)
We can remove theRφ1∗U1φ1term, because there is (trivially) no potential difference betweenφ1andφ1. For simplicity we can defineη=
R
φ∗
1U1φ2. And due to symmetry:
η=R
φ∗
2U2φ1=
R
φ∗
1U1φ2, so we can easily write the results of
φ2 U2 ψ
as well as:
φ1 U1 ψ
=b2ηα(k~), (1.13)
and φ2 U2 ψ
=b1ηα(k~). (1.14)
Using these equations along with equations 1.9 and 1.10 and applying them to Schrodinger
equation stated in Equation 1.7 results in:
E(k~) b1+b2γα(k~)
=b2ηα(k~), (1.15)
and
E(k~) b2+b1γα(k~)
=b1ηα(k~) (1.16)
.
This can be written in matrix form:
E(k~) α(k~)(E(k~)γ−η)
α∗(k~)E(k~)γ E(k~)
1.2. GRAPHENE BAND STRUCTURE CHAPTER 1. INTRODUCTION
Taking the determinant yields:
E2(k~)−α(k~)
2
(E(k~)γ−η)2=0. (1.18)
Expanding:
E2(k~)−α(k~)
2
(E2(k~)γ2+η2−2E(k~)γη) =0. (1.19)
In the tight binding approximationγis small, so we set the terms containingγto zero
and obtain:
E2(k~)−α(k~)
2
η2=
0. (1.20)
The following equations show the steps taken to simplify this result.
E(k~) =±α(k~)
η (1.21)
E(k~) =±ηÇ 1+e−ik~·a~1+e−ik~·a~2 1+eik~·a~1+eik~·a~2 (1.22)
1.3. GRAPHENE PRODUCTION CHAPTER 1. INTRODUCTION
E(k~) =±ηp3+eik~·a~1+e−ik~·a~1+e−ik~·a~2+eik~·a~2+e−ik~·(a~1−a~2)+eik~·(a~1−a~2) (1.24)
From here we use the identity stated in Equation 1.25 to re-write the equation in terms of
cosine functions, as shown:
2 cos(x) =ei x+e−i x (1.25)
E(k~) =±η
Ç
3+2 cos k~·a~1
+2 cos k~·a~2
+2 cos k~·(a~1−a~2)
(1.26)
Equation 1.26 is a good approximation of the band structure of graphene. A plot of this
equation is shown in Figure 1.1. This band structure shows the linear dispersion which is
characteristic of graphene.
1.3
Graphene Production
The dominant methods of graphene production are; Chemical Vapor Deposition (CVD),
de-composition of silicon carbide16and exfoliation. Exfoliation is accomplished by repeatedly
cleaving a Highly Oriented Pyrolitic Graphite (HOPG) crystal using adhesive tape. This was
the first method of obtaining few layer graphene flakes, and the method is useful for several
2-D materials such as hexagonal boron nitride17and molybdenum selenide10. The method
is versatile and inexpensive, producing high quality graphene, however it is particularly
1.3. GRAPHENE PRODUCTION CHAPTER 1. INTRODUCTION
located via optical microscopy, and the size, location and number of layers transferred is
random.
1.3.1
Graphene Production by Chemical Vapor Deposition
The need for wafer scale graphene motivated the development of other methods of graphene
growth. Graphene is most commonly produced commercially by CVD. This process was
first used to grow graphene on a Ni foil18, however copper foil allows more control and
better quality single layer graphene growth19. The CVD process uses a hydrocarbon gas
(such as methane) which is "cracked" by interaction with a hot metal substrate, removing
the hydrogen and leaving behind carbon. The challenges in this method are producing
uni-form graphene with a large grain size. The substrate purity, uniuni-formity, and the particulars
of temperature, pressure, time, and methane flow all have an effect on the growth.
I have produced graphene by this method, and characterized it by AFM and XPS. The
topography of the copper dominates the standard AFM image, making it difficult to resolve
the graphene on the surface, as shown on the left of Figure 1.3. However phase contrast
reveals the graphene by resolving the chemical difference between the graphene and copper
substrate.
For most applications, the CVD graphene needs to be transferred from the metal foil to
a more suitable substrate. A substrate is chosen to electrically isolate and thereby use the
graphene’s electrical properties. The usual choice is a silicon wafer with a significantly thick
thermal oxide layer. To transfer the graphene, a polymer (PMMA or similar) is used to hold
the graphene while the copper is etched away. These polymer and wet chemistry processes
1.3. GRAPHENE PRODUCTION CHAPTER 1. INTRODUCTION
Figure 1.3.AFM of graphene grown on copper foil. The right image shows topography which is dominated by the copper foil morphology. The left image shows the phase contrast indicating regions with graphene in red and without graphene in blue. The graphene is slightly (<1nm) higher in the topography image however the phase contrast is necessary to locate the bound-aries.
graphene. The topography is also affected, as shown in the topography of commercially
obtained graphene on a Si/SiO2wafer (Figure 1.4).
1.3.2
Graphene Production From Silicon Carbide
Graphene films produced by CVD typically have too much disorder and have a grain size
too small to obtain the mobility desired for many applications. Epitaxial growth of graphene
on silicon carbide (SiC) produces higher quality graphene. Since it is an epitaxial growth,
the graphene has no rotational disorder, and since it is a wafer scale process, it has the
potential to produce large, uniform areas of graphene. SiC is also a useful substrate since it
has relatively low conductivity eliminating the need for transfer. An example of graphene
grown on SiC that I have produced at NCSU is shown in Figure 1.5. I produced graphene
1.3. GRAPHENE PRODUCTION CHAPTER 1. INTRODUCTION
Figure 1.4.AFM image of a commercial graphene/Si sample. The roughess shown is attributed to PMMA residue on the graphene.
kept below 1x10−9Torr. This process was developed on our system by Dr. Andreas Sandin20.
At 1300◦C the silicon evaporates leaving behind a carbon rich surface, this process is slower
and more controllable at the silicon rich termination of the SiC wafer as compared to the
carbon terminated surface. The initial carbon layer is still partially bound to the silicon and
is called a buffer layer. When more silicon leaves via evaporation, the buffer layer decouples
from the silicon and becomes graphene. A new “buffer layer” is formed at the interface
between the graphene and the SiC bulk.
This growth method was first used at Georgia Institute of Technology the used to create
graphene films used to demonstrate ballistic transport in graphene, among other unique
electrical properties21,22. It produces high quality graphene that has been used to create
1.4. SPIN RELAXATION IN GRAPHENE CHAPTER 1. INTRODUCTION
Graphene produced in the manner is partially isolated from the silicon carbide but
there are some substrate effects in the graphene’s transport properties. The lattice
mis-match causes ripples in the graphene, affecting it’s electron transport. For use in the most
demanding applications, the buffer layer needs to be passivated. Hydrogen24,25and other
light elements26have been used to bond with the buffer layer and reduce it’s effect on the
graphene. In addition, bilayer graphene has also been used to separate the top graphene
channel from the buffer layer. Two layer graphene on SiC has been shown to exhibit
ex-tremely good spin transport properties27.
Figure 1.5.AFM image of graphene grown on silicon carbide (SiC) by evaporation in ultra high vacuum. Left shows a large scale image. The major features are SiC steps. The image at right is a zoom in showing the single layer and brighter bilayer graphene.
1.4
Spin Relaxation in Graphene
In general spin relaxation occurs when an electron precesses in a non uniform magnetic
1.4. SPIN RELAXATION IN GRAPHENE CHAPTER 1. INTRODUCTION
magnetic field from a magnetic material. But a non-uniform field from impurities,
spin-orbit interactions, or symmetry considerations will cause the spin polarization of charge
carriers to decay.
In many materials the spin-orbit (SO) interactions dominate the spin relaxation process.
The spin orbit interaction can be understood by looking at the electric field from the nucleus
as experienced by an electron. The electron moving in an electric field experiences an
effective magnetic field:
B=
1
c
~
E ×v~ (1.27)
This magnetic field interacts with the electron spin as expected. The primary parameter in
this interaction is the atomic number (Z) of the material. The electric field is proportional
to eZ, and therefore the SO interaction is proportional to the atomic number. There are
extrinsic effects that can cause SO interactions that are not directly related to the electric
field from the nucleus. Two examples are Dresselhaus SO coupling, which comes from
bulk inversion asymmetry28and Rashba SO interactions which appears due to inversion
asymmetry in 2D structures29. Both of these mechanisms may influence spin transport in
graphene devices.
Since graphene, a very low Z material, has a negligible intrinsic SO interaction it should
have a very high spin lifetime. However the measured spin lifetimes in graphene on the
order of 2ns are much shorter than the∼1µs theoretically predicted30. This observed
spin lifetime is comparable to many other semiconductors, however the spin diffusion
length and spin signals are larger in graphene due to the higher mobility30. To understand
1.4. SPIN RELAXATION IN GRAPHENE CHAPTER 1. INTRODUCTION
contributing to the spin relaxation in graphene.
The most often discussed graphene spin relaxation mechanisms are the Elliott-Yafet(EY)
Mechanism31, the Dyakonov-Perel (DP) Mechanism32which are described in Figure 1.6
and in the following sections.
Figure 1.6.Diagram of the Elliott-Yafet (EY) and Dyakonov-Perel (DP) effect. The left side shows the EY effect where the spin ocassionally flips along the conduction path. The green line shows the linear dependence of spin relaxation on number of scattering events characteristic of EY ef-fect. The right side shows the DY effect where the spin gradually precesses along the conduction path. The Red line shows the inverse dependence of spin relaxation on number of scattering events characteristic of DP effect. The spin relaxation in materials with both effect is still under debate as illustrated by the question mark. Diagram from Boross et. al33.
1.4.1
Elliott-Yafet Spin Relaxation
The Elliott-Yafet (EY) mechanism describes the change in spin polarization from
interac-tions with impurities, lattice defects, or phonons. These sources of electric fields create
an effective magnetic field through a spin-orbit interaction as described previously. The
1.4. SPIN RELAXATION IN GRAPHENE CHAPTER 1. INTRODUCTION
EY effect in graphene34. The Elliott relation shown in Equation 1.28 describes the linear
dependence of spin relaxation time (τs) and the momentum scattering time (τp). Theα
term is related to the spin flip probability in a momentum relaxation event34.
τs =
τp
α (1.28)
In graphene this relationship is dependent on the carrier concentration because of the
“absence of a energy gap between the conduction and valenceπbands”34. The relevant
Elliott equation (Equation 1.29) in graphene can be calculated with reasonable
approxima-tions34.
τs '
ε2
F
∆2
S O
τp (1.29)
HereεF is the Fermi energy and∆s ois the spin-orbit splitting energy. Since experimental
evidence35shows a linear dependence between the spin relaxation time and the diffusion
coefficient, matching the dependence predicted by Ochoa et. al34, this is a good model in
some systems. Additionally in CVD based spin valves, the EY mechanism (As described by
Avsar et al.36) can describe the spin relaxation dynamics34. This is attributed to the large
number of grain boundaries in CVD graphene34.
1.4.2
Dyakonov Perel Spin Relaxation
The Dyakonov-Perel (DP) mechanism comes into play in non-centrosymmetric
semicon-ductors and describes spin relaxation occurring between scattering events37–39. The DP
mechanism is proportional to the inverse of elastic scattering time τs p i n∝τ 1 e l a s t i c
1.4. SPIN RELAXATION IN GRAPHENE CHAPTER 1. INTRODUCTION
occurs between scattering events. This is the inverse of the relationship observed in the EY
mechanism τs p i n∝τe l a s t i c. The additional Hamiltonian term describing the DP effect is:
ħ
hΩ~(p~)S~ (1.30)
WhereΩ~(p~)is the effective magnetic field experience by an electron with spinS~. ThisΩ~(p~)
is a function of the momentum as shown:
Ωx ∼px px2−p
2
z
Ωy ∼py pz2−p
2
x
Ωz∼pz
px2−py2 (1.31)
When the electron’s momentum changes by collisions or other scattering processes in
the material, the effective magnetic field changes (since it is a function of the momentum).
This causes the spin to precess in between collisions, meaning that the DP spin relaxation
actually increases with decreasing impurities. Though single layer graphene is a
centrosym-metric 2D material, the substrate and impurity interactions or simply bilayer graphene
break that symmetry, allowing DP spin relaxation to occur.
In graphene a number of experiments show that spin relaxation that does not scale
proportionally to the density of charged impurities. Examples include measurements of
spin transport in by-layer graphene27,40, and measurements where the effects of the density
of charged impurities was examined40. The DP mechanism is also used to explain the effects
of ripples in the graphene on spin transport. Conceptually, the ripples force changes in the
1.4. SPIN RELAXATION IN GRAPHENE CHAPTER 1. INTRODUCTION
the spin-orbit coupling induced by various effects have been conducted by Gmitra et al41,
including the effects of ripples as shown in Figure 1.7.
Figure 1.7.Spectral gap via spin-orbit interactions (2λl) induced by a∆shift of a carbon atom
out of the plane. The red and blue lines show the results with and without including d-orbitals. Figure from Gmitra et al41.
1.4.3
Spin Relaxation Experiments
A informative paper by Han et al.42gives some information about when DP or EY interactions
are the dominant effects. By comparing the results of single and bi-layer graphene at low
temperature they show that the spin lifetime has a different dependence in each case,
meaning that EY interactions are dominant in single layer graphene and DP interactions
are dominant in bilayer graphene. However there is still uncertainty about the details of spin
dynamics of graphene30. The effects of substrates, defects, and other non-uniformities on
1.5. GRAPHENE SPINTRONICS CHAPTER 1. INTRODUCTION
overall spin and electrical transport properties of graphene make it an excellent choice for
spintronics. It is useful to compare graphene to silicon, where the spin transport has been
demonstrated43. Applebaum et al. showed spin signals in the pA range and spin lifetimes
of 1 ns. These figures were only observed at low temperatures (85◦K) where graphene has a
2ns lifetime at room temperature.
1.5
Graphene Spintronics
Graphene is an ideal choice for spin transport due to it’s low spin orbit interaction and high
mobility. Since transporting spin polarized carriers does not require a band gap, the highest
quality and highest mobility graphene can be used. The first graphene spin transport
measurements demonstrated spin diffusion lengths in the micron range44. This would easily
accommodate a device based on the manipulation of electron spin in a graphene channel.
More recent measurements have shown that the use of high quality oxide tunneling barriers
and epitaxial graphene improves the measured spin transport dramatically27. Dublak et.
al27. measured spin diffusion lengths in the 100µm range and spin signals in the MΩrange,
orders of magnitude above previously observed values30. The high quality oxide layers were
used for tunneling junctions for spin injection.
1.5.1
Tunnel Junctions
The first step in any spin dependent device is to have a spin polarized source of electrons.
In a ferromagnetic metal the fermi level is dominated by a single spin polarization, so
1.5. GRAPHENE SPINTRONICS CHAPTER 1. INTRODUCTION
Figure 1.8.Mott model of charge transport across an interface between a ferromagnet and a semiconductor. The spin transport is modeled by separating the current into a spin up (I↑) and a spin down(I↓) channel. The effective resistance (R↑,R↓) in the magnet is spin dependent. The large difference between (R↑- R↓) and RS C makes the spin injection into semiconductors very
inefficient.
interface between the ferromagnetic metal and the spin conducting semiconductor, the
electron transport can be modeled with two transport channels. One path carries spin
"up" electrons and the other carries spin down electrons as shown in Figure 1.8. The
difference in Fermi level for spin up and down electrons in the magnet makes the chemical
potential different for carriers with different spin. Therefore each path has a different
effective resistance. However a semiconductor’s resistance is very large compared to the
difference in resistance between the two spin paths. Since the total resistance of the spin up
and down paths (including both the ferromagnet and semiconductor) are nearly identical,
the spin polarization of carriers injected into the semiconductor is negligible. This problem
is called the conductivity mismatch.
To avoid conductivity mismatch a tunnel junction can be used. Using Quantum
1.5. GRAPHENE SPINTRONICS CHAPTER 1. INTRODUCTION
via tunneling is modeled using Fermi’s Golden Rule:
λi f =
2π ħ h ψi V ψf
ρf(Ef)Ef'Ei (1.32)
Hereλi f is the transition probability per unit time,ψiandψf are the initial and final states
of the tunneling electron,ρf(EF)is the density of final states,Ei is the energy of the initial
state and finally V is the potential due to the oxide barrier.
Equation 1.32 states that the transition probability per unit time (λi f) is proportional to
the density of the initial and final states. The transmission probability is dependent on the
initial states due to theψi
V
ψf
term. Since the initial statesψiis spin dependent in a
ferromagnet, the transmitted electrons preserve that spin asymmetry since the ψiV ψf
term couples the initial and final states. Notice that this process is independent of
con-ductivity in either the metal or semiconductor thereby solving the concon-ductivity mismatch
problem.
1.5.2
Producing Tunnel Barriers on Graphene
The potential barrier necessary for quantum tunneling should be as spatially uniform as
possible. The oxide or other non-conducting material should be chemically and physically
uniform. Since we want significant conduction through tunneling barrier, it must be no
more than a few nanometers thick. MgO and Al2O3 oxide barriers are typical for spin
injection into graphene45. Early attempts at producing oxide films on graphene failed due
to the low sticking coefficient of graphene46. To overcome the low reactivity of graphene,
oxide films have been produced by using fluorine to increase the adhesion of oxide species47.
1.5. GRAPHENE SPINTRONICS CHAPTER 1. INTRODUCTION
to graphene46. The quality of the oxide interface is particularly important to a device as
it directly effects the quality of spin injection and the quality of the graphene after oxide
deposition. We have found that pulsed laser deposition can produce high quality MgO
films quickly and relatively easily, as described in Chapter 3. This process avoids a chemical
modification of the surface, avoiding possible damage to the graphene or detrimental
effects on mobility.
1.5.3
Non Local Spin Valves
Spin diffusion lengths and related quantities can be measured by a non local spin valve48.
It is comparable to a four contact resistance measurement where two contacts are used to
inject current and another two contacts are used to measure a potential drop due to the
resistance of the material. The four contacts are used to eliminate contact resistance from
the measurement. Injection of spin-polarized electrons is accomplished using a
ferromag-netic metal and a tunnel junction. The decay in spin polarization as the electron translate
through the material causes a potential difference which can be measured using another
magnetic contact and tunneling barrier setup. In addition to injection and measuring spin
polarized current, this interface will also probe the spin dependent density of states.
The first reported non local spin valve used to measure spin transport in graphene
is shown in Figure 1.9. This measurement, conducted by Tombros et. al., measured a
spin transport length of 2µm44. Subsequent measurements have claimed a 100µm spin
transport length27. The difference in voltage measured at the contacts 1 and 2 corresponds
to the decay in spin polarization as the electrons are transported in the graphene channel.
1.5. GRAPHENE SPINTRONICS CHAPTER 1. INTRODUCTION
1.5. GRAPHENE SPINTRONICS CHAPTER 1. INTRODUCTION
calculated. The equation modeling the 2-D spin transport44is:
Rn o n−l o c a l =
P2
s λs f
2Wσe
−L
λs f (1.33)
WhereRn o n−l o c a l is the measured resistance,Ps is the spin polarization of the contacts, L
and W are the length and width of the graphene strip between the central electrodes, and
λs f is the spin relaxation length and theσis the conductivity.
1.5.4
Hanle measurment
In a Hanle measurment, a change in the non local resistance is measured in a spin valve
geometry as a result of an applied magnetic field. The measurement is similar to the non
local spin valve, in that spin injection and detection is accomplished with a magnetic
material and a tunneling barrier. However, the addition of a varying magnetic field causes
the electron spins to precess. As shown in Equation 1.34, the length of the channel (L) and
the spin lifetime (τs) and the electron’s precession frequencyωare the dominant factors.
RN L∝ ±
Z ∞
0
1
p
4πD te
−L2 4D t
cos(ωt)e
−t
τs
dt (1.34)
The spin precession frequency is a function of the magnetic field as expected:
ω=gµbB⊥ ħ
h (1.35)
Hanle measurements are an excellent way to quantify spin transport unambiguously49,50
1.5. GRAPHENE SPINTRONICS CHAPTER 1. INTRODUCTION
1.5.5
Datta Das Device and the Rashba Effect
A device that modifies the spin polarization of electron waves was proposed by S. Datta
and B. Das in 198952. To explain the device, they drew an analogy to the modification of
light polarization using a electro-optic material. An electro-optic material has a direction
dependent dielectric constantεx xorεy y so that a 45◦polarized wave of light is shifted from
equal combination of x and y polarized light:
1 0 + 0 1 = 1 1 →
ei kxL
ei kyL
(1.36)
Wherekxandky are momentum of light polarized in the x and y direction and L is the path
length.
The polarization shift is modifiable (in an electro-optic material) by applying an electric
field, which changes the ratio ofkx toky and thereby the polarization of the transmitted
photons. When placed in between a polarizer and analyzer the transmitted photons can be
modulated with the voltage applied to the eletro-optic material.
The analogous device where electron spin polarization is being manipulated instead
of light polarization is shown in Figure 1.10. This device uses the spin orbit interaction
of electrons through a small bandgap semiconductor (such as InGaAs) where the 2DEG
experiences a zero field splitting due to the Rashba effect. The Rashba term in the effective
mass Hamiltonian is:
1.5. GRAPHENE SPINTRONICS CHAPTER 1. INTRODUCTION
Figure 1.10.Datta-Das device, an electron wave analog of the electro-optic modulator52
Whereη is the Rashba coupling,σx,y is the x or y Pauli matrix vector, and Kx,y is the
momentum in the x or y direction.
The Rashba effect occurs in 2-D materials. An electric field (effective or applied) breaks
the symmetry of spin up and down conductors moving in two dimensions, and thereby
breaks the degeneracy of the two states. This interaction is very similar to that of the
standard spin-orbit interactions (B= 1cE~×v~) where in this case the E field comes from the symmetry breaking instead of a nucleus. By applying a electric field across a 2DEG in a
material (or interface) with the proper symmetry, the electron spins can be manipulated.
In most cases this effect is not seen in graphene since it does not show this symmetry,
although there have been attempts at pairing graphene with a suitable material to induce
this effect, such as gold53and nickel54.
1.5.6
Graphene Spin Field Effect Transistor
Instead of using the Rashba effect to manipulate spin, an exchange interaction can be used.
1.5. GRAPHENE SPINTRONICS CHAPTER 1. INTRODUCTION
between spin polarized electrons in a magnetic material and electrons in graphene. This
is necessary when the spin-orbit coupling is negligible, such as in graphene. By using the
effective magnetic field’s interaction with electrons in graphene, the high mobility and low
intrinsic spin orbit interaction will allow for an interesting and useful device. This setup
what proposed by Semenov et. al in 200755with the device structure shown in Figure 1.11
Figure 1.11.Graphene Spin-FET geometry from Semenov et. al.55. A Ferromagnetic Dielectric (FMD) is used to alter the polarized electron spins, injected at the source S, in the graphene channel. The effect is tuned through the gate voltage Vg and detected at the source D. The arrows
in the graphene channel illustrate the spin precession induced by exchange interactions with the FMD.
This device relies on the magnetic exchange interaction with the “Ferromagnetic
Di-electric” (FMD) layer. The exchange interaction depends on the overlap of the graphene
wavefunction with that of the (uncompensated) spins of the magnetic ions in the FMD. The
magnetization in the dielectric is namedS~λ,j designating the spin moment at siteλof unit
cell j. Again following the notation and results of Semenov et. al.55the exchange interaction
1.6. MAGNETOELECTRIC CHROMIA CHAPTER 1. INTRODUCTION
He x=
1
N
N
X
j=1
X
λ
J(R~λ,j)S~λ,j·s~ (1.38)
The interaction between the electron spin,s~, and the magnetization in the material (via
the exchange interactions) causes the spin precession on which the device relies.
1.6
Magnetoelectric Chromia
To realize the device described by Semenov et. al55. requires the selection of a ferromagnetic
dielectric (FMD) that will induce a spin precession in the graphene channel. A
magnetoelec-tric (ME) is an excellent choice for the FMD as it would allow for an elecmagnetoelec-trically switchable
exchange interaction thus providing control over the spin-FET. Chromium Oxide is a very
good choice for this application since it exhibits a significant room temperature ME effect.
1.6.1
Magnetoelectrics
Magnetoelectric materials have a coupling between the the electric and magnetic ordering
in the material. This can be identified by the relevant components of the Landau free energy
expansions:
F(E~,H~) =F0−PisEi−MisHi−
1
2ε0εi jEiEj − 1
2µ0µi jHiHj −αi jEiHj (1.39)
whereF0is the initial free energy,Pis andM s
i are the spontaneous polarization and
mag-netization,E andH are the electric and magnetic field respectively, andεandµare the
1.6. MAGNETOELECTRIC CHROMIA CHAPTER 1. INTRODUCTION
are not uniform in all directions. Theαis the linear magnetoelectric coefficient coupling
the electric and magnetic fields.
The magnetization of a ME material is composed of the static magnetizationMs i, the
response to magnetic fieldHiand the contribution fromαi jEjHj. Theαi j term couples the
magnetization to the electric fieldEi. The “ij” subscripts in theαterm are included because
the value of alpha changes depending on the symmetry of the material, i.e. αx x 6=αz z.
Taking the derivative ddHF
i gives this magnetization:
Mi E~,H~
=−dF
dHi
=MiS+µoµi jHi+αijEi (1.40)
For ME materials, the electric polarization is dependent on the external magnetic field.
By taking the derivative of the Landau free energy expansion shown in Equation 1.39, ddEF i
gives the polarizationPi(E~,H~)which is proportional toαi,jHj.
Only materials with the proper symmetry have anα6=0. The material must have Polarity
(P)·Time (T) symmetry but not P or T symmetry individually. This is shown in Figure 1.12.
Conceptually, time reversal inverts the magnetization, and spacial inversion inverts the
electric polarization. In a material with ME symmetry, the original order is restored only
when both time and space inversion are applied.
Intuitively this effect can be understood by considering Cr2O3, a magnetoelectric,
anti-ferromagnetic and piezoelectric material. The electric field induces a strain in the material
that pushes some chromium atoms out of their sub-lattice by a piezoelectric effect. These
Cr atoms have their spin flipped by an exchange interaction with chromium atoms in the
other sublattice that have an opposing magnetization. The two sub-lattices have opposite