Cation Site Occupancy and Defect Engineering in
Perovskite Heterostructures
by
Eunsoo Cho
Submitted to the Department of Materials Science and Engineering
in partial fulfillment of the requirements for the degree of
DOCTOR OF SCIENCE
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
June 2024
© 2024 Eunsoo Cho. This work is licensed under a CC BY-SA 4.0 license.
The author hereby grants to MIT a nonexclusive, worldwide, irrevocable, royalty-free
license to exercise any and all rights under copyright, including to reproduce, preserve,
distribute and publicly display copies of the thesis, or release the thesis under an
open-access license.
Authored by: Eunsoo Cho
Department of Materials Science and Engineering
May 10, 2024
Certified by: Caroline A. Ross
Interim Department Head, Materials Science and Engineering
Ford Professor of Engineering
Thesis Supervisor
Accepted by: Robert J. Macfarlane
Associate Professor of Materials Science and Engineering
Chair, Departmental Committee on Graduate Studies
2
Cation Site Occupancy and Defect Engineering in Perovskite
Heterostructures
by
Eunsoo Cho
Submitted to the Department of Materials Science and Engineering
on May 10, 2024 in partial fulfillment of the requirements for the degree of
DOCTOR OF SCIENCE
ABSTRACT
Complex oxide thin films and their heterostructures are quintessential materials for next-
generation multiferroic and magnetoelectric devices. Cation and anion point defects collec-
tively affect the structural, electrical, and magnetic properties of complex oxides. At the
same time, such defects allow one to manipulate the behavior and tailor it to desired char-
acteristics. Furthermore, the formation of atomic defects is highly influenced by the growth
condition of these thin films. This thesis assesses how cation and oxygen stoichiometry and
growth dynamics can control the material properties of single- and multi-phase perovskite
oxide thin films.
We start with the generation of ferroelectricity in an antiferromagnet LuFeO3 through
cation antisite defects and concomitant inversion symmetry breaking. We elucidate how the
crystal structure and ferroelectricity depend on cation stoichiometry. We then integrate this
antiferromagnet into a self-assembled nanocomposite with a ferrimagnet CoxFe3−xO4 in order
to demonstrate interfacial magnetic exchange coupling. We discuss how the growth condition
affects cation site occupancy and crystal structure. Next, we describe a nontrivial self-
assembly mechanism of Sr(Co,Fe)O3−δ and CoOx arising from the change in oxygen pressure
during deposition. Electrolyte gating of the nanocomposite can modulate the strain state
and magnetization by removing oxygen vacancies, expanding the pathway of magnetoelectric
coupling. We then discuss a newly discovered layering of oxygen vacancies in perovskite
Sr(Co,Fe)O3−δ and identify the crystal structure using both experimental and theoretical
approaches. We end by investigating the electronic and magnetic properties of SrNiO3−δ
related to oxygen ligand holes, as well as analyzing the effects of cation substitution and
off-stoichiometry in several iron garnets.
All in all, this thesis advances the understanding of the correlation between crystal struc-
ture, atomic defects, and multiferroic and magnetoelectric properties. The observations
encourage further studies of introducing and improving multiferroicity via defect engineer-
ing in underutilized material systems and exploring interfacial phenomena in various types
of complex oxide heterostructures.
Thesis supervisor: Caroline A. Ross
Title: Interim Department Head, Materials Science and Engineering, Ford Professor of En-
gineering
3
4
Acknowledgments
I would like to start by expressing my appreciation to my advisor, Prof. Caroline A. Ross,
for taking me into her group and guiding me throughout my doctoral program. I cannot
emphasize enough how grateful I am for her always making time to discuss results and sug-
gest next steps. I have learned so much from such a wonderful advisor. Her encouragement
enabled me to grow as a researcher.
I also extend my gratitude to my thesis committee members, Prof. Geoffrey S. D. Beach
and Prof. Bilge Yildiz, for all the insightful questions and invaluable comments they gave
throughout the milestones of this program. With their guidance and support, I was able to
make meaningful progress.
I want to thank the staff members of MRL, MIT.nano, and DMSE UGTL, including Dr.
Charlie Settens, Dr. Aubrey Penn, David Bono, and Dr. Brian Neltner, for providing me
with innovative solutions to challenging experimental tasks.
To the past and present Ross group members, thank you for teaching me how things work
in our lab, kindly offering your hand anytime, and sharing your thoughts when I got stuck.
I appreciate my dear friends at MIT, whom I met within the department, at the Warehouse,
and many other places on campus, including my beloved friends from Korea. There are too
many to enumerate, but thank you all for giving me the opportunity to let off steam from
research, as well as motivating me to get out of my comfort zone.
I thank my friends from my hometown and undergrad years who believed in me and showed
endless support throughout my doctorate journey. I also want to say that I am proud of all
of you for where you are now.
Thank you, Jenny Timbas, for welcoming me like a family member and providing me with a
place to relax whenever I needed a getaway. I could not have crossed the finish line without
your support.
Lastly, I would like to express my sincere thanks to my family, especially my parents, for
their unconditional love and support. You are the most important people in my life who
allowed me to grow and shaped me into who I am now. I always have and always will admire
and love you.
5
6
Contents
Title page 1
Abstract 3
Acknowledgments 5
1 Introduction 13
1.1 Overview and motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.2 Multiferroicity and magnetoelectric effect . . . . . . . . . . . . . . . . . . . . 14
1.2.1 The origin of magnetism and ferroelectricity . . . . . . . . . . . . . . 14
1.2.2 Classification of multiferroic materials . . . . . . . . . . . . . . . . . . 15
1.2.3 Magnetoelectric coupling . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.3 Complex oxide crystal structures: perovskite, spinel, and garnet . . . . . . . 19
1.4 Defects in complex oxides . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.4.1 Cation antisite defect . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.4.2 Oxygen vacancy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.5 Self-assembled nanocomposites . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.6 How complex oxides contribute to the field of multiferroics . . . . . . . . . . 23
1.7 Thesis outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2 Methods 26
2.1 Pulsed laser deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.2 X-ray diffraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.2.1 Principles of X-ray diffraction . . . . . . . . . . . . . . . . . . . . . . 28
2.2.2 Reciprocal space mapping . . . . . . . . . . . . . . . . . . . . . . . . 29
2.2.3 X-ray reflectometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.3 Atomic force microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.3.1 Working principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.3.2 Piezoresponse force microscopy . . . . . . . . . . . . . . . . . . . . . 31
2.4 Ferroelectric characterization . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.4.1 Polarization-electric field hysteresis . . . . . . . . . . . . . . . . . . . 32
2.4.2 Pulse measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.5 Magnetic characterization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.5.1 Vibrating sample magnetometry . . . . . . . . . . . . . . . . . . . . . 33
2.5.2 Superconducting quantum interference device magnetometer . . . . . 34
2.6 Scanning transmission electron microscopy . . . . . . . . . . . . . . . . . . . 35
7
2.7 Density functional theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3 Composition-dependent ferroelectricity of lutetium orthoferrite thin films 37
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.2 Growth and structural characterization with varying cation stoichiometry . . 40
3.3 Electrical and magnetic characterization of Lu-rich lutetium orthoferrite . . . 42
3.4 Defect formation energy calculation . . . . . . . . . . . . . . . . . . . . . . . 47
3.5 X-ray absorption analysis of lutetium orthoferrite thin films . . . . . . . . . 48
3.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.7 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.8 Appendices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4 Phase formation and exchange bias in vertically aligned nanocomposites
of lutetium orthoferrite and cobalt ferrite 57
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.2 Effect of pulsed laser deposition target material on structural properties . . . 60
4.2.1 X-ray diffraction based analysis . . . . . . . . . . . . . . . . . . . . . 60
4.2.2 Electron microscopy based analysis . . . . . . . . . . . . . . . . . . . 62
4.3 Magnetic characterization and exchange coupling phenomenon . . . . . . . . 67
4.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.5 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.6 Appendices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
5 Self-assembled multiphase strontium cobalt ferrite thin films with voltage-
controlled magnetism 78
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5.2 Growth mode dependence on pulsed laser deposition process parameters . . . 80
5.3 Manipulating magnetic properties via electrolyte gating . . . . . . . . . . . . 86
5.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
5.5 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
5.6 Appendices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
6 Perovskite-derived layered crystal structure of strontium cobalt ferrite 99
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
6.2 Structural properties of layered perovskite by scanning transmission electron
microscopy and X-ray diffraction . . . . . . . . . . . . . . . . . . . . . . . . 101
6.3 Density functional theory and multislice method based image simulation and
analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
6.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
6.5 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
6.6 Appendices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
7 First principles calculation of oxygen vacancy effects on the magnetic
properties of perovskite strontium nickelate 120
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
8
7.2 Computational methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
7.3 Structural, electronic, and magnetic properties of cubic perovskite strontium
nickelate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
7.4 Electronic structure dependence on oxygen vacancies . . . . . . . . . . . . . 130
7.5 Expanding to other nickelate perovskites and heterostructures . . . . . . . . 135
7.6 Incorporation of nickel in perovskite ferrite thin films . . . . . . . . . . . . . 137
7.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
7.8 Appendices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
8 Effect of cation defects in iron garnets 148
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
8.2 Aliovalent substitution induced material property change in yttrium iron garnets149
8.3 Density functional theory study of yttrium iron garnets with antisite defects 151
8.4 Direct observation of antisite defects in terbium iron garnets . . . . . . . . . 153
8.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
9 Conclusion 155
References 159
9
List of Figures
1.1 Mechanisms of multiferroicity . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.2 Schematic of a noncollinear antiferromagnetic configuration . . . . . . . . . 17
1.3 Complex oxide crystal structures . . . . . . . . . . . . . . . . . . . . . . . . 20
2.1 Schematic of pulsed laser deposition system . . . . . . . . . . . . . . . . . . 26
2.2 Schematic of film growth modes . . . . . . . . . . . . . . . . . . . . . . . . 28
2.3 Description of Bragg’s law . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.4 Example of an X-ray reflectivity scan . . . . . . . . . . . . . . . . . . . . . 30
3.1 Structural analysis of LuFeO3 thin films . . . . . . . . . . . . . . . . . . . . 41
3.2 Magnetic hysteresis of Lu-rich LuFeO3 . . . . . . . . . . . . . . . . . . . . 42
3.3 Piezoresponse force microscopy of LuFeO3 thin films with different growth
conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.4 Electrical characterization of Lu-rich LuFeO3 . . . . . . . . . . . . . . . . . 45
3.5 O K-edge and Fe L-edge XAS results of LuFeO3 with different Lu/Fe . . . 49
3.6 Antiferromagnetic domains of LuFeO3 with different Lu/Fe . . . . . . . . . 50
3.A1 Full range XRD of LuFeO3 thin films . . . . . . . . . . . . . . . . . . . . . 53
3.A2 Characterization of SrRuO3/LuFeO3/SrRuO3 multilayered film . . . . . . . 54
3.A3 Contour plot of the XRD intensity ratio with antisite defects . . . . . . . . 55
3.A4 Crystal structures of defective LuFeO3 . . . . . . . . . . . . . . . . . . . . 55
3.A5 Phase diagram of LuFeO3 from DFT . . . . . . . . . . . . . . . . . . . . . 56
4.1 XRD results of the nanocomposites . . . . . . . . . . . . . . . . . . . . . . 61
4.2 HAADF STEM images of two different nanocomposites . . . . . . . . . . . 63
4.3 EDS mapping of two different nanocomposites . . . . . . . . . . . . . . . . 65
4.4 EELS spectra of two different nanocomposites . . . . . . . . . . . . . . . . 66
4.5 Magnetic hysteresis loops of different nanocomposites . . . . . . . . . . . . 68
4.6 Magnetic hysteresis of as-grown and field-cooled LuFeO3-Co1.2Fe1.8O4 . . . 70
4.A1 Full range XRD results of the nanocomposites . . . . . . . . . . . . . . . . 74
4.A2 SEM images of two different nanocomposites . . . . . . . . . . . . . . . . . 75
4.A3 Additional STEM imaging results . . . . . . . . . . . . . . . . . . . . . . . 76
4.A4 Fully saturated hysteresis of LuFeO3-Co1.2Fe1.8O4 . . . . . . . . . . . . . . 77
4.A5 Crystal structures and magnetic configurations of orthoferrite and spinel . 77
5.1 Schematic of combinatorial growth in PLD . . . . . . . . . . . . . . . . . . 80
5.2 XRD analysis of SrCo1−xFexO3−δ thin films . . . . . . . . . . . . . . . . . . 82
5.3 SEM and STEM images of nanocomposite thin films . . . . . . . . . . . . . 83
10
5.4 STEM images of an electrolyte-gated nanocomposite thin film . . . . . . . 84
5.5 Magnetic hysteresis loop before and after gating . . . . . . . . . . . . . . . 87
5.A1 Film thickness fitting of a nanocomposite . . . . . . . . . . . . . . . . . . . 92
5.A2 XRD of films grown in a different PLD chamber . . . . . . . . . . . . . . . 92
5.A3 SEM images of the nanocomposites . . . . . . . . . . . . . . . . . . . . . . 93
5.A4 Additional XRD of nanocomposites . . . . . . . . . . . . . . . . . . . . . . 93
5.A5 XPS results of SrCo1−xFexO3−δ thin films . . . . . . . . . . . . . . . . . . . 94
5.A6 Additional XRD after gating nanocomposites . . . . . . . . . . . . . . . . . 94
5.A7 Additional HAADF STEM images after gating . . . . . . . . . . . . . . . . 96
5.A8 Rocking curve comparison before and after gating . . . . . . . . . . . . . . 97
5.A9 Magnetic hysteresis of nanocomposites at room temperature . . . . . . . . 97
5.A10 Magnetic hysteresis of gated nanocomposites . . . . . . . . . . . . . . . . . 98
5.A11 XRD of as-grown and gated Co3O4 thin films . . . . . . . . . . . . . . . . . 98
6.1 STEM images and analysis of SrCo0.26Fe0.74O3 thin film . . . . . . . . . . . 101
6.2 XRD and RSM of SrCo0.26Fe0.74O3 thin film . . . . . . . . . . . . . . . . . 102
6.3 Map of reported SrBO3−δ structures (B = Co, Fe, Mn) . . . . . . . . . . . 105
6.4 Crystal structures of possible ordering in SrBO3−δ (B = Co, Fe) . . . . . . 107
6.5 Multislice simulation results of SrBO3−δ (B = Co, Fe) . . . . . . . . . . . . 109
6.A1 Extended STEM images of the layered perovskite structure . . . . . . . . . 112
6.A2 Simulated XRD with varying oxygen occupancy . . . . . . . . . . . . . . . 113
6.A3 STEM images at the grain boundary of layered perovskite . . . . . . . . . 114
6.A4 STEM image of SrCo0.57Fe0.43O3−δ thin film . . . . . . . . . . . . . . . . . 115
6.A5 Compilation of simulated STEM images of different structures . . . . . . . 116
6.A6 STEM image after gating . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
7.1 Crystal structures of SrNiO3−δ . . . . . . . . . . . . . . . . . . . . . . . . . 125
7.2 Projected DOS of SrNiO 2+3−δ and A NiO3−δ . . . . . . . . . . . . . . . . . 127
7.3 Scheme of the electronic structure of Ni and O . . . . . . . . . . . . . . . . 128
7.4 Effect of strain on magnetic moments of SrNiO3 . . . . . . . . . . . . . . . 129
7.5 Magnetic moment dependence on oxygen deficiency . . . . . . . . . . . . . 135
7.6 XRD of SrNixFe1−xO3−δ and YNixFe1−xO3−δ thin films . . . . . . . . . . . 138
7.7 XRD of as-grown and gated SrNixFe1−xO3−δ thin films . . . . . . . . . . . 139
7.A1 Projected DOS calculated using a different pseudopotential . . . . . . . . . 141
7.A2 Crystal structure and DOS of hexagonal SrNiO3 . . . . . . . . . . . . . . . 142
7.A3 Total DOS of different A2+NiO3−δ compounds . . . . . . . . . . . . . . . . 143
7.A4 DOS of heterostructured SrNiO3 . . . . . . . . . . . . . . . . . . . . . . . . 144
7.A5 Additional projected DOS . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
7.A6 Possible vacancy configurations of SrNiO2.75 . . . . . . . . . . . . . . . . . 146
7.A7 Crystal structure and DOS of SrNiO2.625 . . . . . . . . . . . . . . . . . . . 146
8.1 XRD of Ca-substituted yttrium iron garnet thin films . . . . . . . . . . . . 149
8.2 Magnetic hysteresis of Ca-substituted yttrium iron garnet thin films . . . . 150
8.3 Crystal structure and antisite defects of yttrium iron garnet . . . . . . . . 152
8.4 Direct evidence of terbium antisite defects via STEM . . . . . . . . . . . . 153
11
List of Tables
3.1 Defect formation energy of LuFeO3 from DFT . . . . . . . . . . . . . . . . 48
4.1 Growth conditions of different nanocomposites . . . . . . . . . . . . . . . . 60
4.2 Coercivity of as-grown and field-cooled LuFeO3-Co1.2Fe1.8O4 at different
temperatures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
6.A1 Out-of-plane lattice parameters of simulated structures . . . . . . . . . . . 118
6.A2 Unit cell volume difference of strained structures . . . . . . . . . . . . . . . 118
6.A3 Out-of-plane lattice parameter difference of strained structures . . . . . . . 118
6.A4 Formation energy of different structures from DFT . . . . . . . . . . . . . . 119
7.1 Bader charge analysis of SrNiO3−δ . . . . . . . . . . . . . . . . . . . . . . . 129
7.2 Energy and magnetic moment of SrNiO2.75 . . . . . . . . . . . . . . . . . . 133
7.3 Properties of A2+NiO3−δ . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
7.A1 Energies of different magnetic structures . . . . . . . . . . . . . . . . . . . 147
7.A2 Integrated DOS area fraction between eg and t2g . . . . . . . . . . . . . . . 147
12
Chapter 1
Introduction
1.1 Overview and motivation
As the worldwide demand for electronics continuously increases, there has been a surge
of interest in developing devices that are smaller, denser, and more efficient. One approach
is to innovate the conventional complementary metal-oxide-semiconductor (CMOS) archi-
tecture and rebound from the stagnation in Moore’s law. Another approach is to invent
new devices that operate in a different paradigm, where multiferroics and magnetoelectrics
take a part. Multiferroics have classically been defined as materials that exhibit two or more
ferroic properties among ferromagnetism, ferroelectricity, ferroelasticity, and ferrotoroidicity.
Nowadays, the range of multiferroic materials has extended to include antiferromagnetism
and ferrimagnetism as well, and magnetoelectric materials in a broader perspective.1 Multi-
ferroics with magnetoelectric coupling enable the manipulation of magnetism via an electric
field and the control of electric polarization via a magnetic field. This can be utilized in a
number of emergent devices, for example, magnetoelectric spin-orbit device, magnetoelectric
random access memory, and four-state logic device.2 ,3 These differ in the operation mech-
anism from CMOS that may solve the challenges of current computing, such as the von
Neumann bottleneck.
Despite the intensifying interests and promising perspectives of such new device architec-
tures, numerous challenges still need to be solved in this field. The selection of materials that
13
are multiferroic at room temperature with appreciable magnetoelectric coupling is limited.
Elucidating the switching dynamics is important as well because it is common for complex
domains to appear in multiferroic materials. From the perspective of manufacturing, stable
and robust ferroic materials with low switching force and low leakage need to be developed.
Furthermore, processing methods to integrate them at the chip or wafer level without loss of
functionality in nanoscale devices are required. Among these questions yet to be answered,
this thesis aims to investigate how defect engineering can successfully maneuver novel prop-
erties in complex oxide thin films for multiferroic applications. In addition, the studies in the
following chapters provide a meticulous analysis of the fundamental understanding of mate-
rial properties depending on the growth condition, which can serve as a basis and provide
guidance in future production.
1.2 Multiferroicity and magnetoelectric effect
1.2.1 The origin of magnetism and ferroelectricity
The quantum mechanical origin of magnetism is the presence of localized electrons with
corresponding magnetic moments. The idea of spontaneous spin alignments due to the
exchange interactions of electrons between bonded atoms was first proposed by Heisenberg.4
The Heisenberg Hamiltonian is expressed by
∑
Hex = −2 JijSi · Sj (1.1)
i>j
where Jij is the exchange constant, and Si and Sj are spin angular momentum of the adjacent
atoms. Jij > 0 will favor ferromagnetism (parallel alignment of spins) and Jij < 0 will favor
antiferromagnetism (antiparallel alignment of spins). Asymmetric exchange, also known as
the Dzyaloshinskii-Moriya interaction (DMI),5 ,6 which the Hamiltonian can be represented
14
by
HDM = Dij · (Si × Sj) (1.2)
occurs for canted Si and Sj (Dij is the DMI vector) in crystals with fewer symmetry opera-
tions, resulting in a weak net ferromagnetic moment.
The origin of ferroelectricity can quite differ between materials, such as hybridization,
geometric factors, and/or spin, charge, or orbital degrees of freedom.3 The best example of
the hybridization effect is BiFeO3, where the ferroelectricity arises from the hybridization
between Bi 6p and O 2p orbitals and the lone electron pairs of Bi. BiFeO3 is not only
ferroelectric but also magnetic, and the weak magnetism arises from canted Fe3+. When
geometric constraints play a role, the asymmetric position of differently charged ions leads
to inversion symmetry breaking and thus polarization, such as B-site cation displacement
in tetragonal perovskites BaTiO3 or Pb(Zr,Ti)O3. Lastly, ferroelectricity can originate from
the long-range ordering of spin, charge, or orbitals, in other words, from magnetism. This
mechanism will be discussed in the next section. It is worth noting that ferroelectric materials
may be divided into two groups: proper ferroelectrics and improper ferroelectrics. Proper
ferroelectrics refer to materials in which polarization is the order parameter that dictates the
ferroelectric transition, whereas improper ferroelectrics refer to materials whose polarization
is a by-product of other order parameters that do not have one-to-one correspondence with
ferroelectric domains.7 Recent inventions of engineered ferroelectricity in originally magnetic
materials usually fall into this category.1
1.2.2 Classification of multiferroic materials
In general, multiferroic materials can be classified into two groups: Type-I, where the
source of ferroelectricity and magnetism is different, and type-II, where ferroelectricity is due
to magnetism.8 The aforementioned example of BiFeO3 is included in type-I. Nevertheless,
due to the nature of type-I that magnetism and ferroelectricity stem from different origins,
15
the coupling between the two properties is weak in these materials, so efforts have been made
to enhance this coupling.
Figure 1.1: The three major mechanisms in type-II multiferroics, partially adapted from (9 ).
In type-II multiferroics, traditionally there have been three major mechanisms on how the
ferroelectricity originates from spin: (1) spin current model, (2) exchange striction model,
and (3) spin-dependent p-d hybridization model (Figure 1.1).9 First, in the spin current
model, the ferroelectricity comes from spiral magnetic ordering. This is a commonly observed
phenomenon in perovskite-structured rare earth manganites (RMnO3). The polarization P
can be expressed as
P ∝ eij × (Si × Sj) (1.3)
and here eij represents the unit vector connecting Si and Sj. Note that this equation stands
for crystal structures with specific symmetry, i.e. cubic or tetragonal, and thus, this is not
the only way for non-collinear magnetic structured systems to exhibit polarization.
Materials with collinear spin configuration can exhibit ferroelectricity as well, and this
leads to the next mechanism, the exchange striction model. Here, the symmetric spin ex-
change interaction may induce striction along a specific crystallographic direction πij. In
this case, the polarization P is given by
P ∝ πij(Si · Sj) (1.4)
For instance, if the up-up-down-down spin ordering appears and the exchange interaction
between the up-up (down-down) pairs and the up-down pairs are inequivalent, the inversion
16
symmetry is broken and polarization is produced by the displacement of ions from their
charge-balanced location. This can be observed in a chain magnet like Ca CoMnO .103 6
Another example of this mechanism is rare earth orthoferrites (RFeO3) at temperatures
in between the spin-reorientation temperature of Fe and the Neel temperature (TN) of R.
For instance, when R is a solid solution between Tb and Dy, Fe has a spin configuration
of GxAyFz (Γ4) and R has a spin configuration of GxAy.11 Here, GxAyFz is the magnetic
configuration represented in Bertaut’s notation. In this particular case, it defines an ordering
in which the spins are G-type antiferromagnetic (nearest-neighbor spins point in opposite
directions) along the x axis, A-type antiferromagnetic (spins point the opposite direction
in consecutive planes) along the y axis, and have a ferromagnetic ordering along the z
axis. Then the common GxAy alignment of Fe3+ 3d and R3+ 4f electrons causes symmetric
exchange interaction, which leads to the displacement of R3+ and thus polarization.
Figure 1.2: (a) Schematic of a GxAyFz (Γ4) type antiferromagnetic configuration and (b)
GxAyFz antiferromagnetic structure in an orthoferrite RFeO3 with unit cell space group
Pbnm, adapted from (7 ). Only Fe3+ moments have been represented as arrows and Gx is
the dominant alignment among GxAyFz.
Lastly, in the spin-dependent p-d hybridization model, a locally polar bond is modulated
by the spin-orbit coupling depending on the spin direction. When the sum of this polar-
ization across all lattices does not cancel out, the system has a net polarization. Here, the
polarization P is proportional to
P ∝ (S · e )2i il eil (1.5)
17
and eil is the unit vector connecting the spin and the ligand. In summary, (1) requires spin-
orbit coupling, (2) requires low symmetry of the lattice and commensurate spin ordering,
and (3) requires both the spin-orbit coupling and low symmetry of the lattice but can occur
at commensurate or incommensurate spin ordering.9
The main objective for multiferroic materials would be to achieve appreciable polariza-
tion and magnetic ordering at the same time at room temperature. However, these type-II
multiferroic materials have small polarization in general and the ferroelectric behavior usu-
ally occurs at cryogenic temperatures, so it would be a pioneering development to find an
engineering pathway that enhances polarization without compromising the magnetism.
1.2.3 Magnetoelectric coupling
Magnetoelectric coupling stands for the coupling of magnetic and electric ordering in a
material. Thermodynamically, the free energy F of a system is given by12
F (E,H) = F − P sE −M s
1 1
0 i i i Hi − ϵ0ϵijEiEj − µ0µijHiHj − αijEiHj2 2 (1.6)
1 1
− βijkEiHjHk − γijkHiEjEk − . . .
2 2
where F0 is the free energy of the ground state, i, j, k refer to the three spatial coordinate
variables, Ei and Hi refer to the electric and magnetic field, P si and M si refer to the sponta-
neous polarization and magnetization, ϵ0 and µ0 refer to the dielectric constant and magnetic
susceptibility of vacuum, ϵij and µij are the components of the second-order tensor of dielec-
tric constant and magnetic susceptibility, αij denotes the components of the second-order
tensor α which is defined as the linear magnetoelectric effect, βij and γij are the components
of the third-order tensor. The reason why α is the linear magnetoelectric effect tensor can
18
be found in the derivation below:
∂F 1
Pi(E,H) = − = P
s
i + ϵ0ϵijEj + αijHj + βijkHjHk + γijkHiEj + . . .∂Ei 2 (1.7)
∂F 1
Mi(E,H) = − = M
s + µ0µijHj + αijEj + βijkHjEi + γijkEjEk + . . .
∂H ii 2
αij couples Pi and Hj, Mi and Ej in a linear manner.
However, it is challenging to find a single-phase material with such a high α. In order to
overcome this difficulty, heterostructures such as nanocomposites were developed.3 Incorpo-
rating two different phases, one with ferroelectricity and the other with magnetism, allows
magnetoelectric coupling mediated by the strain between the two phases. In other words, the
coupling originates from the piezoelectric effect (polarization and strain) from the former and
the magnetostrictive effect (magnetization and strain) from the latter. A representative ex-
ample of this heterostructure would be a vertically aligned nanocomposite composed of pillars
embedded in the matrix.13 The nanocomposite was grown on (001)-oriented SrRuO3/SrTiO3
with magnetic spinel CoFe2O4 pillars that were embedded in ferroelectric perovskite BaTiO3
matrix.
1.3 Complex oxide crystal structures: perovskite, spinel,
and garnet
The three major complex oxide structures discussed in this thesis are perovskite (ABO3),
spinel (AB2O4), and garnet (A3B5O12). Figure 1.3a shows a cubic perovskite structure with
an A atom coordinated by 12 oxygen ions and a B atom coordinated by 6 oxygen ions
√
(octahedral). The Goldschmidt tolerance factor t defined by t = (rA + rO)/( 2(rB + rO))
can be an indicator of the distortion in the perovskite structure; when A is smaller and
0.7 < t < 0.9 it becomes orthorhombic (Figure 1.3b), and when t > 1 it prefers hexagonal
or tetragonal lattice. Spinel (Figure 1.3c) has the tetrahedral (A, surrounded by 4 oxygen
19
Figure 1.3: Crystal structures of (a) perovskite (space group Pm3̄m), (b) orthorhombic
perovskite (space group Pbnm), (c) spinel (space group Fd3̄m), and (d) garnet (space group
Ia3̄d). Red atoms are oxygen and the corresponding coordination of the cations is labeled.
ions) and octahedral (B) sites in a 1:2 ratio. The strongest superexchange interaction is
antiferromagnetic between 125◦ bonding of A—O—B. The nominal valence state in AB2O4
is A2+ and B3+, however, when the smaller B3+ goes to the tetrahedral site, it is named an
“inverse spinel.” Lastly, Figure 1.3d is a garnet structure with the chemical formula A3B5O12.
Here, there are three dodecahedral A sites, two octahedral B sites, and three tetrahedral B
sites. When A = Y or rare earth (R, La-Lu), the compound is named rare earth iron garnet.
Octahedral and tetrahedral sites couple antiferromagnetically, and if R is a magnetic rare
earth, its moment aligns parallel to the octahedral sites.
1.4 Defects in complex oxides
1.4.1 Cation antisite defect
In complex oxides with two or more cations, antisite defects refer to defects where the
cations switch their position. According to the Kroger-Vink notation,14 A cation in the B
20
site is represented as AB and B cation in the A site is represented as BA. A common system
for antisite defects that affect magnetization is the spinel system. For instance, in ZnFe2O4,
the presence of Zn′Fe changed the antiferromagnetic ordering into a weak ferromagnet.15
Also, in a perovskite La ×0.8Sr0.2MnyO3±δ, LaMn can form to account for Mn deficiency.
16 The
formation energy of the antisite defects can be relatively low; for example FeY, YFe,Oct, and
Y 3+Fe,Tet in a garnet Y3Fe5O12. Replacing a magnetic Fe with non-magnetic Y3+ changes
the net magnetic moment in the ferrimagnetic system.17
The power of epitaxy comes in when one tries to stabilize films with off-stoichiometry.
The growth of thin films with cation compositions prohibited in the bulk phase diagram
is possible for various complex oxide systems including perovskite and garnet structures,
as shown in recent studies the author has contributed to.18–20 Antisite defects give rise to
perturbation in magnetic ordering or crystal symmetry, altering their properties from that
of bulk.
1.4.2 Oxygen vacancy
Oxygen vacancies are the most common defect in any oxide system; they have a signifi-
cant impact and bring about emergent properties. First of all, oxygen vacancies can dictate
the magnetic properties of complex oxides because the superexchange and double-exchange
mechanisms involve oxygen ions. For instance, magnetic exchange interactions, Curie tem-
perature (TC), and/or coercive field (Hc) can all depend on oxygen deficiencies in various
perovskite systems such as SrCoO3−δ, La0.33Sr0.67CoO3−δ, SmBaCo2O3−δ, EuBaCo2O3−δ, or
La0.7Sr0.3MnO3−δ.21–23 Oxygen vacancies can also be the source of ferroelectricity; for exam-
ple, they can break the inversion symmetry by accumulating at the interface of superlattices
due to the difference in diffusivity,24 or they can distort the bonds locally that produce
polarization.25
When the vacancy concentration further increases, they can show a particular ordering
and alter the crystal structure. For example, a structure named brownmillerite with the
21
stoichiometry of ABO2.5 can emerge from the topotactic phase transition of perovskites
when the oxygen vacancy channel order in planes. This is commonly observed in SrCoO 263−δ
and SrFeO .27 ,283−δ In addition, perovskite-derived Ruddlesden-Popper (RP) phase with the
formula An+1BnO3n+1 or infinite layer structures like ABO2 can be produced at oxygen-
deficient conditions or by adding a reduction step.29 ,30
The control of oxygen stoichiometry can be performed after growth via electrolyte gating.
When the film is immersed in the electrolyte (ionic liquid) and a voltage is applied, an electric
double layer is formed and the O2− or H+ ions accumulate at the film surface depending
on the bias direction. As a consequence, the chemical potential difference acts as a driving
force to push those ions into the film. If the bias is applied between the ionic liquid and
the substrate, the electrical field across the film can act as an additional driving force on
the transport of ions. A common way to implement this configuration will be using a
conductive substrate such as Nb-doped SrTiO3 or depositing a conductive buffer layer such
as La0.7Sr0.3MnO3−δ or SrRuO3 before the deposition of the active material. SrCoO 313−δ and
SrFeO 283−δ are representative systems studied with this method because of its capability of
topotactic transition from perovskite to brownmillerite.
1.5 Self-assembled nanocomposites
As mentioned in Section 1.2.3, there has been a great interest in synthesizing nanocompos-
ites with two or more phases with different functionalities. This occurs when the nucleation
of the immiscible phases happens separately on the substrate. The energies involved during
growth are surface energy of the substrate, surface energy of the growing phases, interface
energy between different phases, and the elastic strain energy from lattice mismatch. If
the sum of the surface energy of the growing phases and interface energy between different
phases exceeds the surface energy of the substrate, the arriving flux of materials dewet on
the surface and creates islands (discussed further in Section 2.1) leading columnar growth of
22
the two (or more) immiscible materials. Since different crystallographic planes have different
surface/interface energies, this phenomenon is affected by growth orientation and the result-
ing morphology can also differ. Also, the more the lattice mismatch is, the more faceted the
morphology to decrease elastic strain energy.
1.6 How complex oxides contribute to the field of multi-
ferroics
One of the key challenges in multiferroicity is to overcome the contradiction known as
the “d0 vs. dn problem.”8 Transition metal ions in ferroelectric materials usually have empty
d orbitals. However, when d orbitals are occupied and thus exhibit magnetism, ferroelec-
tricity disappears, stymieing the development of multiferroics. In turn, researchers took
alternate pathways to circumvent this dilemma. For single-phase multiferroics, one can in-
troduce polarization in a magnetic material via charge ordering,32 geometric distortion,33 or
as a byproduct of spin configuration.34 Nevertheless, it is beneficial to induce ferroelectricity
independent of the characteristics of the parent magnetism and to develop a generalizable
method. Atomic defects including cation antisite can be a good candidate in this aspect,
especially considering the potential for magnetoelectric coupling from the structural rela-
tionship.
The objective of this thesis is to elucidate how cation and oxygen stoichiometry of pulsed-
laser-deposited single- and multi-phase films can collectively affect the structural, magnetic,
ferroelectric, and magnetoelectric properties of thin films, towards the goal of enriching the
candidates for multiferroic and magnetoelectric materials. Perovskite-structured complex
oxides have a lot of degrees of freedom in both geometry and composition, which allows for
the design and tailoring of material properties. Both A-sites and B-sites can be a mixture
of two or more cations that bring about novel phenomena including magnetism.35 ,36 In ad-
dition, they are capable of accommodating various symmetry-breaking orderings: the bond
23
angle can deviate from 180◦, the bond length can be different in a, b, c directions, α, β, γ
can deviate from 90◦, and the BO6 octahedra can rotate, which all results in the lowering
of the crystal symmetry towards ferroelectric space groups. Moreover, in thin films, strain
from substrates or growth orientations can be imposed to adjust the magnetic and ferroelec-
tric properties further,37 along with the capability of accommodating off-stoichiometry.20 In
addition, perovskites are compatible in heterostructures with two or more phases, either as
multilayers38 or vertically aligned nanocomposites,13 which additionally benefit from inter-
facial effects. All in all, oxide perovskite compound is an attractive system of interest in the
field of multiferroics. However, the multifunctionality of complex oxide thin films is strongly
influenced by the growth condition, which gives the grounds for studying how growth param-
eters can affect the resulting thin film. Therefore, this study will also highlight the structural
property dependence on growth conditions, particularly in nanocomposite heterostructures,
and demonstrate the interaction and coupling between the different phases.
1.7 Thesis outline
Followed by the introduction in Chapter 1, we begin with describing the experimen-
tal and computational methods used in this thesis in Chapter 2. Chapter 3 discusses the
rare-earth antisite defect mechanism of ferroelectricity in a weakly ferromagnetic orthofer-
rite LuFeO3. Chapter 4 presents the magnetic exchange coupling between self-assembled
nanocomposites that consist of perovskite LuFeO3 from Chapter 3 and spinel CoxFe3−xO4.
Chapter 5 describes the growth of CoOx and perovskite SrCo1−xFexO3−δ nanocomposites
despite stoichiometry and the manipulation of their magnetic properties from oxygen inser-
tion via electrolyte gating. Chapter 6, which is a follow up of a discovery in Chapter 5,
elucidates the structural properties and identifies the oxygen coordination of layered per-
ovskite SrCo1−xFexO3−δ. Chapter 7 elaborates a theoretical work on the role of oxygen in
the magnetism and electronic structure of cubic perovskite SrNiO3−δ. Chapter 8 compiles
24
the work the author has contributed to engineering cation site occupancy in various iron
garnets. Lastly, Chapter 9 will summarize the previous chapters and provide an outlook on
future research directions.
25
Chapter 2
Methods
2.1 Pulsed laser deposition
Figure 2.1: Schematic of a pulsed laser deposition system.
Pulsed laser deposition (PLD) is a physical deposition method where short pulses of laser
(typically nanoseconds of ultraviolet light; the author used a KrF excimer laser from Coherent
with the wavelength of 248 nm) intermittently ablates the target material and generates a
plasma plume consisting of energetic species that absorbed the energy that eventually arrives
26
and impinges onto the substrate surface.39 This is a powerful method to explore complex
oxide thin films with various compositions, defects, and strain states.
Figure 2.1 describes the critical components of a PLD chamber. During deposition, the
laser beam enters through the laser window and hits the target that is rotating with respect
to its center and by the carousel motor at the same time. This is in order to ablate the target
surface as evenly as possible. Also, the window is cleaned every time using a diamond paste
prior to deposition to avoid contamination and absorption of laser energy. The substrate is
rotated as well to ensure uniform thickness of the deposited film. The film can be deposited
at elevated temperatures using the substrate heater; however, the heater’s set point is not
the substrate’s actual temperature, and the difference varies depending on the chamber. The
author used the PLD system manufactured by Neocera.
Different growth modes of PLD are represented in Figure 2.2, which depends on the film
and substrate surface and interface energy relationship. Layer-by-layer mode (Figure 2.2a)
is activated when the film wets the substrate, i.e. the sum of the film surface and interface
energy is lower than the substrate surface energy. In contrast, when the sum of the film
surface and interface energy is larger than the substrate surface energy and the film does
not wet the substrate, island growth (Figure 2.2b) will be promoted. Often in epitaxial
film growth, a transition from layer-by-layer to island growth can occur (Figure 2.2c) as a
mechanism of strain relaxation.
Process parameters of PLD include laser fluence, laser repetition rate, background gas
and its partial pressure, substrate temperature, and substrate-to-target distance. For ex-
ample, laser fluence plays an essential role in dictating the lattice constant, defects, and
stoichiometry of complex oxides.40 Furthermore, there exists a critical value of fluence to
initiate sufficient ablation and traverse of the target material.
For the growth of complex oxides, the background gas is typically oxygen. After achieving
the high vacuum level base pressure (5 × 10−6 Torr), oxygen partial pressure during growth is
controlled by the mass flow controller and the turbo pump. At low oxygen pressures, plumes
27
Figure 2.2: Schematic of film growth modes: (a) Frank-Van der Merwe (layer-by-layer), (b)
Volmer–Weber (island), (c) Stranski–Krastanov, and (d) step flow. Adapted from (39 ).
are more directional and there is less scattering, and vice versa at high oxygen pressures. In
addition to the plume shape, there exists preferential scattering between elements that can
affect the final film stoichiometry, i.e. one element can scatter more than the other, leading to
off-stoichiometry of the deposited film.18 ,20 Such behavior can depend on the oxygen partial
pressure as well.
2.2 X-ray diffraction
2.2.1 Principles of X-ray diffraction
X-ray diffraction (XRD) is a crucial technique in characterizing the structural properties
of epitaxial crystalline thin films since it can provide a myriad of information such as lattice
parameters, composition, strain relaxation, mosaicity, tilt, thickness, density, and more.
This is based on Bragg’s law (Figure 2.3), where the constructive interference occurs when
λ = 2dhklsinθ fulfills. Here, λ is the wavelength of the X-ray, dhkl is the distance between
(hkl) planes, and θ is the angle between (hkl) planes and the incoming X-ray beam.41
XRD in this thesis mainly refers to the symmetric coupled scan. Here, ω (from now on
defined as the incident angle) and 2θ (angle between the incident and the diffracted X-ray)
28
→− →−
Figure 2.3: Description of the Bragg’s law. k and k′ are the wave vectors of incident and
→− →−
diffracted X-ray. | k | = |k′ | = 2π/λ. →−s , the scattering vector, is parallel to the reciprocal
space vector corresponding to (hkl) plane, →−g hkl.
move simultaneously so that the scattering vector is kept consistent parallel to the normal
of the sample surface. We first align the substrate peak and then perform a coupled scan
in a range of desired 2θ values to obtain information about the film. This method provides
information about film planes that are parallel to the sample surface. In addition to the
coupled scan, one can measure the intensity over a range of ω, namely a rocking curve, to
measure the quality of the crystallinity. The sharper the rocking curve, the fewer defects
(e.g., dislocations, mosaicity, tilts) are present in the film.
2.2.2 Reciprocal space mapping
Scanning around reciprocal lattice points provides rich information about the in-plane
strain state otherwise not accessible through coupled XRD scans. The source and detector
move to the asymmetric ω and 2θ particular to the desired reciprocal lattice point taking
account of the plane tilt from the sample surface. Then, detector scans at constant ω or
rocking curves at constant 2θ are repeated to generate collective data that will constitute a
two-dimensional “map” around the substrate and film reflection. Based on the position of
the film peak with respect to the substrate peak, as well as the stretch of the film peak, one
can deduce information about strain relaxation, tilt, and mosaicity within the film.
29
2.2.3 X-ray reflectometry
Figure 2.4: An example XRR of a trilayered film structure, SrRuO3/LuFeO3/SrRuO3 grown
on Nb-doped SrTiO3. Both SrRuO3 layers are ∼10 nm, thinner than LuFeO3 of ∼70 nm.
Detailed analysis on the properties can be found in Figure 3.A2.
X-ray reflectometry (XRR) is performed at small incident angles (below ∼5◦) and utilizes
the refractive index difference between layers (film-substrate or multilayered films). Above
the critical angle, there exists a path difference between the X-ray that is reflected at the
air-film interface vs. the film-substrate interface (more interfaces including film-film for a
multilayered structure). This produces oscillations in the intensity, which is a function of
layer thickness. Figure 2.4 is a representative example of what an X-ray reflectivity scan
looks like. Two different periodicities exist: a wider one from the thinner layer and a narrower
one from the thicker layer. One can fit the curve using density, roughness, and thickness as
variables. The amplitude of the oscillation can be affected by the film roughness and density
difference.
2.3 Atomic force microscopy
2.3.1 Working principle
Atomic force microscopy (AFM) utilizes a mechanical cantilever with a sharp tip to reach
the proximity or contact with the sample surface to image topography and characterize
30
various properties including electrical or magnetic responses. The conventional operation
mode for imaging is called the tapping mode. Here, the cantilever oscillates at a resonance
frequency and intermittently touches the surface, and the laser beam on the cantilever is
reflected and detected at a two-dimensional photodetector. As the tip reaches the surface,
the shift in the reflected signal, caused by the attractive or repulsive interaction between the
tip and the surface, is tracked. The difference in phase and amplitude is compared with the
original drive signal, which is corrected through a feedback loop by adjusting the Z piezo of
the cantilever, giving information about the height.
2.3.2 Piezoresponse force microscopy
Piezoresponse force microscopy (PFM) is a technique that uses AFM to image ferroelec-
tric (ferroelectric materials are piezoelectric by nature because of the space group relation-
ship) domains concurrently with topography. In contrast to typical topography mapping
modes of AFM, PFM is operated with a conductive tip being in contact with the surface
at a constant force. This is because the electric field (bias) can be applied when the tip
is touching the sample, unlike magnetic force microscopy (MFM), where the detection of
stray field does not require contact. During PFM, an AC electrical bias is applied to the
cantilever, which induces strain from the inverse piezoelectric effect. The strain then deflects
the cantilever and this deflection is the opposite for polarization parallel and antiparallel to
the bias. Thus, the phase and amplitude response of the cantilever can give information
about the polarization direction and the piezoelectric coefficient of the material. It is com-
mon to drive the cantilever to a negative vertical deflection but maintain the set point to
zero to provide additional force on the tip pressing the sample surface.
Switching spectroscopy PFM (SS-PFM) is another method to probe ferroelectricity using
PFM.42 At a single position, a triangular DC bias on top of the AC bias is applied in a
stepwise manner. In between each step, the DC bias is turned back to zero to measure the
remnant effect of the DC bias. Such triangular bias gives rise to the nucleation and reversal
31
of a single domain; the amplitude goes to zero when the domain is switching and the phase
simultaneously changes by ±180◦.
When the tip is in contact, the resonance frequency (namely the contact resonance)
becomes higher by 3-4 times vs. the resonance in air. It is desirable to perform the imaging
near contact resonance; however, this needs caution because operating exactly at the contact
resonance can amplify non-piezoelectric responses and mislead interpreting results. Different
AFM manufacturers tackle this challenge in various ways. For example, AFM from Oxford
Instruments supports a Dual AC Resonance Tracking (DART) mode,43 driving the cantilever
slightly above and below the resonance to track the frequency shift as the tip moves on the
surface. Nevertheless, there are potential sources for artifacts in PFM, such as electrostatic
or electromechanical responses.44 One must perform PFM carefully; for instance, check the
electrical connection and grounding are done properly, increase the AC voltage for higher
amplitude and thus improve signal-to-noise ratio, and watch that the signal does not entirely
follow the topography.
2.4 Ferroelectric characterization
2.4.1 Polarization-electric field hysteresis
Polarization vs. electric field hysteresis (namely P-E) is one of the most commonly used
methods to track the ferroelectric response of a material. Here, a triangular wave of voltage
is applied to a device with two terminals and the current flow through the circuit is measured.
This current is integrated over the time interval and converted into charge, which then gives
polarization when divided by the area of the electrode. This can be done by a custom
setup that applies voltage and measures current or by commercially available equipment
from vendors like Radiant Technologies. For ferroelectrics that require a “wake up” like
(Hf,Zr)O ,452 the voltage cycling can be repeated many times prior to actual measurements.
Akin to PFM, ferroelectric hysteresis measurements are prone to artifacts46 and need to
32
be performed meticulously. One of the most common mistakes for measurements using thin
films could be to misinterpret the lens-like hysteresis shape as a ferroelectric response, which
in fact could be the result of a capacitor being in parallel with a resistor. Most of all, it is
important to have a film with minimal leakage current, although it could be mitigated by
operating at higher frequencies or with smaller electrodes. Otherwise, one can look for other
indicators, such as switching current or peak in capacitance, or repeat the measurement at
different frequencies and with varying maximum voltage to observe a systematic change in
the hysteresis loops.
2.4.2 Pulse measurements
Positive-up-negative-down (PUND) measurement is another way to avoid misinterpreting
results of P-E loops.47 Here, after the first “positive” triangular voltage pulse, another “up”
pulse is followed, then a “negative” pulse is applied, followed by the final “down” pulse.
The shapes and magnitude of positive/up and negative/down pulses are generally the same.
The response of the first pulse includes both switchable and nonswitchable effects, whereas
the second pulse will only cause a nonswitchable effect (leakage, for example) because it
has already been switched. Subtracting the two will solely give the switchable response, in
other words, remnant polarization. One can also perform P-E measurements in a PUND-like
manner (triangular pulses in a PUND sequence) and reconstruct a remnant hysteresis loop.
2.5 Magnetic characterization
2.5.1 Vibrating sample magnetometry
Vibrating sample magnetometer (VSM), invented at the Lincoln Lab, utilizes Faraday’s
law to measure the magnetic moment of a given sample. The magnetized sample vibrates
vertically between the pickup coils, which measures the induced voltage due to the motion
33
of a magnetized object (sample) and the following change in the magnetic flux, which is
then converted into magnetization. The electromagnet can apply a varying magnetic field to
the sample, allowing one to measure the hysteresis loop of magnetization vs. magnetic field
(M-H). Before measurements, the gaussmeter, sample position, and pickup coil sensitivity
are calibrated with a Ni disc with a known magnetic moment. In-plane and out-of-plane
measurements can be done by rotating the lollipop-shaped Pyrex sample holder by 90◦.
Hence, the field direction is parallel or perpendicular to the sample surface. The measured
raw data is superposed with the linear background, which is subtracted by fitting the linear
portion at higher applied fields of the M-H curve. Sample temperature can be controlled
down to ∼ −100 ◦C using liquid nitrogen and up to ∼400 ◦C or higher using an oven. For
temperatures higher than the Kapton tape can withstand (i.e., field cooling in Chapter 4),
the author manufactured a sample holder out of mica.
2.5.2 Superconducting quantum interference device magnetometer
Superconducting quantum interference device (SQUID) magnetometer shares the mea-
surement principle as VSM in the sense that the equipment measures the change in the
magnetic flux from the sample. However, here, the sample moves inside a hole consist-
ing of two semicircle superconductors separated by a thin insulating layer (together called
the Josephson junction). While a fixed current flows through the Josephson junction, a
magnetic flux will interfere and affect the voltage oscillation across the junction, which is
then converted into the magnetic moment through the system’s internal calibration. The
SQUID magnetometer from Quantum Design offers access to higher magnetic fields (±7
T) and cryogenic temperatures. Details and difficulties on the measurements using SQUID
magnetometer can be found in (48 ).
34
2.6 Scanning transmission electron microscopy
Scanning transmission electron microscopy (STEM) is a transmission electron microscopy
operated with a focused electron beam rastering across the sample. The advancement in
aberration correction of STEM allows imaging in sub-Å resolution.49 Another benefit of
STEM is that it can utilize multiple detectors, sometimes simultaneously. One of the most
widely used detection modes is high-angle annular dark-field (HAADF) imaging. Here,
electrons that passed close to the nucleus and scattered in high angles (Rutherford scattering)
are collected, providing Z-contrast in the resulting image.50 ,51 In addition, spectroscopies
such as energy-dispersive X-ray spectroscopy (EDS) and electron energy loss spectroscopy
(EELS) can be performed in conjunction with imaging.
2.7 Density functional theory
Density functional theory (DFT) calculation, in a nutshell, is to approximately solve the
Schrödinger equation
h̄2
− ∇2Ψ(r) + V (r)Ψ(r) = EΨ(r) (2.1)
2m
for electrons. The many-body Schrödinger equation can be reduced by local density approx-
imation (LDA) or generalized gradient approximation (GGA) into a self consistent Kohn-
Sham equation
[ ]
h̄2
− ∇2 − Vion(r) + VH(r) + Vxc(ρ(r)) ϕ(r) = ϵϕ(r) (2.2)
2m
where each term in the square bracket corresponds to kinetic energy, Coulomb interac-
tion between electrons and nucleus, Coulomb interaction between electrons, and exchange-
correlation potential, which depends on the electron density ρ(r).
First, a well-defined unit cell with atoms and corresponding coordinates needs to be
35
created. This cell does not necessarily have to be the crystallographic unit cell but can
be a supercell with diluted defect concentration or a slab with vacuum for surface energy
calculation. This unit cell is repeated so that a periodic potential can be assumed. Then,
according to the Bloch theorem, under the translation by lattice vector R, the wave function
Ψ will change by a phase factor; in other words, this can be expressed as
Ψ(r+R) = eik·RΨ(r) (2.3)
where r and k are position (real space) and wave vector (reciprocal space), respectively. This
equation then becomes
Ψ(r+R) = eik·Ru(r) (2.4)
with u being a function with the same periodicity as the unit cell. Vion(r) is approximated
(so-called the “pseudopotential”) below a critical cutoff radius from the core since the local
core electrons do not participate in chemical bonding, although some magnetic interaction
will be ignored. k in the first Brillouin Zone should generate independent equations of
Equation 2.4 which will be computationally consuming. Hence, this is approximated by
sampling the Brillouin Zone and reducing it further according to crystal symmetry.
The author used the Vienna Ab initio Simulation Package (VASP) for DFT calculations
in this thesis. The program solves the Kohn-Sham equation, updates ρ(r), and compares
with the initial value to check self-consistency by total energy. The solution is yielded when
the convergence is met. In the previous paragraph, we described the three basic input files of
VASP: POSCAR (unit cell), POTCAR (pseudopotential), and KPOINTS (sampled Brillouin
Zone). The last necessary input file is INCAR, which contains various control parameters,
including the cutoff energy, convergence criteria, etc.
36
Chapter 3
Composition-dependent ferroelectricity
of lutetium orthoferrite thin films
This chapter is based on a publication which the author wrote and published in Advanced
Electronic Materials (2023).52
3.1 Introduction
Oxides of rare earths and iron can form a range of crystal structures depending on their
rare earth (R) to Fe ratio, including RFe2O4, R3Fe5O12 (rare earth iron garnet), R2Fe3O7,
orthorhombic RFeO3, and hexagonal RFeO3. These materials exhibit a diverse variety of use-
ful properties such as magnetism,53 ferroelectricity,54 multiferroicity,9 magnetoresistance,55
magnetooptical activity,56 and catalytic activity.57 The electrical, magnetic, optical, and
transport properties of these and other complex oxides are strongly influenced by point de-
fects resulting from a non-ideal cation stoichiometry.16 ,19 ,58–61 Such point defects can be
present in much higher concentrations in thin films compared to bulk, due to epitaxial sta-
bilization of a crystal structure with a composition deviating from bulk.60 ,61
The orthoferrites (RFeO3) are perovskite-derived structures in which tilting of the Fe
octahedra yields an orthorhombic lattice with four formula units (f.u.) per unit cell.7
37
Fe3+ cations are coupled antiferromagnetically with Néel temperature TN ∼640 K. The
Dzyaloshinskii–Moriya interaction (DMI) between neighboring Fe3+ leads to a small canting
angle and a net moment ∼0.05 µB per f.u. at room temperature. The antiferromagnetic
configuration of Fe3+ (Γ1, Γ2, or Γ4 in Bertaut notation) and the magnetic transition temper-
atures between different antiferromagnetic states varies with the rare earth. For a nonmag-
netic R3+ such as Lu3+, the configuration remains Γ4 (GxAyFz, i.e. G-type antiferromagnet
with spins oriented along x, with secondary A-type antiferromagnetic arrangement along y,
and a net moment along z ) from 0 K up to T 62 ,63N . Some orthoferrites can exhibit multifer-
roicity at cryogenic temperatures when R3+ ions are magnetically ordered.64 The mechanism
for such ferroelectricity is the exchange interaction between R3+ and Fe3+, which leads to
ionic displacement, breaks the symmetry, and thus leads to a polarization on the order of
0.1 µC cm−2.11 ,64
Introducing room-temperature ferroelectricity into antiferromagnetic orthoferrites is an
appealing strategy to expand the range of room temperature multiferroic materials. The
most widely studied multiferroic is BiFeO3 (BFO), which is a rhombohedral or tetragonal
structure. Its ferroelectricity is derived primarily from the Bi3+ lone pairs, and it exhibits a
high ferroelectric Curie temperature (TC = 1103 K).12 As in the rare earth orthoferrites, the
magnetism in BFO arises from the canting of antiferromagnetically ordered Fe3+ spins with
TN = 643 K. Rare earth orthoferrites lack the lone pairs of BFO, and the centrosymmetric
orthorhombic space group (Pbnm) prohibits ferroelectricity. However, YFe antisite defects
in Y-rich yttrium orthoferrite YFeO3 (YFO) thin films bring about a noncentrosymmetric
structural distortion and induce ferroelectric polarization of ∼10 µC cm−2 at room temper-
ature.20 ,65 ,66 According to first principles calculations, such an antisite defect mechanism is
expected to be applicable in other rare earth rich orthoferrites, with the polarization increas-
ing with decreasing ionic radius of R3+; LuFeO3 (LFO) lies at the higher end of predicted
polarization.20 It is therefore of interest to investigate the effect of antisite defects on the
ferroic properties of LFO.
38
Previous studies of the stoichiometric orthoferrite LFO report that remnant polarization
(Pr) ranges from several nC cm−2 (polycrystalline LFO)67–69 up to ∼15 µC cm−2 (thin
film LFO)70 at room temperature. The ferroelectricity in polycrystalline LFO was first
explained by exchange striction at magnetic domain walls,67 or by charge disproportionation
from increased covalency,69 and the higher polarization in thin films was ascribed to the
tetragonally strained structure with an out-of-plane to in-plane lattice parameter ratio of
1.045.70 Polycrystalline Y1−xLuxFeO3 also showed unsaturated polarization vs. electric field
(P-E) loops, and the hysteresis was attributed to symmetry breaking from the mixing of
Lu and Y ions.71 A metastable hexagonal LFO phase was found to be ferroelectric due
to its noncentrosymmetric crystal structure, with TC = 1020 K and Pr up to ∼10 µC
cm−2 at room temperature.33 ,72 ,73 Both orthorhombic and hexagonal LFO phases have been
grown in the form of thin films, using pulsed laser deposition (PLD),70 ,72 ,74 radio frequency
magnetron sputtering,75 ,76 or molecular beam epitaxy.33 When the two phases are grown
together, the film can show multiferroicity with a magnetization of ∼0.24 µB per f.u. from
the orthorhombic phase and a polarization of ∼5 µC cm−2 from the hexagonal phase.73
Orthorhombic LFO is a good candidate for defect engineered multiferroic material be-
cause the magnetic order persists above room temperature (TN = 623 K), in contrast to
hexagonal LFO with T 33N = 155 K. However, so far, orthorhombic LFO has not received as
much attention as hexagonal LFO, and the stoichiometry range this material can tolerate as
a single-phase film has not been explored.
Here, we grow epitaxial orthorhombic LFO thin films with Lu/Fe ratio varying from 0.6
to 1.5, and characterize their structural, magnetic, and ferroelectric properties at room tem-
perature. LFO films with different orientations grown on SrTiO3 (STO) form a single-phase
perovskite despite their non-ideal cation stoichiometry, which in bulk would favor the for-
mation of secondary phases. The observation of ferroelectric hysteresis by different methods
(piezoresponse force microscopy (PFM), P-E loops, and positive-up-negative-down (PUND)
measurements) all support the hypothesis of antisite-defect-mediated ferroelectricity in or-
39
thoferrites with non-stoichiometric R:Fe ratio.
3.2 Growth and structural characterization with varying
cation stoichiometry
LFO thin films with Lu/Fe ratios of 0.6, 0.8, 1.0, 1.2, and 1.5 were grown on (001)-
oriented STO (insulating) and Nb-doped STO (NSTO, conductive) substrates with a =
3.905 Å (Figure 3.1a). There was no secondary phase observed in the wide angle range X-
ray diffraction (XRD) scans (15-80◦, Figure 3.A1) for any of the films, despite the fact that
according to the Lu-Fe-O ternary phase diagram,77 LFO does not accommodate excess Lu
or Fe, instead forms additional phases such as LuFe2O4, Lu3Fe5O12, Lu2Fe3O7, hexagonal
LuFeO3, or binary oxides when Lu/Fe differs from 1.0. Due to the ionic radius difference
between Lu3+ (86 pm) and Fe3+ (60 pm), as the film becomes Lu-rich, the XRD peak position
2θ of (002)p (p denotes the pseudocubic notation) moves to a lower angle and the out-of-
plane lattice parameter cp increases from 3.789 Å (Lu/Fe = 0.6), 3.793 Å (Lu/Fe = 0.8),
3.800 Å (Lu/Fe = 1.0), 3.820 Å (Lu/Fe = 1.2), to 3.846 Å (Lu/Fe = 1.5). The monotonic
trend in the peak position differs from that observed in YFO, when the 2θ angle increased
then decreased as the film composition moved from Y-rich to Fe-rich.20
The reciprocal space mapping (RSM) data of the (103)p reflection, shown in Figure 3.1d
for Lu/Fe = 1.2 and Figure 3.1e for Lu/Fe = 1.0, indicates cube-on-cube epitaxy of the LFO
on STO substrates. Although the crystal quality of Fe-rich films was not good enough to
produce a peak in RSM (Figure 3.1f), the peak positions of the other films indicate in-plane
incoherency. In-plane pseudocubic lattice parameters extracted from RSM are 4.017 Å and
3.987 Å for Lu/Fe = 1.2 and Lu/Fe = 1.0, respectively, compared with out-of-plane lattice
pseudocubic lattice parameters 3.820 and 3.800 Å. Unit cell volumes are greater by ∼10 %
compared to bulk LFO, which is consistent with a greater oxygen deficiency in thin films
vs. bulk. In contrast, the YFO film unit cell volume was only 1 % greater than that of
40
Figure 3.1: XRD around the substrate (002) and film (002)p reflections for samples with
(a) different Lu/Fe ratio, (b) different PO2 during growth, and (c) different substrates. Re-
ciprocal space mapping around the (103) substrate and (103)p film reflections for Lu/Fe =
(d) 1.2, (e) 1.0, and (f) 0.8 films on STO. (g) AFM scan of Lu/Fe = 1.2 film. (h) Epitaxy
scheme of LFO growth on perovskite substrates (o stands for orthorhombic and p stands for
pseudocubic).
bulk YFO.20 Atomic force microscope (AFM) images of the Lu/Fe = 1.2 film, Figure 3.1g,
shows island-like roughening common to PLD-grown films; the root mean square roughness
41
is 0.460 nm for a ∼30 nm thick film.
3.3 Electrical and magnetic characterization of Lu-rich
lutetium orthoferrite
Figure 3.2: Magnetic hysteresis loop of Lu/Fe = 1.2 film measured at 300 K. The inset shows
the low-field regime.
Figure 3.2 is the magnetic hysteresis loop of the Lu/Fe = 1.2 film measured at room
temperature, with the magnetic field applied in both out-of-plane (OP) and in-plane (IP)
direction. The out-of-plane moment saturates at around 1 T with a saturation magnetization
Ms ≈ 0.8 kA m−1, which is about an order of magnitude lower than that of bulk.62 However,
the in-plane moment does not saturate up to 2 T. This suggests that the orthorhombic c axis
in Pbnm notation is perpendicular to the film, because the weak ferromagnetism from DMI
is along the c axis. Based on the magnetic properties and structural analysis, we conclude
that the epitaxy of LFO films on perovskite substrates can be represented as Figure 3.1h.
The orthorhombic a and b axes lie in-plane, 45◦ rotated with respect to the pseudocubic
unit cell, and the orthorhombic c axis lies out-of-plane with lattice parameter = 2cp.
In order to characterize the ferroelectric properties, we performed switching spectroscopy
via PFM (SS-PFM) and used the PFM tip to write polarization patterns on LFO thin films.
Figure 3.3a,b,c shows the SS-PFM amplitude and phase hysteresis of LFO films with Lu/Fe
of 1.2, 1.0, and 0.8, along with the remnant polarization after writing a “box-in-box” pattern,
42
i.e. polarizing a 1 µm square region at −8 V tip bias then a smaller square region at +8
V. The Lu-rich film shows ferroelectricity which can be interpreted using the antisite defect
mechanism predicted in (20 ). SS-PFM hysteresis was observed in several Lu-rich films grown
with different orientations or at different PO2 (Figure 3.3d,e). Moreover, films grown with
top and bottom conductive SrRuO3 (SRO) layers also exhibit ferroelectric switching (Figure
3.A2a,b,c). With the SRO top layer, the tip is in contact with a conductive layer which rules
out surface charging artifacts of PFM.
All the Lu-rich, stoichiometric, and Fe-rich LFO films showed hysteretic behavior in PFM,
although the amplitude of SS-PFM and the phase contrast produced by domain writing is
much smaller for Lu/Fe = 1.0. This behavior differs from that of YFO films where the
ferroelectric response was limited to Y-rich compositions. The ferroelectricity in Y-rich YFO
was explained by antisite defects YFe on the octahedral sites which lower the symmetry to a
noncentrosymmetric R3c structure.20 If the same mechanism exists in LFO, it can account
for the ferroelectric response in the Lu-rich composition. For the film with Lu/Fe = 1.0, we
hypothesize that the (weaker) ferroelectricity may also be caused by antisite defects, even
though the composition would suggest that few such defects need be present. The XRD peak
intensities provide support for the presence of antisite defects (i.e. both LuFe and FeLu) in the
film with Lu/Fe = 1.0. Using a pseudocubic ABO3 perovskite unit cell, the structure factors
Fhkl of the (001) and (002) reflections are F001 = fA − fB − fO and F002 = fA + fB + 3fO.
The peak intensities Ihkl are then:
I001 ∝ f
2
A + f
2 2
B + fO − 2fAfB − 2fAfO + 2fBfO (3.1)
I ∝ f 2002 A + f
2
B + 9f
2
O + 2fAfB + 6fAfO + 6fBfO (3.2)
The (θ-dependent) values of fA and fB are modified by the presence of antisites, FeLu on the
A sites and LuFe on the B sites. By weighting fA and fB with the antisite concentration (i.e.,
fA = fLu× (fraction of Lu sites occupied by Lu) +fFe× (fraction of Lu sites occupied by Fe),
43
Figure 3.3: SS-PFM hysteresis and phase contrast after writing domain structures with a
bias of ± 8 V on (001)p oriented LFO films with Lu/Fe stoichiometry of (a) 1.2, (b) 1.0, and
(c) 0.8. SS-PFM of Lu/Fe = 1.2 film grown at different conditions, (d) PO2 = 50 mTorr and
(e) in (111)p orientation. (f) SS-PFM results of different orthoferrites performed with the
same tip. BFO, LFO, and YFO films are 23 nm, 33 nm, and 40 nm, respectively.
etc.), we calculate the intensity ratio IR = I(001)/I(002) shown as a contour plot in Figure
3.A3. For the case with minimum antisites, i.e. none for Lu/Fe = 1.0, and LuFe only for
excess Lu, we find that IR(Lu/Fe = 1.0) = β IR(Lu/Fe = 1.2) where β = 1.4. However, the
experimental data gives a ratio of β = 1.2. This is inconsistent with zero defects in the film
with Lu/Fe = 1.0, and instead suggests the presence of LuFe and FeLu according to Figure
44
3.A3. This analysis was not applied to the data from the Fe-rich films because the poorer
crystallinity itself lowers the peak intensities (both (103)p in RSM and (003)p in XRD are
absent), and excess Fe also leads to other defects such as dislocations or cation vacancies
observed in Fe-rich YFO which may also occur in Fe-rich LFO.
In order to qualitatively compare the magnitude of polarization among different orthofer-
rites, SS-PFM was performed using the same tip on Lu-rich LFO, Y-rich YFO, and BFO, and
the results are shown in Figure 3.3f. The PFM amplitude is the highest in BFO, followed by
LFO, and YFO. In a ferroelectric material with 4mm symmetry, the piezoelectric coefficient
d33 and polarization P3 are related by d33 = 2ϵ0ϵ33Q3333P3 where ϵ0 is the permittivity of
vacuum, ϵ33 is the relative permittivity and Q3333 the electrostrictive coefficient,78 therefore
we expect the PFM amplitude to be proportional to the polarization. We also note that
perovskite ferroelectrics tend to show higher PFM amplitude when they are thicker.79–81
Assuming LFO and YFO have similar ϵ33 and Q3333 and considering that the LFO sample
is thinner than YFO, the measurements of PFM amplitude, which is proportional to d33,
implies that LFO has a higher polarization than YFO (as predicted in (20 )), but lower than
BFO.
Figure 3.4: (a) P-E hysteresis (10 kHz) overlaid with the measured current and (b) PUND
result (pulse width 1.5 ms, pulse delay 100 ms) of a 73 nm thick Lu/Fe = 1.2 LFO/NSTO
film with Pt top electrodes. (c) Impedance spectroscopy measured from 1 MHz to 1 Hz
(open circuit condition).
P-E loop and PUND results of Lu-rich LFO with a Pt top electrode are shown in Figure
3.4a,b. In Figure 3.4a, a clear ferroelectric saturation feature is observed at high positive
bias and switching current peaks are observed for both bias directions; however, the P-E loop
45
at negative bias shows a more resistive-like behavior. Current leakage is also confirmed by
the increase in the current at higher bias, similar to that observed in Hf0.5Zr O .820.5 2 The P-E
loop shape indicates that the LFO has a lower resistivity compared to other ferroelectrics (i.e.
PbZrxTi1−xO3, PZT) and a diode-like behavior from the asymmetric electrode structure. The
resistivity of LFO was 1.5×108 Ω cm (Figure 3.4c), measured by impedance spectroscopy
to minimize the contribution of contact resistance. This value agrees with the resistivity
calculated from the current flow at 1 V bias. The LFO resistivity is two or more orders
of magnitude lower than that of PZT.83 We attempted to increase the resistivity by using
Ti4+ as a dopant under the assumption that LFO is a p-type conductor.84 Although Ti was
successfully incorporated in the LFO without producing any secondary phases, the resistivity
decreased to ∼106 Ω cm, implying the leakage was due to a different mechanism.
Asymmetric P-E loops (ferroelectric hysteresis for one direction of bias and resistive
for the other) have been reported in BFO and PZT. VO accumulation at the interface in
BFO85–88 caused such loops, and different diode characteristics at the interface may have
contributed to asymmetry as well. We attempted to reduce the leakage and its directionality
by using SRO as both top and bottom electrodes, but the asymmetry and leakage were
still present as shown by the rounded hysteresis loop shape (Figure 3.A2d). This may be
a result of inequivalent growth condition of the SRO, in which the bottom layer is grown
directly on the substrate whereas the top layer is grown on the strain-relaxed LFO film.
These results, along with studies on BFO or PZT,72 ,83 highlight the importance of process
parameters in ferroelectric film growth. During P-E measurements, some electrodes exhibited
capacitive behavior at first, but applying a pulse similar to or higher than the coercive
field could activate the film into showing ferroelectric behavior. A similar phenomenon was
observed in PZT by removing pinned domains,89 although the mechanism for LFO may
not be the same, considering that LFO is an improper ferroelectric whereas PZT (or BFO)
are proper ferroelectrics. The primary order parameter for proper ferroelectric transitions
is polarization, while in improper ferroelectrics, polarization is derived from other order
46
parameters (e.g. structural distortion arising from Lu ).7 ,72Fe
PUND measurements were performed to exclude parasitic contributions from the P-E
hysteresis and to extract the switchable polarization of LFO. The difference between the rem-
nant polarization from the switching pulse (P ∗r ) and the nonswitching pulse (Pr̂) is equivalent
to the net switchable polarization (QSW , or 2Pr). QSW and −QSW derived from Figure 3.4b
are 10.8 µC cm−2 and −15.8 µC cm−2, respectively, and the difference presumably rises from
the asymmetric electrode configuration. The average value of 13.3 µC cm−2 is higher than
that of Q −2SW = 7.2 µC cm for YFO, which is in line with the hypothesis that Lu-rich LFO
would have a higher polarization than Y-rich YFO.20 Nevertheless, high ±Pr̂ in the PUND
measurement indicates that the leakage which is concurrent with ferroelectric switching is
significant when the field approaches and overcomes the coercivity, and it is also consistent
with the leaky behavior in the P-E loop in Figure 3.4a. Reducing the leakage of LFO films is
necessary to improve the performance of LFO as a room temperature multiferroic material.
3.4 Defect formation energy calculation
The feasibility of finding both LuFe and FeLu antisite defects in LFO films (as mentioned
in Section 3.3) is supported by density functional theory (DFT) calculations. The defect
energies are shown in Table 3.1, which were calculated using the equation
∆E = E(defective LFO) + (µ of defect species)− E(defect-free LFO) (3.3)
under the assumption of Fe-rich and O-rich conditions and with in-plane pseudocubic lattice
parameter of 3.905 Å (crystal structures are shown in Figure 3.A4). While the formation
energies of cation or anion vacancies range from 4.74 to 6.97 eV, the formation of antisite
defects are energetically more favorable suggesting abundant distribution of these defects
in thin films. This highlights the feasibility of antisite defect formation in LFO, which may
explain the presence of ferroelectricity in LFO of various compositions including Lu/Fe = 1.0
47
via the mechanism presented for Y-rich YFO.20 It is worth noting that a negative formation
energy of the FeLu defect is associated with intrinsic instability of bulk orthorhombic phase
LFO. In agreement with (77 ), the thermodynamic phase diagram (Figure 3.A5) constructed
based on the calculated enthalpies of Fe2O3, Lu2O3, and LFO oxides indicates that the
orthorhombic LFO structure is not stable, consistent with the depiction as a point in the
ternary phase diagram, and can be stabilized by epitaxial growth.
√ √
Table 3.1: Formation energies (∆E) per orthorhombic 2 × 2 × 1 supercell (in-plane
lattice = 7.810 Å) of different point defects in LFO. ∆E was calculated with respect to the
defect-free LFO strained on STO as the ground state.
Defect VO (IP) VO (OP) VLu VFe LuFe FeLu LuFe + FeLu
∆E (eV) 4.74 5.10 5.57 6.97 2.89 −1.43 1.32
3.5 X-ray absorption analysis of lutetium orthoferrite thin
films
X-ray absorption spectroscopy (XAS) utilizes the excitation of the core electrons to the
empty states at elemental characteristic energies, thus providing a useful tool for probing
the unoccupied density of states. In LuFeO3 thin films, antisite defects particularly affect
the local chemical environment of Fe—O and Lu—O bondings and thus the resulting XAS
spectra Figure 3.5. O K-edge refers to the excitations from O 1s orbital to O 2p orbital,
and Fe L-edge refers to the excitations from Fe 2p orbital to 3d orbital. In orthoferrites, the
unoccupied states of O K-edge can be categorized into three stages in the order of ascending
energy: Fe 3d, R 5d, and Fe 4sp, all hybridizing with O 2p.90 In a perfect perovskite (i.e.,
tolerance factor, t = 1), the A-site will be coordinated by 12 oxygen atoms. Depending on the
size of the rare earth, the nominal oxygen coordination varies because t < 1 in orthoferrites; in
late rare earths where the ionic radii become even smaller, it can be considered as surrounded
48
by 6 oxygens similar to a prism with a three-fold symmetry.91 ,92 Therefore, the Fe—O
bonding of Fe in octahedral sites vs. prismatic R sites (i.e., antisites) will differ, resulting in
the emergence of an additional state or the intensity change between t2g and eg orbitals, as
observed in Figure 3.5b. Fe L-edge in Figure 3.5c rather conventional. The left L3 and right
L2 peaks correspond to L + S and L − S selection rules and each of them shows t2g (lower
in energy, dxy, dxz and dyz) and eg (higher in energy, dz2 and dx2−y2) splitting which arises
from the octahedral coordination and crystal field splitting.
Figure 3.5: X-ray absorption spectra of LuFeO3 with different Lu/Fe ratios: (a) O K-edge,
(b) enlarged area of O 2p and Fe 3d hybridization range, and (c) Fe L-edge. The spectrum
was collected at the ALBA BL29 - BOREAS beamline.
X-ray magnetic linear dichroism (XMLD) refers to the difference in XAS of linearly
polarized X-rays. The absorption is higher when the Néel vector is parallel to the polarization
than when the two are perpendicular. XMLD photoemission electron microscopy (PEEM)
exploits this phenomenon to image magnetic domains of antiferromagnets by collecting the
emitted secondary electrons.93 The data (intensity per pixel as a function of X-ray energy)
can be processed in many ways, such as using the ratio of the intensities, subtracting the
two orthogonal linear polarizations, or calculating the difference over the sum. Figure 3.6
represents the antiferromagnetic domains of LuFeO3 thin films with two perpendicular Néel
vectors that are ± 45◦ to the pseudocubic in-plane axes (Figure 3.1h) distinguished by bright
and dark contrast. In this case, the data was obtained by subtracting the intensity between
the t2g peak and valley between t2g and eg of L2 at each pixel. Reversing the polarization
from linear horizontal (LH) to linear vertical (LV ) switches the contrast of the domains. The
49
shape of the domains resembles another orthoferrite LaFeO 943, and the size does not show
significant dependence on the Lu/Fe ratio.
Figure 3.6: Antiferromagnetic domains of LuFeO3 with different Lu/Fe ratios imaged by
XMLD-PEEM. The data was collected at the MAX IV MAXPEEM beamline. The incident
light was perpendicular to the film surface.
3.6 Conclusion
This study presents the structural, magnetic, and ferroelectric behavior of epitaxial LFO
thin films grown using PLD, particularly revealing a ferroelectricity depending on Lu/Fe sto-
ichiometry which is attributed to an antisite-defect-mediated mechanism. The out-of-plane
magnetic hysteresis of LFO films shows a small canted ferromagnetism and thus corresponds
to the orthorhombic c axis in Pbnm notation. Using SS-PFM, particularly focusing on
the Lu-rich composition with Lu/Fe = 1.2, we demonstrate room temperature ferroelectric
behavior over a range of growth conditions. Moreover, we attribute the presence of ferro-
electricity across all compositions to the abundance of antisite defects in LFO, which have a
low formation energy according to DFT. The remnant polarization obtained from the PUND
measurement is 13.3 µC cm−2 for Lu-rich LFO, higher than that of Y-rich YFO. However,
the ferroelectric performance is limited by the low resistivity of LFO. This work realizes
room temperature multiferroicity in films of an orthoferrite, LFO, and supports the model
of ferroelectricity derived from antisite defect engineering in rare earth orthoferrites.
50
3.7 Methods
Epitaxial LFO films were grown by PLD using a 248 nm wavelength KrF excimer laser
with a fluence of ∼2 J cm−2 from oxide targets prepared with two different Lu/Fe ratios, 0.6
(Lu3Fe5O12, the stoichiometry of garnet) and 1.4. The Lu/Fe stoichiometry of the targets
and film end members were measured by wavelength dispersive spectroscopy (WDS) using
a JEOL-JXA-8200 Superprobe. The Lu/Fe = 0.6 target resulted in a film with the same
Lu/Fe ratio, however, the Lu/Fe = 1.4 target resulted in a film enriched in rare earth (Lu/Fe
= 1.5). A similar result was also seen in the growth of Y-rich YFO from a stoichiometric
YFO target.20 The Lu/Fe ratio of the film was varied by adjusting the shot ratio between
the two targets based on their growth rates, to achieve the desired stoichiometry. The
substrate heater temperature setpoint was 900 ◦C (actual substrate temperature ∼650 ◦C)
and PO2 was 10 to 150 mTorr during growth. For comparison, YFO was grown at the same
deposition conditions as LFO from an oxide target with Y/Fe = 1 resulting in a film with
Y/Fe ≈ 1.2, and BFO was grown in a different chamber using a Bi/Fe = 1.2 oxide target
with substrate heater setpoint of 700 ◦C (actual substrate temperature ∼550 ◦C). For some
samples, conductive layers of SRO were grown in-situ, above and below LFO, at the same
PO2 and temperature as LFO.
Structural characterization by XRD was done using a Rigaku Smartlab high resolution
diffractometer with a Cu Kα1 (wavelength 1.5406 Å) X-ray source. RSM was performed
with a Bruker D8 Discover diffractometer, also using a Cu Kα1 source. Magnetic hystere-
sis was measured in a Quantum Design MPMS3 SQUID magnetometer at 300 K. AFM
imaging was performed using a Bruker Dimension Icon XR scanned probe microscope, and
PFM was performed using an Asylum Research Cypher VRS AFM in dual AC resonance
tracking (DART) mode. Both probe microscopies were done using Pt-coated Si probes from
MikroMasch (HQ:NSC18/Pt). SS-PFM data was processed with a simple harmonic oscilla-
tor model using built-in software. A Radiant Technologies Precision Premier II tester was
51
used for P-E hysteresis and PUND pulse measurements. Prior to electrical measurements,
Pt electrodes with a diameter of 200 µm were deposited by sputtering using a shadow mask,
and for samples with SRO layers, the top SRO layer was patterned by photolithography (Hei-
delberg MLA 150) and ion milled, while the underlying SRO layer below the LFO remained
continuous.
DFT studies were done using the Vienna Ab initio Simulation Package (VASP).95 ,96 Gen-
eralized gradient approximation (GGA) parameterized by Perdew-Burke-Ernzerhof (PBE)97
functional was used for the exchange-correlation term. The plane-wave cutoff energy was
500 eV. The rotationally averaged Hubbard correction with an effective Ueff = 4 eV for Fe
3d electrons was used. Lu 5p65d16s2, Fe 3p63d74s1, and O 2s22p4 electrons were treated as
valence. Enthalpies of oxide formation and defect formation energies were adjusted, taking
into account of O2 overbinding and the use of Hubbard U correction scheme according to
√ √
(98 ) and (99 ), respectively. In the case of LFO, a 2 × 2 × 1 supercell with 8 f.u. was
used with a 4 × 4 × 4 Monkhorst-Pack k-point grid. For defect formation energy calcula-
tions, the structures were epitaxially strained to the lattice parameter of STO (a = 3.905
Å), but relaxed in the out-of-plane direction. The formation enthalpy of each compound was
calculated by
∆HFe O = EFe O − 2EFe − 3µ2 3 2 3 O
∆HLu O = ELu O − 2ELu − 3µ (3.4)2 3 2 3 O
∆HLuFeO = E3 LuFeO − E3 Lu − EFe − 3µO
where µO = 1/2EO . The formation enthalpies of Fe2O3, Lu2O3, and LFO were −8.472
eV, −20.67 eV, and −14.55 eV per f.u., and our choice of parameters with Fe Hubbard U
correction99 showed a good agreement with the experimental formation enthalpies of binary
oxides Fe2O3 and Lu O .1002 3
52
3.8 Appendices
Figure 3.A1: XRD of LFO films grown on (001), (110), and (111) oriented NSTO substrates.
(001) result corresponds to Figure 3.1a. The atomic structure is the orthorhombic unit cell
of LFO viewed along the (101)o orientation to show that (101)o is equivalent to (111)p.
53
Figure 3.A2: (a) XRD of SRO/LFO/SRO structured film with Lu/Fe = 1.2. SRO and LFO
layers were ∼10 nm and ∼70 nm thick, respectively. (b) PFM geometry and (c) SS-PFM
result with the tip placed on SRO top electrode with a diameter of 50 µm. (d) P-E measured
at 10 kHz with 50 µm diameter SRO electrode.
54
Figure 3.A3: Calculated I(001)/I(002) ratio mapped over 0-20 % LuFe and 0-20 % FeLu
antisite defects. The numbers next to the dotted lines correspond to Lu/Fe ratio of the
total structure. For example, the dotted line for Lu/Fe = 1.0 represents FeLu = LuFe. The
minimum amount of defects is zero (FeLu = LuFe = 0) and I(001)/I(002) = 0.17. When Lu/Fe
= 1.2, the minimum concentration of defects is 0.1 (FeLu = 0; LuFe = 0.1) and I(001)/I(002)
= 0.12. For this minimum-defect comparison, β = 0.17/0.12 = 1.4. In comparison, the
experimental β = 1.2, which means that there must be some antisite defects present for
Lu/Fe = 1.0.
Figure 3.A4: Relaxed crystal structures LFO from DFT calculations. Green, brown, and
red atoms correspond to Lu, Fe, and O, respectively.
55
Figure 3.A5: Phase diagram of LFO, where the blue, red, and yellow lines represent µLu =
µFe+∆HLuFeO −∆H , µ = µ −∆H +∆H , and µ3 Fe2O3 Lu Fe LuFeO3 Lu2O3 Lu+µFe+µO = ∆HLuFeO3
respectively. µLu and µFe values used for defect formation energy calculation is marked by
an arrow in the enlarged inset.
56
Chapter 4
Phase formation and exchange bias in
vertically aligned nanocomposites of
lutetium orthoferrite and cobalt ferrite
This chapter is based on a publication which the author wrote and published in ACS Applied
Electronic Materials (2024).101
4.1 Introduction
Vertical nanocomposite thin films consist of two or more materials codeposited epitaxially
on a substrate, such as pillars of one phase within a matrix of another.102–105 Coupling
between different phases leads to emergent properties; for instance, the strain-mediated
interplay between the piezoelectricity of perovskite BiFeO3 and the magnetoelasticity of
spinel CoFe2O4 leads to magnetoelectric coupling between the electrical polarization of the
BiFeO3 and the magnetization of the CoFe 103 ,1062O4. In nanocomposites where both phases
have magnetic order, a magnetic exchange coupling between the phases could provide an
additional mechanism for interaction at the interface, besides strain transfer. For example,
exchange coupling between the uncompensated spins at the interface between a ferro- or
57
ferrimagnet and an antiferromagnet can lead to a shift (exchange bias) in the magnetic
hysteresis loop107 and can yield useful behavior such as stabilizing a specific magnetization
configuration.
Exchange bias has been studied in oxide-oxide systems such as layered structures of
Bi FeO3-La0.7Sr MnO 1080.3 3 or LaFeO3-La0.7Sr0.3MnO ,1093 self-assembled nanocomposites of
BiFeO -La Sr MnO 110 ,1113 0.7 0.3 3 BiFeO3-Fe3O ,1124 or NiO-NiFe2O4,113 or even within a single-
phase La 114 1150.67Sr0.33MnO3 or YFeO3 that exhibit structural variations. In particular,
for vertical self-assembled nanocomposites, an exchange bias of ∼100 mT in a 1 µm thick
BiFeO3-La0.7Sr0.3MnO3 at 5 K, ∼4 mT in 250 nm thick BiFeO3-Fe3O4 at room temperature,
and ∼91 mT in 180 nm thick NiO-NiFe O at room temperature have been reported.111–1132 4
Orthoferrites (perovskite-structured RFeO3, R: rare earth, Y) are especially attractive anti-
ferromagnets due to their high Néel temperatures ∼650 K and their compatibility in form-
ing self-assembled nanocomposites with spinels. We previously showed that both YFeO3
and LuFeO3, though nonferroelectric in bulk, can exhibit defect-mediated ferroelectricity at
room temperature.20 ,52 ,66 We therefore expect that nanocomposites consisting of antiferro-
magnetic, ferroelectric orthoferrites and ferrimagnetic spinels, here LuFeO3 and CoFe2O4
respectively, could show the effects of both strain coupling and exchange bias.
Oxide nanocomposites are usually made by pulsed laser deposition (PLD), either by
codeposition from the two targets of desired phases116 or from a single target with the aver-
aged composition.117 The resulting phases in the film are not necessarily the same as those
present in the target(s). For instance, a single SrTi0.75Cu0.25O3−δ target yielded a three-
phase thin film consisting of a Cu-SrO-Sr(Ti,Cu)O3−δ core-shell-matrix,118 and codeposition
from SrCoO3−δ and SrFeO3−δ targets produced a film consisting of a Sr(Co,Fe)O3−δ ma-
trix and CoOx pillars.119 Deposition from a La0.7Sr0.3MnO3 target formed (La,Sr)Ox and
Sr3Mn4O7 nanoparticles in addition to La0.7Sr0.3MnO3 because of lattice mismatch,120 ,121
and deposition from a NiCo2O4 target formed spinel nanocolumns of different compositions
due to spinodal decomposition.122 Phase formation in PLD is governed by multiple factors,
58
including the thermodynamic stability of competing phases, differences in surface energy and
lattice parameters, nucleation and growth kinetics, and diffusion. Such factors all depend on
the gases present in the chamber and their pressure during growth, laser energy, substrate
temperature and other growth conditions. For example, insufficient energy of the arriving
species can result in metastable phases,122 and oxygen vacancies can promote diffusion and
induce segregation of metallic phases in an oxide matrix.123 Although the target material is
vaporized during PLD, the bond energies and stoichiometry of the target affect the types
and energies of the species present in the plume, hence influencing the phases that form in
the growing film.
Here we describe the phase formation and magnetic coupling in PLD-grown vertical
nanocomposites of lutetium orthoferrite (LuFeO3, LFO), a canted antiferromagnet with
defect-mediated ferroelectricity,52 and cobalt ferrite (CoxFe3−xO4, CFO), a ferrimagnet. We
first show the effect of target composition and growth kinetics on the phase formation in
LFO-CFO nanocomposites. Depending on the selection of the targets, the nanocomposite
consisted of two (perovskite and spinel) or three (perovskite, spinel, and rocksalt) crystal
structures with varying compositions, characterized by scanning transmission electron mi-
croscopy (STEM). We then show that field cooling can induce exchange bias between the
antiferromagnet LFO and the ferrimagnet CFO. For a 31 nm thick film consisting of ∼57 %
of LuFeO3 and ∼43 % of Co1.2Fe1.8O4 by volume, an exchange bias of −5.3 mT was estab-
lished at room temperature by field cooling at an external field of +1 T. This work highlights
the importance of PLD targets in phase formation and unveils the presence of exchange cou-
pling between a canted antiferromagnet and a ferrimagnet, ascribed to the crystallographic
relationship of the interface.
59
4.2 Effect of pulsed laser deposition target material on
structural properties
4.2.1 X-ray diffraction based analysis
Nanocomposite thin films were deposited on (001)-oriented SrTiO3 (STO, lattice param-
eter a = 3.905Å) substrates using PLD. Four different types of targets were used: LF14,
which contains LuFeO3 and Lu2O3 such that the average atomic ratio of Lu/Fe = 1.4; LF06,
which contains LuFeO3 and Fe3O4 such that Lu/Fe = 0.6; CoFe2O4 (C1F2); and Co3O4
(C3). Two or three of the targets were selected for each codeposition. The combination
and shot numbers of the targets are described in detail in Table 4.1, and film growth and
characterization are described in the Methods section.
Table 4.1: Six different growth conditions used for depositing nanocomposites and their
structural properties (thickness and out-of-plane lattice parameters).
Target and shot number Phase and lattice parameter (Å)
Cycles Thickness (nm)
Sample C1F2 C3 LF14 LF06 Perovskite Spinel Rocksalt
a 25 30 80 240 88 3.800
b 60 180 50 44 3.805
c 90 90 60 31 3.801 8.383
d 12 200 150 80 3.801
e 60 120 60 20 3.795 8.3268.544
8.351
f 90 90 60 31 3.974 8.488 On surface
8.532
When films are grown from LF14 or LF06 targets onto a perovskite substrate such as STO,
the film forms a single-phase epitaxial perovskite structure even though the composition is
60
Figure 4.1: XRD results of the six nanocomposite thin films around the (002) reflection of
the STO substrate. Labels a-f corresponds to the samples in Table 4.1.
far from the stoichiometric 1:1 ratio.52 However, when a film is grown by codeposition of
LF14 or LF06 with C1F2 or C3, we find that a minority second phase of spinel, CFO,
forms with volume fraction and Co:Fe ratio that depend on the targets used and the ratio of
shots between the targets. Figure 4.1 shows the X-ray diffraction (XRD) results around the
STO (002) and LFO (002)p reflection (p denotes the pseudocubic notation) for the target
61
combinations of Table 4.1, where panels (a-f) correspond to samples a-f. The perovskite
(orthoferrite) peaks appear at 2θ ≈ 48◦. The spinel (004) peak is observed at 2θ ≈ 43◦
for larger Co contents, when the shot number ratio of C1F2 or C3 to LF14 or LF06 is 0.5
or higher (samples c, e, f). However, magnetic measurements discussed below suggest the
presence of magnetic CFO nanoparticles in other films (samples a, b, d) despite the absence
of a visible spinel peak in XRD.
The perovskite LFO (002)p peak position is similar for the two-phase films made with
the Fe-rich LF06 target (samples d, e, f) and the Lu-rich LF14 target (samples a, b, c). In
contrast, prior work on single-phase LFO films showed that the peak position depends on
Lu:Fe ratio.52 This suggests that the Lu:Fe composition of the LFO in the two-phase films is
not simply defined by the target composition because the second phase spinel acts as a sink
for Fe and Co. Although several spinel peaks are present in samples e and f, it is challenging
to determine the Co:Fe ratio of the spinel based on XRD because the lattice parameter of
CFO varies only slowly with Co:Fe composition, and is affected by the presence of strain.
The bulk (004) peaks of Fe3O4, CoFe2O4, and Co ◦ ◦ ◦3O4 appear at 43.05 , 43.06 , and 44.82
respectively and the (002) reflection of rocksalt CoO is at 42.37◦124–127 making them hard to
distinguish in the XRD scans of thin films. For sample f, an additional peak at 61.52◦ was
observed (Figure 4.A1), which can correspond to either (440) of spinel or (220) of rocksalt.
This will be discussed further below, based on STEM analysis.
4.2.2 Electron microscopy based analysis
To better understand the microstructure of these nanocomposites, we performed scan-
ning electron microscopy (SEM) and cross-sectional high angle annular dark field (HAADF)
STEM imaging of films made from two different target combinations of Table 4.1, sample
c: C1F2(90)-LF14(90) and sample f: C3(90)-LF06(90), where the number in parentheses
represents the number of shots per cycle from the target. First, the surface morphology
from SEM indicates drastic differences between the two (Figure 4.A2); the top surface of
62
Figure 4.2: Cross-sectional STEM images of the nanocomposite films. Overview of (a)
C1F2(90)-LF14(90) and (b) C3(90)-LF06(90) samples, where dark phases correspond to
spinel pillars and the bright phases correspond to the perovskite matrix. Higher magnifi-
cation of (c) C1F2(90)-LF14(90) and (d) C3(90)-LF06(90) samples, highlighting the crystal
structures of each region. The panels on the right are magnified images of corresponding
areas, and the contrast has been enhanced to better visualize the atoms. The scale bars
correspond to 1 nm. Yellow, red, and blue atoms correspond to dodecahedral, octahedral,
and tetrahedral sites. A spinel antiphase boundary is also visible in (c), marked by an arrow.
C1F2(90)-LF14(90) shows an irregular granular structure, whereas C3(90)-LF06(90) shows
square or rectangular crystals with surface facets likely (111) planes commonly seen in the
spinel phase of perovskite-spinel nanocomposites.103 ,128 The elongated features are believed
to correspond to the (110)-oriented crystals seen in the XRD scan in Figure 4.A1. Figure
4.2a and 4.2b show STEM images of the C1F2(90)-LF14(90) and C3(90)-LF06(90) nanocom-
posites, revealing a pillar and matrix geometry in the cross-section. As the contrast in the
HAADF STEM images is approximately proportional to the atomic number, Z1.7,51 LFO
appears brighter than CFO because of the heavy rare earth element Lu (Z = 71) compared to
Co (Z = 27). The vertical microstructures revealed in cross-section are typical of vertically
63
aligned nanostructures with a range of thicknesses.66 ,129 ,130
Figure 4.2c shows the presence of spinel CFO and perovskite LFO in C1F2(90)-LF14(90)
which meet in (111)/(111)p interfaces oriented at 55◦ to the substrate, analogous to other
perovskite-spinel nanocomposites.66 In contrast, C3(90)-LF06(90) (Figure 4.2d), shows three
crystal structures: perovskite, spinel, and rocksalt. Rocksalt preferentially forms on the top
surfaces of spinel pillars (Figure 4.2d and 4.A3a). The conversion of spinel to rocksalt
preserves the f.c.c. oxygen arrangement but the Fe and Co ions occupy octahedral instead
of tetrahedral plus octahedral sites, with a lattice expansion. The transition from spinel to
rocksalt may be a strain relaxation mechanism or be driven by a reaction of the spinel phase
with oxygen. However, the spinel would already be under compressive strain from epitaxy
with the substrate or the LFO, and this would be increased by conversion to rocksalt, arguing
against a strain relaxation mechanism. We also note that spinel with (011) orientation, which
produced a peak at 61.52◦ in XRD (Figure 4.A1f), was observed along the [211̄] zone axis in
Figure 4.A3a.
Figure 4.3a depicts the energy dispersive spectroscopy (EDS) maps of sample c, C1F2(90)-
LF14(90) where LFO (brighter contrast) and CFO are visible. Electron energy loss spec-
troscopy (EELS) analysis (Figure 4.4a) highlights the different chemical environments in
the perovskite LFO and spinel CFO. The O K-edge in orthoferrites contains three peaks
corresponding to the hybridization between O 2p and the transition metal 3d (∼531 eV),
rare earth 5d (∼539 eV), and transition metal 4sp (∼561 eV),131 whereas in spinel, the pre-
peak and main peak correspond to O 2p and transition metal 3d (∼531 eV) and 4sp (∼541
eV) hybridization.132 The peak associated with the rare earth 5d orbital is broader than
the peaks coming from hybridization between transition metals. The shape of Fe L-edges
in Figure 4.4a share the characteristics of BiFeO and CoFe O ,133 ,1343 2 4 indicating that the
valence state of Fe is likely 3+ in both LFO and CFO.
In contrast, EDS maps of sample f, C3(90)-LF06(90) (Figure 4.3b and Figure 4.A3c)
revealed that the spinel/rocksalt pillars exhibit a range of different cation compositions:
64
Figure 4.3: EDS maps of Lu, Fe, and Co, corresponding to (a) C1F2(90)-LF14(90) and
(b) C3(90)-LF06(90). In (b), pillar A corresponds to the HAADF image in Figure 4.2d,
and details of pillar B are described in Figure 4.A3a. The intensity corresponds to atomic
content of each element.
mainly Co (e.g. pillars B, C, D), mainly Fe (e.g. pillar E), and both Co and Fe (e.g.
pillars A, F). Pillar A from Figure 4.2d forms rocksalt at the surface; however, no noticeable
change in the Co and Fe composition at the spinel-rocksalt transition is observed through
EDS. We have also seen the coexistence of spinel and rocksalt structures in CoOx embedded
in Sr(Co,Fe)O 1193−δ. EELS analysis in Figure 4.4b further confirms that pillar B can be
identified as spinel Co3O4 instead of rocksalt CoO because of the presence of the pre-peak
around 531 eV.132 ,135 Therefore, we can conclude that the spinel pillar phase composition
CoxFe3−xO4 in this nanocomposite can range from x = 0 to 3. From the O K-edges, it is
notable that the main peak energy is lower in CFO compared to Co3O4, although the pre-
peak onset energy is the same. The larger gap between the two peaks in Co3O4 compared
to CFO implies that oxygen vacancies are more abundant and the average valence state
is lower in CFO compared to Co O 1363 4. In addition, the LFO of C3(90)-LF06(90) is more
oxygen-deficient compared to that of C1F2(90)-LF14(90) because the first peak (O 2p and
65
Fe 3d) is absent and the second peak (O 2p and Lu 5d) has broadened.
Figure 4.4: EELS results of (a) C1F2(90)-LF14(90) and (b) C3(90)-LF06(90) nanocompos-
ites. The left HAADF image shows the area where each spectrum was collected (the number
of pixels was kept constant for different regions), followed by O K-edge, Fe L-edge, and Co
L-edge (only for (b)). Transition metal L3 and L2 edges are labeled in (a) as a representa-
tive. Scanned region in (b) is the same as Figure 4.3b (pillars A and B are labeled). The
Lu spectrum was not measured because the peak energies were out of the detectable range.
A small amount of Co and Fe has been detected in the L edges of LFO and Co3O4 in (b),
which is likely due to the dense population of the pillars causing different grains to overlap
through the thickness of the lamella.
To summarize, the nanocomposite synthesized from Lu-rich LFO (LF14) and CoFe2O4
(C1F2) targets (sample c) consists of two phases, LFO and CFO. The other sample made
from C1F2-LF14 (sample b) and one made from three targets C1F2-LF14-LF06 (sample a)
also consist of perovskite and spinel. The LFO (002)p peak position is similar for samples a,
b, c regardless of the shot ratios, but the spinel peak intensity is highest for sample c with the
largest shot ratio from the C1F2 target. In contrast, when Fe-rich LFO (LF06) and Co3O4
(C3) targets were used (sample f), the film consisted of LuFeO3, spinel CoxFe3−xO4 with a
range of composition x, and rocksalt that transitioned from the parent spinel phase. Samples
d and e were also made from C3-LF06 targets and show the LFO (002)p peak shifting towards
a higher angle and increasing spinel or rocksalt peaks with increasing fraction of C3.
To explain the phase formation, in the C1F2-LF14 samples b and c, we infer that Fe
66
arriving from the C1F2 target had sufficient mobility to diffuse into the LFO and compensate
for the Fe-deficient flux from the LF14 target. This would produce LFO with a composition
closer to the 1:1 stoichiometry, and a spinel containing both Co and Fe. In contrast, for
C3-LF06 samples d, e, f, the Fe and Co are delivered by different laser pulses and the excess
Fe from the LF06 target is incorporated into spinel. We hypothesize that nuclei of Co-rich
spinel and Fe-rich spinel are formed separately from the fluxes arriving from the two targets
leading to the growth of spinel crystals with a range of compositions.
4.3 Magnetic characterization and exchange coupling phe-
nomenon
Magnetic hysteresis loops of the six nanocomposites are presented in Figure 4.5. Although
only three out of the six nanocomposites showed the (004) peak of spinel, the magnetization
vs. magnetic field (M-H) loops suggest that spinel is present in all the samples because the
antiferromagnetic LFO does not produce a hysteresis loop. The samples without a clear
spinel peak, a, b, d have lower Co content and lower saturation magnetization than samples
c, e, f. Sample a presents nearly isotropic hysteresis loops; however, from b to f, in-plane
(IP) and out-of-plane (OP) hysteresis loops differ, and saturation is not achieved at fields of
1 T. The M-H loops of samples e and f were measured using higher fields, and saturated at
1.3-2 T.
The magnetization of the pillars is expected to be preferentially oriented along the OP
direction due to magnetoelastic anisotropy. CoFe2O4 has a large and negative magnetostric-
tion constant λ of −670 × 10−6,137100 and is under OP compressive strain due to lattice
match with the LFO matrix which has a smaller lattice parameter. The magnetic soft
and hard phases in sample f likely correspond to (001)- and (110)-oriented grains, which
differ in magnetocrystalline anisotropy energy (higher by 5 × 104 J m−3 along <110> vs.
<001>138 ), in shape anisotropy, and in magnetoelastic energy (magnetostriction coefficient
67
λ = (λ + 3λ )/4 = −77.5 × 10−6 based on λ = 120 × 10−6).137110 100 111 111 A detailed analysis
of the anisotropy of nanocomposites only containing (110)-oriented CoFe2O4 can be found
in (138 ). The Ms of sample f is smaller than that of sample e, suggesting a greater fraction
of paramagnetic Co3O4 vs. magnetic CoFe2O4 or Fe3O4 as the shot ratio of the C3 target
increases.
Figure 4.5: In-plane (IP) and out-of-plane (OP) M-H hysteresis of the six nanocomposites.
Panels (a)-(f) correspond to the samples in Table 4.1, and M has been normalized over the
whole volume of the film. For M-H loops of (e) and (f), the magnetization did not saturate
within the maximum field of 1 T in vibrating sample magnetometer (VSM), so they were
measured using superconducting quantum interference device (SQUID) at higher fields, 4 T
and 3 T, respectively.
We particularly focused on sample c: C1F2(90)-LF14(90) for field cooling and exchange
bias experiments because it had the highest saturation magnetization (Ms) at 1 T, a relatively
68
low IP saturation field, and its spinel composition is more uniform than that of samples e,
f. The Co:Fe ratio x of CoxFe3−xO4 in this nanocomposite can be estimated assuming that
Lu/Fe = 1 in the LFO phase. The balance of cations is:
nLFOLu1.2Fe0.8O3 + nCFOCoFe2O4 → n1LuFeO3 + n2CoxFe3−xO4 (4.1)
where nLFO and nCFO on the left can be derived from the growth rates of the LF14 and
C1F2 targets (the LF14 target yields Lu/Fe = 1.5 films52 ). We neglect the oxygen balance
due to the supply of O2 during deposition. Using the molar volume of each phase to convert
n1 and n2 into volume fractions, we obtain x = 1.2 ± 0.1 and volume fractions of 43 ±
7 % of CFO and 57 ± 7 % of LFO in the nanocomposite, which is reasonably consistent
with the volume observations in cross-sectional STEM images. The Ms of Co1.2Fe1.8O4 is
estimated from (139 ) and based on the volume fraction we predict a magnetization of 150 ±
30 kA m−1 at room temperature for the nanocomposite. The measured IP Ms of 125 ± 1 kA
m−1 (fully saturated at 7 T, Figure 4.A4) is an acceptable match given the approximations
and the likelihood of non-stoichiometric transfer, and the presence of cation disorder, small
crystallite sizes, or oxygen vacancies. Even if the Lu/Fe ratio of LFO was in the range
from 0.8 to 1.2, the CFO composition x would vary by ± 0.2, and pillar and matrix volume
fractions, and predicted magnetization all are expected to stay within the aforementioned
ranges.
We performed field cooling of sample c (now described as a LuFeO3-Co1.2Fe1.8O4 nanocom-
posite) and compared the as-grown and field-cooled hysteresis loops (Figure 4.6). Field cool-
ing was done by elevating the temperature to 633 K (10 K above the LuFeO3 bulk Néel
temperature 623 K7 ), applying a field of +1 T, and then cooling down to room tempera-
ture in the field. Then the hysteresis loops were remeasured at room temperature and below.
Field cooling resulted in a loop shift, a coercivity increase, and a decrease in Ms. The change
in Ms is attributed to a reduction in the switchable volume of CFO due to pinning of the
69
spins nearest to the antiferromagnetic interface140 ,141 or to a change in oxygen content at
elevated temperatures. A coercivity increase in field-cooled loops has also been observed in
BiFeO3-Fe3O4 nanocomposites.112
Exchange bias (EB) was defined as the difference between the loop offsets (Hc) of the
as-grown (AG) and field-cooled (FC) loops, calculated from
EB = Hc,FC −Hc,AG = [(+Hc of FC) + (−Hc of FC)]/2− [(+Hc of AG) + (−Hc of AG)]/2.
(4.2)
The results are shown in Table 4.2. AG samples exhibited a uniform Hc of 1.7 mT on
average independent of the temperature. In contrast, Hc of FC samples is larger, negative,
and increases with decreasing temperature. The exchange bias field is −5.3 ± 0.2 mT at 300
K and −9.0 ± 0.7 mT at 200 K.
Figure 4.6: As-grown (AG) and field-cooled (FC) IP M-H loops of C1F2(90)-LF14(90) film
measured at (a) 200 K, (b) 250 K, and (c) 300 K. The sample used in this field cooling
experiment was grown during a separate run from the sample used in Figure 4.5c. XRD did
not show any difference, but the IP was slightly higher in this sample.
The origin of the exchange coupling can be explained by the interfacial relationship be-
tween ferrimagnetic CFO and antiferromagnetic LFO depicted in Figure 4.2c. The (111)p
facets of LFO are the equivalent of the (101)o or (011)o planes (subscript o stands for or-
thorhombic), assuming that LFO grows such that the film normal is along the orthorhombic
c direction as in a single-phase film.52 As represented in Figure 4.A5a, these planes in a
G-type antiferromagnetic perovskite consist of uncompensated spins (i.e., the spins all point
70
Table 4.2: Coercivity (Hc) extracted from M-H loops in Figure 4.5. +Hc and −Hc refers to
the coercivity of the ascending and descendng branches of the hysteresis loop, and Hc is the
loop offset.
AG FC
T (K) EB (mT)
+Hc (mT) −Hc (mT) Hc (mT) +Hc (mT) −Hc (mT) Hc (mT)
300 47.8 −44.2 1.8 73.7 −80.7 −3.5 −5.3
250 66.6 −63.0 1.8 128.0 −134.9 −3.5 −5.3
200 132.3 −129.4 1.4 227.1 −242.3 −7.6 −9.0
in the same direction, unlike (100)o, (010)o, or (001)o planes that are compensated). Fur-
thermore, spinel (111) planes solely consist of octahedral cation sites (Figure 4.A5b) which
also form an uncompensated surface. These uncompensated arrangement of spins at the
LFO (111)p/CFO (111) interface enables EB by pinning the spins in the CFO, resulting in
a net negative EB. However, we cannot exclude the possibility of other mechanisms such
as Dzyaloshinskii-Moriya interaction (DMI)-driven exchange coupling109 ,142 simultaneously
existing at compensated facets of LFO.
The process for establishing the exchange bias requires the Curie temperature of the CFO
to exceed the Néel temperature of the LFO. This is the case for bulk CFO and LFO, but the
higher Co content, possible cation disorder, and small crystal size of the CFO may reduce its
Curie temperature and hence its magnetization making the exchange biasing less effective.
The magnetic coupling might be increased further in a nanocomposite with perpendicular
magnetic anisotropy, so that all the interfaces contribute, or in a nanocomposite with vertical
(111) interfaces, or in a thicker nanocomposite with greater interface area.
71
4.4 Conclusion
This study elucidates the different structural and magnetic characteristics of vertically
aligned LFO-CFO nanocomposites which were grown using various target combinations by
PLD. Given a perovskite substrate, either Lu-rich or Fe-rich LFO targets can yield a per-
ovskite LFO film. When films are grown by codeposition from a Lu-rich or Fe-rich LFO
target with a Co-containing spinel target, nanocomposites form consisting of a LFO ma-
trix with spinel/rocksalt pillars. The selection of target material dictates the surface and
interface morphology as well as the crystal structure and chemical properties, particularly
yielding pillars with a range of Co:Fe ratios when Co3O4 and Fe-rich LFO targets are used,
which corresponds to the Co and Fe being delivered by different laser pulses. The magnetic
anisotropy of the nanocomposites is associated with the pillar shape and the strain imposed
by the LFO matrix and the substrate. Owing to the magnetic coupling at the interface, a
nanocomposite of a thickness of 31 nm consisting of antiferromagnetic LFO and ferrimagnetic
Co-rich CFO shows an exchange bias of −5.3 mT at room temperature and a saturation mag-
netization consistent with that of the bulk spinel composition. The negative exchange bias
is attributed to the crystallographic relationship between the perovskite and spinel interface,
where both (111) spinel and (111)p LFO facets expose uncompensated spins. These results
highlight the importance of selecting the correct PLD targets to obtain nanocomposites with
desired phases, and how interfaces of the oxide phases can contribute to magnetic coupling,
particularly in a vertically aligned heterostructure with a G-type canted antiferromagnet.
The demonstration of magnetic interactions at the interfaces enhances the opportunities for
coupling the properties of the two phases in a multiferroic nanocomposite that contains an
orthoferrite.
72
4.5 Methods
Films were grown by PLD. The heater temperature setpoint was 900 ◦C leading to an
actual substrate temperature ∼650 ◦C, and the pressure during growth was 10 mTorr of
oxygen. The temperature difference results from the gap between the heating coil and
the substrate holder, limiting heat conduction, and was calibrated by thermocouple on the
substrate holder. The laser fluence and repetition rate were 2 J cm−2 (spot size 0.12 cm2) and
10 Hz, respectively. These parameters were kept consistent among different codepositions
with the conditions elaborated in Table 4.1. The growth rates of LF14, LF06, C1F2, and
C3 targets were calibrated beforehand to ensure that the growth in each cycle is less than
one unit cell or less of each material to promote intermixing of the arriving species. The
rates of LF14, LF06, and C1F2 were 4.3 nm, 1.2 nm, and 4.6 nm per 1000 shots, and C3
was assumed to be the same as C1F2. The time interval between target changes was ∼3
s and the targets were rotated during ablation to even out the erosion. The thickness of
the nanocomposites agrees with that predicted from the individual target deposition rates
within ∼24 % for the films grown with C3 and ∼27 % for those grown using C1F2. STO
substrates were purchased from MTI Corp. and used as received. The distance between
target and substrate was 70 mm. High-resolution XRD was done using a Rigaku Smartlab
high-resolution diffractometer and Ge-(220) double-bounce monochromator with a Cu Kα1
X-ray source (wavelength = 1.5406 Å).
SEM was performed using a Zeiss Merlin high-resolution scanning electron microscope.
Cross-sectional lamellae for STEM imaging were prepared using a Raith VELION FIB-
SEM instrument with Au+ ion beam. HAADF STEM imaging along with EDS and EELS
were done using an aberration-corrected Thermo Fisher Scientific Themis G3 equipped with
Super-X EDS detectors and a Gatan continuum EEL spectrometer. Imaging and spec-
troscopy were performed at 200 kV, with probe convergence semi-angle 18.9 mrad and beam
current ranging from 40 to 120 pA. Dwell times for HAADF and EDS imaging were 0.5 to
73
2 µs. EELS data were processed in Digital Micrograph. Spectral background was removed
by power law fitting and the effects of plural scattering has been minimized by Fourier-ratio
deconvolution. No further denoising was done. M-H hysteresis measurements at room tem-
perature and field cooling treatments were done using a Digital Measurement System 7035B
VSM, and M-H loops before and after field cooling were measured using a Quantum Design
MPMS3 SQUID magnetometer.
4.6 Appendices
Figure 4.A1: Full range (20-70◦) XRD scans of the nanocomposite films.
74
Figure 4.A2: SEM images showing the difference in surface morphology between the two
nanocomposite films, C1F2(90)-LF14(90) (left) and C3(90)-LF06(90) (right).
75
Figure 4.A3: Additional STEM analysis of C3(90)-LF06(90) nanocomposite. (a) HAADF
image of the right side of pillar B from Figure 4.3b, which is oriented along the <112>
direction. Fourier transforms of the selected squares are shown in the insets, and the derived
lattice spacing of rocksalt and spinel are 0.26 nm and 0.48 nm, respectively. (b) (112) facets
of rocksalt and spinel structures, showing how spinel has a wider planar spacing at the
same projection as rocksalt. Red and blue correspond to octahedral and tetrahedral sites
respectively. (c) EDS maps of a different region, showing three different transition metal
compositions of the pillar phase: mainly Co (C, D), mainly Fe (E), and both Co and Fe (F).
(d) EELS comparison of spinel Co3O4 and perovskite LFO, showing prominent differences
in the oxygen electronic structure.
76
Figure 4.A4: FC M-H loops of C1F2(90)-LF14(90) at maximum field of ±7 T, measured at
(a) 200 K, (b) 250 K, and (c) 300 K.
Figure 4.A5: (a) Description of the antiferromagnetic spin structure of LFO in an orthorhom-
bic unit cell. For simplicity, oxygen atoms are removed, and only the dominant Gx ordering
of Bertaut notation GxAyFz is represented in the spin structure. The spins lying on (011)o
and (101)o planes, which are equivalent to (111)p planes, are pointing in the same direction.
(b) Schematic of the (111) family planes in a spinel unit cell. Magenta and blue correspond
to octahedral (Oct) and tetrahedral (Tet) sites, respectively. (111) planes consist of octahe-
dral sites and planes of tetrahedral sites are also parallel with a small offset. The strongest
superexchange interaction in CFO spinel is antiferromagnetic between octahedral and tetra-
hedral sites.
77
Chapter 5
Self-assembled multiphase strontium
cobalt ferrite thin films with
voltage-controlled magnetism
This chapter is based on a publication which the author wrote and published in ACS Applied
Nano Materials (2022).119
5.1 Introduction
Among the vast range of perovskite oxides, SrCo1−xFexO3−δ (SCFO) stands out as a
mixed conductor useful in electrochemical applications such as solid oxide fuel cells, oxy-
gen sensors or oxygen separation membranes.143 ,144 SCFO represents a solid solution of
SrCoO3−δ (SCO) and SrFeO3−δ (SFO). The end members SCO and SFO show topotac-
tic phase transitions from perovskite (P) to brownmillerite (BM, δ = 0.5) which can be
induced by annealing,145 an electrical stimulus,138 ,146 or electrochemically via liquid elec-
trolyte gating147 ,148 where an electric field is used to drive migration of O2− or H+ ions into
a film.23 ,28 ,149 Perovskite SCO (P-SCO) is a ferromagnet with Curie temperature TC = 222
K150 and P-SFO is an antiferromagnet with a Néel temperature of TN = 130 K,151 whereas
78
BM-SCO and BM-SFO are antiferromagnets with TN = 570 K and 715 K, respectively.152 ,153
SCFO with Co:Fe ratios near 1:1 undergoes a reversible transition from ferromagnetic P to
paramagnetic BM phases at room temperature by ionic liquid gating.147
The functionality of perovskite films can be extended by incorporation into self-assembled
nanocomposites consisting of pillars or crystals of one phase, typically with a length scale
of ∼10–100 nm, embedded in a matrix of a second phase. Nanocomposites exhibit the
properties of both phases and often reveal interfacially-mediated cross-coupling between
those properties, for example magnetoelectric coupling in a nanocomposite consisting of a
ferroelectric phase and a magnetic phase.103 ,104 ,117 ,154–157 Nanocomposites have been syn-
thesized from many combinations of materials, most commonly perovskite (e.g. BiFeO3,
BaTiO3, PbTiO3, SrTiO3, La0.8Sr0.2CoO3) and spinel (e.g. CoFe2O4, MgFe2O4, NiFe2O4,
Mn Zn Fe O ),103 ,130 ,158–1610.5 0.5 2 4 but have also included ZnO,162 Sm 1632O3, Cu oxides,164 and
metal nanowires or nanoparticles.118 ,165 In our prior work, pulsed laser codeposition from
SCO and CoFe2O4 targets was used to form a nanocomposite consisting of a SCFO matrix
and Co spinel (Co3O4) pillars of diameter ∼50 nm, where the coherent interfaces enable
strain transfer between the two phases.147
We show here that SCFO can be grown either as a single-phase film of perovskite or
brownmillerite, or as a two-phase nanocomposite with an additional cobalt oxide (CoOx)
phase present as <30 nm diameter pillars through control of oxygen pressure PO2 during
pulsed laser deposition (PLD), with concomitant effects on its structure and magnetic proper-
ties. Single-phase films of SCFO form at intermediate PO2 whereas two-phase nanocomposite
films form at low or high PO2. The formation of the nanocomposites is unexpected because
the overall Sr:(Co+Fe) ratio of the film is 1:1, and the presence of CoOx therefore implies
that the SCFO phase is enriched in Sr. Finally, we describe how the structure and magnetic
properties of the resulting films are modified by ionic liquid gating. The nanocomposite films
may be useful in spintronic applications requiring electrically-controlled magnetization, and
the spontaneous formation of more than one oxide phase enables the introduction of multiple
79
functionalities to the film.
5.2 Growth mode dependence on pulsed laser deposition
process parameters
In this study PO2 and film composition x are the primary variables affecting the film
structure and phase formation. PO2 plays an important and complex role during PLD,
determining the film lattice parameter, strain state, cation ratio and oxygen vacancy con-
centrations21 ,39 ,166 and differential scattering of the species in the plume.167 PO2 also affects
the surface roughness of the film;39 influences the kinetic energy and oxidation state of arriv-
ing ions168 and alters the surface of the substrate;29 furthermore, annealing in oxygen after
growth can reduce28 or oxidize169 ,170 the film. In this work, films with different Co:Fe ratio,
i.e. x, are produced by combinatorial growth, and two different growth modes (single- or
two-phase films) can be obtained by controlling PO2, as depicted in Figure 5.1.
Figure 5.1: Visual description of PLD combinatorial growth process, illustrating how PO2
during growth affects the structure of thin films.
To study the effect of PO2 on phase formation, we first vary PO2 from high vacuum (HV)
80
to 150 mTorr during the growth of SCFO with x ≈ 0.5, i.e. SrFe0.5Co0.5O3−δ (Figure 5.2a).
Films were deposited on SrTiO3 (STO) substrates held at a temperature of ∼700 ◦C, using
5 Hz repetition rate and laser fluence 1.3 J cm−2. For a total number of 12k laser shots,
half from each target, the thickness of the film grown at 150 mTorr was 43.4 ± 1.1 nm
based on X-ray diffraction (XRD) fitting (Figure 5.A1). At PO2 = 20 mTorr a single-phase
BM film is obtained, but at the extremes in pressure, i.e. HV and 150 mTorr, XRD peaks
are observed from both the BM phase and a weaker spinel phase at 2θ = 42◦.147 At other
pressures (10 mTorr, 50 mTorr, and 100 mTorr), the BM peak is split but no secondary
phases are seen. The BM peak position moves to a higher angle when PO2 increases, clearly
evident by comparison of the 150 mTorr and the HV samples, which is consistent with fewer
oxygen vacancies, a higher cation oxidation state and a smaller unit cell volume at higher
P 171O2.
These results indicate a process window for formation of a single-phase SCFO film with
x ≈ 0.5 at an optimum PO2 of 20 mTorr. We also prepared films at a lower temperature
substrate temperature (∼500 ◦C) and a higher fluence (∼2 J cm−2), and found that SCFO
grown at 20 mTorr and then annealed at PO2 = 0.1 mTorr or without O2 also showed a two-
phase film. Further description of these films can be found in Figure 5.A22. This shows that
the transition between forming single-phase and two-phase films depends on a combination
of process parameters.
A set of 150 mTorr samples was then grown to investigate the structure and composition
of the two-phase film as a function of Co:Fe ratio in SrCo1−xFexO3−δ (Figure 5.2b). These
films (grown at ∼700 ◦C, 5 Hz, 1.3 J cm−2) showed the presence of BM plus spinel phases
for x = 0 to 0.59, perovskite plus a very weak spinel peak at x = 0.74, and only P phase for
x = 1. Therefore, films with higher Co content favor two-phase microstructures.
As the Fe content increases, the lattice parameter of the BM phase becomes smaller.
This observation is consistent with a larger population of higher valence, smaller radius Fe
cations and fewer oxygen vacancies. (If Fe and Co had the same valence state, films richer
81
Figure 5.2: XRD results of (a) SCFO with x ≈ 0.5 grown with PO2 = HV, 10, 20, 50, 100,
150 mTorr, (b) SCFO at 150 mTorr with different x, (c) summary of phases present in films
as a function of composition, oxygen pressure and ionic liquid gating (circled samples were
analyzed by STEM), (d) XRD after −2 V gating of nanocomposite films grown at 150 mTorr,
and (e) XRD of SCFO with x = 0.30: as-grown, −2 V gated, and etched (x refers to the
nominal composition).
in Fe would have a larger lattice parameter because the ionic radius of Fe is larger than that
of Co.) These trends contrast with observations in single-phase BM-SCFO films grown at
0.1 mTorr, where the out-of-plane lattice parameter increased with increasing Fe content.147
The spinel peak also shifts slightly with varying composition, consistent with some level of
interfacial coherency between the two phases.
To determine the morphology of the two-phase films, we etch with 10 % HCl solution to
remove the matrix phase. For the samples of x = 0.30 and x = 0.59, etching leaves free-
standing rectangular or square based pillars growing vertically through the film thickness
with average edge length ∼17 nm (range of 8-30 nm), as shown in Figure 5.3a and Figure
82
Figure 5.3: (a) Scanning electron microscope (SEM) image of the etched x = 0.30 film
with 30◦ tilt angle showing pillars; (b,c) HAADF STEM images of (b) as-grown x = 0.43
film with pillars (marked by arrows) surrounded by the matrix, and (c) as-grown x = 0.74
film showing the nucleation of Co3O4 between the substrate/film interface and film grain
boundary, marked by the arrow.
5.A3a. Smaller crystals with size below 10 nm were also present on the substrate, indicating
that a higher density of crystals nucleated at the start of film growth, but not all crystals
continued to grow as high as the film thickness. By analogy with other spinel-perovskite
composites grown on STO, the crystals are identified as spinel and the matrix as perovskite,
i.e. SCFO. The as-grown x = 0.43 sample was imaged in cross-section using high angle
annular dark field (HAADF) scanning transmission electron microscopy (STEM) (Figure
5.3b) along the [110] axis, and showed typical (111) facets of the spinel phase. The x = 0.74
sample (Figure 5.3c) in cross-section showed nucleation of small pillars at the substrate and
film interface, consistent with the barely detectable spinel peak shown in Figure 5.2b. Its
perovskite matrix exhibits a layered structure which is attributed to oxygen vacancy ordering
and is described in detail elsewhere.172
These results indicate that the SCFO samples grown at 150 mTorr are vertically aligned
nanocomposites made up of a BM (for high Co content) and/or P (for high Fe content)
matrix and an increasing amount of spinel pillars for higher Co content. As a comparison,
we also grew films at 100 mTorr and these only exhibited BM without detectable amounts
of spinel phases (Figure 5.A4). The phases present in the films are summarized in Figure
5.2c as a function of Co:Fe ratio and PO2.
83
Figure 5.4: (a) STEM EDS elemental maps of an x = 0.74 film after −2 V gating, (b)
HAADF STEM image showing a pillar at center and the layered oxygen-deficient P matrix,
and (c) magnified image of the outlined region in (b) with atomic distance of Co3O4 and
CoO indicated.
84
Compositional mapping of the x = 0.74, 150 mTorr sample in Figure 5.4a indicates
that the pillars contain little or no Fe, and therefore consist of cobalt oxide, CoOx. The
preferential incorporation of Co in the spinel and Fe remaining in the perovskite have been
seen in other studies, such as the formation of CoO and Co3O4 during the deposition of
a composite of La0.6Sr0.4CoO3 and CoFe 1732O4, and SCFO and Co3O4 phases produced
during codeposition from SCO and Fe3O4 targets.147 Moreover, comparing Figure 5.A3b
and Figure 5.A3c, the number and area density of pillars exposed on the surface decrease
as the Fe content increases. We conclude that the key factor promoting the formation of a
nanocomposite from SCO+SFO codeposition at high PO2 is the presence of Co.
X-ray photoelectron spectroscopy (XPS) survey scan results for the nanocomposite films
grown at PO2 = 150 mTorr show that the Sr/(Co+Fe) ratio is close to 1 (Figure 5.A5).
Moreover, quantitative analysis of the STEM energy dispersive X-ray spectroscopy (EDS)
scan (Figure 5.4a) gives Sr of 48 ± 9 %, Co of 12.54 ± 2.4 %, and Fe of 38.87 ± 5.7 %,
consistent with Sr/(Co+Fe) = 1 within the error bars. We note that wavelength dispersive
spectroscopy (WDS) was not appropriate for Sr/(Co+Fe) determination because of the Sr
contribution from STO substrate. This implies that a stoichiometry change due to differen-
tial scattering between the species is not likely to be responsible for the emergence of the
spinel phase. A pseudo-ternary phase diagram of SrO-CoO-FeO at 900 ◦C in air174 shows
that SCFO exhibits complete solid solubility across the range of Co:Fe ratio, but a defi-
ciency in Sr promotes the formation of secondary phases Co3−yFeyO4 and Sr4CozFe6−zO13.
Furthermore, the phase diagram in (174 ) implies that SCFO cannot coexist with CoFe2O4
thermodynamically at 900 ◦C in air, whereas SCFO and Co3−yFeyO4 can readily coexist up
to 900 ◦C depending on the cobalt content. The high Co content of the spinel phase of
the nanocomposites is consistent with these phase equilibria. Therefore, we attribute the
formation of the two-phase nanocomposite in our study to the thermodynamic stability of
the spinel Co3O4 phase. Nucleation of the stable CoOx could be facilitated by columnar
growth of the film, which is favored at high P 166O2.
85
The CoOx formation implies an excess of Sr in the remaining SCFO matrix phase. Con-
sidering the approximate volume fraction of the pillars (1 %) in the x = 0.30 sample, the
SCFO matrix has a Sr:(Co+Fe) ratio of roughly 1:0.95. This may be accommodated by SrFe
or SrCo antisite defects (associated with additional oxygen vacancies (VO) for charge bal-
ance) analogous to the behavior of Ti-deficient STO epitaxial films,175 or as B-site vacancies
(here V 176Fe or VCo) as in La-rich LaAlO3 films. Larger Sr excess may lead to formation of
Ruddlesden-Popper layered phases (Srn+1(Co,Fe)nO3n+1, with n an integer) which has been
reported for highly Ti-deficient STO,177 but the Sr content here is presumably not large
enough to produce these phases.
5.3 Manipulating magnetic properties via electrolyte gat-
ing
We performed −2 V ionic liquid gating on the five nanocomposite samples grown at 150
mTorr to oxidize them and to promote ferromagnetic exchange interaction between Co/Fe.
According to the XRD in Figure 5.2d, gating converts the BM to P while preserving the
spinel peak, and does not produce a spinel peak in other samples that did not originally have
one. The film shows a metallic finish as the insulating BM converts to conductive P. Gating
changes the out-of-plane lattice parameter of the matrix by −3 %, and that of the pillars by
around −1 % due to strain transferred from the matrix (Figure 5.2e). In (147 ) single-phase
BM-SrCo1−yFeyO2.5 was oxidized into P-SrCo1−yFeyO3, and different initial compositions,
y, yielded different final lattice parameters after gating. In contrast, here we observe that
the lattice parameter of the gated SCFO matrix is the same for x = 0.16 to x = 0.74. This
may reflect the influence of vertical epitaxy on constraining the lattice parameter of the
matrix, because reciprocal space mapping suggests that the matrix is strained in-plane on
the substrate.172 The bias voltage did not play a significant role in gating as long as it is
beyond the threshold for the water molecules to decompose and provide the oxygen that
86
Figure 5.5: In-plane M-H loops of (a) x = 0.30 sample measured at 173 K as-grown, −2
V gated, and etched. (b) Two x ≈ 0.5 samples measured at room temperature after −2
V gating. These samples were grown at 20 mTorr (green: single-phase, TC above room
temperature) and 150 mTorr (yellow: nanocomposite, matrix has higher Fe content and
lower TC).
enables the phase change. Applying +2 V after −2 V gating to the nanocomposite turns P
back to BM (Figure 5.A6), but 30 min was not sufficient to fully transform the matrix unlike
the single-phase shown in (147 ).
HAADF STEM measurements of the gated x = 0.74 sample (Figure 5.A7) show that the
top region of the matrix is mostly P. The matrix closer to the substrate retains its as-grown
structure which is layered P, and BM exists near the spinel pillars where it is assumed to
accommodate the structural distortion caused by the vertices of the (111) facet of the pillars,
since BM has a larger lattice parameter. The BM does not produce a measurable subpeak
in XRD either before or after gating; it may have been present in the film even before gating
but at such a low fraction that it does not yield peaks in XRD. In addition, the ionic liquid
gating induced defects in the matrix phase evident as dark lines parallel to the substrate
separating and distorting the lattice planes. This is also evident in the deterioration of the
rocking curve (Figure 5.A8). Furthermore, we find that the Co-rich pillar includes regions
of rocksalt CoO as well as spinel (Figure 5.4b and Figure 5.4c), but these could not be
distinguished in XRD. The direction of gating would oxidize CoO into Co3O4, thus the CoO
87
was likely present as grown. Therefore, for the x = 0.74 sample grown at 150 mTorr, while
the XRD indicates spinel and perovskite are present before and after gating, STEM indicates
a more complex structure in which the matrix consists of perovskite, layered perovskite, and
a small amount of brownmillerite with pillars consisting of Co3O4 and CoO.
The Co-rich composition of the spinel phase is confirmed by magnetization measure-
ments. The as-grown nanocomposite has no room temperature magnetic moment (Figure
5.A9), which excludes the composition of the spinel phase being CoFe2O4 or Fe3O4, both of
which are magnetic at room temperature. Comparing with magnetic data from (Co,Fe)3O4
nanoparticles,139 the upper bound for Fe content in our pillars is Co1.6Fe1.4O4, and it is likely
much lower (close to zero Fe) according to the compositional analysis.
Figure 5.5a shows the in-plane hysteresis loops measured at 173 K for as-grown, −2
V gated, and etched nanocomposites with x = 0.30 grown at PO2 = 150 mTorr. The
magnetic moment appears on gating but disappears on etching the matrix indicating that
the magnetism originates from the gated SCFO phase and not the pillars.147 The saturation
magnetization (Ms) of the gated nanocomposite was 200 kA m−1 at 173 K but 20 kA m−1 at
room temperature suggesting that its TC is just above room temperature. Similarly, the x =
0.59 nanocomposite has Ms of 300 kA m−1 at 173 K after gating but negligible moment at
room temperature (Figure 5.A10). −2 V gating does not produce significant magnetism at
room temperature, which implies that the matrix Co:Fe ratio has deviated from 1:1 because
of the precipitation of CoO .147x In comparison, single-phase film (PO2 = 20 mTorr) after
gating show a maximum M −1s of 100 kA m at room temperature and TC = 340 K for x =
0.5 (Figure 5.5b), and both Ms and TC fall as compositions deviate from x = 0.5; at 173 K
a single-phase gated sample with x = 0.3 gave Ms ≈ 150 kA m−1.147 The higher Ms of the
gated nanocomposite at 173 K suggests that its matrix has x > 0.3, which is consistent with
the Co being present in the pillar phases and Fe being enriched in the matrix compared to the
average composition. These results show that spinel-perovskite nanocomposites consisting
of a magnetic matrix and paramagnetic pillars are created by gating SCFO grown at high
88
PO2.
5.4 Conclusion
This work shows that codeposition from SCO and SFO targets can produce either a
single-phase SCFO film or a two-phase nanocomposite film, depending on the oxygen pres-
sure during growth. The formation of the nanocomposite is unanticipated because the film
cation ratio is stoichiometric for perovskite and brownmillerite, i.e. Sr/(Co+Fe) = 1, so it
is not driven by the presence of excess Co and Fe as in the case of nanocomposites formed
by codeposition from spinel and perovskite targets. Nanocomposite formation in our case
appears to be favored by the thermodynamic stability of cobalt oxide phases, and the volume
fraction of pillars increases with Co content. The excess Sr is likely incorporated into the ma-
trix via antisite defects and additional oxygen vacancies. The SCFO forms a brownmillerite
matrix phase at high Co content and perovskite phases at high Fe contents.
STEM reveals that the nanocomposites contain several phases, in which oxygen-deficient
P and BM coexist in the matrix, and spinel and rocksalt Co oxides in the pillars. The Co:Fe
composition of the matrix defines its response to ionic liquid gating and its magnetic prop-
erties. Gating converts BM into P for all compositions, but the P formed by gating contains
defects where the layered structure is disrupted, and some BM remains near the pillars. The
magnetic moment of the nanocomposites originates from the matrix, and increases on gating
as BM is converted to P; however, the magnetic moment differs from that found by gating of
single-phase BM-SCFO to P-SCFO because the matrix in the nanocomposite film contains
more Fe than the nominal composition. Strain coupling between the matrix and pillar is
evident by the shift in XRD peak position upon gating.
This work shows that self-assembled oxide nanocomposites, typically formed by the code-
position of two different phases, can instead be formed in a film with a stoichiometric per-
ovskite cation ratio by selecting the processing parameters. The oxygen pressure during
89
growth, PO2, has a profound influence on the oxide film morphology and the phase stability
of nanocomposite thin films. These nanocomposites may be favorable for applications that
require multifunctionality in a single film, or electrical control of magnetism, e.g. magneto-
electric memory.
5.5 Methods
SCFO films and nanocomposites were grown on (001) oriented STO substrates by PLD
using a KrF excimer laser (λ = 248 nm) at 5 Hz repetition rate with a fluence 1.3 J cm−2.
SCO and SFO targets were ablated at up to 60 shots each cycle, ensuring that the thickness
of each deposited sublayer is less than that of a single unit cell. Ablation from two targets
was used for convenient control of the Co:Fe ratio. The heater setpoint was 850 ◦C with the
actual substrate temperature ∼150 ◦C lower at ∼700 ◦C, and the PO2 ranged from HV (PO2
< 5×10−6 Torr) to 150 mTorr. Some samples were grown at 750 ◦C setpoint temperature
(substrate temperature ∼500 ◦C) and 2 J cm−2 in a different chamber. The films were cooled
down to room temperature in the same PO2 at 20 ◦C min−1 cooling rate.
The crystal structure was characterized by high-resolution XRD using a Rigaku Smartlab
high-resolution diffractometer with Ge-(220) double-bounce monochromator and a Cu Kα1
(λ = 1.5406 Å) X-ray source. The morphology of the film was imaged with a Zeiss Merlin
high-resolution SEM. Nanocomposite films were etched in 10 % HCl solution for 5 s to
remove the matrix and leave the pillars. The chemical composition was obtained by WDS
using a JEOL-JXA-8200 Superprobe and Thermo Scientific K-Alpha + XPS system with Al
Kα (1486.6 eV) as a source. The samples were coated with carbon for conductivity when
performing WDS analysis. Before the XPS measurement, the sample surface was cleaned
with a cluster ion beam for 30 s. Lamella samples of the films along the [110] axis (indexed
with respect to the axes of the cubic unit cell of P and the pseudocubic unit cell of BM)
before and after ionic liquid gating were prepared for electron microscopy either using a Ga+
90
ion beam in a FEI Helios Nanolab 600/660 Dual Beam system or by mechanical polishing.
The samples were further thinned using Ar+ ion milling. STEM imaging and spectroscopy
analysis were performed on a probe corrected Thermo Fisher Scientific Titan G3 60-300 kV
microscope operated at 200 kV at a beam current of about 60 pA and convergence semi-angle
of 19.2 mrad. HAADF images were collected with a collection semi-angle range of 65-200
mrad.
Ionic liquid gating was carried out to modify the structure and magnetic properties.147
Experiments were done with 1-hexyl-3-methylimidazolium bis-(trifluoromethylsulfonyl)imide
(HMIM-TFSI) as the electrolyte (Sigma-Aldrich, USA). The films were immersed in the ionic
liquid and a Pt probe was placed in direct contact with the sample surface, and a spiral Pt
wire served as a counter electrode. A Hewlett-Packard 6632A power supply was used to apply
the voltage during gating. The samples were gated at a negative bias of −2 V for 30 min to
insert oxygen. The magnetic hysteresis loops were measured using a Digital Measurement
System 7035B vibrating sample magnetometer (VSM) at room temperature and 173 K. Low
temperature was obtained by flowing cold nitrogen gas past the sample from a liquid nitrogen
source.
91
5.6 Appendices
Figure 5.A1: Thickness fitting using XRD Laue fringes of SCFO film with x = 0.43 (PO2 =
150 mTorr). The simulation yields a thickness of 43.4 ± 1.1 nm.
Figure 5.A2: (a) XRD of SCFO films grown in a different chamber with higher laser fluence
(400 mJ and 300 mJ corresponds to ∼2 J cm−2 and ∼1.9 J cm−2, respectively). SEM images
of nanocomposite films grown at (b) 400 mJ and (c) 300 mJ, and both scale bars correspond
to 200 nm. PO2 window for nanocomposite formation has shifted to a lower value (∼20
mTorr) and a slight reduction in fluence results in different shapes such as rectangles or
triangles indicating a different crystal orientation.
92
Figure 5.A3: SEM images of SCFO with (a) x = 0.59 after etching, (b) x = 0.16 as-grown,
and (c) x = 0.43 as-grown films. In (b,c) the volume of pillars decrease as the film becomes
Fe-rich (higher x).
Figure 5.A4: XRD of SCFO films grown at 100 mTorr, for three compositions x.
93
Figure 5.A5: XPS survey scan results comparing spectra from films grown at (a) PO2 = 150
mTorr and (b) PO2 = 10 mTorr. In both conditions, the Sr/(Co+Fe) ratio is close to 1,
determined by the XPS semiquantitative analysis using built-in relative sensitivity factors.
Figure 5.A6: XRD of SCFO nanocomposite of as-grown, −2 V gated for 30 min, and then
+2 V gated for 30 min state with composition (a) x = 0.16 and (b) x = 0.43.
94
95
Figure 5.A7: HAADF STEM image of SCFO nanocomposite with x = 0.74 after −2 V ionic
liquid gating, showing (a) the matrix, including two regions at higher magnification with
layered perovskite structure; (b) a lower magnification view around the pillar in which the
layered matrix is clearly visible; and (c) a direct comparison between as-grown and gated
matrix.
96
Figure 5.A8: Rocking curve intensity difference between as-grown and −2 V gated film with
x = 0.74. The peak intensity decrease after gating is attributed to interlayer defects.
Figure 5.A9: In-plane M-H loops of as-grown SCFO films at room temperature, indicating
lack of magnetic hysteresis or remanent magnetization.
97
Figure 5.A10: In-plane M-H loops after −2 V gating of (a) x = 0.16 to x = 0.74 nanocom-
posite films measured at room temperature, and (b) x = 0.59 film measured at 173 K.
Figure 5.A11: XRD of as-grown, −2 V gated, and +2 V gated Co3O4/STO. The film peak
is not clearly defined because spinel does not wet well on perovskite substrates.
98
Chapter 6
Perovskite-derived layered crystal
structure of strontium cobalt ferrite
This chapter is based on a publication which the author wrote and published in Advanced
Materials Interfaces (2024).172
6.1 Introduction
In oxide perovskites (ABO3) with transition metal B-site cations, the high degree of com-
positional and structural tunability enables engineering of a vast range of desirable properties,
including ferroelectricity,178 magnetism,21 ,147 ,179 conductivity,143 and electrochemistry.180
The oxygen coordination of the cation is a particularly important factor that dictates the
properties of perovskites: electronic, magnetic, and optical properties of perovskites are all
affected by oxygen stoichiometry.181–186 Moreover, oxygen vacancies can form an ordered ar-
rangement which affects the ionic coordination and the structure, symmetry, and properties
of the material.181 ,187 Perovskite-derived ordered structures may exhibit interesting physical
properties such as magnetoelectric coupling188 and superconductivity.189 ,190
Ordering of oxygen ions and oxygen vacancies (VO) in ABO3−δ depends on the cation
composition as well as the oxygen deficiency, δ.182 ,191–196 In SrBO3−δ with B = Co, Fe, Mn
99
and 0 ≤ δ ≤ 0.5, a number of oxygen-ordered structures have been reported. SrCoO3, SrFeO3,
and SrMnO3 with δ ∼ 0 form a Pm3̄m cubic perovskite (P) at room temperature.145 ,151 ,197 In
contrast, Sr(Co,Fe)O2.5 and Sr(Fe0.95Mn0.05)O2.5 form brownmillerite (BM), in which the VO
order in atom columns leading to alternating BO and BO polyhedra.145 ,153 ,198 ,1996 4 SrMnO2.5
forms a distinct V -ordered structure200 ,201O while SrFe0.5Mn0.5O2.5 and SrFe0.5Mn0.5O2.75
exhibit no V ordering.202O SrFeO3−δ (δ = 0.125, 0.25)203–205 and SrMnO3−δ (δ = 0.286,
0.4)200 exhibit a characteristic VO-ordered structure. SrCoO2.75 thin films are expected
to prefer random VO locations from density functional theory (DFT) modeling,206 how-
ever, a unique ordering can be observed when δ = 0.25 and 0.18 in SrCoO .2073−δ A 314-
type ordering named after the cation stoichiometry in Sr3Y1(Co,Fe)4O10.5 has been ob-
served in SrFe 2080.25Co0.75O2.63, and VO ordering in layers has been observed in Nb-doped
Sr(Co,Fe)O3−δ, where Nb5+ is believed to prevent BM formation,209 both of them in the
bulk form.
Here, we report a new oxygen-ordered perovskite-derived structure in epitaxial SrCo1−xFex
O3−δ (SCFO) thin films grown by pulsed laser deposition (PLD) that differs from the struc-
tures described above. The alternating oxygen deficiency in B-site layers leads to alternate
expansion and contraction of the lattice planes along the out-of-plane direction, observed
using scanning transmission electron microscopy (STEM) imaging. We use DFT to obtain
relaxed perovskite structures with different oxygen arrangements, comparing the results with
STEM image simulations of possible structures. The oxygen layering observed in our SCFO
thin films is accompanied by a change in B-site coordination from BO6 to BO4/BO5 as
proposed in two possible crystal structures whose oxygen deficiency δ is 0.5 or 0.625.
100
6.2 Structural properties of layered perovskite by scan-
ning transmission electron microscopy and X-ray diffrac-
tion
Figure 6.1: (a) STEM image overviewing the x = 0.74 SCFO (film), STO (substrate), and
their interface. The alternating layers in the SCFO are visible throughout the film thickness.
(b) Enlarged region of (a) at the interface showing the epitaxial growth. (c) Simultaneously
collected ADF and iDPC STEM images revealing oxygen ordering in alternate B-site atom
column layers of the x = 0.74 film. (d) Cropped region of (c), with the arrow in the iDPC
pointing out the absence of oxygen in corresponding layers of the ADF image. (e) Sr—Sr
distance map and (f) B-site (Co and Fe site) intensity deviation map obtained from the ADF
image in (c), clearly highlighting the different but consistent characteristics of the dim and
bright layers.
SCFO forms a solid solution with a P or BM structure, and can be reversibly transitioned
between these structures by controlling the oxygen stoichiometry.147 We previously investi-
gated the growth of SCFO on SrTiO3 (STO) substrates by PLD as a function of composition
x and oxygen pressure (P ).119O2 For films grown at PO2 = 150 mTorr, we observed X-ray
101
peaks characteristic of P when x ≥ 0.74 (i.e. Fe-rich) or BM when x < 0.74. However,
electron microscopy revealed a “layered P” structure plus a small amount of BM at the grain
boundaries when x = 0.74, and a majority of BM with a small coexistence of “layered P” and
P when x = 0.43. Furthermore, the films included a small (few percent) fraction of CoOy
pillars embedded in the SCFO matrix. The Sr:(Co+Fe) ratio in the film remained close to
1.0 despite the segregation of Co into the CoOy pillars, and the volume fraction of pillars
diminished with increasing Fe content, suggesting that the formation of the “layered P” does
not necessarily correlate with the formation of the CoOy pillar phase. In contrast, films
grown at low PO2 (20 mTorr) did not exhibit a CoOy phase. Here, we focus on analyzing
the “layered P” crystal structure in films synthesized at 150 mTorr.119
Figure 6.2: (a) XRD results of SCFO films grown at 150 mTorr with x = 0.74 and 0.43. (b)
RSM of (103) reflection of the as-grown x = 0.74 film. Vertically aligned substrate and film
peaks indicate that the film is fully strained.
The “layered P” of an x = 0.74, PO2 = 150 mTorr film viewed along the [110] axis of
the pseudocubic unit cell is shown in Figure 6.1. Simultaneous annular dark field (ADF)
and integrated differential phase contrast (iDPC) STEM images (Figure 6.1c) reveal layers
parallel to the substrate. The ADF STEM image from the film shows an intensity variation
of the transition metal (B-site) but does not provide direct imaging of the oxygen atom
columns. Instead, the corresponding iDPC image shows that the layers of dim and bright
102
B-site atom columns correspond to oxygen-deficient and oxygen-rich layers, as shown in
Figure 6.1c,d, implying that the oxygen stoichiometry change is predominantly present in
the layers containing B-site and O, rather than in the layers of Sr and O. The alternating
intensities of the B-site atom columns are also accompanied by expanded and contracted
Sr—Sr distances of 455 ± 0.3 pm and 319 ± 0.3 pm along the out-of-plane direction, giving
a distance ratio of 1.42 (Figure 6.1e). The periodic intensity of the B-site layers yields a
difference of about 20 % between dim and bright atom columns (Figure 6.1f) which is too
large to be explained even by complete ordering of Fe and Co between the atom columns.
Instead, the large variation in B-site atom column intensity is attributed to the large static
displacements induced by VO ordering, as in the case of Nb-doped Sr(Co,Fe)O3−δ.209
From now on we identify this structure as layered perovskite (LP), which comprises the
majority of the x = 0.74 film and has a uniform appearance throughout the film (Figure
6.A1). The average out-of-plane lattice parameter of the LP determined from STEM images
was 3.87 Å, which is comparable to the out-of-plane lattice parameter obtained by X-ray
diffraction (XRD) (Figure 6.2a), 3.81 ± 0.4 Å. A simple model shows that LP and P are
difficult to distinguish by XRD (Figure 6.A2), explaining why the XRD by itself suggests
that the film consists only of P. In this x = 0.74 sample, a small amount of BM was also
observed near low-angle grain boundaries (Figure 6.A3); but is too little to yield a detectable
peak in XRD.
Reciprocal space mapping (RSM) (Figure 6.2b) shows that the films are fully strained
in-plane to match the STO substrate (a = 3.905 Å), consistent with the STEM images.
The SCFO films have smaller lattice parameters than the STO substrate (bulk P-SCO and
P-SFO lattice parameters are 3.836 Å150 and 3.850 Å151 respectively) and therefore the
SCFO is under in-plane tensile strain. The layered structure observed here differs from that
of La0.6Sr0.4CoO3−δ films on STO subject to in-plane tensile strain, where layers of oxygen
ordering oriented out-of-plane are believed to accommodate the strain.210
The presence of the LP structure is dependent on the Co:Fe ratio. The x = 0.43 film
103
consists of a majority BM phase which produces its characteristic (0010) superlattice peak
in XRD (Figure 6.2a), unlike the x = 0.74 sample that consists mainly of LP. However,
microscopically, cross-sectional HAADF STEM image (Figure 6.A4) of the x = 0.43 film
indicates that three different structures coexist in a single film: BM, LP, and P. Even though
STEM reveals the presence of LP and P in addition to BM, the main peaks (near the
substrate peaks) of these three phases are not readily distinguishable in XRD. Therefore,
XRD suggests only BM for the x = 0.43 sample whereas STEM indicates that BM, LP, and
P are all present. The LP structure itself appeared the same for x = 0.43 and x = 0.74
according to the STEM images.
The LP structure observed in the x = 0.74 and x = 0.43 films differs from typical
oxygen-deficient perovskite structures with random VO distribution, but it is not likely to
be one of the Ruddlesden-Popper phases because the B-site cation positions observed in
STEM images are not interleaved. Furthermore, the oxygen ordering and periodic inten-
sity change in the out-of-plane direction are inconsistent with the SrFeO2.75 or SrFeO2.875
structures that have been reported.204–206 ,211 The mechanism of oxygen ordering in LP
is different from La 212–2150.5Sr0.5CoO3−δ because SCFO does not have two different A-site
cations. A structure with alternating B-site intensities in its STEM image was reported for
SrCo Fe Nb O ,2090.7 0.2 0.1 2.72 attributed to vacancy ordering leading to octahedral tilting.
6.3 Density functional theory and multislice method based
image simulation and analysis
Figure 6.3 summarizes the experimental and theoretical structures reported on SrBO3−δ
compounds with respect to two variables, δ from 0 to 0.5 and B = Co, Fe, or Mn.145 ,147 ,151 ,153 ,197–209
In order to understand the ordering of oxygen ions in our SCFO films and compare with
the structures reported in Figure 6.3, we considered the simulated STEM images of 8 dif-
ferent structures named SrBO3, PL-SrBO2.75, BM-SrBO2.75, 314-SrBO2.625, BM-SrBO2.625,
104
Figure 6.3: Map of observed SrBO3−δ structures as a function of oxygen deficiency (δ, x axis)
and B-site cation chemistry (y axis). Orange and teal colors in the map represents bulk and
thin films, respectively. The structural information was drawn from (145 , 147 , 151 , 153 ,
197–209 ).
PL-SrBO2.5, BM-SrBO2.5, and SrBO2, as depicted in Figure 6.4. These structures were se-
lected because they have periodic oxygen deficiency in the B-coordinated layers along the
[001] direction (i.e. film growth direction). We use the term PL to distinguish the structures
from the one with random VO distribution and to underscore a specific ordering of oxygen
ions in an oxygen-deficient P.
In Figure 6.4, SrBO3 was a reference cubic perovskite without oxygen deficiency. PL-
SrBO2.75 was derived from a stoichiometric P structure but consists of layers of alternating
pyramidal and octahedral B-site coordination. We excluded the structure of bulk Sr4Fe4O11
(δ = 0.25)216 because it has apical VO in its A-site layers instead of the B-site layers. PL-
SrBO2.5 consisted of octahedral and square planar coordination layers, also derived from
removing oxygen from a stoichiometric P structure. (An alternative SrBO2.5 consisting of
layers of square planar coordination of BO4 would not have a difference in oxygen con-
tent between the two B-site layers.) The 314-SrBO2.625 structure was obtained from that
of Sr3Y1Fe4O 21710.5 by replacing the Y with Sr. It consists of alternating layers composed
of octahedral and a mixture of pyramidal and tetrahedral coordination. BM-SrBO2.5 is a
105
typical brownmillerite structure with alternating planes of octahedrally and tetrahedrally co-
ordinated B-sites. BM-SrBO2.625 was constructed by adding one oxygen to one of the planes
of the tetrahedrally coordinated B-sites of BM-SrBO2.5, and BM-SrBO2.75 was obtained by
adding another oxygen in the remaining tetrahedrally coordinated layer. Lastly, in SrBO2,
the two alternating layers consist of BO6 and BO2.
DFT calculations were performed to relax the 8 structures with B = Co and Fe while
constraining the in-plane lattice to be epitaxially matched to that of STO. Neutral oxy-
gen vacancies were assumed during simulation because it is energetically favorable for the
transition metal to reduce upon increasing oxygen deficiency rather than form charged va-
cancies.216 ,218 The out-of-plane lattice parameters of the strained structure (Table 6.A1),
unit cell volume and lattice parameter difference from the fully relaxed state (Table 6.A2
and Table 6.A3), and formation energies of the strained state (Table 6.A4) are summarized
in the Supporting Information. SrCoO3−δ structures in general showed a lower formation
energy compared to SrFeO3−δ. When B = Co, there is an increase in lattice parameter with
increasing δ, whereas there is less change in lattice parameter with increasing δ when B =
Fe.
We then simulated the corresponding ADF and iDPC images using multislice-based219
STEM imaging with the structures obtained from DFT. The comprehensive results are shown
in Figure 6.A5, and images corresponding to selected structures SrFeO3, 314-SrFeO2.625, BM-
SrFeO , PL2.625 -SrFeO2.5, BM-SrFeO2.5 are shown in Figure 6.5a. Even before simulating the
STEM images, we were able to exclude the possibility of BM-SrBO2.625 and SrBO2, because
the former showed different B cation positions from those in the STEM images and the
latter was highly unstable based on DFT calculations (Table 6.A3). We can also rule out the
possibility of BM-SrBO2.5, because the XRD did not show the characteristic BM superlattice
peaks in Figure 6.2a, and BM could be clearly distinguished as a different structure in Figure
6.A3. Lastly, the ordering observed in iDPC of BM-SrBO2.75 in Figure 6.A5 is lacking in our
experimentally observed images, eliminating this structure from the candidates. This leaves
106
Figure 6.4: Crystal structures of (a) P-SrBO3, (b) PL-SrBO2.75, (c) BM-SrBO2.75, (d) 314-
SrBO2.625, (e) BM-SrBO2.625, (f) PL-SrBO2.5, (g) BM-SrBO2.5, and (h) SrBO2 viewed along
the pseudocubic [110] axis with only B-site cation polyhedra visible. Blue corresponds to
BO6, red corresponds to BO5, and yellow corresponds to BO4 coordination. The BO4 co-
ordination in (f) is square planar with slight distortions. In (h) SrBO2, the polyhedra in
oxygen-deficient layers (between the octahedral layers) are absent because there are no oxy-
gen atoms in that lattice plane.
PL-SrFeO2.75, 314-SrFeO L2.625, and P -SrFeO2.5 as the most plausible structures.
We further compare the structures by determining the Sr—Sr distance ratio between the
alternating planes along the out-of-plane direction (Figure 6.5b) and the B-site intensity
deviation from the mean (Figure 6.5c) from the simulated STEM images. Out of the 8
structures, both 314-SrBO L2.625 and P -SrBO2.5 most resemble the experimental results in
Figure 6.1c in three respects. First, the alternating Sr—Sr distance length ratio of these two
structures is ∼1.2, at the upper end of all the structures except for BM-SrBO2.5 and SrBO2.
Second, the B-site intensity variation is ∼10 %, which is the next highest after the unstable
SrBO2. Finally, the [110] and [11̄0] projections are equivalent ([11̄0] projections are shown in
Figure 6.A5). This is an important criterion because all of the LP seen in the STEM images
has the same appearance; a structure with inequivalent [110] and [11̄0] projections would
be expected to show the two different variants in different parts of the sample. Kinematic
X-ray diffraction calculations indicate that the superlattice reflection intensity is about 2.0
% of the main peak for 314-SrBO2.625 and 4.5 % for PL-SrBO2.5 compared to 7.5 % of BM-
107
SrBO2.5, i.e. superlattice peaks of 314-SrBO2.625 and PL-SrBO2.5 are expected to be present
but would be weaker than those of BM-SrBO2.5. However, in the x = 0.74 film, superlattice
peaks were absent despite the clear periodicity of the Sr—Sr distance. We speculate that
this is due to the mosaicity of the crystallites or disorder in the structure which would lower
the peak intensities below the calculated values.220
We also consider the stability of the proposed layered structures with respect to a P
unit cell with randomly placed VO, namely R-SrBO2.625 which was generated by introducing
three VO in a 2 × 2 × 2 perovskite supercell analogous to (221 ). 314-SrFeO2.625 is stable by
−87.4 meV f.u.−1 compared to R-SrFeO2.625, whereas 314-SrCoO2.625 is unstable by +61.9
meV f.u.−1 compared to R-SrCoO2.625, suggesting that Fe stabilizes the ordering of VO. (We
did not make a comparison for δ = 0.5 because of the large number of possible configurations
of R-SrBO2.5.)
Considering the comparisons between experimental data and simulations, we cannot say
decisively that the LP phase is an exact match for 314-SrBO L2.625 or P -SrBO2.5. Although
314-SrBO2.625 is energetically more stable than PL-SrBO2.5 based on DFT, we do not observe
the expected slight modulation of the B-site cation positions in the iDPC image shown
in Figure 6.5a, suggesting that the structure is closer to that of PL-SrBO2.5. However,
the composition may have fewer VO than PL-SrBO2.5 because Fe typically favors a higher
oxidation state compared to Co. It is possible that the experimental result has characteristics
of both 314-SrBO2.625 or PL-SrBO2.5, i.e. a structure with alternating layers of octahedral
and mixed B-site coordination. The discrepancy between the observed Sr—Sr distance ratio
∼1.4 and the calculated values ∼1.2 may be affected by epitaxial strain: the Sr—Sr distance
ratio calculated from bulk unstrained 314-structured SrFe0.25Co 2080.75O2.63 is ∼1.1, smaller
than our calculated value of ∼1.2 for tensile-strained supercells.
We marked our best description of the LP structure on Figure 6.3, highlighting that
the ordering of oxygen ions observed in the thin films of this work is atypical among the
series of BM Sr(Co,Fe)O2.5 solutions. We represent its δ as a range between that of the two
108
Figure 6.5: (a) Simulated ADF (left) and iDPC (right) images of SrFeO3, 314-SrFeO2.625,
BM-SrFeO L2.625, P -SrFeO2.5, and BM-SrFeO2.5, along the (pseudo)cubic [110] axis. His-
togram of (b) Sr—Sr distance ratio vs. structure and (c) B-site intensity deviation from
mean vs. structure analyzed from simulated ADF images.
best-fit structures, 314-SrBO L2.625 and P -SrBO2.5. As a final comment on the LP structure,
in (119 ) we reported that ionic liquid gating can introduce oxygen into the layered struc-
ture causing it to separate between the layers, yielding distorted layers with wider spacing
(Figure 6.A6). Upon application of electrostatic driving force, the oxygen-deficient layers
may provide sites for oxygen insertion, producing linear defects between the planes due to
Coulomb repulsion.119
109
6.4 Conclusion
In summary, we report an oxygen-deficient layered perovskite structure in SCFO films
grown by combinatorial PLD. The perovskite-derived structure of SCFO with x = 0.74
shows clear evidence of layers with alternating oxygen deficiency parallel to the film growth
direction, resulting in the expansion (oxygen-deficient) and contraction (oxygen-rich) of the
out-of-plane Sr—Sr distance, concurrent with the B-site ADF intensity deviation (dim and
bright, respectively). This can be described as the layering of BO6 and BO4/BO5 coordinated
B-site cations resembling 314-SrBO or PL2.625 -SrBO2.5, neither of which has been previously
identified in thin films. The layered structure is dominant and consistent throughout the film
when x = 0.74 and is present as a minority phase at a different Co:Fe ratio, x = 0.43. The
stability of this crystal structure, and the remaining ambiguity in the exact oxygen ordering
encourages further investigation of its origin, as well as the effects of the interplay between
kinetics and thermodynamics in PLD associated with the oxygen pressure during growth,
and the prospect of achieving yet more ordered structures in ABO3−δ. This study also
motivates the exploration of the anisotropic magnetic, electrical, and transport properties of
the oxygen vacancy ordered perovskite structure.
6.5 Methods
SCFO films with different Co:Fe ratios were deposited through PLD on (001) oriented
STO substrates by ablating SCO and SFO targets with a KrF excimer laser (λ = 248
nm) at a fluence of 1.3 J cm−2 and 5 Hz repetition rate. PO2 during growth was 150
mTorr, the heater setpoint was 850 ◦C and the actual substrate temperature was around
700 ◦C. Characterization of the film crystal structure was performed by high-resolution
XRD using a Rigaku Smartlab high-resolution diffractometer with Ge-(220) double-bounce
monochromator and Cu Kα1 (λ = 1.5405 Å) as the X-ray source. The Co:Fe ratio was
110
determined using wavelength dispersive spectroscopy (WDS, JEOL-JXA-8200 Superprobe).
Cross-sectional STEM samples of SCFO films were prepared by mechanical wedge polishing
or by using a Ga+ ion beam in a FEI Helios Nanolab 600 Dual Beam system. Mechanically
polished samples were further thinned down by Ar ion milling. A probe corrected Thermo
Fisher Scientific Titan G3 60-300 kV microscope operating at 200 kV was used for STEM
imaging. The beam current was 60 pA and the convergence semi-angle was 19.2 or 24.8
mrad. The collection semi-angle range for ADF and iDPC images were 25-153 and 6-24
mrad, respectively.
DFT calculations were performed with the Vienna Ab-initio Simulation Package (VASP),
with generalized gradient approximation (GGA) Perdew-Burke-Ernzerhof (PBE) functional.
The GGA+U method was used with the Hubbard U value of 4 eV (U = 4 eV, J = 0
eV) for both Co and Fe.222–224 The valence configurations used were Sr (4s24p65s2), Co
(3s23p63d84s1 for BM-SrCoO and 3p63d84s1 for the others), Fe (3s23p62.5 3d74s1 for BM-
SrFeO and 3p63d74s12.5 for the others), and O (2s22p4), and the energy cutoff used was 600
eV. The different valence configuration for BM is because including 3s2 for both Co and
Fe gave a better match to the known G-type antiferromagnetic state. A 2 × 2 × 2 cubic
supercell with 8 formula units (f.u.) was used for SrBO3, PL-SrBO L2.75, R-SrBO2.625, P -
√ √
SrBO2.5, and SrBO2 structure calculations, a pseudocubic 2 2 × 2 2 × 4 supercell with
√ √
32 f.u. was used for 314-SrBO2.625 calculations, and a pseudocubic 2 × 2 × 4 unit cell
with 8 f.u. was used for BM-SrBO2.75, BM-SrBO2.625, and BM-SrBO2.5 calculations. The
corresponding Monkhorst-Pack k-point grids corresponding to each supercell were 6 × 6 × 6,
2 × 2 × 2, and 6 × 6 × 2. STEM ADF and iDPC images were simulated using the multislice
approach with a custom Python-based STEM image simulation software. Supercells were
constructed using the relaxed structure with varying amount of oxygen vacancy from DFT.
Simulated images were further convolved with a Gaussian with a full-width at half-maximum
of 80 pm to account for the finite source size.
111
6.6 Appendices
Figure 6.A1: Additional STEM images highlighting the consistency and uniformity of the
“layered perovskite” (LP) structure.
112
Figure 6.A2: Powder diffraction pattern simulated using ISODISTORT,225 corresponding to
perfect perovskite ABO3 (a-c) and a structure with alternating oxygen occupancies in the
BO2 layer (d-f). The oxygen concentration was increased and decreased in alternating layers
to maintain an overall stoichiometry of ABO3. (a) and (d) are structures used for simulation;
(b) and (e) are calculated diffraction patterns with 2θ ranging from 25◦ to 80◦; (c) and (f) are
enlarged views around the (002) (pseudocubic) reflection. The (002) peak position was not
affected by the alternating oxygen occupancies. The arrow in (f) indicates a negligible peak
due to the oxygen superlattice which would be below the detection limit in XRD considering
its intensity vs. (002). The simulation showed that vacancy ordering did not alter the (002)
reflection position compared to P, and produced only a slight intensity change. No significant
additional peaks were produced. The insensitivity of the XRD to oxygen vacancy ordering
is consistent with results from Ti-deficient STO175 or SrCo0.7Fe 2090.2Nb0.1O2.72, due to the
cubic symmetry of the parent structure. This suggests that XRD cannot easily distinguish
between LP and P structures in the film.
113
Figure 6.A3: ADF STEM image of x = 0.74 film. The blue box indicates the BM structure
and the red box indicates the LP structure.
114
Figure 6.A4: HAADF STEM image of x = 0.43 film. The green, red, blue boxes indicates
P, LP, and BM structures.
115
Figure 6.A5: Collection of simulated ADF and iDPC STEM images of SrBO3−δ with B =
Fe, Co, viewed along (pseudo)cubic [110] and [11̄0] axes.
116
Figure 6.A6: Lattice disruption (dark lines within the layers) in the LP structure as an effect
of ionic liquid gating.
117
Table 6.A1: Out-of-plane (pseudo)cubic lattice parameters (Å) of each structure, when epi-
taxially strained on STO, from DFT.
SrBO PL- SrBO BM- 314- BM- P
L- BM-
3 2.75 SrBO2.75 SrBO
SrBO2
2.625 SrBO2.625 SrBO2.5 SrBO2.5
Co 3.86 3.91 3.90 3.92 3.99 3.97 3.96 4.13
Fe 3.92 3.96 3.94 4.00 3.98 4.01 3.98 4.03
Table 6.A2: Unit cell volume difference (%) of strained structures with respect to the fully
relaxed state.
PLSrBO - BM- 314- BM- P
L- BM-
3 SrBO2.75 SrBO2.75 SrBO2.625 SrBO2.625 SrBO2.5 SrBO
SrBO2
2.5
Co 0.66 0.25 −1.68 0.52 −0.80 −0.78 1.24 2.18
Fe −0.28 0.15 −0.76 −0.80 −1.36 −2.34 1.33 −0.55
Table 6.A3: Out-of-plane pseudocubic lattice parameter difference (%) of the strained struc-
tures with respect to the fully relaxed state.
L
SrBO P - BM- 314- BM- P
L- BM-
3 SrBO SrBO22.75 SrBO2.75 SrBO2.625 SrBO2.625 SrBO2.5 SrBO2.5
Co −0.55 −0.30 1.69 −0.52 1.09 1.40 −0.51 −1.29
Fe 0.17 −0.39 0.62 0.89 2.38 2.96 −0.60 0.29
118
Table 6.A4: Formation energy ∆E (meV f.u.−1) of each structure with respect to stoichio-
metric SrBO3, calculated using the equation ∆E = E(SrBO3−δ) + δµO − E(SrBO3). All
structures are epitaxially strained on STO. The formation energy for BM-SrBO2.5 has been
obtained with SrBO3 energy calculated using the same pseudopotential (including 3s2). We
also note that PL-SrBO2.5 was unstable compared to BM-SrBO2.5 (for both Co and Fe)
when the same pseudopotential was used. A ferromagnetic (FM) state was more stable for
314-SrCoO2.625 and SrCoO2. BM-SrFeO2.75 did not stabilize to either FM or G-type antifer-
romagnet (G-AFM) but the formation energy was calculated with the more stable structure
(FM except for one spin). Other structures were stabilized in the G-AFM configuration.
PL- BM- 314- BM- PL- BM-
SrBO2.75 SrBO2.75 SrBO2.625 SrBO
SrBO2
2.625 SrBO2.5 SrBO2.5
Co −153.0 −135.8 −6.90 −180.5 −82.7 212.6 1470.3
Fe 47.3 39.3 99.0 98.5 210.7 672.8 2261.5
119
Chapter 7
First principles calculation of oxygen
vacancy effects on the magnetic
properties of perovskite strontium
nickelate
This chapter is based on a publication which the author wrote and published in Physical
Review Materials (2021).221
7.1 Introduction
Nickelates exhibit distinct physical properties compared to oxides of lighter transition
metals. The delicate interplay between structural distortions and the various energies (the
charge transfer energy, Ni and O bandwidths, and Coulomb interaction U) leads to a range
of complex and fascinating electronic behavior. Rare-earth perovskite nickelates, RNiO3
are particularly well studied because the change in size of the R cation allows structural
effects to be examined systematically.226 ,227 The O 2p and Ni 3d bands are close in energy,
120
ligand to metal charge transfer energies are small or negative, and the excitation with lowest
energy is p-d electron transfer in which an electron moves from oxygen to the cation leaving
a ligand hole L.226–231 The ground state includes a significant contribution from dn+1L states
hybridized with dn states (n = 7 for Ni3+) and the materials are charge transfer insulators
at low temperatures. Furthermore, perovskite RNiO3 (except for R = La) are noncollinear
antiferromagnetic insulators at low temperatures with a monoclinic distortion that produces
two inequivalent Ni sites due to bond disproportionation.226 ,227 ,232 Nickelates can exhibit
metal-insulator transitions,233–235 charge disproportionation,236 charge localization,237 and
superconductivity.238–241
While rare-earth nickelates have been widely studied, fewer syntheses of alkaline-earth
nickelates have been reported, and their electronic structure is not as well characterized.
For a divalent A such as Ba2+ or Sr2+ and in the absence of oxygen vacancies, A2+NiO3
would require a formal valence of Ni4+. There have been several studies of bulk hexagonal
nickelates with divalent A such as SrNiO 242 ,2433 and BaNiO .244–2473 Hexagonal SrNiO3 has
been synthesized by calcination242 or by heating in 50–2000 atm oxygen (whereas 1 atm
oxygen produced SrNiO 2432.5). SrNiO2.5, Sr4Ni3O9, Sr5Ni4O11, and Sr Ni O 248–2509 7 21 have
also been synthesized. High valence Ni was proposed in Sr4Ni3O 2509 and in alkali metal
nickelates such as a monoclinic Li NiO 251 ,2522 3 and KNiIO 253 ,2546. Despite these works, the
presence of Ni4+ in oxides remains controversial. Spectroscopic studies reveal considerable
hybridization between Ni and O, the presence of O ligand holes, and Ni valence less than
4+,247 ,251 analogous to the behavior found in the rare-earth nickelates.
The large ionic radii of divalent Sr and Ba lead to Goldschmidt tolerance factors exceeding
1, which explains the stability of the hexagonal SrNiO3 and BaNiO3 compounds compared
to the cubic perovskite structure. However, cubic perovskite-structured compounds have
been stabilized if there is only partial substitution of Ni, such as in cubic SrFe1−xNixO3
with x = 0-0.5.35 ,255 Recently, a single unit cell layer of epitaxial cubic perovskite SrNiO3
was stabilized by charge redistribution in a SrNiO3/(LaFeO3)n superlattice synthesized by
121
molecular beam epitaxy,256 but another recent study to synthesize epitaxial cubic SrNiO3
yielded phase separation into Sr2NiO and SrNi O .2573 2 3
Further analysis of cubic perovskites A2+NiO3 is motivated by the interesting electronic
and magnetic properties reported to date. Half metallicity and ferromagnetism of perovskite
SrNiO3 was first predicted by ab initio simulations,258 where the ground state was ferromag-
netic and the hybridization between dz2 (dx2−y2) and pz (px and py) induces large magnetism
at the O sites. Ni was proposed to be in the intermediate spin state of d6 and the Ni—Ni inter-
action was explained by a double-exchange mechanism. A first principles study of hexagonal
SrNiO3 also predicts robust half-metallic properties and ferromagnetism, governed by the
semicovalent bonding between Ni and O.259 Ni was predicted to induce ferromagnetism in
antiferromagnetic SrFeO ,260 ,2613 and experimentally SrFe1−xNixO3 shows room temperature
magnetism when x ≥ 0.4.35
Here, we investigate the electronic and magnetic properties of the cubic perovskite
SrNiO3−δ (SNO) with δ = 0, 0.125, 0.250, and 0.375 through density functional theory
(DFT) modeling. We focus on clarifying the electronic and magnetic configuration of Ni and
the role of oxygen vacancies and strain on the electronic structure. Density of states (DOS)
and projected density of states (PDOS) reveal the presence of oxygen ligand holes instead of
tetravalent Ni. There is ferromagnetic coupling between the Ni cations, and significant mag-
netism originating from the O antiparallel to that of the Ni. This observation is maintained
throughout increasing δ. Moreover, these behaviors extend to other A2+NiO3 compounds
including CaNiO3 and BaNiO3.
7.2 Computational methods
All first principles calculations were performed with the Vienna Ab-initio Simulation
Package (VASP). The Perdew-Burke-Ernzerhof (PBE) functional, a widely used type of
generalized gradient approximation (GGA), was used within the calculation. The valence
122
configurations used were Sr (4s24p65s2), Ni (4s23d8), and O (2s22p4). The GGA+U method
was used to account for the Coulomb interaction of the Ni 3d electrons, with a specific U
value of 6.2 eV (U = 6.5 eV, J = 0.3 eV), where the effective U = 6.2 eV was obtained for
NiO using the simplified rotationally invariant method.262 The energy cutoff used for cell
relaxation and comparison of different symmetry was 600 eV, and for ionic relaxation the
cutoff was 500 eV. K-point grids were maintained to a Monkhorst-Pack263 6 × 6 × 6 mesh
for all cubic structure calculations. We started from a 2 × 2 × 2 cubic supercell consisting
of 8 Sr, 8 Ni, and 24 O atoms, which is the smallest cell size that can take account of
all ferromagnetic and antiferromagnetic ordering configurations. One, two or three oxygen
atoms were removed from the 2 × 2 × 2 supercell to make a formula unit (f.u.) of SrNiO2.875
(δ = 0.125), SrNiO2.75 (δ = 0.250) or SrNiO2.675 (δ = 0.375), respectively. For hexagonal
(BaNiO -like 2H2643 and SrMnO3-like 4H,265 both P63/mmc), rhombohedral (R3̄c and R3̄m),
and orthorhombic (Pnma) structures, a unit cell containing 2 f.u., 4 f.u., 8 f.u., 8 f.u., and
8 f.u. with 8 × 8 × 8, 8 × 8 × 6, 2 × 2 × 6, 4 × 4 × 4, and 6 × 6 × 4 k-point grids was
used, respectively. Structural characterization was performed using VESTA266 and Bader
charge analysis was performed to determine the oxidation state of each atom.267 The oxygen
vacancy formation energy was determined using the following relations:
1
∆E(first VO) = E(one VO)− E(defect-free) + E(O2)
2
1
∆E(second VO) = E(two VO)− E(one VO) + E(O2) (7.1)
2
1
∆E(third VO) = E(three VO)− E(two VO) + E(O2)
2
where the overbinding of the oxygen molecule was corrected in E(O ).982 We also used the
Heyd-Scuseria-Ernzerhof (HSE) functional for non-defective and one-vacancy ferromagnetic
SNO to demonstrate that the electronic and magnetic behavior are qualitatively the same
as the PBE result (Figure 7.A1). We use the notation that majority carriers are spin up and
minority carriers are spin down.
123
We applied in-plane strain to the δ = 0 system to predict the effect of epitaxial growth
on different substrates. The zero-strain lattice parameter was determined first from the fully
relaxed structure. Up to +2 % (−2 %) of in-plane biaxial tensile (compressive) strain was
applied by fixing the lattice in the x and y direction, while it was allowed to fully relax
in the z direction. Moreover, we performed an analogous set of calculations with CaNiO3
(CNO) and BaNiO3 (BNO) to investigate the effect of A cation and discuss the generality
of the behavior of Ni and O. The valence configurations used for Ca and Ba are 3p64s2 and
5s25p66s2 respectively.
7.3 Structural, electronic, and magnetic properties of cu-
bic perovskite strontium nickelate
We start by determining the most stable magnetic ordering of the cubic perovskite SrNiO3
(Figure 7.1a). Four different types of ordering were initialized: ferromagnetic (FM), G-type
antiferromagnetic (G-AFM, which has AFM order on both the xy plane (“in-plane”) and
xz or yz planes (“out-of-plane”), A-type antiferromagnetic (A-AFM, FM order in-plane and
AFM order out-of-plane), and C-type antiferromagnetic (C-AFM, AFM order in-plane and
FM order out-of-plane). FM was the ground state, followed by A-AFM, C-AFM, and G-
AFM (Table 7.A1). The order of stability is consistent with a prior report,258 considering
the fact that the U − J = 6.2 eV used here is in between the 6 eV and 7 eV used in (258 ).
Structural relaxation indicated that atomic position symmetry was maintained for all types
of ordering, so the octahedra did not exhibit distortions, which is attributed to the larger
ionic radius of Sr2+ (118 pm) compared to that of rare-earth ions (La3+-Lu3+ have radii 102-
86 pm). Calculations using a larger supercell or lowering the symmetry with an initial ion
displacement yielded Ni—O—Ni bond angles of 180o and octahedral distortion was absent.
This excludes the possibilities of noncollinear magnetization, bond disproportionation, or
the E, S, or T-type antiferromagnetic orderings found in rare-earth nickelates.268–270
124
Figure 7.1: (a) Crystal structure of SrNiO3. Green, grey and red correspond to Sr, Ni, and
O atoms. One fixed VO location (marked Fix) and the other seven possible positions for δ =
0.250 are labeled. (b) Structure of [222] SNO2.75 with selected atoms for Figures 7.2c and
7.2d, and Table 7.2 labeled.
FM and G-AFM SNO remained cubic, but A-AFM and C-AFM SNO had tetragonally-
distorted unit cells. The ground state cubic FM had a lattice parameter of 3.890 Å. We
compared the energy of hexagonal, rhombohedral, and orthorhombic FM SNO relative to
that of cubic SNO. Whereas 2H hexagonal is the most stable structure (−397.96 meV f.u.−1),
consistent with the experimental observations of hexagonal SNO, the rhombohedral or oc-
tahedral symmetry converged into the cubic. In particular, Pnma and R3̄c structures which
started with the presence of octahedral tilts converged into a cubic structure without rota-
tions. This further suggests that other types of octahedral distortion or symmetry breaking
are not stable. It is worth noting that the hexagonal structure used in (259 ) is a 4H struc-
ture, which is 349.13 meV f.u.−1 higher in energy than the 2H structure used in this study.
The two different structures are visualized in Figure 7.A2. The 2H structure is what was
observed experimentally.243
As reported in a previous study,258 δ = 0 SNO is half-metallic and the DOS at the Fermi
level (EF ) arises from O 2p orbital electrons (Figure 7.2a and 7.A3). In order to explain
the electronic structure of SNO, we first focus on the PDOS of Ni. The eg electrons show
125
significant spin polarization, whereas the polarization of t2g electrons is weak. The bonding-
antibonding splitting of π bonds is 4 eV, which is smaller than that of the σ bonds which is
around 8 eV, and this value is similar to the splitting of Co observed in SrCoO .2713 The σ
bonding can be represented as the occupied spin up electrons of Ni eg hybridizing with O pz
at 6 eV below EF . Between O px (py) orbitals and Ni dxz (dyz) orbitals, there are π bonding
and anti-bonding hybridizations near 1 eV and 5 eV below EF . In the PDOS of O, p orbitals
with π bonding are less spin-polarized, whereas those with σ bonding exhibit polarization.
The DOS exhibits the characteristics of oxygen ligand holes L, where O p orbitals are
partially filled and contribute to the metallic character. The presence of ligand holes is fur-
ther supported by the qualitative similarity in PDOS calculated using the HSE functional,
which also suggests the validity of U − J = 6.2 eV used in this study (Figure 7.A1). The
presence of ligand holes is analogous to findings in RNiO 234 ,272–2753 where Ni has a d8L con-
figuration, although in the case of SNO, we do not observe metal-insulator transitions upon
introducing oxygen vacancies (discussed in section B) as seen in SmNiO3 or NdNiO 234 ,2353.
The presence of ligand holes was also suggested in other ABO3 compounds with divalent A
such as SrCoO or CaFeO ,271 ,2763 3 and the electronic structure observed here has similarities
to those. However, Ni has negligible contribution to the metallicity unlike SrFeO3 or SrCoO3
where Fe, Co, and O all contribute to states near the Fermi level.277
Based on these PDOS, we propose an electronic configuration (Figure 7.3) where Ni has
all six t2g electrons and two spin-up electrons in eg, and for O, px and py are filled but only
a spin-down electron occupies pz. This configuration is further supported by noting that
the Ni PDOS is similar to that calculated for SmNiO 2783 (GGA+U) where Ni would have
a formal charge of 3+, d7 but with a contribution of d8L, as well as being similar to that
of NiO (GGA+U)279 and Li-doped NiO (GGA+U and HSE)280 where Ni should be in the
2+ oxidation state, d8. We have integrated the area of the PDOS below EF to estimate
the ratio of electrons between the eg and t2g orbitals, which shows that the area of eg is
around one third of t2g (Table 7.A2). These results imply that the electronic characteristics
126
Figure 7.2: PDOS of Ni and O bonded along the c axis of Ni in FM SNO, when (a) δ = 0,
(b) δ = 0.125, (c) octahedrally-coordinated δ = 0.250, (d) pyramid-coordinated δ = 0.250,
(e) CaNiO3, and (f) BaNiO3. Because the selected O is bonded in the z direction, the pz
orbital participates in σ bonding, and px and py orbitals participate in π bonding. In (a),
(b), (d), (e), (f) O, the px and py PDOS overlap.
127
Figure 7.3: Schematic of the proposed electronic structure of Ni and O in SNO, with the
dashed circle indicating the ligand hole in O.
of Ni in SNO are close to d8L2 or a mixture of d7L and d8L2. The d8L2-like configuration
can also explain how the FM ordering of ground state SNO can be observed, because the
spin-up electron from either Ni eg orbital adjacent to O can hop to the O pz orbital. In other
words, the coupling of the magnetic moments is FM between Ni—Ni, FM between O—O,
and AFM between Ni—O. This is also consistent with the sign of the magnetic moment of Ni
and O, which are 1.343 µB and −0.395 µB respectively. One interesting observation is that
even though Ni orders ferromagnetically with a non-zero moment, the net magnetization of
the system is nearly zero because the oxygens have an average magnetic moment about one
third that of Ni. Such half-metallicity and magnetic ordering of SNO would have interesting
consequences in tunnel magnetoresistance, x-ray magnetic circular dichroism, or neutron
diffraction experiments.
The results are consistent with a d8L2-like configuration of Ni rather than the d6 of Ni
with a nominal valence state of 4+. However, we expect the electron not to be localized
at a specific site, but to be delocalized due to the covalency of the bonds. Bader charge
analysis (Table 7.1) was performed to help clarify the electronic state of O. Oxygen has
a charge of −1, instead of −2 expected when the bonding is fully ionic, also supporting
the presence of a ligand hole in O. Such covalent bonding and strong hybridization of Ni
3d and O 2p is commonly observed in RNiO3 through X-ray absorption spectroscopy.281
The covalency between Ni and O is also reported in NdNiO3, as the oxygen contribution
in the electronic states is appreciable.282 Prior DFT work on SNO258 ,259 did not discuss
oxygen ligand holes, but showed that perovskite SNO and 4H-SNO have similar electronic
128
and magnetic properties.
Table 7.1: Bader charge analysis results of SNO3, SNO2.875, and [222] SNO2.75. Oxygen
charge and magnetic moments (δ) are averaged over the supercell.
Sr O—Ni—O O—Ni—VO O
Charge µ (µB) Charge µ (µB) Charge µ (µB) Charge µ (µB)
SrNiO3 +1.630 0.015 +1.414 1.330 −1.015 −0.451
SrNiO2.875 +1.594 0.013 +1.423 1.363 +1.261 1.310 −1.037 −0.391
[222] SrNiO2.75 +1.596 0.011 +1.388 1.448 +1.226 1.330 −1.055 −0.330
Figure 7.4: Strain effects on cubic FM SrNiO3, showing the total energy and the magnetic
moment of Ni and O as functions of the in-plane biaxial strain on the (001) plane.
We selected FM SNO, the ground state ordering, to apply epitaxial strain in the (001)
plane, distorting the cubic structure into tetragonal. The magnitude of both Ni and O
moments increases as tensile strain is applied, but the total magnetization remains nearly
zero for all strain states because of the AFM coupling between Ni and O (Figure 7.4). The
increase in the Ni moment with strain is analogous to an experimental finding in SrCoO3−δ,
129
where the Co valence state increased as epitaxial films were grown on substrates with larger
lattice parameters, observed from the increase in oxygen deficiency.283 Applying strain in
both (001) and (111) orientations of bulk SNO did not stabilize the cubic structure over
the hexagonal structure. However, when two layers of (111) cubic SNO are sandwiched in
between (111) cubic SrTiO3, the energy is more than 2 eV lower than that of two layers of
(0001) 2H hexagonal SNO sandwiched within SrTiO3. Moreover, the magnetic moments of
the Ni in the stabilized (111) SNO layer were 1.414 µB, and the electronic structure was
similar to the bulk result (Figure 7.A4), suggesting strategies for realizing cubic SNO within
a heterostructure.
7.4 Electronic structure dependence on oxygen vacancies
The behavior of perovskites is dramatically affected by oxygen vacancies (VO), which
lower the valence state of the cations, affect the number and mobility of charge carriers
and lead to structural distortions.284 Structural effects can include a change in the crystal
structure from perovskite to brownmillerite, ABO 26 ,272.5, on removal of sufficient oxygen. The
effects of VO have been extensively studied for ABO3−δ cobaltates, ferrites and manganates
with A = Sr, Ba and B = Mn, Fe and Co.30 ,200 ,271 ,285–290 The formation energy of VO
decreased as the radius of the B-site cation decreased, but the magnetism was not analysed
in detail.291–293 VO interactions were studied in SrFeO3−δ, and VO formation and transport
were investigated in (La,Sr)FeO3−δ and (Ba,Sr)(Fe,Co)O 289 ,290 ,294 ,2953−δ. Magnetic exchange
interactions and Curie temperature were calculated for SrCoO3−δ with δ ≤ 0.15.30 The
magnetization of Sr0.67La0.33CoO3−δ drops drastically at δ = 0.25 because of the change in
coupling.22 V may be expected to occur readily in A2+O NiO3 perovskites, but DFT studies
in SNO258 ,259 have not addressed their effects.
We investigated the structure and electronic properties of SrNiO2.875 (δ = 0.125) by
introducing one O vacancy in a 2 × 2 × 2 SNO supercell. Considering all four types
130
of magnetic ordering, we considered six nonequivalent locations where the vacancy can be
placed. For FM and G-AFM, the vacancy was placed in a site connecting octahedra along the
c axis of the crystal. Putting the vacancy along the a axis for FM does not yield qualitatively
different results. However, for A-AFM and C-AFM, there are two possible cases which are
the vacancy being located between Ni with the same spins (case 1) or opposite spins (case
2), and we considered both. Even with one O vacancy, FM was still the most stable ordering
(Table 7.A1). The lattice of FM SNO2.875 expanded in the direction where the vacancy was
located as a consequence of the increase in the distance between the Ni cations adjacent to
the vacancy, and octahedral tilts due to the vacancy were observed. The formation energy
of one neutral oxygen vacancy per 8 f.u. is 0.34 eV for FM SNO. This is lower than other
perovskites such as SrTiO 296 or SrFeO ,2943 3 and similar to that of SrNiO3 reported in (291 ).
In the FM state of SrNiO2.875 the total magnetization increased to 0.240 µB f.u.−1. For
Ni, the average moment increased to 1.356 µB, and for O, the magnitude of the average
moment decreased to −0.342 µB. The increase in net moment f.u.−1 is primarily due to the
removal of one O atom and its moment; the change in the Ni contribution is only 0.013 µB
out of 0.240 µB. To further investigate how the oxygen vacancy affects the magnetism in
FM SNO2.875, we compared the PDOS of Ni and O along the Ni—O—Ni—VO—Ni direction
with that of the non-defective perovskite SNO (δ = 0). Comparing the PDOS in Figure
7.2b with Figure 7.2a, a more hole-like feature in O pz appears near the peak just above
EF , because the absence of O in the other direction prohibits further exchange, and as a
consequence, the σ bonding and the delocalization of the spin-up electron is weakened. In
contrast, equatorial oxygens that do not have broken Ni—O bonds have qualitatively the
same PDOS properties as those of SrNiO3 (Figure 7.A5).
The PDOS of Ni remains almost the same as in SrNiO3, supporting the d8L2-like elec-
tronic structure of Ni, i.e. the change in oxygen stoichiometry mainly affects the ligand hole
characteristic of O rather than the valence state of Ni. The hybridization between the O pz
orbital and Ni eg orbital near −6 eV in the spin up state has increased on adding the VO, and
131
this explains how the polarization of O electrons and hence the magnitude of the O moment
was reduced. This is the opposite of what is expected for a SrCoO3−δ system, where the
unoccupied peak of Co is shifted to a higher energy, explained by the ligand hole in e .271g
Also, the mechanism for an increase in magnetization is different from that of SrTiO3, where
an oxygen vacancy leads to a magnetic moment due to the emergence of spin up states in
Ti atoms that were formerly S = 0.111
Two O vacancies in the 2 × 2 × 2 supercell system gives the formula SrNiO2.75 (δ =
0.250). When one vacancy position is fixed, there are seven possible locations for the second
one (Figure 7.1a). Each configuration is labeled by the crystallographic direction between the
two vacancies. Considering spin directions of Ni as well as the vacancy locations for C-AFM
and A-AFM further increases the number of possible configurations. However, we focused on
only one configuration for each case based on the results from SNO2.875, because the addition
of the first vacancy did not affect the order of energy (FM < A-AFM < C-AFM < G-AFM).
Among all possible magnetic ordering and vacancy position configurations, the FM [222]
structure illustrated in Figure 7.1b was the lowest energy state. Other configurations are
visualized in the Figure 7.A6. All of the AFM configurations were unstable and SNO shows
a robust stability in its FM ordering. At δ = 0.250, the net magnetization is about 0.5
µ −1B f.u. regardless of the vacancy position (Table 7.2), so these results provide a reference
for experimentally synthesized SNO in which the oxygen vacancy locations would likely be
random. The formation energy of the second oxygen vacancy depends on the configuration.
It is higher than the first vacancy on average, but the most stable configuration has a lower
formation energy (0.24 eV).
Comparing the 1-vacancy and 2-vacancy cases, as we add oxygen vacancies, the magnetic
moment of Ni increases and the absolute moment of O decreases (Figure 7.5). These two
effects combine to increase the total magnetization of the system. A similar behavior was
empirically observed in NdNiO3, where inducing oxygen vacancies increases the saturation
magnetization from the canted as-grown state.235 This was explained not only because of a
132
Table 7.2: Relative total energies (E), magnetic moment (µ) of Ni and O, total magnetic
moment per formula unit (f.u.) of FM SNO δ = 0.250, and the second oxygen vacancy
formation energy (∆Ef ) depending on the vacancy configurations. Magnetic moments of Ni
and O are averaged over the supercell.
Config. E (meV f.u.−1) µ of Ni (µB) µ of O (µB) Total µ f.u.−1 (µB) ∆Ef (eV)
[001] 99.14 1.368 −0.284 0.492 1.04
[010] 10.65 1.389 −0.289 0.485 0.33
[011] 13.30 1.371 −0.285 0.490 0.35
[121] 16.76 1.379 −0.287 0.492 0.38
[022] 46.69 1.384 −0.290 0.492 0.63
[220] 57.17 1.366 −0.285 0.488 0.70
[222] 0 1.388 −0.290 0.493 0.24
change in Ni valence state but also due to the overlap between the trapped electrons in the
vacancy and Ni d orbitals and the formation of a Ni—VO—Ni complex.
In order to understand the hybridization between Ni and O, the ground state configura-
tion [222] was chosen for the PDOS analysis. Figures 7.2c and 7.2d shows the PDOS of Ni
in octahedral (Ni(Oct)) or pyramid coordination (Ni(Pyr)), and adjacent O atoms along the
c axis. Both octahedra and pyramids are slightly compressed in the z direction by about
5 %, so there will be a small energy splitting within the t2g and eg orbitals. In particular,
dz2 will have a higher energy than dx2−y2 , and dyz and dzx will have a higher energy than
dxy, however, the grouping of t2g and eg is maintained. The peak around −3 eV of Ni eg
and O pz relates to the spin down hybridization between Ni—O. For Ni(Pyr), a peak around
−6.5 eV in the spin up states stands out for both Ni eg and O pz PDOS in comparison
with Ni(Oct). This explains how O(Pyr) (−0.174 µB) has a lower magnetic moment than
O(Oct) (−0.388 µB), and thus the total magnetization increases. The magnetic moment of
Ni slightly changed as well, where Ni(Oct) (1.460 µB) is higher than Ni(Pyr) (1.315 µB).
Even though the volume change observed in the direction of the oxygen vacancy is the same,
133
the behavior is different from that of Co in SrTi0.75Co0.25O3−δ, which has a higher magnetic
moment when the coordination of oxygen atoms decreases.297
A similar behavior is observed in the spin polarization as was seen for δ = 0.125. The
PDOS for Ni either in Ni(Oct) or Ni(Pyr) are almost unchanged, which could indicate that
the excess electrons resulting from the removed O atoms are delocalized.285 O participating
in the octahedral coordination (Ni—O—Ni—O—Ni) has an electronic structure similar to
that in non-defective SNO (Figure 7.2a), and O along the Ni—O—Ni—VO—Ni direction
participating in the pyramid coordination exhibits a distinct ligand hole characteristic from
the peak increase above EF . This trend was not only observed for the [222] configuration, but
also observed for other configurations as well. The Bader charge (Table 7.1) of Ni adjacent
to VO is slightly smaller than that of Ni surrounded by six O atoms, which could indicate
that the extra electron generated by the removal of one O has a preference to locate near
the VO site, but because all Ni and O have very small changes in their charge, it would be
more plausible that the extra electron delocalizes.
A third oxygen vacancy was added to the supercell, and the results are shown in the
Figure 7.A7. Because there are more vacancy configurations available with the third one,
the third position was selected to keep the distance between the other two vacancies as far as
possible, based on the finding for the two-vacancy case that the [222] configuration was the
most stable case. The trend in the electronic and magnetic properties upon introducing the
third vacancy is similar to the one- and two-vacancy cases: FM ordering remains the most
stable and the total magnetization increases due to the increase in Ni moment and decrease
in the net O moment. The PDOS of Ni remains nearly the same as the two vacancy case
and the PDOS of O depends whether it is in a Ni—O—Ni—VO—Ni (O1) or a Ni—O—Ni—
O—Ni (O2) configuration. The vacancy formation energy with respect to [222] SNO2.75 was
0.80 eV.
Overall, these observations support the model that Ni in SNO3 has an initial state close
to 2+ with ligand holes on O (d8L2 mixed with d7L), and the removal of O appears to
134
have more effect on the ligand hole of O than Ni. Observing the trend from δ = 0 to δ
= 0.375, the half-metallic property is not affected by the O vacancy; however, the states
in the conduction band mainly originate from the electrons of oxygens participating in the
π bonding. Moreover, the covalency between Ni—O bond decreases as oxygen vacancy
concentration increases.
Figure 7.5: Total, Ni, and O magnetic moment dependence on δ. δ = 0.250 Ni moments
were averaged over all seven FM configurations.
7.5 Expanding to other nickelate perovskites and het-
erostructures
In order to generalize the unusual behavior of SNO to perovskite A2+NiO3 with other
divalent A cations, we performed an analogous analysis on CNO and BNO. Ca and Ba were
chosen because they are one period lower or higher than Sr. Table 7.3 shows a summary
of the structure, electronic and magnetic properties of A2+NiO3. All of them were metallic
and stable in the ferromagnetic configuration. Both the magnetic moment of Ni and the
magnitude of the moment of O increases as the ionic radius of A2+ increases. The mag-
netic behavior of Ni and O in CNO and BNO were both analogous to that of SNO; Ni is
135
antiferromagnetically coupled with O, which have a substantial moment, and as a result
the net magnetization is negligible. This similarity also appears in the PDOS of Ni and O
shown in Figures 7.2e and 7.2f. The occupancy of t2g and eg are the same, in which eg is
spin-polarized whereas t2g is not. Moreover, the ligand hole feature in the O p orbital with
σ bonding appears distinctly in all three cases. These results give an interesting insight
into nickelate compounds, suggesting that oxides with higher formal valence Ni include O
ligand holes instead of Ni 3+ or 4+. We expect these characteristics will appear not only
in perovskite A2+NiO3 but also in 4H hexagonal A2+NiO3 based on the similarity between
perovskite SNO and 4H SNO.259
Table 7.3: Shannon ionic radius (r), tolerance factors, E(cubic) − E(hexagonal) per formula
unit (∆E f.u.−1), electronic and magnetic properties of A2+NiO3 (A = Ca, Sr, Ba).
r ∆E f.u.−1A Electronic Magnetic µ of A µ of Ni µ of O(pm) t τ (meV) Property Ordering (µB) (µB) (µB)
Ca 100 0.903 4.594 344.06 Half-metallic FM 0.024 1.208 −0.360
Sr 118 0.970 4.383 384.72 Half-metallic FM 0.011 1.343 −0.395
Ba 135 1.034 4.356 864.50 Metallic FM 0.007 1.528 −0.437
To compare the structural stability among A2+NiO3, we consider the predictions of the
tolerance factor. The Goldschmidt tolerance factor t is given by
rA + rO
t = √ (7.2)
2(rNi + rO)
where ri represents the ionic radius of ion i, and the perovskite structure is expected for 0.9
< t < 1. A more recently defined tolerance factor τ is given by298
( )
rA
rO r
τ = − n n − Nia A (7.3)
rNi ln(
rA )
rNi
where nA is the oxidation state of A cation, which is fixed to 2 in this case. The criterion
136
for the perovskite structure to form is τ < 4.18. Table 7.3 contains the tolerance factors
calculated using the Shannon ionic radii of Ca2+, Sr2+, Ba2+, Ni4+, and O2−. According
to t, the perovskite structure is more likely to form for CNO than for SNO or BNO due
to the lower ionic radius of Ca2+, but experimentally, other factors such as epitaxial strain
can be used to stabilize perovskite structures despite their tolerance factors predicting that
other structures are favored. This trend is opposite for τ , which arises from the fact that
t decreases with decreasing rA whereas τ increases with decreasing rA. In order to clarify
this discrepancy, we compared the energy difference between cubic and hexagonal structures,
∆E, defined as (energy of cubic A2+NiO3 per f.u.) − (energy of hexagonal A2+NiO3 per
f.u.) for non-defective structures. The results in Table 7.3 indicate that the energy difference
becomes smaller for Ca2+, which is consistent with the trend in the Goldschmidt tolerance
factor. Although the applicability of tolerance factors in compounds with strong covalent
character remains questionable, both t and τ will be lowered towards the perovskite regime
when O has a ligand hole character and as a consequence rO decreases.
7.6 Incorporation of nickel in perovskite ferrite thin films
As discussed throughout this chapter, nickelates are highly interesting because of their
novel and emergent properties, which differ from those of ferrites or cobaltates. While it may
be difficult to grow a single-phase cubic perovskite nickelate through pulsed laser deposition
(PLD) because phase separation into binary oxides can occur, Ni can substitute Fe sites in
ferrites, for example, in SrFeO3 and YFeO3, which will be described below.
Figure 7.6a depicts the X-ray diffraction (XRD) scan results of SrNixFe1−xO3 (SNFO)
thin films deposited by codeposition of SrFeO3−δ and SrNi0.5Fe0.5O3−δ targets at 850 ◦C of
substrate temperature setpoint. Ni content was measured by wavelength dispersive spec-
troscopy (WDS) which gave x = 0.07. When x increases, binary oxides such as NiO and
SrO2 begin to segregate. Lower laser fluence tends to suppress the formation of secondary
137
Figure 7.6: XRD results of (a) SrNi0.07Fe0.93O3 films grown on SrTiO3 (STO), (LaAlO3)0.3-
(Sr2AlTaO6)0.7 (LSAT), and LaAlO3 (LAO) substrates (panel on the right is an enlarged
plot around substrate (002) reflection), and (b) YNi0.05Fe0.95O3 films deposited on STO with
(001) and (111) growth directions. Film peaks are marked by stars.
phases. Ni substitution can take place in orthoferrite YFeO3 as well. Figure 7.6b shows XRD
results of x ≈ 0.05 YNixFe1−xO3 (YNFO) films grown via codeposition of Y3Fe5O12, Y2O3,
and YNiO3 targets on (001)- and (111)-oriented STO substrates. It forms a single-phase in
both growth directions.
SNFO films can also undergo structural change through the ionic liquid gating like
SrCo1−xFexO3−δ mentioned in Chapter 5. As shown in Figure 7.7, the brownmillerite peaks
(marked by asterisks) disappear and the film peak shifts to a higher 2θ, indicating a decrease
in the out-of-plane lattice parameter and thus oxidation upon negative gating voltage. It is
of interest whether Ni and Fe will demonstrate an antiferromagnetic coupling according to
the Goodenough-Kanamori rule,299 ,300 and how the total magnetization and Curie temper-
ature would change if an ordering between Ni and Fe is accompanied. Substituting Ni in Fe
sites will be a good starting point in experimentally exploring the characteristics of cubic
perovskite nickelates.
138
Figure 7.7: XRD results of as-grown and −2 V gated SNFO films with 7 % Ni substitution on
(001)-oriented (a) STO and (b) LSAT substrates. The substrate temperature setpoint was
750 ◦C and the growth temperature affects the orientation of octahedral- and tetrahedral-
coordinated layers in brownmillerite structure,301 resulting in additional peaks in contrast
to the results in Figure 7.6a.
7.7 Conclusion
We used DFT to predict the electronic and magnetic properties of stoichiometric and
oxygen-deficient cubic perovskite SrNiO3−δ. The results are consistent with the presence
of ligand holes, and Ni cations exhibiting a mixture of d8L2 and d7L states with a high
contribution of d8L2. Ferromagnetism mediated by O and half-metallicity are robust for
SNO with δ = 0, 0.125, and 0.250 or for epitaxially strained SNO with δ = 0. In particular,
for δ = 0.250, the magnetic ordering is little affected by the vacancy positions. Although Ni
is ferromagnetically ordered, because of the antiferromagnetic coupling between Ni and O
and the magnetic moment of the O due to its ligand holes, the total magnetization of the
system is nearly zero at δ = 0. However, the magnetism increases as SNO becomes oxygen
deficient, corresponding to the removal of the O moment, and the hybridization between Ni
139
and O becomes weaker as an oxygen vacancy is introduced. Other A2+NiO3 compounds,
CaNiO3 and BaNiO3, show analogous behavior.
SNO itself has proven to be challenging to synthesize with a cubic perovskite structure,
with a hexagonal structure typically forming instead of the cubic. However, a smaller divalent
cation as in CaNiO3 reduces the Goldschmidt tolerance factor towards the perovskite regime
and may be easier to synthesize. Moreover, cubic SNO in a heterostructure has a possibility
to be stabilized over hexagonal SNO without losing its properties. Some of the computational
results for 4H hexagonal SNO, including the FM configuration of the Ni, the half-metallicity
and the PDOS of Ni and O are analogous to those found in our study of cubic SNO. This
work demonstrates the importance of ligand holes in oxides with formally high valence Ni,
and motivates further growth studies and characterization of the electronic and magnetic
properties of nickelates with divalent A-site cations.
140
7.8 Appendices
Figure 7.A1: PDOS of Ni and O in FM SNO3 and SNO2.875 using HSE06. For SNO3, µNi =
1.079 µB and µO = –0.336 µB. For SNO2.875, average µNi = 1.112 µB and µO = –0.285 µB.
A 2 × 2 × 2 Monkhorst-Pack k-mesh was used.
141
Figure 7.A2: The structure of (a) 2H and (b) 4H hexagonal SrNiO3, TDOS, Ni 3d and O 2p
states of (c) 2H-SNO and (d) 4H-SNO. 2H-SNO is more stable by 349.13 meV f.u.−1 than
4H-SNO. Ni is nonmagnetic in this structure; all FM and AFM ordering converged into the
same state. HSE calculation of 2H-SNO also gives qualitatively the same results.
142
Figure 7.A3: Total density of states of SrNiO3, SrNiO2.875, SrNiO2.75, CaNiO3, and BaNiO3.
143
Figure 7.A4: (a) Slab structure of (111) cubic SNO with 14 layers of STO, (b) slab structure
of (0001) hexagonal SNO with 13 layers of STO, (c) Ni 3d and O 2p states of selected atoms
in (a) (dashed lines are Ni 3d and O 2p states of bulk SNO for reference), and (d) Ni 3d and
O 2p states of selected atoms in (b). Magnetic moments of Ni in (111) SNO were 1.414 µB
and (0001) SNO were zero. Because the slab structure of (0001) SNO has one less STO layer
than (111) SNO slab structure, we added the energy of bulk STO to compare their energies.
(111) SNO slab is 2494.35 meV lower than (0001) SNO slab structure.
144
Figure 7.A5: PDOS of Ni (identical to Figure 7.2b), O on a axis (O(x)), b axis (O(y)), and
c axis (O(z), identical to Figure 7.2) for SrNiO2.875. In O(z), the px and py PDOS overlap.
145
Figure 7.A6: Six remaining configurations of SrNiO2.75 in addition to [222] shown in Figure
7.1b.
Figure 7.A7: (a) The structure of SrNiO2.625 used in the calculation and (b) PDOS of selected
atoms in the figure.
146
Table 7.A1: Relative energies (E) per formula unit (f.u.), lattice parameters, and octahedral
tilts of δ = 0 and δ = 0.125 SNO, depending on the magnetic configurations. Glazer notation
has been used to describe octahedral distortions.
Config. E (meV f.u.−1) a (Å) b (Å) c (Å) Octahedral distortion
δ = 0 FM 0 3.890 a0a0a0
G-AFM 539.48 3.881 a0a0a0
A-AFM 205.59 3.881 3.878 a0a0c0
C-AFM 316.85 3.899 3.891 a0a0c0
δ = 0.125 FM 0 3.881 3.902 a−a−c0
G-AFM 401.44 3.888 3.907 a−a−c0
A-AFM 135.50 3.897 3.886 3.866 a0b−c−
A-AFM 149.93 3.871 3.918 a−a−c0
C-AFM 299.93 3.896 3.915 a−a−c0
C-AFM 270.56 3.914 3.894 3.881 a0b−c−
Table 7.A2: Area fraction eg/t2g derived by PDOS integration performed using the trape-
zoidal integration function (trapz) in MATLAB.
eg/t2g
δ = 0 34.7 %
δ = 0.125 35.3 %
δ = 0.250 Ni(Oct) 35.4 %
δ = 0.250 Ni(Pyr) 35.4 %
147
Chapter 8
Effect of cation defects in iron garnets
8.1 Introduction
Iron garnets have the composition of R3Fe5O12, where R can consist of Y or rare earth.
It has a complicated unit cell with 8 formula units in total, but each formula unit, three
tetrahedral Fe3+ sites and two octahedral Fe3+ sites couple antiferromagnetically, resulting
in a ferrimagnetic system. The dodecahedral R sites tend to align parallel to the octahedral
sites. The balance between two different sublattices impacts the magnetization, as well as
the Curie and the compensation temperatures. Iron garnet is an attractive material system
for research in spintronics applications because one can utilize different magnetic anisotropy
and systematically change or mix the R sites.53 This chapter will focus on how point defects
in dodecahedral sites can be realized in iron garnet thin films and how they can be used to
alter the material characteristics.
148
Figure 8.1: XRD of (a) Ca0.3Y3Fe4.7O12 grown on (111) GGG, (b) Ca0.6Y2.4Fe5O12 grown
on (111) GGG, (c) Ca0.3Y2.7Fe5O12 grown on (111) GGG, and (d) Ca0.3Y2.7Fe5O12 grown on
(001) GGG. No secondary phases are present throughout the entire scan range. The panel
on the right shows an enlarged plot around the selected substrate reflection.
8.2 Aliovalent substitution induced material property change
in yttrium iron garnets
Rare earth iron garnets are typically insulators; however, conductive and magnetic gar-
nets possess a number of advantages that motivate the development of such thin films. There
have been reports of Ca2+ or Si4+ doping in yttrium iron garnets (YIG, Y3Fe5O12) in order to
induce mixed valence states of Fe and increase the number of charge carriers.302–304 Doping
aliovalent Ca2+ results in a mixture of Fe3+ and Fe4+, corresponding to p-type conduction,
whereas it is n-type conduction when Fe3+ and Fe2+ coexist. Since oxygen vacancies can
149
act as donors, the former mechanism may compensate for increasing oxygen deficiencies in
the film. This section will demonstrate the effect of incorporating Ca2+ into YIG (CaYIG
in short).
Figure 8.2: In-plane (IP) and out-of-plane (OP) M-H loops of (a) Ca0.3Y2.7Fe5O12 grown on
(001) GGG, (b) Ca0.3Y3Fe4.7O12 grown on (111) GGG (OP loop is omitted because it did
not saturate before the paramagnetic background from GGG became dominant), and (c)
Ca0.3Y3Fe4.7O12 grown on (111) GSGG. All curves were measured at room temperature.
CaYIG can be grown with various Ca:Y ratio, either by codeposition of “Ca3Fe5O12” (a
mixture of CaFe2O4 and Ca2Fe2O5) and YIG or using a single target, without secondary
phases on both (001) and (111) orientations of Gd3Ga5O12 (GGG) substrates (Figure 8.1).
The ionic radii difference between Ca2+ and Y3+, mixed valence state of Fe3+ and Fe4+,
and oxygen deficiency in the film collectively affect the lattice parameter. The film shows
as low as ∼0.5 nm RMS roughness, especially when it is grown at lower oxygen pressure
(50 mTorr). The surface is smooth, and the in-plane epitaxial strain is preserved enough to
result in the same structural and magnetic properties when a second rare earth garnet layer
is grown consecutively.
By substituting Y3+ by Ca2+, the resistivity decreased down to the semiconducting
regime. The saturation magnetization at room temperature is around 100-120 kA m−1,
which has decreased about 14-29 % from that of YIG. As illustrated in Figure 8.2, while
YIG typically does not exhibit perpendicular magnetic anisotropy (PMA) on GGG sub-
strates, CaYIG on Sc-doped GGG (Gd3Sc2Ga3O12, GSGG) exhibits PMA with a interme-
diate anisotropy in-plane. This can be attributed to the coexistence of mixed valence states
of Fe and enhanced magnetoelasticity. Magnetization vs. temperature measurements and
150
molecular field coefficient (MFC) modeling305 will provide more insight into determining the
site occupancy of antisites and vacancies within these films.
8.3 Density functional theory study of yttrium iron gar-
nets with antisite defects
This section is based on a publication which the author contributed to and coauthored in
Physical Review Materials (2021).19
Antisite defects influence the properties of not only perovskites or orthoferrites but also
iron garnets. Y-rich YIG with Y:Fe ratio of 1:1, deviated from the nominal Y:Fe = 3:5, can
be grown epitaxially on GGG substrates albeit the non-bulk stoichiometry. The Y antisites
modify the magnetization and Curie temperature of YIG, and this section particularly focuses
on the energetics and changes in structural and magnetic properties obtained from density
functional theory (DFT) calculations.
Excess Y in the single-phase garnet films may be incorporated within the structure as
antisite defects where Y occupies a Fe site or via the presence of Fe and O vacancies. Prior
modeling work17 indicates that antisite defects YFe, FeY and oxygen vacancies VO have lower
formation energies than cation vacancies VFe, VY or Frenkel defects, and YFe is therefore
likely to be the dominant defect in the Y-rich YIG. We modeled the magnetic and structural
effects of YFe by replacing one Fe with one Y in the YIG unit cell (8 f.u., Figure 8.3a)
yielding Y/Fe = 0.641. The results show that the unit cell volume expanded for Y-rich
YIG by 0.72 % and 0.74 % for octahedral (YFe,a) and tetrahedral (YFe,d) antisite defects,
respectively. Compared with stoichiometric YIG (Figure 8.3b), Y antisite defects induce
structural distortion (Figures 8.3c and 8.3d). The Y—O bond length of the dodecahedral
site decreases by a small amount near the defect in common for both YFe,a and YFe,d. For
YFe,a, the Fe—O bond length of the O shared with the tetrahedral site is slightly extended,
whereas for YFe,d, the Fe—O bond length for the O shared with the octahedral site is slightly
151
contracted. Here, YFe,a is more stable than YFe,d by 133 meV f.u.−1, indicating octahedral
site preference for Y to occupy the larger octahedral sites, which may be expected from the
greater ionic radius of Y3+ (110 pm) than Fe3+ (65 pm). This differs from prior work17
which predicted a lower formation energy for YFe,d than the YFe,a.
In stoichiometric YIG, the strongest superexchange coupling is antiferromagnetic between
the tetrahedral and octahedral site Fe3+. With three tetrahedral sites and two octahedral
sites Fe3+ per formula unit, the material is ferrimagnetic with a net moment corresponding
to one Fe3+, i.e., 5 µB f.u.−1 at low temperatures. Because Y3+ has almost zero moment,
the antisite YFe is equivalent to removing an Fe3+ moment, which increases (decreases) the
magnetization by 5 µB f.u.−1 when it is in the octahedral (tetrahedral) site. The effect of
antisite defects on the magnetism agrees with (17 ).
Figure 8.3: (a) Unit cell of stoichiometric YIG. Green, gold, and purple correspond to
dodecahedral, tetrahedral, and octahedral sites, respectively. Enlarged structure of YIG
with (b) no defects, (c) a YFe,a antisite defect, and (d) a YFe,d antisite defect. Insets are
bond lengths in angstroms.
The replacement of Fe3+ with Y3+ leads to small changes in the Y and O moments as
well. Particularly, Y3+ in a or d sites has slightly larger moments than Y3+ in c-sites, and
the small moment of YFe,a (YFe,d) is parallel to that of tetrahedral site (octahedral site) Fe3+.
Average moments of O increase (decrease) with YFe,a (YFe,d). However, these changes in the
Y and O moments are minor, and the main contribution to the magnetization is from the
removal of Fe3+.
152
8.4 Direct observation of antisite defects in terbium iron
garnets
This section is based on a publication which the author contributed to and coauthored in
Small (2023).18
Figure 8.4: Atomic resolution HAADF STEM images and STEM EDS elemental mappings
of Tb and Fe, viewed along [111] zone axis (a) and [110] zone axis (b). Arrows indicate
the location of Tb antisites. The legend at the top identifies the sites in the crystal models
superposed on the HAADF images. The intensity profiles in the rightmost panel of (a) show
the contrast seen when antisite defects are present (Line 1) versus absent (Line 2).
Excess Tb in Tb-rich terbium iron garnets (TbIG) can be incorporated as antisite defects,
with the majority in octahedral sites because of the larger ionic radius, accompanied by a
minute quantity of Fe vacancies. It is important to reiterate how these cation antisite defects
can affect the structure and magnetism of complex oxides, realizing distinctive properties
of thin films compared to bulk. This section provides direct evidence of Tb antisite defects
in TbIG via scanning transmission electron microscopy (STEM) energy dispersive X-ray
spectroscopy (EDS) measurements.
A plan view TEM sample of the TbIG thin film on (111) GGG was prepared via me-
153
chanical polishing and Ar+ ion milling. Figure 8.4a depicts the atomic resolution high angle
annular dark field (HAADF) STEM image and EDS elemental maps of Tb and Fe along the
[111] projection, normal to the film plane. Along this zone axis, the columns of dodecahe-
dral, octahedral, and tetrahedral sites are separately visible. We use the non-overlapping
Tb M and Fe L peaks in the X-ray spectrum to obtain the EDS elemental maps in Fig-
ure 8.4. The Tb elemental map shows evidence of Tb present in the octahedral Fe atomic
columns as marked by arrows. From the EDS intensity, a larger amount of Tb antisites are
found to occupy the octahedral sites in the center of six Tb atom columns as compared to
the octahedral sites in between three Tb columns. However, identification of Tb antisites
in tetrahedral sites is less straightforward in the [111] projection, because the tetrahedral
Fe atomic columns appear close to the neighboring Tb atomic columns. To investigate the
presence of tetrahedral Tb antisites, we tilted the sample to align the beam along a [110]
zone axis. STEM EDS measurements performed along the [110] projection reveal the occu-
pation of Tb in one of the tetrahedral Fe columns marked by an arrow in Figure 8.4b. These
results clearly indicate that Tb preferentially occupies the octahedral atomic sites versus the
tetrahedral sites, as expected from the larger size of the octahedral interstice.
8.5 Conclusion
To summarize, cation substitution and rare earth antisite defects can be utilized to tune
the material property of iron garnet thin films. Aliovalent substitution simultaneously al-
ter the structure, electrical conductivity, and magnetism. Larger R antisites preferentially
occupy octahedral Fe sites, predicted according to the energetics from DFT calculation and
observed directly via atomic resolution STEM imaging. On top of the discussions through
Chapters 3 and 7, these findings highlight how point defects are crucial in tailoring desired
properties in other complex oxides as well. Furthermore, it is important to note that the
non-bulk composition can be stabilized due to epitaxy in thin films.
154
Chapter 9
Conclusion
In this thesis, the structural, magnetic, and ferroelectric property evolution upon engi-
neered atomic defects of cation antisites and oxygen vacancies and the importance of con-
trolling the pulsed laser deposition (PLD) growth kinetics were investigated. As discussed in
Chapter 1, defect engineering of materials provides a framework for provoking advantageous
functionalities such as multiferroicity and magnetoelectric effects that are crucial for next-
generation energy-efficient memory and logic applications. Accompanied by the advanced
characterization methods elaborated in Chapter 2, this thesis provides a thorough analysis
of point defects in epitaxial thin films comprised of perovskite structured LuFeO3 (LFO),
SrCo1−xFexO3−δ (SCFO), their self-assembled nanocomposites, or iron garnets, as well as a
theoretical study of magnetism in cubic perovskite SrNiO3−δ related to oxygen ligand holes.
Detailed discoveries and achievements are summarized as follows.
Chapter 3 characterizes the structural, ferroelectric and magnetic properties of epitaxial
LFO thin films with Lu/Fe ratio ranging from 0.6 to 1.5. LFO is antiferromagnetic with
a weak canted moment perpendicular to the film plane. PFM imaging and switching spec-
troscopy reveal room temperature ferroelectricity in Lu-rich and Fe-rich films, whereas the
stoichiometric film shows little polarization. Ferroelectricity in Lu-rich films is present for
a range of deposition conditions and crystallographic orientations. PUND measurement on
155
a Lu-rich film yields ∼13 µC cm−2 of switchable polarization, although its electrical leak-
age needs to be improved for it to be a useful ferroelectric. The ferroelectric response is
attributed to antisite defects analogous to that of Y-rich YFO, yielding multiferroicity via
defect engineering of a rare earth orthoferrite.
Chapter 4 discusses the synthesis of nanocomposite thin films comprising antiferromag-
netic and ferroelectric LFO and ferrimagnetic CoxFe3−xO4 (CFO) are synthesized by a code-
position method using PLD from two or three targets of different composition. The phases
that form (perovskite, spinel, and rocksalt), their compositions, and the magnetic anisotropy
of the nanocomposites are strongly influenced by the selection of targets from the group of
Lu-rich LFO, Fe-rich LFO, CoFe2O4, and Co3O4. When the Co and Fe are delivered from
different targets, the spinel crystals have a range of compositions. Owing to the crystal-
lographic relationship between the perovskite LFO and spinel CFO, a negative exchange
bias of −5.3 mT at room temperature is demonstrated upon field cooling a 31 nm thick
LuFeO3-Co1.2Fe1.8O4 film.
Chapter 5 describes the growth of SCFO forming either a single-phase or a two-phase
self-assembled nanocomposite film depending on PO2. A high PO2 of 150 mTorr during
growth promotes phase separation of SCFO into a nanocomposite comprising CoOx pillars
embedded epitaxially in a Fe-rich SCFO matrix made up of brownmillerite and oxygen-
deficient perovskite, despite the average composition of Sr/(Co+Fe) = 1. The SCFO matrix
in the nanocomposite consists of a brownmillerite structure at high Co content and a layered
perovskite at high Fe content. In contrast, single-phase SCFO films grow at 20 mTorr. Elec-
trolyte gating of the nanocomposites oxidizes brownmillerite into perovskite with interlayer
defects, and renders the matrix phase magnetic with a saturation magnetization of 200 kA
m−1 at 173 K when x = 0.30.
Chapter 6 elucidates the unconventional layering of oxygen ions in SCFO thin film grown
epitaxially on STO at PO2 = 150 mTorr. STEM reveals alternating layers of oxygen deficiency
along the growth direction, with the oxygen-rich layer correlated with the neighboring Co,Fe-
156
site intensity, and contraction of the Sr—Sr distance. DFT calculations and STEM image
simulations support the emergence of periodic (Co,Fe)O6 and (Co,Fe)O4/(Co,Fe)O5 layers,
an ordering which is also sensitive to the Co:Fe ratio.
Chapter 7 presents a theoretical work on the electronic and magnetic properties of per-
ovskite SrNiO3−δ with an oxygen deficiency δ up to 0.375 using DFT. Because of the strong
covalency and negative charge transfer energetics, the structure is predicted to exhibit lig-
and holes L, with Ni present as d8L2 or d7L and significant magnetic moment at the oxygen
sites. The ground state for δ = 0 to 0.375 consists of ferromagnetically ordered Ni with
the Ni and O coupled antiferromagnetically, and the removal of oxygen increases the net
magnetization. These behaviors are also predicted for other A-site cations such as Ca and
Ba. This chapter demonstrates the importance of ligand holes in oxides with formally high
valence Ni, including their influence on the magnetic properties.
Chapter 8 summarizes the feasibility and outcomes of cation defect engineering in mate-
rials with garnet crystal structures. Nontrivial cation composition can be achieved in vari-
ous iron garnet systems such as Ca-substituted Y3Fe5O12−δ, Y-rich Y3Fe5O12, and Tb-rich
Tb3Fe5O12. The substitution and off-stoichiometry of cations affect not only the structural
property but also the electrical and magnetic behavior. Epitaxial stabilization in thin films
is crucial in producing films with unusual cation compositions, allowing one to engineer
properties that deviate from bulk.
There exists a significant amount of potential work to be done in the field of defect engi-
neering of complex oxides, mainly divided into three aspects as an outlook of this thesis: (1)
Improving ferroelectricity and magnetism of orthoferrites towards application. The effect of
antisite defects on the magnetism of orthoferrites has yet to be explored, which can initiated
by the molecular field coefficient modeling analogous to rare earth iron garnets.305 ,306 It
might be possible to control the magnetic anisotropy of ferroelectric orthoferrites by sub-
stituting with different rare earths or applying shear strain through substrates from lower
symmetry (such as (110) or (111)) and mismatch between the cubic vs. orthorhombic sym-
157
metry. Antiferromagnets have been on the spot for spintronics lately,307 ,308 and it will be
meaningful if the magnon transport can be correlated with polarization. (2) Expanding the
scope of cation site occupancy engineering. The antisite defect mechanism of ferroelectricity
has only been demonstrated in orthoferrites so far. However, other perovskite systems, such
as nickelates and aluminates, or rare earth iron garnets, all have the potential to be subject
to a local distortion and symmetry breaking upon introducing antisite defects, which will
result in a more diverse family of multiferroics. Moreover, in line with the rare earth substi-
tution mentioned previously, designing a solid solution of different B-site transition metals
based on the Goodenough-Kanamori rule299 ,300 will allow one to develop multiferroic mate-
rials with desired magnetic properties. (3) Utilizing decent interfacial quality of PLD-grown
thin films. The author first demonstrated the feasibility of growing smooth trilayered oxide
thin films using the lab’s PLD system, which then has been revealed to be applicable in
different perovskite and garnet multilayers. This allows one to investigate interfacial effects
in heterostructures further than exchange bias, for example, charge transfer between the
layers and how it can contribute to multiferroicity in a different approach.
158
References
1. N. A. Spaldin et al., Nat. Mater. 18, 203–212 (2019).
2. S. Manipatruni et al., Nature 565, 35–42 (2019).
3. J. Kreisel et al., Europhys. News 40, 17–20 (2009).
4. G. F. Dionne, Magnetic Oxides (Springer, 2009).
5. I. Dzyaloshinsky, J. Phys. Chem. Solids. 4, 241–255 (1958).
6. T. Moriya, Phys. Rev. 120, 91–98 (1960).
7. E. Bousquet et al., J. Phys. Condens. Matter 28, 123001 (2016).
8. D. Khomskii, Physics 2, 20 (2009).
9. Y. Tokura et al., Reports Prog. Phys. 77, 076501 (2014).
10. Y. J. Choi et al., Phys. Rev. Lett. 100, 047601 (2008).
11. Y. Tokunaga et al., Nat. Phys. 8, 838–844 (2012).
12. K. F. Wang et al., Adv. Phys. 58, 321–448 (2009).
13. H. Zheng et al., Science 303, 661–663 (2004).
14. F. A. Kroger et al., J. Phys. Chem. Solids. 5, 208–223 (1958).
15. A. Sutka et al., Phys. Status Solidi A 210, 1892–1897 (2013).
16. F. M. Chiabrera et al., JPhys Energy 4 (2022).
17. S. J. Tan et al., J. Appl. Phys. 128, 183904 (2020).
18. E. Rosenberg et al., Small 19, 2300824 (2023).
19. T. Su et al., Phys. Rev. Mater. 5, 094403 (2021).
20. S. Ning et al., Nat. Commun. 12, 4298 (2021).
21. S. Kumari et al., Sci. Rep. 10, 3659 (2020).
22. B. Raveau et al., in Cobalt Oxides: From Crystal Chemistry to Physics (Wiley,
2012), pp. 129–178.
23. N. Lu et al., Nature 546, 124–128 (2017).
24. R. He et al., ACS Appl. Mater. Interfaces 12, 45602–45610 (2020).
25. J. L. Lin et al., Acta Mater. 199, 9–18 (2020).
26. H. Jeen et al., Adv. Mater. 25, 3651–3656 (2013).
159
27. A. Khare et al., Adv. Mater. 29, 1606566 (2017).
28. M. S. Saleem et al., ACS Appl. Mater. Interfaces 11, 6581–6588 (2019).
29. K. T. Kang et al., Appl. Phys. Lett. 117, 151601 (2020).
30. M. Hoffmann et al., Phys. Rev. B 92, 094427 (2015).
31. B. Cui et al., J. Phys. Condens. Matter 32, 344001 (2020).
32. K. Yamauchi et al., J. Phys. Condens. Matter 26, 103201 (2014).
33. S. M. Disseler et al., Phys. Rev. Lett. 114, 217602 (2015).
34. T. Kimura et al., Nature 426, 55–58 (2003).
35. H. Seki et al., Angew. Chem. Int. Ed. 55, 1360–1363 (2016).
36. S. Majumdar et al., J. Phys. D Appl. Phys. 47, 034010 (2014).
37. D. Sando et al., Nat. Mater. 12, 641–646 (2013).
38. Y. Jiang et al., Small Methods 7, 2201246 (2023).
39. R. Eason, Pulsed laser deposition of thin films (Wiley, 2006).
40. T. Ohnishi et al., Applied Physics Letters 87, 241919 (2005).
41. M. De Graef et al., Structure of Materials: An Introduction to Crystallography,
Diffraction and Symmetry (Cambridge University Press, 2012).
42. S. Jesse et al., Appl. Phys. Lett. 88, 062908 (2006).
43. A. Gannepalli et al., Nanotechnology 22, 355705 (2011).
44. A. Gruverman et al., Nat. Commun. 10, 1661 (2019).
45. H. J. Kim et al., Nanoscale 8, 1383–1389 (2016).
46. J. F. Scott, J. Condens. Matter Phys. 20, 021001 (2008).
47. H. Naganuma et al., Appl. Phys. Express 1, 061601 (2008).
48. M. Buchner et al., J. Appl. Phys. 124, 161101 (2018).
49. M. Varela et al., Annu. Rev. Mater. Res. 35, 539–569 (2005).
50. C. B. Carter et al., Eds., Transmission Electron Microscopy: Diffraction, Imaging,
and Spectrometry (Springer, 2016).
51. J. J. Liu, Microsc. Microanal. 27, 943–995 (2021).
52. E. Cho et al., Adv. Electron. Mater. 9, 2300059 (2023).
53. E. R. Rosenberg et al., Adv. Electron. Mater. 7, 2100452 (2021).
54. J. D. Groot et al., Phys. Rev. Lett. 108, 187601 (2012).
55. G. R. Hoogeboom et al., Phys. Rev. B 103, 134406 (2021).
56. T. Fakhrul et al., Adv. Opt. Mater. 9, 2100512 (2021).
57. T. Arakawa et al., Mater. Res. Bull. 16, 97–103 (1981).
58. M. D. Scafetta et al., Phys. Chem. Chem. Phys. 19, 10371–10376 (2017).
160
59. R. Gao et al., Chem. Mater. 29, 6544–6551 (2017).
60. W. Noun et al., Phys. Rev. B 81, 054411 (2010).
61. J. Santiso et al., ChemNanoMat 9, e202200495 (2022).
62. J. S. Zhou et al., Phys. Rev. B 102, 104420 (2020).
63. R. L. White, J. Appl. Phys. 40, 1061–1069 (1969).
64. Y. Tokunaga et al., Nat. Mater. 8, 558–562 (2009).
65. A. Kumar et al., Adv. Funct. Mater. 32, 2107017 (2021).
66. A. Kumar et al., Adv. Electron. Mater. 8, 2200036 (2022).
67. U. Chowdhury et al., Appl. Phys. Lett. 105, 052911 (2014).
68. U. Chowdhury et al., Appl. Phys. Lett. 109, 092902 (2016).
69. U. Chowdhury et al., Phys. Rev. B 100, 195116 (2019).
70. Y. Ahn et al., Appl. Phys. A 122, 531 (2016).
71. X. Yuan et al., J. Appl. Phys. 111, 053911 (2012).
72. P. Barrozo et al., Adv. Mater. 32, 2000508 (2020).
73. S. Song et al., Adv. Mater. 28, 7430–7435 (2016).
74. T. Katayama et al., Thin Solid Films 642, 41–44 (2017).
75. D. V. Mamedov et al., J. Phys. Conf. Ser. 903, 012014 (2017).
76. R. C. Rai et al., 264, 209–216 (2018).
77. M. S. V. Kumar et al., J. Mater. Res. 23, 2996–3005 (2008).
78. M. Davis et al., J. Appl. Phys. 101, 054112 (2007).
79. J. L. Zhao et al., Phys. B Condens. Matter 407, 2258–2261 (2012).
80. K. P. Kelley et al., Phys. Rev. Mater. 4, 024407 (2020).
81. D. A. Kiselev et al., Thin Solid Films 619, 214–219 (2016).
82. Z. Gao et al., J. Semicond. 43, 014102 (2022).
83. S. Saremi et al., Adv. Mater. 28, 10750–10756 (2016).
84. X. Qi et al., Appl. Phys. Lett. 86, 062903 (2005).
85. S. Y. Yang et al., Appl. Phys. Lett. 95, 062909 (2009).
86. H. Matsuo et al., J. Appl. Phys. 118, 114101 (2015).
87. H. Matsuo et al., Appl. Phys. Lett. 108, 032901 (2016).
88. Y. Noguchi et al., Sci. Rep. 9, 4225 (2019).
89. M. Kohli et al., Appl. Phys. Lett. 72, 3217–3219 (1998).
90. J. R. Hayes et al., J. Phys. Condens. Matter 23, 465502 (2021).
91. M. Marezio et al., Acta Crystallogr. B 26, 2008–2022 (1970).
92. S. Cao et al., J. Phys. Condens. Matter 28, 156001 (2016).
161
93. M. Ghidini et al., Adv. Electron. Mater. 8, 2200162 (2022).
94. F. Nolting et al., Nature 405, 767–769 (2000).
95. G. Kresse et al., Comput. Mater. Sci. 6, 15–50 (1996).
96. G. Kresse et al., Phys. Rev. B 59, 1758–1775 (1999).
97. J. P. Perdew et al., Phys. Rev. Lett. 77, 3865–3868 (1996).
98. L. Wang et al., Phys. Rev. B 73, 195107 (2006).
99. A. Jain et al., Phys. Rev. B 84, 045115 (2011).
100. O. Kubaschewski et al., Materials thermochemistry. 6th ed., rev.enl. (Pergamon
Press, 1993).
101. E. Cho et al., ACS Appl. Electron. Mater. (2024).
102. O. J. Lee et al., APL Mater. 9, 030904 (2021).
103. D. H. Kim et al., J. Mater. Chem. C 7, 9128–9148 (2019).
104. S. Misra et al., Mater. Horizons 8, 869–884 (2021).
105. D. M. Cunha et al., MRS Bull. 46, 152–158 (2021).
106. Z. Mo et al., APL Mater. 9, 041109 (2021).
107. W. H. Meiklejohn et al., Phys. Rev. 102, 1413–1414 (1956).
108. M. Huijben et al., Adv. Mater. 25, 4739–4745 (2013).
109. F. K. Olsen et al., Phys. Rev. B 99, 134411 (2019).
110. W. Zhang et al., Nanoscale 7, 13808–13815 (2015).
111. Y. Zhang et al., Phys. Chem. Chem. Phys. 17, 27136–27144 (2015).
112. E. M. Choi et al., Appl. Phys. Lett. 102, 012905 (2013).
113. R. Wu et al., ACS Appl. Mater. Interfaces 10, 42593–42602 (2018).
114. M. Saghayezhian et al., Phys. Rev. Res. 1, 033160 (2019).
115. J. Scola et al., J. Appl. Phys. 110, 043928 (2011).
116. G. Tian et al., Adv. Electron. Mater. 5, 1900012 (2019).
117. N. M. Aimon et al., Appl. Phys. Lett. 100, 092901 (2012).
118. D. H. Kim et al., Adv. Funct. Mater. 25, 3091–3100 (2015).
119. E. Cho et al., ACS Appl. Nano Mater. 5, 14646–14653 (2022).
120. P. Abellán et al., Appl. Phys. Lett. 98, 041903 (2011).
121. C. Moreno et al., Adv. Funct. Mater. 19, 2139–2146 (2009).
122. X. Chen et al., ACS Appl. Electron. Mater. 2, 3964–3970 (2020).
123. J. H. Kim et al., J. Appl. Phys. 92, 6066–6071 (2002).
124. H. E. Swanson, Standard X-ray diffraction powder patterns (US Department of
Commerce, National Bureau of Standards, 1967), vol. 25.
162
125. H. E. Swanson, Standard X-ray diffraction powder patterns (US Department of
Commerce, National Bureau of Standards, 1971), vol. 25.
126. X. Liu et al., Phys. Chem. Miner. 17, 168–172 (1990).
127. S. Sasaki et al., Proc. Jpn. Acad. Ser. B 55, 43–48 (1979).
128. N. M. Aimon et al., ACS Appl. Mater. Interfaces 7, 2263–2268 (2015).
129. C. Zhang et al., Appl. Phys. Lett. 113, 223105 (2018).
130. D. H. Kim et al., J. Appl. Phys. 121, 163902 (2017).
131. J. R. Hayes et al., J. Phys. Condens. Matter 23, 465502 (2011).
132. Y. Wei et al., Chem. Mater. 35, 4461–4470 (2023).
133. W. Gao et al., Nature 575, 480–484 (2019).
134. B. Rivas-Murias et al., Nano Lett. 23, 1688–1695 (2023).
135. D. Barreca et al., CrystEngComm 12, 2185–2197 (2010).
136. M. Varela et al., Phys. Rev. B 79, 085117 (2009).
137. R. C. O’Handley, Modern Magnetic Materials: Principles and Applications (Wiley,
2000).
138. J. Tian et al., Adv. Mater. 31, 1903679 (2019).
139. A. F. Jr. et al., J. Magn. Magn. Mater. 320, 709–713 (2008).
140. P. Lv et al., J. Appl. Phys. 111, 013910 (2012).
141. P. W. Huang et al., J. Phys. Condens. Matter 23, 415801 (2011).
142. S. Dong et al., Phys. Rev. Lett. 103, 127201 (2009).
143. N. Grunbaum et al., J. Solid State Chem. 177, 2350–2357 (2004).
144. B. Ma et al., J. Am. Ceram. Soc. 85, 2641–2645 (2002).
145. H. Jeen et al., Nat. Mater. 12, 1057–1063 (2013).
146. S. K. Acharya et al., ACS Appl. Mater. Interfaces 8, 7902–7911 (2016).
147. S. Ning et al., ACS Nano 14, 8949–8957 (2020).
148. C. Ge et al., Adv. Mater. 31, 1900379 (2019).
149. H. B. Li et al., Adv. Sci. 6, 1901432 (2019).
150. H. Taguchi et al., J. Solid State Chem. 29, 221–225 (1979).
151. J. B. MacChesney et al., J. Chem. Phys. 43, 1907–1913 (1965).
152. H. Takeda et al., J. Phys. Soc. Jpn. 33, 970–972 (1972).
153. J. C. Grenier et al., J. Solid State Chem. 58, 243–252 (1985).
154. J. Ma et al., Adv. Mater. 23, 1062–1087 (2011).
155. A. Chen et al., Adv. Mater. 31, 1803241 (2019).
156. J. Huang et al., MRS Bull. 46, 159–167 (2021).
163
157. N. Dix et al., Appl. Phys. Lett. 95, 062907 (2009).
158. D. H. Kim et al., Adv. Funct. Mater. 24, 2334–2342 (2014).
159. Q. Zhan et al., Appl. Phys. Lett. 89, 172902 (2006).
160. Y. Yang et al., J. Mater. Chem. 19, 4998–5002 (2009).
161. Y. Chen et al., CrystEngComm 18, 7745–7752 (2016).
162. T. Fix et al., Nano Lett. 13, 5886–5890 (2013).
163. J. MacManus-Driscoll et al., APL Mater. 3, 062507 (2015).
164. Y. Wang et al., Sci. Rep. 7, 11122 (2017).
165. J. Huang et al., Nanoscale 9, 7970–7976 (2017).
166. A. P. Chen et al., J. Appl. Phys. 114, 124101 (2013).
167. S. Lafane et al., Appl. Phys. A 112, 159–164 (2013).
168. C. Aruta et al., Appl. Phys. Lett. 97, 252105 (2010).
169. Y. Wang et al., Phys. Rev. X 10, 021030 (2020).
170. S. Hu et al., Chem. Mater. 29, 708–717 (2017).
171. J. Chen et al., J. Phys. D: Appl. Phys. 49, 045201 (2015).
172. E. Cho et al., Adv. Mater. Interfaces 11, 2300807 (2024).
173. J. Hoffmann et al., J. Mater. Res. 27, 1462–1470 (2012).
174. A. Fossdal et al., Solid State Ion. 143, 367–377 (2001).
175. D. Fuchs et al., J. Appl. Phys. 88, 1844–1850 (2000).
176. L. Qiao et al., Phys. Rev. B 83, 085408 (2011).
177. R. J. D. Tilley, J. Solid State Chem. 21, 293–301 (1977).
178. D. I. Khomskii, J. Magn. Magn. Mater. 306, 1–8 (2006).
179. F. Aziz et al., J. Magn. Magn. Mater. 517, 167338 (2021).
180. S. Bao et al., Sci. Rep. 4, 4726 (2014).
181. M. Moreau et al., Sci. Rep. 7, 4386 (2017).
182. L. Ruiz-Gonzalez et al., Open Inorg. Chem. J. 1, 37–46 (2007).
183. A. Marthinsen et al., MRS Commun. 6, 182–191 (2016).
184. T. Jia et al., Phys. Chem. Chem. Phys. 21, 20454–20462 (2019).
185. S. Y. Smolin et al., Chem. Mater. 28, 97–105 (2016).
186. K. H. L. Zhang et al., ACS Appl. Mater. Interfaces 10, 17480–17486 (2018).
187. C. Bloed et al., ACS Appl. Energy Mater. 2, 6140–6145 (2019).
188. X. K. Wei et al., J. Appl. Phys. 111, 073904 (2012).
189. J. D. Jorgensen et al., Phys. C 153, 578–581 (1988).
190. A. Ourmazd et al., Nature 329, 425–427 (1987).
164
191. C. Bernuy-Lopez et al., Materials 9, 154 (2016).
192. E. L. Rautama et al., Chem. Mater. 21, 102–109 (2009).
193. S. Streule et al., J. Phys. Condens. Matter 17, 3317–3324 (2005).
194. C. Frontera et al., Chem. Mater. 17, 5439–5445 (2005).
195. K. Vidyasagar et al., J. Chem. Soc., Chem. Commun. 1, 7–8 (1985).
196. A. I. Becerro et al., Phys. Chem. Chem. Phys. 2, 3933–3941 (2000).
197. O. Chmaissem et al., Phys. Rev. B 67, 944311 (2003).
198. P. D. Battle et al., J. Solid State Chem. 76, 334–339 (1988).
199. F. Ramezanipour et al., J. Mater. Chem. 22, 9522–9538 (2012).
200. L. Suescun et al., J. Solid State Chem. 180, 1698–1707 (2007).
201. S. Kobayashi et al., Appl. Phys. Lett. 102, 231911 (2013).
202. F. Ramezanipour et al., Inorg. Chem. 50, 7779–7791 (2011).
203. M. Reehuis et al., Phys. Rev. B 85, 184109 (2012).
204. B. C. Tofield et al., Mater. Res. Bull. 10, 737–745 (1975).
205. M. Schmidt et al., J. Phys. Condens. Matter 15, 8691–8701 (2003).
206. C. Cazorla, Phys. Rev. Appl. 7, 044025 (2017).
207. R. L. Toquin et al., J. Am. Chem. Soc. 128, 13161–13174 (2006).
208. J. Fernández Sanjulián et al., Dalt. Trans. 46, 1624–1633 (2017).
209. S. V. Rompaey et al., Z. Kristallogr. Cryst. Mater. 228, 28 (2013).
210. Y. P. Ivanov et al., Commun. Mater. 1, 25 (2020).
211. E. Heifets et al., Phys. Chem. Chem. Phys. 21, 3918–3931 (2019).
212. X. Rui et al., Appl. Phys. Lett. 114, 233101 (2019).
213. J. Walter et al., Phys. Rev. Mater. 4, 091401 (2020).
214. S. Inkinen et al., Phys. Rev. Mater. 4, 046002 (2020).
215. J. Gazquez et al., APL Mater. 1, 012105 (2013).
216. J. P. Hodges et al., J. Solid State Chem. 151, 190–209 (2000).
217. A. M. Abakumov et al., J. Solid State Chem. 183, 2845–2854 (2010).
218. C. Ricca et al., Phys. Rev. Res. 2, 023313 (2020).
219. E. J. Kirkland, Advanced Computing in Electron Microscopy (Springer, 2010).
220. A. Guinier, X-Ray Diffraction: In Crystals, Imperfect Crystals, and Amorphous
Bodies (W.H. Freeman, 1963).
221. E. Cho et al., Phys. Rev. Mater. 5, 094413 (2021).
222. S. Mehmood et al., Mater. Chem. Phys. 196, 222–228 (2017).
223. T. Jia et al., RSC Adv. 7, 38798–38804 (2017).
165
224. S. Yao et al., Phys. Chem. Chem. Phys. 21, 25397–25405 (2019).
225. B. J. Campbell et al., J. Appl. Crystallogr. 39, 607–614 (2006).
226. S. Middey et al., Annu. Rev. Mater. Res. 46, 305–334 (2016).
227. J. Catalano et al., Reports Prog. Phys. 81, 046501 (2018).
228. T. Mizokawa et al., Phys. Rev. B 52, 13865–13873 (1995).
229. M. N. Grisolia et al., Nat. Phys. 12, 484–492 (2016).
230. J. A. Alonso et al., Phys. Rev. Lett. 82, 3871–3874 (1999).
231. M. T. Fernández-Díaz et al., Phys. Rev. B 64, 144417 (2001).
232. M. Hepting et al., Nat. Phys. 14, 1097–1102 (2018).
233. H. Park et al., Phys. Rev. Lett. 109, 156402 (2012).
234. M. Kotiuga et al., Proc. Natl. Acad. Sci. 116, 21992–21997 (2019).
235. L. Wang et al., ACS Appl. Mater. Interfaces 8, 9769–9776 (2016).
236. S. Johnston et al., Phys. Rev. Lett. 112, 106404 (2014).
237. G. Coslovich et al., Nat. Commun. 4, 2643 (2013).
238. B.-X. Wang et al., Phys. Rev. Mater. 4, 084409 (2020).
239. D. Li et al., Phys. Rev. Lett. 125, 027001 (2020).
240. M. Osada et al., Nano Lett. 20, 5735–5740 (2020).
241. J. Nanjundaswamy et al., Phys. C 166, 361–371 (1990).
242. E. Ksepko, Int. J. Hydrogen Energy 39, 8126–8137 (2014).
243. Y. Takeda et al., J. Inorg. Nucl. Chem. 34, 1599–1601 (1972).
244. J. J. Lander et al., J. Am. Chem. Soc. 73, 2452–2454 (1951).
245. H. Shibahara, J. Solid State Chem. 69, 81–92 (1987).
246. J. G. Lee et al., Chem. Commun. 52, 10731–10734 (2016).
247. R. Gottschall et al., Inorg. Chem. 37, 1513–1518 (1998).
248. M. Zinkevich, J. Solid State Chem. 178, 2818–2824 (2005).
249. J. Lee et al., J. Solid State Chem. 93, 267–271 (1991).
250. F. Abraham et al., J. Mater. Chem. 4, 1763–1764 (1994).
251. M. Bianchini et al., Chem. Mater. 32, 9211–9227 (2020).
252. V. W. Bronger et al., Z. Anorg. Allg. Chem. 333, 188–200 (1964).
253. A. N. Mansour et al., J. Electrochem. Soc. 141, L69 (1994).
254. A. N. Mansour et al., J. Phys. Chem. A 102, 65–81 (1998).
255. H. Seki et al., J. Jpn. Soc. Powder Powder Metall. 63, 609–612 (2016).
256. L. Wang et al., Adv. Mater. 32, 2005003 (2020).
257. L. Wang et al., Sci. Adv. 7, eabe2866 (2021).
166
258. G. Chen et al., presented at the 2014 International Conference on Mechatronics,
Electronic, Industrial and Control Engineering, ed. by L. Chang et al., pp. 746–749.
259. G. Y. Chen et al., J. Solid State Chem. 233, 438–443 (2016).
260. S. Mehmood et al., J. Electron. Mater. 49, 3780–3790 (2020).
261. F. Fan et al., Phys. Rev. B 94, 214401 (2016).
262. S. Dudarev et al., Phys. Rev. B 57, 1505–1509 (1998).
263. H. J. Monkhorst et al., Phys. Rev. B 13, 5188–5192 (1976).
264. Y. Takeda et al., Acta Cryst. 32, 2464–2466 (1976).
265. R. N. Song et al., Front. Phys. 10, 321–326 (2015).
266. K. Momma et al., J. Appl. Crystallogr. 44, 1272–1276 (2011).
267. M. Yu et al., J. Chem. Phys. 134, 064111 (2011).
268. J. Varignon et al., npj Quantum Mater. 2, 21 (2017).
269. G. M. Dalpian et al., Phys. Rev. B 98, 075135 (2018).
270. A. Mercy et al., Nat. Commun. 8, 1677 (2017).
271. J. Lim et al., Phys. Rev. B 98, 085106 (2018).
272. A. Hampel et al., Phys. Rev. B 96, 165130 (2017).
273. V. Bisogni et al., Nat. Commun. 7, 13017 (2016).
274. S. R. Barman et al., Phys. Rev. B 49, 8475–8478 (1994).
275. T. Mizokawa et al., Phys. Rev. B 61, 11263–11266 (2000).
276. P. C. Rogge et al., Phys. Rev. Mater. 2, 015002 (2018).
277. S. Mathi Jaya et al., Phys. Rev. B. 43, 13274–13279 (1991).
278. C. Lan et al., J. Comput. Electron. 19, 905–909 (2020).
279. R. S. Li et al., Chin. J. Phys. 56, 2829–2836 (2018).
280. H. Chen et al., Phys. Rev. B 85, 115127 (2012).
281. M. L. Medarde, J. Phys. Condens. Matter 9, 1679–1707 (1997).
282. B. Geisler et al., Phys. Rev. B 102, 020502 (2020).
283. J. R. Petrie et al., Adv. Funct. Mater. 26, 1564–1570 (2016).
284. P. Walke et al., J. Phys. Chem. Solids 123, 1–5 (2018).
285. R. Merkle et al., J. Electrochem. Soc. 159, B219 (2011).
286. T. Yokota et al., Vacuum 84, 663–665 (2009).
287. R. K. Hona et al., Polyhedron 167, 69–74 (2019).
288. Q. Ji et al., Energy Environ. Sci. 13, 1408–1428 (2020).
289. T. Das et al., Phys. Chem. Chem. Phys. 22, 9723–9733 (2020).
290. C. Zhang et al., RSC Adv. 3, 12267–12274 (2013).
167
291. M. T. Curnan et al., J. Phys. Chem. C 118, 28776–28790 (2014).
292. A. A. Emery et al., Sci. Data 4, 170153 (2017).
293. H. Y. Su et al., J. Mater. Sci. 50, 1701–1709 (2015).
294. T. Das et al., ECS Trans. 78, 2807–2814 (2017).
295. T. Das et al., J. Mater. Chem. A 5, 25031–25043 (2017).
296. L. Zhang et al., Comput. Mater. Sci. 118, 309–315 (2016).
297. J. M. Florez et al., Appl. Phys. Lett. 100, 252904 (2012).
298. C. J. Bartel et al., Sci. Adv. 5, eaav0693 (2019).
299. J. B. Goodenough, Phys. Rev. 117, 1442 (1960).
300. J. Kanamori, J. Phys. Chem. Solids 10, 87–98 (1959).
301. F. Yan et al., Adv. Funct. Mater. (2024).
302. T. Merceron, Phys. Status Solidi A 15, 539–544 (1973).
303. A. Teurtrie et al., Adv. Funct. Mater. 29, 1904958 (2019).
304. Z. Zhang et al., J. Solid State Chem. 290, 121530 (2020).
305. M. J. Gross et al., Physical Review Applied 21, 014060 (2024).
306. K. M. Watson, Physical Review 135, A1336–A1338 (1964).
307. J. Han et al., Nat. Nanotechnol. 15, 563–568 (2020).
308. S. Das et al., Nat. Commun. 13, 6140 (2022).
168