close

Вход

Забыли?

вход по аккаунту

?

Molecular Beam Epitaxy Integration of Magnetic Ferrites with Wide Bandgap Semiconductor 6Hydrogen-Silicon carbide for Next-generation Microwave and Spintronic Devices

код для вставкиСкачать
Molecular Beam Epitaxy Integration of Magnetic Ferrites
with Wide Bandgap Semiconductor 6H-SiC for Next-generation
Microwave and Spintronic Devices
A Dissertation Presented
By:
Zhuhua Cai
To
The Department of Chemical Engineering
In partial fulfillment of the requirements
For the degree of
Doctor of Philosophy
In
Chemical Engineering
Northeastern University
Boston, Massachusetts
May, 2010
UMI Number: 3443799
All rights reserved
INFORMATION TO ALL USERS
The quality of this reproduction is dependent upon the quality of the copy submitted.
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.
UMI 3443799
Copyright 2011 by ProQuest LLC.
All rights reserved. This edition of the work is protected against
unauthorized copying under Title 17, United States Code.
ProQuest LLC
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, MI 48106-1346
Acknowledgements
I am very grateful for the opportunity that I had to study at Northeastern University and
for all of my experiences with the students and faculty of the Chemical Engineering
Department. I would especially like to thank my advisor, Dr. Katherine Ziemer for inspring me
and supporting me to step into the previously unfamiliar field of material science. I am very
grateful for her rigorous and scientific attitude, and the knowledge she has given to me has
prepared me for my future endeavors even more than I ever imagined. My grateful thanks
are extended to Dr. Albert Sacco Jr., Dr. Elizabeth Podiaha-Murphy, Dr. Nian Sun, and Dr.
Vlado Lazarov for serving on my committee and for their insightful comments and
suggestions. I would especially like to thank Dr. Lazarov for his help with TEM analysis and
his intellectual advice and discussion.
I am also appreciative of the assistance and support from my labmates. I would like
to thank Joseph Tsai for helping me to get started on the UHV and MBE system. I would like
to thank Dr. Trevor Goodrich, who trained me how to operate and maintain the MBE system.
He always encouraged me when things were not working and helped me unconditionally in
the past many years. Without his tireless support, this thesis would not have been possible. I
would also like to thank my other labmates, Bing Sun, Alex Avekians, Natalia Maximova,
Bo Zhou, and Ghulam Uddin for their support and assistance around the lab. Especially Bing,
whose lovely and humorous character provided a pleasant atmosphere at work.
I would like to thank staff in Center for Microwave Magnetic Materials and
Integrated Circuits under the direction of Dr. Vince Harris. Dr. Vince Harris always treated
me as his own student and offered me full access to all of his equipments. His generosity and
assistance made this research possible. I greatly appreciate Dr. Yajie Chen for giving me
invaluable suggestions from his profound knowledge and experience in magnetism and
ferrite materials. Furthermore, I appreciate Dr. Zhaohui Chen and Dr. Aria Yang for training
me on the PLD system and for their friendships.
I would also like to thank the staff and graduate students at the Center for Advanced
Microgravity Materials Processing under the direction of Dr. Al Sacco, Jr. for use of their
equipments and expertise. I would like to thank Dr. Juliusz Warzywoda for training me on
their SEM and AFM. I greatly appreciate my lifelong friend Dr. Zhaoxia Ji, whose diligent
and rigorous attitude of research has been and will always be my model.
I am deeply indebted to Guangdong Zhang for his endless love, support, and
encouragement. He always motivated me to be the very best and was always there for me. I am
truly grateful to Lin Wang, who kept me company and brought a lot of happiness into my life.
I am also greatly indebted to Ming Liu for his continuous help and support. His insights, ideas,
and excitement for research are very motivating. My sincere thanks are extended to my
many other friends both in China and in the United States. Their friendships made my Ph.D.
experiences colorful and memorable.
Finally, I would like to dedicate this thesis to my family; my parents Yongkang Zhao
and Shufeng Cai, my sister Youli Zhao, and my love Trevor Goodrich. Thank you all for
loving me and supporting me unconditionally. Without you, I would never have made it this
far in life. With all of my heart, I thank all of you, my mentors, friends, and family!
ii
Abstract
Ferrite/ferroelectric heterostructures have attracted much attention in recent years
because of their unique ability to potentially enable dual magnetic and electric field tunability.
The simultaneous magnetic and electric tunability in such structures can be applied in a wide
range of microwave planar devices (e.g., tunable phase shifters, resonators, and delay lines)
and spintronics (e.g., magnetic tunneling junctions for magnetic sensors and nonvolatile
magnetic memories). However, the attempts to engineer ferrite/ferroelectric heterostructures
to operate at the frequencies higher than 5 GHz are limited. Barium hexaferrite (BaM,
BaFe12O19) is an ideal candidate for high frequency microwave device applications because of
its strong uniaxial anisotropy (HA ~17 kOe) and can be tuned to ferromagnetic resonance
(FMR) at frequencies higher than 40 GHz with relatively small applied magnetic fields. Spinel
ferrite Fe3O4 has a high Curie temperature of 858 K and is predicted to possess ~ 100% spin
polarization, which can lead to ultrahigh tunneling magnetoresistence even at room
temperature. The performance of today’s ferrite-based microwave communication and
spintronic devices would be enhanced and next-generation monolithic microwave integrated
circuit (MMIC) would be possible if ferrite/ferroelectric heterostructures can be integrated
with wide band gap semiconductors (e.g., SiC or GaN), which can function in
high-temperature, high-power, and high-frequency environments. The goal of this work is to
use molecular beam epitaxy (MBE) to understand nucleation and film growth mechanisms
needed to integrate magnetic ferrites (BaM and Fe3O4) with SiC, and subsequently understand
the material chemistry and structure influences on forming functional interfaces (i.e.,
interfaces that enable effective ferrite/ferroelectric coupling).
The study of chemistry, structure, and magnetic properties of three generations of BaM
films grown by pulsed laser deposition shows a MBE-grown single crystalline MgO template
promotes the c-axis alignment through formation of an oxygen bridge at the interface and
minimizes the interface mixing, which enables the effective heteroepitaxy of device quality
BaM on 6H-SiC. Epitaxial single crystalline BaM film with strong c-axis perpendicular
alignment, high HA (16.2 kOe) and magnetization (4.1 kG) was also successfully grown by
MBE for the first time on 6H-SiC. Through MBE, further study of the chemistry and structure
evolution at the BaM//SiC interface suggests the 10 nm MgO template not only functions as a
diffusion barrier, but also forms a spinel transition layer that is structurally similar to BaM.
The high quality BaM film on SiC is compatible with MMIC and can also function as a
magnetic layer in BaM/ferroelectric multiferroic heterostructures for electrostatic FMR
tuning. Through MBE, single crystalline, epitaxial Fe3O4 (111) films and Fe3O4/BaTiO3/Fe3O4
heterostructures were successfully integrated with 6H-SiC. The Fe3O4 film exhibits high
strucutrual order with sharp interfaces and an easy axis in-plane magnetization with a
coercivity of 200 Oe. In the Fe3O4/BaTiO3/Fe3O4 heterostructure, the magnetoeletric coupling
is demonstrated at room-temperature by an electric field induced magnetic anisotropy field
change. The Fe3O4/BaTiO3/Fe3O4 heterostructure has the potential application in multiferroic
tunneling junction used in novel information storage. Understanding the ferrite growth
mechanisms and interface functions through this research, is an important contribution
toward the realization of a next-generation, multifunctional device.
iii
Table of Contents
1.0 Introduction..............................................................................................................1 2.0 Background ..............................................................................................................5 2.1 Magnetic Ferrite Materials ................................................................................5 2.1.1 Classification of Materials by Magnetic Properties.....................................6 2.1.2 Spinel Ferrite Fe3O4 .....................................................................................9 2.1.3 Hexagonal BaFe12O19 (BaM) .....................................................................12 2.1.4 Magnetic Properties ...................................................................................14 2.1.4.1 Anisotropy Field .....................................................................................15 2.1.4.2 Magnetic Hysteresis................................................................................15 2.1.4.3 Magnetic Domains ..................................................................................17 2.1.4.4 Microwave Properties .............................................................................18 2.2 Spintronics .......................................................................................................21 2.3 Multiferroics and Ferrite/Ferroelectric Heterostructure ..................................23 2.4 Wide Bandgap Semiconductor ........................................................................27 2.5 Thin Film Nucleation and Growth ...................................................................29 2.5.1 Epitaxy and Atomic Mismatch ..................................................................29 2.5.1.1 Lattice Mismatch ....................................................................................31 2.5.1.2 Thermal Mismatch ..................................................................................34 2.5.2 Thin Film Growth Mechanisms .................................................................35 2.5.3 Substrate Surface Impacts..........................................................................41 2.6 Thin Film Deposition Methods ........................................................................42 2.6.1 Traditional Deposition Methods for BaM Thin Film Growth ...................42 2.6.2 MBE Growth Process ................................................................................45 3.0 Critical Literature Review .....................................................................................48 3.1 Barium Ferrite Thin Film Growth ...................................................................49 3.1.1 The Influences of Substrates on BaM Film Growth ..................................50 3.1.2 The Influences of Various Templates for BaM Film Growth ....................53 3.1.3 The Influences of Temperature on BaM Film Growth ..............................61 3.1.4 The Influences of Deposition Pressure and O2 Pressure on BaM Film .....69 3.2 Iron Oxides Growth by Molecular Beam Epitaxy ...........................................76 3.2.1 Tunability of Iron Oxides Growth via MBE Differentiate Iron Oxides by XPS
…………………………………………………………………………...77 3.2.2 The Choice of Oxidizing Gas for Iron Oxides Growth O vs. O2 ...........80 3.2.3 The Choice of Substrate for Iron Oxide Growth .......................................81 3.2.4 The Choices of MBE Processing Parameters (T, Po2, plasma) for Iron Oxide
Growth ……………………………………………………………………………85 3.2.5 Integrate Fe3O4 with Semiconductor for Spintronics ................................88 3.3 Ferrite/Ferroelectric Multiferroic Heterostructure ...........................................90 3.3.1 ME and CME in Laminate Multiferroic Heterostructure ..........................90 3.3.2 Multiferroic Tunnel Junction (MFTJ) in Laminate Multiferroic Heterostructure
……………………………………………………………………………95 iv
3.4 Summary ..........................................................................................................97 4.0 Experimental ........................................................................................................100 4.1 Thin Film Preparation ....................................................................................100 4.1.1 Thin Film Growth: Molecular Beam Epitaxy (MBE) .............................101 4.1.1.1 The Ultra High Vacuum Set Up for MBE System ................................101 4.1.1.2 BaM and Fe3O4 on 6H-SiC by MBE ....................................................105 4.1.2 Thin Film Growth: Pulsed Laser Deposition (PLD)................................108 4.2 Thin Film Characterization ............................................................................ 110 4.2.1 X-ray Photoelectron Spectroscopy .......................................................... 111 4.2.2 Reflection High Energy Electron Diffraction .......................................... 113 4.2.3 Atomic Force Microscopy ....................................................................... 115 4.2.4 Scanning Electron Microscopy ................................................................ 116 4.2.5 X-ray Diffraction ..................................................................................... 117 4.2.6 Transmission Electron Microscopy and Energy Dispersive X-ray
Spectroscopy.. .......................................................................................................... 118 4.2.7 Vibrating Sample Magnetometry ............................................................. 118 4.2.8 Ferromagnetic Resonance ........................................................................ 119 5.0 Results and Discussion ........................................................................................120 5.1 Interface Engineering of BaM on 6H-SiC by PLD .......................................120 5.1.1 Generation I and Generation II BaM Films Grown by PLD ...................121 5.1.2 Generation III BaM film on SiC by PLD with MBE Grown MgO .........134 5.1.2.1 MgO Template Grown by MBE ...........................................................135 5.1.2.2 The Impact of O2 Pressure on the Gen III BaM Quality ......................139 5.1.2.3 The Impact of MBE-grown MgO Thickness ........................................143 5.1.2.4 The Impact of Post-deposition Annealing ............................................155 5.1.2.5 Summary ...............................................................................................161 5.2 Integration of BaM with 6H-SiC by Molecular Beam Epitaxy .....................162 5.2.1 Establishing Operating Conditions Window ...........................................163 5.2.2 The Impact of MgO Template .................................................................167 5.2.3 The Optimization of O Pressure ..............................................................172 5.2.4 Demonstration of the MBE-grown BaM as a Seed Layer for Thick BaM
Growth by PLD....…………………………………………………………………184 5.2.5 Interface Study for the Potential Applications of the BaM-ferroelectric
Heterostructures .......................................................................................................186 5.3 Integration of Fe3O4 film with SiC by MBE for Potential Spintronics Applications
……………………………………………………………………………...198 5.3.1 Epitaxial Growth of Fe3O4 Film on SiC ..................................................198 5.3.2 Epitaxial Growth of Fe3O4 /BaTiO3/Fe3O4 Multiferroic Heterostructure on
SiC
………………………………………………………………………….201 6.0 Conclusions..........................................................................................................210 7.0 Recommendations................................................................................................215 8.0 References............................................................................................................217 v
List of Figures
Figure 1: The inverse Fe3O4 spinel crystal structure [29]. One unit cell consists of four (001)
layers, each layer contains the oxygen anions (large white circles) and the octahedral
iron ions (small black circles), the tetrahedral sites (small grey circles) are located
halfway between these layers. ........................................................................... .10 Figure 2: A schematic drawing of Fe3O4 (001) plane[37], where the 90o superexchange
interaction at the octahedral site is shown as solid black line, the superexchange
interaction between iron ions on the octahedral and tetrahedral sites is shown as
dashed black line, and the double exchange between iron ions on the octahedral
sites is shown as solid grey line. ......................................................................... 11 Figure 3: The unit cell of BaFe12O19, courtesy to Dr. Aria Yang. One unit cell consists of 10
oxygen layers, The oxygen ions (small white spheres) form a hexagonal or cubic
closed packed lattice along the [001] direction. ................................................. 13 Figure 4: A typical hysteresis loop for a ferro- or ferri- magnetic material [46]. Point “a” and
“d” indicate the satuation magnetization (Ms), where the material has reached its
magnetization saturation; point “b” and “e” indicate the level of residual magnetism
(Mr) in the material; point “c” and “f” indicate the force (Hc) required to remove the
residual magnetism from the material. ............................................................... 15 Figure 5: Ferrimagnetic resonance spectra for ~50m thick BaM film on (111) GGG
substrate at 58GHz [48], the reasonably narrow FMR linewidth was ~0.068 kOe.
............................................................................................................................ 20 Figure 6: The relationship between multiferroic and magnetoelectric materials [63]. The
intersection (red hatching) represents the multiferroic materials that are
ferromagnetic and ferroelectric........................................................................... 25 Figure 7: Schematic illustration of the relationships between different functionalities of the
materials [60,64,65]. Where E: electric field, H: magnetic field, P: dielectric
polarization, M: magnetization, : stress, and : strain. ..................................... 26 Figure 8: Different 2D stacking sequences for SiC. (a) 3C-SiC, (b) 4H-SiC, (c) 6H-SiC, the
black dots represent the carbon atoms and the hollow dots represent the silicon
atoms [67]. .......................................................................................................... 28 Figure 9: Schematic illustration of lattice-matched, strained, and relaxed heteroepitaxial
structures. Lattice matched heteroepitaxy is structurally similar to homoepitaxy
[68]. ..................................................................................................................... 30 Figure10: Schematic illustration of substrate surface reconstruction of hexagonal
close-packed crystal structure [68]. .................................................................... 32 Figure 11: Schematic illustrate of 4:3 lattice matching. The black boxes indicate the lattice
cells for GaAs and MgO. While there is a large difference between the size of the
cells, 4 MgO lattices provides a good match to 3 GaAs lattices, as shown by the red
line [69]. .............................................................................................................. 32 Figure 12: (a) (111) plane in MgO Rock-Salt lattice (b) Pseudo-hexagonal MgO (111) plane,
only O atoms are involved in these two figures. ................................................ 33 Figure 13: Schematic illustration of lattice matching for BaM growth on MgO (111) and
vi
6H-SiC,only O atoms involved........................................................................ 34 Figure 14: Schematic illustration of basic atomic nucleation of impinging atoms on a
substrate through vapor deposition [68]. ............................................................ 36 Figure 15: Basic modes of thin film growth [72], (a) layer by layer growth (2D), (b) layer plus
island growth (Stranski-Krastanov), (c) island growth (3D). ............................. 37 Figure 16: Schematic drawing of a generic MBE system. The MBE growth chamber consists
of effusion cells sources, and an oxygen plasma source. This chamber also includes
a 15keV RHEED system for structural monitoring during film growth............. 47 Figure 17: C-axis dispersion in the film and lattice strain in terms of change in lattice spacing
(△d/dbulk) as a function of film thickness. The inset shows how c-axis dispersion in
the film changes with the lattice strain [76]........................................................ 51 Figure 18: Schematic illustration of the matching between rhombohedral sapphire and
hexagonal barium ferrite; (a) before deformation; (b) one way of deforming the
hexagonal structure to fit on the rhombohedral structure; the area change for this
case is -25.7Å2 (c) another way of fitting the hexagon on the rhombohedral
structures, which yields a smaller area (-3.58 Å2 ) change of the hexagon [76]. 52 Figure 19: X-ray diffraction diagrams of the BaM film: (a) without template, (b) with
template [91]. ...................................................................................................... 55 Figure 20: XPS depth profiles on the concentration of BaM films: (a) without template, the
depth of diffused layer is about 700 Å; (b) with template, the depth of diffused layer
is about 300 Å [91]. ............................................................................................ 56 Figure 21: The XRD results of BaM films prepared by a) the ex-situ process at various
annealing temperatures; b) the in-situ process at various substrate temperatures [78].
............................................................................................................................ 63 Figure 22: a) The dependence of the spontaneous magnetization of films prepared by the
ex-situ process on annealing temperatures; b) the dependence of Ms of films
prepared by the in-situ process on substrate temperatures [78]. ......................... 63 Figure 23: The dependence of coercivity, Hc, a) on annealing temperature, Ta, for films
prepared by the ex-situ process; b) on substrate temperature, Ts, for films prepared
by in-situ process [78]. ....................................................................................... 65 Figure 24: SEM image of films, a) prepared by the in-situ process at Ts of 550 ºC; b) prepared
by the ex-situ process at Ta of 850 ºC for 2 h [78]. ............................................. 65 Figure 25: Schematic representation of the changes in the film structures with the substrate
temperatures varying from 400 to 920C during deposition (TD). Starting at TD;300–
400 °C, the film is amorphous; with TD;550– 600 °C, the grains are too small to
observe by XRD but grow upon annealing; with TD around 700 °C, film is granular;
and only at TD around 900 °C the film becomes continuous and exhibits a spiral
growth[51]. ......................................................................................................... 66 Figure 26: Parallel (dashed line) and perpendicular (solid line) hysteresis loops for the films
with (a) Ts=300, Ta=850, (b) Ts=400, Ta=850, (c) Ts=575, Ta=850, and (d)
Ts=920 °C [51]. ................................................................................................... 67 Figure 27: The AFM micrographs of barium ferrite films deposited on (001) sapphire
vii
substrates with (a) as deposited at 920C; (b) deposited at 920C and post-annealed
at 1000C under flowing oxygen of 300 mTorr for 2 hours [51]. ...................... 68 Figure 28: SEM surface morphologies of the BaM films deposited at (a) 300 mTorr and (b) 20
mTorr [77]. .......................................................................................................... 70 Figure 29: Out-of-plane (solid line) and in-plane (dashed line) VSM hysteresis loops for the
BaM films with (a) 300 mTorr, (b) 20 mTorr [77]. ............................................ 70 Figure 30: TEM cross sections of film grown in (a) oxygen atmosphere and (b) vacuum [101].
A sharp interface existed without any diffusion between the film grown in oxygen
atmosphere and the substrate; however, the quality of interface became deteriorated
when grown under vacuum. ................................................................................ 73 Figure 31: Fe 2p XPS spectra of the of the 80 and 4 Å thick iron oxide films, grown by MBE
on Al2O3 (0001) at 250 °C. For comparison, three reference spectra of Fe2O3,
Fe3O4, and FeO are included. The Fe 2p3/2 main peak and satellite corresponding to
Fe3+ and Fe2+ cations are illustrated by the arrows respectively in Fe2O3 and FeO
spectra [109]. ...................................................................................................... 79 Figure 32: Phase diagram for the growth of iron oxides by OPA-MBE [123]. ................ 87 Figure 33: Configuration of FeBsiC (metal glass)/PZT fiber multiferroic heterstructure (left)
and the ME coefficient as a function of frequency (right) [132]. ....................... 92 Figure 34: Corss-section (c) of micro ME device (a), magnetic hysteresis loop of FeGa film
(b), and the ME coefficient vs. Si cantilever thickenss (d) [133]. ...................... 93 Figure 35: Giant electric field tuning of FMR in Fe3O4/PMN-PT and Fe3O4/PZN-PT [134].
............................................................................................................................ 94 Figure 36: Tuanble resonator YIG/PMN-PT (left) and tunable filter YIG/PZT (right) [135].
............................................................................................................................ 95 Figure 37: Schematic diagram of a tunneling junction [130], which consists of two electrodes
separated by a nanometer-thick ferroelectric barrier layer. ................................ 96 Figure 38: Effects of ferroelectricity in the SrRuO3/BaTiO3/SrRuO3 MFTJ [136]. (a)
Schematic double-well potential for the MFTJ (solid line) and for the bulk BaTiO3
(dashed line). (b) Cell averaged electrostatic potential energy profile for
polarization to the right P→ (blue) and left P←(red) states and the interfaces are
indicated by vertical dashed lines. ...................................................................... 97 Figure 39: UHV system consisting of two interconnected chambers. The chambers are
separated by a UHV compatible gate valve. The growth chamber consists of a
remote oxygen atom source, solid source effusion cells (Mg, Ba, Fe), a Ti sublimator,
and RHEED system. The analysis chamber consists of a XPS hemispherical
analyzer and an AES single pass CMA. ........................................................... 102 Figure 40: Schematic illustration of diffraction resulting from different surface features [144].
(a) ideal, single crystalline, flat surface, (b) polycrystalline surface, (c) single
crystalline with 3-D island features, and (d) single crystalline with 2-D features.
.......................................................................................................................... 114 Figure 41: Diffraction spot shape due to transmission through different sized 3-D features
(left) and chevron diffraction characteristic of surface faceting (right)............ 114 viii
Figure 42: Schematic diagram of three intermediate stages barium ferrite PLD deposition
process, the estimated thicknesses are around 2nm, 9nm, and 20nm respectively.
.......................................................................................................................... 122 Figure 43: VSM hystersis loops of (a) Gen I, and (b) Gen II BaM films grown by PLD; Gen II
films showed more c-axis perpendicular alignments, resulting in lower coercivity in
both perpendicular and in plane directions. ...................................................... 123 Figure 44: SEM and AFM images of Gen I BaM film (a & b) and Gen II BaM film (c & d)
grown by PLD on 6H-SiC. ............................................................................... 124 Figure 45: X-ray θ-2θ diffraction pattern for BaM films grown by PLD on (a) 6H-SiC, and (b)
6H-SiC with MgO interwoven template. .......................................................... 125 Figure 46: XPS O 1s photoemission spectra obtained with Al k X-rays for BaM films grown
on 6H-SiC (a) without MgO interwoven layer (b) with MgO interwoven layers126 Figure 47: XPS depth profile of BaM films grown by PLD on (a) 6H-SiC, and (b) 6H-SiC
with MgO modified template showed the effectiveness of the MgO templates in
preventing the Si diffusion................................................................................ 128 Figure 48: XPS elemental scan of Si2p for different stage growth. (a) 1st, (b) 2nd, (c) 3rd, (d)
thick BaM on SiC with MgO interwoven layers, and (e) thick BaM on SiC without
MgO interwoven layers. ................................................................................... 130 Figure 49: SEM images of different stage growth (a - c) with MgO interwoven layers, (d - f)
without MgO interwoven layers. ...................................................................... 131 Figure 50: (a) AFM image of the 3rd stage BaM growth on SiC with MgO interwoven layers,
(b) Profile along the lines in (a) , (c) AFM image of the 3rd stage BaM growth on SiC
without MgO interwoven layers. ...................................................................... 133 Figure 51: XPS survey scans of H2 cleaned 6H-SiC: (a) with 8% O; (b) with 14% O; insets
are the corresponding RHEED patterns along <11 2 0>. The L1/3 and L2/3 rings from
reconstruction are not clear for SiC with 14% O. ............................................. 137 Figure 52: XPS tight scans of the O 1s spectra for MgO films grown on H2 cleaned 6H-SiC:
(a) with 8% O; (b) with 14% O. ....................................................................... 138 Figure 53: Cross sectional HRTEM images of the MgO films deposited at 150oC under the
oxygen pressure of 5×10-6 Torr on: (a) 6H-SiC with 8% O; (b) 6H-SiC with 14% O
after H2 cleaning. .............................................................................................. 139 Figure 54: XRD patterns for Gen III BaM films grown by PLD with 10nm MBE-grown MgO
at (a) 20mTorr, (b) 100mTorr, and (c) 200mTorr ; increased background O2 pressure
in the PLD chamber deteriorated the dc magnetic characteristics of the deposited
BaM film. .......................................................................................................... 141 Figure 55: Hysteresis loops for Gen III BaM films grown by PLD with 10 nm MBE-grown
MgO at (a) 20 mTorr, (b) 100 mTorr, and (c) 200 mTorr ; increased background O2
pressure in the PLD chamber deteriorated the dc magnetic characteristics of the
deposited BaM film. ......................................................................................... 142 Figure 56: AFM pattern for Gen III BaM films grown by PLD with 10 nm MBE-grown MgO
at (a) 20 mTorr, (b) 100 mTorr, and (c) 200 mTorr show a deterioration of desirable,
ix
smooth morphology with the increase in PLD chamber O2 pressure. .............. 143 Figure 57: The thicker MgO single crystal film enabled a more aligned BaM film with no sign
of misoriented (107) peak as seen in the X-ray θ-2θ diffraction patterns of BaM
films grown by PLD (a) with 4 nm MgO grown at 150 oC by MBE, inset AFM
showed evidence of misaligned crystals, and (b) with 10nm MgO grown 150 oC by
MBE. * designated the second phase of spinel structure.................................. 144 Figure 58: Hysteresis loop of BaM grown on (a) 10nm, and (b) 4 nm MBE-grown MgO/SiC
obtained by VSM with a maximum applied field of 12,500 Oe aligned parallel (open
square) and perpendicular (solid square) to the film plane............................... 145 Figure 59: XPS depth profile of BaM films grown by PLD on (a) 10 nm MgO (111)/SiC(0001)
template, and (b) 4 nm MgO(111)/SiC(0001) template. This comparison shows a
sharper interface and limited Si diffusion with the 10 nm MgO template. ...... 148 Figure 60 (a) Bright-field cross-section TEM image of BaM thin films grown on 6H-SiC with
a 10nm MgO template. A line with altered contrast between the BaM and SiC
indicates presence of interface phases. (b) SAD pattern from the substrate and film
in [11 2 0] zone axis. BaM film is single crystal and epitaxially grown on SiC.151 Figure 61: High resolution TEM cross sections of (a) the interface of BaM film grown on
6H-SiC with 10nm MgO template, (b) the interface between BaM and the transition
layer MgFe2O4, insets are the digital diffractograms of the BaM (top) and MgFe2O4
(bottom). ........................................................................................................... 152 Figure 62: Chemical composition of BaM film interface region by EDX. SiOx layer is formed
closest to the SiC substrate and the final film contains thicker than 40 nm transition
layer of MgxBa1-xFe2O4..................................................................................... 153 Figure 63: Power derivative as a function of applied magnetic field in the region near the
ferromagnetic resonance at 53 GHz. The linewidth measured as the peak-to-peak
power derivative is 96 Oe for the post annealed BaM film deposited on 10 nm
MgO(111)//SiC(0001) [141]. ............................................................................ 156 Figure 64: (a) AFM image of the annealed BaM on SiC with MgO template, (b) measured
step heights, and (c) profile along the lines in (a)............................................. 157 Figure 65: XPS depth profile of annealed BaM films grown by PLD with 10 nm MgO grown
at 150 oC by MBE. ............................................................................................ 158 Figure 66: XRD patterns of BaM film before (a) and after (b) anneal. * designated W type
BaM and ** designated MgFe2O4 (111) ........................................................... 159 Figure 67: The region is shown where a Mg 1s peak will appear if Mg is present in the surface
of the film. (a) As deposited BaM grown on 10 nm MBE-grown MgO, (b) two step
annealing of two-minute duration each, (c) two step annealing of seven-minute
duration each. .................................................................................................... 160 Figure 68: Fe 2p photoemission spectra obtained with Mg k X-rays for 200nm BaM grown
at O/Fe flux ratio of (a) 20, and (b) 40. ............................................................ 164 Figure 69: RHEED patterns of the BaM films deposited at 600, 800, and 900 °C on 6H-SiC
respectively. ...................................................................................................... 165 x
Figure 70: AFM images of the BaM films deposited at 600, 800, and 900 °C on 6H-SiC
respectively. ...................................................................................................... 166 Figure 71: XPS photoemission spectra obtained with Mg k X-rays for 2 nm BaM films
grown on 6H-SiC (a) Si 2p with 2nm MgO template, (b) Si 2p without 2nm MgO
template, (c) O 1s with 2nm MgO template and (d) O 1s without 2nm MgO template.
.......................................................................................................................... 168 Figure 72: XPS Si 2p (a) and O 1s (b) photoemission spectra obtained with Mg k X-rays for
around 1.5 nm BaM films grown directly on 6H-SiC at two different electron
emission angles: 90° (normal emission) and 30° (grazing emission). .............. 170 Figure 73: 3D AFM images obtained for 200 nm BaM films grown by MBE on (a) 6H-SiC (b)
6H-SiC with 10nm MgO template. ................................................................... 171 Figure 74: RHEED patterns at for (a) as received 6H-SiC, (b) H2 cleaned 6H-SiC (0001), (c)
10nm MgO (111) , (d) 200nm BaM film grown on MgO (111) at PO2 = 3×10-6 Torr
along the [11 2 0] and [10 1 0] azimuths. The primary beam energy was 12.5keV.
.......................................................................................................................... 173 Figure 75: RHEED patterns for 200 nm BaM films grown on MgO (111) template at Po2 =
6×10-6 Torr along the (a) [11 2 0] and (b) [1010] azimuths, and Po2 = 1.5×10-6 Torr
along the (c) [11 2 0] and (d) [1010] azimuths. The primary beam energy was 12.5
keV. ................................................................................................................... 175 Figure 76: Pressure impact on VSM Hysteresis loops for BaM films grown by MBE with 10
nm MgO templates at (a) Po2 = 1.5×10-6 Torr, (b) Po2 = 3.0×10-6 Torr, and (c) Po2 =
6.0×10-6 Torr. The maximum applied field of 12,500 Oe aligned parallel and
perpendicular to the film plane. ........................................................................ 177 Figure 77: X-ray diffraction pattern for Ba-hexaferrite (M-type) film grown under (a)
deficient oxygen environment, (b) excess oxygen environment, and (c) optimal
oxygen environment. ........................................................................................ 179 Figure 78: Pole figure obtained for a fixed value of 2= 30.6o. The single dominant peak
corresponding to =90 o and  = 0 o corresponds to the <008> reflection. The six
dots exhibiting six-fold symmetry correspond to the closely spaced <107> type
reflections illustrating the epitaxial relationship between the BaM and SiC.... 180 Figure 79: 2m×2m (a) AFM and (b) MFM images of BaM film with 10nm MgO template
grown at optimal oxygen pressure. ................................................................... 181 Figure 80: 5m×5m MFM images of BaM film with 10nm MgO template on SiC grown by
(a) MBE and (b) PLD (Gen III BaM film). ...................................................... 182 Figure 81: (a) Model of BaM crystal structure in [11 2 0] orientation, (b) high angle annular
dark field (HAADF) image of BaM grown on 6H-SiC with a 10nm MgO template,
white lines indicate Ba atomic planes. Inset is the EDX line scan for three spots in
the film. Regions marked as red circles show the Ba atomic columns, suggesting
xi
high quality film with limited stacking faults (The waviness of the Ba atomic planes
is artificial, due to instabilities of the STEM scanning coils). .......................... 183 Figure 82: Hysteresis loop of BaM grown on a 10nm MBE-grown BaM obtained by VSM
with a maximum applied field of 12,500 Oe aligned parallel (open red square) and
perpendicular (solid black square) to the film plane. Inset is the AFM image for this
film. ................................................................................................................... 185 Figure 83: (a) Bright-field cross-section TEM image of BaM thin films grown on 6H-SiC
with a 10nm MBE-grown BaM seed layer, inset is the SAD pattern from the
substrate and film in [11 2 0] zone axis; (b) High resolution TEM of the BaM film,
inset is the digital diffractogram. ...................................................................... 186 Figure 84: RHEED patterns along the [11 2 0] azimuth of bulk SiC, for epitaxial BaM (0001)
on MgO (111) template at PO2 = 3×10-6 Torr as a function of film thickness, taken at
12.5 keV: (a) H2 cleaned 6H-SiC (0001), (b) 10nm MgO (111) template, (c) 10 Å, (d)
40 Å, (e) 100 Å, and (f) bulk BaM (0001)........................................................ 188 Figure 85: High-energy-resolution Fe 2p core-level XPS spectra for epitaxial BaM (0001) on
10nm MgO (111) template on 6H-SiC (0001) as a function of BaM film thickness.
For comparison, a Fe3O4 spectra is included, where the broad Fe 2p 3/2 peak
indicates the presence of both Fe3+ and Fe2+ cations. ....................................... 190 Figure 86: (a) Bright-field image of cross-section TEM of BaM grown on 6H-SiC with a
10nm MgO template, (b) SAED pattern from the SiC-film in [1100] zone axis,
BAM is epitaxial with respect to SiC besides the presence of the amorphous SiOx at
the interface. (c)The interface of BaM film grown on 6H-SiC with 10nm MgO
shows existence of an amorphous SiOx layer (~3 nm) followed by a spinel transition
layer (~20nm), (d) Higher resolution TEM shows an abrupt interface between BaM
and the spinel transition layer, both structures are crystalline and epitaxial as shown
by the inset digital diffractograms, Mg-ferrite is in [11 2 ] and BaM in [1 1 00]
orientation. ........................................................................................................ 192 Figure 87: Elemental profile across the SiC-film interface. a) HAADF image from the
SiC-film interface, vertical line denotes the regions from which the EDX is taken, as
shown in (b). ..................................................................................................... 193 Figure 88: Fe 2p photoemission spectra obtained with Mg k X-rays for 5 nm BaM grown on
SiC (a) with MgO template, and (b) without MgO template. ........................... 196 Figure 89: High resolution cross-section TEM of the interface of BaM grown on 6H-SiC,
shows existence of an amorphous SiOx layer (~7 nm) followed by a spinel Fe3O4
transition layer. ................................................................................................. 197 Figure 90: XRD pattern of a 50 nm thick Fe3O4 (111) film. The right inset: RHEED patterns
of the Fe3O4 (111) thin film when the incident electron beam is along the [11 2 0] and
[1100] azimuths of 6H-SiC respectively. ......................................................... 199 xii
Figure 91 : Magnetic hysteresis loops by VSM measurement of 50 nm Fe3O4 (111) thin films
at T ~300K after the subtraction of the substrate diamagnetic contribution..... 200 Figure 92: High resolution cross-section TEM of the interface of Fe3O4 grown on 6H-SiC,
shows a sharp interface without indicating formation of SiOx between the film and
the substrate. ..................................................................................................... 201 Figure 93: RHEED patterns along [11 2 0] and [1 1 00] azimuths respectively from: H2
cleaned SiC (a & b), 50 nm FO deposited on SiC(c & d), 3 nm BTO deposited on FO
(e & f), and another 50 nm FO deposited on BTO (g & h). ............................. 203 Figure 94: 2m×2m AFM images of (a) FO/BTO(3nm)/FO/SiC (0001), and (b)
FO/BTO(50nm)/FO/SiC(0001). ....................................................................... 203 Figure 95: XRD patterns of (a) FO/BTO (3nm)/FO/SiC, and (b)FO/BTO (50nm)/FO/SiC
heterostructure; * designates peaks from the Al sample stage. ........................ 204 Figure 96: Magnetic hysteresis loops by VSM measurement of FO/BTO (4nm)/FO/SiC
(black higher curve) and FO/BTO (50nm)/FO/SiC (red lower curve) heterostructure
at T ~300K after subtraction of the substrate diamagnetic contribution. ......... 205 Figure 97: Ferromagnetic resonance absorption spectra of FO/BTO/FO/SiC with BTO layer
thickness ranging from 0 to 50 nm. .................................................................. 207 Figure 98: Magnetization curves taken before (black higher curve), and after (red lower curve)
electrical poling of the FO/BTO (50nm)/FO/SiC. ............................................ 208 List of Tables
Table 1: Classification of Materials according to Magnetic Properties [24,25] ................... 8
Table 2: Comparison of the electrical properties of Silicon, 3C-SiC, 6H-SiC, and 4H-SiC [66]
............................................................................................................................ 29
Table 3: Coercivity values of BaM thin films with various templates [93]........................ 58
Table 4: Comparison of deposition conditions for BaM/TiO2/Si by Wee and Kim et al.... 59
Table 5: Curve fitting parameters for XPS Fe 2p3/2 spectra [109] .................................... 80
Table 6: Lattice constant and crystal structure of iron oxides compared with Al2O3, MgO and
Pt ......................................................................................................................... 82
Table 7: A list of photoelectron energies and sampling depths for selected elements that are of
interest for this thesis. ....................................................................................... 112
Table 8: Three generations of PLD deposited BaM films with various templates ........... 121
Table 9: Magnetic property comparison of BaM films grown on SiC with and without MgO
interwoven layers .............................................................................................. 123
xiii
1.0 INTRODUCTION
Ferrite/ferroelectric heterostructures have attracted much attention in recent years
because of their ability to realize dual magnetic and electric field tunability; i.e., a dielectric
polarization variation in response to an applied magnetic field, or magnetization induced by
an external electric field [1,2,3,4]. The simultaneous magnetic and electric tunability in such
structures can be potentially applied in a wide range of applications including non-volatile
memory, sensors, controllers, and RF/microwave communication systems [ 5 , 6 , 7 , 8 ].
However, the attempts that have been made to extend the capabilities to millimeter wave
band frequencies are very limited [9,10]. The proper choice of a high anisotropy ferrite
component is critical to millimeter wave applications (frequency higher than 30 GHz). In
addition, the ferrite components used in today’s transmit/receive (T/R) modules in
microwave devices are bulky, costly, and excess energy consuming. A longstanding
objective of the ferrite device community is developing planar ferrite devices that send,
receive, and manipulate electromagnetic radiation and efficiently couple these devices to
complementary metal-oxide semiconductor (CMOS)-based integrated circuits [11].
Barium hexaferrite (BaM, BaFe12O19) is an ideal candidate for high frequency
microwave device applications because of its strong uniaxial anisotropy (HA~17 kOe) and
can be tuned to ferromagnetic resonance (FMR) at frequencies higher than 40 GHz with
relatively small applied magnetic fields [12,13]. The performance of today’s ferrite
devices would be enhanced and next-generation monolithic microwave integrated circuit
(MMIC) would be possible if ferrite films such as BaM were compatible with CMOS
processing. This goal can be realized by integrating BaM with wide bandgap
1
semiconductors (e.g. SiC or GaN), which can function in high-temperature, high-power,
and high-frequency environments [14,15,16].
Magnetite (Fe3O4) as a well-known spinel ferrite, has recently attracted considerable
attention in spin electronic (spintronic) applications, because of its high Curie temperature
(Tc = 858 K) and close to 100% spin polarization [17,18,19]. Using Fe3O4 as the magntic
layer in magnetic tunneling junctions (MTJs) can lead to ultrahigh tunneling
magnetoresistence (TMR) even at room temperature since TMR is proportional to the spin
polarization of the electrons [20]. Incorporating semiconductors in spintronic devices would
allow a greater flexibility in design and device performance by taking advantage of the high
degree of controlling carrier concentrations, gate voltages, and band offsets as well as the
optical properties of semiconductors [21]. However, very few works have focused on
integration of Fe3O4 on semiconductor substrates such as Si and GaAs, and no work has
been reported on wide bandgap semiconductors such as GaN and SiC. In order to develop
next-generation spintronic devices for high-temperature, high-power, and high-frequency,
the integration of Fe3O4 with wide bandgap semiconductor substrates is necessary. These
next-generation spintronic devices that take advantage of both magnetic materials and wide
bandgap semiconductors are expected to be non-volatile, fast and capable of simultaneous
data storage and consuming less energy [22].
The program goal of this research is to develop ferrite/ferroelectric multiferroic
heterostructures for next generation microwave communication and spintronic devices in
order to provide the fundamental understandings needed for designing a commercial
process in the future. In order to achieve this goal, an effective interface must be engineered
2
and the atomic level interations during film growth must be understood. Molecular beam
epitaxy (MBE) is a perfect tool to effectively understand the nucleation and growth
mechanisms because of its capability of the atomic level control. Thus the strategy of this
work is to use MBE to first grow magnetic ferrites (BaM and Fe3O4) on the wide bandgap
semiconductor SiC, and subsequently understand the growth mechanism and interface
function, and then develop ferrite/ferroelectric multiferroic heterostructure on SiC.
The specific objectives included: understanding the effects of surface/interface and
processing parameters (e.g., temperature and O2 pressure) on thick BaM film growth on
6H-SiC by pulsed laser deposition, building correlations between processing parameters
and film magnetic properties to understand atomic-level mechanisms of BaM growth on
6H-SiC for nanoscale films and applications by MBE, studying the structure and
composition evolution at the BaM/SiC interface in order to engineer a more effective
interface
for
more
efficient
coupling
interactions,
and
demonstrating
ferrite
(Fe3O4)/ferroelectric(BaTiO3) multiferroic heterostructure on SiC to realize the tunning of
magnetic properties by applying an electric field.
Through this research, the integrations of thick device quality BaM film with
6H-SiC has been achieved by PLD through understanding and improving the interface
properties and the MBE integrations of BaM, Fe3O4, and Fe3O4/BaTiO3/Fe3O4 with
6H-SiC have been successfully demonstrated the first time. The high quality BaM film has
the potential to be compatible with the microwave monolithic integrated circuits and can
also function as a magnetic layer in BaM-ferroelectric multiferroic heterostructures for
electrostatic FMR tuning. Furthermore, the successful demonstration of electric field
3
tuning of magnetism in Fe3O4/BTO/Fe3O4/SiC at room-temperature opens new avenues
for introducing ferrite/ferroelectric multiferroics into tunable microwave magnetic devices
and spintronics. In summary, the successful realization of integrating BaM and Fe3O4 on
SiC in this research would lead to develop and optimize the next-generation microwave
communication and spintronic device architectures for high-power, high-frequency
applications in harsh environments.
4
2.0 BACKGROUND
Effective integration of functional oxides with wide bandgap semicondutors is
defined as the multipal oxide components with different functionalities need to be
independently integrated into a single substrate and an efficient coupling can be achieved
between those different functional oxides. The many materials and processing challenges
that must be overcome to effectively integrate magnetic ferrite films (i.e., BaM and Fe3O4)
with 6H-SiC, including lattice mismatches ranging from 3.77 % to 4.76 %, interface mixing
both at the substrate and beween functional layers, stoichiometry control to achieve desired
functional film properties, and structure control to engineer both film properties and
coupling effects. In order to overcome these challenges, it is necessary to incorporate known
characteristics of the magnetic ferrites, mechanisms of thin film growth and processing
influences on those mechanisms, and the current understanding of functionalities (e.g.,
magnetic and electric) of each material as well as expected coupling between those
material porperties.
2.1
Magnetic Ferrite Materials
Magnetic phenomena have been known and studied for many centuries, and are
closely connected to people’s everyday lives today in information technologies ranging
from personal computers to mainframes use magnetic materials to store information on hard
disks. Magnetic materials have attracted attention in a wide range of consumer and
industrial electronics applications, including recording medium, transformers, microwave
communication devices, and spintronics.
5
Ferrites are a group of magnetic oxide materials that have been widely used in
microwave and millimeter wave device applications. Ferrites are ferrimagnetic ceramic
materials with iron oxide as their principle compoments. Ferrites only referred to spinel
structure crystals in the 1950s [23]. Now, they involve all of the iron oxides and a variety of
crystal structures, such as spinels, garnets, hexaferrites, and orthoferrites. The unit cell of
spinel and garnet ferrites is a cubic crystal structure whereas the unit cell of hexaferrites is
hexagonal. The different unit cell structures available in ferrite materials expands the
potential integration oppotunites with semiconductors which can be either cubic (Si, GaAs,
β-SiC) or hexagonal (GaN, 6H-SiC).
2.1.1
Classification of Materials by Magnetic Properties
In terms of magnetic behavior, all materials can be classified into five categories
depending on their bulk magnetic susceptibility (). , defined as
M
is the degree of
H
magnetization (M) of a material in response to a magnetic field (H). Most elements in
periodic table are non-magnetic and classified as diamagnetic and paramagnetic. The atoms
in diamagnetic material have no magnetic moments and no spontaneous magnetization. The
susceptibility of diamagnetic material is small and negative. In comparison, the atoms of
paramagnetic material have magnetic moments that are randomly oriented and the
susceptibility positive, but still small.
Magnetic materials are classified as ferromagnetic, antiferromagnetic, and
ferrimagnetic. For the ferromagnetic material, atoms have parallel aligned magnetic
moments. The susceptibility of ferromagnetic material, such as Iron is large and positive,
and is a function of applied magnetic field. In an antiferromagnetic material, such as Cr, the
6
susceptibility is small and positive and the atomic magnetic moments couple in anti-parallel
arrangements with zero net moment, rather than parallel as in a ferromagnet. The atoms in a
ferrimagnetic material have anti-parallel aligned but unequal magnetic moments, thus the
net moment is not zero. Ferromagnetism and antiferromagnetism can be found in both pure
elements and compounds. However, ferrimagnetism cannot be observed in any pure
element but only in compounds, such as ferrites. Both Ferromagnetic and ferrimagnetic
materials have spontaneous magnetization, which contribute to wide applications at
microwave and millimeter wave frequency communication devices, as well as at magnetic
recording devices .
The five main classes of materials and the criteria by which they are differentiated
from each other are summarized in Table1. Based on the classification, barium hexaferrite
(BaO•6Fe2O3) is a kind of ferrimagnetic material with around 3 at the initial point.
7
Table 1: Classification of Materials according to Magnetic Properties [24,25]
Type of
Magnetism
Diamagnetism
Magnitude of
Susceptibility
Atomic / Magnetic Behavior
Example /
Susceptibility
Small and negative
~ -10-6 to -10-5
Au
Cu
-2.74x10-6
-0.77x10-6
Paramagnetism
Small and positive
~ +10-5 to +10-3
β-Sn
Pt
Mn
0.19x10-6
21.04x10-6
66.10x10-6
Ferromagnetism
Large and positive,
function of applied field,
microstructure dependent.
Fe
~100,000
Antiferromagnetism
Small and positive, as
paramanetism
Cr
3.6x10-6
Ferrimagnetism
Large and positive,
function of applied field,
microstructure dependent
Ba
ferrite
~3
As shown in Table 1, the difference between ferromagnetic and ferrimagnetic
materials lies in the spin alignment of the magnetic ions. For a ferromagnetic material, all
the magnetic moments align spontaneously in one direction, however for a ferrimagnetic
material, the magnetic ions that occupy different crystal sites in a unit cell may, even if ions
of the same element, align in opposite directions. The opposing moments in a ferrimagnetic
material are unequal, and thus a spontaneous magnetization remains. In ferrimagnetic
materials, ions that have both the same crystallographic sites and magnetic moment in a unit
cell repeat with the period structure of the lattice and form magnetic sublattices. Each
sublattice has the magnetic ions align in one direction.
8
2.1.2
Spinel Ferrite Fe3O4
The spinel ferrite has a general formula of Me2+O· Fe3+2O3, where Me can be any
divalent magnetic ion (e.g., Ni, Co, Mn, or Zn) and Fe can be replaced by other trivalent
magnetic ions (e.g., Cr, Co, or Ni). Spinels present a cubic structure, which can be
subdivided into two magnetic sublattices whose atoms occupy tetrahedral sites (A sites) and
octahedral sites (B sites).
The magnetic moment of cations within each sublattice align parallel to each other
and anti-parallel to the other sublattice. The normal spinel is described as the inversion
parameter
δ
equals
zero,
where
δ
in
the
formular
is
expressed
as
(Me1-δFeδ)tet[MeδFe2−δ]OctO4. In the normal spinel, 8 divalent metal cations occupy the 64
possible A sites and 16 trivalent metal ions occupy the 32 possible B sites. Typically, natural
spinel materials (e.g., Fe3O4, CoFe2O4, NiFe2O4, etc.) present an “inverse” structure as the
inversion parameter δ equals one, in which 8 divalent cations occupy B sites and 16 trivalent
cations distribute equally among A and B sites. The cation distribution in most spinel
ferrites is neither purely inverse nor normal but some mixture of both. The magnetism in
these spinel structures arises from a super-exchange mechanism [26] and the net moment is
determined by the sum of individual moments of all sites. For an inverse spinel, the
magnetic moments of the trivalent magnetic ions in sublattice A cancel those of the trivalent
magnetic ions in sublattice B. Therefore, the net magnetic moment of the inverse spinel
ferrites is reduced from the normal spinel structures, if one was possible.
Magnetite, Fe3O4, as one of the most popular spinel has the inverse spinel structure
with a lattice constant of 8.396 Å [27,28]. The inverse spinel structure consists of a
9
face-centered-cubic (fcc) oxygen sublattice, with Fe3+ ions filling A sites and equal amounts
of Fe2+ and Fe3+ ions filling the B sites. As shown in Figure 1, the Fe3O4 unit cell consists of
four (001) layers. Each layer contains the oxygen anions and the B site iron cations, and the
A site iron cations are located halfway between these layers.
Figure 1: The inverse Fe3O4 spinel crystal structure [29]. One unit cell consists of four
(001) layers, each layer contains the oxygen anions (large white circles) and the
octahedral iron ions (small black circles), the tetrahedral sites (small grey circles) are
located halfway between these layers.
A model of a (001) surface plane is shown in Figure 2. The B sites align in strings
along the <110> directions. The spins of the Fe2+ and Fe3+ cations at the B sites are aligned
and coupled antiparallel to the Fe3+ cations at the A sites via superexchange from
overlapping charge distributions with the nonmagnetic oxygen ion. This leads to a net
magnetic moment close to 4 B per each Fe3O4 formula unit [30]. Electron hopping happens
between the Fe2+ ions and Fe3+ ions, contributing to a high conductivity (ca. 100Ω-1cm-1) at
room temperature [31]. Around 120 K the so-called Verwey transition (Tυ) occurs [32]. The
structure distorts from the cubic symmetry at Verwey transition temperature and thus a
charge ordering occurs at the B sites, which reduces the conductivity by a factor of ~ 100
10
[33,34]. The exact transition temperature is dependent on the crystallinity and phase purity
of the magnetite crystals [35,36], again exemplifying the need for exact stoichiometric and
structure control for tuned behaviors.
Figure 2: A schematic drawing of Fe3O4 (001) plane[37], where the 90o superexchange
interaction at the octahedral site is shown as solid black line, the superexchange
interaction between iron ions on the octahedral and tetrahedral sites is shown as
dashed black line, and the double exchange between iron ions on the octahedral sites
is shown as solid grey line.
Recently, Fe3O4 has attracted a lot of interests because it has been predicted as a half
metallic ferromagnet (HMF) [38]. Half metals show only minority spin (spin down)
electrons presented at the Fermi level [38,39]. This implies that these materials act as a
metal (conducting) for one spin orientation, while acting as an insulator (non-conducting)
for the opposite orientation. As a consequence, these materials have 100% spin polarized
conduction electrons in the 3d conduction bands of minority spins [40]. Since only one spin
direction if conductive, then an applied magnetic field will align all spins that act as a way to
11
regulate transfer of electrons. At an applied magnetic field, tunneling of the high spin
polarized conduction electrons will lead to significant tunneling magnetoresistance (TMR)
in ferromagnetic metal-insulator heterostructure, which is very important for spintronic
devices (will be introduced in section 2.2). However, in order to be compatible with the
semiconductor technique and process, these materials must meet the practical requirements
such as high Curie temperature, low deposition temperature, and thinness technique [39].
Fe3O4 is a very promising candidate for TMR devices due to its high Curie temperature (850
K) in comparison with other half-metal oxides (e.g., ~360 K for La0.7Sr0.3MnO3, ~395 K for
CrO2) [40]. Using Fe3O4 electrodes in magntic tunnel junctions (MTJ’s) brings up new
performances in spin electronics, leading to potential TMR values much higher than those
obtained with usual ferromagnetic electrodes [38].
2.1.3
Hexagonal BaFe12O19 (BaM)
Hexaferrites have lower crystal symmetry compared to cubic spinels and several
types have been discovered and studied [41,42]. Most hexaferrites are classified base on
their composition into one of four types called M, W, Y, and Z. It is the M type barium
hexaferrite (BaFe12O19 or BaM) that is focus of this work. The M type hexaferrites have a
formula of BaFe12O19, where Ba can be substituted by Pb or Sr, and Fe can be replaced by
trivalent ions such as Sc, Al, or Ga. BaM has the magnetoplumbite structure, where the unit
cell can be expressed as RSR*S*. The “*” denotes a 180º rotation of the R or S building
blocks around the c-axis. The S block is a spinel block with the (111) plane transverse to the
BaM c-axis and the the chemical formula Fe6O8. The R block is hexagonal and has a
chemical formula of BaFe6O11. As shown in Figure 3, one unit cell consists of 10 oxygen
12
layers. Five oxygen layers make up one molecular formula so the unit cell contains two
BaFe12O19 formulas rotated 180º with respect to each other around the c-axis. One of the
oxygen ions in every five layers is replaced by a divalent cation of Ba2+ (solid circle) which
has a similar ionic radius to the oxygen. The iron ions occupy octahedral, tetrahedral and
bipyramidal interstitial sites [43]. Like spinels, the ferrimagnetic properties in BaM are
resulted from the exchange interaction of Fe3+ between different sites. The strongest
exchange interaction is the Fe-O-Fe bond that is the closest to 180o, which determines the
spin distribution in each block [44]. In the S block, the strongest exchange interaction is the
superexchange of Fe3+ between the tetrahedral site and octahedral site, while in the R block,
the strongest exchange lies between the bypyramidal site and octahedral site.
c=23.2
a=5 88
Figure 3: The unit cell of BaFe12O19, courtesy to Dr. Aria Yang. One unit cell consists
of 10 oxygen layers, The oxygen ions (small white spheres) form a hexagonal or cubic
closed packed lattice along the [001] direction.
The lattice constants of the BaM unit cell are 23.2 Å in the c-direction and 5.89 Å in
13
the a-direction [45]. The uniaxial symmetry found in hexaferrites due to two trigonal
bipyramidal sites contributes to high magnetic anisotropy fields (17 kOe). The high cubic
symmetry of spinels leads to low anisotropy field (0 ~ 2 kOe) for spinels. The higher
anisotropy fields of the hexaferrites compared to spinels make them more useful in the
microwave frequency applications. In addition, the pure 3+ oxidation state of Fe ions in
BaM contributes to better insulating properties compared to spinels, which provides BaM
advantages over spinels in the microwave/RF applications that require low eddy current
losses.
Because of the complexity of the crystal structures of hexaferrites, structural
disorder, ranging from cation disorder to grain boundaries, has a measurable effect on the
electronic and magnetic properties that can significantly impact device performance.
Moreover, these ferrites allow easy diffusion of cations and oxygen ions because of the open
crystal structures. As will be demonstrated in the literature review, the complexity and
openness of the BaM structure provide severe challenges in achieving integrated magnetic
films for microwave frequency device or for multifunctional heterostructures.
2.1.4
Magnetic Properties
The magnetic properties of a material include both static and dynamic (microwave)
properties. The static properties can be measured by vibrating sample magnetometry (VSM),
while the dynamic properties can be deduced from the ferromagnetic resonance (FMR)
measurements. These methodologies enable understanding of the magnetic behavior of the
magnetic material, and then determine its utility for a particular application. The important
magnetic properties include the saturation magnetization (4Ms), the magnetic anisotropy
14
field (HA), the coercivity (Hc), and the FMR linewidth (ΔH).
2.1.4.1 Anisotropy Field
Magnetic Anisotropy in a crystalline magnetic material refers to the preference for the
magnetization to depend on the crystallographic direction in which the magnetic dipoles are
aligned in the absence of an external magnetic field. BaM has the anisotropy field (easy axis)
along c direction, while Fe3O4 has the anisotropy field along <110> directions.
2.1.4.2 Magnetic Hysteresis
Hysteresis is well known in ferromagnetic or ferrimagnetic materials. When an
external magnetic field is applied to a ferro- or ferri-magnet, the relationship between
magnetic field strength (H) and magnetic flux density (B) is not linear. In such materials, the
changes of B always lag behind the change of H. This phenomenon is called magnetic
hysteresis. Figure 4 shows a typical hysteresis loop.
Remanence
Applied Field
0
Hc
Applied Field
Figure 4: A typical hysteresis loop for a ferro- or ferri- magnetic material [46]. Point
“a” and “d” indicate the satuation magnetization (Ms), where the material has
reached its magnetization saturation; point “b” and “e” indicate the level of residual
magnetism (Mr) in the material; point “c” and “f” indicate the force (Hc) required to
remove the residual magnetism from the material.
15
The loop is generated by measuring the magnetic flux (B) of a ferromagnetic material
while the applied field H is changed. If a magnetic field is applied to a thoroughly
demagnetized ferri- or ferro-magnetic material, its magnetic flux density will gradually
increase, following the dashed line from “o” to “a”. As the line demonstrates, the greater the
amount of applied field, the stronger the magnetic field in the component. At point "a"
practically all of the magnetic domains are aligned and an additional increase in the applied
field will produce very little if any increase in magnetic flux. The material has reached its
magnetization saturation, Ms. Ms is a measure of the maximum amount of magnetic field
that can be generated by a material. It will depend on the strength of the dipole moments of
the atoms that make up the material and how densely the atoms are packed together. Iron
saturates at about 1.6 T while ferrites normally saturate between about 0.2 T and 0.5 T [47].
The saturation magnetization of barium ferrite bulk material is 0.48 T at room temperature
[43]; however thin films usually have a lower saturation magnetization than that of the bulk
material.
When H is reduced back down to zero, the curve will move from point "a" to point
"b." At this point, it can be seen that some magnetic flux remains in the material even though
the external applied field is zero. This indicates the remanence or level of residual
magnetism in the material. Some of the magnetic domains remain aligned but some have
lost their alignment. The magnetization remaining in the sample when the applied field is
zero is called remanence magnetization Mr.
As the external applied field force is reversed, the curve moves to point "c", where
the flux has been reduced to zero. This is called the point of coercivity on the curve. The
16
reversed applied field has flipped enough of the domains so that the net flux within the
material is zero. The force required to remove the residual magnetism from the material, is
called the coercive force or coercivity, Hc.
As the magnetizing force is increased in the negative direction, the material will
again become magnetically saturated but in the opposite direction (point "d"). Reducing H
to zero brings the curve to point "e." It will have a level of residual magnetism equal to that
achieved in the other direction. Increasing H back in the positive direction will return B to
zero. Notice that the curve did not return to the origin of the graph because some force is
required to remove the residual magnetism. The curve will take a different path from point
"f" back the saturation point where it completes the hysteresis loop.
The most important static magnetic parameters (e.g., Ms, Mr, and Hc) that
characterize a magnetic material are related to its hysteresis loop. The hysteresis loop can be
measured by vibrating sample magnetometry (VSM), and is very helpful to determine the
utility of a material for a particular application. In this study, the target applications require
high Ms and low Hc.
2.1.4.3 Magnetic Domains
Many
ferrimagnetic
materials
of
macroscopic
dimensions
exhibit
zero
magnetization when an applied external field is absent. This is because these ferrimagnetic
materials are composed of many small regions, each of which is magnetized to saturation
but is randomly oriented with respect to each other. Each of these regions is called a
magnetic domain, and within these areas the magnetic moments of the atoms are grouped
together and aligned. The net magnetic moment is the vector sum of all local magnetic
17
moments of each domain.
The regions separating magnetic domains are called domain walls. At these walls,
the magnetization rotates coherently from the direction in one domain to that in the next
domain. The domain wall separates the “up” and “down” magnetic moments by a certain
distance, which is a consequence of the competition between the anisotropy and exchange
energies. The anisotropy energy is minimized when the spin is parallel to the easy axis of
magnetization and the exchange energy is minimized when the spins are parallel to one
another. The width of the domain wall is determined by minimizing both the anisotropy and
exchange energies within the wall. Magnetic domains can be detected using Magnetic Force
Microscopy (MFM).
2.1.4.4 Microwave Properties
Microwave properties of a magnetic material may be deduced from the
ferromagnetic resonance (FMR) measurement. In a microwave field, the magnetic moments
precess around the internal magnetic field. If the frequency of an applied microwave field
coincides with the frequency of precession, the magnetic resonance occurs. This
phenomenon is called FMR.
The fundamental physics of FMR originates from gyromagnetic motion, described
as:
dM
  M  H eff
dt
Where M is magnetic moment, and γ =
Eqn. 1
g
e
is gyromagnetic ratio, e and m are
2mc
the electron charge and mass respectively, and g is Lande g-factor. H eff is the effective
18
internal magnetic field, which is the sum of external field, shape anisotropy field and
magnetocrystalline anisotropy field, and can be described as:
H eff  H  H d  H A
Eqn. 2
Here, H is the external field, H d = N  M , and H A = NA  M , N and NA are
tensors that denote the shape anisotropy and magnetocrystalline symmetry, respectively.
When the microwave signal is introduced into ferrite thin films, with the assumption of
having both external field and magnetocrystalline anisotropy field normal to the film, the
FMR frequency condition may be simplified and given as:
f 

( H  HA  4Ms )
2
Eqn. 3
where Ms is the saturation magnetization.
The saturation magnetization and magnetocrystalline anisotropy field determine the
natural resonant frequency through the magnitude of the internal field in the absence of an
external applied field. At different external fields, the resonant frequency will shift to higher
or lower frequency bands required for certain device applications. In each FMR
measurement, a specific and fixed microwave frequency (40 GHz to 60 GHz) was chosen,
and the external magnetic field was swept from a particular range depending on the
measurement purpose.
In a polycrystalline ferrite, the orientations of each crystal are random, and FMR
spectra are a summation over the FMR absorption of each grain in the polycrystalline
material. From experiment data such as Figure 5, the overall FMR linewidth can indicate the
degree of crystalline alignment in a thin film. The narrowest linewidth represents the most
uniform FMR response throughout the sample and would be expected for a single crystal
19
material. The poorer the alignment of the crystal structure, the greater the number of
separate magnetic domains and the broader the FMR adsorption for the material.
Linewidths in polycrystalline material may be as high as 4Ms. Even for an oriented
hexaferrite, the FMR linewidth is much higher than that of single crystal because not all
crystallites are oriented along one direction. Other structure defects in the film (oxygen
vacancies, chemical phase inclusions, substrate diffusion, etc.) may also increase the
linewidth of the FMR data.
Figure 5: Ferrimagnetic resonance spectra for ~50m thick BaM film on (111) GGG
substrate at 58GHz [48], the reasonably narrow FMR linewidth was ~0.068 kOe.
FMR spectroscopy has been widely employed to characterize high frequency losses
in magnetic materials [49,50]. The FMR linewidth (∆H) represents the rate at which energy
is transferred from the precessing spins to the lattice [51], causing loss of efficiency in a
microwave frequency device. The linewidth can be affected by compositional
inhomogeneities, defects, spin wave excitations, strain, etc [49,50, 52 ]. Although the
broadening of the FMR line has not been fully understood, it is clear that homogeneous and
defect-free materials can minimize the extrinsic contributions to the linewidth. Until
20
recently [53,54], the best reported value of the peak-to-peak linewidth was about 27 Oe at
60.3 GHz for 0.85 m BaM films on sapphire and ~ 50 Oe for 45 ~ 80 m BaM films on
garnet. Bulk BaM single crystals yield values of ∆H of 6 – 30 Oe [51,55].
For microwave frequency device applications, the needed magnetic properties of
barium ferrite films are: low coercivity (Hc), high saturation magnetization (4Ms) and
squareness (Mr/Ms), narrow FMR linewidth (∆H), and high magnetic anisotropy field
(HA).Usually, lower Hc will decrease the magnetic loss, whereas higher magnetization and
HA will allow for higher operational frequency, and higher loop squareness (Mr/Ms)
indicates the degree to which the material is self-biased.
2.2
Spintronics
Spin-based electronics, spintronics, has attracted a lot of attention recently because
it has the potential to dramatically reduce the power and time needed to drive transistors and
other electronic devices as well as exponentially enhance the information processed.
Spintronics is the device that either functions or processes information by using both the
flow of electrons as in strandard electrical devices and the quantum spin of electrons in a
current flow. Spin is a purely quantum phenomenon roughly similar to the directional
behavior of a compass needle, which point either “up” or “down” in relation to a magnetic
field. The movement of spin, like the flow of charge, can also transmit information among
devices. One advantage of spin over charge is that spin can be easily manipulated by
external applied magnetic fields, a property already used in magnetic storage technology.
Another important property of spin is its long coherence, or relaxation, time—once created
21
it tends to stay that way for a long time, unlike charge states, which are easily destroyed by
scattering or collision with defects, impurities or other charges [56].
The most common application of this spin-related effect is a giant magnetoresistance
(GMR) device. A typical GMR device consists of at least two layers of ferromagnetic
electrodes separated by a paramagnetic material. When the two magnetization vectors of the
ferromagnetic electrodes are aligned (can be thought as “on”), the spin aligned current
propagated through the paramagnetic can retain their majority status in the other
ferromagnetic electrode producing low electrical resistance. However, if the magnetic fields
of two ferromagnetic electrodes are antiparallel to each other or “off”, the spin aligned
current encounters a high number of carriers with opposite spin alignment in the other
electrode. This produces high spin scattering and a high resistence to current flow.
Another common spintronic device is magnetic tunnel junction (MTJ), in which
tunnel magnetoresistance (TMR) can occur. MTJ consists of two ferromagnets separated by
a thin insulator, typically with thickness of a few nanometers. The insulating layer is so thin
that electrons can tunnel through the barrier if a bias voltage is applied between the two
ferromagnetic electrodes. In MTJs the tunneling current depends on the relative orientation
of magnetizations of the two ferromagnetic layers, which can be changed by an applied
magnetic field. This phenomenon is called TMR.
The GMR and TMR effects in thin magnetic layers are both related to
spin-dependent transport, and can be potentially exploited in a variety of advanced devices
such as highly sensitive magnetic sensors (e.g., read heads for magnetic recording), and
nonvolatile magnetic memories (MRAM’s). There are many challenges still to be met
22
before the full potential of spintronic applications can be realized, including the challenge of
integrating semiconductor materials into spintronic devices.
2.3
Multiferroics and Ferrite/Ferroelectric Heterostructure
Multiferroics have been defined as materials that simultaneously exhibit more than
one primary “ferroic” order parameters and have coupling between these order parameters
[57,58]. By this original definition, only a single phase multiferroic [59] is a material that
simultaneously possesses two or more of the primary “ferroic” order parameters, which are
ferroelectricity, ferromagnetism, and ferroelasticity. However, the definition of
multiferroics has been expanded as to include non-primary order parameters (e.g.,
antiferromagnetism or ferrimagnetism). By this modified convention, a whole class of
composites and multilayer structures, such as ferrite/ferroelectric heterstructure, can be
considered as multiferroics. Such multiferroic materials have received a lot interests due to
the large achievable stress/strain-mediated magnetoelectric (ME) coupling. This coupling
can be realized as, for example, a dielectric polarizaiotn variation in response to an applied
magnetic filed, or as an induced magnetization from an external electric field. The promise
of coupling between magnetic and electronic order parameters and the potential
applications in novel microwave magnetic devices has attracted the attentions of researchers
worldwide. To better understand magnetoelectrics and multiferroics, the different types of
order parameters and coupling that can occur in these materials need to be investigated.
There are 20 (of 32) crystal classes of materials exhibit electrical polarity when
subjected to a stress and are called piezoelectric materials [60]. Of those 20 piezoelectric
23
crystal classes, 10 are called polar because they show a unique polar axis, and thus have a
spontaneous polarization [61]. A ferroelectric material (ferroelectrics) is a material that has
spontaneous electric polarization in the absence of an electric field and can be reversed by
an external electric field. Ferroelectrics can undergo a phase transition from a
high-temperature phase that behaves as an ordinary dielectric to a low temperature phase
that process a spontaneous polarization whose direction can be switched by an external
electric field [60].
The most common type of ferroelectrics is the ABO3 perovskite ferroelectrics like
PbTiO3 and BaTiO3. In those perovskites, the Ti 3d – O 2p orbital hybridization is
important for stabilizing the ferroelectric distortion [60]. It is also reported that most
perovskite ferroelectrics have B-site ions that are formally d0 in nature and thus the lowest
unoccupied energy levels are the d states and they tend to hybridized with the O 2p orbitals
resulting in the double well potential [62].
A ferroelastic material (ferroelastics) is a material that has two or more orientation
states (spontaneous strain) in the absence of a mechanical stress (and electric field) and can
be shifted from one to another of these 4 states by mechanical stress. It is necessary that two
of the orientation states are identical or enantiomorphous in crystal structure and different in
mechanical strain tensor at absence of mechanical stress (and electric field) [60]. The
magnetic materials are introduced previously in section 2.1.
The multiferroic materials simultaneously display two or all three of the properties
ferroelectricity, ferromagnetism, and ferroelasticity. Figure 6 shows the overlap required of
ferroic materials to be considered as multiferroic [63]. Only small subsets of magnetically
24
and electrically polarizable materials are either ferromagnetic or ferroelectric and even
fewer can simultaneously process both order parameters and thus be classified as
multiferroics. In these multiferroics where multiple order parameters exist, it is possible that
not only the electric fields can realign the polarization and the magnetic fields can induce
the magnetization, but also there is a possibility to couple an electric field to magnetization
and a magnetic field to electric polarization.
Figure 6: The relationship between multiferroic and magnetoelectric materials [63].
The intersection (red hatching) represents the multiferroic materials that are
ferromagnetic and ferroelectric.
Figure 7 [64,65] illustrates the concept of multifunctional heterostructures and
multiferroics. Using a single phase multiferroics bismuth ferrite (BFO) as an example, since
BFO is ferroelectric and antiferromagnetic, the induced stress/strain from an external
electric field will impact the magnetic properties. However, such single phase materials in
which ferromagnetism and ferroelectricity arise independently are rare [62], which can be
understood by investigating such factors as symmetry, electronic properties, and chemistry.
25
There are only 13 point groups can give rise to multiferroic behavior [60]. In addition,
ferroelectrics are insulators by definition while ferromagnets are often metallic in nature.
There are only a few antiferromagnetic ferroelectrics despite the fact that both phases are
typically good insulators. A number of ferrimagnets and weak ferromagnets that are in fact
insulating in nature can be identified. Lastly, as discussed in section 2.1.1, many perovskite
oxide ferroelectrics like PbTiO3 and BaTiO3 have B-site ions with formal d0 electron
configurations that this was necessary for the formation of a B-site driven ferroelectric
distortion in these materials [62]. This means there are no d electrons to create magnetic
moments in these materials and thus creating both magnetic and ferroelectric order is
difficult. Therefore, the contradiction between the conventional mechanism of off-centering
in ferroelectrics (which requires an empty d orbital) and the formation of magnetic order
(which results from partially filled d orbitals) result in the scarcity of single phase
multiferroics [62].
Figure 7: Schematic illustration of the relationships between different functionalities
of the materials [60,64,65]. Where E: electric field, H: magnetic field, P: dielectric
polarization, M: magnetization, : stress, and : strain.
That, in turn, has resulted in a shift in focusing from the single phase multiferroics to
26
composites heterostructure multiferroic. Among those multilayer heterostructure,
ferrite/ferroelectric heterostructures have recently became a subject of intense interest due
to the wide applications of those magnetic field tunable ferrite materials in microwave and
millimeter wave planar devices, such as tunable phase shifters, resonators, and delay lines
[5]. However, the attempts that have been made to extend the capabilities to frequencies
higher than 5GHz are very limited [9,10]. The proper choice of a high anisotropy ferrite
component is critical to millimeter wave applications. BaM is an ideal candidate for high
frequency microwave device applications because of its strong uniaxial anisotropy
(HA~17kOe) and can be tuned to ferromagnetic resonance (FMR) at frequencies higher than
40GHz with relatively small applied magnetic fields [12,13]. The performance of current
ferrite devices would be enhanced and next-generation monolithic microwave integrated
circuit (MMIC) would be possible if ferrite films such as BaM were compatible with
complementary metal-oxide semiconductor (CMOS) processing. This goal can be realized
by integrating BaM with wide bandgap semiconductors (e.g., SiC or GaN), which can
function in high-temperature, high-power, and high-frequency environments.
2.4
Wide Bandgap Semiconductor
Wide bandgap semiconductors are semiconductors that have a bandgap energy
greater than the 1.1 eV bandgap of silicon [66]. Because of wide bandgap, they are
compatible with high-temperature, high-power and high-frequency applications. Silicon
carbide (SiC) is a wide bandgap semiconductor which has over two hundred different
27
polytypes (same chemical formula, but different crystalline structure), each with specific
physical and electrical properties. Those polytypes are characterized by their stacking
sequence, which is determined by the repeating pattern of the atomic pairs and is
responsible for the unit cell structure.
The three most common types of silicon carbide are 3C-SiC, 6H-SiC and 4H-SiC.
As for the nomenclature, the number indicates the number of atomic stacks per unit cell and
the letter indicates the crystal structure. Figure 8 below illustrates the difference in stacking
sequence between 3C-, 4H-, and 6H-SiC, which are designated as ABCABC, ABCBABC,
and ABCACBA, respectfully. For general characterization, the nomenclature A, B, and C
represent different stacking options. With a base layer, A, the two subsequent stacking
options are B, or C. Assuming the second layer consists of B, the two stacking options for
the next layer will be A, or C.
(a)
(b)
(c)
Figure 8: Different 2D stacking sequences for SiC. (a) 3C-SiC, (b) 4H-SiC, (c) 6H-SiC,
the black dots represent the carbon atoms and the hollow dots represent the silicon
atoms [67].
Comparing some key electrical characteristics of the three common SiC polytypes
and Si (Table 2), it is found that the fundamental difference between silicon and those three
SiC polytypes is the larger bandgap of SiC. Just because wide bandgap provides a greater
28
energy barrier to the thermal excitation of the atoms, SiC can be operated in higher
temperature environments than Si electronic devices. The higher thermal conductivity and
saturated electron drift velocity make SiC be more compatible than Si in high power, high
frequency applications. In addition, the strong covalent Si-C bond provides this material
high chemical inertness. A wide range of electronic, communication and sensory devices
require the capability in harsh environment, where the wide bandgap semiconductors (e.g.,
SiC) can function. These kinds of devices have attracted much attention in aerospace,
automobile, and military applications.
Table 2: Comparison of the electrical properties of Silicon, 3C-SiC, 6H-SiC, and
4H-SiC [66]
Property
Si
3C-SiC
6H-SiC
4H-SiC
Bandgap (eV)
1.1
2.2
2.9
3.2
Electron Mobility (cm2/V s)
1500
1000
600
1000
Hole Mobility (cm2/V s)
600
40
24
120
Thermal Conductivity (W/cm K)
1.5
5
5
5
1
2.5
2
2
Saturated Electron Drift Velocity (x 107cm/s)
2.5
Thin Film Nucleation and Growth
In order to understand the complexities of forming the BaFe12O19/SiC interface and
growing a BaM thin film, it is necessary to understand the basic thermodynamics, kinetics
and material considerations of film nucleation and growth. By understanding the variables
influencing the nucleation and growth, it is possible to design processes that can form an
interface and resulting film of specifically desired properties.
2.5.1
Epitaxy and Atomic Mismatch
Epitaxy refers to the growth of an extended single crystal film on top of a
29
crystalline substrate [68]. There are two types of epitaxy; homoepitaxy and heteroepitaxy.
Homoepitaxy refers to film growth where the film and substrate are the same material. An
example of this is the epitaxial growth of MgO on an MgO substrate, denoted
(MgO/MgO). Heteroepitaxy refers to the case where film and substrate are composed of
different materials. An example of heteroepitaxy is the growth of BaM on a SiC substrate,
denoted (BaM/SiC), which is the primary focus in this thesis.
In homoepitaxy, the film (also called an epilayer) is generally free of defects and
purer than the starting substrate. In heteroepitaxy, the growth process is more complex
because there are many incompatibilities existing between the film and substrate.
Differences between the two basic types of epitaxy are shown in Figure 9.
FILM
SUBSTRATE
MATCHED
STRAINED
RELAXED
Figure 9: Schematic illustration of lattice-matched, strained, and relaxed
heteroepitaxial structures. Lattice matched heteroepitaxy is structurally similar to
homoepitaxy [68].
There will be no strain in homoepitaxy since the epilayer and substrate crystal are
the same and the lattice parameters are matched entirely. In heteroepitaxy, only when the
lattice mismatch is very small, the heterojunction interfacial structure is like that for
homoepitaxy; otherwise either the strained-layer epitaxy or relaxed-layer epitaxy will
30
happen. The strain that arises in the film is thought to occur due to lattice mismatch or
thermal mismatch between the depositing film and the substrate.
2.5.1.1 Lattice Mismatch
Lattice mismatch is caused by the different lattice parameters of the thin film and the
substrate and is defined by the following equation:
f=
( as  a f )
 100%
af
Eqn. 4
where: f = Lattice mismatch in %,
as = Lattice constant of the substrate, Å, and
af = Lattice constant of the film, Å.
These lattice constants refer to the unstrained lattice parameters of the substrate or
the film and are defined as the length of one unit cell of the crystalline material. A positive f
means the epitaxy film is under tensile strain, and a negative f implied the film is under
compressive strain.
Lattice mismatch can be compensated by rearrangement of the substrate surface
atoms. The crystal structure of a surface is usually different as that of the bulk material
because the loss of periodicity in the vertical direction leaves energetic broken bonds. A
“critical thickness” where the surface structure differs from the bulk occurs when the atoms
with uncompensated energetic bonds seek to lower their energy by reorganizing. As shown
in Figure 10, the selvedge layer is a reconstruction of the top four atomic layers of the
substrate. There is a change in the electron density and electron state in the selvedge layer,
which can cause a difference in the electron properties between the selvedge layer and the
bulk semiconductor substrate, as well as a difference in atomic spacing creating a
potentially different “lattice constant” on the surface than the bulk crystal structure would
31
suggest.
SELVEDGE
Figure 10: Schematic illustration of substrate surface reconstruction of hexagonal
close-packed crystal structure [68].
Lattice mismatch can also be compensated by unit cell arrangement. Equation (2)
only relates lattice constants to a lattice mismatch parameter, assuming a 1:1 unit cell line up.
However, it is possible to have other configurations that reduce lattice mismatch, when the
net lattice spacing between a serious of atoms is similar. As illustrated in the Figure 11, four
unit cells of cubic film line up with three unit cells of a cubic substrate. For example, MgO
film, with a lattice constant of 4.213 Å, has a 34.1% mismatch with GaAs substrate, with a
lattice constant of 5.65 Å. This large lattice mismatch can be resolved when 4 MgO atoms
line up with 3 GaAs atom at the interface, which results in a mismatch of 0.58 % for the 4:3
pattern [69].
MgO
GaAs
Figure 11: Schematic illustrate of 4:3 lattice matching. The black boxes indicate the
lattice cells for GaAs and MgO. While there is a large difference between the size of
the cells, 4 MgO lattices provides a good match to 3 GaAs lattices, as shown by the red
line [69].
32
In addition, lattice strain can be absorbed by use of a template. Templates are thin
films of a few nanometers thick, which serve as a bridge between the substrate and film
The lattice parameter of a template material can often lie between that of substrate and film.
An example of this is BaM on 6H-SiC, where MgO (111) can act as a template. 6H-SiC has
lattice parameters of a=3.08 Å and c=15.11 Å, while that of BaM is a=5.89 Å and c=23.2 Å.
The lattice mismatch between these two materials is -4.76 %. Since two-dimensional
growth requires very small lattice mismatch (~1%) [68], MgO (111) template may be
required to reduce the strain between BaM and 6H-SiC in order to produce high quality thin
film. The lattices parameters of cubic MgO (001) is a=4.21 Å and the equivalent lattice
spacing in the MgO pseudo hexagonal structure is 5.958 Å ( 2 a) (Figure 12). Therefore,
the lattice mismatch between the template and film is reduced to -1.3 %. Figure 13 is the
illustration of lattice matching for BaM on MgO and 6H-SiC.
(111) plane
a(111)=
2 ao=5.96 Å
ao
(a)
(b)
Figure 12: (a) (111) plane in MgO Rock-Salt lattice (b) Pseudo-hexagonal MgO (111)
plane, only O atoms are involved in these two figures.
33
3.08Å
2.98Å
2.94Å
O atoms from BaM
O atoms from MgO
Si atoms from 6H-SiC
Figure 13: Schematic illustration of lattice matching for BaM growth on MgO (111)
and 6H-SiC,only O atoms involved.
2.5.1.2 Thermal Mismatch
The thermal mismatch is caused by the different thermal expansion coefficients of
the film and the substrate. This kind mismatch results in problems upon cooling from the
film growth temperature to room temperature. The film and substrate materials experience
dimensional changes proportional to their coefficient of linear expansion and the change in
temperature. If the coefficients of expansion are seriously mismatched and the temperature
changes are large, then the resulting stress caused by the different dimensional changes of
the film and substrate can result in defects and cracking. The thermal expansion coefficient
of 6H-SiC (s) is 4.310-6/ºC [70], and that of BaM (f) is 1.010-5/ºC [71], which result
in a large thermal mismatch (132 %) between the film and substrate.
When thermal expansion coefficientof substrate s) is smaller than that of film (f),
the film will sustain biaxial tensile stress upon cooling from the deposition temperature to
room temperature. Otherwise, the film will sustain biaxial compressive stress. The stress is
proportional to the thermal coefficient difference of the substrate and film (∆ and the
change in temperature (∆T) [68]. Reducing the temperature change and the thermal
34
coefficient difference will reduce the strain in the film, and then produce high quality film.
2.5.2
Thin Film Growth Mechanisms
The growth of BaM on SiC involves gas-solid interactions. When a molecule or an
atom hits a surface, it can jump off or become adsorbed to the surface. The residence time (τs)
that the adatom or molecule remains on the surface is given by:
τs = (1/υ)exp(Edes/kBT).
Where:
Eqn. 5
υ = Vibrational energy of the adatom on the substrate, s-1,
Edes = Desorption energy, eV, and
kB = Boltzmann’s constant.
This adatom or molecule that stays on the surface can either physically adsorbs on
the surface (physisorption), or chemically bonds to the surface (chemisorption). If the atom
or molecule is stretched or bent but retains its identity, and van der Waals forces bond it to
the substrate, then it is called physisorption. Chemisorption happens when the atoms or
molecules changes its identity though ionic or covalent bonding with substrate atoms.
Adsorption heats or energies (Ep and Ec) are used to quantitatively distinguish the two kinds
of sorptions. Physisorption is considered weak with Ep on the order of less than 0.25 eV,
while chemisorption is stronger with Ec around 1-10eV [68].
During the limited residence time, those adsorbed species can migrate along the
surface, and then either find a favorable site bond to the surface or desorb back into the gas
phase. The probability that a gas molecule that hits a surface will stay permanently on the
surface and involve in the film growth is called a sticking coefficient.
Once the bond is formed with the substrate, film nucleation and growth mechanisms
are controlled by the chemical interaction between these adatoms, the substrate atoms, and
35
the vapor phase atoms. These interactions are highly dependent on surface free energy and
surface chemistry, which is illustrated in Figure 14.
Figure 14: Schematic illustration of basic atomic nucleation of impinging atoms on a
substrate through vapor deposition [68].
The growth modes resulting from this nucleation process can be described using the
capillary theory of heterogeneous nucleation. The Capillarity theory provides a
conceptually simple qualitative model of film nucleation process while this model is
quantitatively inaccurate [68]. After formation of film nuclei with mean dimension r, the
free energy change is given by:
ΔG = a3r3 ΔGV + a1r2γfv + a2r2γfs – a2r2γsv
Eqn. 6
Where: ΔG = Free-energy change of nucleation, mJ,
ΔGV = Chemical free-energy change per unit volume, mJ/m3,
a1 = Geometric constant 2π(1-cosθ),
a2 = Geometric constant πsin2θ,
a3 = Geometric constant π/3(2-3cosθ+cos3),
γ = Interfacial tensions, mJ/m2 and the subscripts f, s, v represent the
film, substrate and vapor respectively, and
θ = Wetting angle, which depends on the surface properties of the
materials and the interface only.
When the system attains mechanical equilibrium, the interfacial tensions or forces
balance surrounding the nucleus yields Young’s equation:
γsv = γfs + γfvcosθ
Eqn. 7
36
This equation helps to better understand the three types of film growth shown
schematically in Figure 15.
Figure 15: Basic modes of thin film growth [72], (a) layer by layer growth (2D), (b)
layer plus island growth (Stranski-Krastanov), (c) island growth (3D).
In layer by layer, or Frank-Van der Merwe, growth, the depositing film “wets” the
substrate and θ is approximately equal to zero, giving a result for Young’s equation:
γsv  γfs + γfv.
Eqn. 8
In this growth mode, the atoms of the depositing film are preferentially bound to the
substrate while not to each other. This growth is also called two-dimensional growth.
Crystalline growth or epitaxial growth of high quality film requires this growth mode for
flatness and uniformity of thickness.
The opposite characteristic growth is island (or Volmer-Weber) growth, in which the
atoms are more tightly bound to each other than to the substrate. In this growth regime, θ is
greater than zero and,
γsv < γfs + γfv.
Eqn. 9
If the film/substrate interfacial tension γfs is negligible, the surface tension of the
film will cause atoms to cluster together that result in three-dimensional growth. Metal and
37
semiconductor films grow on oxide substrates fall into this kind growth mode.
The third method of growth regime or Stranski-Krastanov (S-K) growth is a
combination of the two modes discussed before. The initial formation of the film is
two-dimensional, however after approximately 5-6 monolayers [68] deposition, a transition
to three-dimensional growth occurs. It is considered that the strain energy accumulating in
the growing film is released and, this high energy at the depositing intermediate-layer
interface may trigger island formation [68].
The film characteristics and thin film growth modes are also determined by the
kinetics of film growth. In the model of Figure 14, the process goes as follows: energetic
vapor atoms impinge on the solid surface of the substrate, then they may desorb
immediately or remain on the surface for a period of time given by (τs) given by Eqn.5.
While on the surface, they can migrate across the surface until they meet a nucleus
and attach to it, meet a reactive site on the surface and attach to it, or desorb. As the nuclei
grow, they can coalesce with other nucleating regions to form larger regions of film. When
this coalescence continues, exposure of previously covered substrate allows further
nucleation and the increases in size of the coalescing regions allow more adatoms adding to
their bulk.
Quantitatively, the surface density of these adatoms impinging on the substrate
surface, na, is given by the product of the vapor atoms impingement rate Φ and the residence
time on the surface τs:
na= τsΦ =
sPNA
2MRT
Eqn. 10
Where: Φ = Flux to the surface, molecules/cm2-s,
38
NA = Avogadro’s number, 6.021023/mole
P = Pressure above the substrate, torr,
M = Molecular weight of the gas, g/mole
R = Ideal gas constant, and
T = Absolute temperature in the gas, K.
These adatoms then diffuse across the surface to add themselves to nucleation sites.
The nucleation rate, Ň (nuclei/cm2-s), is basically proportional to the product of three terms
given by:
Ň = N*A*ω
Eqn. 11
Where: N* = Equilibrium concentration of stable nuclei, nuclei/cm2,
A* = Critical area, cm2, and
ω = Rate of impingement of adatoms onto the nuclei, (cm2-s)-1.
The critical area is defined as the area of critical radius, r*. Below this critical radius,
the free energy of the surface decreases as the radius decreases, in essence, allowing the
nucleus to shrink until it vanishes. Above this critical radius, the nuclei can grow. The
critical area is given by:
A* = 2πr*aosinθ.
Eqn. 12
In order to impinge on the nucleus surface area A*, the adatoms are required to
diffuse across the surface with a frequency given by υexp(-ES/kBT), where ES is the
activation energy for surface diffusion. The rate of impingement of adatoms on the surface
of the nuclei is the product of the jump frequency and na:
ω=
sPNA
- Es
υexp(
)
KBT
2MRT
Eqn. 13
The equilibrium concentration N* is given by:
N* = nsexp(-ΔG*/kBT)
Eqn. 14
Where: ns = Total nucleation site density, and
ΔG* is the surface free energy evaluated at the critical radius,
The total rate of nucleation is:
39
Ň = 2πr*aosinθ
PNA
2MRT
nsexp(
Edes - Es - G *
).
KBT
Eqn. 15
Thus the total rate of nucleation is a function of the nucleation energetics which are
strongly correlated to the term ΔG*, and a function of the pressure of the atoms in the gas
phase above the substrate, the temperature of the gas phase and the substrate. A high
nucleation rate contributes to a fine-grained or even amorphous structure whereas a low
value of Ň results in a coarse-grained deposition.
The film nucleation and growth mechanisms discussed previously are just basic
conceptions that apply to very simple system, such as a metal-metal deposition. The lack of
detailed atomistic assumptions makes those mechanisms an attractive broad generality with
the ability of creating useful connections among such variables as substrate temperature,
deposition rate, and critical film nucleus size. However, when considering such multiple
systems as BaM grown on SiC, the growth mechanisms may become more intricate.
In BaM growth, in addition to the interactions between gas atoms and substrate
surface, the three different kinds of gas atoms (Ba, Fe, and O) will interact with each other,
react and then bond together. There will have more kinds of interfacial tensions on the
surface than those in Figure 14. Since different atoms have different sticking coefficients on
the surface, the atomic densities of those three elements on the surface will be different, and
thus influence the nucleation and growth process. Furthermore, there may be some reactions
that happen between the gas atoms and substrate atoms, which change the chemistry and
structure of the interface. Subsequently the interactions between the gas atoms and substrate,
which now has a “new” surface, will be affected. Because of various interactions in this
BaM growth system, the growth mode will become more complex than any of those three
40
basic growth modes based on Young’s equation.
2.5.3
Substrate Surface Impacts
In this work, surfaces are thought of as a few atomic layers on the top of a solid and
represent an atomically clear interfacial separation between condensed-phase and gas-phase
atoms. Associated with these interfaces or surfaces are interfacial energies. Because the
atoms at free surfaces have fewer bonds with surrounding atoms than that of bulk atoms,
they tend to be less constrained and more energetic. The difference of the interatomic energy
of atoms at these two locations is the cause of surface energy. The basic definitions of
surface energy is related to the reversible work (dW) on a material when its surface area (A)
is increased, or dW=dA. The change in either A or can change the surface energy. It is
very important to study the nature of substrate surface energy; because it determines the
interactions between the gas-surface, the reactivity of the surface, and thus the subsequent
film growth.
As discussed in 2.5.1.1, substrate surface reconstruction leads to a different crystal
structure at the surface than that of the bulk material, and then alters the surface free energy.
Different kinds of reconstructions lead to different starting point for the growth process and
different initial template for subsequent film growth. The reconstruction may also impact
surface reactivity [66]. In addition to surface reconstruction, anything that modifies the
energy of the surface will affect the reactivity. One surface modification affecting
gas-surface interactions could be atomic steps or ledges on the surface that can provide
different reaction probabilities for an adsorbed species than an atomic plateau [66],
another is the chemistry of the surface. Although SiC substrate surfaces are scratched and
41
contain both oxygen and carbon contamination, a cleaning procedure has been developed
by our laboratory to create well-controlled, clean, and smooth SiC surfaces with atomic
steps for BaM thin film growth [73].
2.6
Thin Film Deposition Methods
Thin-film deposition is the technique of depositing a thin film of material onto a
substrate or onto previously deposited layers. "Thin" is just a relative expression. The
majority of deposition techniques are capable of controlling layer thickness within a few
tens of nanometers, and some techniques can even allow one layer of atom deposition at a
time. There are many different deposition methods that have been used for BaM thin film
growth, such as rf-sputtering, pulsed laser deposition (PLD), sol-gel, and liquid phase
epitaxy (LPE). In the proposed study, molecular beam epitaxy (MBE) will be employed, for
the first time, to understand the nucleation and growth mechanisms of BaM.
2.6.1
Traditional Deposition Methods for BaM Thin Film Growth
Radio frequency (RF) sputtering uses a plasma (usually a noble gas, such as Argon)
to knock material from a "target" a few atoms at a time. The relatively low target
temperature makes sputtering one of the most flexible deposition techniques. The sputtered
atoms ejected into the gas phase, are not in their thermodynamic equilibrium state, so they
tend to condense back into the solid phase on the substrate surface. This technique is
especially useful for deposition of compounds or mixtures, where different components
would otherwise need to evaporate at different rates. This phenomenon is widely employed
in the semiconductor industry to deposit thin films of various materials onto silicon wafers.
42
However, the resulting film is often amorphous or polycrystalline, thus post deposition
annealing is required to promote the crystallinity and orientation. Post annealing process
will cause a lot of problems such as the softening of substrate, the increasing of thermal
mismatch, and the interdiffusion between the substrate and film. In addition, sputtering does
not offer precise control over the species fluxes necessary stoichiometric film growth due to
different sputtering rates for different atoms. Therefore, for complex systems such as BaM,
it is not as straight forward to grow a stoichiometric, single crystal thin film successfully by
rf sputtering, while for an element film or an alloy of similar atoms, sputtering can be a
relatively low-cost, low maintenance, effective process.
Similar to RF sputtering, pulsed laser deposition (PLD) uses a pulsed laser beam to
impinge on a target in order to deposit desired material as thin films. A high vacuum
chamber (base pressure ~10-7 Torr) is commonly necessary for the processing. Pulses of
focused laser light transform the target material directly from solid to plasma; the resulting
plume of plasma is thrown perpendicularly away from the surface by thermal expansion. As
expansion cools the plume, it will revert to a gas, but sufficiently high vacuum (10-3~10-2
Torr) will allow momentum to carry this gas to the substrate, where it condenses to a solid
state and the growth happens. Similar to sputtering, the grown film is usually amorphous or
polycrystalline, and thus a post deposition anneal is needed for improved crystallinity. The
crystal structure and orientation of films grown by PLD are highly dependent on
thermodynamic stability. In PLD, the precise control of the processing parameters (such as
species fluxes) is also difficult to realize, which is helpful for understanding the nucleation
and growth mechanisms of thin films.
43
The sol-gel process involves the transition of a system from a liquid solution (the
colloidal “sol") into a solid (the "gel") phase. The solution is made of solid particles with a
diameter of few hundred of nm, usually inorganic metal salts, suspended in a liquid phase.
In a typical sol-gel process, the precursor undergoes a series of hydrolysis and
polymerization reactions to form a colloidal suspension. Then the particles condense in a
new phase, the gel, which is a solid macromolecule immersed in a solvent. After the “gel”
film is crystallized, the substrate and film are placed in a furnace to be annealed. The desired
film thickness is obtained by repeating the process. Sol-gel may have the capability of
controlling the nucleation of film; however the crystal structures of films grown by this
technique are controlled primarily by thermodynamics.
Liquid phase epitaxy (LPE) is another common thin film material growth technique
that has capable of producing high quality BaM thick film [74]. In LPE, a crystalline film
precipitates from a supersaturated melt onto a substrate that serves as the template for
epitaxy and the physical support for the heterostructure [68]. For BaM growth, the melt can
be prepared form a mixture of iron oxide (Fe2O3), barium carbonate (BaCO3), and boron
oxide (B2O3) powders [74]. The powders are ground and mixed homogeneously. The ratio
between the powder weights of Fe2O3 and BaCO3 is set based on the chemical formula of
BaFe12O19. After the hexaferrite composition for growing is completely formed as liquid at
its melting temperature, the substrate can be dipped into the melt. By maintaining the
temperature of the melt at a slowly decreasing rate during growth, BaM can be deposited
onto the substrate. LPE technique can produce BaM films of controlled composition and
thickness, but poorer thickness uniformity and rougher surface morphology than MBE
44
technique [68]. With the use of BaM seed layer (~0.5 m) grown by PLD or sputtering, the
thickness of BaM films grown by LPE can be increased greatly [74].
All the techniques that have been used to grow BaM thin film proceed under ambient
pressure or in the mTorr range, which can result in substrate contamination prior to
deposition or film contamination during growth from residual gasses or compounds. In
order to improve crystallinity, a post annealing process may needed, which can result in
thermodynamic dependence of the crystal structure and orientation, as well as such
problems as interdiffusion and increasing lattice mismatch between the film and the
substrate. In addition, these techniques do not offer the level of tunable control over the
processing parameters necessary for understanding the nucleation and growth mechanisms.
2.6.2
MBE Growth Process
In molecular beam epitaxy (MBE), atomic or molecular beams are directly
deposited at a carefully prepared single-crystal substrate in an ultrahigh vacuum (UHV)
system (≤10-9Torr). In MBE, thin films grow by reactions between thermal-energy
molecular or atomic beams of the constituent elements and a substrate surface, which can be
maintained at a different temperature. The composition of the deposited film depends on the
relative arrival rates of the constituent elements, which consequently depend on the
evaporation rates and sticking coefficient of the appropriate materials. The growth rate is
typically 1 monolayer/s (2~3Å/s) [75], which is low enough to ensure surface movement of
the impinging species on the growing surface and can result in very smooth and uniform
film. In our system, the rate for growing MgO on 6H-SiC ranges from 0.05 to 0.5 Å/s, which
is highly dependent on the Mg and O fluxes.
45
Compared with other epitaxial growth techniques, MBE has distinctive advantages.
MBE allows for more precise control, literally atomic level control of the atomic or
molecule fluxes and the surface state of the substrate. Under UHV (10-9Torr), there are
large mean free path, long electron life, and very low contamination rates
(50mins/monolayer contamination). Also because of UHV, MBE growth can proceed under
conditions far from the thermodynamic equilibrium and be governed mainly by the kinetics
of the surface processes occurring when the impinging beams react with the outermost
atomic layers of the substrate crystal [75]. However, other epitaxial growth techniques, such
as liquid phase epitaxy or vapor phase epitaxy, which proceed at conditions near
thermodynamic equilibrium, are commonly controlled by diffusion processes occurring in
the crystallizing phase surrounding the substrate crystal.
Because of UHV environment, the film deposition can be controlled in-situ by
surface analytical methods such as Reflection High-Energy Electron Diffraction (RHEED),
Auger Electron Spectroscopy (AES) and X-ray Photoelectron Spectroscopy (XPS). These
powerful techniques for control and analysis eliminate much of the guesswork in film
growth, help to understand the nucleation and growth mechanisms of thin films, and
consequently enable the fabrication of sophisticated device structures using this growth
technique. Different types of MBE involve gas source MBE, solid source MBE, and metal
organic MBE, in which solid source MBE is the most common type. A schematic drawing
of a generic MBE system is presented in Figure 16.
46
O2 plasma
Figure 16: Schematic drawing of a generic MBE system. The MBE growth chamber
consists of effusion cells sources, and an oxygen plasma source. This chamber also
includes a 15keV RHEED system for structural monitoring during film growth.
An MBE growth chamber is usually equipped with a substrate heater, effusion cells,
an oxygen plasma source, and a RHEED system. The growth chamber is evacuated by a
turbo molecular pump capable of maintaining a pressure in 10-9 Torr. Effusion cells are the
key components of an MBE system, because they must provide excellent flux stability and
uniformity, and material purity. Furthermore, being the parts that must withstand the highest
temperatures (up to 1300ºC) for the long periods of time, those cells are placed on a source
flange, and are co-focused on the substrate heater, to optimize flux uniformity. Typically a
specific flux is between 1013~1016 atoms/cm2sec. A detailed description of the MBE system
used in this study will be introduced in Chapter 4.
47
3.0 CRITICAL LITERATURE REVIEW
Integration of magnetic ferrite film (e.g., BaM and Fe3O4) with wide bandgap
semiconductor (e.g., SiC) can not only meet the increasing requirements of minimizing the
device size and improving the power efficiency, but also can open a new performance in
next-generation microwave and spintronic devices with dynamic tunability and diverse
functionality. The properties of microwave and spintronic devices connected to magnetic
anisotropy field (HA), coercivity (Hc), saturation magnetization (4Ms), and FMR linewidth
(∆H) are highly dependent on the crystal structure, chemistry and orientation of the grown
films. The magnetoelectric coupling between the ferrite/ferroelectric heterostructure is
sensitive to the quality of the interface. Both theory and experiment suggest the structure,
chemistry, and orientation of the film are determined by the interface properties and
processing parameters. Therefore, understanding the influences of the film processing
conditions and the surface/interface on the propeties of BaM and Fe3O4 films on wide
bandgap semiconductor substrates is important for engineering an effective process to
achieve next-generation multifunctional devices.
Current BaM films for microwave device application are mostly grown by PLD,
LPE, sol-gel, and rf sputtering on non-semiconducting substrate (e.g., MgO and Al2O3),
which are not suitable for the planar devices that compatible with CMOS-based integrated
circuits. A key step to realize the CMOS-based integrated circuits is integration single
crystal BaM films on semiconductor substrates. However to date, there is no BaM thin-film
based integrated microwave device obtaining any success in either commercial or military
48
microwave market because BaM growth involves high temperature processing and oxygen
environment that are not compatible with Si or GaAs based CMOS processing [11]. In order
to address this issue, it is necessary to develop a new material system that is integrated wih
wide bandgap semiconductors, such as SiC and GaN.
It is hypothesized that understanding the nucleation and growth mechanisms of
single crystal BaM films is the first step to successfully engineer the BaM films with wide
bandgap semiconductor for next-generation microwave devices. More specifically, the
influences of the substrate, template, temperature, pressure, and source flux on the film
stoichiometry, structure and orientation need to be understood. In addition, it is important to
understand how crystal structure, orientation, and stoichiometry affect magnetic properties
of the BaM film in order to engineer the thin films suitable for next-generation microwave
devices. MBE processing allows the atomic level controlled environment necessary to study
these influences. Utilizing the benefits of UHV and MBE to understand the nucleation and
growth mechanisms, it will be possible to integrate high quality magnetic ferrite (e.g., BaM
and Fe3O4) on 6H-SiC with precise stoichiometry, single crystallinity, and preferred
orientation necessary for next-generation microwave and spintronic applications.
3.1
Barium Ferrite Thin Film Growth
BaM films have been grown on a variety of substrates by such methods as PLD, LPE,
sol-gel, and rf sputtering. Much progress has been made in increasing crystallinity and
improving morphology, while chemical analysis to determine the impact of processing
conditions on stoichiometry has been limited. Much morphological analysis to determine
49
the impact of processing conditions on structure has been performed. The effects of
substrate surface, temperature, O2 pressure, and film stoichiometry on the morphology,
orientation, and magnetic property of the BaM film are still the major issues that need to be
analyzed and understood.
3.1.1
The Influences of Substrates on BaM Film Growth
Barium ferrite thin films have been grown on many substrates, such as Al2O3 [76],
MgO [77], Si/SiO2 [78], and some templates such as Pt [79], Si3N4 [80], ZnO [81]. All these
studies have shown that the orientation of BaM depends highly on the crystal structure and
orientation of the substrates or the templates. Thus, exploring the effect of lattice mismatch
strains on the magnetic and structural properties of these films is very important.
Among the literature on BaM films, single-crystal (001)-Al2O3 substrates were used
most often to obtain good quality films. Because of the different crystal structures of
sapphire and BaM, the lattice mismatch cannot be ascertained directly. However, by
comparing the areas of the sapphire and BaM oxygen planes, one obtains a 7% lattice
mismatch at room temperature, with sapphire being smaller [82]. As a result, an epitaxial
film is under stress, at least near the film– substrate interface.
In addition, high growth temperatures (~900C) [74] or post-annealing temperatures
(800~1000C) [51,83] are required for many techniques (e.g., PLD, rf Sputtering, and LPE)
in order to get crystalline BaM. The differences in the thermal expansion coefficients of
BaM and the substrates ultimately become very important for the BaM film qualities at such
high temperatures. For example, the thermal expansion coefficients of barium ferrite and
sapphire are 1.0×10-5 /C [84] and 7.8×10-6 /C [85] respectively. Such differences can
50
distort crystal lattice structure, stress the interfaces, and consequently cause the formation of
defects that affect the FMR linewidth. It is believed that choosing suitable substrates and
lowering growth temperatures or annealing temperatures can help reduce the stresses
caused by the lattice mismatches and thermal expansion differences.
Shinde and Ramesh [76] studied the effect of substrate-induced lattice strains on the
structural and magnetic properties of epitaxial BaM thin films. From measuring the lattice
strain △d/dbulk (△d =d
bulk-d observed,
d is the lattice spacing), they found the lattice strain
was a function of thickness. As shown in Figure 17, the lattice strain became increasingly
more important below a thickness of about 1000 Å and the c-axis dispersion [FWHM of the
rocking curve for the (0008) peak] became serious for ~100Å thin films.
Figure 17: c-axis dispersion and lattice strain (shown as change in lattice spacing
△d/dbulk) in the film as a function of film thickness. The inset shows how c-axis
dispersion in the film changes with the lattice strain [76].
Shinde and Ramesh explained that the deformation was caused by the nature of
matching BaM lattice to the sapphire lattice. Sapphire has a rhombohedral crystal structure
with a = 5.128Å and an angle  of 5522’, whereas barium ferrite has a hexagonal structure
with a = 5.88Å and c = 23.2Å. Therefore, the barium ferrite structure has to be sheared to
match these lattices (Figure 18). The shear causes the change of the lattice surface area on
51
the a-b plane. Therefore, both shear and compressional strains appear in these films. Since
these strains take their largest values at the interface of BaM films and sapphire substrate,
little relief is possible when films are very thin (thinner than 0.1m). Figure 18 showed the
schematic illustration of such mismatches of the hexagonal and rhombohedral structures.
Figure 18: Illustration of the matching between rhombohedral sapphire and
hexagonal barium ferrite; (a) before deformation; (b) one way of deforming the
hexagonal structure to fit on the rhombohedral structure, resulting an area change of
-25.7Å2 (c) another way of fitting the hexagon on the rhombohedral structures,
showing a smaller area (-3.58 Å2 ) change of the hexagon [76].
The large stresses induced by the differences between the lattice parameters and
coefficients of thermal expansion of the film and underlying substrate, may cause film
cracking and delamination. Therefore in practice, the maximum possible thickness of BaM
(001) films grown on Al2O3 (001) substrates is less than 20m [86]. This is too thin for
practical application in low-loss microwave devices operating at frequencies below 40 GHz,
where a minimum thickness of ~100m is required [86].
In order to develop the thick oriented hexaferrite films required for microwave
devices, Oliver et al. [87] chose MgO (111) as the substrate to deposit BaM films by PLD.
Since the ionic MgO crystalline lattice has a rock-salt structure, the MgO (111) crystal plane
is close packed and retains the continuity of oxygen planes from the (001) BaM film.
Consequently from straightforward geometrical considerations it is expected that the BaM
[ 1120 ] crystal axis will lie collinear to the [ 110 ] axis in the MgO [111] crystal plane. The
lattice parameter of MgO is =4.21Å, and the equivalent lattice spacing in the MgO pseudo
52
hexagonal structure is 5.96Å. Thus a lattice mismatch between atoms on the MgO [111]
plane and those of BaM [001] is expected to be f= (BaM -MgO)/BaM= -0.013, which is
improved compared with that of BaM grown on Al2O3 (f=0.07).
Another important fact to be realized is that, the thermal expansion coefficient of
MgO (1.36×10-5/C) from 0 to 1000 °C [85] is greater than that of BaM (1.0×10-5/C), while
that of Al2O3 (7.8×10-6/C ) is smaller. As a result there will be some biaxial compressive
stress in the BaM film on MgO in contrast with tensile stress on Al2O3 upon cooling.
Consequently the eventual film thickness will be governed by the fracture strength of the
MgO substrate instead of the BaM film. Because of the improvement of lattice match, the
BaM films grown on MgO (111) did not crack or delaminate to a thickness of at least 28m.
Furthermore, the XRD pattern and hysteresis loop showed excellent c-axis orientation
normal to the film plane and comparable magnetic properties to bulk BaM values [87].
From the discussions above, the quality and thickness of the BaM films can be
influenced by the crystal structure (lattice constant) and physical properties (e.g. thermal
expansion coefficient) if the surface chemistry effect is excluded. Properly choosing a
substrate can minimize the difference of lattice parameters and thermal expansion
coefficients, thus avoid film cracking and delamination.
3.1.2
The Influences of Various Templates for BaM Film Growth
Comparing with the film processing parameters that can be varied systematically,
such as substrate temperature, pressure, and species flux, the quantity of substrates that have
physical parameters matching to the BaM is highly limited due to crystallographic and
economic considerations. Especially when a semiconductor substrate is requied in the
53
compact and integrated devices, the growth of high quality BaM is difficult because of the
significant interdiffusion of the heterogeneous atoms at high temperature and the large
lattice mismatch between the semiconductor substrates and the films.
According to the report of Hylton et al. [88], the interdiffused layer between BaM
film and the SiO2 substrate was nonmagnetic and strongly affected the magnetic properties
(Ms). In addition, Kakizaki and Hirastuka [89] found that it was problematic to directly
deposit BaM on Si wafers with a thermally oxidized surface layer (SiO2/Si) or a fused quartz
slide and achieve an out of plane uniaxial magnetic anisotropy. Thus a relatively thin
(5-100nm) template grown between the film and substrate was needed to block interface
diffusion.
So far there are many different kinds of templates that have been investigated, such
as Pt [79], amorphous Al2O3 and TiO2 [80], and low-Ba content BaM layer [90]. General
criteria for selection of template materials are low cost, chemical stability, mechanical
hardness, the ability to withstand high temperature annealing, and limited interdiffusion
while providing good adhesion with the BaM film [88].
Morisako et al. [91] has shown that the microstructure and magnetic properties of
BaM thin films can be improved by using an amorphous Ba-Fe-O template. In their study,
thin films were prepared on SiO2/Si by a sputtering system. A thin amorphous Ba-Fe-O film
(~20nm) was deposited as a template for the c-axis oriented hexagonal BaM film growth.
Morisako et al. found that this kind of templates can lower the substrate temperature for
BaM growth. Comparing BaM films grown with and without templates (Figure 19), it was
seen that the crystallinity of the BaM films without template was not improved by raising
54
substrate temperature (Ts), the intensity of diffraction lines was very weak even at Ts as high
as 600ºC; however, for films grown with template of Ba-Fe-O, a preferential c-axis
orientation was seen at Ts as low as 500ºC, and the c-axis orientation was improved with the
increase of Ts.
Figure 19: X-ray diffraction diagrams of the BaM film: (a) without template, (b) with
template [91].
Moreover, the amorphous template prevented the deep interdiffusion between film
and substrate. During a postannealing process, Si can diffuse into the thin film and Ba, Fe
can diffuse into the substrate, and the Fe atoms diffused deeper into the substrate than Ba
atoms. Consequently, the Ba/Fe ratio of BaM film was changed and a non-magnetic
template was formed, which caused the relatively lower Ms compared with bulk materials.
Their results were confirmed by XPS depth profile of BaM thin film grown with and
without template, as shown below in Figure 20. However, the authors were not clear
whether this diffused layer would affect the crystallographic characteristics of the BaM
layer that was deposited on SiO2/Si, or whether such a thin template was sufficient to
55
prevent deep diffusion. Since the amorphous Ba-Fe-O films were used as a template, it is
hard to relate the lattice mismatch influence to the quality of BaM films.
Figure 20: XPS depth profiles of the BaM films: (a) without template, the depth of
diffused layer is about 700 Å; (b) with template, the depth of diffused layer is about
300 Å [91].
In addition, suitable templates can promote c-axis perpendicular orientation of BaM
films as reported by Zhuang et al. [79]. In their study, the use of Pt templates greatly
improved the magnetic properties and crystallographic characteristics of BaM films because
the nucleation sites for perpendicularly oriented grains were relatively increased with Pt
templates compared with no templates. BaM films and Pt templates were deposited by rf
diode sputtering onto a thermally oxidized silicon (SiO2/Si) substrate. Without Pt templates,
the perpendicular c-axis orientation was getting worse with an increase of BaM film
thickness; while with two Pt templates (500 Å thick), a perpendicular remanent squareness
of 0.95 can still be achieved for a 900-Å-thick BaM film.
The enhancement in perpendicular c-axis orientation for films with Pt templates was
related to the suppression of the growth of in-plane and/or randomly oriented acicular grains.
There were two kinds of nucleation sites for crystallization of BaM during ex-situ annealing.
One kind of nucleation sites were at the interfaces formed between Pt and BaM interfaces,
56
which believed to favor perpendicularly oriented grains [92]. In addition, there were also
some randomly oriented nucleation sites in the bulk of BaM films. The total number of
random nucleation sites increased as the film thickness increased, thus resulted in the
deterioration of perpendicular orientation with an increase in film thickness. Pt templates in
BaM films had the effect of increasing the number of perpendicular nucleation sites at the
interfaces, thus effectively improved the perpendicular c-axis orientation. However, the
authors did not report how to measure the number of nucleation sites; therefore it is hard to
prove whether the explanation is reasonable.
Other various amorphous templates such as Al2O3, Cr2O3, SiO2, CuO, and TiO2
were employed by Wee et al. [80] on a Si substrate to optimize the characteristics of BaM
thin films for magnetic recording applications. They reported that sputtered Al2O3
amorphous templates may be a suitable candidate for growing BaM with high coercivity
(5kOe), high squareness (Mr/Ms about 0.92), and small grain size as needed for magnetic
recording applications. Another interesting result should be noticed that a c-axis
perpendicularly oriented texture was not observed in films grown on amorphous Al2O3,
whereas only c-axis in-plane orientation texture was developed. In addition, BaM film
grown on a TiO2 template had no crystallinity, which resulted in the lowest coercivity
(353Oe). According to the nucleation theory of Zhuang et al., amorphous Al2O3 may result
in the relative increase of the nucleation sites for in-plane oriented grains, while amorphous
TiO2 may not be able to provide any nucleation sites for both in-plane and out of plane
oriented grains .
Kim et al. [93] showed some different results when they used Fe, Cr, Al2O3,
57
ZnFe2O4, and TiO2 as templates for BaM on (100) oriented bare Si substrate by rf/dc
magnetron sputtering. The coercivity values of BaM films with various templates are
summarized in Table 3. All films experienced nearly the same coercivity in both the
in-plane and the out-of-plane directions except for BaM/Fe/Si film, which meant all the
films did not possess crystalline magnetic anisotropy.
Table 3: Coercivity values of BaM thin films with various templates [93]
Coercivity(kOe)
Template
Longitudinal(Hc//)
Perpendicular(Hc┴)
Fe
0.2
1.1
Cr
4.4
4.6
Fe2O3
3.3
3.1
Al2O3
3.6
3.6
ZnFe2O4
2.3
1.9
TiO2
4.7
4.4
In contrast with Wee’s results [80], Kim et al. reported that BaM/TiO2 /Si films
exhibited the highest coercivity of all films with templates, while the lowest coercivity was
achieved in the BaM/Fe/Si film. Wee and Kim used different deposition conditions for BaM
film and TiO2 template sputtering (Table 4), such as Ar/O2 pressure and anneal temperature.
It is hypothesized that the function of templates can be influenced by those deposition
conditions, and thus contributes to different BaM grain orientation and crystallinity, which
ultimately are responsible for the different magnetic properties of BaM/TiO2/Si by two
groups. Kim et al. also reported that the microstructure of BaM in BaM/TiO2/Si was
strongly dependent on the total Ar/O2 gas pressure and microstructure of the TiO2 template,
which proved the hypothesis to some extent.
58
Table 4: Comparison of deposition conditions for BaM/TiO2/Si by Wee and Kim
et al.
Deposition conditions
Ar/O2 pressure for BaM sputtering
Wee et al.
0.3mTorr Ar/1.7mTorr O2
Kim et al.
Ar+10% O2
5 or 10mTorr for total
Ar+30% O2
Ar/O2 pressure for TiO2 sputtering
60mTorr Ar/none O2
Substrate Temperature (Ts)
Room temperature
Room temperature
Anneal procedure
600ºC for 5h
750ºC or 850ºC
Followed by 800ºC for 3h
For 10 min
5 or 10mTorr for total
It was reported by Kim et al. that by using ZnFe2O4 as a template, the interdiffusion
of Si from substrate was prohibited to some degree, which was proved by extremely weak Si
peak from XRD pattern. However, concerning several microns penetration depth of XRD
measurement, it is really hard to get the point that whether Si peak came from the
interdiffusion layer. It is possible that the Si peak still came from the substrate because the
total thickness of BaM film and ZnFe2O4 template was only a tenth of a micron, which is
thin enough for x-ray to get through.
Regarding the effect of template thickness on coercivity, the Fe2O3 template was
thinner than all other templates, yet exhibited almost the same coercivity as those films with
other templates. This suggests that about 3.3 kOe of coercivity was achievable with the 50
nm thick Fe2O3 template compared to other materials that needed 100 nm thickness as an
template to produce 3.3kOe coercivity. According to Doh et al. [94], the secondary Fe2O3
phase can act as a useful inhibitor to abnormal acicular grain growth in BaM films by
pinning the grain boundary movement of the BaM phase in a later crystallization state.
Because of the grain size refinement of the magnetic BaM phase, the intrinsic coercivity of
BaM film was enhanced from 1.9kOe to 4.4kOe for high density recording media.
Concerning the effect of some templates that may introduce new diffused layers,
59
which contain some undesirable magnetic grains, Chen et al. [90] used a barium ferrite layer
with low-Ba content as a template, and then deposited high-Ba content barium ferrite thin
film by sputtering. They reported that, the nucleation rate was dependent on the Ba content
and high-Ba content resulted in high nucleation rate. Because of lower nucleation rate, the
low-Ba content barium ferrite template allowed a diffused layer at its interface, prevented
the grains from nucleating in the interdiffused layer. The suppression of nucleation in the
interdiffused layer resulted in higher coercivity squareness and thus, the magnetic properties
were dramatically improved.
In addition, Chen et al. concluded that the top high Ba-content layer should be thick
enough to provide a significantly higher nucleation rate relative to the template. Since with
the increase of the top layer thickness, the number of nucleation sites also increased and thus
resulted in reduction of grain size and improvement of coercivity value. More importantly,
the template did not introduce a new diffused layer at its interface with the top layer.
From all the discussions above, it is proposed that the orientation and magnetic
properties of the BaM films can be influenced by use of different templates. Proper
templates can effectively prevent the interdiffusion between the substrate and the film,
promote c-axis orientation, and improve the crystallinity at relative lower temperature.
However, it is still not clear what the optimal thickness of a template needs to prevent deep
diffusion and how processing parameters influence the function of a template. In addition,
many templates employed previously were amorphous, and may not be capable of
dispersing thermal and lattice stain as well as forming an effective, abrupt interface. It is
possible that the amorphous templates used by many authors [80,91] may be a result of the
60
limitation of the sputtering growth method used. Through atomic level control of the
surfaces and reactants in MBE, it will be possible to determine the cause of interface
characteristics and then engineer an effective, abrupt interface for c-axis oriented BaM film
growth. It is suggested that in MBE growth, the templates of single crystalline oxide with
lattice constant matches with BaM (e.g., MgO) may be employed since MgO can act to
disperse thermal and lattice stain between the BaM and the SiC. In addition, a high quality
MgO thin film can also be used as a high-k dielectrics on semiconductor such as Si [95] and
tunnel barriers for magnetic tunnel junctions with giant tunneling magnetoresistence
[96,97].
3.1.3
The Influences of Temperature on BaM Film Growth
Crystalline BaM films are usually obtained by two kinds of fabrication processes.
One is the in-situ process, in which crystallized BaM thin films directly grow on the
substrate heated above 500ºC. The other is the ex-situ process; this is a two step process
involving deposition of an amorphous film at room temperature and then crystallization of
the film by postannealing. Both methods can produce BaM thin films with high degree of
epitaxy, good stoichiometry, and desired magnetic properties [51,78, 98 ]. The ex-situ
process has been mostly applied in fabrication of BaM as recording media material, and is
considered a good process for the mass production of this kind of recording media.
Liu et al. [78] discussed the crystal growth, magnetic properties and surface
morphology of the films prepared by the ex-situ and in-situ processes at different substrate
temperatures (Ts) and postannealing temperatures (Ta). In their study, about 180 nm thick
thin films were prepared by facing target sputtering (FTS) system on thermally oxidized
61
silicon wafers (SiO2/Si). The O2 partial pressure was fixed at 0.2 mTorr and the Ar/O2 total
pressure was 2 mTorr during the sputtering process. For the ex-situ process, post-annealing
was carried out in a conventional electronic furnace in air. Thin films were annealed for 2 h
at various temperatures.
From XRD results (Figure 21a) of films prepared by the ex-situ process at various
annealing temperatures, the authors found those films have preferential c-axis orientation,
but many imperfections or crystal defects from lower (00l) peak intensity. The
crystallographic properties could not be improved with the increase of the annealing
temperature. However as seen in Figure 21b, with the increase of substrate temperature, the
crystallinity for BaM phase was improved and fewer imperfections of the films existed.
Films deposited at temperatures higher than 500 ºC showed preferential c-axis orientation,
and the crystal defects were decreased dramatically at a Ts of 550 ºC. The half maximum
width of the (008) peak also decreased with the increase of substrate temperature, which
meant the grain size increased with the increase of substrate temperature.
62
Figure 21: The XRD results of BaM films prepared by a) the ex-situ process at various
annealing temperatures; b) the in-situ process at various substrate temperatures [78].
From Figure 22a, it was observed that the BaM phase started to crystallize when the
annealing temperature was elevated up to 600ºC. The spontaneous magnetization, Ms,
hardly increased above the annealing temperature of 650ºC; however Ms increased with
increasing substrate temperature (Figure 22b). Films deposited at 550ºC had Ms Value about
290 emu/cm3, which was relatively lower than that of bulk BaM materials (375 emu/cm3).
Figure 22: a) The dependence of the spontaneous magnetization of films prepared by
the ex-situ process on annealing temperatures; b) the dependence of Ms of films
prepared by the in-situ process on substrate temperatures [78].
According to Hylton et al. [88], this may be caused by the reaction between the film
63
and the substrate. During the high temperature annealing process, Si can diffuse into the thin
film and Ba, Fe can diffuse into the substrate, and the Fe atoms diffused deeper into the
substrate than Ba atoms. Consequently, the Ba/Fe ratio of BaM film was changed and a
non-magnetic template was formed, which caused the relatively lower Ms compared with
bulk materials. The existence of interdiffusion layer was confirmed by their XPS depth
profile. Another reason for the lower Ms in BaM films was due to the inadequate
crystallinity in the films in agreement with the results of XRD, where relatively higher peak
intensity have higher Ms and lower peak intensity have lower Ms.
As for temperature influences on the Hc (Figure 23), Liu reported Hc increased with
increasing of Ta, and with decreasing of Ts. In addition, films prepared by the ex-situ process
had relatively larger grain size (200 nm) and the mean grain size kept similar at the Ta range
of 700 ºC ~900 ºC, whereas films prepared by in-situ process had relatively small grain size
of about 20 to 30 nm (Figure 24). High ex-situ annealing temperature at 850 ºC may
promote the coalescence of the grains and result in bigger grain size, compared with in-situ
deposition at 550 ºC. There was an interesting result that Hc was dependent on the grain size
for films prepared by in-situ process, but not correlated to grain size for films prepared by
ex-situ process. This may be attributed to different mechanisms of magnetization reversal
for films prepared by two processes. For films prepared by the ex-situ process, the low
crystallinity may strongly influence the magnetization reversal process; whereas for films
prepared by the in-situ process, the magnetization reversal process is determined by the
grain size and the magnetostatic coupling between the grains.
64
Figure 23: The dependence of coercivity, Hc, a) on annealing temperature, Ta, for
films prepared by the ex-situ process; b) on substrate temperature, Ts, for films
prepared by in-situ process [78].
Figure 24: SEM image of BaM films prepared by, a) the in-situ process at Ts of 550 ºC;
b) the ex-situ process at Ta of 850 ºC for 2 h [78].
In PLD growth, Shinde and Ramesh [51] also found that the substrate temperature
played an important role in the crystallinity and morphology of barium ferrite film. In the
study, the films were grown on single-crystalline (001) sapphire using KrF excimer laser
with energy density of 2 J/cm2 under oxygen pressure ranging from 75 to 500 mTorr. The
substrate temperature ranged from 300 to 920C and post annealing temperature of those
films were from 850 to 1035C under flowing oxygen for 2 hours. The films were 0.5 to 0.6
m in thickness.
As reported by Shinde and Ramesh, the as-grown films were amorphous when
deposition (TD) was below 550C, and the films became polycrystalline after annealing at
65
850C for 2 hours. However when TD increased to 575C, the films were found
single-crystalline with c-axis perpendicular to the substrate after annealed at the same
conditions. It was assumed that there were small c-axis oriented “nuclei” in the as-grown
films when TD was raised to 575C, and grew during subsequent heat treatment. When TD
increased gradually, those “nuclei” grew bigger and a granular structure became discernible.
Typically, films grown at TD=700C had granular structure with grain sizes of 40 - 60 nm.
As TD kept increasing to 920C, the films were found to be continuous, but with spiral
growth, which were known to originate from screw dislocations [99]. The BaM film depth
had steps up to 6-8; each step was about 11 to 12Å thick, which was about half of the BaM
lattice constant on c-axis. Figure 25 collected together the results of the structural studies.
Figure 25: Schematic representation of the changes in the film structures with the
substrate temperatures varying from 400 to 920C during deposition (TD). Starting at
300– 400 °C, the film is amorphous; with TD of 550– 600 °C, the grains are too small to
observe by XRD but grow upon annealing; with TD around 700 °C, film is granular;
and only when TD reaches 900 °C the film becomes continuous and exhibits a spiral
growth[51].
In addition, the magnetic hysteresis loops (Figure 26) confirmed the film quality was
improved by increasing TD. For the films deposited at 300°C and 400°C and subsequently
66
annealed at 850°C, the M-H loops were almost the same in both geometries because of the
random orientation of crystallites. The perpendicular direction began to be the preferred
direction (easy axis) of magnetization in the film deposited at 575 °C and annealed at 850°C.
A much higher substrate temperature (920°C) made the film to behave magnetically like a
single crystal and produce the anisotropy field of 16 ±1 kOe as bulk material. Since the
films deposited at higher temperature (near 900°C) were highly oriented with reduced
defects concentration, the FMR linewidth was very narrow, usually narrower than 200Oe,
4M (arb.
which meant lower resonance loss in microwave application.
H (kOe)
H (kOe)
Figure 26: Parallel (dashed line) and perpendicular (solid line) hysteresis loops for the
films with (a) Ts=300 °C, Ta=850 °C, (b) Ts=400 °C, Ta=850 °C, (c) Ts=575 °C,
Ta=850°C, and (d) Ts=920 °C [51].
The thermal anneal was also studied by Ramesh et al. in order to improve the film
quality. They chose films that were deposited at deposition temperatures (TD) of 920C and
annealed at temperature ranging from 920~1035C (Ta) for 2 hours under flowing oxygen of
67
300 mTorr. Typically, post annealing at Ta=1000C caused ∆H to reduce about 100Oe, and
the Atomic Force Microscope (AFM) micrograph (Figure 27) showed a significant reduce
of grain boundaries of the island. Comparing with the as-grown film, the surface roughness
of 1000C post annealing was suppressed because of the growth of bigger islands.
(a)
TD = 920C
(b)
Figure 27: The AFM micrographs of BaM films deposited on (001) sapphire
substrates: (a) deposited at 920C; (b) deposited at 920C and post-annealed at
1000C under flowing oxygen of 300 mTorr for 2 hours [51].
Comparing with the substrate temperature (500°C) used by Liu et al. to start getting
c-axis oriented BaM film, higher starting point of Ts (575°C) needed for PLD to get BaM
“nuclei”. This difference can be attributed to many factors, such as the substrate and
pressure of the ambient gas. A similar result was obtained by both groups that with the
increase of deposition temperature, the coercivity of BaM films decreased and the grain size
became bigger. This may be explained as the growth rate is higher than nucleation rate at
higher deposition temperature. It is regretful that neither groups reported any chemistry
information of the BaM films, so no correlations of film stoichiometry to deposition
temperature can be obtained.
Higher deposition temperatures and post annealing temperatures contribute to
68
highly c-axis oriented films with good stoichiometry, reduced defects concentration, and
narrow FMR linewidth. However, in order to apply these films to MMIC devices, it is
required to lower the deposition temperature as well as the post annealing requirement. One
of the potential advantages of MBE growth is relatively lower growth temperature and no
need of post annealing process. Through the use of MBE, it may be possible to engineer an
effective interface that grow BaM film at lower temperature and control the film’s chemistry
and structure more precisely.
3.1.4
The Influences of Deposition Pressure and O2 Pressure on BaM Film
Other than temperature, the literature shows that the deposition pressure and O2
partial pressure also play an important role on BaM film stoichiometry, morphology, and
grain size [77,100,101]. Yoon et al. [77] investigated the material characteristics and
magnetic properties of BaM films deposited on (111) MgO at different O2 growth pressure
by using PLD. The films were deposited at Ts at 925 ºC for 40 min then followed by anneal
in air flow at 1000ºC for 1 h. The thickness of films was typically ranging from 1~2 m. The
O2 pressure of the deposition in the chamber was varied from 10mTorr to 300mTorr.
Comparing SEM images (Figure 28) of films grown at oxygen background
pressures of 300 mTorr and 20 mTorr, they found films deposited at higher PO2 showed
significant outgrowth from the surface, while films deposited at lower PO2 showed less
outgrowth. This result meant that in plane growth rate for the outgrowth platelets was faster
than the on-plane hexagonal platelet growth rate at higher O2 pressure. The presence of
those outgrowth platelets influenced the crystallographic properties of the film by
disturbing epitaxy, and adding voids and cracks.
69
Figure 28: SEM of the BaM films deposited at (a) 300 mTorr and (b) 20 mTorr [77].
The results were confirmed by the magnetic hysteresis loop, shown in Figure 29.
The films grown at higher O2 pressure had worse magnetic orientation than the films grown
at lower O2 pressure. In addition, the authors reported that FMR linewidth (∆H) decreased
with decreasing PO2, which indicated that films grown at lower PO2 have better microwave
loss properties than films grown at higher PO2. Nothing was reported by the authors
regarding the chemistry of BaM films, therefore it is unknown that whether the films with
outgrowth platelets have different stoichiometry as those with hexagonal platelet, or
whether the films are O rich or not.
Figure 29: Out-of-plane (solid line) and in-plane (dashed line) VSM hysteresis loops
for the BaM films with (a) 300 mTorr, (b) 20 mTorr [77].
Comparing the processing parameters reported by Yoon et al. with that of Ramesh et
70
al. [51], there is some confusion about the optimum Po2. The two groups deposited BaM at
Ts ~ 920 C and Ta ~ 1000 C by PLD, but Ramesh et al. annealed films that grown on (001)
Al2O3 1 hour longer than films grown on MgO (111) by Yoon et al.. In addition, a film
thickness of 0.5 ~ 0.6m was typically obtained. According to Ramesh et al., 300 mTorr
was the optimal value for their BaM films grown because highly c-axis orientation (Figure
27) and the narrowest FMR width were obtained. If the films were cooled to room
temperature in reduced Po2, the magnetic characteristics deteriorated noticeably (∆H and Hc
increase) with unchanged structure properties. Ramesh et al. hypothesized that at lower
oxygen pressure during deposition, the BaM film became oxygen-deficient and some of the
Fe3+ ions transformed into Fe2+ ions, causing vacancies, thereby enhanced the
inhomogeneities and influenced the film’s magnetic properties. It was discussed previously
that the increase of film thickness can result in the deterioration of perpendicular orientation.
Since both groups grew BaM films with different thickness, it is hard to determine which
factor (substrate, post annealing time, or film thickness) mostly contribute to the difference
of the optimum Po2 for growing BaM with high c-axis orientation and narrowest FMR
linewidth. In addition, neither groups reported any chemistry information of the growth
BaM films, which may help understand the different influences of Po2 on film structure and
prove reliability of the hypothesis.
Huang et al. [100] showed some similar and different results of oxygen partial
pressure influences on the orientations of barium ferrite films. In their study, barium ferrite
films were grown on large lattice constant (LLC) garnet substrates using PLD technique.
The films were grown at 800 C with oxygen partial pressure at 50, 180 and 400 mTorr with
71
thickness of 0.23 m. They reported that increasing O2 pressure enhanced both
crystallization and c-axis in-plane orientation, while suppressed the epitaxial growth in the
perpendicular direction. The barium ferrite films started to crystallize normal to the
substrate when the oxygen pressure increased from 50 to 180 mTorr, and lost the c-axis
orientation when the oxygen pressure increased from 180 to 400 mTorr. The optimum Po2 in
their study was around 180 mTorr that produced a crystallized film with predominantly
perpendicular anisotropy. The optimum Po2 reported by Huang et al. was different with the
results of previous groups [51,77], who concluded that films grown around 20 mTorr or 300
mTorr by PLD showed the highest quality BaM film with high c-axis orientation and
narrowest FMR linewidth. However, there are also some agreements obtained in between
those three groups. They all concluded that a high Po2 (> 300 mTorr) adversely affect the
magnetic properties of BaM films, which can be observed by broad ∆H, reduced intensity of
perpendicular XRD peak, and outgrowth plates in SEM. On the other side, certain amount
of oxygen can improve film crystallization to some extent.
The results reported by Lisfi et al. [101], agreed with the result that certain amount
of oxygen can improve film crystallization. In their study, 0.3 m thick BaM films were
grown on (001) sapphire using PLD technique at fixed substrate temperature 770C. The
films were grown under two different conditions of O2 pressure, one was under vacuum
(~10-6 Torr) and the other was under controlled Po2 from 20 to 200 mTorr. The results
showed that with the existence of oxygen, the barium ferrite films exhibited higher c-axis
orientation compared with those grown under vacuum, which was randomly oriented. It is
hypothesized that certain amount of oxygen can increase the perpendicular nucleation sites
72
more than the in-plane nucleation sites. More importantly, Lisfi et al. found that the
presence of oxygen helped reduce the stress region (~40 nm thick) caused by lattice
mismatch and thermalexpansion coefficient difference between the barium ferrite and
sapphire lattices as shown by transmission electron microscope (TEM) cross section (Figure
30).
Figure 30: Cross sectional TEM of BaM film (a) grown in oxygen atmosphere showing
a sharp interface, (b) grown under vacuum showing the quality of interface became
deteriorated [101].
Lisfi et al. interpreted the result from the enhancement of plasma reactivity at the
target by introducing oxygen during laser deposition. In contrast with the low kinetic energy
of atoms in sputtering (~ 3 eV) and evaporation (~ 0.1 eV), the ions of PLD plume could
have much higher kinetic energy(> 200 eV). Such high energy of ions in the PLD plume can
damage the smoothness of the substrate at an earlier stage of growth, thus rough growth
with deteriorated interface may occur and a random distribution in the crystallite orientation
was observed (Figure 30). However, introducing O2 during deposition can enhance the
plasma’s reactivity, and high energetic atomic ions maybe transformed into monoxides with
low energy due to reactive collisions, and then smooth growth with a sharp interface can be
73
achieved.
All the results discussed above were connected to PLD growth. As for BaM films
grown by RF diode sputtering, Zhuang et al. [83] reported similar results that the c-axis
orientation of BaM thin film was very sensitive to the oxygen partial pressure. In their study,
a 500Å thick Pt template was first deposited on to a thermally oxidized silicon substrate,
then 10wt % barium rich nonstoichiometric barium ferrite thin films were deposited. The
total pressure was fixed at 5.7 mTorr with the change of Ar and O2 mixture flow rate ratio.
It was reported that all films deposited with O2 had a random c-axis texture, as
indicated by existence of both weak (00l) and (106) peaks. However, the films grown
without oxygen showed excellent perpendicular c-axis orientation. TEM results confirmed
that films grown with O2 had a mixture of acicular (needle-like, with average size of 600 Å
by 250 Å) and platelet grains (300 Å). The platelet grains were of c-axis perpendicular
orientation and the acicular grains oriented in random directions, mostly in the film plane
[83,102]. While the films grown without O2 were dominated by the platelet grains that were
c-axis out of plane orientation. Their results indicated that oxygen partial pressure may
enhance the growth of the in-plane and randomly oriented acicular grains, but worsen the
perpendicular c-axis orientation for Barium rich nonstoichiometric films. Zhuang et al. [103]
also reported that in order to optimize the perpendicular c-axis orientation, stoichiometric
and nonstoichiometric BaM films (BaFe12O19) required different oxygen partial pressure.
The partial pressure ratio of Ar/O2 around 5.0/0.7 was necessary to grow the stoichiometric
BaM films with perpendicular c-axis orientation, while the nonstoichiometric films didn’t
need O2 to get perpendicular c-axis orientation.
74
Zhuang reported that zero oxygen partial pressure can optimize the perpendicular
c-axis orientation for barium rich film. This result is different compared with results
obtained by PLD growth that a small amount of oxygen can promote perpendicular
orientation. It is hard to explain the difference of O2 function because of the differences in
the sputtering and PLD system. In addition, there are many other factors that may cause this
kind of difference, such as the different kinetic energy of ions in the sputtering and PLD
system, the use of Pt template that can promote the perpendicular orientation, and the use of
Ar in sputtering system. It is hard to say which effect is overriding based on the current
knowledge.
Considering all the discussions above, it is proposed that the orientation, crystal
structure, and chemistry of the BaM films can be affected by the O2 partial pressure. The
existing BaM growth methods control over O2 pressure in the mTorr to ambient pressure
range, where the oxygen fluxes are much higher comparing to the fluxes of 10-2 and 10-3
mTorr in MBE. Any remaining O2 could oxidize the substrate prior to deposition, forming
an interface layer that would affect the orientation and structure of the BaM film. By
utilizing the benefits of UHV, which include substrate surface preservation and low
contamination rates, it is hypothesized that through precise control of the O2 pressure and
other chemical species (e.g., Ba, Fe) by MBE, most importantly, by study the chemistry of
growth BaM films, it will be possible to fully understand the effects of O2 so that a process
for an abrupt interface can be engineered.
For practical microwave device fabrication, a BaM film with thickness of the order
of 100 m is required [104]. Many efforts have been made to achieve thicker films by using
75
liquid phase epitaxy (LPE), such as choosing substrate properly [87,105], using seed layer
produced by PLD [104] or ion beam sputtering [106], and so on. Based on the research of
Wang et al. [104], the seed layer preparation was the most important way to grow thicker
films. It is proposed that by using the benefits of UHV and MBE, an effective, abrupt
interface can be engineered on an atomic level to grow a high-quality BaM film with desired
c-axis orientation, precise stoichiometry, smooth, and uniform film surface. The
high-quality BaM film grown by MBE can not only be used as a seed layer for subsequent
thick BaM film grown by PLD or LPE, but also can function as a magnetic layer in
BaM-ferroelectric heterostructures for electrostatic FMR tuning.
3.2
Iron Oxides Growth by Molecular Beam Epitaxy
The development of MBE has provided a means of preparing thin epitaxial
structures of much higher quality for device application [75]. Using MBE techniques,
surface contamination and defects on films can be reduced by controlled growth under UHV
conditions. However, the MBE growth of oxides involves significant challenges compared
with its conventional uses for semiconductor. For example, the oxides need oxygen
sufficient to compose their crystalline structure, but the quantitative control of oxygen
molecules is so difficult that the surface has always been over oxidized or oxygen deficient.
In addition, the insulating nature of many oxide materials always produced difficulties to
study the oxide surfaces during MBE growth.
In order to utilize MBE successfully, some principal scientific issues of MBE
76
growth should be considered adequately, such as strain accumulation and relief resulting
from lattice mismatch, true “defined” epitaxial growth in layer-by-layer mode to form a
single crystal film, as compared to the island growth mode resulting in the evolution of
“textured” films, interdiffusion between the film and substrate material, and the oxidation of
metal species [107]. An ideal film would grow in a layer-by-layer fashion, forming an
atomically abrupt interface, and the film surface would maintain the bulk stoichiometry
upon termination of the growth. However, in practice, a number of mitigating factors (e.g.,
lattice mismatch and interdiffusion at interface) can prevent the formation of an ideal film.
Furthermore, such MBE processing parameters as substrate temperature, oxygen pressure
and species (O vs. O2), can influence the properties of oxide film greatly.
In order to determine the nucleation and growth mechanism of BaM (BaO6Fe2O3)
by MBE, it is useful to understand the influences of different MBE processing parameters
on the nucleation and growth mechanism of iron oxide, since the iron oxidation states (e.g.,
Fe2+ or Fe3+) determine the magnetic properties of the BaM film. In the way of separating
BaM structure, the complexity of developing a tunable BaM growth process via MBE
decreases. By fully understanding the influences of processing parameters on MBE growth
of iron oxide, it will help to grow BaM film with controlled properties by MBE.
3.2.1
Tunability of Iron Oxides Growth via MBE Differentiate Iron Oxides by XPS
Iron oxides films such as Fe2O3, Fe2O3 and Fe3O4 have attracted much interest,
because of technological applications in heterogeneous catalysis, magnetic recording and
integrated microwave devices [108]. Hematite (-Fe2O3, corundum-like structure) and
magnetite (Fe3O4, inverse spinel) are two most common and stable phases. There is also a
77
metastable phase of Fe2O3 (-Fe2O3, maghemite), which is nearly isostructural with
magnetite and easily transforms into Fe2O3 when heated above 325ºC. Finally, there is a
rocksalt phase, wustite (Fe1-xO), which is not stable in the bulk at temperatures below 580ºC.
Magnetite is a conductor at room temperature, and thus is amenable to the full suite of
charge-particle surface science probes, including STM. However, FeO and both phases of
Fe2O3 are insulators.
Many groups have worked on the growth of thin, crystalline iron oxide films by
MBE [109,110,108]. The effects of substrate surface, temperature, O2 pressure, and source
flux on the stoichiometry, crystal structure and orientation are being studied. As with most
oxides, the stoichiometry and structure difference produce many different magnetic and
electrical properties.
In order grow BaM (BaO.6Fe2O3), pure Fe3+ oxidation state should be involved in
the films. Thus an analysis technique that has strong ability to determine film chemistry
should be employed. XPS is such a kind of surface analysis that used by many groups
[108,109,110] to qualitatively and quantitatively investigate the oxidation state of iron and
other elements in single crystal surfaces and epitaxial films.
The Fe 2p core level spectrum is mostly used to differentiate the oxidation state of
iron. The Fe 2p line shape in iron oxides is rather complex and inherently broad, and the
chemical shift between Fe3+ and Fe2+ peaks is too small to be resolved. However, there are
some features in these spectra which are useful to diagnose the Fe oxidation state  the
width at half maximum (FWHM) of the Fe 2p peak and the satellite feature associated with
Fe3+ photoemission.
78
As shown in Figure 31 and Table 5 [109], a difference in the Binding Energy (BE) of
3.5 eV is observed when comparing the satellite of FeO (BE=715.5 eV, purely Fe2+) and
Fe2O3 (BE=719 eV, purely Fe3+). As for Fe3O4, the presence of both satellites results in a
smeared, unresolved structure between the spin-orbit components [101,106]. This result can
be supplemented by the measurement of the FWHM and BE of the Fe 2p3/2 main peak. The
2p3/2 peak (FWHM = 3.5 eV) at 711 eV of Fe2O3 corresponds to the presence of only Fe3+
cations. The 2p3/2 peak (FWHM = 3.8 eV) at 710 eV of FeO corresponds to the presence of
only Fe2+ cations. In contrast, the broader 2p3/2 peak (FWHM = 4.1 eV) at 710.5 eV of Fe3O4
corresponds to the presence of both Fe3+ and Fe2+ cations.
Figure 31: Fe 2p XPS spectra of the 80 Å and 4 Å thick iron oxide films, grown by
MBE on Al2O3 (0001) at 250 °C. The three spectra of Fe2O3, Fe3O4, and FeO are
included as references. The Fe 2p3/2 main peak and satellite corresponding to Fe3+
and Fe2+ cations are illustrated by the arrows respectively in Fe2O3 and FeO
spectra [109].
79
Table 5: Curve fitting parameters for XPS Fe 2p3/2 spectra [109]
Only after the Fe oxidation states are identified clearly by XPS, the chemistry of iron
oxide film can be studied, and thus the control growth of specific iron oxide can be realized.
For example, Figure 31 also showed the Fe 2p XPS spectra corresponding to 80 and 4 Å
thickness iron oxide films grown on alumina at 250 °C by Gota et al., where Fe3+ satellite
was clearly observed, also the FWHM and BE of the Fe 2p3/2 (Table 5) confirmed the
existence of Fe3+ cations in those films. The peak shifting due to surface charging is usually
corrected by fixing the O 1s BE at 530.1 eV, because O 1s BE is independent of the
stoichiometry of the iron oxide [111,112],
3.2.2
The Choice of Oxidizing Gas for Iron Oxides Growth O vs. O2
According to Chambers [107], different oxide films can be obtained at different
metal oxidation rates and film growth rates under MBE, which may be influenced by
surface free energy, lattice mismatch, substrate temperature, and species fluxes. At high
metal oxidation rates, there will be less “intermediates” (e.g., suboxides and metal clusters)
involved in film formation. According to Chambers et al. [107], Fe may be fully oxidized to
be -Fe2O3 or -Fe2O3 only when Fe deposition rate is slower than the kinetics of the oxygen
dissociation reaction O2→2O and the oxidation reaction 2Fe+3O→Fe2O3. Suboxides Fe3O4
80
or FeO may form if the deposition rate is slightly faster than the rate of O2 dissociation.
Moreover, if the atomic oxygen production rate is extremely lower than the Fe delivery rate,
Fe metal will nucleate and may form clusters if the kinetics permits.
It had been shown by many groups [113,114] that FeO and Fe3O4 can be readily
grown in a high-vacuum (≤10-6Torr) environment when molecular O2 is used as the oxidant.
Fe2O3 and Fe2O3, however, can only be prepared with the use of much higher O2
partial pressures (≥10-3 Torr) under MBE [113,115]. That is because the rate of oxygen
dissociation reaction O2→2O is very slow, then the atomic O is not adequately active to
fully oxidize Fe atoms at the surface even at very low growth rates [107].
Atomic oxygen produced from pure O2 [108,109] or NO2 [116,117] have been used
to grow iron oxide successfully via MBE. The use of activated O from either of the two
sources has made greater control over the oxidation and growth process. The use of
activated O from either of the two sources has allowed greater control over the oxidation
and growth process when compared to molecular O2 growth [107], because of greater
oxidation ability (reactivity) of atomic oxygen. Consequently, the literature suggests that the
dissociation of oxygen turns out to be a major complicating factor in determining the
formation of Fe2O3 and Fe2O3.
3.2.3
The Choice of Substrate for Iron Oxide Growth
Epitaxy essentially refers to an extended single crystal film formation on top of a
crystalline substrate. Consequently, the crystallographic properties of the substrate,
especially the crystal symmetry and in-plane lattice constant are major influences on epitaxy
film quality [107]. A variety of substrates have been used to grow thin, crystalline iron oxide
81
films [108,109,110], in which Pt (111), MgO, and Al2O3 were mostly commonly used
substrates.
Al2O3 and Fe2O3 are the corundum structure; MgO, FeO, and Pt are the FCC
(face-centered) cubic structure; Fe3O4 and Fe2O3 are cubic inverse spinel. The three
structures can all be pictured as closed packed hexagonal (111) planes (cubic phases) and
(0001) planes (corundum phases) with different arrangements of cations in between [118].
This kind of description is suited to the oxide-on-oxide epitaxy, which is guided by the
mismatch between the oxygen sublattices [119]. The O-O distance for the different iron
oxide is slightly different (Table 6), so for growing different iron oxide on Al2O3 as an
example, there are different mismatches with the Al2O3 O-O distance of 2.75Å.
According to Gota et al. [109], the deposited iron oxide films showed different symmetries
and orientations: (11) for Al2O3, (2/
3 2/
3 ) R30º for Fe3O4 (111), and
(1/ 3 1/ 3 ) R30º for FeO (111), neglecting the differences of processing parameters.
Table 6: Lattice constant and crystal structure of iron oxides compared with
Al2O3, MgO and Pt
Crystal
Lattice
NND1 O-O
Lattice
structure
constant (Å)
distance(Å)
Mismatch2 (%)
corundum
a=5.128
2.75
Al2O3
MgO
Pt
MgO
cubic
a=4.212
2.98
(0001)
(001)
(111)
Pt
cubic
a=2.77
--
Fe2O3
corundum
a=5.03
2.90
+5.4
-2.68
+4.69
Fe2O3
inverse spinel
a=8.352
2.95
+7.3
-0.85
+6.5
Fe3O4
spinel
a=8.396
2.97
+8.0
-0.34
+7.22
FeO
cubic
a=3.04
3.04
+10.5
+2.01
+9.75
Material
Substrate
Film
1
Al2O3
2
NND: nearest neighbor distance, Based on O-O distance
From the lattice mismatch shown in Table 6, MgO (001) is an ideal substrate for the
MBE growth of Fe2O3, Fe3O4 and FeO, however, Chambers [107] reported the binary
spinels MgxFe3-xO4 are thermodynamically stable and form at the interface if the film is
82
grown at high temperature or annealed after growth. However, no growth or anneal
temperature range was reported as well as the other growth conditions, which could be
responsible for the formation of MgxFe3-xO4. Thevuthasan et al. reported that the
outdiffusion of Mg atoms into Fe3O4 film can occur at temperature higher than ~ 450ºC
followed by Fe interdiffusion starting at ~ 800ºC [120]. Based on Rutherford backscattering
(RBS), the Mg concentration in the film was 8 at. %, while the associated Fe concentration
in the substrate was 3.4 at. % after annealed at 800ºC.
An advantage of using Pt as a substrate is scanning tunneling microscopy (STM) can
be easily achieved because of sufficient conductivity. The primary disadvantage reported by
Chambers was that only one film orientation ([0001]Fe2O3 ) could be obtained on Pt [107],
but the author did not give the explanation. As for growing iron oxide film on Al2O3 by
MBE, the advantages are multiple film orientations and no interdiffusion occurring
compared with Pt and MgO respectively, however, it is impossible to use such techniques as
STM when growing Fe2O3 on Al2O3 because both the film and the substrate are
nonconductive. Therefore interface structure can be either obtained by AFM technique
which has lower resolution than STM; or by use of Pt (111) template for the iron oxide
growth on Al2O3 to make STM measurement feasible. In addition, the use of Pt template
can reduce lattice mismatch and improve film quality when growing Fe2O3 on Al2O3,
because the Pt (111) plane is 0.7% larger and 4.54% smaller than the basic-plane unit cells
of Al2O3 and Fe2O3 respectively [108]. According to Chambers et al. [121], the
relatively large compressive lattice mismatch (+5.4%) for Fe2O3 on Al2O3 led to the
formation of a heavily strained wetting layer of fully stoichiometric Fe2O3 that was
83
distorted along the [11 2 0] direction, provided that the first several tens of angstrom units
were grown at an exceedingly slow rate (~1 Å/min) at a substrate temperature of 450ºC.
After coverage of a few monolayers, the strained layer underwent relief by transforming to
3D nanocrystals.
Selecting a suitable substrate or engineering a suitable surface on a desired substrate
is an important step to grow high quality iron oxide films. The issues of crystal symmetry
and lattice mismatch should be carefully scrutinized before growing iron oxides as well as
BaM. In addition, substrate surface morphology and interface chemistry will influence the
grown iron oxide film as well. As discussed previously, one of the major advantage of MBE
and UHV is the ability to characterize and preserve substrate surface chemistry because of
the low contamination rate and large mean free path at pressure order of 10-9Torr. Under
MBE growth condition, it may be possible to study the influence of starting substrate
surface characteristics (e.g. roughness and chemistry) on the crystal structure, orientation
and growth mode of the resulting film.
Although many literature reported the cleaning procedure of the substrates
[108,109,122], the film growth literature has not generally commented about the effective of
different clean MgO or Al2O3 surfaces as long as they are clean surface. Typically, oxide
substrates are usually degreased followed by exposed to activated oxygen from a plasma
source in the UHV chamber, in order to make the surface free of C contamination before
growth, as judged by XPS or AES, and well ordered as judged by RHEED or LEED
(Low-energy electron diffraction).
84
3.2.4
The Choices of MBE Processing Parameters (T, Po2, plasma) for Iron Oxide
Growth
Chambers et al. [108] reported selective growth of Fe2O3, Fe2O3 and Fe3O4 thin
films by careful control of growth conditions and choice of substrate via
oxygen-plasma-assisted (OPA) MBE. In this study, Fe2O3 (0001) films were grown on Pt
(111) template, which was deposited on Al2O3 prior to growth of Fe2O3. The oxygen
partial pressure was ~4 × 10-5 Torr and electron cyclotron resonance (ECR) plasma source
was running at 250W to produce activated oxygen, however they did not reported how
much percent atomic O were produced under this condition.Fe2O3 (001) and Fe3O4 (001)
films were both grown on MgO (001), but the former were under similar growth conditions
to those of Fe2O3, whereas the latter were prepared under lower oxygen pressure (~3x10-6
Torr) and ECR plasma watts (200W), which supposed to produce lower activated O than
that used to grow Fe2O3 The substrate temperatures were kept in the range of 250 to
500ºC for all those films. The growth rate of Fe3O4 (0.6 ~ 0.8 Å/s) was controlled higher
than those of Fe2O3 and Fe2O3 (0.2 ~ 0.3 Å/s), which implies that the Fe/O flux ratio
used in Fe3O4 growth was 30 times more than that used in Fe2O3 and Fe2O3 growth.
However, there was no Fe/O flux ratio reported for each growth.
Chambers et al. reported that comparing the initial 3D growth of Fe2O3 on Pt,
with the 2D growth modes for Fe2O3 and Fe3O4 on MgO (001) was revealed by RHEED.
The RHEED patterns also showed sharp elongated streaks and radial Kikuchi lines for all
the Fe2O3, Fe2O3, and Fe3O4 growth films, inactive of flat film surfaces and good near
surface crystallinity. Moreover, XPS spectra were obtained to confirm the existence of the
85
Fe2O3 and Fe3O4 phases, but XPS could not distinguish Fe2O3 and Fe2O3 because both
phases only contain the trivalent Fe and thus the XPS Fe 2p peaks are the same. There was
no crystal structure information was reported by Chambers et al., so that the different crystal
structure of Fe2O3, and Fe2O3 are unknown, which may help distinguish Fe2O3 and
Fe2O3 as well.
Comparing with Chamber et al.[108], Gota et al. [109] reported much lower
deposition pressure (4.510-7 Torr) to grow Fe2O3 (0001) on Al2O3 (0001) via OPAMBE. The substrate temperature was fixed at 250ºC and the O/Fe flux ratio of ~30 was used,
at which allowed obtaining films of Fe2O3 composition even when the alumina substrate
was kept at room temperature. The growth rate was ~0.035 Å/s, much lower than that
discussed previously. The low growth rate may correlate with the lower pressure (4.510-7
Torr) used in contrast to Chamber’s study (410-5 Torr).
It was also reported by Gota et al. that at the very first stages (1-2 ML), the iron
oxide film showed an expanded FeO (111) like in-plane parameter. However, from XPS
results, the iron oxide phase was unusual because it contained Fe3+ cations instead of the
Fe2+. The thickness of phase was independent of the substrate and specific to ~2 ML. The
growth mode was conceived as starting by a 3D mode followed by a 2D mechanism after
the deposition of several oxide elementary cells. Large mismatch between alumina and
hematite (5.4%) may be a strong drive for 3D growth as means of reducing the total free
energy at the early stages. For a longer deposition time, those epitaxial islands of Fe2O3
began to nucleate, the amount of the hematite phase increased while that of the wustite
phase decreased. However, Fujii et al. [117] reported Stranski-Krastanov (S-K) growth
86
mode when depositing hematite films on Al2O3 (0001) by NO2-assited MBE, regardless of
the large lattice mismatch. They found period oscillations in the specular beam intensity
with a period corresponding to 1ML (-Fe-O-Fe-), which appeared at the beginning of
growth. The oscillations disappeared for a thickness of about 6 ML. This result was different
with Gota’s finding. It is hypothesized that in the case of out-of-equilibrium MBE
conditions, the use of different oxidants (atomic O or NO2) can change the kinetics
conditions, which led to very different growth modes took place for the same system.
From the discussions above, the full range of iron oxide stoichiometries can be
obtained by OPA-MBE by adjusting the relative Fe and activated oxygen fluxes at the
substrate. Generally, ferric iron oxides (Fe2O3, Fe2O3) prefer to grow at low growth rate
under oxygen-rich conditions, whereas Fe3O4 and FeO can be prepared at high growth rate
under oxygen-poor conditions (Figure 32) The choice of growth temperature is flexible,
ranging from 250ºC to 600ºC .
Figure 32: Phase diagram for the growth of iron oxides by OPA-MBE [123].
The substrate crystal symmetry and lattice mismatch play a key role in iron oxide
film quality and growth mode. In addition, substrate surface morphology and interface
87
chemistry will influence the grown iron oxide film. Selective growth of iron oxides has been
achieved by controlling the growth conditions as well as the use of appropriate substrate. By
successfully growing single crystal and stoichiometric Fe2O3 and Fe3O4, and fully
understanding the influences of MBE processing parameters on the film characteristics and
growth modes, integreting BaM and Fe3O4 on SiC with desired qualities can be realized.
3.2.5
Integrate Fe3O4 with Semiconductor for Spintronics
One of the major steps in developing next-generation spintronic devices is the
synthesis of magnetic/semiconductor hybrid materials with high efficient spin injection and
the Curie temperature above room temperature [124]. Half-metallic Fe3O4 has attracted
great attention recently for spintronics as it has high polarization at the Fermi level [39, 40]
and relatively high electronic conductivity at room temperature [31], which can benefit the
injection of spin carriers into the semiconductors [124].
It is hard to grow epitaxial Fe3O4 films directly on Si due to large lattice mismatch
(~29.3%) [125]. Reisinger et al. [126] used a double buffer layer of tinanium nitride (TiN)
and magnesium oxide (MgO) to obtain epitaxial magnetite thin film on Si (001) by pulsed
laser deposition. However, the double layer method is not very practical for real device
application because of the complex processing. Lu et al. [124] first realized integration of
half-metallic Fe3O4/semiconductor hybrid structure, which is one of the most promising
spintronic materials. Using GaAs substrate, the lattice mismatch between film and substrate
is decreased to around 5.0% if the Fe3O4 cell is rotated by 45° relative to GaAs cell in the
(100) plane. They grew epitaxial Fe films on GaAs substrates first and then oxidizing the
films to form Fe3O4 in a UHV chamber. However, RHEED pattern showed transmission
88
spots indicating the film has the three dimentional island growth. In addition, the post
oxidation method may cause impure phase in a thick film from insufficient oxidation deep
in the film or overoxidation on the surface. Using Si or GaAs subtrates, Fe3O4 (001) films
are grown in the (001) orientation.
Since epitaxial Fe3O4 (100) film surface is polar, the surface reconstruction usually
happens to minimize the surface energy [127]. The surface reconstruction may change the
electronic band structure and thus deteriorate the half-metallic properties. Vescovo et al.
studied the electronic structure of Fe3O4 (100) films grown on MgO (100), and only
observed ~ –50% spin polarization [128]. However, in the epitaxial Fe3O4 (111) film
grown on W (110), Dedkov et al. observed a spin polarization of – (80±5) % near Fermi
energy EF by using the spin-resolved photoemission spectroscopy [129]. Furthermore, the
Fe3O4 (111) surface is more stable and all the magnetic moments in the Fe3O4 (111) film lie
in the film plane.
SiC is a promising semiconductor material to grow Fe3O4 for next-generation
spintronic devices. SiC has wide bandgap (3.03eV), high breakdown field strength
(2.4106 V/cm), high saturation velocity (~2.0107cm/s), and high thermal conductivity
(4.9W/cmK), which are requied for high-power, high-temperature, and high-frequency
applications. Furthermore, since SiC has a hexagonal structure, Fe3O4 will grow in (111)
plane that has pseudo hexagonal structure matches with the hexagonal SiC substrate with
lattice mismatch of 3.77%. The (111) plane of Fe3O4 has been shown to have higher
polarization and be more stable than the (001) plane. These next-generation spintronic
devices that take advantage of both (111) oriented Fe3O4 and wide bandgap semiconductor
89
6H-SiC are expected to be non-volatile, fast and capable of simultaneous data storage and
consuming less energy [22].
3.3
Ferrite/Ferroelectric Multiferroic Heterostructure
In multiferroics, ferroic orders would cross interact with each other and creat
magnetoelectric (ME) coupling, defined as dielectric polarization varied with external
magnetic field, or electric field induced magnetization. Generally, multiferroics is classified
as single phase and composite heterostructure. In intrinsic single phase multiferroics, large
ME coupling at room temperature has been chased in order to realize electric field control of
magnetism, towards spintronics and multiple-state memory applications. However, in
composite multiferroics, especially in ferromagnetic/ferroelectric laminate heterostrucutre,
strong strain or stress mediated ME coupling has been demonstrated and used to design
novel magnetic field sensor and tunable microwave devices. In addition, by combining
magnetic tunneling junction (MTJ) and ferroelectric tunneling junction (FTJ), the
multiferroic tunneling junction (MFTJ) laminate structure is becoming a new research
frontier, towards spintronics and information storages [130].
3.3.1
ME and CME in Laminate Multiferroic Heterostructure
Given the great design flexibility and large ME coupling, ferro(ferri)magnetic/
ferroelectric laminate multiferroic heterostructures have drawn significant interests because
of their potential applications in magnetic sensors and RF tunable devices. By combining
piezoelectric and magnetostrictive effects, magnetic or electric field could induce strain or
stress in magnetic or ferroelectric phase that can be transferred to ferroelectric or magnetic
90
phase, therefore producing an electric displacement or a magnetic anisotropy. They are
called strain or stress mediated magneto-electric (ME) or conversed magneto-electric (CME)
coupling. To obtain strong ME coupling, high permeability, large magnetostriction constant,
and large piezoelectric coefficient are desired; to obtain large CME coupling, large
magnetostriction constant, large piezoelectric coefficient, and small 4πMs are necessary.
One of great application by using ME coupling in multiferroic heterostructure is
magnetic field sensor. A small ac or dc magnetic field can be determined by observing an
obvious electrical response through strain or stress mediated ME coupling defined as dE/dH
[131,132]. In this application, magnetic metal, piezoelectric ceramic, and polymer magnetic
metal/piezoelectric ceramic (e.g., metal glass/PZT, FeGa/PZT, Terfenol-D/PMN-PT) are
quite often used, because of the high permeability of magnetic metal, the large piezoelectric
coefficient of ceramic, and the small Yong’s modulus of polymer. Moreover, considering the
geometry and the design of heterostructure, mechanical resonance effect would improve
ME coefficient and sensitivity of magnetic field sensor by two order of magnititude if the
frequency of ac magnetic field matches with the intrinsic mechanical resonance frequency
of device. For example, Dong et al. investigated laminate metal glass/PZT fiber multiferroic
heterostructure as shown in Figure 33, illustrating a large ME coefficient up to 400V/cm Oe
at mechanical resonance frequency and towards a nano-gauss magnetic field sensor [132].
91
Figure 33: Configuration of FeBsiC (metal glass)/PZT fiber multiferroic
heterstructure (left) and the ME coefficient as a function of frequency (right) [132].
The reported magnetic field sensors have been mainly focused on the bulk form with
compatible magnetic and ferroelectric volume ratio, which puts several limits on integrating
sensors into MEMs. The challenge to design a micro size ME sensor is so-called clamping
effect, which refers to the substrate impacts of preventing the mechanical deformation of the
multiferroics layer. Very recently, by etching silicon substrate down to micro scale, a
micro-size ME sensor of FeGa/PZT/Si cantilever beam was demonstrated as shown in
Figure 34. The obvious thickness dependence of ME coefficient shows the feasibility of
being used in MEMs [133].
92
(d)
Figure 34: Schematic illustration of the fabricated ME device structure (a), with
cross-sectional SEM image shown in (c); the magnetic hysteresis loop of Fe0.7Ga0.3 film
is shown in (b), and the ME coefficient vs. Si cantilever thickenss is shown in (d) [133].
Electric field tuning of magnetism through converse ME coupling in multiferroic
heterostrucutres becomes a new exciting frontier due to its potential applications in novel
electrostatically tunable microwave magnetic devices such as filters, resonators, inductors
and phase shifters. The magnetic anisotropy can be modulated in multiferroic
heterostructure through magnetoelastic and piezoelectric effects by applying electric field
on ferroelectric phase. It would realize electric field rotating of magnetization by 90 degree.
Generally, any effect or phenomena involved with magnetic anisotropy change can be
manipulated by an electric field that facilitates to realize tunable magnetic device. For
example, ferromagnetic resonance (FMR), as a fundamental property for high frequency
magnetic applications, is highly related to magnetic anisotropy. Therefore, FMR can be
93
dynamically tuned through electric field induced magnetic anisotropy, and thus a shift of
permeability spectrum can be observed. In order to obtain a large FMR tunable range,
laminate ferrite/ferroelectric heterostructure structures with large piezoelectric and
magnetostiction coefficients were often chosen, such as, Fe3O4/PZN-PT, YIG/PMN-PT,
NiFe2O4/PZT, and BaM/BSTO. So far, the largest FMR tunable range by applying electric
field was obtained in Fe3O4/PZN-PT heterostructure up to 860 Oe with a FMR linewidth of
330-380 Oe as shown in Figure 35 [134].
Figure 35: Ferromagnetic resonance absorption spectra of Fe3O4/PZN-PT showed
giant electric field tunability [134].
So far, several simple microwave tunable devices, such as phase shifter, resonator
and filter have been demonstrated as shown in Figure 36 [135,136]. As well known, for
microwave application that highly relies on the FMR effect, a low linewidth of magnetic
material is critically desired. However, the difficulties to find out a ferrite with low FMR
linewidth and large magnetostriction constant limit its applicaiton. Alternative multiferroic
heterostructure consists of magnetic metal and ferroelectric ceramic has been considered,
such as metal/piezoelectric FeGaB/PZN-PT, exhibiting a low linewidth and a large FMR
94
tunable range. However, the magnetic metal film has to be thinner than skin depth to avoid
eddy current effect. In addition, ions doped spinel and garnet ferrites were considered to be
involved in tunable multiferroic microwave devices, since by doping ions, such as Co and
Mn, the magnetostriciton constant would be enhanced without loss of FMR linewidth.
Figure 36: Tuanble resonator YIG/PMN-PT (left) and tunable filter YIG/PZT (right)
[135].
3.3.2
Multiferroic Tunnel Junction (MFTJ) in Laminate Multiferroic
Heterostructure
The phenomenon of electron tunneling has been known since the advent of quantum
mechanics. One type of important tunneling junction effect is magnetic tunnel junction in
which the tunneling current depends on the relative magnetization orientation of the two
ferromagnetic electrodes. Recently, magnetic tunneling junctions have attracted
considerable interest due to their large tunneling magnetoresistant effect and potential
application in spin-electronic devices such as magnetic field sensors and magnetic random
access memories. In magnetic tunneling junction, the insulator tunneling barriers usually
used are non-polarized dielectric oxides, such as Al2O3 and MgO. There is another concept
of tunneling junction, ferroelectric tunneling junction (FTJ), in which the tunneling current
95
depends on the relative orientation of spontaneous polarization of ferroelectric barrier layer
as shown in Figure 37 [130]. By applying electric field to switch ferroelectric spontaneous
polarization, two states of resistance can be obtained.
Figure 37: Schematic diagram of a tunneling junction [130], which consists of two
electrodes separated by a nanometer-thick ferroelectric barrier layer.
From above, it is possible to combine ferroelectric tunneling effect and magnetic
tunneling effect to realize four resistance states and apply to new generation of spintronic
device and information storage. Today, this becomes a very excited research frontier for
multifunctional tunneling junction with multiferroics involved. In multiferroic tunneling
junction, ferro (ferri) magnetic materials are used as electrodes and a ferroelectric thin film
is used as a tunneling barrier. By manipulating the relative orientation of magnetization and
spontaneous polarization, four logical resistant states can be achieved. To our best
knowledge, so far, only simulation results have been done to predicte this MFTJ as shown in
96
Figure 38, and no experimental results were reported [136]. It is hypothesized that by using
MBE, Fe3O4/BaTiO3/ Fe3O4 heterostructure with sharp interface can be actually prepared to
study the MFTJ effect and understand its potential application as a MFTJ.
Figure 38: Effects of ferroelectricity in the SrRuO3/BaTiO3/SrRuO3 MFTJ [136]. (a)
Schematic double-well potential for the MFTJ (solid line) and for the bulk BaTiO3
(dashed line). (b) Cell averaged electrostatic potential energy profile for polarization
to the right P→ (blue) and left P←(red) states and the interfaces are indicated by
vertical dashed lines.
3.4
Summary
From studies of BaM film growth by such methods as PLD, rf sputtering, and sol-gel,
a commonality is that a higher growth temperature or post annealing temperature is needed
to produce highly c-axis oriented films with good stoichiometry, reduced defects
concentration, and narrow FMR linewidth. However, in order to apply BaM films to
integrated circuit processing for microwave devices, it is required to lower the deposition
temperature as well as eliminate if possible the post annealing requirement. One of the
potential advantages of MBE growth is relatively lower growth temperature and no need of
97
the post annealing process. With MBE, by atomically controlling the source flux, it will be
possible to engineer growth processes for BaM thin films with desired crystal structure and
orientation and magnetic properties.
Successful integration of single crystalline BaM on 6H-SiC may be influenced by
many different parameters and conditions, such as substrate surface structure, growth
temperature, O2 pressure, and growth species. Proper templates can effectively prevent the
interdiffusion between the substrate and the film, promote c-axis orientation, and improve
the crystallinity at relative lower temperature. However, many templates employed
previously were amorphous, and may not be capable of dispersing thermal and lattice stain
as well as forming an effective, abrupt interface.
By understanding of BaM grown via other methods, it is hypothesized that the
formation of a surface oxide may lead to the deterioration of morphology and magnetic
properties of BaM film and a suitable template may be helpful to mediate interfacial stress
and lesson the density of defects at the interface. The formation of a functional interface on
6H-SiC may be the limitations of the growth methods such as PLD, sputtering, and LPE. By
utilizing the benefits of UHV and MBE, which include substrate surface preservation, low
contamination rates, and atomic level control of BaM growth, it will be possible to
understand the nucleation and growth mechanisms of BaM film, then engineer an effective
process that grow BaM film on 6H-SiC with desired properties for microwave device
applications.
For iron oxide grown by MBE, the literature shows that substrate crystal symmetry,
lattice constant and surface reactivity play a key role in oxide film quality and growth mode.
98
Selective growth of iron oxides has been achieved by controlling the growth conditions as
well as the use of appropriate substrate. A full range of iron oxide stoichiometries can be
obtained by oxygen plasma assisted MBE by adjusting the relative Fe and activated oxygen
fluxes at the substrate. Generally, ferric iron oxides (Fe2O3, Fe2O3) prefer to grow at
low growth rate under oxygen-rich conditions, whereas Fe3O4 and FeO can be prepared at
high growth rate under oxygen-poor conditions. The choice of growth temperature is
flexible, ranging from 250ºC to 600ºC.
One of the major steps in developing next-generation spintronic devices is the
synthesis of magnetic/semiconductor hybrid materials with high efficient spin injection and
the Curie temperature above room temperature. Integration of half-metallic Fe3O4 with SiC
can not only produce (111) plane of Fe3O4 that has been shown to have higher polarization
and be more stable than the (001) plane, but also can offer great potential for high-power,
high-temperature, and high-frequency spintronics applications.
Because magnetoelectric coupling between layers is extremely sensitive to the
quality of the interface, by understanding how the initial structure and composition of the
interface evolve with film thickness by MBE, engineering an effective interface for more
efficient ferrite-ferroelectric coupling can be realized. Using this ferrite/ferroelectric
heterostructure with sharp interface, it is possible to combine ferroelectric tunneling effect
and magnetic tunneling effect to realize four resistance states and apply towards new
generation of spintronic devices and information storage.
99
4.0 EXPERIMENTAL
In order to successfully integrate single crystalline, epitaxial ferrite (BaM and Fe3O4)
on 6H-SiC with a controlled orientation and stoichiometry, it will be necessary to determine
and understand the roles and interactions of each individual species flux and processing
parameters on the nucleation and growth mechanism of the oxide films. In this chapter, a
detailed description of the thin film growth methods and the techniques used to analyze the
films will be introduced.
4.1
Thin Film Preparation
All thin oxide films in this dissertation were deposited by molecular beam epitaxy
(MBE) and Pulsed laser deposition (PLD) in the Interface Engineering Laboratory and
CM3IC center at Northeastern University. In PLD, high energy Excimer laser pulses,
impinging upon a homogeneous target, produce a vapor flux, also known as a plume. The
content of the plume, including ions, atoms, molecules, clusters, and other species, is
condensed on the nearby substrate to form the film. In MBE, a molecular beam of metal
species is created by heating the material, which then evaporates and condensates on a
substrate. To grow metal oxides with this method, for instance BaM, the Ba and Fe fluxes
and a flux of oxygen (O & O2) are employed simultaneously. Comparing to PLD of growing
oxide film by directly transferring the stoichiometry from the target to the film, MBE is
more complicated because the right stoichiometry can only be achieved by controlling
individual flux precisely. However, a detailed chemical and structural evolution study of the
interface structure is usually not feasible due to the limitations of characterization capability
100
in the PLD system. MBE has the advantages to study the interface by in-situ analysis tools,
such as x-ray photoelectron spectroscopy (XPS) and reflection high energy electron
diffraction (RHEED).
4.1.1
Thin Film Growth: Molecular Beam Epitaxy (MBE)
4.1.1.1 The Ultra High Vacuum Set Up for MBE System
The MBE system used for film growth and characterization in this dissertation
consists of two interconnected custom-built UHV chambers, which are shown in Figure 39
as growth chamber and analysis chamber. Prior to analysis and growth, all samples are
loaded through a small load lock chamber that is attached directly to the analysis chamber
but separated by a UHV-compatible manual gate valve. The load lock chamber is
independently pumped by a mini turbo pump (Leybold, Turbovac 50). Once a pressure
lower than ~5×10-7 Torr (as measured by an Ionivac hot cathode pressure gauge) is reached,
the samples are transferred into the analysis chamber onto a manipulation stage (with x, y, z,
and  movement) by using a magnetic linear transfer arm. The analysis chamber is pumped
by a Varian 500181B ion pump, which maintains a background pressure of ~2×10-9 Torr (as
measured by an ion gauge set with a nitrogen gas emission factor). Comparing to a turbo
pump, the ion pump is an entrapment pump that can minimize vibrational noise during XPS
and AES characterization. The analysis chamber is equipped with AES, XPS, SED, Ar ion
gun, and an electron flood gun. The analysis chamber is connected to another UHV growth
chamber, separated by a UHV-compatible manual gate valve that allows transferring sample
between the two chambers without breaking vacuum and thus protecting sample from
101
contamination. Another magnetic linear transfer arm is used to realize the sample transfer
between the growth chamber and analysis chamber.
XPS
Growth
chamber
Analysis
chamber
AES
RHEED
Figure 39: UHV system consisting of two interconnected chambers. The chambers are
separated by a UHV compatible gate valve. The growth chamber consists of a remote
oxygen atom source, solid source effusion cells (Mg, Ba, Fe), a Ti sublimator, and a
RHEED system. The analysis chamber consists of a XPS hemispherical analyzer and
an AES single pass CMA.
The growth chamber is usually maintained to a pressure of ~2×10-9 Torr by a
Leybold Turbovac 600C turbo pump with a pumping speed of 560 L/sec and a Leybold
Trivac D25B rotary vane roughing pump. There are a custom built heater, a SPECS
Scientific Instruments dual source, low-temperature effusion cell (for Ba and Mg), a Veeco
high temperature effusion cell (for Fe), a modified Varian Ti-Ball (for Ti), an Ar ion gun, an
Oxford Applied Research remote oxygen rf-plasma source (model HD25), and a Staib
RHEED system (RH15) equipped with the growth chamber. The custom built substrate
heater can be moved in and out (z movement) as well as rotate ( rotation). The heater is
102
capable of heating the sustrate from room temperature up to 900 oC, as measured by a type C
thermocouple in direct contact to the molybdenum puck. The more accurate temperature
measurement was performed using a two-color optical pyrometer, which typically read 50 –
75 oC below the thermocouple reading (refer to Brian Doyle’s thesis) [69].
For film growth, a SPECS Scientific Instruments low-temperature effusion cell,
with two independently controlled crucibles supplies magnesium (99.98%, Alpha Aesar)
for MgO films and barium (99.9%, Electronic Space Products International) for BaM films.
The iron (99.999%, Sigma-Aldrich), for BaM films, was supplied using a Veeco high
temperature effusion cell. The Mg, Ba, and Fe fluxes were calculated based on the
thermocouple temperature reading and vapor pressure calculations based on the following
equation:  
3.51 10 22 PA
L2 MT
Where:  = atomic flux (#/cm2sec)
P = vapor pressure (Torr)
A = aperture area (cm2)
L = distance from source to substrate (cm)
M = molecular weight (g/mol)
T = temperature (oC)
Changing the temperature of the appropriate cell (thermocouple resolution of 1 oC),
would change the flux of that metal. By having the capability of controlling a cell
temperature to ± 1 oC, it was possible to adjust the individual fluxes to obtain desired film
chemistry. The effusion cell temperatures were typically run at temperatures several
hundred oC below the melting point, since the ceramic crucible would crack if the material
was melt and then cooled again. It is worth to keep in mind that for the Vecco high
temperature cell, the temperature reading from the thermocouple is around 200 oC lower
than the actual crucible temperature. Since the melting point of Fe is around 1540 oC, the
103
operating temperature should be maintained below 1300 oC to prevent breaking the crucible.
During the growth that Fe is used, it is necessary to check the cell from the top viewport of
the chamber to prevent crucible failure. Typical fluxes for film growth ranged from 1013 –
1015 atoms/cm2 sec with the maximum achievable flux dependent on the melting point and
vapor pressure of the each material. The titanium metal used for BTO growth was supplied
using a Varian mini Ti-ball, which was originally designed for use as a titanium sublimation
source for an ion pump and is based on the design by Theis et al. [137]. A detailed
description of Ti-ball can be found in Dr. Goodrich’s dissertation [73].
The atomic oxygen used for film growth was generated from an Oxford Applied
Research remote oxygen rf-atom source model HD25. The plasma was generated within an
Al2O3 discharge tube and exited through an Al2O3 aperture plate with 276 holes that were
0.2 mm in diameter. The size, quantity, and dispersion of the through holes can minimize the
line-of-site effect associated with UHV. The oxygen pressure of 1.0 × 10-6 to 1.0 × 10-5 Torr
correlates to an equivalent molecular oxygen flux of 3.6 × 1014 /cm2 s and 3.6 × 1015 /cm2 s.
Although the oxygen environment is a combination of oxygen species, the flux is
represented on the basis of O2 molecules. The oxygen flux was calculated from the
following equation.
O
2
 3.513  10 22 P 

 
MT


Where:
= flux (#moluculars/cm2 s)
P = chamber pressure (Torr)
M = molecular weight (g/mol)
T = gas temperature (K)
The source was built with ion filter bias plates that were located at the end of the
104
discharge tube and consisted of two parallel plates aligned parallel to the flow of oxygen.
One plate was connected to ground and the other was connected to a voltage source. By
supplying a voltage bias between the two plates, all the charged species (O*, O2*, etc) can
be deflected and removed from the neutral species. By turning the deflection plate bias on
and off allowed for the use of neutral or neutral/charged oxygen. Typically, only neutral
species with lower kinetic energy were involved in film growth. By varying the rf power, it
was possible to vary the relative amount of O2 that was cracked into different atomic oxygen
and oxygen radical species. Direct measurement of the amount of atomic oxygen was made
possible through the use of an optical emission detector (OED), which was aligned
line-of-site to the discharge tube where the plasma was being generated. By using an in-line
optical filter equipped with the OED, only photons with λ of 844 nm can be collected, which
allows qualifying the amount of neutral atomic oxygen generated within the plasma. A
higher emission reading corresponds to higher oxygen generation. For all the films grown in
this dissertation, an OED usually reads between 30 and 450 mV.
4.1.1.2 BaM and Fe3O4 on 6H-SiC by MBE
The n-type research grade 6H-SiC substrates used in this study were on-axis (± 0.5o)
supplied by Cree Inc. The SiC were first degreased in heated solvents of trichloroethylene,
acetone, and methanol and then loaded in a custom-built hydrogen furnace. A typical H2
cleaning procedure consists of the following steps.
1. Remove loading section, install a Ta boat between Cu clamps and tighten down the Cu
clamps
105
2. Place the SiC sample (< 0.7 × 0.7 cm2 ) at the center of the Ta boat and secure loading
section
3. Purge furnace with hydrogen for 5 minutes (ignite exhaust burner) and turn on cooling
water
4. Adjust hydrogen to a desired flow rate (11.4 lpm, reading of 35-45 from the flow meter)
5. Turn on power supply (Sorenson DCR 20-115B) and increase current 5 A/30 sec
(100oC/30 sec); monitor temperature with optical pyrometer (two color) with filter on
6. Once desired temperature (1600oC, [1725oC as of 06/2007]) is reached, maintain
desired hydrogen flow rate and temperature for 30 minutes. Typical current is around
70-85A, depending on the width of Ta boat
7. Reduce power (2.5 A/30 sec, [remove power immediately as of 12/2008]) while
maintaining H2 flow.
8. Turn off power supply, hydrogen, and cooling water; remove loading section and
retrieve sample
The hydrogen cleaned 6H-SiC (0001) has an atomically smooth, stepped surface
with removed surface contamination and scratches from mechanic polishing, and is
terminated by a silicate adlayer [138]. The characterization of cleaned 6H-SiC (0001) is
described in Dr. Goodrich’s dissertation [73]. The samples were then mounted on a
molybdenum puck with conductive silver paint, immediately loaded into a load lock, and
then transferred to the analysis chamber with a base pressure of 2.0 × 10-9 Torr, where they
were checked for chemistry with XPS. After XPS was done, the samples were transferred to
the growth chamber, where the substrate crystallinity was checked with RHEED.
106
All films were grown by depositing metal from effusion cells, with the
simultaneous oxidation of the metal atoms by atomic oxygen, which is generated from a
remote oxygen rf-plasma source. The oxygen pressures reported here are the chamber
background pressures from the remote oxygen plasma source. Although atomic oxygen
versus molecular oxygen has not been fully characterized due to equipment limitations, the
optical photodiode filter on the plasma discharge tube enables a relative measure of atomic
oxygen species, which increases with chamber pressure even at constant plasma power. For
example, an increase in oxygen pressure from 1.0 × 10-6 to 1.0 × 10-5 Torr at a power of 200
W resulted in an increase in the photomultiplier reading from 125mV to 400mV, indicating
an increase of atomic oxygen in the chamber. Based on literature characterizations of remote
rf-plasma sources showing a 36% atomic oxygen content at 300 W [ 139 ] and our
rudimentary characterization of the source, we expect less than 15 % atomic O in the
background oxygen pressure.
In all experiments, the MgO templates were deposited on 6H-SiC at a substrate
temperature of 150°C in a constant oxygen plasma environment with an oxygen pressure of
5.0 ± 0.2 ×10-6 Torr with the plasma source running at 100W and a photodiode reading of 35
mV. Typical operating temperatures for Mg source were between 320oC-360oC, which
would result in the growth rates ranging from 1 Å/min to 1 nm/min.
BaM films were deposited by MBE at a substrate temperature between 300°C and
900°C. With plasma power held constant at 200W, the oxygen partial pressures ranged from
1.0 × 10-6 to 1.0 × 10-5 Torr to give a photodiode reading from 125 mV to 400 mV. The
optimal pressure for BaM growth is found to be 3.0×10-6 Torr with a photodiode reading of
107
210 mV, where optimal was defined as achieving stoichiometric films with aligned single
crystals and a smooth surface. The typical operating temperature for Fe was between
1180-1280 oC and for Ba was between 480-540 oC. The growth rates for BaM films
presented in this work were typically 0.2- 0.3 Å/s.
Fe3O4 films were deposited at a substrate temperature of 350 °C. The oxygen plasma
was running with a photodiode reading from 40 mV to 60 mV at pressure of (1.0 ± 0.2) ×
10-6 Torr and power of 100 ± 5 W. The typical operating temperature for Fe was between
1250-1280 oC and the growth rate is ~ 0.1 Å/s.
BaTiO3 films were deposited at a substrate temperature of 750 °C. The oxygen
pressure of (1.0 ± 0.2) × 10-6 Torr with plasma power running at 100 ± 5 W gives a
photodiode reading ~ 35 mV during growth. The typical operating temperature for Ba was
520 to 545 oC to match the Ti flux from a Ti ball running at 44.6 A. The BTO growth rate is
typically ~ 0.08 Å/s.
4.1.2
Thin Film Growth: Pulsed Laser Deposition (PLD)
Three generations of BaM film have been grown on wide bandgap 6H-SiC by PLD,
the improvement in magnetic properties between different generations of BaM films is
linked to the initial stages of BaM film growth. These initial stages of growth are highly
influenced by the template used between the SiC and the BaM. Through three generations of
templates from none (Gen I) [11], to a PLD deposited MgO/BaM interwoven structure (Gen
II) [140], to a MBE-grown 10 nm template of MgO (Gen III) [141], the magnetic quality of
the BaM improved (as measured by FMR linewidth).
The Gen I BaM films were grown on the n-type 6H-SiC substrates directly without
108
using any templates. The n-type 6H-SiC used were also supplied by Cree Inc. The SiC
surface preparation is the same as described in chapter 4.1.1.
The Gen II BaM films were grown on an interworven multilayers consist of
alternating MgO and BaM layers. The interworven multilayers consist of alternating MgO
and BaM layers, represented by, [MgO(An)/BaM(Bn)]n, where An is the thickness of the nth
MgO layer, Bn is the thickness of the nth BaM layer, and n is the number of repeated pairs of
alternating layers. Each layer thickness is controlled by the number of laser pulses or shots
incident upon each target, where a single laser shot provides ~0.01 nm of MgO or BaM. The
first interwoven layer is the thickest at A1 = 80 shots MgO B1 = 150 shots BaM. Each
subsequent layer has 10 less MgO shots than the layer before. Thus the 8th layer is 10 shots
of MgO followed by pure BaM deposition until a 0.3 ~ 0.5 m thick BaM layer is grown.
The Gen III BaM films were grown on a MBE-grown MgO template. Crystalline
MgO (111) film was grown by MBE at a temperature of 150oC on the silicon face of
6H-SiC by using a low-temperature Mg effusion cell and a remote oxygen plasma source.
The oxygen plasma was held constant at 100 W and a chamber pressure of 5×10−6 Torr.
The Mg:O flux ratio was ~1:20 which resulted in single crystalline MgO films that grew
conformal on the 6H-SiC atomic steps with the epitaxial relationship MgO (111)/6H-SiC
(0001). The MgO film thickness was controlled from 2 nm to 12 nm by changing the
growth time. After the MgO template was deposited on SiC by MBE, the sample was
immediately loaded to the PLD chamber.
All three generation BaM Film deposition was carried out by PLD (NanoPLDTM
system, PVD Inc.) with a KrF Excimer laser. The PLD chamber was first evacuated to a
109
base pressure of 5×10-7 Torr before deposition. The Tui KrF Excimer laser delivered a beam
at a wavelength of 248 nm with energy per pulse of 250 mJ, which was focused by a lens
system on a homogeneous BaFe12O19 target. During deposition, the BaFe12O19 target was
rotated to ensure uniform wear. The substrate was located at a distance of 5.5 cm away from
target. All depositions were carried out at 915 ºC with an O2 pressure ranging from 20 mTorr
to 300 mTorr. During deposition, the growth rate was controlled by the laser pulse repetition
rate. In the first 5 minutes of deposition, the pulse rate was slowly raised from 2 Hz to 16 Hz
to promote good adhesion and epitaxy and then fixed at 16 Hz for the remaining film
deposition time (~ 30 mins). The thickness of all final BaM films was 0.4 ± 0.1m.
Post-deposition annealing of the films occurred in air at 1050C in two steps ranging in time
from 2-minutes to 7-minutes for each step. It was found that the optimum annealing time for
BaM grown on SiC is two steps of 2-minute duration each, which is much shorter than the
hours of annealing for BaM grown on such substrates as sapphire [76].
4.2
Thin film Characterization
The characterization techniques involved in this thesis consist of various in-situ and
ex-situ techniques. In-situ reflection high energy electron diffraction (RHEED) was used
real-time to monitor the crystallinity, crystal structure, and surface reconstruction. X-ray
photoelectron spectroscopy (XPS) was used to characterize the chemistry and bonding
states of the films. Atomic force microscopy (AFM), scanning electron microscopy (SEM),
and x-ray diffraction (XRD) were used to characterize the surface morphology (roughness)
and crystal structure (orientation). Transmission electron microscopy (TEM) and Energy
110
Dispersive X-ray (EDX) analysis was performed at university of York (UK) to examine the
interface structure. Magnetic characterization on the BaM and Fe3O4 were performed by
vibrating
sample
magnetrometry
(VSM)
and
ferromagnetic
resonance
(FMR)
measurements. Magnetic force microscopy (MFM) was used to image the magnetic domain
structure.
4.2.1
X-ray Photoelectron Spectroscopy
X-ray Photoelectron Spectroscopy (XPS), also known as Electron Spectroscopy for
Chemical Analysis (ESCA), is a surface analysis technique to determine the composition
and bonding information of sample surface. In this method, soft X-rays (1486.6eV from an
Al anode or 1253.6eV from an Mg anode) ionize the atoms and lead to the ejection of
electrons. The Photoejected electrons enter a hemispherical analyzer where they are counted
and their kinetic energy is measured. From the measured kinetic energy and the incident
x-ray energy, the electron’s binding energy is determined by the following equation.
BE =h KE s
Where BE is the binding energy of the photoelectron, his the energy of the
incident photon, KE is the kinetic energy of the photoelectron, and s is the work function.
The work function is the minimum energy required to eject an electron from the highest
occupied level into vacuum level. Since the binding energy of the photoelectron is
characteristic to a specific atom regardless of incident energy and instrumentation, it can be
used to identify the elements present. As a result, the position and shape of the photoelectron
spectra can be used to obtain the bonding states of specific elements.
Table 7 lists the sampling depths of several photoelectrons that are involved in this
111
thesis. These sampling depths were derived from NIST database 82: Electron Effective
Attenuation Length (EAL). The film thickness was estimated using the TPP-2M and Gries
equations [142] and the XPS signal attenuation of the Si 2p photoelectrons. These thickness
calculations have an inherent error around 10% due to transport approximations, surface
roughness, surface excitations, and surface refraction.
Table 7: A list of photoelectron energies and sampling depths for selected elements
that are of interest for this thesis.
Element
(emission)
KE (eV):
Al 
Mg 
Depth (nm)
Al 
Mg 
Mg
(2p/1s)
Ba
(3d5)
Ti
(2p3)
Fe (2p3)
1202.0 955.0
969.0 722.0
1436.9/282.6
1203.9/---
706.9
473.9
1032.8
799.8
779.9
546.9
8.6
7.3
9.9/3.1
8.7/---
5.8
4.3
7.7
6.3
6.2
4.8
Si
(2p)
C
(1s)
1387.5
1154.5
9.7
8.4
O
(1s)
7.2
5.9
The XPS system consists of a dual source, non-monochromated x-ray source (Phi
model 04-548) and a hemispherical analyzer (Phi model 10-360). The two x-ray options are
Mg (1253.6 eV) and Al  (1486.6 eV) operated at 300 W. Careful calibration and
fitting of gold and the Au4f7 photoelectron peak set the system’s minimum full width at half
maximum (FWHM) of 1.2 eV with an 80% Gaussian/Lorentzian distribution, when run at a
pass energy of 35.75 eV. Background subtraction was performed using the integrated
Shirley method, which is the preferred method for inorganic crystalline solids [143]. The
analysis spot size for XPS is controlled by a manually set aperture and was typically set at a
diameter of 1.1 mm. XPS data collection and processing was performed using RBD
Instruments (formerly RBD Enterprises) AugerScan software version 3.22.
The 04-303 Ion Gun with the 11-065 Ion Gun Control was also used in this thesis to
generate a reproducible ion beam for sputter cleaning and depth profiling. The beam voltage
112
supply can go from 0 to 5 keV. A high beam voltage will result in high etch rate but more
surface damage. For the depth profiling, the beam voltage of 4 keV was used to ionize argon
of 15mPa. However for the sputter cleaning, in order to minimize damaging the surface,
much lower beam voltage of 0.5 keV was used to ionize argon of 5mPa. The raster size for
depth profiling is typically 4×4 mm2 area, and 10×10 mm2 for surface dusting. Sputtering
rate measurements are performed on a Si substrate with a 1000 Å SiO2 thin film, and was
determined to be ~ 10 Å/min at our depth profiling conditions.
4.2.2
Reflection High Energy Electron Diffraction
Reflection high-energy electron diffraction is an in-situ, real time analysis technique
that is used to characterize surface structure in ultra high vacuum environment. RHEED is
based on the scattering effect of a high energy electron beam (5-100keV) that incident on the
sample surface at a glancing angle less than 2º. RHEED is also a surface sensitive analysis
technique because the glancing angle of the incident beam has a penetration depth on the
order of a few monolayers.
There are two types of diffractions, one is flat surface and the other is 3D crystal
diffraction. But most diffraction pattern is a result of both 3D and surface diffraction,
because most surfaces are not perfectly flat. Figure 40 illustrated the diffraction pattern
resulting from different surface features. The ideal single crystal flat surface results in
distinct spots located on the Laue rings. The more diffuse the pattern, the less crystalline
order. Amorphous RHEED patters have a completely diffuse pattern with no observed
diffraction spots. This is due to the fact that diffraction maxima require long range order for
the constructive interference of the diffracted electrons. Polycrystalline films result in only
113
Laue rings and no distinct spots. 3D growth results in distinct spots, which does not match
with the Laue rings. Whereas 2D growth results in distinct spots with vertical elongation
that match with Laue rings.
(a)
(c)
(b)
(d)
Figure 40: Schematic illustration of diffraction resulting from different surface
features [144]. (a) ideal, single crystalline, flat surface, (b) polycrystalline surface, (c)
single crystalline with 3-D island features, and (d) single crystalline with 2-D features.
Interpretation of 3-D features and islands is significantly more involved. Once the
surface becomes three dimensional, transmission effects begin to influence the RHEED
pattern. Figure 41 illustrates how the shape of the 3-D features can impact the shape of the
transmission pattern. Further, if the 3-D features are faceted, the transmission pattern will
exhibit chevron tails, which are a result of transmission through to the symmetric planes of
the faceted feature (right figure in Figure 41).
Figure 41: Diffraction spot shape due to transmission through different sized 3-D
features (left) and chevron diffraction characteristic of surface faceting (right).
114
In addition to determining growth mode, RHEED can used to monitor growth rate
through intensity oscillation. The intensity oscillation will have a distinct repeating pattern
for 2D growth, which can be used to determine growth rate. However, there is no distinct
repeating pattern during 3D growth. RHEED interpretation can also be helpful for
determining surface reconstructions and relaxation mechanisms.
A Staib Instruments RH-15 RHEED system with a maximum accelerating voltage of
15 keV was used in this thesis. Typical RHEED images were obtained using incident beam
energy of 12.5 keV and 1.5A at a glancing angle of 2o. The RHEED images were digitally
captured and processed using a k-Space Associates KSA 400 digital CCD camera and OEM
software version 4.75.
4.2.3
Atomic Force Microscopy
AFM characterization of the substrates and films is very important for
characterizing the surface morphology and roughness. AFM characterization was
performed using an Ambios Technology 2SAAVO USPM. Typically, AFM characterization
was performed in wave mode (non-contact/tapping) with the cantilever frequency around
~186 kHz and a tip radius of curvature less than 10 nm. The scan resolution was set at
1024×1024 and collected at a scan frequency of 1 Hz. AFM characterization of the surfaces
was limited to the 40 m lateral movement of the piezotube. Therefore, it is possible that the
images obtained by AFM would not accurately represent the surface characteristics.
Multiple AFM scans could be collected across the sample surface in order to ensure an
accurate representation.
In addition to traditional morphology characterization, the USPM could also be used
115
for magnetic force microscopy (MFM) to study the structure of domains in the magnetic
films. The USPM can record either the amplitude (- Amp) of the cantilever vibration or the
phase (-Phase) shift of the cantilever vibration as the probe is rastered across the surface.
Both topography and magnetic structure have been obtained using the “follow the terrain”
method (T-). In such measurements, the probe passes over each scan line twice. In the first
pass the topology of the surface is recorded with tapping mode, then in the second pass the
feedback is disabled and magnetic force data (cantilever phase shift) are recorded as the
probe scans at a constant vertical distance above the surface. The gradient of magnetic force
introduces a phase shift of the cantilever vibration, and can be recorded as the feedback
control signal. T-Phase mode was used in this thesis to collect MFM images since the phase
signal generally provides the best results [145].
4.2.4
Scanning Electron Microscopy
A Hitachi S-4700 field emission SEM was used to be in conjunction with AFM to
accurately characterize morphology of the substrate and film surfaces, since SEM can
image much larger area of the surface than AFM. Under high vacuum, a high energy
electron beam (1-10keV) is incident on a sample surface to generate scattered electrons. The
scattered electrons consist of secondary, backscattered, and Auger electrons as well as
characteristic x-rays. Usually, SEM images are obtained by detecting the secondary or
backscattered electrons.
By separating the secondary electrons from backscattered electrons, it is possible to
create different resolution images. Secondary electron imaging contributes to higher
116
resolution and better contrast than backscattered electron because of the different generation
mechanisms. The resolution of SEM imaging is decided by the spot size, beam voltage,
beam current, and working distance. In this thesis, secondary electron imaging at 12 mm
working distance, 2 kV accelerating voltage, and 10 A beam current was used to obtain the
SEM images for uncoated samples. High resolution SEM imaging was performed at a
working distance of 3 mm.
4.2.5
X-ray Diffraction
XRD is an important characterization technique for determining the orientation and
epitaxy of the crystalline materials. However, XRD is limited to films that are typically
greater than 10 nm in thickness. This is due to the requirement of many diffraction planes
for the constructive interference of Bragg’s law, which is necessary for the detection of
diffracted x-rays. Although characterization of thin (less than 10 nm) was not possible, XRD
was used to characterize thicker films in order to determine the orientation and epitaxy of
the films. XRD data was collected using a Cu  (40 kV, 20 mA) on a Bruker D5005 :2
Bragg-Brentano diffractometer equipped with a curved graphite crystal diffracted beam
monochromator and a NaI scintillation detector. The system was equipped with a
divergence and receiving slit set to 1.0 mm and a detector slit of 0.6 mm. Due to system
limitations, only -2 scans could be collected. Therefore, it was not possible to
characterize in-plane stress or rotation. However, in conjunction with RHEED, a complete
epitaxy picture could be determined. RHEED could be used to determine the orientation of
the growing film as well as any in-plane rotation of the crystal lattice. XRD could be used to
confirm the out-of-plane orientation and epitaxy. Although RHEED can be used to
117
determine the in-plane lattice spacing (i.e. lattice strain) precise calibration using known
standards was not performed, thus limiting the quantitative lattice spacing to relative
differences. Knowing the orientation and epitaxial relationship between the film and
substrate is important for understanding the alignment and growth of functional oxides on
the wide bandgap semiconductors.
4.2.6
Transmission Electron Microscopy and Energy Dispersive X-ray Spectroscopy
TEM and EDX analysis were performed using JEOL 3000F (S) TEM at 300keV to
investigate the interface structure and chemistry of BaM/SiC, and thus help to understand
the function of the MgO templates. Cross-sectional TEM specimens were prepared using
conventional grinding and polishing methods followed by low angle Ar ion milling in order
to achieve specimen electron transparency. With TEM, both diffraction patterns and images
of the sample can be obtained. Such microstructural defects as dislocations, grain
boundaries, twin structures, and antiphase boundaries can be revealed.
4.2.7
Vibrating Sample Magnetometry
Static (DC) magnetic properties were measured using the vibrating sample
magnetometry (VSM). The principle of operation for the VSM is the detection of an
induced dipole field from an oscillating magnetic sample. The sample is placed in a uniform
magnetic field provided by a permanent magnet or electromagnet and vibrated vertically by
a loudspeaker cone.
The shape of the hysteresis loop is determined by the orientation of magnetic
domains and the exchange energy within the material. From the hysteresis curve, the
118
saturation magnetization (4πMs), the remnant moment (Mr), and the coercive field (Hc),
can be determined. For thin films, the paramagnetic or diamagnetic substrate contribution
was subtracted from the measured hysteresis loop.
4.2.8
Ferromagnetic Resonance
The ac magnetic properties as the most important parameters of the film for
microwave device application are measured by using the ferromagnetic resonance (FMR)
technique. Typically, the BaM film is placed in a shorted waveguide and the differential
absorption signal is detected to give a FMR linewidth. However, due to the complex
structure of BaM/SiC and semiconducting nature of SiC substrates, the resonant method
was employed to improve the signal/noise ratio.
119
5.0 RESULTS AND DISCUSSION
All oxide films used in this study were deposited by PLD or MBE technique. By
engineering a more effective interface and optimizing the growth parameters, high quality
magnetic ferrite films (i.e., BaM and Fe3O4) have been successfully integrated with wide
band semiconductor (6H-SiC) and have the potential for next-generation spintronics and
microwave device applications. There are three major sections in this chapter: interface
engineering of BaM on 6H-SiC by PLD, integration of BaM with 6H-SiC by MBE, and
integration of Fe3O4 with 6H-SiC by MBE.
5.1
Interface Engineering of BaM on 6H-SiC by PLD
Barium hexaferrite (BaM) has a complicated magnetoplumbite structure and can
coexist with quite a few different compounds, such as BaFe2O4 (BaO·Fe2O3), Ba2Fe2O5
(2BaO·Fe2O3), etc. [146,147]. Molecular beam epitaxy has never been utilized to grow BaM
on any substrates. In order to understand how the substrate surface, temperature, and O2
pressure impact the integration of BaM with wide bandgap semiconductor 6H-SiC, PLD
was used first in this research since PLD is relatively a mature technique for complex oxide
film deposition and has been demonstrated successfully to grow BaM on substrates as MgO,
Al2O3, Si/SiO2, and GGG [54, 76, 77, 78].
Three generations of BaM films on 6H-SiC have been deposited by PLD in this
work. Through three generations of interface engineering using template, from none [11], to
a PLD deposited MgO/BaM interwoven structure [140], to a MBE-grown 10 nm template of
MgO [141], the magnetic quality of the BaM improved as measured by VSM and FMR. The
120
descriptions of the various generations of BaM film samples in this PLD section are listed in
Table 8.
Table 8: Three generations of PLD deposited BaM films with various templates
Template
Gen I [11]
Gen II [140]
Gen III [141]
None
16nm interwoven
BaM/MgO layers grown
by PLD
10nm MgO template
grown by MBE
Anisotropy
kOe
16.0
16.3
4πMs*
kOe
4.2
4.3
FMR Linewidth
Oe
>2000
500
16.9
4.4
220
* The film magnetization (4πMs) is normalized to volume, where the film thickness is estimated by a stylus
profilometer, which gives an error of approximately 10%.
5.1.1
Generation I and Generation II BaM Films Grown by PLD
The first generation (Gen I) BaM films were grown on SiC without using any
template, while the second generation (Gen II) BaM films were grown on SiC with
MgO/BaM interwoven multilayers. Film deposition of both generations was carried out by
PLD with a KrF Excimer laser operating at a wavelength of 248 nm. The substrate
temperature was maintained at 925 ºC and the O2 pressure was kept constant at 20 mTorr.
Two targets (MgO and BaM) were used and the final BaM films are typically 0.3 ~ 0.5 m
thick.
The interworven multilayers consist of alternating MgO and BaM layers,
represented by, [MgO(An)/BaM(Bn)]n, where An is the thickness of the nth MgO layer, Bn is
the thickness of the nth BaM layer, and n is the number of repeated pairs of alternating layers.
Each layer thickness is controlled by the number of laser pulses or shots incident upon each
target, where a single laser shot provides ~0.01 nm of MgO or BaM. The first interwoven
layer is the thickest at A1 = 80 shots MgO B1 = 150 shots BaM. Each subsequent layer has 10
121
less MgO shots than the layer before. Thus the 8th layer is 10 shots of MgO followed by pure
BaM deposition until a 0.3 ~ 0.5 m thick BaM layer is grown.
In order to understand the formation of BaM and the function of interwoven MgO
layers at the BaM/SiC interface, three different intermediate stages of film deposition were
analyzed. The first stage consisted of the first layer of MgO and BaM
([MgO(A1)/BaM(B1)]1), the second stage of deposition consisted of 4 layers MgO
interwoven with 4 layers BaM ([MgO(A1)/BaM(B1)]1 through [MgO(A4)/BaM(B4)]4), and
the third stage consisted of all 8 layers of alternating MgO/BaM and an additional 400 shots
of BaM. This final structure with approximate thicknesses is shown in Figure 42.
3rd , ~20nm
BaM
2nd, ~9nm
MgO
1st, ~2nm
6H-SiC
Figure 42: Schematic diagram of three intermediate stages barium ferrite PLD
deposition process, the estimated thicknesses are around 2nm, 9nm, and 20nm
respectively.
BaM films deposited on interwoven layers of MgO/BaM (Gen II) demonstrated
improved magnetic properties. As shown in Figure 43a, the VSM measurement confirmed
that there is some random orientation of crystallites or defects that exist in the Gen I films,
as shown from high Hc in both perpendicular and parallel directions. It is hypothesized that
the large lattice mismatch (4.38%) that exists between BaM and SiC, leads to defects
122
(off-stoichiometric material, dislocations, and vacancies) at the interface, which produces a
lower quality BaM film. Thus suitable interwoven layers may be required to mediate
interface stress and reduce the density of defects at the interface, which can promote higher
quality BaM films. As shown in Figure 43b, the film grown on the MgO/BaM interwoven
layers showed most crystals are c-axis perpendicular aligned, which results in a BaM film
with higher Ms and lower coercivity. Table 9 listed the detailed magnetic properties of Gen I
and Gen II films.
4000
(a)
(b)
Perpendicular
In plane
4M (Gauss)
2000
0
-2000
-4000
-15000 -10000 -5000
0
5000 10000 15000
Applied Field (Oe)
Figure 43: VSM hystersis loops of (a) Gen I, and (b) Gen II BaM films grown by PLD;
Gen II films showed more c-axis perpendicular alignments, resulting in lower
coercivity in both perpendicular and in plane directions.
Table 9: Magnetic property comparison of BaM films grown on SiC with and without
MgO interwoven layers
Sample ID
Direction
Ms (kG)
Mr (kG)
Hc (Oe)
S (Mr/Ms)
Gen I (Without MgO
⊥
4.2
1.7
1240
0.4
interwoven layers)
//
3.3
0.7
1030
0.2
Gen II (With MgO
⊥
4.3
1.3
950
0.3
interwoven layers)
//
3.3
0.4
390
0.1
123
Gen II films also showed improved surface morphologies compared to Gen I films.
SEM and AFM images (Figure 44a and b) of a Gen I film showed some c-axis perpendicular
orientation. Large hexagonal crystals with typical size of 0.5 m were clearly observed, but
the crystals were not uniform, and contributed to a non-continuous, porous, and
polycrystalline BaM film. With the MgO interwoven layers, as shown in Figure 44c and d, a
Gen II film consisted of hexagonal crystals that appeared to be more continuous with c-axis
perpendicular aligned, which resulted in a BaM film with higher Ms and lower coercivity, as
shown in Table 9.
(a)
(b)
1 μm
(c)
400nm
(d)
1m
Figure 44: SEM and AFM images of Gen I BaM film (a & b) and Gen II BaM film (c
& d) grown by PLD on 6H-SiC.
XRD analysis confirmed that Gen II BaM films have improved c-axis orientation
124
compared to the Gen I films under the same processing conditions, as revealed by the
(0,0,2n) diffraction peaks shown in Figure 45. The diffractions peaks of Gen I film showed
the presence of strong peaks assigned to random orientation (106) and an impurity phase of
Fe2O3 (Figure 45a). In addition, the intensity of the main peak (008) was much lower than
the Gen II films.
3000
008
Intensity (arbitrary units)
2500
SiC 0006
(b) Gen II
BaM_SiC with
interwoven MgO
006
2000
0014
1500
SiC 0007
0010
1000
Fe2O3(012) 106
500
SiC 0008
(a) Gen I
BaM_SiC without
interwoven MgO
107
0
20
30
40
50
60
2θ
Figure 45: X-ray θ-2θ diffraction pattern for BaM films grown by PLD on (a) 6H-SiC,
and (b) 6H-SiC with MgO interwoven template.
XPS analysis of the two films indicated a small silicon peak in the Gen I BaM film
deposited directly on SiC but not in the Gen II BaM film deposited on the MgO/BaM
interwoven layers. The typical O1s spectra for BaM films grown directly on SiC (Figure
46a), showed a shoulder shifted 2.1 ± 0.1 eV to a higher binding energy relative the lattice
125
O1s peak (~530 eV) [148], which suggests the formation of Si-Ox bonding in the film. For
films grown with interwoven MgO/BaM layers, there was no sign of Si-Ox bonding, as
illustrated in Figure 46b. This suggests silicon diffusion from the substrate into the BaM
film directly deposited on the SiC substrate. However, with the use of interwoven layers, the
silicon diffusion all the way into the film was prevented.
O 1s Photoemission Intensity (arb.units)
Lattice O
Si-OX
536
534
532
530
528
526
Binding Energy (eV)
Figure 46: XPS O 1s photoemission spectra obtained with Al k X-rays for BaM films
grown on 6H-SiC (a) without MgO interwoven layer (b) with MgO interwoven layers
To evaluate the degree of silicon diffusion, XPS depth profiles were performed on
both films. Figure 47 compares the atomic concentrations of each film as a function of etch
depth. Note that the depth is an equivalent depth based on etching rate of SiO2. Both films
showed over-oxidation and Fe deficiency in the BaM surface upon exposure to air, which is
consistent with the results reported by Kamzin et al [149]. The ratio of Fe/Ba at the surface
is around 6 and increases with the etching. After 40-50 mins etch, the Fe/Ba ratio reaches a
value close to the stoicheometry of 12. The Gen I film grown directly on SiC (Figure 47a)
126
indicated Si diffusion throughout the film, which resulted in non-stoicheometric Fe/Ba
around 9, as determined by XPS. An intermixing layer of Ba, Fe, O, and Si existed between
the film and the substrate. According to Tang et al [150], the reaction between Fe and SiC is
thermodynamically favored at 925oC and might produce Fe-Si bond at the interface.
However, XPS analysis did not show Si-Fe bonds instead of only Si-O bonds from Si 2p
peak in a thin BaM film (~ 1.3 nm) grown directly on SiC. Since oxygen has an
electronegativity value approximately twice that of those values observed for iron and
silicon [ 151 ], Si-O bonds are more favored than Si-Fe bonds [138, 152 ]. Check the
thermodynamic calculations; the reaction between the O2 and SiC is feasible at 925oC with a
△G of ~ -850 kJ/mol. This reaction can result in the formation of Si-Ox bond and drive Si
diffusion into the film, which will be further confirmed in Figure 48e. However, Gen II
films deposited on the MgO/BaM interwoven layers show a much sharper interface with no
evidence of Si diffusion all the way into the film.
127
60
(a)
Ba
Fe
O
Si
C
Composition %
50
40
30
20
10
0
0
50
100
150
200
250
300
350
400
450
500
O
Mg
Si
C
Depth (nm)
(b)
60
Ba
Fe
Composition %
50
40
30
20
10
0
0
50
100
150
200
250
300
350
Depth (nm)
Figure 47: XPS depth profile of BaM films grown by PLD on (a) 6H-SiC, and (b)
6H-SiC with MgO modified template showed the effectiveness of the MgO templates
in preventing the Si diffusion.
In order to fully understand the formation of BaM and the function of MgO/BaM
interwoven layers at the interface, three staged samples were studied by XPS, SEM and
AFM. Figure 48 shows the evolution of the Si 2p XPS peak shape for thin films grown at
different stages compared to the ~0.4 m thick films deposited with and without the
MgO/BaM interwoven layers. The high binding energy side Si 2p peak represents Si-Ox,
128
which suggests the SiC substrate was oxidized at the interface during the first layers of
deposition, as seen in Figure 48a. Since the estimated film thicknesses (~2 nm) was within
the Si 2p photoelectron attenuation depth (~8 nm) in XPS analysis [148], the right side peak
shifted ~ 1.5 eV to the lower binding energy is assigned to the Si-C bond from the substrate.
With the use of MgO/BaM interwoven layers, the intensity of Si-Ox peak decreases at
increased film thickness from Figure 48a to c, and is not evident in the thick film, Figure 48d.
Since the Si-Ox signal is still evident in the 3rd stage BaM film with thickness thicker than
the Si 2p attenuation depth of ~ 8nm, the Si is diffused into the BaM film at the initial stages
even with the protection of the MgO/BaM interwoven layers. However, compared to the
BaM film grown on SiC directly, the interwoven layers trapped the diffusing Si and
prevented the migration of Si all the way in the film. It is worthwhile to notice that the Fe/Ba
ratios at the initial stages are around 5 to 6, suggesting the formation of some other Ba-Fe-O
compounds than just BaM at the starting growth. Since the film thicknesses at the initial
growth are too thin for XRD, it is hard to indentify the impurity phases.
129
(e) BaM on SiC w/o MgO
Si 2p Intensity (Arb.Units)
(d) BaM on SiC w/ MgO
(c) 3rd stage
(b) 2nd stage
Si-Ox
Si-C
(a) 1st stage
108
106
104
102
100
98
96
Binding Energy(eV)
Figure 48: XPS elemental scan of Si2p for different stage growth. (a) 1st, (b) 2nd, (c) 3rd,
(d) thick BaM on SiC with MgO interwoven layers, and (e) thick BaM on SiC without
MgO interwoven layers.
Figure 49a-c shows the SEM images of the progression of MgO/BaM interwoven
layers for the 1st, 2nd, and 3rd stages. The evolution of the grain size and shape were
compared for films grown with the same number of respective BaM shots but no MgO shots
(Figure 49d-f). After the first stage, illustrated in Figure 49a and d, the grain sizes were in
the range of 0.05 - 0.1 m, with no obvious difference resulting from the use of MgO in the
interwoven layer. However, after stage 2, the dimensions of grains with the MgO/BaM
interwoven layer were larger and more hexagonal or triangular shape compared to the grains
without MgO. This may suggest that the integration of MgO may promote preferred BaM
130
c-axis orientation with an increase in two-dimensional spread rate and a decrease in
three-dimensional nucleation rate. After the third stage growth, the grain size with MgO
interwoven increased up to 0.5m, and the grains took on more hexagonal character with
sharp edges. However, the grain size of BaM film without MgO was around 0.2 m with
more disorientation. This supports the proposed function of the MgO interwoven layers to
improve c-axis perpendicular orientation and two-dimensional growth mode.
(a)
(d)
1 μm
1 μm
(b)
(e)
1 μm
(c)
1 μm
(f)
1 μm
1 μm
Figure 49: SEM images of different stage growth (a - c) with MgO interwoven layers,
(d - f) without MgO interwoven layers.
131
AFM characterization was used to study the growth mechanism by evaluating the
morphology of the grains. Characterization of a crystalline BaM film deposited on the third
stage of the MgO/BaM interwoven layers showed a typical hexagonal structure with
indications of 2-D, layer-by-layer growth mode. A layered structure was clearly observed on
the hexagonal face of the crystallites. The height of each layer was measured to be around 2
nm (Figure 50a), which is close to the 2.32 nm c-axis dimension of the BaM unit cell [45].
However, the grains that formed on the third stage with no MgO showed more random
orientation and smaller dimensions (Figure 50b). From these observations, a possible
mechanism of the BaM film grown on SiC with the integration of the MgO/BaM
interwoven layers can be proposed. During BaM growth on the MgO/BaM interwoven layer,
the BaM nucleates on the surface and grows along the substrate surface faster than in the
direction perpendicular to the SiC substrate. Since the growth rate of BaM in the a-b
direction exceeds the growth rate in the c-direction [153], the growth mode of BaM
promotes 2-D growth of thin, hexagonal plates parallel to the substrate surface with c-axis
orientated perpendicular to the substrate (Figure 50a).
132
(a)
(c)
200nm
200nm
Figure 50: (a) AFM image of the 3rd stage BaM growth on SiC with MgO interwoven
layers, (b) Profile along the lines in (a) , (c) AFM image of the 3rd stage BaM growth on
SiC without MgO interwoven layers.
Both chemical and structural differences are observed in the evolution of Gen I BaM
films deposited by PLD on bare 6H-SiC substrates compared to Gen II BaM films deposited
on MgO/BaM interwoven layers on 6H-SiC substrates. XPS analysis shows that the use of
the MgO/BaM interwoven layers in the early stages of film growth traps diffusing Si and
prevents the migration of Si all the way into the film that was evident in films deposited
directly on SiC. XPS also shows evidence of Si-Ox bonding at the initial stages of growth
that decrease through the MgO/BaM interwoven layers. SEM shows improved
crystallographic texture on the MgO/BaM interwoven layer during the initial stages of
growth that lead to improved texture in the final 0.3 to 0.5 μm thick films. AFM results
suggest that the MgO/BaM interwoven layer promotes increased growth rate of BaM in the
a-b direction producing larger, two-dimensional crystallites in earlier stages of growth. By
comparing the chemical and structural evolution of PLD deposited BaM on bare 6H-SiC
133
and on MgO/BaM interwoven layers on 6H-SiC, it is concluded that the interwoven layers
effectively eliminate silicon diffusion into the bulk BaM film. In addition, the initial stages
of film growth on the MgO/BaM interwoven layers have a greater number of c-axis
perpendicularly aligned crystals, and larger crystals than films deposited with the same
number of BaM shots on a bare SiC substrate. This suggests that the interwoven layers also
act to relieve strain between the SiC substrate and the BaM film to promote aligned,
two-dimensional growth more quickly.
Although the MgO/BaM interwoven layers addressed the interface mixing and
lattice mismatch quite well, the FMR linewidth of 500 Oe does not meet the device
requirement of an FMR linewidth less than 150 Oe [154]. Post-deposition heat treatments
were found to be ineffective in further reducing the FMR linewidth. In addition, the
deposition process for interwoven layers is quite complex and not practical for real device
manufacture. In next section, the combined use of MBE and PLD technique to further
improve the BaM film quality will be discussed.
5.1.2
Generation III BaM Film on SiC by PLD with MBE Grown MgO
The third generation (Gen III) BaM was grown on SiC by PLD with a MBE-grown
MgO template. The combined use of MBE to grow a high quality MgO template followed
by the growth of a thick film of BaM both techniques is proven to be a simple and successful
method to allow effective integration of BaM with SiC for microwave device applications.
This approach is shown to result in device quality epitaxial films with FMR linewidths of
less than 100 Oe.
134
5.1.2.1 MgO Template Grown by MBE
Dr. Goodrich has explored the growth and characterization of MgO (111) on 6H-SiC
in his Ph.D study [73]. Before MgO film growth, the 6H-SiC substrate was first degreased
and then cleaned in a custom-built hydrogen furnace. For more details of the H2 cleaning
procedure, refer to the experimental section. In Dr. Goodrich’s work, such variables as
cleaning temperature, time of H2 exposure, and H2 flow rate have been studied and
optimized to produce an atomically smooth stepped SiC surface for subsequent MgO
growth. The contact between the SiC and Ta boat and the cooling rate are very crucial to
determine the chemistry of the SiC substrate. In order to make sure good contact between
the SiC and Ta boat, the SiC substrate size is limited to 0.7× 0.7 cm2 because of the
limitation of the H2 furnace set up. The heating rate was fixed at ~5A every 30 second to
minimize the deformation of Ta boat. A fast heating may cause the sudden deformation of
the Ta boat, and thus cause the sample to move away from the center of the Ta boat. The
movement will result in bad contact, non-uniform heating, and rough surface. However, for
different batch SiC wafers purchased from Cree., the optimum cooling rate used was
different. It was found out in later research that slow cooling (~2.5 A every 30 second) did
not work well for some batch SiC wafers, the cleaned SiC end up with high O concentration
(>10%, determined by XPS), and fainter ( 3 × 3 ) R30° reconstruction (determined by
RHEED). In order to obtain good cleaning for such wafers, fast cooling rate with the sudden
removal of current was required. The reason for the non-effective cleaning from the slow
cooling is not very clear at this point.
The SiC surface after H2 cleaning is free of surface contamination and polishing
135
scratches, and terminated by a silicate adlayer [138]. Once good cleaning (<10% O
composition, sharp surface reconstruction, and smooth surface) is obtained, high quality,
single crystalline MgO can be grown on this H2 cleaned SiC. Dr. Goodrich reported the
MgO growth has a magnesium adsorption controlled growth mechanism under our growth
condition [73] and the surface structure of the MgO films consisted of an OH stabilized (1×1)
structure, which has even smaller surface energy, and therefore is more stable than MgO
(100) [73]. Both XPS and RHEED show the amount of OH decreases for smoother MgO
film (less 3-D features), which is grown at much higher temperatures (i.e. 650oC) [73]. The
current work shows that the O composition of the H2 cleaned SiC has a significant impact on
the MgO film quality. Figure 51 shows XPS survey scans and RHEED patterns along
<11 2 0> of two H2 cleaned 6H-SiC samples with different oxygen compositions. One
sample has 8% oxygen on the surface and the other has 14% oxygen after cleaning as
determined by XPS. RHEED of SiC with 8% show very sharp and crisp ( 3 × 3 ) R30°
reconstruction pattern, as determined from the presence of two intermediate rings
designated L1/3 and L2/3 [73]. However, the reconstruction pattern is much fainter for SiC
with 14% O, without showing two clear intermediate rings. Higher O composition than 10%
means unsuccessful cleaning, which can be resulted from bad contact with Ta, sample
movement during cleaning, or slow cooling rate.
136
14% O
8% O
Si 2s
C 1s
Si 2p
Intensity (Arb. Units)
O 1s
(b)
(a)
800
600
400
200
0
Binding Energy (eV)
Figure 51: XPS survey scans of H2 cleaned 6H-SiC: (a) with 8% O; (b) with 14% O;
insets are the corresponding RHEED patterns along <11 2 0>. The L1/3 and L2/3 rings
from reconstruction are not clear for SiC with 14% O.
On these two SiC substrates with different O compositions, two MgO films (5-6 nm
thick) were prepared respectively at the same growth conditions. XPS characterization of
the ratio of OH to MgO, as indicated by the O 1s peak (Figure 52), indicated an OH: MgO
ratio of 0.20 for the film grown on the SiC with 14% O and 0.10 for the film grown on the
SiC with 8% O. The significant decrease in the relative amount of OH is mostly likely due to
higher quality film with smoother surface and less defects. Cross-sectional TEM
characterization was performed through collaboration with Dr. Lazarov at York University
to further investigate the MgO/6H-SiC interface and the MgO film quality.
137
O 1s Photoelectron Intensity (Arb. Units)
O-Mg
H-O-Mg
MgO grown
on 14% SiC
(b)
MgO grown
on 8% SiC
538
536
(a)
534
532
530
528
526
524
Binding Energy (eV)
Figure 52: XPS tight scans of the O 1s spectra for MgO films grown on H2 cleaned
6H-SiC: (a) with 8% O; (b) with 14% O.
Figure 53 shows the high resolution interface lattice images of two MgO films
grown on H2 cleaned 6H-SiC with 8% O and 14% O respectively. TEM of MgO grown on
low O SiC (Figure 53a) has shown a sharp MgO/SiC interface and a uniform MgO film of 6
nm thick without evidence of the SiOx formation and columnar growth. No grains with
other orientations or big crystal defects is found, implying the high quality of the (111)
oriented MgO film grown on low O SiC substrate. However, TEM of the MgO grown on a
high O SiC (Figure 53b) shows very rough SiC surface and the existence of a ~1 nm
amorphous Si-Ox layer at the MgO/SiC interface. Some surface facets and small grain size
can also be observed in this MgO film. The excess O on the SiC surface after H2 will form
other bonds (Si-Ox or C-Ox) than just the silicate adlayer (Si3O5) [73], and thus cause the
partially reconstructed and rough SiC surface, as indicated by the fainter reconstructed rings
138
from RHEED (inset of Figure 51b). The non uniform and rough starting surface can cause
the crystal defects and grains with different orientations in the subsequent MgO film.
Therefore, in order to grow high quality single crystal MgO film with low defects, the O
composition of the starting SiC surface should be controlled at not higher than 10%.
(a)
Mg
SiC
(b)
Mg
SiO
SiC
Figure 53: Cross sectional HRTEM images of the MgO films deposited at 150oC under
the oxygen pressure of 5×10-6 Torr on: (a) 6H-SiC with 8% O; (b) 6H-SiC with 14% O
after H2 cleaning.
5.1.2.2 The impact of O2 pressure on the Gen III BaM quality
The oxygen environment during PLD deposition plays a crucial role on the
stoichiometry, morphology, and thus magnetic properties of the BaM film because the O2
139
environment pressure changes the chemistry and reactivity of the barium ferrite plumes.
The optimal O2 pressures for growing highly c-axis oriented, single crystalline BaM film by
PLD are different as reported by different groups. Shinde et al. reported 300 mTorr was the
optimal value for growing BaM on Sapphire [51] and Liu et al. maintained oxygen pressure
of 200 mTorr to 250 mTorr for BaM growth on Si [102, 155]. Yoon et al., however, reported
20 mTorr as the optimal oxygen pressure for growing BaM on MgO [77]. The difference for
the reported optimal O2 pressure may be attributed to the different laser energy intensity,
chamber geometry, and the substrate used by different groups.
For Gen III BaM films grown on 10nm MBE-grown MgO templates on SiC, the
effect of O2 pressure was explored at growth temperatures of 915ºC. It was found that when
BaM was grown on 10 nm MBE-grown MgO/SiC in vacuum with no oxygen present, the
spinel Fe3O4 (111) oriented structure was obtained, as determined by XRD. This suggests
the coexistence of Fe2+ and Fe3+ ions in the oxygen deficient environment. However, excess
oxygen will cause misorientations of the BaM film. XRD pattern (Figure 54) confirmed that
the existence of the in plane orientation, shown as the (106) peaks for films grown at high
pressure, i.e., 100 mTorr and 200 mTorr. In comparison, film grown at 20 mTorr only
exhibited (0, 0, 2n) peaks, suggesting highly c-axis orientation. The optimum oxygen
pressure of 20 mTorr to deposit the BaM films on a 10 nm MBE-grown MgO template with
desired properties is the same as the optimum value for a film grown on bare SiC substrate
[11] and bare MgO (111) substrate [77], suggesting that the substrate may not be a crucial
factor to determine the value of optimal pressure but rather the laser system and the chamber
geometry is what determines optimal O2 pressure. More specifically, the overriding impact
140
is from balance of the laser energy and O2 pressure that is sensitive to any kind of
maintenance, outage, etc. Therefore, more controlled environment like MBE system may
help better understanding the growth and nucleation mechanisms of BaM, which will be
discussed in the later sections.
008
Intensity (Arb. Units)
006
(c)
106
*
SiC 006
20 mTorr
100 mTorr
200 mTorr
SiC 007
0010
0012
40
50
0014
(b)
(a)
20
30
60
2 Theta (degree)
Figure 54: XRD patterns for Gen III BaM films grown by PLD with 10nm
MBE-grown MgO at (a) 20mTorr, (b) 100mTorr, and (c) 200mTorr ; increased
background O2 pressure in the PLD chamber deteriorated the dc magnetic
characteristics of the deposited BaM film.
The Fe/Ba ratio decreased from 11.1 for film grown at 20 mTorr to 10.0 for film
grown at 100 mTorr and finally to 7.5 for film grown at 200 mTorr. This decrease in Fe/Ba
ratio correlated to a decrease in magnetic properties and a noticeable impact on the surface
morphology. The hysteresis loops and AFM images are shown in Figure 55 and Figure 56 as
a function of oxygen pressure and with the corresponding Fe/Ba ratio. The hysteresis loop
141
showed highly c-axis oriented perpendicular to the substrate for films grown at 20 mTorr.
With the increase of oxygen pressure, the hysteresis loop showed deteriorated perpendicular
orientation and an increase in-plane orientation. No apparent correlation was found in the
value of perpendicular coercivity (Hc) as the oxygen pressure (Po2) was increased, but the
value of in-plane Hc decreases systematically as a function of Po2. The films grown at 200
mTorr still maintained a predominantly easy axis normal to the film plane and a hard
direction in–plane. However for these films grown at high Po2, the in-plane magnetization
curve was much broader with an in-plane coercivity of 1200 Oe, suggesting misaligned
crystal grains and/or the existence of a secondary phase.
Figure 55: Hysteresis loops for Gen III BaM films grown by PLD with 10 nm
MBE-grown MgO at (a) 20 mTorr, (b) 100 mTorr, and (c) 200 mTorr ; increased
background O2 pressure in the PLD chamber deteriorated the dc magnetic
characteristics of the deposited BaM film.
AFM image of the BaM films grown on 10nm MBE grown MgO template at 20
mTorr, shown in Figure 56a, reveals a smooth surface with hexagonal and triangular
patterns, which suggests strong c-axis orientations of the BaM crystals. However, AFM
images of films grown at 100 mTorr and 200 mTorr, shown in Figure 56b and c exhibited
acicular crystals with sizes ranging from 0.4 m to 0.8 m oriented in random directions
over the entire surface of the films.
142
Figure 56: AFM pattern for Gen III BaM films grown by PLD with 10 nm
MBE-grown MgO at (a) 20 mTorr, (b) 100 mTorr, and (c) 200 mTorr show a
deterioration of desirable, smooth morphology with the increase in PLD chamber O2
pressure.
Papakonstantinou et al. reported only neutral Fe and Ba, and singly ionized Ba
species were emitted from laser ablation of a BaM target without evidence of oxidized
species in the gas phase such as BaO and FeO under an O2 environment of 0.05 to 0.5
mbar (66.5 mTorr to 665 mTorr) [156]. Through reactive collisions, the oxygen in the
background pressure of the PLD chamber changes the chemistry of the BaM plume by
oxidizing the highly energetic atomic species to oxides with much lower energy, thus
preventing surface damage by high-energy species. However, the velocities of the atoms
and ions at high O2 pressure can become so slow that there will not be enough surface
activation for c-axis only oriented growth, and disoriented grains and rough surfaces
result.
5.1.2.3 The Impact of MBE-grown MgO Thickness
While BaM films deposited by PLD on a 10 nm thick MBE-grown MgO template
in SiC at O2 pressure of 20 mTorr and 915oC showed all of the grains oriented, further
experimentation showed that the thickness of the MBE-grown MgO impacts the
orientation of the subsequently deposited BaM. BaM films grown on 4 nm MgO template
143
had similar XRD pattern to Gen II films, with most peaks indexed to (0,0,2n) diffraction
peaks, but a small (107) peak showing the presence of some random orientation (Figure
57a). However, for Gen III BaM films grown on 10 nm MgO template, the XRD pattern
revealed no existence of the (107) peak combined with strong (0,0,2n) diffraction peaks,
which suggest that the films have strong c-axis orientation perpendicular to the substrate
plane. Comparing the inset AFM patterns for the corresponding diffraction patterns, the
film grown on the 4 nm MgO template was significantly rougher over the same scan area,
and showed apparent misalignment of crystal grains, consistent with the XRD result that
showed some random orientation peaks.
Figure 57: The thicker MgO single crystal film enabled a more aligned BaM film with
no sign of misoriented (107) peak as seen in the X-ray θ-2θ diffraction patterns of
BaM films grown by PLD (a) with 4 nm MgO grown at 150 oC by MBE, inset AFM
showed evidence of misaligned crystals, and (b) with 10nm MgO grown 150 oC by
MBE. * designated the second phase of spinel structure.
144
Because of the random orientations and misalignment, the BaM film grown on a 4
nm MBE-grown MgO template showed much worse magnetic properties, shown as higher
coercivity and opener loop at the in plane direction than the film grown on a 10 nm MgO
(Figure 58). For BaM film grown on 10 nm MgO, the anisotropy field is estimated as the
linear extrapolation of the hard loop to the saturation region of the easy loop and is
approximated to be 16.9 ± 0.2 kOe, which is in agreement with the published results of the
single crystal BaFe12O19 value of 17 kOe. In addition, the 4.4 ± 0.3 kG magnetization (4Ms)
of the film compares well to the bulk material value of 4.48 kG. However, for BaM film
grown on 4 nm MgO, the estimated anisotropy field is around 16.0 kOe and the 4Ms is
only 4.0 ± 0.3 kG, suggesting deteriorated magnetic properties.
4000
(a)
(b)
4M (G)
2000
0
-2000
-18
In plane
Perpendicular
In plane
Perpendicular
-4000
-12
-6
0
6
Applied Field (kOe)
12
18-18
-12
-6
0
6
12
18
Applied Field (kOe)
Figure 58: Hysteresis loop of BaM grown on (a) 10nm, and (b) 4 nm MBE-grown
MgO/SiC obtained by VSM with a maximum applied field of 12,500 Oe aligned
parallel (open square) and perpendicular (solid square) to the film plane.
As both the 10 nm and 4 nm films are single crystalline (111) oriented MgO smooth
films, this suggests that either the template is degrading with time and allowing diffusion
or other defects to form which are hidden underneath the bulk of the BaM film, or the
145
surface of the thick MgO is different from the surface of the thin MgO in such a way that
minimizes defects in the BaM film and hence improves its quality. Compared to the 4 nm
MgO film, the 10 nm MgO film shows an increase of three dimensional characteristics in
the RHEED pattern from formation of MgO islands. Original cross-sectional STEM
analysis illustrated the existence of columnar grains after an MgO film thickness greater
than 3 nm. Additional exploration of the MgO/6H-SiC interface has shown that the MgO
films can be as thick as 10 nm without evidence of columnar growth. AFM
characterization of the MgO films with thickness ranging from 2 nm to 12 nm showed no
measurable difference in surface roughness. Thus, the most likely hypothesis for the 4 nm
MgO template not working as well as 10 nm MgO template is that the MgO template
degrades during the subsequent deposition of BaM. Hence XPS depth profiles and TEM
analysis were used to examine the interface between the SiC and the BaM film.
XPS depth profiles were performed on identically deposited BaM films on a 4 nm
MBE-grown MgO/6H-SiC template and a 10 nm MBE-grown MgO/6H-SiC template. Both
BaM films were deposited under the same PLD processing conditions and processing time.
As published previously, films grown without the MgO template showed Si diffusion
throughout the film, and a > 300 nm intermixing layer between the film and the substrate.
The XPS depth profiles in Figure 59 show that the Si diffusion to the surface was prevented
by using the MgO templates of both the 4 nm and 10 nm thicknesses. Both film surfaces
showed the expected over-oxidation and Fe deficiency in the BaM surface upon exposure to
air per Kamzin et al [149]. However, the shape and width of the interface region is different
for each film, with the 10 nm MgO template (Figure 59a) showing a sharper interface and
146
more limited Si diffusion. The interface region is defined as the depth at which Mg appears
to the depth at which only Si and C are seen. It is noted that this is the broadest possible
definition of the interface region, as the Ar+ etching is known to create roughness in the
surface and to favor removal of lighter elements over heavier elements. For the 10 nm
template, Mg appeared at a depth of around 460 minutes of etch time, forming the boundary
between the BaM film and beginning of the interface layer. The width of the interface region
from this point to only Si and C remaining on the surface is 220 minutes of etch time, which
is approximately 110 nm, based on the etch rate of SiO2. Comparing Figure 59a with b, the
BaM grown on 4 nm MgO template showed a much broader interface region of
approximately 140 nm between the BaM and SiC. As the magnetic properties are
normalized to volume, this may have contributed to the deteriorated magnetic properties as
the presence of this broad interface was not known and hence was not taken into account.
147
Figure 59: XPS depth profile of BaM films grown by PLD on (a) 10 nm MgO
(111)/SiC(0001) template, and (b) 4 nm MgO(111)/SiC(0001) template. This
comparison shows a sharper interface and limited Si diffusion with the 10 nm MgO
template.
For the BaM film deposited on the 10 nm MgO template, the Si and C signal
appeared at the same sample time in the depth profile. However, the Si/C ratio from XPS at
this sample time is around 2.0, suggesting preferential Si diffusion at the interface, although
preferential etching of carbon could also contribute to an increase Si/C ratio. Considering
148
the approximately 10 nm etch rate for the data collection time interval and the sampling
depth of 8 nm and 7 nm for Si and C through MgO respectively, the SiOx formed at the
MgO/SiC interface as silicon diffuses is less than 17 nm thick. However, for the BaM film
grown on the 4 nm MgO template, the Si signal appeared more than 60 minutes earlier than
the appearance of the C signal, showing approximately 30 nm of mixed interface with SiOx
and implying significantly greater silicon diffusion than with the 10 nm MgO template, even
with an effect from preferential etching of the carbon. These results confirmed that the 10
nm MgO minimizes interface breakdown and mixing more efficiently than the 4 nm MgO
template. However, the 10 nm MgO interface does show evidence of breakdown, suggesting
a kinetically driven breakdown mechanism at the growing film/MgO template interface
during PLD. This is supported by TEM data discussed later.
The XPS depth profiles also showed that a small amount of diffused Mg segregated
on the surface of each film. However, there was no evidence of Mg diffusion throughout the
“bulk” of the film, since no Mg peak was discernible from signal-to-noise after 1 min Ar+
etch. Based on the 2 nm sampling depth for Mg 1s electrons through BaM, the content of
Mg if assumed distributed through the top layers of film was estimated to be 3-4 atomic %
by XPS.
The interface mixing may be driven by the known interdiffusion between MgO and
iron oxide, which can occur at temperatures in excess of ~450 oC [120]. It is also known that
the spinel structure MgFe2O4 can be initiated at 820 oC through the reaction between MgO
and Fe2O3 [ 157 ]. Thus, the deterioration of the BaM film may be attributed to the
degradation of the MgO template at the growing BaM film/MgO template interface at the
149
growth temperature of 915 oC. As such, during the PLD deposition time, the 4 nm MgO
template is fully degraded to a point that permits further degradation of the SiC and then Si
diffusion. Hence the resulting films are non-stoichiometric with Fe:Ba ratios of less than
11, contain random orientation and a three dimensional morphology, and show a coercivity
larger than 1000 Oe. However, the BaM film deposited on the thicker 10 nm MBE-grown
MgO template during the same PLD processing time, retains enough of the MgO template at
the SiC surface to limit Si diffusion and result in a BaM film with much more desirable
chemical, structural, and magnetic properties.
To more fully study the interface mixing and avoid the preferential sputtering and
film decomposition issues that accompany Ar+ etching, cross-section TEM was used to
study the BaM/10 nm MgO/SiC interface structure. Figure 60 is a cross-sectional bright
field TEM image of the BaM film grown on the 10 nm MgO/SiC substrate. It shows that the
BaM film surface is smooth with very low roughness, consistent with the AFM results
shown in Figure 57b. A region with different contrast between the BaM and SiC is also
clearly seen in Figure 60, indicating the presence of interfacial layer(s) with different
structures than that of the substrate or the film.
150
Figure 60 (a) Bright-field cross-section TEM image of BaM thin films grown on
6H-SiC with a 10nm MgO template. A line with altered contrast between the BaM and
SiC indicates presence of interface phases. (b) SAD pattern from the substrate and
film in [11 2 0] zone axis. BaM film is single crystal and epitaxially grown on SiC.
Indeed the high resolution TEM of the interface (Figure 61) reveals two thin layers
with different structures showing the formation of transition layers between the SiC
substrate and the BaM. The first layer closest to the substrate has as amorphous structure
(Figure 61) revealed by EDX scan line (Figure 62) to be SiOx. Thus even with the 10 nm
MgO template, the SiC substrate is oxidized after 30 minutes exposure to the PLD
high-energy plume in 20 mTorr oxygen at 915 °C. The resulting SiOx with thickness varying
from 10nm to 20 nm is consistent with the XPS depth profile result (Figure 59a). Figure 61a
shows clearly a crystalline transition layer thicker than 40 nm above the SiOx amorphous
151
layer and before the crystalline BaM layer. Figure 61b shows this transition layer to be
mostly of a spinel structure, as confirmed by the digital diffractogram of this layer. The
crystallographic relation between the film and substrate is as follows: [0001](SiC) ||
[0001](BaM) and [11 2 0](SiC) || [11 2 0](BaM).
Figure 61: High resolution TEM cross sections of (a) the interface of BaM film grown
on 6H-SiC with 10nm MgO template, (b) the interface between BaM and the
transition layer MgFe2O4, insets are the digital diffractograms of the BaM (top) and
MgFe2O4 (bottom).
EDAX line scan (Figure 62) suggests the final BaM film is layered on a transition
layer that contains mixed Mg, Fe, Ba, Si, and O atoms. As shown in Figure 62, the sudden
increase of O marks the start of the interface region above the SiC. In the previous work
[158], interface breakdown between the MgO/SiC was shown to occur at processing
temperatures higher than 400°C when exposed to active oxygen, in that case an oxygen
plasma. The PLD deposition temperature of 915°C with a background oxygen pressure of
20 mTorr, seems to have enabled active O species to diffuse (potentially along MgO grain
boundaries or due to template breakdown from a reaction with iron oxides) to the substrate
152
and initiate the interface breakdown, as is consistent with the previous work [158]. The 10
nm MgO template seems to minimize the amorphous SiOx layer to less than 20 nm thick,
per TEM image (Figure 61a) and EDX (Figure 62). On top of this oxide layer, the transition
layer with mixed Mg, Fe, Ba, Si, and O atoms is observed. The formation of the transition
spinel structure is a result of the reaction between MgO template and the Fe and Ba atoms
from the BaM plume in O2 environment.
Figure 62: Chemical composition of BaM film interface region by EDX. SiOx layer is
formed closest to the SiC substrate and the final film contains thicker than 40 nm
transition layer of MgxBa1-xFe2O4.
The digital diffractograms (insets in Figure 61b) show epitaxial growth of BaM with
respect to the transition layer. Both the transition layer and BaM film are fully epitaxial with
respect to SiC substrate despite the presence of an interfacial amorphous SiOx layer (as
shown by SAD pattern in Figure 60b), which implies that the amorphous layer is forming
153
during BaM film deposition. TEM image of MgO/SiC (Figure 61) confirmed that the less
than 20nm SiOx formed during BaM growth, since the MgO/SiC interface before BaM
growth is pretty sharp with no indication of SiOx formation. It is also found out that the
formation of ~ 20 nm SiOx requires the presence of active oxygen that comes from the BaM
plume. When 4 nm MgO/SiC was heated to 915 oC in the PLD chamber at 20 mTorr for the
same amount of growth time without exposed to BaM plume, XPS only showed only 8 Å
formation of SiOx. This is consistent with the results shown by Dr. Goodrich that the
MgO/SiC interface breakdown in MBE system [73] requires unbound (atomic) oxygen
from the plasma source. Once the unbound O reaches the MgO/SiC interface through grain
boundaries of the MgO film, the SiC substrate will be oxidized to form SiOx.
Figure 61b shows the locally sharp interface between the transition spinel layer and
BaM in the final film. The average spacing, measured from the digital diffractograms, for
d0002 BaM atomic planes is 1.17 nm, which matches well to the bulk values of BaM (1.16
nm). The spacing of the (111) planes in the spinel transition layer is 0.481 nm, which is very
close to the 0.483 nm of the bulk spinel MgFe2O4 {111} planes. The selected area
diffraction patterns from the film and substrate (Figure 61b) confirm the single crystal phase
of the BaM film.
Several groups have reported that BaM films tend to grow epitaxially onto a spinel
structure [159,160,161]. Li et al. reported the formation of transition layer ZnFe2O4, when
they used a ZnO template of thicker than 20 nm to grow BaM on a Si substrate by sputtering
[159]. They confirmed the formation of ZnFe2O4 as a reaction product of BaM and ZnO,
since the transition layer grew during the ex situ annealing at the expense of ZnO and BaM,
154
as observed by HRTEM and XRD analysis. BaM has the magnetoplumbite structure that
can be viewed as stacks of cubic spinel blocks having <111> orientation (called S blocks)
and R blocks containing the barium ion [162,163]. The formation of BaM on a spinel phase
is structurally preferred due to the close relation and structural similarity between BaM and
spinel phase [159]. This also explains why the thickness of MgO layer is important. Since
the MgO template is degrading at the initial film growth due to the reaction with Fe and Ba
neutrals from the plume to form a transition layer for the subsequent BaM growth, the 4 nm
MgO template is not sufficient to provide enough Mg to form the transition layer and
provide enough of a remaining MgO template to reduce the interfacial strain and prevent the
Si diffusion. Using the reaction rate of a MgO-Fe2O3 powder system as reported by Shimada
et al.[157], it was estimated that at 915 oC, around 5.6 nm of MgO film would be lost during
40 minutes of PLD growth. While this is not inconsistent with our findings, due to
significant differences between the powder system and the PLD process, further study is
needed to determine MgO degradation rates under PLD conditions and a critical MgO
thickness required for high quality BaM film growth.
5.1.2.4 The Impact of Post-deposition Annealing
After deposition, the BaM films grown on 10 nm MgO template were annealed in air
at 1050C in two steps of two-minute duration each [141]. The FMR linewidth was
significantly reduced by the post-deposition heat treatment. The as-deposited films’ FMR
linewidths of ~220 Oe was consistently reduced to ~100 Oe after the post-deposition
annealing (see Figure 63). A ΔH value of 96 Oe compares well with values reported for
BaM films grown on lattice matched non-semiconducting substrates, such as MgO and
155
sapphire [53,76,87]. The improvement of the magnetic properties of films after annealing
indicates assorted defects within the film have been eliminated by the annealing process.
XPS does not show evidence of silicon diffusion at the surface of the annealed films.
Figure 63: Power derivative as a function of applied magnetic field in the region near
the ferromagnetic resonance at 53 GHz. The linewidth measured as the peak-to-peak
power derivative is 96 Oe for the post annealed BaM film deposited on 10 nm
MgO(111)//SiC(0001) [141].
A typical AFM image of the layer structured film after annealing is shown in Figure
64a. In comparison to the as-grown film (Figure 56a), the film annealed at 1050 °C shows
much bigger triangle/hexagonal crystallites with several steps, which are also seen clearly in
the AFM line scans (Figure 64c). The measured step heights (Figure 64b) at three points
were around 1.3 nm, roughly half the BaM unit cell dimension along c axis. Compared to
the as grown film, the surface roughness of ~ 2 nm was not improved.
156
(a)
(b)
400nm
(c)
Figure 64: (a) AFM image of the annealed BaM on SiC with MgO template, (b)
measured step heights, and (c) profile along the lines in (a).
The interface between the BaM and MgO, however, became much broader (~ 150
nm) as shown in Figure 65 compared to the as-grown film (~ 110 nm) in Figure 59a. The
broader interface indicates that the annealing promotes the reaction between the MgO
template and BaM film, which is consistent with the results reported by Li et al. that the high
temperature and extended anneal can make the spinel transition layer grow thicker [159].
157
Figure 65: XPS depth profile of annealed BaM films grown by PLD with 10 nm MgO
grown at 150 oC by MBE.
As shown in Figure 66, the as-deposited spectrum reveals all (0, 0, 2n) diffraction
peaks indicating a pure phase magnetoplumbite crystal structure having strong c-axis
orientation perpendicular to the substrate plane. An unknown peak appears near 2~28o
signaling the presence of a secondary phase, which was originally indentified as -BaFe2O4.
However with closer analysis, this peak is more likely due to formation of hexagonal
W-type barium ferrite with chemical formula of BaMg2Fe16O27, since this peak only existed
when BaM was grown on MgO substrate or template (Mg atoms involved), not on
substrates such as Al2O3, SiC, or Si. If the second phase is BaFe2O4, we should see it on
other substrates at the identical growth condition. Evidence for the existence of this phase is
eliminated after the annealing procedure, which may suggest decompose of the W type
158
ferrite to form M type and spinel ferrite at 1050 oC. After anneal, a peak around 18.4o starts
to show up. This peak is designated from the spinel MgFe2O4 phase, which is the reaction
products from BaM reacting with MgO and decomposition of the W type BaM. The XRD
result is consistent with the XPS depth profiling result, which suggests annealing would
promote the interface reaction and thus cause the broadening of the interface.
Intensity (Arb. Units)
SiC 006
Pre Anneal
Post Anneal
008
006
SiC 007
(b) **
(a)
15
0014
0012
0010
*
20
25
30
35
40
45
50
55
60
2 Theta (degree)
Figure 66: XRD patterns of BaM film before (a) and after (b) anneal. * Designated W
type BaM and ** designated MgFe2O4 (111)
Even though with more Mg diffusion and broader interface compared with the
as-grown film, the enhanced magnetic properties for the annealed BaM film may suggest
the importance of the crystallinity and orientation improvements of “bulk” BaM film
formed from the spinel transition layers, which provide a good template for the early BaM
formation. In our experiment, we also found the two steps of two-minute duration annealing
is the optimum post-anneal treatment condition for our BaM film on SiC, which is different
159
from most groups’ reports that annealed BaM film for hours long [51,77]. Figure 67
demonstrated the typical changes in Mg 1s peak of BaM sample surface after annealed for
different time. The concentration of surface segregated Mg in the ~2nm sampling volume
ranged from 3.9 % of as-grown film to 4.8 % of 4 minute annealed film, then 12.5 % for 14
minute annealed film. After 14 minute annealing, the film showed reduced magnetic
properties with saturation magnetization of only 1.6 kG, which is much lower than 4.4 kG of
as-grown film. The long time annealing destroyed the crystal orientation and drove more
Intensity (arb. units)
interdiffusion at the BaM/SiC interfaces.
1312
1310
1308
1306
1304
1302
1300
Bonding Energy (eV)
Figure 67: The region is shown where a Mg 1s peak will appear if Mg is present in the
surface of the film. (a) As deposited BaM grown on 10 nm MBE-grown MgO, (b) two
step annealing of two-minute duration each, (c) two step annealing of seven-minute
duration each.
Yoon et al. grew BaM on MgO (111) substrate, and annealed film at 1000 °C in air
160
for 1 hour [77]. After anneal, the FMR linewidth decreased 45-60%, however, the FMR
linewidth became much broader after anneal longer than 1 hour. Hence, the optimization of
anneal time is a key to improve magnetic properties of BaM films, especially for films
grown on such substrates as SiC or GaN, which is easily oxidized in oxygen environment.
The annealing time should be long enough to help improving the misorientation and
crystallinity, but short enough to prevent serious interface breakdown, especially for BaM
grown on substrate that is easily oxidized or react with the BaM.
5.1.2.5 Summary
Three generations of interface engineering for the integration of barium hexaferrite
(BaM) films on 6H-SiC has culminated with BaM film deposited by PLD showing a
ferromagnetic resonance linewidth of less than 100 Oe, which can be used for integrated
devices. The highest quality BaM deposited by PLD was on a 10nm single crystalline MgO
template grown by MBE on the SiC substrate. A systematic study of the interfaces for the
three generations shows that the chemistry, crystallinity, morphology and magnetization of
the BaM films deposited by PLD highly depends on the quality of the surface prior to PLD
deposition, and, for the third generation, the thickness of the MgO templates.
In contrast to a 10 nm MgO template, a 4 nm MgO template results in
non-stoichiometric BaM films with random orientation, three-dimensional film properties,
and a coercivity larger than 1000 Oe. Through the interface analysis, the deterioration of the
BaM films on the 4 nm template may be attributed to the reaction between the MgO and the
BaM at the growth temperature of 915 oC that depletes the MgO template and then allows Si
diffusion into the BaM. Cross-section TEM reveals that the BaM films with the best
161
magnetic properties were epitaxially aligned on a transition layer of mostly a spinel
MgFe2O4 structure which formed between the BaM and the MgO template at the PLD
growth temperature of 915 oC. Thus the 10 nm MgO template is necessary and not only
forms a spinel transition layer that is a similar structure to BaM, but is also thick enough to
provide a diffusion barrier throughout the PLD deposition process in order to achieve device
quality BaM on SiC.
The combined use of MBE to provide a high quality template followed by the
growth of a thick film of BaM by PLD is proven to be a simple and effective method to
integrate device quality BaM with SiC. However, the impact of high temperature processing
on the continuing reaction between the BaM and the MgO/Mg-ferrite layer suggests that a
more chemically inert template is needed for the extended high-temperature processing
required for thick BaM films.
5.2
Integration of BaM with 6H-SiC by Molecular Beam Epitaxy
In the previous section, the magnetic and microwave properties of PLD-grown BaM
on SiC have been improved through engineering increasingly effective interfaces. The
highest quality BaM deposited by PLD was on a 10nm single crystalline MgO template
grown by MBE. The increased order and alignment enabled by the MBE-grown MgO layer
suggests that MBE deposition of a high quality BaM seed layer has the potential to be an
even more effective template for near-perfect thick BaM film deposition by PLD or for
enabling the tight stoichiometric control and near-perfect structure properties needed for
162
effective property coupling in ferrite/ferroelectric multiferroic heterostructures. The section
is to demonstrate well-ordered, stoichiometric, high-purity epitaxial BaM films on 6H-SiC
deposited by MBE.
5.2.1
Establishing Operating Conditions Window
Comparing to the growth methods as PLD and sputtering to grow oxide film by
directly transferring the stoichiometry from the target to the film, MBE is more complicated
because the right stoichiometry can only be achieved by controlling individual flux
precisely. In the Ba-Fe-O system, a lot of ternary compounds have been reported, such as
BaFe2O4 (BaO·Fe2O3), Ba2Fe2O5 (2BaO·Fe2O3), BaFe12O19 (BaO·6Fe2O3), etc. [146,147].
The first step in achieving high quality BaM was to understand the relative fluxes necessary
to grow stoichiometric BaFe12O19 with the proper oxidation states of both Fe and Ba.
Molecular oxygen, while being the most convenient to oxidize metal Ba to Ba 2+, is
not sufficiently active to allow all metal Fe to form ferric iron (Fe 3+) even at O2 pressures on
the order of 10-4 Torr [113,115], a pressure which is not conducive to the MBE environment.
Atomic oxygen produced from pure O2 is needed to grow -Fe2O3 or -Fe2O3 phases [107]
in MBE environment. For the Fe flux of 6×1013/cm2s in this work, the use of molecular
oxygen in the 10-6 Torr only enables the Fe3O4, which is consistent with Chambers et
al.[107]. Therefore, for BaM growth, O atoms from the remote oxygen plasma are needed
to fully oxidize metal Fe into Fe3+. Once in the correct oxidation state, the Fe3+ can occupy
the needed crystallographic positions, namely octahedral (12k, 4f2, 2a), tetrahedral (4f1),
and bipyramidal (2b) coordination.
With the Fe flux and plasma power held constant, O2 pressure and Ba flux were
163
varied to achieve the correct stoichiometry and oxidation states as measured by XPS. The
oxygen pressure of 1.0 × 10-6 to 1.0 × 10-5 Torr correlates to an equivalent molecular oxygen
flux of 3.6 × 1014 /cm2 s and 3.6 × 1015 /cm2 s. At constant plasma power of 200 W, it was
found that the ratio of total oxygen/Fe flux lower than 20 could not fully oxidize metal iron
to Fe3+, resulting in mixture of Fe2+ and Fe3+ in BaM film as confirmed by XPS Fe 2p peak.
Fully oxidized Fe3+ can be achieved at total oxygen/Fe flux ratio of 21 or higher. As shown
in Figure 68, the Fe 2p spectrum for the BaM films grown at O/Fe flux ratio of 40 reveal a
narrow peak (FWHM 2.16 eV) around 711 eV and a shake-up satellite peak at around 719
eV, which is characteristic of Fe3+ in BaM [28]. In contrast, the Fe 2p spectrum for the BaM
film grown at O/Fe flux ratio of 15 shows a broad peak (FWHM 3.56 eV around 710.8 eV
and no satellite. The broad peaks are attributed to existence of dual iron oxidation states
(Fe2+ and Fe3+) that have different, but non-resolvable binding energies.
Figure 68: Fe 2p photoemission spectra obtained with Mg k X-rays for 200nm BaM
grown at O/Fe flux ratio of (a) 20, and (b) 40.
After achieving the right stoichiometry and Fe3+ oxidation state, the effect of
164
substrate temperature deposition (Ts) during deposition was explored. All these films were
grown at constant O pressure of 3 × 10-6 Torr with plasma running at 200 W. For Ts < 600 °C,
the as-grown films appeared to be amorphous as determined by RHEED. As shown in
Figure 69, very weak RHEED implying poor film crystallinity was observed at 600 °C.
When Ts was increased to 800 °C, RHEED became very streaky, suggesting single
crystalline film. However at 900 °C, spotty RHEED from both surface diffraction and
transmission was observed, implying growth of three dimensional island features.
6H-SiC
600 °C
800 °C
900 °C
Figure 69: RHEED patterns of the BaM films deposited at 600, 800, and 900 °C on
6H-SiC respectively.
AFM micrographs also show changes in the film morphology with varying Ts.
Figure 70 shows that for Ts at 600 °C, the films have granular structure with grain sizes of
20–30 nm. As Ts is increased, these nuclei grow much bigger and then have average size of
400 nm at 900°C. The RMS roughness of those films increase dramatically, from
sub-nanometer at 600°C to 3-4 nm at 800°C, and then 30-40nm at 900°C because of the
165
formation of those big islands.
SiC
600 °C
200nm
200nm
800 °C
900 °C
400nm
400nm
Figure 70: AFM images of the BaM films deposited at 600, 800, and 900 °C on
6H-SiC respectively.
The increase of grain size can be interpreted by the common nucleation theory [164].
For a deposition performed at low temperature, the surface mobility of the deposited species
on the substrate surface is very low, which leads to a high nucleation density and small
nucleation site size and thus a small grain size while the nucleation sites finally turn into
grain. With the improved surface mobility at an elevated deposition temperature, the growth
rate of the crystals increase much faster than the nucleation rate and therefore a large grain
size will form [165]. However, high deposition temperature will cause interface mixing,
such as Si diffusion in our case, and thus result in deteriorated crystal quality. For BaM
growth, the optimum growth temperature is around 800 oC, where RHEED and AFM
showed single crystal film with uniform grains and smooth surface.
166
5.2.2
The Impact of MgO Template
It was shown previously in section 5.1 that the highest quality BaM with a
perpendicular anisotropy field of 16.9 kOe, a 4πMs of 4.3 kG, and a FMR linewidth at 53
GHz of 220 Oe (improved to < 100 Oe with flash anneal) is obtained on a high quality 10
nm MBE-grown single crystal MgO template. The function of the MgO film is to provide a
diffusion barrier, form a spinel transition layer, and enable the alignment of BaM crystallites
early in the PLD deposition process, which are necessary to achieve device-quality BaM on
SiC. Then the question comes to if the MgO is still necessary for the BaM film grown in
MBE system, which has better control capability than the PLD system. In order to address
this question, two thin BaM films were grown by MBE on SiC and MgO/SiC respectively.
When BaM was deposited by MBE directly on SiC, we see evidence of an interface
breakdown in the form of oxidized silicon between the BaM and SiC (similar to what was
observed in the PLD system). A 2 nm single crystal MgO (111) film deposited by MBE on
the SiC prior to BaM deposition by MBE, however, shows minimal if any increase of
substrate oxidation. In Figure 30, the left peaks are XPS Si 2p peaks and right peaks are XPS
O 1s peaks for 2 nm BaM films deposited by MBE on SiC with a 2 nm MgO template
(Figure 71a and c), and without an MgO template (Figure 71b and d). The Si 2p spectra
reveal two peaks; the left peak is shifted 1.5 eV to a higher binding energy relative to the Si
2p from the SiC substrate, which is assigned to Si-Ox [148]. The O 1s spectra also reveal 2
peaks. The right peak represents the lattice O 1s peak (i.e., the O bound to Ba or Fe in BaM
and O bound to Mg in MgO), while the peak shifted 1.8 eV to a higher binding energy is
indicative of Ox-Si. The relative amount of Si-Ox to Si-C of 0.1 seen in the Si 2p peak of the
167
film grown on the MgO template (Figure 71a), is approximately the same as the relative
amount of Si-Ox to Si-C present after MgO deposition. This suggests very little, if any, Si
oxidation at the interface after BaM deposition. In contrast, the Si diffusion at the interface
is severe when the BaM is deposited without the MgO template, as shown by the increased
intensities of Si-Ox in both Si and O peaks in XPS (Figure 71b and d).
Figure 71: XPS photoemission spectra obtained with Mg k X-rays for 2 nm BaM
films grown on 6H-SiC (a) Si 2p with 2nm MgO template, (b) Si 2p without 2nm MgO
template, (c) O 1s with 2nm MgO template and (d) O 1s without 2nm MgO template.
To determine the structure of the interface breakdown when BaM is grown by MBE
without an MgO template, angle-resolved XPS was used to investigate film chemistry of the
near-surface region for BaM films thinner than 2 nm. Measurements were carried out at
take-off angle of 90° and 30° to detect Si photoelectrons roughly from the top 8 nm and 4 nm
and O photoelectrons from the top 5 nm and 2.5 nm of the sample respectively. Due to the
less than 2 nm thickness of the BaM films, even at a takeoff angle of 30°, the substrate/film
168
interface is detected. Because the sampling volume is constant, at the 30° angle the film to
interface ratio is higher so that the bonding information in the spectra is more representative
of the film than the interface. This is best illustrated by the change in Si-C bonding in the Si
2p peak in Figure 72a. At 90°, the Si 2p peak is dominated by the Si-C bonding of the
substrate. At 30°, where less interface and substrate is within the sampling volume, the area
under Si-C bond peak intensity decreases 70% while the area under Si-Ox keeps almost the
same. The Si-C bond peak of the substrate is still visible at 30° through the film. If all the
attenuation of the Si 2p Si-C peak is assumed to be due to BaM, then an equivalent film
thickness of 1.5 nm is calculated. In Figure 72b, the O peak intensity of higher energy peak
(Ox-Si) deceases at glancing angle of 30o (i.e., the angle that increases the film to interface
ratio), suggesting that the layers near the SiC interface contain the Si-Ox. The absolute peak
area of the higher energy peak relating to Si-Ox in the Si 2p scan did not change
significantly at glancing angle of 30°. This further suggests that the SiOx is located directly
at the SiC substrate interface, and not scattered throughout the film.
169
Figure 72: XPS Si 2p (a) and O 1s (b) photoemission spectra obtained with Mg k
X-rays for around 1.5 nm BaM films grown directly on 6H-SiC at two different
electron emission angles: 90° (normal emission) and 30° (grazing emission).
Comparing the 3D AFM images for two thick BaM grown on SiC and MgO/SiC
respectively, the film grown on 6H-SiC directly is significantly rougher over the same scan
area, and shows apparent misalignments of crystal clusters (Figure 73a). However, the AFM
image of the BaM films grown on a 10nm MgO template, shown in Figure 73b, reveals a
uniformly smooth surface with indications of a granular structure with homogeneous size
distribution. The root-mean-square (RMS) roughness as measured by AFM over a 5 m × 5
µm area was 3.0 nm, much lower than the RMS of 10.6 nm for films grown on SiC directly.
This result suggests by using the MgO template to minimize the Si diffusion, more oriented
and uniform BaM film will form, which is consistent with the results for the PLD-grown
BaM films. But the BaM film grown directly on SiC by MBE showed much lower
roughness (~ 10 nm) comparing to the ~ 50 nm for the PLD-grown BaM (Gen I), suggesting
170
there might be less Si diffusion for the films grown in the MBE system than the PLD
system.
(a) RMS: 10.6nm
(b) RMS: 3.0nm
Figure 73: 3D AFM images obtained for 200 nm BaM films grown by MBE on (a)
6H-SiC (b) 6H-SiC with 10nm MgO template.
171
5.2.3
The Optimization of O Pressure
Once the MgO (111)//SiC (0001) interface was determined to be necessary to
prevent interface breakdown, further optimization of structure and chemistry of the BaM
film could be achieved. RHEED patterns in two high-symmetry azimuths are shown in
Figure 74 for each processing step for integration of BaM on 6H-SiC (0001) with a 10 nm
MgO template. The Figure 74a and b show typical RHEED patterns along [11 2 0] and
[10 1 0] on as-received 6H-SiC and H2 cleaned 6H-SiC, which has ( 3 × 3 ) R30° relative to
the as-received SiC (0001)-(1×1) surface [138]. Figure 74c patterns are a typical 10 nm
MgO (111)-(1×1) template grown on H2 cleaned SiC (0001) for subsequent BaM film
growth [158]. Optimal BaM (0001) growth pressure is around 3.0×10-6 Torr, which enables
a single crystalline BaM phase as shown by streaky RHEED (Figure 74d), a pure (0,0,2n)
XRD pattern, and a c-axis oriented VSM loop (shown and discussed later). BaM film
RHEED patterns grown under this optimal oxygen pressure differ from MgO RHEED
patterns due to the factor-of-two difference in lattice parameter, as shown in Figure 74d
compared to c. RHEED of MgO (111) along [1010] is dominated by coherent diffraction
between adjacent rows of Mg and O atoms, which are which are separated by
2 a/2 = 2.98
Å, while the [11 2 0] direction is dominated by rows of Mg or O atoms separated by
3 ×2.98
= 5.16Å [30]. The streak spacing ratio in the zeroth-order Laue zone for MgO
(111), h [10 1 0]/h[11 2 0] =
3
is exactly the reciprocal of the row spacing ratio in real space.
Since the lattice constant for BaM (5.89 Å) is almost twice that of the 2.98 Å lattice spacing
for MgO (111), the BaM diffraction pattern showed half-order streaks in the zeroth-order
Laue zones as the underlying MgO (111) pattern in both directions. The streak spacing of
172
BaM from RHEED is 2.9 % bigger than half of the MgO (111) pattern. The bulk lattice
constant of BaM is 1.3% only larger than that of MgO (111). This reveals that the BaM film
is grown epitaxially under tension. The presence of streaks suggests two-dimensional
growth.
Figure 74: RHEED patterns at for (a) as received 6H-SiC, (b) H2 cleaned 6H-SiC
(0001), (c) 10nm MgO (111) , (d) 200nm BaM film grown on MgO (111) at PO2 = 3×10-6
Torr along the [11 2 0] and [10 1 0] azimuths. The primary beam energy was 12.5keV.
173
O pressure played an important role for determining the surface structure of BaM. If
BaM is grown under an oxygen environment of 5×10-6 Torr and hence an excess oxygen
environment relative to the optimal, the streaky RHEED pattern initially formed will fade
within 2 minutes as film thickness increases. As shown in Figure 75a and b, the pattern
eventually evolves into spotty streaks and rings after 30 minutes of growth (~50 nm thick),
suggesting formation of a polycrystalline film from a three-dimensional growth. The
mechanism of the 3D growth under high O2 pressure can be interpreted in three aspects.
First, the velocities of the Fe and Ba atoms can become so slow that there will not be enough
surface activation for c-axis only oriented growth, and disoriented grains and rough surfaces
result. Second, harsh oxygen plasma (450 mV) contains more active O species that can drive
more Si diffused into the film. Dr. Goodrich showed severe MgO/SiC interface breakdown
at oxygen plasma of 125 mV [73]. And last, the XPS result showed a Fe/Ba ratio of 8 for
the BaM grown under excess O, implying formation of other phases with much lower Fe/Ba
ratio than 12. This hypothesis will be confirmed by the XRD analysis shown later.
174
(a)
(b)
(c)
(d)
Figure 75: RHEED patterns for 200 nm BaM films grown on MgO (111) template at
Po2 = 6×10-6 Torr along the (a) [11 2 0] and (b) [10 1 0] azimuths, and Po2 = 1.5×10-6
Torr along the (c) [11 2 0] and (d) [10 1 0] azimuths. The primary beam energy was 12.5
keV.
For BaM grown under a deficient oxygen environment relative to the optimal,
however, a different sharp and streaky RHEED pattern (Figure 75c and d) was observed
from a film grown under the optimal oxygen environment (Figure 75d). The
oxygen-deficient diffraction patterns exhibited half-order streaks in the zeroth-order zone
and the half-order Laue zones. The clear 2×2 reconstruction relative to MgO (111)-(1×1)
suggests formation of spinel-like unit-cell periodicity [108]. In oxygen deficient
environment, Fe metal can not be fully oxidized to Fe3+ and hence spinel Fe3O4 would form.
The streak spacing between the high intensity streaks and the half-order streaks for the
spinel structure is equal to half those of the MgO template, reflecting the near factor-of two
175
difference in lattice parameter between the film and MgO. The broad Fe 2p photoemission
spectrum without the Fe3+ satellite feature confirmed the existence of Fe3O4 phase in this
BaM film grown under deficient oxygen environment. Just because of the existence of the
Fe3O4 phase, XPS showed a Fe/Ba ratio of 14, greater than the stoichiometric ratio of 12.
The hysteresis loops, shown in Figure 76, essentially confirm the findings of the
crystal structural studies from RHEED. At a deficient O2 environment, VSM shows
anomalous loops with easy axis aligned in plane and low coercivity in both directions. This
is due to the formation of the spinel Fe3O4 (111), which has an easy axis in the film plane due
to the the shape anisotropy [38]. At an excess oxygen environment, both loops exhibit large
hysteresis and high coercively, again suggesting the magnetic anisotropy is either randomly
oriented in the films or there are some other Ba-Fe-O magnetic compound formed. However,
for the film grown at optimum pressure (3×10-6 Torr), VSM confirm that the easy magnetic
axis of the BaM film is aligned perpendicular to the film plane, which is consistent with the
crystallographic c-axis aligning perpendicular to the sample plane. The perpendicular loop
exhibits a large coercivity (Hc~2000Oe) and a remanence of 0.7. The in-plane
magnetization is quasi linear and the estimated anisotropy field of 16.2 ± 0.2 kOe is in
agreement with the published results of the single crystal BaFe12O19 value of 17 kOe. In
addition, the magnetization of the film was estimated with VSM to be 90% from the bulk
material value of 4.48 kG [166]. XPS of the as-deposited film under the optimal oxygen
pressure showed the surface composition to 11.2, close to the expected stoichiometric Fe to
Ba ratio of 12. No Si diffusion was observed in this BaM film by XPS when intermittently
analyzed during growth of a 20 nm film at approximately 5, 10 and 20 nm thicknesses.
176
Figure 76: Pressure impact on VSM Hysteresis loops for BaM films grown by MBE
with 10 nm MgO templates at (a) Po2 = 1.5×10-6 Torr, (b) Po2 = 3.0×10-6 Torr, and (c)
Po2 = 6.0×10-6 Torr. The maximum applied field of 12,500 Oe aligned parallel and
perpendicular to the film plane.
177
The change in crystal structure of BaM films grown at different oxygen pressures
can be seen in the x-ray diffraction patterns of the films as well. Figure 77 shows  -2 XRD
scans of three BaM samples grown under different oxygen pressure using Cu K radiation.
BaM films grown in a deficient oxygen environment (1.5×10-6 Torr) showed mixture of
hexagonal BaM, spinel Fe3O4, and cubic BaO phases. As shown in Figure 77a, the BaM in
the film was aligned with the c-axis normal to the substrate. Both Fe3O4 and BaO were
aligned in (111) orientation, matching the MgO (111) and the hexagonal SiC substrate. The
XRD pattern is consistent with RHEED, XPS, and VSM results showing the existence of
spinel phase in the BaM film that grown in a deficient oxygen environment. It is interesting
to note that films grown under excess oxygen environment, showed a small amount of
impurity phase of hexagonal -BaFe2O4 oriented in (100) orientation, shown in Figure 77b.
It has been found in calcination and solid-state reaction that this BaFe2O4 phase usually
coexists with BaFe12O19 and Fe2O3, along with other metastable phases [167,168]. In sol-gel
precursors, BaFe2O4 , orFe2O3 have always been reported to crystallize before the
formation of pure BaM and the completion of the phase transition to the hexagonal barium
ferrite occurs at 1050°C [169]. In our MBE growth, BaM grown at excess O may promote
formation of BaFe2O4, but the growth temperature of 800°C is not high enough for this
phase to transition into BaM. Although XRD did not show random oriented (1, 0, n) BaM
peaks, the existence of random orientation can not be ruled out because it is possible that the
small amount of random oriented BaM may not be registered by XRD.
178
Figure 77: X-ray diffraction pattern for Ba-hexaferrite (M-type) film grown under (a)
deficient oxygen environment, (b) excess oxygen environment, and (c) optimal oxygen
environment.
Figure 77c shows strong (0, 0, 2n) diffraction peaks indicating epitaxial, oriented
growth of BaM with the c-axis perpendicular to the substrate under optimized oxygen
pressure (3.0× 10-6 Torr). Except for the SiC substrate and Al holder lines, there are no
additional peaks in this spectra, suggesting no secondary phases in the film. To fully
characterize the texture of the BaM film grown under optimal oxygen pressure, an in-plain
pole figure was obtained by tilting the sample and rotating it in a spiral pattern for a fixed
value of 2 at 30.6° representing the <008> plane. Figure 78 clearly illustrates a single
dominant peak and six dots in the middle suggesting hexagonal symmetry of the BaM
film. The single dominant peak at  = 90° and  = 0 corresponds to the <008> reflection,
which is consistent with the XRD -2 data in Figure 77 and suggests low c-axis
dispersion. The six dots represents the <107> reflection which exhibits a d-spacing very
179
close to that of the <008> peaks (2= 30.6o for <008> as compared to 2= 32.0o for
<107>). The interplanar angle between the <107> plane and <008> plane is measured as ~
33.8o in pole figures, which is very close to 33.1o estimated by the lattice parameters of
BaM (a=5.89 Å and c=23.2 Å) [45]. The <107> type reflections are observed as well
defined dots rather than a ring, which confirm the strong epitaxial relationship between the
BaM and the SiC.
Figure 78: Pole figure obtained for a fixed value of 2= 30.6o. The single dominant
peak corresponding to =90 o and  = 0 o corresponds to the <008> reflection. The six
dots exhibiting six-fold symmetry correspond to the closely spaced <107> type
reflections illustrating the epitaxial relationship between the BaM and SiC.
A commercial MFM (Ambios Technology USPM model 2SAAVO) was applied to
study the structure of domains in the BaM films. Figure 79 shows the 2 m×2 m AFM and
MFM images of BaM grown at the optimal growth condition and whose magnetic
properties are shown in Figure 76b. The AFM image reveals a uniformly smooth surface
180
with indications of a granular structure with homogeneous size distribution. The grains are
round shape with mean size of 80 nm in diameter. The RMS roughness as measured by
AFM over a 2 m × 2 µm area was 1.4 nm with a variation in roughness on the sample and
from area-to-area of ± 0.3 nm. This is compatible with the best BaM films grown by PLD
with 10 nm MBE grown MgO templates. The MFM image (Figure 79b) is of the sample in a
remnant magnetization state, and shows a cluster-like domain structure consisting of up and
down magnetization corresponding by the dark and white contrast respectively. According
to Lisfi et al., the up and down magnetic domain orientations enable minimization of the
magnetostatic energy in the perpendicular orientation of the anisotropy of the BaM film.
Figure 79: 2m×2m (a) AFM and (b) MFM images of BaM film with 10nm MgO
template grown at optimal oxygen pressure.
The cluster-like structure for this film in a 5 m × 5 µm scale (Figure 80a) is similar
to the structure observed by Lisfi et al. in BaM film grown by PLD on both Al2O3 and
SiO2/Si substrates [170]. However, the Gen III film grown on SiC by PLD showed a more
defined domain structure, more like strip domains (Figure 80b). Since the VSM loop of
MBE grown BaM showed large coercivity without any defined shoulder (Figure 76b), the
cluster-like domains usually form [171]. Strip domains only exist for a system with low
181
coercivity and hysteresis loop with a well defined shoulder [171], which is close to the VSM
loops of PLD grown BaM shown in Figure 58a.
(a)
(b)
1 m
1 m
Figure 80: 5m×5m MFM images of BaM film with 10nm MgO template on SiC
grown by (a) MBE and (b) PLD (Gen III BaM film).
The crystalline structure of the BaM grown under optimal pressure was also
analyzed by the high angle annular dark field (HAADF) microscopy. For thin samples, the
contrast in HAADF image is strongly dependent on the atomic number Z thus the image
contrast is readily interpretable in comparison to HRTEM. The HAADF contrast is ~Z1.7
sensitive, which would imply the Ba columns should be the brightest in the film. Figure 40
is the high resolution HAADF image of the BaM film, in which the white lines are the
regions where Ba atomic planes are located. The white dots in the image (outlined with red
circles) represent the Ba atomic columns suggesting a high quality BaM film with limited
stacking faults. The Ba atomic columns contain both Ba and O atoms in both [10-10] and
[11-20] directions, as shown in Figure 81a. The distance of Ba atomic planes along the c
axis (measured from Figure 81b) is 1.17 nm, which matches well to half of the bulk unit cell
of BaM (2.32 nm) [45]. EDX analysis of the bulk BaM layers gives the composition of
BaFe11.6O18.0 which confirms on average, within the accuracy of EDX technique, the correct
182
stochiometry of the film. This is also consistent with previous XPS analysis of the surfaces
of the MBE-grown BaM films, suggesting uniform chemistry through the film.
Figure 81: (a) Model of BaM crystal structure in [11-20] orientation, (b) high angle
annular dark field (HAADF) image of BaM grown on 6H-SiC with a 10nm MgO
template, white lines indicate Ba atomic planes. Inset is the EDX line scan for three
spots in the film. Regions marked as red circles show the Ba atomic columns,
suggesting high quality film with limited stacking faults (The waviness of the Ba
atomic planes is artificial, due to instabilities of the STEM scanning coils).
By using a 10nm thick MgO template, single crystalline, epitaxial barium ferrite
films have been successfully grown by molecular beam epitaxy for the first time on 6H
silicon carbide substrates. Oxygen pressure is very critical for determining the chemistry
and surface structure of BaM. An oxygen deficient or rich environment will cause impurity
phase of Fe3O4 or - BaFe2O4 respectively. BaM films grown under the optimal oxygen
environment has a six fold symmetry with anisotropy highly oriented and perpendicular to
the substrate plane. XPS analysis showed that the use of the thin MgO template in the early
stages of film growth trapped the diffusing Si and prevented the migration of Si into the film.
183
AFM and MFM revealed uniform smooth surface with cluster-like magnetic domains. The
high quality BaM grown by MBE has the potential to function as a more effective template
for subsequent near-perfect thick BaM grown by PLD or LPE and to enable the tight
stoichiometric control needed for more effective property coupling in ferrite/ferroelectric
multiferroic heterostructures.
5.2.4
Demonstration of the MBE-grown BaM as a Seed Layer for Thick BaM
Growth by PLD
Using a high-quality MBE grown-BaM film as a seed layer, the quality of the
subsequent thick BaM film grown by PLD is further improved compared to the Gen III BaM
film grown on a 10nm MBE-grown MgO template. The hysteresis loop (Figure 82) showed
highly c-axis oriented perpendicular to the substrate for the BaM film grown on a 10nm
MBE-grown BaM seed layer. Comparing to the Gen III BaM film (Figure 58a), the
anisotropy field of around 17 kOe and 4Ms of 4.4 kG of this BaM are comparable to the
Gen III BaM. The main difference in the magnetic properties is that this film exhibits lower
coercivities at both in-plane (180 Oe vs. 300 Oe) and perpendicular directions (680 Oe vs.
800 Oe), suggesting better c-axis orientation and less crystal defects than the Gen III films.
AFM image of this BaM film, shown in the inset of Figure 82, reveals a smooth surface
(RMS roughness of 1.3 nm) with hexagonal patterns, which are also comparable to the best
Gen III films.
184
4000
4M (G)
2000
500nm
0
-2000
In plane
Perpendicular
-4000
-18
-12
-6
0
6
12
18
Applied Field (kOe)
Figure 82: Hysteresis loop of BaM grown on a 10nm MBE-grown BaM obtained by
VSM with a maximum applied field of 12,500 Oe aligned parallel (open red square)
and perpendicular (solid black square) to the film plane. Inset is the AFM image for
this film.
Figure 83a is a cross-sectional bright field TEM image of the BaM film grown on a
10 nm MBE-grown BaM seed layer. Similar to the Gen III film shown in Figure 60, a region
with different contrast between BaM and SiC is clearly observed in Figure 83a, indicating
presence of a transition spinel layer and an amorphous SiOx layer. The inset SAED of
Figure 83a shows fully epitaxial growth of BaM on SiC. Comparing to the Gen III film, this
film shows more uniform film thickness with thinner amorphous SiOx layer. In Gen III film,
the thickness of the SiOx layer varies from 10 to 20 nm and is non-uniformly distributed.
However, for this film, only 5-7 nm SiOx formed at the interface, suggesting less interface
breakdown. The thinner SiOx formed in this film may contribute to the improved magnetic
properties shown in Figure 82. Figure 83b confirms the well ordered high quality BaM
185
structure on top of the spinel and SiOx layers.
BaM
BaM
SiC
(a)
(b)
Figure 83: (a) Bright-field cross-section TEM image of BaM thin films grown on
6H-SiC with a 10nm MBE-grown BaM seed layer, inset is the SAD pattern from the
substrate and film in [11 2 0] zone axis; (b) High resolution TEM of the BaM film,
inset is the digital diffractogram.
5.2.5
Interface Study for the Potential Applications of the BaM-ferroelectric
Heterostructures
The elemental composition and the structure of the interface are important for
engineering effective and efficient ferrite-ferroelectric coupling because magnetoelectric
coupling between layers is extremely sensitive to the quality of the interface. Most of the
work developing the BaM-ferroelectric heterostructures have been utilizing PLD or
sputtering. So far, very few studies have addressed interface issues between BaM and any
substrates, since a detailed chemical and structural evolution study of the interface structure
was not feasible due to the limitations of characterization capability in the sputtering or the
PLD system. In this section, we examined the structure and composition at the interface
186
between an MBE-grown BaM films on SiC using in-situ XPS, RHEED, and ex-situ
cross-section TEM. By understanding how the initial structure and composition of the
interface evolve with film thickness, engineering an effective interface for more efficient
BaM-ferroelectric coupling can be realized.
RHEED patterns in [11 2 0] azimuth are shown in Figure 84 as a function of
coverage for epitaxial BaM on 6H-SiC (0001) with a 10 nm MgO template. Figure 84a and
b show typical RHEED patterns of a H2 cleaned 6H-SiC and a 10 nm MgO (111)-(1×1)
template along the [11 2 0] azimuth of bulk SiC. Figure 84f is the pattern measured for a
2000 Å-thick BaM grown under optimal oxygen pressure of 3.0×10-6 Torr. As described in
the previous section, this pattern was verified to be BaM through XRD and XPS, thus this
pattern is used as a reference surface for bulk-like BaM (0001). Comparing the bulk-like
BaM pattern in Figure 84f to the MgO template in Figure 84b, the line spacing differs from
each other by a factor of 2 in the [11 2 0]BaM//[1 1 0]MgO. This is expected, since the lattice
constant for BaM (5.89 Å) is almost twice that of the 2.98 Å lattice constant for MgO (111).
As a result of this difference in surface structure, the RHEED pattern of BaM can be
considered as (1×1) relative to the MgO (111)-(1×1). However, the initial layers of BaM
film growth produces a different pattern from the bulk-like BaM. For the film thickness of
~10 Å, the high intensity streaks (Figure 84c) are twice the spacing as the MgO (111) pattern,
Figure 84b. Thus, the ~10Å film shows a 2×2 surface relative to the MgO (111)-(1×1).
Weaker intensity streaks are apparent in between the higher intensity streaks of the ~10 Å
film. These weaker streaks increase in intensity as the film grows until a clear 1×1 BaM
surface structure is visible at ~100Å, as shown in Figure 84e. The intensity modulations are
187
not visible along the RHEED streaks during film growth, suggesting two-dimension
growth.
 1210 
SiC
MgO
~10 Å
~40 Å
~ 100 Å
~ 2000 Å
Figure 84: RHEED patterns along the [11 2 0] azimuth of bulk SiC, for epitaxial BaM
(0001) on MgO (111) template at PO2 = 3×10-6 Torr as a function of film thickness,
taken at 12.5 keV: (a) H2 cleaned 6H-SiC (0001), (b) 10nm MgO (111) template, (c) 10
Å, (d) 40 Å, (e) 100 Å, and (f) bulk BaM (0001).
Zhang et al. and Robbert et al. reported the 2×2 phase of -Cr2O3 when they grew by
MBE on hexagonal structure Pt (111) to film thickness thinner than 1 nm [172,173]. They
have interpreted this 2×2 pattern as due to the formation of inverse spinel Cr3O4 (111). This
assignment was confirmed by the expected spinel surface structure and broad XPS spectrum
of Cr 2p due to the coexistence of Cr2+ and Cr3+. For the same observation, Chambers et al.
interpreted the ~1 nm -Cr2O3 (2×2) structure grown on Pt (111) as mostly due to the
188
formation of a metastable - Cr2O3 (111), which has nearly the same structure with Cr3O4
(111) [174]. However the 2×2 reconstruction during our BaM film at small thickness can not
be interpreted as either of those two cases. Spinel - Fe2O3 phase is not likely to form in our
experiment because the growth temperature of 750°C is much higher than phase transition
temperature of 325°C from - Fe2O3 to - Fe2O3 [175]. A shake-up satellite peak at ~719 eV
indicates that the film of ~ 10 Å thick (Figure 85) contains essentially Fe3+ instead of a 1:1
ratio of mixed Fe2+ and Fe3+ as would be expected in the spinel Fe3O4 phase [28]. But since
satellite peak is not very pronounced at film thickness of 10 Å, it is possible that a small
amount of Fe3O4 forms during the initial stage of growth. With the increase of film thickness,
Fe maintains as Fe3+ oxidation state and the characteristic satellite peak becomes more
pronounced, as shown in Figure 85. The observed 2×2 RHEED pattern relative to MgO
(111)-(1×1) in our ~10Å film, suggests formation of another spinel structure similar to
Fe3O4 (111) with a lattice constant of two times that of MgO, since the streak spacing of this
phase is equal to half those of the MgO template.
189
Fe 3+
satellite
Fe 2p 1/2
~2000Å
Fe 2p 3/2
Fe 2p Photoemission Instensity (arb. units)
~100 Å
~40 Å
~10 Å
Fe3O4
732
727
722
717
712
707
702
Binding Energy (eV)
Figure 85: High-energy-resolution Fe 2p core-level XPS spectra for epitaxial BaM
(0001) on 10nm MgO (111) template on 6H-SiC (0001) as a function of BaM film
thickness. For comparison, a Fe3O4 spectra is included, where the broad Fe 2p 3/2
peak indicates the presence of both Fe3+ and Fe2+ cations.
For a 200 Å thick BaM film, XPS shows a small Mg 2p peak situated at 49.6 ± 0.1
eV. Since the film thickness exceeds the attenuation depth of Mg 2p photoelectron (~ 80 Å).
This suggests that the magnesium signal is not from the MgO template but from the BaM
film. These results indicate that a certain number of magnesium atoms diffuse into the BaM
layer at the early stage of growth. Since RHEED suggests the existence of a spinel structure,
it is quite possible that ternary compounds such as (MgxFe1-x)Fe2O4 form at the interface.
Identical to Fe3O4, MgFe2O4 has an inverse spinel structure where the divalent cations are
localized in the octahedral sites [176]. To fully study the interface mixing and formation of
the spinel phase, cross-section TEM was used to study a 160nm BaM/10 nm MgO/SiC
interface structure.
190
Figure 86a is a cross-sectional bright field TEM image of the BaM film grown on the
10 nm MgO/SiC template. A region with different contrast between BaM and SiC is clearly
observed in Figure 86a, indicating presence of interfacial layer(s) with different structures
than that of the substrate and the film. Figure 86b shows the overlaid SAED patterns of both
the BaM film and the SiC substrate. The orientation relationship between BaM and SiC can
be briefly described as [0001]BaM||[0001]SiC and [11-20]BaM|| [11-20]SiC without evidence of
30o plane rotation. Between the BaM layer and the SiC substrate, there are two different
structures. Just beneath the BaM film, a band of around 20 nm thick with cubic symmetry
forms (Figure 86c), which compositionally contains Mg, Fe and O (supported by EDX data
that is discussed later), and structurally corresponds to a spinel Mg-ferrite (lower inset,
Figure 86c) with a small amount of Fe3O4. The formation of this Mg-ferrite suggests the
degradation of 10 nm MgO template during the early stages of BaM growth. Between this
layer of cubic symmetry and the SiC substrate is an amorphous structure of approximately 3
nm, which is deduced to be SiOx. The SAED (Figure 86b) shows epitaxial growth of BaM
on SiC even with the presence of an interfacial amorphous SiOx layer that is apparent after
growth. This implies that the disordered layer is caused by interface breakdown during
growth, most likely due to oxygen diffusion from the BaM and MgO film into the SiC
substrate during BaM film deposition, and not prior to BaM film growth. In Figure 61, the
presence of SiOx layer (> 10 nm) at interface of the PLD-grown BaM is also observed by
TEM. The thicker SiOx formed in the PLD-grown BaM film may be attributed to more
excess O flux and higher deposition temperature in PLD growth (2×10-2 Torr & 915°C ) than
the MBE growth (3×10-6 Torr & 800°C). Figure 86d shows a sharp interface between the
191
transition spinel layer and BaM in the final film. The d (111) of the spinel transition layer
measured from 5 different SEAD images (lower insert, Figure 86d) is 0.471 ± 0.008 nm,
which is very close to the reported value of 0.483 nm [177] for the spinel MgFe2O4 {111}
plane, again supporting the presence of the Mg-ferrite spinel. Figure 86c and d confirm the
high
quality
BaM
structure
on
top
of
the
Mg-ferrite
(b)
(a)
and
SiOx
layers.
[1-100]
BaM
100 nm
SiC
(d)
(c)
BaM
BaM
MgFe2O4
SiC
Mg-ferrite spinel
SiOx
Figure 86: (a) Bright-field image of cross-section TEM of BaM grown on 6H-SiC with
a 10nm MgO template, (b) SAED pattern from the SiC-film in [1 1 00] zone axis, BAM
is epitaxial with respect to SiC besides the presence of the amorphous SiOx at the
interface. (c)The interface of BaM film grown on 6H-SiC with 10nm MgO shows
existence of an amorphous SiOx layer (~3 nm) followed by a spinel transition layer
(~20nm), (d) Higher resolution TEM shows an abrupt interface between BaM and the
spinel transition layer, both structures are crystalline and epitaxial as shown by the
inset digital diffractograms, Mg-ferrite is in [11 2 ] and BaM in [1 1 00] orientation.
192
The chemistries of the transition layer and BaM layer are revealed by EDX line
scans. Figure 87a shows the SiC-film interface and the region (vertical line) along which is
taken an EDX scan line. The compositional profile obtained from the EDX scan line (Figure
87b) corresponds well to the image contrast in Figure 87a. The sudden increase of O
composition marks the start of the interface region, and presence of SiOx at the interface.
Small outdiffusion of Si and no presence of Ba is observed in the Mg-ferrite interfacial band.
The presence of Mg, Fe and O is clearly seen in the region that has spinel structure,
confirming that these layers (Figure 86d) are indeed Mg-ferrite.
Figure 87: Elemental profile across the SiC-film interface. a) HAADF image from the
SiC-film interface, vertical line denotes the regions from which the EDX is taken, as
shown in (b).
An 800°C growth temperature with oxygen plasma intensity of 210 mV (determined
by photomultiplier), can enable active (unbound) O species to diffuse along MgO grain
boundaries and initiate the formation of amorphous SiOx at the SiC interface. On top of this
oxide layer, the transition layer with mixed Mg, Fe, and O atoms is observed, which
confirms the existence of spinel MgFe2O4. Based on the EDX line scan result, the chemistry
of this ternary spinal compound is determined to be (Mg0.7Fe0.3)Fe2O4. This again confirms
193
that the MgFe2O4 is the dominant spinel phase at the interface. Because most iron ions are in
3+ valance state, the Fe3+ satellite feature is pronounced in XPS spectra (shown in Figure
87), instead of a broad Fe peak without any satellite feature, as would be expected in Fe3O4.
The formation of the transition spinel structure is a result of the reaction between MgO
template, the Fe atoms, and the active O from the plasma source. It is very interesting that
the Ba atom is not incorporated into this spinel structure (or is at least under the 1%
detection limit of EDX), and only exists in the final BaM film grown on this spinel phase.
Although the formation of BaFe2O4 (BaO·Fe2O3) and BaFe12O19 (BaO·6Fe2O3) at
temperature of 800oC are both thermodynamically favorable (with negative ΔG), the lack of
Ba in the transition layer suggests there is a continuous competition at the MgO template
interface for Fe to react and then form an oxide. RHEED patterns (Figure 84c and d) reveal
that the formation of spinel oxide is favorable in the early stage of BaM growth.
From a structure point of view, BaM has the magnetoplumbite structure that can be
viewed as stacks of cubic spinel blocks having <111> orientation (called S blocks) and R
blocks (shown in Figure 3). S block has a chemical formula of Fe6O8, and R block has
BaFe6O11. Ba ions only exist in R blocks, replacing an oxygen ion [162,163]. It has been
suggested that the spinel forms at an interface consisting of the following steps; an ion
approaches an interface, crosses it, becomes supersaturated in the reactant, and then
precipitates to form reaction product [178,179]. During MBE deposition of BaM on MgO
template, the O and O2 flux is more than 20 times of Fe flux, and the Fe to Ba flux ratio is ~
10. At the initial stage of growth, when Fe atoms approach the interface, here is MgO, the
reaction between MgO, Fe, and O to form MgFe2O4 would be dominant due to the excess
194
MgO and very limited Ba flux (less than 10% of Fe flux). Just because of the low Ba flux,
the reaction product of BaFe6O11 that incorporate Ba may be hard to reach supersaturation,
and can not start to form until the concentration of BaO is comparable to the MgO at the
interface. The initial stage of BaM growth by XPS shows that the Fe/Ba ratio decreased
from ~20 at film thickness of 1 nm to ~ 11 at film thicker than 20 nm, which suggests more
Ba atoms are incorporated into the crystal structure with the increase of film thickness. This
is consistent with the RHEED pattern evolution in Figure 84, which shows that the
formation of BaM starts after the film thickness reaches 10 nm.
Thus, (Mg0.7Fe0.3)Fe2O4 may be kinetically preferred to BaM under the initial
conditions for nucleation and growth due to the excess MgO and very limited Ba flux, and
the formation the (Mg0.7Fe0.3)Fe2O4 phase at the interface is structurally favorable for BaM
deposition because of the close relation and structural similarity of the spinel
(Mg0.7Fe0.3)Fe2O4 to the BaM film [159]. However, TEM (Figure 86c) shows a spinel
transition layer of around 20 nm, which is inconsistent with the RHEED (Figure 84e)
showing BaM starts to form at film thickness of around 10 nm.
The inconsistence in transition layer thicknesses from TEM and RHEED suggests
the formation of the transition spinel structure is not only from the initial reaction between
the MgO and Fe flux, but also from the continuous breakdown between the MgO and the
formed BaM film. Li et al. reported the formation of transition layer ZnFe2O4, when they
used a ZnO template of thicker than 20 nm to grow BaM on a Si substrate by sputtering
[159]. They confirmed the formation of ZnFe2O4 as a reaction product of BaM and ZnO,
since the transition layer grew during the ex situ annealing at the expense of ZnO and BaM,
195
as observed by HRTEM and XRD analysis. As discussed earlier in 5.1.2.3, for growing
BaM on SiC with MgO template by PLD, we found that the MgO is degrading due to the
reaction with BaM, and thus the 4 nm MgO template does not function as effectively as 10
nm MgO template to form the transition layer for subsequent BaM growth and mean time
remain certain amount at the interface to limit the Si diffusion. It is also found that if BaM is
deposited directly on SiC (without a 10 nm MgO) by MBE, both XPS and TEM show
evidence of Fe3O4 forming at the interface. Figure 88a shows a 5 nm BaM film grown on
SiC with a 10nm MgO template, Fe is mainly in the 3+ oxidation state revealed by the
pronounced satellite peak at around 719 eV. However, for a 5 nm BaM film grown at the
same conditions on SiC without the MgO template, XPS Fe 2p peak shows mixed oxidation
states of 2+ and 3+ without any satellite feature (Figure 88b), indicating the existence of a
Fe3O4 phase at the initial stage growth.
Fe 2p Photoemission Intensity
(Arb. Units)
Fe 2p3/2
Fe 2p1/2
(b)
(a)
Fe3+ Sat.
Fe3+
Mixed Fe2+ & Fe3+
735
730
725
720
715
710
705
700
Binding Engergy (eV)
Figure 88: Fe 2p photoemission spectra obtained with Mg k X-rays for 5 nm BaM
grown on SiC (a) with MgO template, and (b) without MgO template.
196
In Figure 89, the high resolution cross-section TEM confirms the formation of a
Fe3O4 transition structure, which is crystalline and epitaxial as shown by the inset digital
diffractograms. Comparing to a 3 nm SiOx amorphous layer in the BaM film grown on SiC
with a 10 nm MgO template (Figure 86c), there is a thicker SiOx layer (7 nm) formed at the
interface of the BaM film grown directly on SiC. This again confirms the function of the
MgO template to minimize Si diffusion. For the BaM film grown on SiC with or without a
MgO template, the XPS and TEM results further support that the spinel structure is
favorable and can facilitate the subsequent BaM formation, because of the structural
similarity discussed previously. Since the formation of the spinel phase at the interface is
structurally favorable for subsequent BaM film growth and the degradation of the MgO
template, using a spinel template (e.g. Fe3O4, and MgFe2O4) to replace MgO is promising to
further improve the quality of the BaM film.
SiOx
SiC
Fe3O4
Figure 89: High resolution cross-section TEM of the interface of BaM grown on
6H-SiC, shows existence of an amorphous SiOx layer (~7 nm) followed by a spinel
Fe3O4 transition layer.
197
5.3
Integration of Fe3O4 film with SiC by MBE for Potential Spintronics
Applications
Integration of half-metallic Fe3O4 film with SiC (0001) can take advantage of high
spin polarization of Fe3O4 and wide bandgap semiconductor for future multifunctional
spintronic devices. In addition, epitaxial Fe3O4 film grown on SiC substrate can be used as a
magnetic layer in SiC-based ferrite/ferroelectric multiferroic heterostructures, which would
open up new applications on many tunable magnetic devices.
5.3.1
Epitaxial Growth of Fe3O4 film on SiC
Fe3O4 has a cubic inverse spinel structure, with a room temperature lattice constant
of 8.396 Å. The (111) plane has pseudo hexagonal structure matches with the hexagonal SiC
substrate with lattice mismatch of 3.77%. Figure 90 shows a typical XRD pattern of a 50nm
Fe3O4 film grown on SiC. No planes other than (lll) are detected, suggesting pure magnetite
phase has the crystallographic order perpendicular to the (111) growth direction. Inset
RHEED patterns show no evidence of transmission spot pattern, and reveal 2D growth
mode. The Kikuchi lines existed along the [11 2 0] direction of SiC substrate indicates
atomically flat surface of the Fe3O4 film. AFM confirms that the Fe3O4 surface is smooth,
with an average surface roughness of 6 Å over 10 ×10 m2 scan areas. XPS measurements
show the presence of both Fe2+ and Fe3+ without any satellite features, similar to the peak
shown in Figure 68a. These data ruled out the growth of the competing maghemite phase (Fe2O3, also a cubic inverse spinel with a = 8.322Å).
198
Fe3O4
Intensity (arb. units)
(111)
[11 2 0]
SiC
(0006)
[1100]
Fe3O4
(222)
Fe3O4
Al
stage
10
20
30
40
Fe3O4
(333)
50
60
(444)
70
80
2 Theta (degree)
Figure 90: XRD pattern of a 50 nm thick Fe3O4 (111) film. The right inset: RHEED
patterns of the Fe3O4 (111) thin film when the incident electron beam is along the
[11 2 0] and [1 1 00] azimuths of 6H-SiC respectively.
The magnetic hysteresis loops of the 50 nm Fe3O4 (111) film has also been measured
at room temperature using a VSM with an external magnetic field up to 12.5 kOe. The field
was applied parallel and perpendicular to the [0001] axis of the SiC (0001) substrate (see
Figure 91, where the [-3 kOe – 3 kOe] part of the loop is shown). The diamagnetic
contribution from substrate and VSM holder has been subtracted from raw data. As shown
in Figure 91, typical shape anisotropy induced easy (in-plane) and hard (out-of-plane)
magnetic hysteresis loops can be observed, showing the observed in-plane and out-of-plane
coercivity of 500 Oe and 200 Oe respectively, which are consistent with the results of Fe3O4
(111) film grown on Al2O3 reported by Moussy et al. [38]. However, the coercivities in both
directions are much higher than the thick Fe3O4 film grown on PZT by spin-spray (110 Oe
199
out-of-plane and 90 Oe in-plane) [4]. This can be explained by the lattice mismatch induced
stress, which has more impact on the thin films than the thick films.
1.0
4M (arb. units)
0.5
0.0
-0.5
-1.0
Out of plane
In plane
-3
-2
-1
0
1
Applied Field (kOe)
2
3
Figure 91 : Magnetic hysteresis loops by VSM measurement of 50 nm Fe3O4 (111) thin
films at T ~300K after the subtraction of the substrate diamagnetic contribution.
It is also interesting to find that out-of-plane magnetic moment is hard to be
saturated even though at an external filed of 12.5 kOe. The unsaturated feature at high
magnetic fields also has been observed in the Fe3O4 grown on MgO [180,181] and Al2O3
[38], which is due to the effect of antiphase boundaries (APBs). APBs are structural defects
resulting from the nucleation of islands at the early stages of film growth, resulting in the
formation of a stacking fault in the iron cation sublattice [182]. Eerenstein et al. showed a
systematic decrease in the APB density with the increase of film thickness t that tends to
vary as t-1/2 [183]. Because of the decrease of the magnetization at the APB, the measured
magnetization is usually lower than the bulk value. The Ms of this film is estimated to be
400 ± 20 emu/cc, while the bulk value is around 470 emu/cc. The magnetic interaction at the
200
APBs can also explain that the saturation is more difficult to achieve in the thin films
compared to a single crystal without any APBs. The existence of APB or other defects in
this Fe3O4/SiC film needs to be studied systmaticlly in the future.
The high resolution cross-section TEM (Figure 92) shows a sharp interface between
the Fe3O4 film and the SiC substrate without obvious formation of SiOx amorphous layer,
which is attributed to the low growth temperature (~350oC) and moderate O plasma
environment (~50mV). The orientation relationship between Fe3O4 and SiC can be briefly
described as [111]Fe3O4||[0001]SiC and [110] Fe3O4|| [11 2 0]SiC.
Fe3O4 in [1-10] zone
SiC in [11-20]
Figure 92: High resolution cross-section TEM of the interface of Fe3O4 grown on
6H-SiC, shows a sharp interface without indicating formation of SiOx between the
film and the substrate.
5.3.2 Epitaxial Growth of Fe3O4 /BaTiO3/Fe3O4 Multiferroic Heterostructure on
SiC
As well known, BaTiO3 holds a well-developed ferroelectric order and has been
201
widely used in multiferroics study [184,185,186]. Thus Fe3O4/BTO/Fe3O4 would be a
good candidate to realize multiferroic tunneling junction. In this thesis, highly epitaxial
Fe3O4(FO)/BaTiO3(BTO)/Fe3O4(FO) multiferroic heterostructure were successfully
integrated with wide bandgap semiconductor SiC by MBE, which may provide a new
route for the introduction of the ferrite/ferroelectrics into spintronics, such as multiferroic
tunneling junction.
A 50 nm FO layer was first grown on SiC substrate, and then a BTO layer with
varied thickness was grown on this FO layer followed by growing another 50 nm FO layer
on top of the BTO layer. Two films with this ferrite/ferroelectric/ferrite structure have been
grown on SiC in this thesis, one is FO/BTO (3nm)/FO and the other is FO/BTO
(50nm)/FO. We carefully ensured that the only difference between these two films is the
BTO thickness, all the layers being grown under the same processing parameters.
The epitaxy and growth mode of FO/BTO/FO heterostructure were studied by
RHEED during the film deposition in [11 2 0] and [1100] directions of the SiC substrate.
As shown in Figure 93, the patterns of the first FO layer show streaks and sharp Kikuchi
lines (c & d), suggesting well ordered and smooth surface with 2D growth mode. The
RHEED patterns for the subsequent BTO layer show several Bragg-reflection spots
superposed on sharp streak diffraction patterns, suggesting that the BTO film deposited on
FO was grown in the Stranski-Krastanov growth mode, which is different as the 3D island
mode indicated by the spot-like diffraction pattern for the BTO films grown on SiC with a
MgO template in our previous work [187]. The lattice constant of Fe3O4 (8.396 Å) is
slightly smaller than twice that of MgO (4.212 Å), resulting in smaller lattice mismatch
202
with BTO (4.9% vs. 5.3%). Therefore, it is estimated that the first FO layer can function as
a template to enhance the crystallization properties of the BTO film, as well as a magnetic
layer for ME coupling. The second FO layer grown on BTO layer shows similar RHEED
patterns as the first FO layer grown on SiC, suggesting smooth FO surface.
(a)
(c)
(e)
(g)
(b)
(d)
(f)
(h)
Figure 93: RHEED patterns along [11 2 0] and [1 1 00] azimuths respectively from: H2
cleaned SiC (a & b), 50 nm FO deposited on SiC(c & d), 3 nm BTO deposited on FO (e
& f), and another 50 nm FO deposited on BTO (g & h).
The surface morphology of two FO/BTO (3nm)/FO and FO/BTO (50nm)/FO
heterostructures was investigated by AFM, as shown in Figure 94. It can be observed that
the surface of both films are smooth, dense and uniform, with RMS roughness of 1.1 nm ±
0.1 nm. The grain size of top FO film is about 80 nm for both films.
(a)
(b)
400 nm
400 nm
Figure 94: 2m×2m AFM images of (a) FO/BTO(3nm)/FO/SiC (0001), and (b)
FO/BTO(50nm)/FO/SiC(0001).
203
XRD characterizations were carried out to investigate the crystal orientation of the
FO/BTO/FO heterostructure. Figure 95 is the XRD –2 scanning results of FO/BTO
(3nm)/FO and FO/BTO (50nm)/FO on SiC substrates. Both films show highly (lll)
oriented FO and BTO layers without any evidence of any other secondary phases or
misorientations. The BTO (111) peak can be barely observed for the 3 nm BTO,
suggesting high quality epitaxial film. XRD confirms that the epitaxial structure of
FO/BTO/FO/SiC(0001) is in the mode of FO(111)/BTO(111)/FO(111/SiC(0001).
Fe3O4
111
Fe3O4/50nm BTO/Fe3O4/SiC
222
SiC 0006
Intensity (arb. units)
Fe3O4/3nm BTO/Fe3O4/SiC
Fe3O4
BTO
111
Fe3O4
333
*
Fe3O4
444
(b)
*
SiC 0012
(a)
10
20
30
40
50
60
70
80
2 Theta (degree)
Figure 95: XRD patterns of (a) FO/BTO (3nm)/FO/SiC, and (b)FO/BTO
(50nm)/FO/SiC heterostructure; * designates peaks from the Al sample stage.
The magnetic hysteresis loops of these two FO/BTO/FO heterostructures with
varied BTO thickness were characterized by VSM. As shown in Figure 96, both films
show in-plane easy axis and out-of-plane hard axis with coercities of ~ 200 Oe and ~ 450
Oe respectively, which are similar to the behavior of a single layer FO grown on SiC.
204
However, the in-plane magnetizaion process of the FO/BTO/FO heterostructure displays an
obvious BTO layer thickness dependence of the hysteresis loops, which may be resulted
4M (arb. Units)
from the strain or stress induced magnetic anisotropy.
FO/BTO(3nm)/FO
FO/BTO(50nm)/FO
Figure 96: Magnetic hysteresis loops by VSM measurement of FO/BTO (4nm)/FO/SiC
(black higher curve) and FO/BTO (50nm)/FO/SiC (red lower curve) heterostructure
at T ~300K after subtraction of the substrate diamagnetic contribution.
As a well known phenomenon, imposing a stress on a magnetic material can affect
magnetization process and produce an effective magnetic anisotropy field which is
expressed by H eff 
3
, where λ is magnetostriction constant; σ is applied stress which
Ms
could be compressive or tensile; Ms is magnetization. It can be concluded from this
equation that a positive λσ makes magnetization process easier and a negative λσ makes
magnetization harder. In our work, (111) oriented FO has a positive in-plane anisotropic
magnetostriction constant and a negative magnetostriction of -89 ppm alone out-of-plane
205
direction. BTO has a perovskite tetragonal structure and is ferroelectric between 278K and
393K, with lattice constants a = 3.992Å and c = 4.0361Å (293K) [184]. The FO layer
grown on BTO will be grown under a compressive stress, which can be further increased
with the increase of BTO film thickness. Considering a positive in-plane magnetostriction
constant of FO and an enhanced compressive stress as increasing the thickness of BTO,
the magnetization process becomes harder due to the negative λσ, which is consistent with
the experimental results shown in Figure 96.
The different magnetization behavior for these two films from the lattice mismatch
induced stress can cause FMR field shift, as shown in Figure 97. Here, a microwave cavity
operating at TE102 mode at X-band (9.53 GHz) was used to perform FMR measurements
of these FO/BTO/FO multiferroic heterostructure. Obviously, FMR field shifts from 1.1
kOe for pure FO/SiC to 1.6 kOe for FO/BTO (50nm) /FO/SiC structure (H = 500Oe).
This FMR field shift can be explained by the stress induced in-plane magnetic anisotropy
field Heff. The in-plane FMR frequency can be expressed by the well known Kittel
equation:
f /   ( H r  H k  Heff )( H r  H k  4M s  Heff )
With the appearance of a new effective magnetic field induced by the stress from
the lattice mismatch between the ferrite and ferroelectric layers, the FMR field has to
change accordingly to meet the Kittel equation.
206
FO/SiC
FO/BTO(3nm)/FO/SiC
FO/BTO(50nm)/FO/SiC
dP/dH (arb.units)
-1.2
-1.6
-2.0
-2.4
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
External magnetic field (kOe)
Figure 97: Ferromagnetic resonance absorption spectra of FO/BTO/FO/SiC with
BTO layer thickness ranging from 0 to 50 nm.
To demonstrate the magnetoelectric coupling in the FO/BTO (50nm)/FO
heterostructure, the structure was electrically poled with 5 V. First, the magnetic hysteresis
loops of an unpoled FO/BTO (50nm)/FO were measured by VSM at room temperature
(black curves in Figure 98). Afterward, the heterostructure was electrically poled along
out-of-plane with electric field of 5 V for 100 seconds. After electrical poling, the
magnetizaiotn of the film was measured by VSM again (shown as red curves in Figure 98).
As the result of electrical poling, the hysteresis loops shows a lower magnetization up to
about 5 kOe as the field is ramped up.
207
4M (arb. Units)
0.7
0.0
Pre pole
Post pole
-0.7
-5000
0
5000
Applied Field (kOe)
Figure 98: Magnetization curves taken before (black higher curve), and after (red
lower curve) electrical poling of the FO/BTO (50nm)/FO/SiC.
This electric poling dependence of magnetic hysteresis loop can be explained by
strain or stress mediated magnetoelectric coupling. Without poling, the spontaneous
polarization is randomly distributed showing an initial stage. However, after poling, the
spontaneous polarization is aligned along the poling direction and an in-plane compressive
stress can be achieved because of piezoelectric effect and a negative in-plane piezoelectric
coefficient [188]. This compressive stress can be transferred to FO layer and make
magnetization process harder which is consistent with the VSM loops showing in Figure
98.
Single crystalline, epitaxial Fe3O4 films and Fe3O4/BTO/Fe3O4 heterostructure have
been successfully grown by molecular beam epitaxy on 6H-SiC substrates. The successful
demonstration of electric field tuning of magnetism in Fe3O4/BTO/Fe3O4/SiC at
room-temperature opens new avenues for introducing ferrite/ferroelectric multiferroics
208
into magnetic tunable devices and spintronics. Particularly, the Fe3O4/BTO/Fe3O4/SiC
with thin BTO (2-3 nm) layer shows great prospects for application in multiferroic
tunneling junction (MTJ), which is combined by ferromagnetic tunneling junction and
ferroelectric tunneling junction, In MTJ, the tunneling resistance can be determined by
relative magnetization orientation and spontaneous polarization direction, thereby creating
four logical states of resistance towards novel information storages and spintronic devices.
However, the Fe3O4/BTO/Fe3O4 with thick BTO layer (> 50nm) can be potentially used to
modulate the magnetic anisotropy of a magnetite film (Fe3O4) by using the piezoelectric
response of BTO, and thus introduce more freedom in the multeferroic system.
209
6.0 CONCLUSIONS
The goal of this research was to effectively understand the nucleation and growth
mechanisms of magnetic ferrites (BaM and Fe3O4) on 6H-SiC in order to develop
ferrite/ferroelectric multiferroic heterostructure for next-gerneration microwave and
spinstronics devices by molecular beam epitaxy (MBE). The specific objectives included;
understanding the effects of processing parameters on the nucleation and growth
mechanisms of BaM on 6H-SiC by pulsed laser deposition, growth of single crystalline,
stoichiometric, and c-axis oriented BaM film on 6H-SiC by MBE, studying the structure
and compostion evolution at the BaM/SiC interface for engineering a more effective
interface,
and
to
demonstrate
ferrite
(Fe3O4)/ferroelectric(BaTiO3)
multiferroic
heterostructure with SiC and characterize the functional properties. By successful
demonstration of the multiferroic heterostructure, next-generation microwave and
spintronic device architectures can be developed and optimized for high-power,
high-frequency applications in harsh environments.
Three generations of interface engineering for the integration of barium hexaferrite
(BaM) films on 6H-SiC has culminated with BaM film deposited by PLD showing a
ferromagnetic resonance linewidth of less than 100 Oe, which can be used for integrated
devices. The highest quality BaM deposited by PLD was on a 10 nm single crystalline MgO
template grown by MBE on the SiC substrate. A systematic study of the interfaces for the
three generations shows that the chemistry, crystallinity, morphology and magnetization of
the BaM films deposited by PLD highly depends on the quality of the surface prior to PLD
210
deposition, and for the third generation, the thickness of the MgO templates. In contrast to a
10 nm MgO template, a 4 nm MgO template results in non-stoichiometric BaM films with
random orientation, three-dimensional film properties, and a coercivity larger than 1000 Oe.
Through the interface analysis, the deterioration of the BaM films on the 4 nm template may
be attributed to the reaction between the MgO and the BaM at the growth temperature of
915 oC, which depletes the MgO template and allows Si diffusion into the BaM.
Cross-section TEM reveals that the BaM films with the best magnetic properties were
epitaxially aligned on a transition layer of mostly a spinel MgFe2O4 structure that formed
between the BaM and the MgO template at the PLD growth temperature of 915 oC. Thus the
10 nm MgO template is necessary and not only forms a spinel transition layer that is a
similar structure to BaM, but is also is thick enough to provide a diffusion barrier throughout
the PLD deposition process in order to achieve device quality BaM on SiC. The combined
use of MBE to provide a high quality template followed by the growth of a thick film of
BaM by PLD is proven to be a simple and effective method to integrate device quality BaM
with SiC.
By using a 10 nm thick MgO template, single crystalline, epitaxial barium ferrite
films have been successfully grown by molecular beam epitaxy for the first time on 6H
silicon carbide substrates. Oxygen pressure is very critical for determining the chemistry
and surface structure of BaM. An oxygen deficient or rich environment will cause impurity
phase of Fe3O4 or -BaFe2O4 respectively. BaM films grown under an optimal oxygen
environment have a six-fold symmetry with the anisotropy highly oriented and
perpendicular to the substrate plane. XPS analysis showed that the use of a thin MgO
211
template in the early stages of film growth trapped the diffusing Si and prevented the
migration of Si into the film. AFM and MFM revealed a uniform smooth surface with
cluster-like magnetic domains. The high quality BaM grown by MBE has the potential to
function as a more effective template for subsequent near-perfect thick BaM grown by PLD
or LPE and to enable the tight stoichiometric control needed for more effective
property-coupling in ferrite/ferroelectric multiferroic heterostructures.
Using a high-quality MBE grown-BaM film as a seed layer, the quality of the
subsequent thick BaM film grown by PLD is further improved compared to the Gen III BaM
film grown on a 10 nm MBE-grown MgO template. This film exhibits lower coercivities at
both in-plane (180 Oe vs. 300 Oe) and perpendicular directions (680 Oe vs. 800 Oe),
suggesting better c-axis orientation and less crystal defects than the Gen III films.
In order to effectively integrate BaM-ferroelectric heterostructures on wide bandgap
semiconductors, the interface properties between the MBE-grown BaM and single-crystal
6H-SiC with a 10 nm MgO template have been investigated by RHEED, XPS, and
cross-section TEM. The in-situ RHEED and XPS suggest the formation of a spinel structure
during the initial stage of growth. The ex-situ cross-section transmission electron
microscopy and energy dispersive x-ray analysis confirm the high quality BaM film grown
epitaxially on the spinel structure with chemical composition of (Mg0.7Fe0.3)Fe2O4. The
formation of the spinel phase at the interface may be kinetically preferred due to the excess
MgO and relatively limited Ba flux during the initial growth conditions, and this spinel
structure is structurally favorable for subsequent BaM film growth. However, the impact of
high temperature processing on the continuing reaction between the BaM and the
212
MgO/Mg-ferrite layer suggests that a more chemically inert template is needed for the
extended high-temperature processing required for thick BaM films.
Single crystalline, epitaxial Fe3O4 films and Fe3O4/BTO/Fe3O4 heterostructure have
been successfully grown by molecular beam epitaxy on 6H-SiC substrates. The Fe3O4 film
exhibits high structrual order with sharp interfaces and an easy axis in-plane magnetization
with a coercivity of 200 Oe. In the Fe3O4/BTO/Fe3O4 heterostructure, the magnetoeletric
coupling is demonstrated at room-temperature by an electric field induced magnetic
anisotropy field change. The FO/BTO/FO/SiC with thin BTO (2-3 nm) layer shows great
prospects for application in multiferroic tunneling junction (MFTJ), which is combined by
ferromagnetic tunneling junction and ferroelectric tunneling junction. In MFTJ, the
tunneling resistance can be determined by relative magnetization orientation and
spontaneous polarization direction, thereby creating four logical state of resistance towards
novel information storages and spintronic devices. In addition, the FO/BTO/FO/SiC with
thick BTO layer (>50 nm) can be used to potentially modulate the magnetic anisotropy of
magnetite film (FO) by using the piezoelectric response of BTO, and thus introduce more
freedom in the multeferroic system by integrated with wide bandgap semiconductor.
This work demonstrates the integration of BaM and Fe3O4 with 6H-SiC by
molecular beam epitaxy the first time. By engineering an effective interface, high-quality
BaM on 6H-SiC has been achieved. The high quality BaM film have the potential to be
compatiable with the monolithic microwave integrate circuits and can also function as a
magnetic layer in BaM-ferroelectric multiferroic heterostructures for electrostatic FMR
tuning. Furthermore, the successful demonstration of electric field tuning of magnetism in
213
Fe3O4/BTO/Fe3O4/SiC at room-temperature opens new avenues for introducing
ferrite-ferroelectric multiferroics into tunable microwave magnetic devices and
spintronics.
214
7.0 RECOMMENDATIONS
Throughout the understandings of the research in this dissertation, this section will
discuss recommendations towards further understanding of the ferrite/ferroelectric
interface to engineer higher quality functional oxide layers, and thus achieve more
efficient coupling between each layer.
The impact of high temperature processing on the continuing reaction between the
BaM and the MgO template suggests that a more chemically inert template is needed for
the extended high-temperature processing required for thick BaM films. The thin spinel
templates such as Fe3O4 and MgFe2O4 can be considered to replace MgO for studying the
growth mechanism of spinel transition layer and further improve the BaM film quality by
optimizing both MBE and PLD parameters.
Explore the use of a conducting layer as a growth template for BaM to minimize
lattice mismatch and Si diffusion in order to incorporate the BaM//SiC film into the real
microwave device fabrication. The platinum (Pt) template is a good candidate to be used as
a conducting layer since Pt has been proven by other researchers [79] to promote the c-axis
oriented BaM growth and minimize the substrate diffusion into the film.
There are also several other works need to be continued in order to realize the
multiferroic tunneling junction for the Fe3O4/BTO/Fe3O4/SiC multilayer structure: (1) The
spontaneous polarization switching needs to be demonstrated in the multilayer structure
with BTO of 2-3 nm thick. Electric field switching of polarization in ferroelectric has been
well-studied in bulk. However, for ultra-thin ferroelectric film, whether the spontaneous
215
polarization does exist or can be switched is still unknown. Very recently, some simulation
and experimental results provide the edidence that the spontaneous polarization does exist
and can be switched by electric field in 2 nm BTO. (2) In order to realize the real
ferromagnetic tunneling junction, an anti-ferromagnetic pinning layer has to be considered
for coupling one Fe3O4 layer. One good candidate is anti-ferromagnetic Fe2O3, since it
should be easy to grow Fe2O3 and Fe3O4 by adjusting the atomic oxygen fluxes (through
changing O pressure and plasma power) in our MBE system. (3) The interface of
Fe3O4/BTO/Fe3O4/SiC has to be studied by TEM in order to well understand the grain
structures in different layers, since the defects (e.g., antiphase boundary in Fe3O4 layer) and
the interface mixing can impact the functionalities, and thus affect the tunneling effects and
the coupling between each layer. (4) The tunneling magnetoresistance has to be measured
through collaborations to find out the resistance change due to the electric field tuning and
the magnetic field tuning.
216
8.0 REFERENCE
1 Venevtsev, Y. N. and V. V.Gagulin, Ferroelectrics 1994, 162, 23.
2 Scott J. F., Multiferroic memories, Nature Materials, 6, pp. 256-257, 2007
3 Zheng, H., J. Wang, S. E. Lofland, Z. Ma, L. Mohaddes-Ardabili, T. Zhao, L.
Salamanca-Riba, S. R. Shinde, S. B. Ogale, F. Bai, D. Viehland, Y. Jia, D. G. Schlom, M.
Wuttig, A. Roytburd, and R. Ramesh, Multiferroic BaTiO3-CoFe2O4 Nanostructures,
Science, 303, pp. 661-663, 2003
4 Liu, M., O. Obi , J. Lou, Y. Chen, Z. Cai, S. Stoute, M. Espanol, M. Lew, X. Situ, K. S.
Ziemer, V. G. Harris, and N. X. Sun, Giant electric tuning of magnetic properties in
multiferroic ferrite/ferroelectric heterostructures, Adv. Funct. Mater., 19, pp. 1-9, 2009
5 Song, Y., J. Das, P. Krivosik, N. Mo, and C. E.Patton, Electric field tunable 60 GHz
ferromagnetic resonance response in barium ferrite-barium strontium titanate multiferroic
heterostructures, Appl. Phys. Lett. 94, pp. 182505, 2009
6 Kim, W. J., W. Chang, S. B. Qadri, H. D. Wu, J. M. Pond, S. W. Kirchoefer, H. S.
Newman, D. B. Chrisey, and J. S. Horwitz, Electrically and magnetically tunable
microwave device using (Ba, Sr) TiO3/Y3Fe5O12 multilayer, Appl. Phys. A: Mater. Sci.
Process., 71, pp. 7-10, 2000
7 Ustinov, A.B., V. S. Tiberkevich, G. Srinivasan, A. N. Slavin, A. A. Semenov, S. F.
Karmanenko, B.A. Kalinikos, J.V. Mantese, and R. Ramer, Electric field tunable
ferrite-ferroelectric hybrid wave microwave resonators: Experiment and theory, J. Appl.
Phys. 100, pp. 093905, 2006
8 Fetisov, Y. K. and G. Srinivasan, Electrically tunable ferrite-ferroelectric microwave
delay lines, Appl. Phys. Lett, 87, pp. 103502, 2005.
9 Semenov, A. A., S. F. Karmanenko, B. A. Kalinikos, G. Srinivasan, A. N. Slavin, and J. V.
Mantese, Dual-tunable hybrid wave ferrite-ferroelectric microwave resonator, Electron.
Lett. 42, pp. 641-642, 2006.
10 Heindl, R., H. Srikanth, S. Witanachchi, P. Mukherjee, T. Weller, A.S. Tatarenko, and G.
Srinivasan, Structure, magnetism, and tunable microwave properties of pulsed laser
deposition grown barium ferrite/barium strontium titanate bilayer films, J. Appl. Phys., 101,
pp. 09M503, 2007.
11 Chen, Z., A. Yang, S.D. Yoon, K.S. Ziemer, C. Vittoria, and V. G. Harris, Epitaxial
growth of Ba-hexaferrite films on single crystal 6-H SiC, J. Magn. Magn. Mater., 301, pp.
166-170, 2006.
12 S. Srinath, N.A. Frey, R. Heindl, H. Srikanth, K.R. Coffey, N.J. Dudney, Growth and
characterization of sputtered BSTO/BaM multilayers, J. Appl. Phys., 97, pp. 10J115, 2005.
13 Das, J., B. A. Kalinikos, A. R. Barman, and C. E. Patton, Multifunctional dual-tunable
low loss ferrite-ferroelctric heterostructures for microwave devices, Appl. Phys. Lett. 91,
pp. 172516, 2007
217
14 W. Wondrak, E. Niemann, K. Held, R. Constapel, and G. Kroetz, SiC devices for power
and high-temperature applications, Industrial Electronics, Proceedings. ISIE '98., 1, pp.
153-156, 1998.
15 Yorito Ota, High-frequency power applications using wide-bandgap semiconductors,
Electronics and Communications in Japan (Part II: Electronics), 81, pp. 54-61, 1998.
16 E Janzén, O Kordina, A Henry, W M Chen, N T Son, B Monemar, E Sörman, P Bergman,
C I Harris, R Yakimova, M Tuominen, A O Konstantinov, C Hallin and C Hemmingsson,
SiC – a semiconductor for high-power, high-temperature and high-frequency devices, Phys.
Scr., T54, 283-290, 1994.
17 Yanase, A. and K. Siratori, Band Structure in the High Temperature Phase of Fe3O4,
J. Phys. Soc. Jpn. 53, pp. 312-317, 1984
18 Zhang, Z. and S. Satpathy, Electron states, magnetism, and the Verwey transition in
magnetite, Phys. Rev. B 44, pp. 13319-13331, 1991.
19 Dedkov, Y., U. Rudiger, and Guntherodt, G., Weak ferromagnetism and excitonic
condensates, Phys. Rev. B 65, pp. 014428, 2002.
20 Huang, P.H., C.H. Lai, and R.T. Huang, Room-temperature growth of epitaxial (111)
Fe3O4 films with conductive Cu underlayer, J. Appl. Phys. 97, 10c311 pp 1-3, 2005.
21
22 Lisauskas, V., K. Sliuziene, R. Butkute, S. Tamulevicius, M. Andrulevicius, and B.
Vengalis, Study of p-Fe3O4/n-(GaAs, Si) heterostructures: Fabrication and physical
properties, Acta physica polonica A, 113 (3), pp1055-1058, 2008.
23 Suzuki, Y., Epitaxial spinel ferrite thin films, Annu. Rev. Mater. Res. 31, pp.265-289,
2001.
24 Jakubovics, J. P., Magnetism and Magnetic Materials, 2nd edition, The Institute of
Materials, London, 1994, Chapt. 2.
25 http://www.aacg.bham.ac.uk/magnetic_materials/type.htm
26 H. A. Kramers, Physica, 1, 182 (1934); L. Néel, Ann. Phys. (Paris), 3, 137 (1948)
P. W. Anderson, Phys. Rev., 79, 350 (1950); J. H. van Vleck, J. Phys. Rad., 12, 1178.
27 Wyckoff, R., Crystal Structure, Wiley, New York, 1963.
28 Kim, Y. J., Y. Gao, and S. A. Chambers, Selective growth and characterization of pure,
epitaxial Fe2O3 and Fe3O4 (001) films by plasma-assisted molecular beam epitaxy,
surface science 371, pp. 358-370, 1997.
29 Eerenstein, W., Spin-dependent transport across anti-phase boundaries in magnetite
films, Thesis (Ph. D.), University of Groningen, 2003.
30 Cox, P.A., Transition Metal Oxides, An Introduction to Their Electronic Structure and
Properties, Clarendon Press, Oxford, 1992.
31 Samsonov, G., The Oxide Handbook, Plenum, New York, 1982.
32 Verwey, E.J.W., Electronic Conduction of Magnetite (Fe3O4) and its Transition Point
at Low Temperatures, Nature, 144, pp. 327-328, 1939
33 Verwey, E.J.W., P.W. Haayman, and F.C. Romeijan, Physical Properties and Cation
Arrangement of Oxides with Spinel Structures II. Electronic Conductivity, J. Chem. Phys.
15, pp. 181-187, 1947
218
34 Wright, J.P., J.P. Attfield, and P.G. Radaelli, Long Range Charge Ordering in
Magnetite Below the Verwey Transition, Phys. Rev. Lett. 87 pp. 266401, 2001.
35 Kundig, W. and R.S. Hargrove, Electron hopping in magnetite, Solid State Comm. 7,
pp. 223-227, 1969.
36 Arag´on, R., D. J. Buttrey, J.P. Sheperd, and J.M. Honig, Influence of nonstoichiometry
on the Verwey transition, Phys. Rev. B, 31, pp. 430-436, 1985.
37 Eerenstein, W., Spin-dependent transport across anti-phase boundaries in magnetite
films, Thesis (Ph. D.), University of Groningen, 2003.
38 Moussy, J.-B., S. Gota, A. Bataille, M.-J. Guittet, M. Gautier-Soyer, F. Delille, B. Dieny,
F. Ott, T.D. Doan, P. Warin, P. Bayle-Guillemaud, C. Gatel, and E. Snoeck, Thickness
dependence of anomalous magnetic behavior in epitaxial Fe3O4(111) thin films: Effect of
density of antiphase boundaries, Phys. Rev. B 70, pp174448, 2004.
39 Liu, H., E.Y. Jiang, R.K. Zheng, and H. L. Bai, structure and magnetic properties of
polycrystalline Fe3O4 films deposited by reactive sputtering at room temperature, Phys.
Stat. Sol. (a) 201, No4, pp 739-744, 2004.
40 Jain, S., A.O. Adeyeye, D.Y. Dai, Magnetic properties of half-metallic Fe3O4 films, J.
Appl. Phys., 95 (11), pp. 7237-7239, 2004.
41 Vohlfarth, E.P., Ferromagnetic Materials, vol. 3, North-Holland Publishing Company,
New York, 1982.
42 Landolt-Bornstein, Numerical data and functional relationships in science and
technology, Vol 4, Part b, Editors: K.H. Hellwege and A.M. Hellwege, Springer-Verlag,
Berlin, 1970.
43 Robert C. O'Handley, Modern magnetic materials : principles and applications, Wiley,
New York, 2000.
44 Shi, P., Development of microwave and millimeter wave integrated
semiconductor-ferrite devices, Thesis (Ph. D.),Northeastern University, 2000.
45 Hellwege,K. H., Magnetic and Other Properties of Oxide and Related
Compounds, Landolt-Börnstein, Vol. 4, edited by K.-H. Hellwege, Springer, New York, pp.
562. 1970.
46 http://www.ndt-ed.org/EducationResources/CommunityCollege/MagParticle/Physics/H
ysteresisLoop.htm
47 http://www.ee.surrey.ac.uk/Workshop/advice/coils/mu/
48 Yoon, S.D. and C. Vittoria, Preparation of High-quality hexaferrite thick films by an
improved liquid vapor epitaxy deposition technique, IEEE trans Mag., 30(5), pp. 3163-3165,
2003
49 Hurben, M.J., D. R. Franklin, and C. E. Patton, Angle dependence of the ferromagnetic
resonance linewidth in easy-axis and easy-plane single crystal hexagonal ferrite disks, J.
Appl. Phys. 81(11), pp. 7458-7467, 1997.
50 Karim,R., S. D.Ball, J. R. Truedson, and C. E Patton, Frequency dependence of the
ferromagnetic resonance linewidth and effective linewidth in maganese substituted single
crystal barium ferrite, J. Appl. Phys. 73(9), pp. 4512-4515, 1993.
51 Shinde, S. R., S. E. Lofland, C. S. Ganpule, S. B. Ogale, S. M. Bhagat, T.Venkatesan,
and R. Ramesh, Realization of epitaxial barium ferrite films of high crtystalline quality with
219
small resonance loss, J. Appl. Phys., 85(10), pp.7459-7466, 1999.
52 Sparks, R., R. Loudon, and C. Kittel, Ferromagnetic Relaxation. I. Theory of the
Relaxation of the Uniform Precession and the Degenerate Spectrum in Insulators at Low
Temperatures, Phys. Rev. 122, pp. 791-803, 1961.
53 Song, Y. Y., S. Kalarickal, and C. E. Patton, Optimized pulsed laser deposited barium
ferrite thin films with narrow ferromagnetic resonance linewidths, J. Appl. Phys., 94, pp.
5103-5110, 2003.
54 Yoon, S. D. and C. Vittoria, Thick M-type barium hexaferrite films grown on garnet
substrate, J. Appl. Phys., 96 (4), pp. 2131-2135, 2004.
55 Karim, R., K. D. McKinstry, J. R. Truedson, and C. E. Patton, Frequency dependence of
the FMR linewidth in single crystal barium ferrite platelets,IEEE Trans. Magn. 28, pp.
3225-3227, 1992.
56 Sarma, S. D., Spintronics, American Scientist, 89, pp. 516-523, 2001.
57 Ramesh, R. and N. A. Spaldin, Multiferroics: progress and prospects in thin films,
Nature Mater. 6, pp. 21-29, 2007.
58 Cheong, S.-W. and M. Mostovoy, Multiferroics: a magnetic twist for ferroelectricity,
Nature Mater. 6, pp. 13-20, 2007.
59 Schmid, H., Multi-ferroic magnetoelectrics, Ferroelectrics 162, pp. 317-338, 1994.
60 Martin, L. W., Engineering Multiferroic Materials and New Functionalities in
Materials, Thesis (Ph. D.), University of California, Berkeley, 2008.
61 Aizu, K., Possible Species of Ferromagnetic, Ferroelectric, and Ferroelastic Crystals
Phys. Rev. B, 2, pp. 754-772, 1970.
62 Hill, N. A., Why are there so few magnetic ferroelectrics? J. Phys. Chem. B 104,
pp. 6694-6709, 2000.
63 Eerenstein, W., N. D. Mathur, and J. F. Scott, Multiferroic and magnetoelectric
materials, Nature, 442, pp. 759-765, 2006.
64 Heckmann, Die gittertheorie der festen korper, Ergeb. Exakten. Naturwiss, 4, pp.
100-153, 1925.
65 Nye, J.F., Physical properties of crystals. Oxford: Oxford University Press, 1985.
66 Ziemer, K. S., “Studies of the initial stage of silicon carbide growth on silicon”, Ph.D.
Dissertation, Morgantown, WV: West Virginia University, 2001.
67 http://neon.mems.cmu.edu/skowronski/Stacking%20sequences.htm
68 Ohring,M., Materials Science of Thin films Deposition and Structure, 2nd edition,
Academic Press, London, 2002.
69 Doyle, B., Investigation of MgO Film Growth on GaAs as a Potential Spin-Aligned
Tunnel Barrier, Master Thesis, Boston, MA, Northeastern University, 2005.
70 http://www.ioffe.rssi.ru/SVA/NSM/Semicond/SiC/thermal.html
71 Dorsey, P. C., S. B. Qadri, J. L. Feldman, J. S. Horwitz, P. Lubitz, D. B. Chrisey, and J.
B. Ings, Coefficients of thermal expansion for barium hexaferrite, J. Appl. Phys., 79, pp.
3517-3520, 1996.
72 www.fkf.mpg.de/ga/starke/ lehre/2dphysik/kap3/kap31ff.pdf
73 Goodrich, T.G., “Atomistic investigation into the interface engineering and
heteroepitaxy of functional oxides on hexagonal silicon carbide through use of a
220
magnesium oxide template layer for the development of a multifunctional heterostructure”,
Ph.D. Dissertation, Boston, MA, Northeastern University, 2008.
74 Wang, S. G., S. D. Yoon, and C. Vittoria, “Microwave and magnetic properties of
double-sided hexaferrite films on (111) magnesium oxide substrates”, J. Appl. Phys., 92,
pp.6728-6732, 2002.
75 Herman, M.A., Sitter, H., Molecular Beam Epitaxy Fundamentals and Current Status,
2nd, Rev. and Updated Ed., Springer, New York, 1996.
76 Shinde, S. R., R. Ramesh, S. E. Lofland, S. M. Bhagat, S. B. Ogale, R. P. Sharma, and T.
Venkatesan, Effect of lattice mismatch strains on the structural and magnetic properties of
barium ferrite films, Appl. Phys. Lett., 72, pp. 3443-3445, 1998.
77 Yoon, S. D., Vittoria C., and Oliver, S.A., Variation in material parameters for
hexagonal barium ferrite films on (111) magnesium oxide as a function of oxygen pressure,
J. Appl.Phys., 92, pp. 6733-6738, 2002.
78 Liu, X., Jianmin Bai, Fulin Wei, Yang ZhengLiu, A. Morisako, and M. Matsumoto, The
study of sputtered barium ferrite thin films”, Phys. stat. sol.(a), 174, pp. 389-394, 1999.
79 Zhuang, Z. L., M. Rao, D. E., Laughlin, and Kryder M. H., The effect of Pt templates on
the magnetic and structural properties of perpendicularly oriented barium ferrite media, J.
Appl. Phys., 85, pp. 6142-6144, 1999.
80 Wee, A.T.A., J. P. Wang, A.C.H. Huan, L. P. Tan, R. Gopalakrishnan, and K. L.Tan,
Template and substrate effects in RF-Magnetron Sputtered Barium Ferrite thin film media,
IEEE Transactions on Magnetics, 33, pp. 2986-2988, 1997.
81 Hoshi, Y., Y. Kubota, and M. Naoe, Control of orientation and crystallite size of barium
ferrite thin films in sputter deposition, IEEE Trans.Magn. 31, pp.2782-2784,1995.
82 Yuan, M. S., H. L. Glass, and L. R. Adkins, Epitaxial barium hexaferrite on sapphire by
sputtering deposition, Appl. Phys. Lett., 53, pp.340-341, 1988.
83 Zhuang, Z. L., M. Rao, R. M. White, D. E. Laughlin, and M. H. Kryder, Barium ferrite
thin film media with perpendicular c-axis orientation and small grain size, J. Appl. Phys.,
87, pp.6370-6372, 2000.
84 Dorsey, P. C., S. B. Qadri, J. L. Feldman, J. S. Horwitz, P. Lubitz, D. B. Chrisey, and J.
B. Ings, Coefficients of thermal expansion for barium hexaferrite, J. Appl. Phys., 79,
pp.3517-3520, 1996.
85 Shackelford, J. F., CRC Materials Science and Engineering Handbook, 1994.
86 How, H., S. A. Oliver, S. W. McKnight, P. M.Zavracky, N. E.McGruer, C. Vittoria, and
R. Schmidt, Theory and experiment of thin-film junction circulator, IEEE Trans.
Microwave Theory Tech., 46, pp. 1645-1653,1998.
87 Oliver, S. A., S. D. Yoon, I. Kozulin, M. L. Chen, and C. Vittoria, Growth and
characterization of thick oriented barium hexaferrite films on MgO (111) substrates, Appl.
Phys. Lett., 76, pp. 3612-3614, 2000.
88 Hylton, T. L., Parker, M. A., Ullah, M., Coffey, K. R., Umphress, R., and Howard, J. K.,
Ba-ferrite thin film media for high-density longitudinal recording, J. Appl. Phys., 75, pp.
5960-5965, 1994.
89 Hiratsuka, N., K. Yazawa, and K. Kakizaki, High coercivity force of sputtered nickel
ferrite film, J.Magn.Soc. Jpn. 21, pp. 425-428, 1997.
221
90 Chen, Y. J., M. Rao, R. M. White, D. E. Laughlin, and M. H. Kryder, Barium ferrite thin
films without a dead layer, IEEE Transactions on Magnetics, 34, pp.1648 – 1650, 1998.
91 Morisako, A., X. X. Liu, M. Matsumoto, and M. Naoe, The effect of template for
Ba-ferrite sputtered films on c-axis orientation, J. Appl. Phys., 81, pp. 4374-4376, 1997.
92 Chen, Y., Bain, J.A., and Kryder, M.H., J.Magn. Soc. Jpn. 21, Supp., 65 , 1997.
93 Kim, D. H., I. T. Nam, Y. K. Hong, S. H. Gee, and M. H. Park, Microstructure and
magnetic properties of hexagonal barium ferrite thin films with various templates, J. Appl.
Phys., 91, pp. 8751-8753, 2002.
94 Doh, S. J., S. D. Choi, J. H. Je, D.Y. Noh, and T. S. Cho, Inhibition of abnormal acicular
grain growth of nanostructured Ba-ferrite films by secondary Fe2O3 phase, J. Appl.
Phys., 91, pp. 7869-7871, 2002.
95 Yan, L., C. M. Lopez, R. P. Shrestha, and E. A. Irene, Magnesium oxide as a candidate
high- gate dielectric, Appl. Phys. Lett., 88, pp.142901, 2006.
96 Yuasa, S., T. Nagahama, A. Fukushima, Y. Suzuki, and K. Ando, Giant
room-temperature magnetoresistance in single-crystal Fe/MgO/Fe magnetic tunnel
junctions, Nature. Mater. 3, pp868-871, 2004.
97 Fork, D.K., F. A. Ponce, T. C. Tramontana, and T. H. Geballe, Epitaxial MgO on Si(001)
for Y-Ba-Cu-O thin-film growth by pulsed laser deposition, Appl. Phys. Lett., 58, pp. 2294,
1991.
98 Cho, T. S., S. J. Doh, J. H. Je, and D.Y. Noh, Thickness dependence of the crystallization
of Ba-ferrite films, J. Appl. Phys., 86, pp. 1958-1963, 1999.
99 Hawley, M., I. Raistrick, J. Beery, and R. J. Hulton, Growth Mechanism of Sputtered
Films of YBa2Cu3O7 Studied by Scanning Tunneling Microscopy, Science 215, pp.
1587-1589, 1991.
100 Huang, F., J. J. Wolfgang, T. M. Le, D. N. Lambeth, and D. D. Stancil, Oxygen
pressure dependence of laser deposited barium ferrite films on LLC(111), IEEE
Transactions on Magnetics, 31, pp. 3826-3828, 1995.
101 Lisfi, A., J. C. Lodder, E. G. Keim, and C. M. Williams, Evidence of stress anisotropy
and role of oxygen pressure in growth of pulsed-laser-deposited hexaferrite films, Appl.
Phys. Lett., 82 (1), pp. 76-78, 2003.
102 Liu, X. H., M. H. Hong, W. D. Song, G. X. Chen, J. F. Chong, J. P. Wang, Y. H. Wu,
and T. C. Chong, Pulsed laser deposition of oriented barium ferrite (BaFe12O19) thin films,
Appl. Phys. A 78, pp. 423-425, 2004.
103 Zhuang, Z. L., B. Bian, M. Rao, R. M. White, and D. E. Laughlin, Effect of
stoibhiometry on the magnetic and structural properties of perpendicular barium ferrite
thin film media, IEEE Transactions on Magnetics, 36, pp. 2405-2407, 2000.
104 Wang, S. G., S. D. Yoon, and C. Vittoria, Microwave and magnetic properties of
double-sided hexaferrite films on (111) magnesium oxide substrates, J. Appl. Phys., 92, pp.
6728-6732, 2002.
105 Yoon, S. D. and C. Vittoria, Thick M-type barium hexaferrite films grown on garnet
substrate, J. Appl. Phys., 96, pp. 2131-2135, 2004.
106 Yuan, M. S., H. L. Glass and L. R. Adkins, Epitaxial barium hexaferrite on sapphire by
sputter deposition, Appl. Phys. Lett. 53, pp. 340, 1988.
222
107 Chambers, S.A., epitaxy growth and properties of thin film oxides, Surface Science
Reports, 39, pp.105-180, 2000.
108 Gao, Y. and S. A. Chambers, Heteroepitaxial growth of Fe2O3, Fe2O3 and
Fe3O4 thin films by oxygen-plasma-assisted molecular beam epitaxy, Journal of Crystal
Growth, 174, pp. 446-454, 1997.
109 Gota, S., E. Guiot, M. Henriot, and M.G. Soyer, Atomic-oxygen-assisted MBE growth
of Fe2O3 on Al2O3(0001):Metastable FeO(111)-like phase at subnanometer
thicknesses, Phys. Rev. B, 60, pp.14387-14395, 1999.
110 Niedrig, T.S., W. Weiss, and R. Schlogl, Electronic structure of ultrathin ordered iron
oxide films grown onto Pt(111), Phys. Rev. B, 52, pp.17449-17460, 1995.
111 Mills, P. and J. L. Sullivan, A study of the core level electrons in iron and its three
oxides by means of X-ray photoelectron spectroscopy, J.Phys.D: Appl. Phy., 16, pp.723-732,
1983.
112 Brundle, C.R., T. J.Chuang, and K. Wandelt, Core and valence level photoemission
studies of iron oxide surfaces and the oxidation of iron, Surf. Sci., 68, pp459-468, 1977
113 Weiss, W. and M. Ritter, Metal oxide heteroepitaxy: Stranski-Krastanov growth for
iron oxides on Pt(111)”, Phys. Rev. B, 59, pp. 5201-5213, 1999.
114 Galloway, H.C., P. Sautet, and M. Salmeron, Structure and contrast in scanning
tunneling microscopy of oxides: FeO monolayer on Pt(111), Phys. Rev. B, 54, pp.
R11145-11148, 1996.
115 Shaikhutdinov, S. K. and W. Weiss, Oxygen pressure dependence of the α-Fe2O3(0001)
surface structure, Surf. Sci., 432, pp. L627-634, 1999.
116 Fujii, T., F.M.F.de. Groot, G. A. Sawatzky, F. C. Voogt, T. Hibma, and K. Okada,
In-situ XPS analysis of various iron oxide grown by NO2-assisted molecular-beam epitaxy,
Phys. Rev. B, 59, pp 3195-3202, 1999.
117 Fujii, T., D. Alders, F. C. Voogt, T. Hibma, B. T. Thole, and G. A. Sawatzky, In-situ
RHEED and XPS studies of epitaxial thin α-Fe2O3(0001) films on sapphire, Surf. Sci. 366,
pp. 579-586 ,1996.
118 Lad, R. J. and V. E. Henrich, Structure of α-Fe2O3 single crystal surfaces following Ar+
ion bombardment and annealing in O2, Surf. Sci., 193, pp. 81-93, 1988.
119 Vurens, G. H., M. Salmeron, and G. A. Somorjai, Structure, composition and
chemisorption studies of thin ordered iron oxide films on platinum (111), Surf. Sci., 201, pp.
129-144, 1988; Weiss, W. and G. A. Somorjai, Preparation and structure of 1–8 monolayer
thick epitaxial iron oxide films grown on Pt(111), J. Vac. Sci. Technol.A, 11, pp. 2138 1993.
120 Thevuthasan, S., W. Jiang, D. E. McCready, S. A. Chambers, Rutherford
backscattering and channeling studies of Mg and Fe diffusion at the interface of
-Fe2O3(001)/MgO(001), Surf. Interf. Anal., 27, pp. 194-198, 1999.
121 Yi, S. I., Y. Liang, and S. A. Chambers, Morphological and structural investigation of
the early stages of epitaxial growth of α-Fe2O3 (0001) on α-Al2O3 (0001) by
oxygen-plasma-assisted MBE, Surf. Sci., 443, pp. 212-220, 1999.
122 Chambers, S.A. and S. A. Joyce, Surface termination, composition and reconstruction
of Fe3O4 (001) and Fe2O3 (001), Surf. Sci., 420, pp. 111-122, 1999.
223
123 Gao, Y. Y. J. Kim, and S. A. Chambers, Preparation and characterization of epitaxial
iron oxide films, J. Mater. Res., 13, pp. 2003, 1998.
124 Lu, Y. X., J. S. Claydon, Y. B. Xu, S. M. Thompson, K. Wilson, and G. van der Laan,
Epitaxial growth and magnetic properties of half-metallic Fe3O4 on GaAs(100), Phys. Rev.
B 70, pp. 233304, 2004.
125 Zhou, W.L., Granular growth of Fe3O4 thin films and its antiphase boundaries
prepared by pulsed laser deposition, J. Appl. Phys. 89, pp. 7398, 2001.
126 Reisinger, D., M. Schonecke, T. Brenninger, M. Opel, A. Erb, L. Alff and R. Gross,
Epitaxy of Fe3O4 on Si (001) by pulsed laser deposition using a TiN/MgO buffer layer, J.
Appl. Phys. 94, 3, pp1857-1863, 2001.
127 Mariotto, G., S. Murphy and I.V. Shvets, Charge ordering on the surface of Fe3O4(001),
Phys. Rev. B, 66, pp 245426-245433, 2002.
128 Vescovo, E., H.-J. Kim, J. M. Ablett, and S. A. Chambers, Spin-polarized conduction
in localized ferromagnetic materials: The case of Fe3O4 on MgO(100), J. Appl. Phys., 98,
pp. 084507, 2005.
129 Dedkov, Y.S., U. Rudiger, and G. Guntherodt, Evidence for the half-metallic
ferromagnetic state of Fe3O4 by spin-resolved photoelectron spectroscopy, Phys. Rev. B,
65, pp. 064417-064421, 2002.
130 Tsymbal, E. Y. and H. Kohlstedt, Tunneling Across a Ferroelectric, Science, 313, pp.
181-183, 2006.
131 Zhai, J., Z. Xing, S. X. Dong, J. F. Li, and D. Viehland, Detection of pico-Tesla
magnetic fields using magneto-electric sensors at room temperature, Appl. Phys. Lett., 88,
pp. 062510, 2006.
132 Dong, S. X., J. Y. Zhai, J. F. Li, and D. Viehland, Near-ideal magnetoelectricity in
high-permeability magnetostrictive/piezofiber laminates with a 2-1 connectivity, Appl.
Phys. Lett., 89, pp. 252904, 2006.
133 Zhao, P., Z. Zhao, D. Hunter, R. Suchoski, C. Gao, S. Mathews, M. Wuttig, and I.
Takeuchi, Fabrication and characterization of all-thin-film magnetoelectric sensors, Appl.
Phys. Lett., 94, pp. 243507, 2009.
134 Liu, M., O. Obi, J. Lou, Y. Chen, Z. Cai, S. Stoute, M. Espanol, M. Lew, X. Situ, K.
S. Ziemer, V. G. Harris and Nian X. Sun, Giant Electric Field Tuning of Magnetic
Properties in Multiferroic ferrite/ferroelectric Heterostructures, Advanced Functional
Materials, 19, pp.1826, 2009.
135 Srinivasan, G., A. S. Tatarenko, Y. K. Fetisov, V. Gheevarughese, and M. I.
Bichurin, Microwave magneto-electric Interactions in Multiferroics, Mater. Res. Soc.
Symp. Proc., 966, pp. 0966-T14-01, 2007.
136 Velev, J. P., C. Duan, J. D. Burton, A. Smogunov, M. K. Niranjan, E. Tosatti, S. S.
Jaswal, and E. Y. Tsymbal, Magnetic Tunnel Junctions with Ferroelectric Barriers:
Prediction of Four Resistance States from First Principles, Nano Lett., 9, pp. 427, 2009.
137 Theis, C. D. and D. G. Schlom, Cheap and stable titanium source for use in oxide
molecular beam epitaxy systems, Journal of Vacuum Science Technology A, 14, pp.
2677-2679, 1996.
224
138 Goodrich, T. L., J. Parisi, Z. Cai and K. S. Ziemer, Low temperature growth of
crystalline magnesium oxide on hexagonal silicon carbide (0001) by molecular beam
epitaxy, Appl. Phys. Lett., 90, pp. 042910, 2007.
139 Krstulovic, N., I. Labazan, S. Milosevic, U. Cvelbar, A. Vesel, and M, Mozetic Optical
emission spectroscopy characterization of oxygen plasma during treatment of a PET foil, J.
Phys. D: Appl. Phys., 39, pp. 3799-3804, 2006.
140 Chen, Z., A. Yang, Z. Cai, S. D.Yoon, K.S. Ziemer, C. Vittoria, and V. G. Harris,
Structure and magnetism of Ba-hexaferrite films grown on single crystal 6-H SiC, IEEE
Trans. Magn., 42, pp. 2855, 2006.
141 Z. Chen, A. Yang, A. Gieler, V.G. Harris, C. Vittoria, P.R. Ohodnicki, K.Y. Goh, M. E.
Mchenry, Z. Cai, T. L. Goodrich, and K. S. Ziemer, Epitaxial growth of M-type
Ba-hexaferrite films on MgO (111)||SiC (0001) with low ferromagnetic resonance
linewidths, Appl. Phys. Lett. 91, pp. 182505, 2007.
142 NIST Database 82, U.S. Department of Commerce 2001.
143 Wagner, C.D., Riggs, W.M., Davis, L.E., Moulder, J.F., and Muilenberg, G.E.,
handbook of x-ray photoelectron spectroscopy: Perkin-Elmer Corporation, 1979.
144 http://www.las.inpe.br/~cesar/Infrared/rheed.htm
145 Ambios Technology 2SAAVO USPM Menu.
146 Castro, S., M. Gayoso and A. C. Rodriguez, A Study of the Combustion Method to
Prepare Fine Ferrite Particles, J. Solid State Chem., 134, pp. 227-231, 1997.
147 Rakshit, S. K., S. C. Parida, Z. Singhand, R. Prasad, and V. Venugopal,
Thermodynamic properties of ternary oxides in the system Ba–Fe–O using solid-state
electrochemical cells with oxide and fluoride ion conducting electrolytes, J. Solid State
Chem., 177, pp. 1146-1156, 2004.
148 National Institute of Standards and Technology (NIST) Standard Reference Database,
http://www.nist.gov/srd/
149 Kamzin, A. S., V. L. Rozenbaum, L. P. Ol’khovik, E. D. Kovtun, Diagnostics of the
surface and near-surface region of barium hexaferrite monocrystals, J.Magn. Magn. Mater.,
161, pp. 139-142, 1996.
150 Tang, W. M., Z. X. Zheng, H. F. Ding and Z. H. Jin, A study of the solid state reaction
between silicon carbide and iron, Mater. Chem. Phys., 74, pp 258-264, 2002.
151 Zarpellon, J., H. F. Jurca, J. J. Klein, W. H. Schreiner, N. Mattoso, and D. H. Mosca,
Iron Silicide Formation from Fe Thin-Film Electrodeposition on Hydrogen-Terminated
Si(111), Journal of The Electrochemical Society, 152, pp. C808-C812, 2005.
152 Soukiassian, P. and F. Amy, Silicon carbide surface oxidation and SiO2/SiC interface
formation investigated by soft X-ray synchrotron radiation, J. Electron Spectroscopy and
Related Phenomena, 144-147, pp. 783-788, 2005.
153 Pankov,V., A. Bartholdson, O. Stukalov, S. Smolenchuk, O. Bavushkin, V. Gremenok,
Growth of BaFe12O19 thin films formed by reactive diffusion, J. Crystal Growth., 252, pp.
382-390, 2003.
154 Discussion with Dr. Vince Harris
225
155 Liu, X. H., M. H. Hong, W. D. Song, G. X. Chen, H. M. J. Lam, J. P. Wang, and T.
C. Chong, Barium ferrite (BaFe12O19) thin films prepared by pulsed laser deposition on
MgO buffered Si substrates, Appl. Phys.A, 80, pp. 611-614, 2005.
156 Papakonstantinou, P., M. C. O’Neill, R. Atkinson, R. Al-Wazzan, T. Morrow, and I.W.
Salter, Influence of oxygen pressure on the expansion dynamics of Ba-hexaferrite ablation
plumes and on the properties of deposited thin films, J.Magn. Magn. Mater., 189, pp.
120-129, 1998.
157 Shimada, S., R. Furuichi, and T. Ishii, Effects of additives on solid state reaction. III.
Effects of halide additives on the formation of MgFe2O4, Bulletin of the chemical society of
Japan, 49, pp. 1289-1293, 1976.
158 Goodrich, T. L., Z. Cai, and K. S. Ziemer, Exploring stability of MgO(111) films grown
on 6H-SiC(0001) by molecular beam epitaxy for two-step integration of functional oxides,
Applied Surface Science, 254, pp. 3191-3199, 2008.
159 Li, J., R. Sinclair, S. S. Rosenblum, and H. Hayashi, Reaction-mediated texturing of
barium ferrite magnetic thin films on ZnO underlayer, J. Mater. Res., 10, pp. 2343-2349,
1995.
160 Glass, J. L. and J. H. W. Liaw, Growth and characterization of LPE hexagonal ferrites,
J. Appl. Phys., 49, pp. 1578-1581, 1978.
161 Morisako, A., M. Matsumoto and M. Naoe, c-axis orientation of hexagonal ferrite
films prepared by RF diode sputtering, J. Magn. Magn. Mater., 54-57, pp. 1657-1658.
(1986)
162 Smit, J. and H. P. J.Wijn, Ferrites (Wiley, New York, 1959), pp. 177-183.
163 Chikazumi, S., Physics of Ferromagnetism, 2nd ed. (Oxford Science, New York, 1997),
pp. 210-383.
164 Zhang, Z. and M. G. Lagally, Atomistic Processes in the Early Stages of Thin-Film
Growth, Science, 276, pp. 377-383, 1997.
165 Ning, M., Mi, Y.Y., Ong, C.K., Lim, P.C. and Wang, S.J., Growth studies of (220),
(200), and (111) oriented MgO films on Si (001) without buffer layer”, J. Phys. D: Appl.
Phys., 40, pp. 3678-3682, 2007.
166 Chen, Y., A. L. Geiler, T. Chen, T. Sakai, C. Vittoria and V. G. Harris, Low-loss
barium ferrite quasi-single-crystals for microwave application, J. Appl. Phys., 101, pp.
09M501, 2007.
167 Van Hook H. J., Thermal Stability of Barium Ferrite (BaFe12O19), J. Am. Ceram. Soc.
47, pp. 579-581, 1964.
168 Pullar, R. C. and A. K. Bhattacharya, Crystallisation of hexagonal M ferrites from a
stoichiometric sol–gel precursor, without formation of the α-BaFe2O4 intermediate phase,
Mater. Lett., 57, pp. 537-542, 2002.
169 Zhong, W., W. Ding, N. Zhang, J. Hong, Q. Yan, and Y. Du, Key step in synthesis of
ultrafine BaFe12O19 by sol-gel technique, J. Magn. Magn. Mater., 168, pp. 196-202, 1997.
170 Lisfi, A. and J. C. Lodder, Relation between the microstructure and magnetic
properties of BaFe12O19 thin films grown on various substrates, J. Magn. Magn. Mater.,
242-245, pp. 391-394, 2002.
226
171 Lisfi, A. and J. C. Lodder, Magnetic domains in epitaxial BaFe12O19 thin films with
perpendicular anisotropy, J. Phys.: Condens. Matter., 14, pp. 12339-12352, 2002.
172 Zhang, L., M. Kuhn, and U. Diebold, Growth, structure and thermal properties of
chromium oxide films on Pt(111), Surf. Sci., 375, pp. 1-12, 1997.
173 Robbert, P. S., H. Geisler, C. A. Ventrice, Jr., J. van Ek, S. Chaturvedi, J. A. Rodriguez,
M. Kuhn, and U. Diebold, Novel electronic and magnetic properties of ultrathin chromium
oxide films grown on Pt(111), J. Vac. Sci. Technol. A, 16, pp. 990-995, 1998.
174 Chambers, S. A. and T. Droubay, Role of oxide ionicity in electronic screening at
oxide/metal interfaces, Physical Review B, 64, pp. 075410, 2001.
175 Gao, Y., Y. J. Kim, S. Thevuthasan, S. A. Chambers, and P. Lubitz, Growth, structure,
and magnetic properties of -Fe2O3 epitaxial films on MgO, J. Appl. Phys., 81, pp.
3253-3256, 1997.
176 Wells, A. F., Structural Inorganic Chemistry (3rd edn, Clarandon Press: Oxford, 1962)
pp. 488
177 Passerini, L., Gazz. Chim. Ital., 60, pp. 389, 1930.
178 Schmalzried,H., Polish J. Chem. 67, pp. 167, 1993.
179 Kotula, P. G. and C. B. Carter, Interfacial Control of Reaction Kinetics in Oxides, Phys.
Rev. Lett., 77, pp. 3367-3370, 1996.
180 Eerenstein, W., T. T. M. Palstra, S. S. Saxena, T. Hibma, and S. Celotto, Origin of the
increased resistivity in epitaxial Fe3O4 films, Phys. Rev. B, 66, pp. 201101 (R), 2002.
181 Eerenstein, W., T. T. M. Palstra, S. S. Saxena, and T. Hibma, Spin-Polarized Transport
across Sharp Antiferromagnetic Boundaries, Phys. Rev. Lett., 88, pp. 247204, 2002.
182 Ramos, A.V., J.-B. Moussy, M.-J. Guittet, A.M. Bataille, M. Gautier-Soyer, M. Viret,
C. Gatel, P. Bayle-Guillemaud, and E. Snoeck., Magnetotransport properties of Fe3O4
epitaxy thin films: thickness effects driven by antiphase boundaries, J. Appl. Phys. 100, pp.
103902, 2006.
183 Eerenstein, W., T. T. M. Palstra, T. Hibma, and S. Celotto, Diffusive motion of
antiphase domain boundaries in Fe3O4 films, Phys. Rev. B, 68, pp. 014428, 2003
184 Vaz, C.A.F., J. Hoffman, A –B. Posadas, and C. H. Ahn, Magnetic anisotropy
modulation of magnetite in Fe3O4/BaTiO3 (100) epitaxial structures, Appl. Phys. Lett. 94,
pp. 022504, 2009.
185 Huang, W., Zhu, J., Zeng HZ., Wei, XH, Zhang, Y., Li, YR., Hao JH., Scr. Mater., 58,
pp. 1118, 2008.
186 Zhang, J. X., J.Y. Dai, W. Lu and H. L. W. Chan, Room temperature magnetic
exchange coupling in multiferroic BaTiO3/CoFe2O4 magnetoelectric superlattice, J. Mater.
Sci., 44, pp. 5143-5148, 2009.
187 Goodrich, T. L., Z. Cai, and K. S. Ziemer, Improved epitaxy of barium titanate by
molecular beam epitaxy through a single crystalline magnesium oxide template layer for
integration on hexagonal silicon carbide, J. Vac. Sci. and Technol. B, 26, pp. 1110-1114,
2008.
188 Wada, S., S. Suzuki, T. Noma, T. Suzuki, M. Osada1, M. Kakihana, S-E Park, L. E.
Cross, and T. R. Shrout, Enhanced piezoelectric property of barium titanate single crystals
with engineered domain configurations, Jpn. J. Appl. Phys., 38, pp. 5505-5511, 1999.
227
Документ
Категория
Без категории
Просмотров
0
Размер файла
11 737 Кб
Теги
sdewsdweddes
1/--страниц
Пожаловаться на содержимое документа