Professional Documents
Culture Documents
Ball Diss Final
Ball Diss Final
Double Perovskites
DISSERTATION
Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy
in the Graduate School of The Ohio State University
By
Molly Ball
2017
Dissertation Committee:
Molly Ball
2017
Abstract
At present, electronic devices are reaching their storage and processing limit causing a
major push to find materials that can be used in the next generation of devices. Double
perovskites with A2BB′O6 stoichiometry form one of the leading classes of materials
currently being studied as a potential candidate because of their extremely wide range
and tunability of functional properties, along with economic and highly scalable synthesis
routes. Having a thorough understanding of their electronic and magnetic structure and
their dependence on composition and local structure is the basis for targeted development
of novel and optimized double perovskites. While the body of knowledge and rules
within the field of materials chemistry has enabled many previous discoveries, recent
developments within density functional theory (DFT) allow by now a rather realistic
This thesis details computational work based on DFT within several collaborative studies
to better understand the electronic and magnetic properties of double perovskites and
related materials that show promise for future use in multifunctional devices. First, we
will begin with a general introduction to the double perovskite structure, their properties,
and the computational methods used to study them. In the next section, we will look at
ii
measurements showed that the transition metal ions in the two sublattices undergo
coupling and a dominance of longer-range interactions, which has previously not been
between the ions and explain this unique dominance of the long-range interactions.
Then, we will look at studies done on thin films of Sr2CrReO6, where our experimental
that examine the dependence of MCA on strain and could identify orbital magnetism on
In the last section, we introduce double perovskites as novel lead-free halide solar cell
materials, with current focus on Cs2AgBiBr6 and Cs2AgBiCl6. While organic Pb based
halides that can be synthesized without expensive clean rooms have achieved within
record time efficiencies that rival that of traditional semiconductor based materials,
creating quite a buzz within the field of photovoltaics, their Pb content and lacking air
stability represented severe roadblocks towards market introduction. Here, we show with
band structure calculations that spin-orbit coupling is a much more dominant interaction
than in traditional semiconductors and thus needs to be considered when designing novel
materials for maximum efficiency. The results of this study have given momentum to
order to explore the wide and to date little explored composition space of double
iii
perovskites, one of the currently most promising materials classes for novel devices with
iv
Dedication
To Tom Moe.
v
Acknowledgments
I would like to first acknowledge my advisor, Professor Wolfgang Windl, for the
guidance he has given me throughout my graduate school career. His patience and
encouragement have been vital to my success. I thank him for introducing me to the
incredible research opportunities I have been able to be a part of through the Center for
I am grateful for all members in CEM for their interest in my theoretical work and
led to many more successful publications than I would have ever been able to achieve
alone. Thank you to Dr. Oscar Restrepo for his willingness to answer the overwhelming
amount of questions that I asked on a weekly basis and always being accommodating
when I asked him to explain portions of density functional theory yet again. I am grateful
for the time you spent passing your knowledge onto me.
I would like to thank my research group members, both past and present, not only for
their insightful academic discussions, but also for wonderful company both in and out of
the office. I am truly lucky to have been able to spend so much time with such incredible
people. I am grateful for the friendships that I have made while in Columbus both in and
vi
I would finally like to thank my family, who taught me to be my best and encouraged me
to pursue my higher education goals. Their support has been unwavering for my entire
vii
Vita
Publications
viii
4. E. T. McClure, M. R. Ball, W. Windl, P. M. Woodward, Cs2AgBiX6 (X= Br, Cl): New Visible
Light Absorbing, Lead-Free Halide Perovskite Semiconductors, Chemistry of Materials 28 (5),
1348-1354 (2016).
Fields of Study
ix
Table of Contents
Abstract…………………………………………………………………………………... ii
Dedication………………………………………………………………………………... v
Acknowledgements……………………………………………………………………… vi
Vita………………………………………………………………...…………….…….. viii
Table of Contents………………………………………………………………………… x
Chapter 2. Background.……………..…………………………………………………… 3
2.1 Introduction………………………………………………………………....... 3
x
3.1 Introduction…………………………………………………………………. 17
3.4 Summary……………………………………………………………………. 21
4.1 Abstract……………………………………………………………………... 23
4.2 Introduction…………………………………………………………………. 24
4.5 Conclusions…………………………………………………………………. 44
Epitaxial Films..……………………………………………………………………….... 48
Chapter 7. Cs2AgBiX6 (X = Br, Cl) – New Visible Light Absorbing, Lead-free Halide
7.1 Abstract……………………………………………………………………... 84
7.2 Introduction…………………………………………………………………. 85
7.4 Results………………………………………………………………………. 90
xi
7.4.1 Synthesis……………………………..…………………………… 90
Measurements….…………………………………………………………..139
& SCNO/LSAT……………………………………………………………142
B.4 Atomically Resolved Spin and Orbital Magnetic Moments for SCRO Films
References……………………………………………………………………………... 148
xiii
List of Tables
Table 4.1 gives information concerning the unit cell and goodness of fit parameters for
each refinement on the longer neutron data sets. Figures and tables containing further
Table 5.1 Lattice constants (a, c), tetragonal distortion (c/a), coercivity (Hc) and
Table 7.1 details of the X-ray powder diffraction experiments and Rietveld
refinements……………………………………………………………………………… 93
Table 7.2 Refined atomic positions and atomic displacement parameters for Cs2AgBiCl6
Table A.1 Refined atomic positions and displacement parameters for longer measuring
time data sets where applicable. Os atomic positions for both space groups remain at
xiv
Wyckoff site 2a (0,0,0) while Co atomic positions remain at Wyckoff site 2b (0,0,½) for
both space groups. Ueq is defined as one third of the trace of the Uij tensor, and it is
given for the oxygen atoms for which anisotropic displacement parameters were
refined…………………………………………………………………………………. 117
Table A.2 Select bond lengths and angles for Sr2CoOsO6 as obtained from Rietveld
Table A.3 Relative energy of different magnetic configurations examined in this work,
along with the spin and orbital-moments of Co atoms and OsO6 units obtained for each
configuration…………………………………………………………………………... 126
Table A.4 Spin exchange parameters J ijeff = Si S j J ij for Sr2CoOsO6 (in meV)………... 134
Table A.5 Comparison of energy per f.u. of the different magnetic configurations
Table B.1 DFT-calculated rotation angles of the Cr and Re oxygen octahedra as well as
equatorial (deq) and axial (dax) lengths of the Cr–O and Re–O bonds for the Sr2CrReO6
Table B.2 First principles calculated in-plane (average of [100] and [010] and out-of-
plane ([001]) spin and orbital magnetic moments for Cr, Re, and O atoms in SCRO films
Table C.1 An HSE hybrid functional with spin orbit coupling was used for the
computations…………………………………………………………………………... 145
xv
List of Figures
Figure 2.1 Corner sharing BO6 octahedra create a cage in which the A cation sits;
forming the perfect cubic structure for a) a ternary perovskite and b) a double
perovskite………………………………………………………………………………… 6
Figure 2.2 Low values of the tolerance factor t will lower the symmetry of the crystal
structure. GdFeO3 have tilted octahedra and crystallize in the orthorhombic system. If the
tolerance factor t > 1 as for BaNiO3 due to a large A or a small B ion then hexagonal
variants of the perovskite structure form. In LnMnO3 the octahedra around Mn are Jahn–
Teller distorted…………………………………………………………………………… 8
Figure 2.3 Two possible distortions of an octahedron that remove the degeneracy of
unequally occupied eg orbitals for a high spin d4 ion; (right) elongation of the bonds
along z, and (left) elongation of the bonds in the x-y plane. In both instances the
Figure 4.1 The A2MM′O6 double perovskite structure. The large A cations are shown in
gray, while the M and M′ centered octahedra are shown in red and blue respectively… 25
Figure 4.2 Field cooled (filled circles) and zero field cooled (open circles) magnetic
susceptibility with inverse zero field cooled data plotted against the right axis showing
xvi
Figure 4.3 Specific heat of Sr2CoOsO6 as a function of temperature………………….. 32
Figure 4.5 Evolution of the (a) lattice parameters, (b) Os–O bond lengths, and (c) Co–O
diffraction data. The dashed line separates the tetragonal I4/m phase from the monoclinic
I2/m phase………………………………………………………………………………. 34
Figure 4.6 Total- and atom- resolved density of states of Sr2CoOsO6 from a non-collinear
DFT calculation with UCo = 4.1 eV and UOs = 2.1 eV. The contributions of the Os 5d and
Figure 4.7 Rietveld fits to the high d-spacing regions of the neutron diffraction patterns
(130 K), between TN1 and TN2 (80 K), and below TN2 (12 K). Black symbols represent
observed data while the red curves represent the calculated patterns. The hkl values given
Figure 4.8 (a) Experimentally determined magnetic structure of Sr2CoOsO6, with the Os
shown in red and the Co in blue. (b) A schematic of the superexchange pathways between
Co and Os ions with the exchange constants extracted from the DFT calculations. A
Figure 5.1 XRD scans around the Sr2CrReO6 (004) (left panels) and Sr2CrReO6 (022)
(right panels) peaks for 90-nm Sr2CrReO6 films grown on (a) LSAT, (b) SrTiO3 and (c)
Sr2CrNbO6 buffer layer on LSAT. The off-normal Sr2CrReO6 (022) peaks are measured
xvii
Figure 5.2 Schematics of (a) Cr−O (counterclockwise) and Re−O (clockwise) octahedral
rotations in the Sr2CrReO6 lattice and (b) an oxygen (red) octahedron surrounding Cr or
Re with two different bond lengths, deq and dax. (c) Octahedral rotation angle θ and (d)
bond lengths of Cr−O and Re−O octahedra as a function of the tetragonal distortion c/a of
Figure 5.3 In-plane and out-of-plane magnetic hysteresis loops for the Sr2CrReO6 films
grown on (a) LSAT, (b) SrTiO3 and (c) Sr2CrNbO6/LSAT in magnetic fields up to 7 T at
T = 20, 100 and 300 K. Magnetic anisotropy fields can be obtained from the intercept of
linear extrapolation of the hard-axis hysteresis loop and the saturation magnetization as
Figure 5.4 Differences (∆m) between out-of-plane ([001] direction and in plane (average
of [100] and [010] directions (a) total magnetic moments, (b) total orbital and spin
magnetic moments, and (c) spin and (d) orbital magnetic moments for Re (squares), Cr
(diamonds), and O (circles). The lines are guides for the eye.…………........…………. 60
Figure 6.1 In-plane magnetic hysteresis loop at T = 200 K for films on (001)-oriented
LSAT (green circles), SrTiO3 (black squares), and ~200 nm Sr2CrNbO6 buffer layer on
Figure 6.2 Normalized XANES and XMCD spectra at the (a-c) Cr L, (d-f) Re L, and (g-i)
O K edges. XANES/XMCD spectra are on top/bottom within each figure. Spin and
Figure 6.3 Normalized x-ray absorption spectra for different strains at the (a) Cr L2,3
xviii
Figure 7.1 Rietveld refinements of the XRPD patterns for Cs2AgBiCl6 (upper) and
Cs2AgBiBr6 (lower). The black dots, red and green lines are the experimental pattern,
calculated fit, and difference curve, respectively. Red tick marks at the bottom give the
expected peak positions. The inset shows the fit to the (111) and (200) reflections…… 92
Figure 7.2 Refined crystal structure of Cs2AgBiCl6. The Cs+ ions are shown as gray
spheres, the chloride ions as small green spheres, while the Ag and Bi centered octahedra
Figure 7.3 Diffuse reflectance spectra for Cs2AgBiCl6 and CH3NH3PbCl3 (top), and
Figure 7.4 Band Structure diagrams for Cs2AgBiCl6 (top) and cubic CsPbCl3 (bottom).
Figure 7.5 Band Structure diagrams for Cs2AgBiBr6 (top) and cubic CsPbBr3 (bottom).
Figure 7.6 Atomic partial density of states plots for Cs2AgBiCl6 (top) and Cs2AgBiBr6
(bottom)………………………………………………………………………………... 103
Figure 7.7 UV-Vis diffuse reflectance spectra showing the light stability of Cs2AgBiCl6
Figure 7.8 UV-Vis Diffuse spectra showing the light instability of Cs2AgBiBr6 after two
Figure A.1 Refined powder X-ray diffraction pattern of Sr2CoOsO6. Data was collected
monochromator. Black symbols observed data, while red and blue curves indicate the
symbols represent observed data points, red curves for the refined fit and blue for the
difference curve. Pink hash marks indicate reflections from the vanadium sample can,
black tick marks indicate reflections from the nuclear structure, and green tick marks
Figure A.3 Colorfill plot of observed variable temperature NPD data revealing the
splitting of the (220) peak which is associated with the structural phase transition
lowering the symmetry of the structure from tetragonal I4/m to monoclinic I2/m at
Figure A.4 The structure of Sr2CoOsO6 at 12 K in space group I2/m. The view looking
down [001] on the left shows out-of-phase octahedral rotations about the c-axis, while the
view looking down [110] on the right shows the lack of octahedral tilting about the a- or
b-axes (right). The higher temperature I4/m structure has an identical tilting
pattern............................................................................................................................. 120
temperature……………………………………………………………………………. 122
Figure A.6 16 formula-unit cell of Sr2CoOsO6 showing (a) the experimentally observed
magnetic configuration which is non-collinear and (b) the collinear version of the
experimentally observed magnetic configuration where all the moments have been
aligned along the [001] direction. Only Os (red) and Co (blue) atoms are shown for
clarity. The arrows show the direction of the magnetic moment on every Co and Os
ion……………………………………………………………………………………... 125
xx
Figure A.7 Spin- and atom- resolved density of states with collinear arrangement of
independently ordered Co and Os sublattices obtained using (a) GGA without Hubbard-U
and (b) with UCo = 4.1 eV and UOs = 2.1 eV………………………………………….. 125
Figure A.8 Spin- and orbital-resolved density of states of (a) Co and (b) Os in the 16
and Os sublattices obtained using UCo = 4.1 eV and UOs = 2.1 eV…………………… 127
Figure A.9 The total DOS for the 16 f.u. cell with (a) Exp(non_col), (b) Exp(col)+SOC
Figure A.10 Schematic showing the different exchange interactions between Co and Os
ions…………………………………………………………………………………….. 133
Figure A.11 Schematic showing the spin arrangements in two neighboring ab planes in
Sr2CoOsO6…………………………………………………………………………….. 136
Figure B.1 Raw magnetic data for the 90 nm SCRO/LSAT film taken with a Quantum
Design SQUID for both in-plane and out-of-plane geometries at T = 20 K. The hard axis
loop (field applied perpendicular to the film) shows almost no signal differentiating the
SCRO film from the background of the LSAT substrate, so a reliable anisotropy field
Figure B.2 Raw magnetic data for 90-nm SCRO/STO film taken with a Quantum Design
Figure B.3 Calculated DOS of Re t2g (a) xy and (b) yz and xz suborbitals near the Fermi
energy in SCRO for tetragonal distortions c/a = 1.025 and c/a = 0.99………………... 144
xxi
Figure C.1 Kubleka-Munk plots for the silver-bismuth (blue) and lead compounds (red),
with chloride compounds on the left and bromide compounds on the right…………... 146
Figure C.2 XRPD overlays of Cs2AgBiCl6 (upper) and Cs2AgBiBr6 (lower) showing the
light stability of the compounds after a month of exposure to sunlight. The change in the
powder diffractogram for the bromine compound indicates that structural changes are
xxii
Chapter 1. Introductory Note
This dissertation is organized in the form of chapters, which have been written in varying
styles depending on the peer-reviewed journal they were initially published. Since all
work was done in collaboration with other groups involving graduate students, I will
report my work outside of the general introductory chapters and summary in their
problems with overlap with the theses of my experimental collaborators while still being
able to discuss the interplay between experimental and theoretical results that was at the
heart of my research and that enabled the advances in materials understanding and design
described below.
properties associated with the electronic and magnetic structure. In chapter 3, I will
discuss the methods used to study the electronic structures of these materials. I will
give a basic overview of density functional theory (DFT) and review some of the
assumptions that are vital for this work. Chapter 4 is a journal article published in
the Journal of the American Chemical Society and will discuss the unique magnetic
1
observed in Sr2CrReO6 both in verbatim form as published in Physical Review B. In
chapter 7, I present work done on the lead-free halides, Cs2AgBiBr6 and Cs2AgBiCl6, as
conclusion from the previous chapters and provide an overall synthesis of my work.
2
Chapter 2. Background
2.1 Introduction
The limitations of conventional electronics, which are based on the charge of electrons,
are being reached more and more as devices become smaller. In order to explore devices
that go beyond the dogma of Moore’s law and incorporate materials and mechanisms
outside of charge-switching in silicon, more and more device classes have been studied in
the past decade or two. This includes enhanced research in III-V compound
lighting, which is now dominantly based on LED-based devices. Beyond that, more
complex materials classes further away from the diamond structure are explored, which
not only allow enhanced tunability, but also may allow exploitation of spin-based
typically called spintronics. This requires materials with non-trivial magnetic properties,
such as for example half-magnetism, where one spin channel in a material has a band
gap, which the other one is metallic, resulting in a material where 100% of the
conduction electrons have the same spin – something that is not possible in traditional
simple metals or semiconductors. Among the materials classes that allow half metallicity
while at the same time provide large tunability for the sum of all functional properties,
3
the double perovskites is arguably one of the most prominent ones, if not the front runner.
Many composition regimes within double perovskites are to date unexplored, and most
every new regime that is investigated for the first time provides new, exciting and
unexpected findings, making this a highly rewarding field to study. In this dissertation, I
have studied the electronic and magnetic structure and their effects on the properties of
double perovskites.
Ternary perovskites with stoichiometry ABO3, where A is usually a group two metal and
by oxygen atoms on the face centers of a cubic unit cell with a B atom at its center and
the A atoms sitting outside the octahedra on the corners of the cell [1]. Figure 2.1a shows
a ternary perovskite supercell. Perovskites have been thoroughly investigated due to their
is SrTiO3, which demonstrates good insulating properties [2] and is often employed as a
substrate for film growth. It retains its cubic structure at room temperature.
Able to accommodate a large number of different metallic elements on the A and B sites,
the perovskite structure is very versatile [3]. However, its idealized cubic form is only
stable for certain size ratios of the A and B cations, i.e. when twice the B−O bond length
equals the cell edge and twice the A−O bond distance the face-diagonal of the cube. Thus,
unlike the previously mentioned cubic perovskite, SrTiO3, the vast majority of
perovskites are not able to stably remain cubic in structure. An example for this is
CaTiO3, where the Ca cation is smaller than Sr. The resulting size mismatch can be
4
accommodated by a rotation of the octahedral, which results in the structure becoming
orthorhombic rather than cubic [4]. This flexibility of the structure is not only interesting
from a crystallographic standpoint, but can also have significant implications for the
electronic and magnetic properties. This type of distortion along with the Jahn-Teller
distortion, which occurs within the octahedra and lowers the energy of the structure by
While these distortions can lead to significant changes in the functional properties of the
perovskites, even larger tunability can be achieved by expanding the structure to include
works since the majority of the functional properties associated with perovskites stem
from the BO6 octahedra. The resulting quaternary compounds are referred to as double
perovskites and have ideally alternating B and B′ octahedra in rocksalt ordering along all
that double perovskites can possess beyond what is possible in simple perovskites include
magnetic ordering with high Curie temperatures (well above room temperature), high
5
Figure 2.1 Corner sharing BO6 octahedra create a cage in which the A cation sits;
forming the perfect cubic structure for a) a ternary perovskite and b) a double perovskite
Double perovskites also undergo the same types of distortions as ternary perovskites to
accommodate certain ion substitutions. Nearly 70% of them are known to have structures
where octahedral tilting is present [6]. Even the slightest decrease in symmetry in and
around the octahedra can have significant impact on the electronic and magnetic
structure, as our results in the subsequent chapters show. Overall, these compounds have
As previously stated, perovskites are able to rotate their octahedra from the perfect cubic
structure in order to optimize the local environment around the A site cation. When a
perovskite undergoes this type of tilting, there is a change in the overlapping of orbitals
seen between that of the transition metal d-states and the oxygen 2p-states. Often times,
6
the unique properties associated with this family of materials originate from this and will
Goldschmidt Criterion
One way to gauge how “perfect” a perovskite will be is the Goldschmidt tolerance factor
[7]. In an ideal cubic perovskite, the lattice constant can be geometrically linked to the
size of the atomic radii (rA, rB, and rO). From this, the tolerance factor (t) can be derived,
(rA + rO )
t= [2.1]
2(rB + rO )
Figure 2.2 shows examples of possible structures with different amounts of distortion.
For a cubic perovskite, such as SrTiO3, the tolerance factor will be approximately 1 and
at large deviations from this, the cubic structure becomes more likely to be unstable. In
the case of a tolerance factor being t < 1, as it is with CaTiO3 and GdFeO3, the crystal
will adopt an orthorhombic structure [3,4]. When the structure experiences a tolerance
factor t > 1, a hexagonal variation of the perovskite structure can form which leads to the
7
Figure 2.2 Effect of the Goldschmidt tolerance factor t in the resulting perovskite
structure. Low values t < 1 such as in GdFeO3 will lower the symmetry of the crystal
structure by octahedral tilting, resulting in an orthorhombic crystal structure. Large
values t > 1 as in BaNiO3 result in hexagonal variants of the perovskite structure. On top
of the size effect, electronic states can shift and lower the total energy by a symmetry-
breaking Jahn-Teller distortions as shown for LnMnO3 [3].
The Goldschmidt tolerance factor is not only applicable to ternary perovskites. It can also
be modified and applied to double perovskites. In the case of fully ordered stoichiometric
double perovskites the Goldschmidt tolerance factor can be calculated by the following
equation 2.2:
(rA + rO )
t=
r r [2.2]
2( B + B!! + rO )
2 2
8
Now the equation is adapted to account for the effect that both of the B site atoms have
on the structure. Calculating the tolerance factor helps to evaluate the presence and
The use of a tolerance factor has also been used on mixed A-site perovskite compounds.
It was shown that Sr2FeMoO6 has almost a perfect structure with a calculated tolerance
factor of 0.990 when using experimentally found ionic radii [8]. Kim et al. proposed that
by substituting at the A-site, the tolerance level could be brought closer to 1, which could
pursuits with optimal doping yielding the compound Sr0.2Ba0.8Fe0.5Mo0.5O3. Popov and
Greenblatt also studied how optimizing the tolerance factor could be applied to double
perovskites with mixed A-site cations. [7]. Their work with Ba2-xSrxMnReO6 showed that
optimizing the structure to have a tolerance factor of 1 would maximize the saturation
moment. They found that this was indeed true when the compound was about 36% Sr and
64% Ba. The lowest saturation moment was found when the compound had no Ba
present. This was attributed to the lowering of symmetry from the cubic structure. It is
interesting to see that although the tolerance factor was originally used to predict the
stable crystal structure of a perovskite, it can also help to optimize properties. The
simplicity of the Goldschmidt tolerance factor makes it a popular tool used for
perovskites, but of course, there are many other aspects of the structure that come into
play when investigating properties, which in the end make DFT calculations a very
9
Glazer Notation
While the Goldschmidt tolerance factor is able to detect the presence of octahedral tilting,
it is too simple to also describe the exact nature of the tilting. In order to do that, one can
use the Glazer notation, which was first developed in 1972 [9]. Glazer notation is a
scheme that relates the crystallographic lattice to the distortion of the octahedra and has
become the standard by which these types of distortions are described - simplifying the
usual symmetry analysis. There are three tilting characteristics that Glazer classification
describes:
crystallographic direction
• The degree of tilting in the octahedra that occurs specific to one axis compared to
one another
A group of three letters that are combinations of a, b, and c are assigned for each tilt
system, the order of which corresponds to the x, y, and z axes, respectively. Different or
equal letters correspond to equal or different degrees of tilt. For example, a+a+b+ would
represent a crystal with equal rotation along the x and y axes, but a different magnitude of
tilt along the z axis. Glazer notation also expresses whether adjacent layers are in-phase
(rotated in the same direction), out-of-phase (rotated in opposite directions), or if they are
not rotated at all. These are represented by a superscript; (+) in-phase, (-) out of phase, or
around 70% of all double perovskites undergo tilting. Nearly all of these distorted
structures (~97%), can be group within 5 tilt systems, Fm3m (a0a0a0), I4/m (a0a0c-), R3
These distortions are dependent on the electronic structure of the system. The Jahn-Teller
theorem predicts that “any non-linear molecular system in a degenerate electronic state
will be unstable and will undergo distortion to form a system of lower symmetry and
lower energy, thereby removing the degeneracy.” [12] Because of that, the transition-
metal octahedra will experience either elongation or compression (Figure 2.3), which in
turn have an effect on the amount that the orbitals overlap. In an undistorted octahedron,
the d-states of a transition-metal ion split into three t2g states, which are lower in energy
since they do not point toward the oxygen, and two eg states, which are higher in energy
since they do point towards the oxygen. This splitting is due to the repulsive nature of
like charges between the ligand and the d-orbitals. When these states are partially filled,
the states will move either higher or lower in energy depending on where the electrons
Figure 2.2
11
Figure 2.3 Two possible distortions of an octahedron that remove the degeneracy of
unequally occupied eg orbitals for a high spin d4 ion; (right) elongation of the bonds along
z, and (left) elongation of the bonds in the x-y plane. In both instances the remaining
bonds contract to maintain the same degree of metal–ligand bonding [13].
This section of the dissertation outlines some of the important aspects of electronic and
is largely influenced by the B-site cations. Due to the chemical flexibility, or ability to
interchange different ions in the system, and variations in oxidation states of these
cations, there is a wide range of possible properties that double perovskites can take on,
and even superconducting [16]. Double perovskites also exhibit an extensive amount of
12
various combinations of elements. Many are also known to exhibit high Tc values and
The d-states of transition metal oxides generally overlap that is small with O 2p states,
resulting in localized d-electrons causing strong Coulomb repulsion between them and
the narrowing of bands. These are known as strongly correlated materials and such
materials have inter-atomic interactions (or bandwidth (W)) that are equivalently
This interaction occurs between an electron’s spin and the magnetic field generated by
the orbital rotation of the electron through a varying electrostatic potential around the
nucleus. This coupling between the spin and orbital movement causes shifts in an
electron’s energy levels. These energy shifts can cause dramatic changes in the electronic
proportional to the fourth power of the effective nuclear charge Z. With heavier ions, the
need to consider the effects of spin-orbit coupling becomes more important. This is true
in the case of many double perovskites where an extra spin-orbit coupling term needs to
As previously stated, the role that distortion of a perovskite plays on the electronic
structure and magnetic structure is an important one as it helps to govern the degree of
interaction between each atom. In perovskites, there is some covalency present between
13
the transition-metals and the oxygen ligand, which results in a sharing of valence
electrons. Consequently, this sharing can result in strong “exchange” interactions. The
ligand.
There are several types of dominant exchange interactions that can be seen in the
due to the distance between transition metals, which results in no overlap between their
transfer of electrons and is an indirect coupling through the oxygen ligand [17]. It is also
The electrons present in superexchange are localized unlike the delocalized electrons
seen in double exchange. Also in contrast to the hopping of an electron seen in double
exchange, superexchange occurs when valence electrons are shared between the
transition metals and oxygen ligand [18]. This covalency from shared electrons between
the transition metal and oxygen results in magnetic ordering between the neighboring
determined by the Goodenough-Kanamori rules, a set of rules that takes into account
symmetry and orbital occupancy of the transition metals involved to determine the
coupling between two ions [18]. The strongest interactions are usually between the
nearest- and second-nearest-neighbors and the strength decreases rapidly as the distance
increases. Several of these rules are important to the ordering of insulating perovskites:
14
• The coupling of ions with half-filled orbitals will have antiferromagnetic
moments with respect to each other when the B-O-B bonds are ~180 ͦ
• The coupling of ions with half-filled or empty, or half-filled and fully occupied
orbitals will have ferromagnetic moments with respect to each other when the B-
• The coupling of ions will prefer to have ferromagnetic moments with respect to
For the majority of insulating transition metal oxides, the Goodenough-Kanamori rules
extremely helpful to scientists in their search for new magnetic materials. However, there
are exceptions to the rules, and sometimes variations in orbital overlap from the idealized
cases used for derivation lead to what looks like a breaking of the rule, as in the case of
References
1. Piskunov, S., Heifets, E., Eglitis, R. . & Borstel, G. Comput. Mater. Sci. 29, 165–
178 (2004).
2. Van Benthem, K., Elsa sser, C. & French, R. H. J. Appl. Phys. 90, 6156 (2001).
15
6. Lufaso, M. W. & Woodward, P. M. Acta Crystallogr. Sect. B Struct. Sci. 60, 10–
20 (2004).
7. Popov, G., Greenblatt, M. & Croft, M. Phys. Rev. B 67, 024406 (2003).
8. Kim, B. G., Hor, Y. S. & Cheong, S. W. Appl. Phys. Lett. 79, (2001).
10. Mishra, R., Soliz, J. R., Woodward, P. M. & Windl, W. Chem. Mater. 24, 2757–
2763 (2012).
13. Woodward, P. M., Karen, P., Evans, J. S. . & Vogt, T. Functional Materials.
(Unpublished).
14. Mishra, R., Restrepo, O. D., Woodward, P. M. & Windl, W. Chem. Mater. 22,
6092–6102 (2010).
15. Popov, G., Greenblatt, M. & Croft, M. Phys. Rev. B 67, 024406 (2003).
16. Vasala, S. & Karppinen, M. Prog. Solid State Chem. 43 1-36 (2015).
16
Chapter 3. First Principles Methods
3.1 Introduction
In this chapter, I will discuss the basic fundamental elements of density functional theory
(DFT), which was the method used to do all electronic structure calculations within this
used.
To better understand the influence of distortions on the electronic structure and the
wavefunctions in the system. DFT is able to map a many-body system into a more
interacting with a system of ions and an effective field created by the electron density,
where the neglected many-body (correlation) and exchange effects are included by
17
ℏ
− !! ∇! 𝜓 + 𝑉𝜓 = 𝐸𝜓 (1)
Where the first term is the kinetic energy of the electron of mass, m, 𝜓 is the
wavefunction, and E is the energy. As this is numerically impractical for the full many-
body system, approximations need to be made as mentioned above. Firstly, since the
mass of nuclei is much heavier than that of the surrounding electrons, it can be assumed
that the electrons adjust instantaneously to the motion of the nuclei, and thus the motion
of nuclei and electrons in a system can be separated into independent equations. Within
equation is solved for only the electrons while it is assumed that the nuclei are fixed at a
given location. Also, periodic boundary conditions are frequently applied to considerably
reduce the amount of atoms present in the computational unit cell in order to calculate
bulk properties. Even with that, especially for heavier atoms, there are still a significantly
large number of electrons present in the system. In order to reduce this number, the
explicit calculation is often restricted to valence electrons, while the core electrons are
method [1].
In 1964, Hohenberg and Kohn proved that, for a given external potential, the total energy
of a system is a functional of the ground state electron density [2]. Although this was an
important simplification, the problem of obtaining the full electron density still remained.
In 1965, Kohn and Sham were able to remap the many-body interacting system into
terms of non-interacting electrons to produce the same electron density as the interacting
18
system [3]. While working functionals could be found for most interaction terms, the
functional dependence of the kinetic energy on the electron density has been elusive. In
order to work around this problem, single-electron wavefunctions are introduced, whose
summed up square modulus is set equal to the system’s electron density. For
wavefunctions, the kinetic energy can be easily calculated within the standard
Schrödinger equation.
In practice, the wave functions within a DFT run are determined from an initial
assumption for the electron density (e.g. the superposition of atomic electron densities
which are known), and then a new, updated electron density is calculated from the
resulting wave functions. This procedure is iterated until input and output electron
density agrees within a chosen convergence criterion. Despite the many approximations
that are made, DFT is a widely used and trusted computational method due to its
and use it locally for the system under consideration. This is known as the local density
approximation (LDA) [4]. Although LDA works well for systems with slowly varying
densities, it does not always perform well, especially when it comes to functional
problem when doing calculations for double perovskites. In addition to that, the
19
simplified uniform electron density approximation leads to an overestimation of the
functionals that do not just assume dependence on the local electron density, but also on
the local gradient of density. This approach is known as the Generalized Gradient
Approximation (GGA) and is the method used for the majority of this work. GGA tends
to be an improvement from that of LDA, but it sometimes overcorrects for the binding,
When transition metal oxides come into play, the system will have usually strongly
localized d-electrons. These localized electrons tend to have very strong electron-electron
many body interactions. These can be problematic when using LDA and GGA, which
In order to adjust for these strong correlation effects an additional term is needed in the
calculations. This is known as the on-site Hubbard U [5]. In a simple picture, it can be
thought of is an energy penalty for the localized d and f- state electrons of the strongly
correlated transition metals when they delocalize. It helps to improve the description of
the ground state with very little extra computational cost. This +U correction can be
included in the previously mentioned LDA or GGA. Incorporating the U shifts energy
levels of occupied orbitals by -U/2 and unoccupied orbitals by +U/2. This provides an
adjustable parameter for strongly correlated materials to better model the physical
Still needing to be considered for calculations is the spin of electrons, which is essential
incorporate spin dependence into LDA, two electron densities can be introduced, one
describing spin-up, the other spin-down electrons. This approach is known as Local Spin
Density Approximation (LSDA) [6] and works well for the vast majority of systems,
including double perovskites, provided that the spin interactions are well described by a
3.4 Summary
The flexibility of the perovskite structure opens itself up to the incorporation of numerous
different elements in different combinations. Many of the structures that are possible
have been unexplored and could lead to even more interesting properties. As techniques
in synthesis and growth improve, the list of available structures to investigate increases.
Since the functional properties of perovskites and double perovskites depend sensitively
on the exact geometry of the structure, which in turn is a function of the oxidation states
and size of the elements involved, examining structural distortions and how they affect
the underlying electronic and magnetic structure is key in gaining a better understanding
In order to be able to guide synthesis and explore novel composition systems, we have
shown that computational work can be an indispensable asset towards faster and more
targeted synthesis efforts. As the currently dominant approach, we have introduced the
framework of DFT and briefly discussed its physical framework. Finally, I would like to
21
mention that although DFT is a very useful tool in understanding the phenomena seen in
experiment, there are still limitations, and it is frequently difficult to know a priori which
modeling choices need to be made (such as what value the Hubbard U should have),
collaboration with experiment is still extremely important, which is thus the nature of the
References
4. Sholl, D.S. & Steckel, J.A. Density Functional Theory: A Practical Introduction
(John Wiley and Sons, Hoboken, NJ, 2009).
5. Himmetoglu, B., Floris, A., de Gironcoli, Stefano & Cococcioni, M. Int. J. Quant.
Chem. 114, 14-49 (2014).
22
Chapter 4. Independent Ordering of Two Interpenetrating
Magnetic Sublattices in the Double Perovskite Sr2CoOsO6
Note: This chapter is being presented in the format of a journal article. It was published
as Morrow, R., Mishra, R. Restrepo, O. D., Ball, M. R., Windl, W., Wurmehl, S.,
American Chemical Society. 135 18824-18830 (2013). Rohan Mishra performed the DFT
calculations and performed the fitting for the results of this paper with the assistance of
Oscar Restrepo.
4.1 Abstract
The insulating, fully ordered, double perovskite Sr2CoOsO6 undergoes two magnetic
phase transitions. The Os(VI) ions order antiferromagnetically with a propagation vector
k =(½,½,0) below TN1 = 108 K, while the high spin Co(II) ions order
symmetry to monoclinic I2/m symmetry, which reduces the frustration of the face
centered cubic lattice of Os(VI) ions. Density functional theory calculations show that
23
considerably stronger than the shorter Os−O−Co interactions. The poor energetic overlap
between the 3d orbitals of Co and the 5d orbitals of Os appears to be responsible for this
unusual inversion in the strength of short and long range superexchange interactions.
4.2 Introduction
a given superexchange interaction depends upon the orbitals involved and generally
decreases rapidly as the distance between the magnetic ions increases. The Goodenough-
Kanamori rules1,2 provide a simple, yet accurate, means of determining the sign of the
superexchange interactions These simple concepts are used by scientists to guide their
search for new magnetic materials. In this article we report the crystal and magnetic
structures of Sr2CoOsO6, a quaternary transition metal oxide with the double perovskite
structure. The A2MM′O6 double perovskites are among the most interesting classes of
transition metal oxides.3 Compared to simpler AMO3 ternary perovskites, the presence of
chemical order of M and M′ cations profoundly impacts both the electronic structure and the
magnetic coupling, in part because there are a greater variety of magnetic exchange interactions
at work in the double perovskite structure. The results presented here show that long range
superexchange interactions between ions of the same type (i.e. Co–O–Os–O–Co and Os–
O–Co–O–Os) are much stronger than the shorter nearest neighbor Co−O−Os
superexchange interactions. This result has broad implications for the way we think
The double perovskite structure, with general formula A2MM′O6, consists of a cubic,
corner connected network of MO6 and M′O6 octahedra that alternate in all three
24
directions, so that each MO6 octahedron is connected only to M′O6 octahedra and vice
rules. According to those same rules adding electrons to the eg orbitals of the 3d ion, for
example by replacing Cr with Fe(III), Co(II) or Ni(II), should change the sign of the
superexchange interactions and stabilize an insulating ferromagnet with a high TC—a rare
combination.5 To test this hypothesis as well as to better understand the rules that dictate
prepared by Sleight et al.,6 but whose magnetic and electrical properties were previously
unknown.
Figure 4.1 The A2MM′O6 double perovskite structure. The large A cations are shown in
gray, while the M and M′ centered octahedra are shown in red and blue respectively.
25
4.3 Experimental Section
Batches of at most 1.6 g (3.075 mmol) Sr2CoOsO6 powder were prepared by a solid state
method utilizing stoichiometric quantities of SrO2 (Sigma Aldrich, 98% pure), Co3O4
(Fischer Scientific, 99.8%pure) and Os metal (Alfa Aesar, 99.98% pure). The mixture
was loaded into a capped alumina tube and placed in a quartz tube, of approximately 40
mL volume with 3mm thick walls, along with an additional capped alumina vessel
containing PbO2. The quartz tube was then evacuated and sealed prior to firing in a
furnace to 1000 °C for 48 hours within a fume hood. The PbO2 was reduced to PbO at
without allowing appreciable loss of volatile OsO4.7,8 The best results were obtained
when the amount of PbO2 was enough to produce an excess of one-quarter mole O2 per
Phase purity and chemical ordering was established via XRD utilizing a Bruker D8
26
diffraction pattern for this sample can be found in the supporting information in
Note that when heating osmium metal or binary osmium oxides it is important to confine
the reactants to a sealed vessel and take precautions in case a leak of the vessel occurs,
sample size of approximately 1.6 g contained within an 8 mm vanadium can was used for
referred to as Frame 1.5 and Frame 5 respectively. Long scan times of 1 hour for Frame
1.5 and 2 hours for Frame 5 were collected at 12, 80, 130, and 300 K while brief 5 minute
was performed and magnetic cells were generated utilizing SARAh11, and refinements
were conducted with the Rietveld method as implemented in GSAS EXPGUI.12,13 The
magnetic form factor for osmium was input manually using the parameters given by
Kobayashi et al.14
The magnetic susceptibility of Sr2CoOsO6 was collected within the temperature range of
5–400 K with both zero-field cooled and field cooled conditions under the effect of an
27
Design MPMS SQUID magnetometer. A 46.6 mg sample was contained in a gelatin
capsule and mounted in a straw for insertion into the instrument. No diamagnetic
corrections for this sample holder were taken into account due to the large contribution of
the sample.
Sr2CoOsO6 powder was pressed into a pellet prior to being sintered at 1100 °C and being
cut into an approximate bar shape. The density was calculated to be 66.9% of the
theoretical density of the material. Four-point contacts were secured to the pellet with
silver paint before DC electrical measurements of the sample were conducted within the
Specific heat capacity was measured with a Physical Property Measurement System
(PPMS, Quantum Design), using a relaxation method. Sr2CoOsO6 was measured in zero
magnetic field between 2 K and 298 K using small, disc-like samples of roughly 11 mg.
In order to thermally couple the sample to the calorimeter, Apiezon grease was used. The
specific heat of the Apiezon grease was measured alone first over the whole temperature
range, and then subtracted from the total value of the subsequent measurement which
included the sample, to derive the specific heat contribution from the sample only.
Density functional theory calculations were performed using the Vienna ab-initio
Simulation Package (VASP).15,16 The influence of the core electrons was incorporated
wave cutoff energy was set at 525 eV. The Brillouin zone was sampled using dense
28
Monkhorst-Pack k-point meshes,19 with mesh divisions Ni such that the product of Ni with
the corresponding lattice vectors was as close as possible to 50 Å. The low temperature
(12 K) experimental lattice constants and ionic coordinates obtained from neutron
diffraction studies were used in the calculations without performing any structural or
ionic relaxations. The collinear magnetic calculations were performed on 20 to 160 atoms
supercells. The experimentally observed magnetic structure was studied using a 160 atom
implemented in VASP. 20
To include the strong-correlations in the transition metal elements, we have used the
Hubbard parameter Ueff is used. Given that Ueff is not precisely determined by experiment
or theory, we have examined a range of Ueff values: 0 and 4.1 eV for Co and 0, 2.1 and
4.1 eV for Os. While the order of stability of different magnetic configurations remains
the same for all combinations of UCo and UOs, we find that UCo = 4.1 eV and UOs = 2.1
eV, also reproduce the experimentally observed insulating nature. Hence, we use this
Figure 4.2. As signaled by the cusps in the data, Sr2CoOsO6 appears to undergo two
K. The inverse susceptibility in the paramagnetic regime 200 – 400 K follows the Curie-
29
Weiss law. The Weiss temperature, θ, extracted from the Curie-Weiss fit was found to be
−51 K. The atypical observation that θ is lower than either of the Neél temperatures
superexchange interactions. The effective magnetic moment per formula unit obtained
from the Curie-Weiss fit is µeff = 4.46 µB, which is in reasonably good agreement with the
expected spin only value calculated assuming the presence of Os(VI) and high spin (HS)
Co(II), µspin = [µspin(Co)2 + µspin(Os)2]1/2 = 4.80 µB. However, it should be noted that HS
Co(II) has been reported in related double perovskites to have an effective moment which
is higher than the spin-only value of 3.87 µB due to a constructive contribution from
unquenched orbital moment,23-26 and the orbital contribution if present for Os(VI) would
tend to reduce the moment of that ion from its spin only value of 2.83 µB.27 Thus it is
quite plausible that orbital angular momentum contributes to the moment of both ions,
but the effects average out to give an overall µeff close to the spin-only value.
30
240
1.4e-2
Molar Susceptibility (emu/mol)
200
Inverse Susceptibillity
1.2e-2
160
1.0e-2
120
8.0e-3
80
6.0e-3 40
4.0e-3 0
0 100 200 300 400
Temperature (K)
Figure 4.2 Field cooled (filled circles) and zero field cooled (open circles) magnetic
susceptibility with inverse zero field cooled data plotted against the right axis showing
the linear Curie-Weiss fit.
at 108 K and 70 K (Figure 4.3). These temperatures correspond closely to the cusps seen
normally seen for a second-order phase transition. This observation, taken together with
other hand, the rather symmetric shape of the peak at 108 K suggests a first-order phase
31
Figure 4.3 Specific heat of Sr2CoOsO6 as a function of temperature
with decreasing temperature, as shown in Figure 4.4. The log of the conductivity shows a
linear T−1/4 temperature dependence (see supporting information) which is consistent with
variable range hopping transport, whereby electrons hop between localized states. There
is a continuous deviation which occurs between the phase transitions, leading to two
separate linear fits, above and below the transitions. The mathematical form of variable
!/! )
range hopping conductivity is given as σ=σ0𝑒 (!!/! . The fitted A factors found above
and below this transition temperature are 43.12(2) and 46.14(9), respectively.
32
Figure 4.4 Log resistivity vs temperature.
collected to obtain accurate structural parameters and to probe the magnetic structure. At
room temperature both X-ray and neutron powder diffraction patterns can be fit to a
tetragonally distorted, ordered double perovskite structure with I4/m symmetry and
complete ordering of Co and Os. The tetragonal distortion occurs due to rotations of the
octahedra about the c-axis (a0a0c– tilting)3 and a relatively subtle axial elongation of the
octahedra. This relatively common distortion of the structure is also observed at room
The evolution of the structure was studied with variable temperature neutron powder
diffraction, and the results are shown in Figure 4.5. On cooling below 108 K, where
33
magnetic ordering sets in, an abrupt structural distortion occurs lowering the symmetry to
the monoclinic space group I2/m. The splitting of the (220) reflection signaling the
Figure 4.5 Evolution of the (a) lattice parameters, (b) Os–O bond lengths, and (c) Co–O
bond lengths as a function of temperature, as extracted from variable temperature neutron
diffraction data. The dashed line separates the tetragonal I4/m phase from the monoclinic
I2/m phase.
34
The Co–O and Os–O distances are plotted as a function of temperature in Figure 4.5. In
the high temperature tetragonal structure the axial Co–O(1) and Os–O(1) bonds are
slightly longer than the Co–O(2) and Os–O(2) bonds that lie in the xy-plane. This type of
distortion could be associated with a Jahn-Teller distortion arising from the orbital
degeneracy of either the HS d7 Co(II) ion or the d2 Os(VI) ion. The magnitude of the
distortion is relatively small, but this is not unexpected for a Jahn-Teller distortion arising
In both the tetragonal and monoclinic phases the bond lengths are consistent with the
assignment of high spin Co(II) and Os(VI) oxidation states,29,30 so we can rule out a
change in spin-state or some type of Co–Os charge transfer as the driving force behind
can also rule out octahedral tilting as the driving force for the phase transition. Instead
the transition appears to be most closely linked to distortions of the CoO6 octahedra, as
In the low temperature monoclinic structure the symmetry is lowered in such a way that
there are three unique bond distances within each octahedron. Despite the reduction in
symmetry, the 2 long/4 short distortion of the osmium octahedron is essentially retained
distortion of the cobalt octahedron changes at the phase transition and this change is
responsible for lowering the symmetry to monoclinic. In the tetragonal structure all four
of the equatorial Co–O(2) bonds are the same length, 2.036(1) Å at 130 K. This square
35
base distorts in the monoclinic structure to give two bonds of 2.013(2) Å and two bonds
of 2.056(2) Å. The longer of these two sets of bonds is comparable to the axial Co−O(1)
distance, 2.067(1) Å, so that the octahedron now has 4 long and 2 short bonds. This type
of distortion is unusual for Co(II). A thorough literature search revealed only one other
Table 4.1 gives information concerning the unit cell and goodness of fit parameters for
each refinement on the longer neutron data sets. Figures and tables containing further
details on the structure of Sr2CoOsO6 at different temperatures can be found in the
supporting information in Appendix A.
Co moment 2.90(5) µB - - -
36
Figure 4.6 shows the electronic density of states calculated using density functional
theory (DFT). The 5d t2g orbitals of osmium make the major contribution at the Fermi
level, but correlations open up a small gap in what would otherwise be a 1/3 filled band.
Figure 4.6 Total- and atom- resolved density of states of Sr2CoOsO6 from a non-collinear
DFT calculation with UCo = 4.1 eV and UOs = 2.1 eV. The contributions of the Os 5d and
Co 3d orbitals are shown in red and blue, respectively.
Upon cooling below TN1 = 108 K, an additional set of reflections arise in the neutron
diffraction pattern that cannot be attributed to the structural phase transition, one such
reflection is found at d = 7.7 Å as shown in Figure 4.7. These reflections signal the onset
of antiferromagnetic order, and can be accounted for with a magnetic propagation vector
was also possible to fit the data reasonably well with non-negligible moments on both Co
and Os by fixing the moment on one atom to an intermediate value, such as 0.8 µB, and
refining the moment on the other. However, those refinements always gave slightly
higher Rwp values, and poorly fit one of the magnetic reflections. This observation lends
support to the hypothesis that between 70−110 K only one magnetic sublattice orders.
Figure 4.7 Rietveld fits to the high d-spacing regions of the neutron diffraction patterns
showing the presence/absence of magnetic reflections at three temperatures: above TN1
(130 K), between TN1 and TN2 (80 K), and below TN2 (12 K). Black symbols represent
observed data while the red curves represent the calculated patterns. The hkl values given
in the figure are indexed on the nuclear cell.
38
On cooling below TN2 a second set of magnetic reflections emerge that can be indexed
with a propagation vector k = ½, 0, ½, the strongest of which is the peak at d = 9.1 Å seen
in Figure 4.7. Due to the similarity of the a and b lattice parameters, it is possible to
½, due to powder averaging. The following analysis and conclusions are not affected by
this choice of propagation vector. A symmetry analysis yields only one irreducible
representation, Γ(2), for this second set of magnetic peaks. Although the Γ(2)
representation has three basis vectors, only the two oriented in the a and c directions were
needed to model the intensities of the magnetic reflections. It should also be noted that
the emergence of a second set of magnetic reflections below TN2 does not lead to any
noticeable change in the intensities of the first set of magnetic reflections. A low
scheme on one magnetic sublattice, and the k= ½, 0, ½ scheme on the other sublattice,
leading to a 2 × 2 × 2 magnetic unit cell, shown in Figure 4.8(a). In this structure the
cobalt and osmium spins are non-collinear, with both sublattices adopting
antiferromagnetic order.
While the neutron diffraction experiments provide strong support for independent
ordering of cobalt and osmium ions, it is not immediately apparent which ion orders at
the higher temperature. If we assume that the cobalt ions order at 108 K and the osmium
ions order at 70 K, refinements of the 12 K data set yield atomic moments of 1.99(4) µB
and 2.71(4) µB for these two ions, respectively. These values are not physically
reasonable because a moment of 2.71 µB for the d2 Os(VI) ion is larger than the spin-only
39
value of 2 µB, and any orbital contribution would be expected to reduce, not increase, the
moment. Using the opposite assumption, osmium ordering at 108 K and cobalt ordering
at 70 K, gives moments of 1.81(4) µB and 2.90(5) µB that are quite close to the high field,
spin-only values of 2 µB and 3 µB expected for the Os(VI) and HS d7 Co(II) ions,
respectively. DFT calculations give moments of 1.69 µB and 2.71 µB for osmium and
cobalt, respectively. These values are in close agreement with the experimentally
obtained values, and the agreement strongly supports the conclusion that the osmium ions
40
Figure 4.8 (a) Experimentally determined magnetic structure of Sr2CoOsO6, with the Os
shown in red and the Co in blue. (b) A schematic of the superexchange pathways between
Co and Os ions with the exchange constants extracted from the DFT calculations. A
positive (negative) value of J indicates ferromagnetic (antiferromagnetic) coupling.
The neutron refinements reveal a complex magnetic structure, but they do not provide
direct information as to why this magnetic structure is more stable than simpler
extract the magnetic coupling constants that are responsible for the unusual magnetism
41
Sr2CoOsO6, but in that study they only looked at the ferromagnetic and ferrimagnetic
However, as shown here the actual ground state is an antiferromagnetic insulator, which
lessens the validity of the conclusions made in the earlier computational study.
The relative energies of ten magnetic configurations were calculated and the
unit cell was found to be the lowest energy configuration (see supporting information).
The ferromagnetic configuration was 809 meV per f.u. higher in energy, while the
ferrimagnetic configuration was 776 meV higher in energy. The exchange constants, for
seven exchange pathways of potential importance were extracted from these calculations
and are given in Figure 4.8. To minimize correlations in the fitting procedure the through
bond superexchange constants J1eff, J2eff, and J3eff were constrained to be the same in all
three directions. Because the direct exchange coupling between ions of the same type
across a face of the unit cell can be quite sensitive to octahedral tilting, we have used two
different J values for Co−Co coupling, J4eff, and J5eff, and two for Os−Os coupling, J6eff,
and J7eff.
!""
The calculations reveal very weak coupling, 𝐽! = −1.3 meV for the nearest neighbor
Co–O–Os superexchange pathways that are the shortest interactions and would normally
be expected to be the strongest. The largest superexchange coupling is found for the four
!""
bond Co(↑)–O–Os–O–Co(↓) interaction, which is antiferromagnetic with 𝐽! = – 47.2
!""
meV The four bond Os(↑)–O–Co–O–Os(↑) interaction is also quite strong, with 𝐽! =
20.2 meV, but ferromagnetic. The in-plane Co−Co coupling, J4eff was also relatively
42
strong, 29 meV, but this value was somewhat dependent on the constraints used with
emerge as being much stronger than the shorter nearest neighbor interactions along the
same path, but a closer look at the experimental structure shows that all Os–O–Co–O–Os
interactions are indeed ferromagnetic, while all of the Co–O–Os–O–Co interactions are
(Figure 4.6) shows that there is only a small overlap of the Co 3d and Os 5d states
(particularly for the eg orbitals), which helps to explain the weak nearest neighbor
interaction between the two ions. In other words the poor energetic overlap of the Co 3d
and Os 5d orbitals (once hybridized with oxygen) leaves the two sublattices relatively
uncoupled.
Before finishing let’s return to the unusual 4 long/2 short distortion of the cobalt
octahedron that occurs on cooling below TN1. While there is no way to rationalize this
type of distortion based solely on the electron configuration of the HS Co(II) ions, the
fact that the distortion accompanies ordering of the osmium spins provides an important
clue. In the double perovskite structure the osmium ions sit on a face centered cubic (fcc)
interpenetrating sublattices of Os ions that do not interact with each other via this
exchange route. These sublattices interact through nearest neighbor Os–Os exchange
given as 𝐽! and 𝐽! in Figure 4.8. In the high temperature tetragonal structure, the Os–Os
43
nearest neighbor distances are 5.51 Å (×4) and 5.59 Å (×8). This arrangement is slightly
distorted from a perfect fcc lattice, but still highly frustrated. In the low temperature
monoclinic structure the distances become 5.51 Å (×4), 5.57 Å (×4), and 5.61 Å (×4).
These distances represent the nearest neighbor distances from one Os sublattice to the
other three independent Os sublattices. The two sublattices that are aligned
ferromagnetically with each other are separated by the longest of these three distances,
5.61 Å. Thus the structural distortion that accompanies TN2 reduces the frustration of the
fcc arrangement of osmium ions by shortening the Os–Os distances between the eight
between the four nearest neighbor ferromagnetically coupled Os ions. Based on this
observation we hypothesize that the structural transition at 108 K, and the resulting
spins.
4.5 Conclusions
The double perovskite Sr2CoOsO6 undergoes two magnetic phase transitions on cooling
from room temperature. The two interpenetrating sublattices of Os(VI) and Co(II) each
order antiferromagnetically, but independent of each other. The Os(VI) spins order
below 70 K. The observed magnetic structure suggests and the computational results
Co, are stronger than the nearest neighbor Os–O–Co superexchange interactions. This
revealing a complexity that was not previously appreciated. Further studies on a variety
of A2MM′O6 double perovskites with varying electron count, orbital energies, and
structural distortions are needed to better understand the factors that control the strength
References
[3] Lufaso, M.W.; BarnesP. W.;Woodward, P.M. Acta Cryst. B 2006, 62, 397-
410.
[4] Krockenberger, Y; Mogare, K.; Reehuis, M.; Tovar, M.; Jansen, M.;
Vaitheeswaran, G.; Kanchana, V.; Bultmark, F.; Delin, A.; Wilhelm, F.; Rogalev,
A.; Winkler, A.; Alff, L. Phys. Rev. B. 2007, 75, 020404(R).
[5] Rogado, N. S.; Li, J.; Sleight, A. W.; Subramanian, M.A. Adv. Mater. 2005, 17
(18), 2225-2227.
[6] Sleight, A. W.; Longo, J.; Ward, R. Inorg. Chem. 1962, 1 (2), 245–250.
[7] Macquart, R; Kim, S. J.; Gemmill, W. R.; Stalick, J. K.; Lee, Y.; Vogt, T.; Zur
Loye, H. C . Inorg. Chem. 2005, 44, 9676-9683.
[8] Lufaso, M. W.; Gemmill, W. R.; Mugavero, S. J.; Kim, S. J.; Lee, Y.; Vogt, T.;
zur Loye, H. C. J. Solid State Chem. 2008, 181 (3), 623-627.
[9] Huq A.; Hodges J. P.; Gourdon, O.; Heroux, L. Zeitschrift für Kristallographie
Proceedings 2011, 1, 127-135.
45
[10] Rodriguez-Carvajal, J. Physica B. 1993, 192, 55-69.
[12] Larson, A. C.; Von Dreele, R. B. Los Alamos National Laboratory Report LAUR
86-748 2000.
[13] Toby, B. H. EXPGUI, a graphical user interface for GSAS. J. Appl. Cryst. 1991,
34, 210-213.
[14] Kobayashi, K.; Nagao, T.; Ito, M. Acta Crys. A 1991, 67, 473-480.
[20] Hobbs, D.; Kresse, G.; Hafner, J. Phys. Rev. B 1991, 62, 11556–11570.
[21] Dudarev, S. L.; Botton, G. A.; Savrasov, S. Y.; Humphreys, C. J.; Sutton, A. J.
Phys. Rev. B 1991, 57, 1505–1509.
[22] Anisimov, V. I.; Zaanen, J.; Andersen, O. K. Phys. Rev. B 1991, 44, 943-954.
[23] Viola, M. C; Martinez-Lope, M. J.; Alonso, J. A.; De Paoli, J. M.; Pagola, S.;
Pedregosa, J. C.; Fernandez-Diaz, M. T.; Carbonio, R. E. Chem. Mater. 2003, 15
(8), 1655-1663.
[24] Ivanov, S. A.; Eriksson, S. G.; Tellgren, R.; Rundlöf, H.; Tseggai, M. Mater. Res.
Bull. 2005, 40, 840-849.
[25] Lopez, C. A.; Saleta, M. E.; Curiale, J.; Sanchez, R. D. Mater. Res. Bull. 2012,
47, 1158-1163.
[27] Choy, J. H.; Kim, D. K.; Kim., J. Y. Solid State Ionics. 1998, 108, 159–163.
46
[28] Chmaissem, O.; Kruk, R.; Dabrowski, B.; Brown, D. E.; Xiong, X.; Kolesnik, S.;
Jorgensen, J. D.; Kimball C. W. Phys. Rev. B 2000, 62, 14197-14206.
[31] Currie, R. C.; Vente, J. F.; Frikkee, E.; IJdo, D. J. W. J. Solid State Chem. 1995,
116 (1), 199-204.
[32] Tian,C.; Wibowo, A. C.; Zur Loye, H.-C.; Whangbo, M-H. Inorg. Chem. 2011,
50, 4142-4148.
[33] Wang, J.; Meng, J.; Wu, Z. Chem. Phys. Lett. 2011, 501, 324-329.
47
Chapter 5. Strain-dependent, Extraordinary
Magnetocrystalline Anisotropy in Sr2CrReO6 Epitaxial Films
Note: This chapter is being presented in the format of a journal article. It was published
as Lucy, J. M., Ball M. R., Restrepo, O. D., Hauser, A. J., Soliz, J. R., Freeland, J. W.,
180401(R) (2014). I performed all the DFT calculations including relaxing the structures,
ascertaining the density of states, and determining the magnetic moments. The
subsequent analysis for this work was also done by me with assistance from Oscar
Restrepo.
Abstract
use magnetometry measurements and first principles calculations to determine the atomic
origins of the large anisotropy observed. These techniques elucidate the interplay
48
between structural deformations and magnetic behavior and lay the groundwork for the
applications such as power generation and magnetic data storage. The search for and
study of materials for such applications is of both scientific and technological interest.
There is much focus on ferromagnets such as SmCo5, Nd2Fe14B and FePt which exhibit
high anisotropy due to crystal symmetries and strong spin-orbit coupling [1,2]. Here we
report the discovery of exceptionally large anisotropy fields and strain-tunable MCA in
octahedral rotations of the Sr2CrReO6 lattice which lead to dramatic changes in MCA and
the capability to switch the magnetic easy axis from in-plane to out of plane via strain.
films in recent years, have enabled engineering of materials with desired, and sometimes
exotic, electronic and magnetic properties such as new ferroelectric materials [3,4] and
ferromagnets (FMs) with large MCA [1,2,5]. Most importantly, epitaxial strain offers the
capability to significantly alter the electronic and magnetic properties of the films and
even creates new phenomena [3,4] that do not exist in bulk. Crystal structures that are not
isotropic, such as tetragonal (e.g. CrO2 and FePt) and hexagonal lattices (e.g. Co and Dy),
49
typically result in large MCA due to magnetization-lattice coupling [5-7]. Since spin-
orbit coupling (SOC) generally scales with Z4, where Z is the atomic number, high-
small due to the high symmetry, such as in 3d FM metals and Heusler compounds [8,9].
The ABO3 perovskites are a large family of complex materials that exhibit many
modest MCA. Meanwhile, the A2BB’O6 ferrimagnetic double perovskites have been
shown to possess versatile magnetic properties such as high spin polarization, high TC (up
to 725 K), strong and tunable SOC, and electrical conductivity ranging from insulating to
Sr2CrReO6 epitaxial films [14-16] using off-axis sputtering, which exhibit SOC-enhanced
magnetization. Previous studies of Sr2CrReO6 in bulk and epitaxial film form can be
found in Refs. [10,11,14-16]. It has been recently reported that Sr2CrReO6 films exhibit
elucidated the contributing factors leading to the large MCA in Sr2CrReO6 [18,19]. These
results point toward the possibility of tuning the large MCA in S Sr2CrReO6 films via
strain.
(SCNO) buffer layer on LSAT with lattice constants a = 3.868 Å, 3.905 Å and 3.946 Å,
50
lattice constants of bulk Sr2CrReO6 are ap = 3.907 Å and cp = 3.905 Å [10]. It should be
Sr2CrReO6 films through the use of a piezoelectric substrate. However, at this time no
piezoelectric substrate is available that can apply epitaxial strain within the
Sr2CrReO6.
The 90 nm Sr2CrReO6 films are thin enough to be fully strained to the underlying
substrates and the similarities in film thickness reflect our precise control of deposition
rates. The x-ray diffraction (XRD) scans in Fig. 5.1 show that the films are pure phase
0.0063°. The left panels in Figs. 5.1(a)-5.1(c) show the θ−2θ scans near the Sr2CrReO6
(004) peak for films on LSAT, SrTiO3 and SCNO/LSAT, from which the out-of-plane
7.804 Å, respectively, are obtained. From the off-axis θ−2θ scans for the Sr2CrReO6
(022) peak (right panels in Fig. 5.1), we calculate the in-plane lattice constants a = 7.732
Å, 7.806 Å and 7.876 Å, for films on LSAT, SrTiO3 and SCNO/LSAT, resulting in
tetragonal distortions c/a = 1.025, 1.007 and 0.991, respectively. All three films are fully
strained, compressive (c/a > 1) or tensile (c/a < 1), to the substrates or buffer layer (Table
5.1). Finally, Laue oscillations are observed in all three samples, indicating high
uniformity through the films. From the spacing of Laue oscillations, we obtain
thicknesses of 91.1, 90.2 and 89.6 nm for the Sr2CrReO6 films grown on LSAT, SrTiO3
51
(a) SCRO/LSAT
LSAT(002)
LSAT(011)
SCRO/LSAT
105
Intensity (c/s)
SCRO(004)
SCRO(022)
103
101 = 45o
(b) SCRO/STO SCRO/STO
STO(002)
STO(011)
105
Intensity (c/s)
103
SCRO(022)
SCRO(004)
101 = 45o
(c) 45 45.5 46
SCRO/SCNO/LSAT 46.5 47 SCRO/SCNO/LSAT
LSAT(011)
105 LSAT(002)
Intensity (c/s)
SCNO(004)
103
SCRO(004)
SCNO(022)
SCRO(022)
101 = 45o
45 45.5 46 46.5 47 32 32.5 33
2 (deg) 2 (deg)
FIG. 1
Figure 5.1 XRD scans around the Sr2CrReO6 (004) (left panels) and Sr2CrReO6 (022)
(right panels) peaks for 90-nm Sr2CrReO6 films grown on (a) LSAT, (b) SrTiO3 and (c)
Sr2CrNbO6 buffer layer on LSAT. The off-normal Sr2CrReO6 (022) peaks are measured
at a tilt angle Ψ =45° for in-plane characterization.
the octahedra alters both the bond lengths and bond angles between transition metals and
oxygen, potentially affecting both their electronic and magnetic properties. We utilize the
measured lattice parameters of the Sr2CrReO6 films to determine the rotation of the Cr
52
and Re oxygen octahedra by performing density functional theory (DFT) calculations
within the generalized gradient approximation (GGA) [26] using the Vienna ab initio
pseudopotentials [29]. Correlation effects were treated within the (GGA+U) approach
orbitals [31]. The resulting geometries and structural parameters are shown in Fig. 5.2
equatorial (in-plane) and axial (out-of-plane) bond lengths, deq and dax, respectively, in
the octahedra are considerably smaller than the changes in the lattice constants,
compressive to tensile strain. Given the linear relationship in Figs. 5.2(c) and 5.2(d), we
find that the increase in the in-plane bond lengths is 23% (17%) of the changes in the in-
plane lattice constant a for Cr (Re) octahedra, while the decrease of the perpendicular
bond length to be 25% of the value of the c lattice constant for both octahedra. The
rotation of the octahedra is similar for both cations and changes by 2.3° across our strain
range (Supplemental Material Table B.1 in Appendix B). These structural changes, as
shown below, drastically affect the magnetic behaviors of the Sr2CrReO6 films.
53
(a) 6
(c) Cr
5.5
LSAT
Re
(deg)
5
SCNO/LSAT
4.5
STO
4
3.5
1.99 (d) d eq(Re-O)
d ax(Re-O)
1.95
d eq(Cr-O)
1.93 d ax(Cr-O)
0.99 1 1.01 1.02
compressive c/a tensile
FIG. 2
Figure 5.2 Schematics of (a) Cr−O (counterclockwise) and Re−O (clockwise) octahedral
rotations in the Sr2CrReO6 lattice and (b) an oxygen (red) octahedron surrounding Cr or
Re with two different bond lengths, deq and dax. (c) Octahedral rotation angle θ and (d)
bond lengths of Cr−O and Re−O octahedra as a function of the tetragonal distortion c/a
of the Sr2CrReO6 lattice.
To characterize the MCA in our Sr2CrReO6 films, we measure the magnetic hysteresis
out-of-plane hysteresis loops at temperatures T = 20, 100 and 300 K for the three films
(including the buffer layer) exhibit diamagnetic responses at the temperatures discussed
54
We note three distinct features from these hysteresis loops. First, the magnetic easy axis
changes from in plane for the films on LSAT and STO with compressive strain (c/a > 1)
to out of plane for the film on SCNO/LSAT with tensile strain (c/a < 1). This is similar to
the observation in strained Sr2FeMoO6 films [21] since MCA favors the magnetic easy
axis along a shorter axis of the tetragonally distorted crystal lattice. A recent theoretical
calculation predicts the change in sign of the MCA energy in Sr2CrReO6 films at a tensile
strain of 0.7% using the local-spin-density approximation (LSDA) or 0.3% using the
GGA method [18]. For our Sr2CrReO6 film on SCNO/LSAT with a tensile strain of
0.9%, the magnetic easy axis is indeed out of plane and complements the theoretical
predictions.
55
(a)
1 T = 20 K T = 100 K T = 300 K
H || film
M ( B/f.u.)
0.5
0
H film
-0.5
-1 SCRO/LSAT
-6 -3 0 3 6 9 12 15 18 -6 -3 0 3 6 -6 -3 0 3 6
H (T) H (T) H (T)
(b) H || film
1 T = 20 K T = 100 K T = 300 K
M ( B/f.u.)
0.5
H film Hu
0
-0.5
-1 SCRO/STO
-6 -3 0 3 6 9 12 15 18 -6 -3 0 3 6 -6 -3 0 3 6
H (T) H (T) H (T)
(c)
1 T = 20 K H film T = 100 K T = 300 K
M ( B/f.u.)
0.5
H || film Hu
0
-0.5
-1 SCRO/SCNO/LSAT
-6 -3 0 3 6 9 12 15 18 -6 -3 0 3 6 -6 -3 0 3 6
H (T) H (T) H (T)
FIG. 3
Figure 5.3 In-plane and out-of-plane magnetic hysteresis loops for the Sr2CrReO6 films
grown on (a) LSAT, (b) SrTiO3 and (c) Sr2CrNbO6/LSAT in magnetic fields up to 7 T at
T = 20, 100 and 300 K. Magnetic anisotropy fields can be obtained from the intercept of
linear extrapolation of the hard-axis hysteresis loop and the saturation magnetization as
shown in the left panels.
Second, the MCA indicated by the hysteresis loops along the hard axis is very large for
all three films. In the left panel of Fig. 5.3(b) for Sr2CrReO6/SrTiO3, we extrapolate the
high-field region of the out-of-plane hysteresis loop to find the intercept with the
(Table 5.1). Since the magnetization is almost, but not fully saturated at 7 T [14,32], the
demagnetization field 4πMs of 1260 G [14] is much smaller than the anisotropy field and
can be neglected. We calculate the anisotropy energy by finding the area between the
easy- and hard-axis loops. We find Ku, the uniaxial MCA energy density to be Ku = 4.77
× 106 erg/cm3 (4.77 × 105 J/m3), which is very high for Sr2CrReO6 with a small Ms = 0.85
µB/f.u. (as a reminder, this is for SCRO/STO). For comparison, SmCo5 films exhibit one
of the highest Ku = 7.6 × 107 erg/cm3, while the Ms of SmCo5 is much higher than that of
Sr2CrReO6 [1]. For Sr2CrReO6/LSAT, the MCA is considerably larger than that in
Sr2CrReO6/SrTiO3, as can be seen from the much larger difference between the in-plane
and out-of-plane hysteresis loops in Fig. 5.3(a). The magnetometry data for
Sr2CrReO6/LSAT does not allow a reliable determination of Hu from the hysteresis loops
since there is very little magnetic signal from the Sr2CrReO6 film for the hard-axis loop
(see Supplemental Material Figure B.1 in Appendix B for the raw magnetic data for
Sr2CrReO6 film on SCNO/LSAT, the hard axis is in plane and the anisotropy field Hu = -
11.0 T is obtained from Fig. 5.3(c), where the negative sign indicates out of plane
anisotropy. For this film, we obtain Ku = -1.46 × 106 erg/cm3. Clearly, there is a strong
dependence of the anisotropy fields on the strain, in particular the c/a ratios of the
linear extrapolation of the anisotropy fields for the Sr2CrReO6 films on SrTiO3 (Hu = 18.1
57
T) and SCNO/LSAT (Hu = -11.0 T), we find that the MCA should change sign at 0.30%
tensile strain, which agrees with the value predicted by the GGA calculations [18].
Table 5.1 Lattice constants (a, c), tetragonal distortion (c/a), coercivity (Hc) and
magnetocrystalline anisotropy field (Hu) at T = 20 K and 300 K of the Sr2CrReO6 films
grown on LSAT, SrTiO3 and Sr2CrNbO6/LSAT; as is the in-plane substrate lattice
constant (doubled for LSAT and STO for easier comparison).
SCNO/LS
7.892 7.876 7.804 0.991 1.876 1.128 11.0 11.0
AT
Finally, the anisotropy fields remain essentially unchanged from T = 20 to 300 K, which
is promising for applications. It should be noted, however, that coercivity (Hc) decreases
with increasing temperature (Table 5.1). This can be understood as follows. The MCA is
mainly determined by the magnetic interaction, particularly SOC, and not defects in the
films. Since the magnetization of Sr2CrReO6 changes only slightly from 20 to 300 K
because of the high Tc [14,32], the MCA is essentially unchaged below 300 K.
Meanwhile, the coercivity is sensitive to defects in the films, thus Hc decreases at higher
58
To further understand the origin of the change in easy axis under strain seen in
experiment, we need to examine the difference in total energies for magnetic orientations
along different crystalline axes. This is known as magnetic anisotropy energy (MAE).
The origin of MAE resides in the spin-orbit coupling between the magnetization and the
lattice. The crystalline axis associated with the lowest energy determines the most
MAE and the difference between hard and easy axis orbital magnetic moments, which we
𝜉
MAE = − 𝛼 Δ𝑚!
4𝜇!
where ΔmL is the difference between hard- and easy-axis orbital magnetic moments, µB is
the Bohr magneton, ξ is the spin-orbit coupling constant, and the prefactor α depends on
the electronic structure and is on the order of 0.05-1 [34]. This relation was shown to be
valid when the majority spin is fully occupied [33], which is the case for Sr2CrReO6, and
subsequent studies have confirmed it [35-37]. Eq. (1) suggests that when the moment
anisotropy is calculated for two directions, a sign change in the moment anisotropy
indicates a change in easy axis. Performing GGA+U calculations including spin orbit
coupling, for the different substrate-induced strain states (Supplemental Material Table
B.1 in Appendix B), we indeed find that Δmtot (defined as the difference between the
magnetic moment in the [001] direction and the average magnetic moment in the [100]
and [010] directions) changes sign with increasing c/a ratio [Fig. 5.4(a)], in agreement
with our experiments [Fig. 5.3]. This agreement confirms the validity of our theory
59
within numerical uncertainties. The calculated Δmtot values are much larger, by 1-2 orders
of magnitude, than those found in Ref. [35] for Ni and Fe, demonstrating the robust
nature of the magnetic anisotropy in Sr2CrReO6. However, the Δmtot values are small
enough that the precision required to measure them experimentally is not accessible by
Figure 5.4 Differences (∆m) between out-of-plane ([001] direction and in plane (average
of [100] and [010] directions (a) total magnetic moments, (b) total orbital and spin
magnetic moments, and (c) spin and (d) orbital magnetic moments for Re (squares), Cr
(diamonds), and O (circles). The lines are guides for the eye.
Since the easy axis corresponds to the most energetically favorable axis (lowest energy),
Eq. (1) tells us that the easy axis aligns with the axis where the orbital magnetic moment
60
is the largest [38]. This overall agreement allows us now to examine the origin of the
easy-axis change. For that, we first separate the total moment anisotropy in spin (ΔmS)
and orbital (ΔmL) components [Fig. 5.4(b)] and find that they have approximately linear
strain dependence with opposite slopes, with the orbital moment anisotropy dominating.
Separating these contributions further according to their atomic origins, we find that for
the spin moment anisotropy [Fig. 5.4(c)], Cr and Re contribute at approximately the same
magnitude, but with opposite sign, thus canceling each other and leaving the small
contribution from the hybridizing O atoms as the dominant spin component. We have
K edge (work performed at the Advanced Photon Source, Argonne National Laboratory),
complementing our theoretical findings that the O site carries a substantial magnetic
moment [39]. The overall dominant orbital moment anisotropy, however, is nearly
exclusively caused by the Re atoms, which contribute 93% to the combined ΔmL [Fig.
5.4(d)]. The calculated spin and orbital moments for the Cr, Re, and O atoms in
Sr2CrReO6 for each strain state (and both in plane and out of plane) can be found in
Having identified the Re atoms as the origin of the magnetic anisotropy, we now can
further look into the orbital origin of the observed magnetic anisotropy. For that, we
examine the changes in calculated Re t2g (xy, yz, and xz) density of states (DOS) near the
Fermi level for tensile (c/a = 0.99) and compressive (c/a = 1.025) strain (Supplemental
Material Fig. B.3 in Appendix B). We observe a downward shift in energy for the xy
DOS (Supplemental Material Fig. B.3(a) in Appendix B) and an upward shift in energy
61
for both yz and xz DOS (Supplemental Material Fig. B.3(b) in Appendix B) when
In conclusion, we reveal large magnetic anisotropy and achieve dramatic changes in the
MCA of Sr2CrReO6 films via epitaxial strain. We use DFT calculations to understand the
structural distortions and to elucidate the dependence of MCA on the structure of the
films. A switching of the magnetic easy axis from in plane for compressive strain to out
of plane for tensile strain is observed via superconducting quantum interference device
(SQUID) magnetometry, and Sr2CrReO6 films exhibit some very high anisotropy fields.
Finally, we use first principles calculations to probe the atomic origins of the large and
tunable MCA in Sr2CrReO6, for which the anisotropy is driven primarily by Re orbitals.
A thorough understanding of the atomic magnetic behavior of this complex system can
guide the design of other versatile and applicable materials. Our results suggest that
substitution of other heavy transition metals, such as W or Os, for Re an drastically affect
References
[1] M. Seifert, V. Neu, and L. Schultz, Appl. Phys. Lett. 94, 022501 (2009).
[2] T. Shima, K. Takanashi, Y. K. Takahashi, and K. Hono, Appl. Phys. Lett. 85,
2571 (2004).
[3] J. H. Haeni, P. Irvin, W. Chang, R. Uecker, P. Reiche, Y. L. Li, S. Choudhury, W.
Tian, M. E. Hawley, B. Craigo, A. K. Tagantsev, X. Q. Pan, S. K. Streiffer, L. Q.
Chen, S. W. Kirchoefer, J. Levy, and D. G. Schlom, Nature 430, 758-761 (2004).
62
[4] J. Wang, J. B. Neaton, H. Zheng, V. Nagarajan, S. B. Ogale, B. Liu, D. Viehland,
V. Vaithyanathan, D. G. Schlom, U. V. Waghmare, N. A. Spaldin, K. M. Rabe,
M. Wuttig, and R. Ramesh, Science 299, 1719 (2003).
(2012).
[6] D. R. Behrendt, S. Legvold, and F. H. Spedding, Phys. Rev. 109, 1544 (1958).
[19] X. Chen, D. Parker, K. P. Ong, M. H. Du, and D. J. Singh, Appl. Phys. Lett. 102,
102403 (2013).
[22] H. L. Wang, C. H. Du, Y. Pu, R. Adur, P. C. Hammel, and F. Y. Yang, Phys. Rev.
B 88, 100406(R) (2013).
[30] V. I. Anisimov, J. Zaanen, and O. K. Andersen, Phys Rev B 44, 943–954 (1991).
[31] H.-T. Jeng and G. Y. Guo, Phys. Rev. B 67, 094438 (2003).
64
[34] O. Hjortstam, K. Baberschke, J. M. Wills, B. Johansson, and O. Eriksson, Phys.
Rev. B 55, 15026 (1997).
[35] I. Yang, S. Y. Savrasov, and G. Kotliar, Phys. Rev. Lett. 87, 216405 (2001).
[38] H.-T. Jeng and G. Y. Guo, Phys. Rev. B 65, 094429 (2002).
65
Chapter 6. Magnetic Structure in Epitaxially Strained
Note: This chapter is being presented in the format of a journal article. It was published
as Hauser, A. J., Lucy, J. M., Gaultois, M. W., Ball M. R., Soliz, J. R., Choi, Y.,
Restrepo, O. D., Windl, W., Freeland, J. W., Haskel, D., Woodward, P. M. & Yang, F.
and XMCD. Physical Review B. 89 180402(R) (2014). All of the DFT calculations
including the structural relaxations and partial density of states to find the number of
holes for each system were performed by Molly Ball with assistance from Oscar
Restrepo.
Abstract
function of epitaxial strain using magnetometry and x-ray magnetic circular dichroism
(XMCD) measurements of Cr, Re, and O sites. The in-plane magnetic moments change
significantly when tensile strain is applied. O K edge XMCD indicates the O sites carry
at least a portion of the bulk magnetization. Spin moment values measure for Cr match
calculations incorporating spin-orbit effects, while both spin and orbital moments
66
measure for Re sites are slightly higher than previously predicted. Finally, we discuss
large changes in the x-ray absorption near-edge structure that are observed at the Cr and
Sr2CrReO6 has proven to be a highly interesting double perovskite material due to the
convergence of many attractive properties: a Curie temperature well above 300 K[1-4]
significant spin-orbit interactions and correlated electron behavior due to the presence of
epitaxial films which, combined with high temperature ferrimagnetism, may be useful in
nonvolatile logic devices[3,8]. Progress toward device applications, however, will require
Theoretical modeling has suggested a double perovskite double exchange model, wherein
the Cr 3d and Re 5d orbitals are hybridized via the oxygen 2p orbitals [5,6,9,10]. In these
models, the rocksalt ordered Cr and Re atoms have oppositely aligned spin moments and
almost exclusively account for the magnetization in the system. Previously, element-
polycrystalline bulk samples have been studied [11], but to date such work on highly
characterize the Cr or O sites in bulk or thin film have been reported. The L edge x-ray
applying varying epitaxial strain to Sr2CrReO6 thin films, we can draw conclusions from
the system’s reaction to the induced distortions. These results are needed to properly test
substrates, as well as on LSAT substrate with a fully relaxed Sr2CrNbO6 (SCNO) buffer
relaxed (STO), and 1.09% tensile (SCNO/LSAT) strains upon their respective Sr2CrReO6
films. All films were grown to a thickness of 90 nm, as confirmed by modeling of Laue
Quantum Interference Device (SQUID) magnetometer was used to characterize the in-
plane magnetization.
Cr L, O K and Re L edges were then taken at the soft (Cr, O, 4-ID-C) and hard (Re, 4-ID-
D) x-ray beam lines at the Advanced Photon Source, Argonne National Laboratory. The
soft x-ray measurements were taken in the surface sensitive total electron yield mode,
with the applied magnetic field offset ten degrees from the plane of the film surface. The
hard x-ray measurements were taken in glancing angle fluorescence mode to optimize
probed volume at the higher energy of Re L edges, with the plane of the film tilted 3.8(1)
degrees with respect to the applied magnetic field. Edge energies were calibrated with
standards at known energies (e.g., Cr2O3 for the Cr L edge, and W foil for the Re L edge).
68
Systematic error introduced by the placement of the absorption edge step is the dominant
source of error in our XMCD measurements, so we have analyzed the data for the range
of all reasonable step edge centroid energies, and report the average moment and error
values consummate with the range of possible moments calculated. Within the ranges of
error we report, the choice of placing the step edge at the white line maximum intensity
will generally correspond with the lower magnitude moment values. Placing the step
edge at the maximum derivative of the XAS (center of rising edge) with respect to energy
will yield magnitude moment values in the upper end of the error range. The hole
number values used in this work are the result of theoretical calculations and cannot be
69
0.8
-0.4
LSAT
STO
-0.8 SCNO/LSAT
-8 -6 -4 -2 0 2 4 6 8
Magnetic Field (T)
Figure 6.1 In-plane magnetic hysteresis loop at T = 200 K for films on (001)-oriented
LSAT (green circles), SrTiO3 (black squares), and ~200 nm Sr2CrNbO6 buffer layer on
LSAT (blue triangles).
In-plane magnetic hysteresis loops for the three Sr2CrReO6 films at temperature T = 200
K are shown in Figure 6.1. Experimental limitations of the 4-ID-D beamline at the time
largely saturate each film at 200 K. At 3.5 T (shown in Figure 6.1 as a vertical dashed
line), the magnetization of films grown on LSAT, STO and SCNO/LSAT are 0.76, 0.79,
and 0.36 µB per formula unit (f.u.), respectively. These numbers represent the
“macroscopic” total moment of the Sr2CrReO6 unit cell and will be compared to the
individual Cr and Re moments. It should be noted that the film under large tensile strain
70
significant drop in moment is not due to loss of crystallinity or decreased Cr/Re cation
order as indicated by our previous work [3,12,13], which evidences high crystallinity in
Figure 6.2 shows a full summary of the Cr L edge, Re L edge, and O K edge XANES and
XMCD spectra for all three films. All L edge data is normalized to a 2:1 L3/L2 edge jump
ratio. Spin (mS) and orbital (mL) moments were then extracted by sum rule analysis
[14,15]. Due to the narrow 2p spin-orbit splitting in Cr and the resultant quantum
mechanical mixing of j1/2 and j3/2 excitations, a spin correction factor of 2 was applied for
all three Cr spectra [16]. To analyze these data, we use density functional theory (DFT)
calculations to calculate the number of holes for each system. We use projector
simulation package (VASP) [18,19]. Correlation effects were treated within the
and an exchange parameter J = 0.87 eV for the Cr d-orbitals [21]. We have relaxed the
atomic positions within the Sr2CrReO6 unit cells for the different strains using the
can be obtained by integrating its partial density of states up to the Fermi level, which
subtracted from 10, gives the number of holes. We obtain 5.69 (5.17), 5.72 (5.28), and
Sr2CrReO6/STO, which lead to the calculated averaged spin moments of 1.48 (-0.68),
1.50 (-0.69), and 1.49 (-0.68), respectively. The measured spin and orbital moments for
71
We again note the stark difference in the results of the Sr2CrReO6 films on LSAT and
STO as compared to the film on SCNO/LSAT. On the Re site, both the spin and orbital
moments are reduced by roughly a factor of 3, and the Cr spin moment sharply decreases
in magnitude. Given this information and the shape of the hysteresis loop, it is likely that
a strong anisotropy turns the magnetic easy axis out-of-plane and, consequently, an in-
plane magnetizations, coercive fields, and anisotropy fields, and is the subject of a more
72
10 (a) SCRO/LSAT(001) (b) SCRO/STO(001) (c) SCRO/SCNO/LSAT(001)
Cr L3
Cr L3
Cr L3
Normalized Intensity
L2 L2
L2
5
0
ms = 1.10(1) µB/Cr ms = 1.26(1) µB/Cr ms = 0.58(1) µB/Cr
m = 0.039(2) µ /Cr m = -0.041(2) µ /Cr mL = -0.012(1) µB/Cr
L B L B
-5
575 580 585 590 575 580 585 590 575 580 585 590
Energy (eV) Energy (eV) Energy (eV)
2
Re L3 Re L3
0
ms = -0.76(4) µB/Re x4 m = -0.77(3) µ /Re ms = -0.24(1) µB/Re x4
s B x4
-1 mL = 0.24(1) µB/Re mL = 0.26(1) µB/Re mL = 0.079(4) µB/Re
10.5 10.6 11.9 12.0 10.5 10.6 11.9 12.0 10.5 10.6 11.9 12.0
Energy (keV) Energy (keV) Energy (keV)
2.5
SCRO/LSAT(001) SCRO/STO(001) SCRO/SCNO/LSAT(001)
(g) (h) (i)
2.0
Normalized Intensity
1.0
0.5
0.0
-0.5
525 530 535 540 525 530 535 540 525 530 535 540
Energy (eV) Energy (eV) Energy (eV)
Figure 6.2 Normalized XANES and XMCD spectra at the (a-c) Cr L, (d-f) Re L, and (g-
i) O K edges. XANES/XMCD spectra are on top/bottom within each figure. Spin and
orbital moments are given for the Cr and Re spectra.
Figures 6.2(a) and (b) show the Cr XMCD spectra for films grown on LSAT and STO,
respectively, which indicate strongly quenched orbital moments and spin moments
73
antiparallel to that of Re, as expected in ferrimagnetic Sr2CrReO6. The respective spin
moments of 1.10(1) µB/f.u. (LSAT) and 1.26(1) µB/f.u. (STO) for Cr are smaller than our
calculated values (1.48 and 1.49, µB/f.u., respectively). However, it is important to note
that the calculated values represent a material system at T = 0 K and full magnetic
saturation. It is clear from Figure 6.1 that at T = 200 K, an applied field of 3.5 T does not
fully saturate any film in this study. From previous work, we can conservatively expect a
10-15% loss as compared to the calculated magnetization due to thermal energy, and
another 10-15% loss due to incomplete magnetization [3]. This accounts for the
difference between experiment and theory, even without considering the small decrease
Figures 6.2(d) and (e) show the Re XMCD spectra for films grown on LSAT and STO,
which yield mS and mL values matching closely with previous theoretical predictions
[10,23] as well as our calculations and bulk powder XMCD results [11]. The opposite
signs of the spin and orbital moments are as expected for a less than half-filled shell with
spin-orbit coupling. However, by applying the same considerations in the Cr spectra with
saturated” spin and orbital moment values may be slightly higher than previously
predicted.
When the combined spin and orbital moments for Cr and Re are added together, the
XMCD spectra imply total moments of 0.62(6) and 0.71(5) µB/f.u. for films on LSAT
and STO, respectively. Comparing these values to the macroscopic magnetizations found
74
by SQUID magnetometry, we find that 0.14(6) µB/f.u. (LSAT) and 0.08(5) µB/f.u. (STO)
There are two primary sources for the missing moment. It could either simply reside on
the O 2p orbitals, and/or the neglect of a magnetic dipole moment Tz in the spin sum rule
for Re may contribute to the unaccounted moment [15]. The latter is possible since
sizable spin-orbit interactions are known to occur in 5d elements such as Re [24], and can
contribute to Tz. Although strain is well known to produce effects such as octahedral
rotations, tilting, and Jahn-Teller distortions [25,26], we see a similar large moment
Between these two, our data indicate that magnetic moments on the O 2p orbitals should
oxygen 2p orbitals by measuring the O K edge spectra, as shown in Figures 6.2(g)- 6.2(i)
for all three films. Quantitatively, our DFT calculations yield a total moment of 0.02
µB/oxygen for all three epitaxial strains, which matches well with the equivalent missing
oxygen sites can be understood by the fact that when the double exchange model is
site cations (in this case, Cr and Re) mediated by each intervening oxygen [6]. Our
findings are also consistent with previous work that found that CrO2 had significant spin
uncommon for double perovskites to have oxygen with magnetic moment above this
75
range. For instance, our results are in line with previous work that found that Sr2FeMoO6
has a magnetic moment on oxygen between 0.06 µB/oxygen and 0.09 µB/oxygen [29].
0
575 580 585 590
Energy (eV)
3
(b) t2g
eg*
Normalized Intensity
Re L2
LSAT
STO
SCNO/LSAT
1
11.94 11.96 11.98
Energy (keV)
Figure 6.3 Normalized x-ray absorption spectra for different strains at the (a) Cr L2,3
edges and (b) Re L2 edge.
76
Figures 6.3(a) and (b) show the overlaid Cr L and Re L2 x-ray absorption spectra for all
three films, which result from excitation of 2p electrons to unoccupied d states. The line
shapes are sensitive to changes in coordination geometry, and the edge absorption energy
given by the inflection point is sensitive to changes in bonding and charge density
[30,31].
Figure 6.3(a) shows the Cr L3 edge, where the absorption energy increases by 0.25 eV
under both ~1% compressive and tensile strain when compared to the nominally
unstrained Sr2CrReO6 film grown on STO. The absorption peak maximum exhibits an
identical shift. The change in the Cr L3 edge energy is small compared to shifts seen in
Cr compounds due to changes in Cr formal charge [32], suggesting epitaxial strain has
only a weak influence on the Cr ground state electron density. In general, a decrease in
Cr L edge absorption energy is caused by higher charge density on Cr, which leads to a
lower effective nuclear charge. This destabilizes the electronic levels and decreases the
The Cr L edge also undergoes minor changes in spectral line shape with strain, but the
line shapes of 3d L edge spectra are complicated by multiplet effects. These multiplet
effects are caused by the overlap of partially filled core and valence wave functions in the
final state, which makes interpretation difficult by altering the line shape and intensity
ratios of the resulting spectra [33]. (The large 2p spin-orbit coupling in 5d systems
reduces the influence of multiplet effects in Re 5d L edge spectra discussed later, which
makes interpretation of Re L edge spectra more feasible.) The t2g/eg* intensity ratio in the
Cr L2 edge is ~1.03 in epitaxial Sr2CrReO6 films on LSAT and STO, but 1.11 in films on
77
SCNO/LSAT, a 7.7% change. Similarly, the L3 t2g/eg* intensity ratio increases by 5.2%
on SCNO/LSAT. Although there is a significant change in the line shape of the films
grown under 1.09% tensile strain on SCNO/LSAT, these films are complicated by the
line shape is seen in films grown on LSAT under 0.78% compressive strain. The L3/L2
integrated area branching ratios are 1.47 and 1.48 for films grown on STO and LSAT,
and 1.64 for films grown on SCNO/LSAT. Overall, there are only small changes between
films grown on LSAT and those grown on STO, and although there are slightly larger
changes in films grown on SCNO/LSAT, drawing conclusions from the Cr L edge is not
The Re L edge spectra of Sr2CrReO6 films show more significant changes. The Re L3
edge absorption energy (given by the inflection point) decreases by ~1 eV under ~1%
applied tensile or compressive strain (0.78% compressive strain: ΔE(STO – LSAT) = –1.2 eV,
1.09% tensile strain: ΔE(STO – SCNO/LSAT) = –1.0 eV). The L2 edge line shape experiences a
strained films, shown in Figure 6.3(b), and the L3 edge white line intensity also decreases
by ~45%. The L3/L2 integrated area branching ratios are 1.99 and 2.02 for films grown
states, so the intensity of the primary absorption feature (i.e., the L edge white line) is
sensitive to changes in the electronic structure. Under most circumstances, the decrease in
absorption energy and white line intensity would suggest an increase in the ground state
78
charge density of Re, though the magnitudes of the changes are unexpectedly large given
that there is no change in the formal charge of Re. The stability of the Re valence states
is supported by DFT calculations performed here, which suggest the number of electrons
per Re site increases by 0.1 from films grown on STO to films grown on LSAT.
Furthermore, any change in the Re charge would be difficult to reconcile with other
commensurate with the change in the Re valence state. However, based on the work by
Clancy et al., the change in valence state correlating to the observed absorption energy
shift would require nearly two oxygen vacancies per unit cell [34]. Such a large
material properties of the system, but we see no such effects. For instance, oxygen
vacancies have been previously reported to lead to rapid relaxation of epitaxial strain in
Sr2CrReO6 films [13]. However, off-axis x-ray diffractometry reveals the films in the
present work are coherently strained to the substrate, even at film thicknesses of 90 nm.
In the case of Sr2CrReO6, it is more likely that these changes are the result of the
breaks the degeneracy of unoccupied Re states and leads to broadening of the resulting
transitions. In a simple crystal field interpretation, tetragonal distortion splits the partially
79
a single transition with high intensity in unstrained Sr2CrReO6, there are several
These changes can occur without any change in the Re formal charge. Analogous results
have been observed in Al K-edge investigations of regular and distorted Al3+ octahedral
and splits the primary absorption feature [35]. In fact, we estimate that the integrated
white line intensity shown in that paper decreased by approximately 50% due to
L2,3 edges. We find a significant change in the in-plane magnetic moment by both
techniques when tensile strain is epitaxially applied. For nominally relaxed and
compressively strained films, Cr-site XMCD shows a slightly lower spin moment than
previously predicted, but matches well with calculations incorporating spin-orbit effects.
Re-specific XMCD yields spin and orbital moments that agree well with previous theory
and bulk powder results, but considerations of incomplete magnetic saturation suggest
calculations may underestimate the ideal spin and orbital moment values. Interestingly,
we find a small moment (0.01-0.03 µB/oxygen) at the oxygen site, which is confirmed by
both theory and experiment. Finally, comparative analysis of Cr and Re L edge XANES
80
shows that epitaxial strain leads to large changes in the Re ground state, while Cr is
largely unaffected.
References
[1] D. Serrate, J.M. De Teresa, and M.R. Ibarra, J. Phys. Condens. Matter 19, 023201
(2007).
[3] A.J. Hauser, J.R. Soliz, M. Dixit, R.E.A. Williams, M.A. Susner, B. Peters, L.M.
Mier, T.L. Gustafson, M.D. Sumption, H.L. Fraser, P.M. Woodward, and F.Y.
Yang, Phys. Rev. B 85, 161201 (2012).
[4] H. Asano, N. Kozuka, A. Tsuzuki, and M. Matsui, Appl. Phys. Lett. 85, 263
(2004).
[5] G. Vaitheeswaran, V. Kanchana, and A. Delin, Appl. Phys. Lett. 86, 032513
(2005).
[9] H. Das, P. Sanyal, T. Saha-Dasgupta, and D.D. Sarma, Phys. Rev. B 83, 104418
(2011).
[12] J.M. Lucy, A.J. Hauser, H.L. Wang, J.R. Soliz, M. Dixit, R.E.A. Williams, A.
Holcombe, P. Morris, H.L. Fraser, D.W. McComb, P.M. Woodward, and F.Y.
Yang, Appl. Phys. Lett. 103, 042414 (2013).
81
[13] A.J. Hauser, J.M. Lucy, H.L. Wang, J.R. Soliz, A. Holcomb, P. Morris, P.M.
Woodward, and F.Y. Yang, Appl. Phys. Lett. 102, 032403 (2013).
[14] B.T. Thole, P. Carra, F. Sette, and G. van der Laan, Phys. Rev. Lett. 68, 1943
(1992).
[15] C. Chen, Y. Idzerda, H. Lin, and N. Smith, Phys. Rev. Lett. 75, 152 (1995).
[20] V.I. Anisimov, J. Zaanen, and O.K. Andersen, Phys. Rev. B 44, (1991).
[21] H.-T. Jeng and G. Guo, Phys. Rev. B 67, 094438 (2003).
[25] P.M. Woodward, Acta Crystallogr. Sect. B Struct. Sci. 53, 32 (1997).
[26] P.M. Woodward, Acta Crystallogr. Sect. B Struct. Sci. 53, 44 (1997).
[27] D. Huang, H.-T. Jeng, C. Chang, G. Guo, J. Chen, W. Wu, S. Chung, S. Shyu, C.
Wu, H.-J. Lin, and C. Chen, Phys. Rev. B 66, 174440 (2002).
[29] R. Mishra, O.D. Restrepo, P.M. Woodward, and W. Windl, Chem. Mater. 22,
6092 (2010).
[30] R. Leapman, L. Grunes, and P. Fejes, Phys. Rev. B 26, 614 (1982).
82
[32] T.L. Daulton and B.J. Little, Ultramicroscopy 106, 561 (2006).
[34] J.P. Clancy, N. Chen, C.Y. Kim, W.F. Chen, K.W. Plumb, B.C. Jeon, T.W. Noh,
and Y.-J. Kim, Phys. Rev. B 86, 195131 (2012).
[35] J. van Bokhoven A., H. Sambe, D. Koningsberger C., and D. Ramaker E., J. Phys.
IV Fr. 7, 835 (1997).
83
Chapter 7. Cs2AgBiX6 (X = Br, Cl) – New Visible Light
Note: This chapter is being presented in the format of a journal article. It was published
as McClure, E. T., Ball, M. R., Windl, W. & Woodward, P. M. Cs2AgBiX6 (X = Br, Cl) –
Materials. 28 1348-1354 (2016). I performed all electronic band structure and density of
states calculations. The carrier mobility analysis was also done by me.
7.1 Abstract
The double perovskites Cs2AgBiBr6 and Cs2AgBiCl6 have been synthesized from both
solid state and solution routes. X-ray diffraction measurements show that both
compounds adopt the cubic double perovskite structure, space group Fm 3 m , with lattice
measurements reveal band gaps of 2.19 eV (X = Br) and 2.77 eV (X = Cl) that are
slightly smaller than the band gaps of the analogous lead halide perovskites, 2.26 eV for
CH3NH3PbBr3 and 3.00 eV for CH3NH3PbCl3. Band structure calculations indicate that
the interaction between the Ag 4d-orbitals and the 3p/4p-orbitals of the halide ion
84
modifies the valence band leading to an indirect band gap. Both compounds are stable
when exposed to air, but Cs2AgBiBr6 degrades over a period of weeks when exposed to
both ambient air and light. These results show that halide double perovskite
perovskite semiconductors.
7.2 Introduction
In a remarkably short period of time, metal-halide perovskites have gone from relative
constructed from metal-halide perovskites have risen from 3.8% in a dye-sensitized solar
cell configuration [1], to NREL certified 20.1% in a planar heterojunction cells [2]. The
most studied materials by far are the AMX3 perovskites where A is an alkyl ammonium
cation, such as CH3NH3+, M is Pb2+, and X is a halide ion (I−, Br−, Cl−). The allure of the
organohalide perovskites for photovoltaic and LED applications stems from the fact that
they contain cheap, earth-abundant elements and are amenable to a variety of processing
Metal-halide perovskite solar cells have already reached efficiency levels that are
develop cost effective, large scale manufacturing routes, (ii) improve the moisture
expensive encapsulation methods, and (iii) discover metal-halide alternatives that don’t
contain environmentally harmful elements like lead [3]. The first obstacle is an industrial
85
challenge that lies outside of the scope of fundamental research, but the latter two
challenges fall squarely in the realm of materials chemistry. While there are many reports
of CH3NH3PbX3 perovskites being stable for hundreds of hours when left exposed to the
air in the dark [4,5], their stability plummets when simultaneous exposed to moisture and
sunlight [6]. Regarding the presence of lead there are varying opinions in the scientific
community about the viability of working with lead. Lead is present in everyday items
like car batteries and solder, and equally toxic elements like Cd can be found in
commercial photovoltaic modules [7,8]. At the same time much of the world is trying to
move away from materials containing lead. While there may be a place for applications
of the lead halide perovskites in solar energy conversion devices, there can be little doubt
CH3NH3PbI3 or CH3NH3PbBr3 would be a boon. In short that is the goal of this work, to
studied CH3NH3PbX3 perovskites without incorporating toxic elements like Pb, Cd, Tl, or
Hg.
The Sn2+ ion is most obvious substitute for Pb2+, and tin-halide perovskites with attractive
optical and electrical transport properties can be made [9-11]. Unfortunately, the tin-
halide perovskites are extremely unstable in air, degrading over time even when handled
in an inert atmosphere glovebox. It seems unlikely that the Sn-based perovskites will
ever have the stability to be commercially viable alternatives to the Pb-based perovskites.
The role of the organic CH3NH3+ cation seems to be less critical. The optical band gaps
of CsPbBr3 (2.25 eV) [12] and CsPbCl3 (2.97 eV) [13] are quite similar to their
86
organohalide analogs CH3NH3PbBr3 (2.26 eV) and CH3NH3PbCl3 (3.00 eV), vide infra.
Furthermore, Kulbak, et al. have shown that the efficiencies of solar cells made from
CsPbBr3 are comparable to equivalent cells made from CH3NH3PbBr3, with the
advantage that the all-inorganic CsPbBr3 cells have better thermal stabilities and can
CsPbI3, which has the most suitable band gap of the CsPbX3 family for solar cell
applications, is very difficult to stabilize at room temperature [9]. Hence, the role of the
CH3NH3+ ion is largely to stabilize the perovskite form of APbI3 at room temperature.
Simple combinatorics of ionic charges in the A+M2+X3 formula does not lend much hope
that alternatives to the APbX3 perovskites with comparable band gaps, carrier mobilities,
and good moisture stability can be found. There are few if any suitable 2+ cations with a
d10s2 or d10s0 configuration that are both nontoxic and stable against oxidation.
Fortunately, the situation improves if we expand the search from ternary A+M2+X3
combinations of 1+ and 3+ ions with suitable electron configurations that are quite stable
when exposed to the air, such as Cu+, Ag+, Bi3+, Sb3+, and In3+. In this paper we
document our initial investigations into the structural and optical properties of metal-
= Br, Cl). These compounds have band gaps and moisture stability that are comparable to
87
7.3 Experimental Section
Starting materials CsCl and CsBr were prepared by reacting Cs2CO3 (99+%, Strem
Chemicals) with HCl (Sigma-Aldrich, 37%) and HBr (Fluka, ≥ 48%), respectively. The
solutions were evaporated, and the resulting solids were filtered and washed with ethanol.
AgCl and AgBr were precipitated from as prepared aqueous solutions of AgNO3
(99.9+%, Alfa Aesar) and NaCl (ACS Reagent, GFS Chemicals) or KBr (99+%, Alfa
Aesar). BiBr3 was prepared by reacting Bi2O3 (≥99.0%, J.T. Baker) with HBr (Fluka, ≥
48%). The mixture was heated until fully dissolved, evaporated to dryness, and then
filtered and washed with ethanol. Pure BiCl3 could not be prepared from this method,
hydrohalic acid and hypophosphorous acid. A mixture of 8 mL of 12.1 M HCl (or 8.84 M
HBr) and 2 mL of a 50% solution of H3PO2 was added to a round bottom flask and
heated to 120 °C. Then 1.89 mmol of AgCl and an equal amount of BiCl3 (1.41 mmol
AgBr and BiBr3) were dissolved in the hot solution. Once dissolved and well mixed, 3.78
mmol CsCl (2.82 mmol CsBr) was added to the flask, immediately triggering a
precipitate reaction. The precipitate was collected on filter paper, washed with ethanol,
Polycrystalline Cs2AgBiX6 samples were prepared via a solid state route as well by
mixing cesium, silver, and bismuth halide salts in a 2:1:1 molar ratio. Reagents were
88
ground together for 20 minutes and placed in an alumina crucible as a loose powder. The
samples were heated in a box furnace in air at 210 °C for 10 hours. It was found that at
least two heating cycles with grinding in-between was needed to obtain nearly phase pure
samples.
diffractometer (40 kV, 50 mA, sealed Cu X-ray tube) equipped with an incident beam Ge
111 monochromator and Lynx Eye position sensitive detector. Rietveld refinements of
laboratory XRPD data were carried out using TOPAS Academic software package to
determine the crystal structure [16]. A Rigaku MiniFlex II bench top X-ray powder
diffractometer (30 kV, 15 mA, sealed Cu X-ray tube) with a NaI scintillation detector
was also used for phase identification, pattern indexing, and for probing air and light
UV-Visible diffuse reflectance data were collected over the spectral range of 200-1100
(3648-element linear silicon CCD array). The spectrometer was used in conjunction with
an Ocean Optics DH-2000-BAL deuterium and halogen UV-Vis-NIR light source and a
Electronic band structure and density of states (DOS) calculations were obtained with
[17,18]. These calculations were performed using projector augmented wave (PAW)
potentials [19] based on the PBE exchange-correlation functional [20] within the hybrid
HSE hybrid approach [21-23] with an exact exchange fraction of 0.26. Cutoff energies of
89
250 and 280 eV were used for Cs2AgBiBr6 and Cs2AgBiCl6, respectively. The
7.4 Results
7.4.1 Synthesis
When synthesizing the bromine compound in the solid state, evidence for the perovskite
phase was not seen in the XRPD pattern until the sample was heated to 150 °C. At a
heating temperature of 185 °C the target compound was the dominant phase in the
sample mixture. After heating to 210 °C the yellow-orange sample was nearly phase
pure.
For the solid state synthesis of the chlorine compound, evidence of the perovskite phase
was seen upon room temperature grinding of the reactants. Since bismuth (III) chloride
is hygroscopic, the sample mixture took on moisture from the air while being ground,
causing the sample to clump together in a gray mass that had the appearance of wet clay.
The sample was left to sit on the lab bench for 10-15 minutes, and during that time the
gray mass dried out. Additional grinding led to the formation of a fine powder with a pale
yellow color, similar to the final product. Evidently the moisture taken on by the
reagents promotes reactivity and is then displaced once the double perovskite phase
forms. However, two cycles of heating to 210 °C are still needed to obtain phase pure
samples.
90
7.4.2 Crystal Structure
The refined fits to the XRPD patterns of Cs2AgBiBr6 and Cs2AgBiCl6 are shown in
Figure 7.1. Both compounds adopt the cubic double perovskite structure with Fm 3 m
space group symmetry. Reflections with odd-odd-odd Miller Indices signaling a rock salt
ordering of Ag+ and Bi3+ ions are readily apparent. Close inspection of the XRPD
91
Figure 7.1 Rietveld refinements of the XRPD patterns for Cs2AgBiCl6 (upper) and
Cs2AgBiBr6 (lower). The black dots, red and green lines are the experimental pattern,
calculated fit, and difference curve, respectively. Red tick marks at the bottom give the
expected peak positions. The inset shows the fit to the (111) and (200) reflections.
92
Table 7.1 details of the X-ray powder diffraction experiments and Rietveld refinements.
Cs2AgBiCl6 Cs2AgBiBr6
χ2 1.148 1.163
Structural parameters extracted from the refinements of Cs2AgBiCl6 and Cs2AgBiBr6 are
given in Table 7.1, while atomic coordinates and displacement parameters are given in
Table 7.2. A visualization of the double perovskite structure is shown in Figure 7.2.
connected octahedra, with Cs+ ions occupying the cuboctahedral cavities in the
framework. The double perovskite structure is then obtained by alternating Ag+ and Bi3+
Bond distances and bond valence sums are tabulated in Table 7.3. In both compounds
the Ag+ and Bi3+ ions show complete ordering into a rock-salt supercell of the simple
perovskite structure. The halide ion undergoes small displacements toward the bismuth
93
site, leading to Bi−X distances that are slightly shorter than the Ag−X distances. The
possible presence of Ag/Bi antisite disorder was investigated in the Rietveld refinements,
but the occupancies did not refine to values that indicated anything other than complete
ordering. Given the similarity of the Bi−X and Ag−X distances it is somewhat surprising
that antisite mixing between the two ions is not observed. The relatively low bond
valence sum for the Cs+ ion and rather large values of the displacement parameters (Beq)
suggest that fairly large dynamic rotations of the octahedra are taking place at room
Table 7.2 Refined atomic positions and atomic displacement parameters for Cs2AgBiCl6
(upper) and Cs2AgBiBr6 (lower, in italics).
Wyckoff
Site x y z Beq
site
2.0(2)
Ag 4a 0 0 0
1.8(2)
0.98(7)
Bi 4b 0.5 0 0
1.09(8)
2.68(9)
Cs 8c 0.25 0.25 0.25
3.4(1)
Cl 0.2513(7) 3.0(2)
24e 0 0
Br 0.2503(4) 3.66(8)
94
Figure 7.2 Refined crystal structure of Cs2AgBiCl6. The Cs+ ions are shown as gray
spheres, the chloride ions as small green spheres, while the Ag and Bi centered octahedra
are shown as blue and green polyhedra, respectively.
Ag – X (× 6) 2.708(8) 2.821(4)
Bi – X (× 6) 2.680(8) 2.814(4)
Cs 0.76 0.73
Ag 1.13 1.18
Bi 3.49 3.55
X 1.02 1.03
95
7.4.3 Optical Properties
In order to determine the optical band gaps UV-Vis diffuse reflectance spectra of the
samples were measured (Figure 7.3). Spectra for the methylammonium lead halide
samples are also shown (in red) for comparison. Aside from some minor differences seen
at photon energies above the absorption onset there are striking similarities between the
Cs2AgBiX6 double perovskites and their CH3NH3PbX3 analogs. However, the onset of
absorption is not quite as sharp in the double perovskites. We believe this change can be
attributed to the indirect band gap of the double perovskites. This hypothesis is
96
Figure 7.3 Diffuse reflectance spectra for Cs2AgBiCl6 and CH3NH3PbCl3 (top), and
Cs2AgBiBr6 and CH3NH3PbBr3 (bottom).
In order to extract an optical band gap the reflectance data were transformed to pseudo-
absorption data using the Kubelka-Munk equation, which expresses the absorbance as a
optical absorption coefficient. Kubleka-Munk plots for the silver-bismuth and lead
Method is then applied to the Kubelka Munk transformed data. It utilizes the equation
compositions). Using this approach band gaps of 2.19 eV and 2.77 eV are extracted for
Band structure calculations were performed to further investigate the electronic structure
of the two compounds. Calculations were also performed on lead halide systems for
comparison using Cs+ as the A-site cation for simplicity. The approach is justified by the
establishment fact that neither the CH3NH3+ nor the Cs+ ions make significant
contributions to the band structure near the Fermi level [12,24-26]. The space group
symmetry for the lead compounds was taken to be cubic with Pm-3m symmetry, both for
ease of comparison with the cubic double perovskites and because the average structures
of CH3NH3PbBr3 and CH3NH3PbCl3 are both cubic. The band structure diagrams for the
chlorides are shown in Figure 7.4, and the bromides in Figure 7.5.
98
Figure 7.4 Band Structure diagrams for Cs2AgBiCl6 (top) and cubic CsPbCl3 (bottom).
The Fermi energy is set to E = 0 and denoted with a dashed line.
99
The calculated band gaps are in good agreement with the experimental values (Table
7.4). While the band structures for the Cs2AgBiX6 and lead-halide systems contain many
similarities, there is one significant difference: the double perovskites possess an indirect
band gap as opposed to the direct band gap seen for the lead halide perovskites. This
comes from the fact that the valence band maximum has moved away from the (111)
Brillouin zone boundary (the R point in a primitive cubic cell and the L point in a face
centered cubic cell, respectively) to the X-point. Calculations performed with and without
spin-oribit coupling show that the band degeneracy present at the Γ-point of the
conduction band is lifted when spin-orbit coupling is included. This leads to the
emergence of a narrow heavy electron band that is separated from the two next higher
In order to estimate the carrier mobilities, we determined effective charge carrier masses
from the curvature of the band extrema (assuming parabolic bands). Along the R to X
direction for CsPbCl3 (CsPbBr3) the electron effective mass is 0.41me (0.34me) and the
hole effective mass along the same direction is 0.35me (0.37me). By comparison
(0.37me) along the L to W direction, and a hole effective mass of 0.15me (0.14me) along
the X to Γ direction. These estimates are only a qualitative comparison since they
perovskites have a less disperse conduction band, it is encouraging that estimates of the
hole effective masses for the double perovskites are lighter than their lead analogs.
100
Figure 7.5 Band Structure diagrams for Cs2AgBiBr6 (top) and cubic CsPbBr3 (bottom).
The Fermi energy is set to E = 0 and denoted with a dashed line.
The valence to conduction band transition is primary from filled halogen 3p/4p states to
antibonding Ag 5s and Bi 6p states. This can be seen by looking at the partial density of
101
state diagrams (PDOS) in Figure 7.6. The participation of both Ag and Bi orbitals to the
lower energy conduction bands is a key to maintaining some dispersion of the conduction
band. Although the valence band is largely halogen 3p/4p in character there is extensive
admixture of Ag 4d states, which results in the presence of several relatively flat bands at
energies between −1.5 and −4 eV. The presence of Ag 4d states plays a role in reducing
the band gap and is partially responsible for the indirect band gap of these phases.
102
Figure 7.6 Atomic partial density of states plots for Cs2AgBiCl6 (top) and Cs2AgBiBr6
(bottom).
To probe the stability of the compounds, samples were exposed to an ambient atmosphere
in both light and dark conditions. Each compound was loaded into a flat sample holder
103
used for XRPD measurements and then left exposed to the ambient air. Visual inspection
of those samples that were kept in the dark for two weeks, showed no apparent change.
Subsequent XRPD measurements, confirmed that no change had occurred. When the
samples were left sitting next to a large glass window, thereby exposing them to both
atmosphere and visible light, darkening of the exposed surface was observed, particularly
for Cs2AgBiBr6. It should be noted that, during the month of exposure, the samples only
received about six hours of direct exposure each day. The samples were analyzed by
XRPD and UV-Vis diffuse reflectance spectroscopy after two and four weeks exposure to
sunlight. As can be seen in Figure 7.7, the overall features of the reflectance spectrum
for Cs2AgBiCl6 remain intact, but the total reflectance decreases ever so slightly, in
agreement with the visual darkening of the surface of the sample that was observed. No
apparent change was detected in the XRPD patterns of Cs2AgBiCl6 that were exposed to
104
Figure 7.7 UV-Vis diffuse reflectance spectra showing the light stability of Cs2AgBiCl6
after two and four weeks of light exposure.
The reflectance spectrum for Cs2AgBiBr6, shown in Figure 7.8, changes drastically after
two weeks exposure to light and ambient air, accompanied by the appearance of
unidentified phases in the XRPD (see Supporting Information in Appendix C). These
observations clearly show degradation of the double perovskite phase over time when
105
Figure 7.8 UV-Vis Diffuse spectra showing the light instability of Cs2AgBiBr6 after two
and four weeks of light exposure.
7.5 Conclusions
Given the challenges of finding air-stable, non-toxic alternatives to the lead halide
perovskites, it’s highly encouraging to find double perovskites with band gaps that are
comparable to their CH3NH3PbX3 analogs. Given the intense level of interest and
research activity being devoted to the lead halide perovskites the emergence of double
While the structure and properties of Cs2AgBiX6 double perovskites are exciting as a
proof of concept there are still challenges to be overcome. The mixing of silver 4d
orbitals with the halogen 3p/4p orbitals changes the valence bands sufficiently that an
indirect band gap is observed. While not ideal for solar cell applications it should be
noted that the most widely used material in photovoltaic cells, silicon, has an indirect
106
gap, so this does not completely rule out their use in solar cells. The air and moisture
stability are promising, but the instability of Cs2AgBiBr6 to simultaneous exposure would
necessitate encapsulation for long term use. It should be noted that its stability is similar
to that of CH3NH3PbBr3 and CH3NH3PbI3. Finally, for single junction cells there is a
need to reduce the band gap. However, metal-halide perovskites with band gaps of ~2.2
eV, such as CH3NH3PbBr3 and Cs2AgBiBr6, are of interest for use with silicon in tandem
It is worthwhile to compare the properties of Cs2AgBiBr6 with those of the defect ordered
perovskites A3Bi2I9 and A3Sb2I9 (A = Cs, Rb, K), which have also been investigated as
lead-free alternatives to CH3NH3PbI3 and CH3NH3PbBr3. The band gaps of Cs3Bi2I9 (1.9
eV), Rb3Bi2I9 (2.1 eV), K3Bi2I9 (2.1 eV) and Cs3Sb2I9 (2.05 eV) are slightly smaller but
comparable to Cs2AgBiBr6 [27-29]. While the A3Bi2I9 and A3Sb2I9 compounds have
direct or near direct band gaps, their band dispersions are quite small which will translate
to heavier carrier effective masses than estimated here for Cs2AgBiBr6 [28,29]. Saparov
et al. estimated electron and hole effective masses in the [100] direction where they are
smallest, to be 0.44me and 0.60me, respectively [29]. Both values are heavier than those
estimated for Cs2AgBiBr6, particularly the hole effective mass. Although they have not
been estimated from calculations, the effective carrier masses for phases appear to be
even heavier.
Finally, it is important to remember that Cs2AgBiCl6 and Cs2AgBiBr6 are just two
replacement of Bi3+ with ions like Sb3+ and In3+, or replacing Ag+ with ions like Cu+ and
107
Au+, is an obvious strategy for tuning the electronic structure to narrow the band gap and
move back toward a direct band gap. Double perovskites such as Cs2NaSbCl6 and
Cs2NaInCl6 have been reported; however, little is known about their optical properties
[30], and chemical substitutions such as Ag+ for Na+ and Br− for Cl− have not been
explored. The results reported here provide a compelling motivation to prepare and
References
[1] Kojima, A.; Teshima, K.; Shirai, Y.; Miyasaka, T. Organometal halide
perovskites as visible-light sensitizers for photovoltaic cells. J. Am. Chem. Soc.
2009, 131 , 6050–6051.
[2] National Renewable Energy Labs (NREL) efficiency chart (2015);
http://www.nrel.gov/ncpv/images/efficiency_chart.jpg (accessed 30 October
2015).
[3] Stranks, S. D.; Snaith, H. J. Metal-halide perovskites for photovoltaic and light-
emitting devices. Nat. Nanotechnol. 2015, 10, 391–402.
[4] Kim, H.-S.; Lee, C.-R.; Im, J.-H.; Lee, K.-B.; Moehl, T.; Marchioro, A.; Moon,
S.-J.; Humphry-Baker, R.; Yum, J.-H.; Moser, J. E.; Grätzel, M.; Park, N.-G.
Lead iodide perovskite sensitized all solid state submicron thin film mesoscopic
solar cell with efficiency exceeding 9%. Sci. Rep. 2012, 2, 591.
[5] Noh, J. H.; Im, S. H.; Heo, J. H.; Mandal, T. N.; Seok, S. Il. Chemical
management for colorful, efficient and stable inorganic-organic nanostructured
solar cells. Nano Lett. 2013, 13, 1764–1769.
[6] Leijtens, T.; Eperon, G. E.; Pathak, S.; Abate, A.; Lee, M. M.; Snaith, H. J.
Overcoming ultraviolet light instability of sensitized TiO2 with meso-
superstructured organometal trihalide perovskite solar cells. Nat. Commun. 2013,
4, 2885.
[7] Fthenakis, V. M.; Kim, H. C.; Alsema, E. Emissions from photovoltaic life
cycles. Environ. Sci. Technol. 2008, 42, 2168–2174.
[8] Held, M.; Ilg, R. Update of environmental idicators and energy payback time of
CdTe PV systems in Europe. Prog. Photovoltaics Res. Appl. 2011, 19, 614–626.
108
[9] Stoumpos, C. C.; Malliakas, C. D.; Kanatzidis, M. G. Semiconducting tin and
lead perovskites with organic cations: Phase transitions, high mobilities, and near-
infrared photoluminescent properties. Inorg. Chem. 2013, 52, 9019–9038.
[10] Noel, N. K.; Stranks, S. D.; Abate, A.; Wehrenfennig, C.; Guarnera, S.;
Haghighirad, A.-A.; Sadhanala, A.; Eperon, G. E.; Pathak, S. K.; Johnston, M. B.;
Petrozza, A.; Herz, L. M.; Snaith, H. J. Lead-free organic–inorganic tin halide
perovskites for photovoltaic applications. Energy Environ. Sci. 2014, 7, 3061.
[11] Hao, F.; Stoumpos, C. C.; Cao, D. H.; Chang, R. P. H.; Kanatzidis, M. G. Lead-
free solid-state organic–inorganic halide perovskite solar cells. Nat. Photonics
2014, 8, 489–494.
[12] Stoumpos, C. C.; Malliakas, C. D.; Peters, J. A.; Liu, Z.; Sebastian, M.; Im, J.;
Chasapis, T. C.; Wibowo, A. C.; Chung, D. Y.; Freeman, A. J.; Wessels, B. W.;
Kanatzidis, M. G. Crystal growth of the perovskite semiconductor CsPbBr3: A
new material for high-energy radiation detection. Cryst. Growth Des. 2013, 13,
2722–2727.
[13] Myagkota, S.; Voloshinovskii, A.; Stefanskii, I.; Mikhailik, M.; Pashuk, I.
Reflection and emission properties of lead-based perovskite-like crystals. Radiat.
Meas. 1998, 29, 273–277.
[14] Kulbak, M.; Cahen, D.; Hodes, G. Planar How important is the organic part of
lead halide perovskite solar cells? Efficient CsPbBr3 solar cells. J. Phys. Chem.
Lett. 2015, 6, 2452–2456.
[15] Kulbak, M.; Gupta, S.; Kedem, N.; Levine, I.; Bendikov, T.; Hodes, G.; Cahen,
D. Cesium enhances long-term stability of lead bromide perovskite-based solar
cells J. Phys. Chem. Lett. 2016, 7, 167−172.
[16] Topas Academic, General Profile and Structural Analysis for Powder Diffraction
Data; Bruker AXS: Karlsruhe, Germany 2004.
[17] Kresse, G.; Hafner, J. Ab initio Molecular-Dynamics for Liquid- Metals. Phys.
Rev. B 1993, 47, 558–561.
[18] Kresse, G.; Hafner, J. Ab Initio Molecular-Dynamics Simulation of the Liquid-
Metal Amorphous-Semiconductor Transition in Germanium. Phys. Rev. B 1994,
49, 14251–14269.
[19] Blöchl, P.E. Projector augmented-wave method. Phys. Rev. B. 1994. 50, 17953-
17979.
[20] Perdew, J. P.; Burke, K.; Ernzerhof, M. Generalized gradient approximation made
simple. Phys. Rev. Lett. 1996, 77, 3865–3868.
109
[21] Heyd, J.; Scuseria, G. E.; Ernzerhof, M. Hybrid Functionals Based on a Screened
Coulomb Potential. J. Chem. Phys. 2003, 118, 8207–8215.
[22] Heyd, J.; Scuseria, G. E.; Ernzerhof, M. Hybrid Functionals Based on a Screened
Coulomb Potential. J. Chem. Phys. 2006, 124, 219906.
[23] Paier, J.; Marsman, M.; Hummer, K.; Kresse, G.; Gerber, I. C.; Angyan, J. G.
Screened Hybrid Density Functionals Applied to Solids. J. Chem. Phys. 2006,
125, 249901.
[24] Giorgi, G.; Fujisawa, J. I.; Segawa, H.; Yamashita, K. Small photocarrier
effective masses featuring ambipolar transport in methylammonium lead iodide
perovskite: A density functional analysis. J. Phys. Chem. Lett. 2013, 4, 4213–
4216.
[25] Melissen, S. T. A. G.; Labat, F.; Sautet, P.; Le Bahers, T. Electronic properties of
PbX3CH3NH3 (X = Cl, Br, I) compounds for photovoltaic and photocatalytic
applications. Phys. Chem. Chem. Phys. 2015, 17, 2199-2209.
[26] Yin, W.-J.; Yang, J.-H.; Kang, J.; Yan, Y.; Wei, S.-H. Halide perovskite materials
for solar cells: a theoretical review J. Mater. Chem. A 2015, 3, 8926-8942.
[27] Park, B. W.; Philippe, B.; Zhang, X.; Rensmo, H.; Boshchloo, G.; Johansson, E.
M. J. Bismuth Based Hybrid Perovskites A3Bi2I9 (A: Methylammonium or
Cesium) for Solar Cell Application, Adv. Mater. 2015, 27, 6806–6813.
[28] Lehner, A. J.; Fabini, D. H.; Evans, H. A.; Hebert, C. A.; Smock, S. R.; Hu, J.;
Wang, H.; Zwanziger, J. W.; Chabinyc, M. L.; Seshadri, R. Crystal and Electronic
Structures of Complex Bismuth Iodides A3Bi2I9 (A = K, Rb, Cs) Related to
Perovskite: Aiding the Rational Design of Photovoltaics, Chem. Mater. 2015, 27,
7137−7148.
[29] Saparov, B.; Hong, F.; Sun, J.-P.; Duan, H.-S.; Meng, W.; Cameron, S.; Hill, I.
G.; Yan, Y., Mitzi, D. B. Thin-Film Preparation and Characterization of Cs3Sb2I9:
A Lead-Free Layered Perovskite Semiconductor, Chem. Mater. 2015, 27,
5622−5632.
[30] Morss, L. R.; Siegall, M.; Stenger, L.; Edelstein, N. Preparation of cubic chloro
complex compounds of trivalent metals: Cs2NaMCl6 Inorg. Chem.,1970, 9, 1771-
1775.
110
Chapter 8. Summary
In this dissertation, I have described computational work that has guided and explained
experimentation for functional materials within the perovskites and double perovskites
families. I have tried to convey how a large number of crucial questions can only be
theory (DFT) calculations, and how the calculation results have led to new levels of
understanding and improved routes to discover even better materials. I have also
discussed what level of DFT calculations is necessary in order to produce reliable and
realistic properties, which often require special treatment of spin interactions, spin-orbit
For discovery of novel phenomena and great property tunability, I have then shown that
double perovskites are an ideal family of materials, since a huge number of transition
metal ions can be incorporated in their structure, frequently also alloyed with other ions,
properties, making this one of the most comprehensive materials classes existing. Within
111
• record-breaking magnetocrystalline anisotropy and its strain tunability; and
one of the most exciting new material discoveries in the field of photovoltaics.
For all these materials and effects, we have developed a detailed understanding of the
To discuss these results in more detail, we found novel magnetic behavior in Sr2CoOsO6,
which experiences antiferromagnetic ordering of its two transition metal ions completely
independently from each other. The Co and Os sublattices were found to be almost
entirely uncoupled from each other, which results in a magnetic structure that has a
ground state with non-collinear spins. DFT+U was used to extract the exchange
competing forces in this structure. It was found that the nearest neighbor Co-O-Os are
examined by density of states calculations with DFT, which showed that there is little to
no overlap between the electronic states on Co and Os atoms, leading to weak interaction.
Within the second body of work described within this thesis, I have described how thin
films of Sr2CrReO6 were found to have extraordinarily large anisotropy fields and
substrates, LSAT, STO, and SCNO/LSAT, causing strain ranging from compressive to
tensile causing tetragonal distortions in the structure and the transition metal and oxygen
octahedral to rotate optimizing the local environment. Across this strain range, the
112
magnetic easy axis was found to change from in-plane for LSAT and STO to out-of-plane
for SCNO/LSAT. While it is thus clear that the applied strain and the resulting internal
anisotropy, the exact mechanism could not be understood by experiment alone. Thus,
DFT was used better understand the exact dependencies. The origin of the change in easy
axis was identified by looking at the total energies for magnetic orientations along
different crystalline axes. We have then proposed that a simple relationship between the
magnetic anisotropy energy and the atomic moments, originally developed for metallic
materials, can be also applied to these double perovskites, and that the sign change in
moment anisotropy can be used to identify when the easy axis switches. The difference in
moment values was calculated and found to be an order of magnitude larger than any
the Re atoms were identified to be the origin of the magnetic anisotropy, since all other
calculated DOS that there was an opposite shift in energy between the in-plane and out-
of-plane d-orbital peaks, indicating a shift in electron distribution, which then is at the
The magnetic configuration and its changes with strain in Sr2CrReO6 analyzed were then
theoretical results, especially the large strain dependence of the Re ground state. As an
interesting side result, it was confirmed that a small moment (0.01-0.03 µB/oxygen)
should reside at the oxygen site, which has been also found in our calculations. This was
113
an important result with relevance to many other calculations for magnetic oxide
materials, since it wasn’t clear previously if this was real or an artifact from partitioning
Lastly, I explored the lead-free halides, Cs2AgBiBr6 and Cs2AgBiCl6. These double
perovskites are a part of a very promising class of materials. They were investigated to
help solve several obstacles still remaining in the commercialization of perovskite solar
cells, including high fabrication costs, moisture instabilities, and environmentally harmful
element inclusion. Finding a material that doesn’t have these issue and also remains as
efficient as existing solar cells has been a challenge. DFT calculations including spin-
orbit coupling were performed and optimized to give results very similar to the
experimentally measured electronic properties. Band structures and density of states were
presented and compared to the lead-containing analogs of these materials. While it was
found that the band gaps were comparable in size to that of the lead containing halides,
the materials possessed an indirect band gap, which is not ideal for efficient solar cells.
Since we could show that these materials can be described well within optimized DFTs,
materials with direct band gaps that one day may revolutionize solar cell materials in a
114
Appendix A. Supplementary Information for Chapter 4
Figure A.1 Refined powder X-ray diffraction pattern of Sr2CoOsO6. Data was collected
using a Bruker D8 Advance equipped with a copper source and Ge (111)
monochromator. Black symbols observed data, while red and blue curves indicate the
calculated pattern and difference respectively.
115
A.2 Neutron Powder Diffraction
Neutron diffraction data was used to refine the crystal structure of Sr2CoOsO6 at 300,
130, 80 and 12 K. The magnetic structure was also refined at the latter two temperatures.
Using peak splitting and systematic absences the space group symmetry at 300 and 130 K
The results of the Rietveld refinements are given in Table A.1. Isotropic displacement
parameters were used for Sr, Co, and Os atoms, and anisotropic displacement parameters
for O atoms.
116
Table A.1 Refined atomic positions and displacement parameters for longer measuring
time data sets where applicable. Os atomic positions for both space groups remain at
Wyckoff site 2a (0,0,0) while Co atomic positions remain at Wyckoff site 2b (0,0,½) for
both space groups. Ueq is defined as one third of the trace of the Uij tensor, and it is given
for the oxygen atoms for which anisotropic displacement parameters were refined.
Constraints were used to avoid correlations among the variables that describe the
direction of the magnetic moment. The components of the moments in the z-direction for
both osmium and cobalt were fixed to be zero based on the fact that any moment oriented
117
in this direction produced intensity for Bragg peaks that are not present in the observed
diffraction patterns. In addition, mx1 = my2 = −mx1 = −my2 where subscripts x and y
indicate the component of the magnetic moment along those respective directions, and
the subscripts 1 and 2 refer to the two different osmium atoms in the magnetic unit cell.
Similar constraints were used for the cobalt moments, mx3 = mz4 = −mz3 = −mx4 and my3
= my4 = 0, where subscripts 3 and 4 refer to the two different cobalt atoms in the unit cell.
Table A.2 Select bond lengths and angles for Sr2CoOsO6 as obtained from Rietveld
refinements of the neutron diffraction data.
118
Figure A.2 Rietveld refinement histograms at (a) 300 K and (b) 12K where black
symbols represent observed data points, red curves for the refined fit and blue for the
difference curve. Pink hash marks indicate reflections from the vanadium sample can,
black tick marks indicate reflections from the nuclear structure, and green tick marks
correspond to magnetic reflections.
119
200
150
Temperature (K)
100
50
d spacing (Å)
Figure A.3 Colorfill plot of observed variable temperature NPD data revealing the
splitting of the (220) peak which is associated with the structural phase transition
lowering the symmetry of the structure from tetragonal I4/m to monoclinic I2/m at
approximately 110K.
Figure A.4 The structure of Sr2CoOsO6 at 12 K in space group I2/m. The view looking
down [001] on the left shows out-of-phase octahedral rotations about the c-axis, while the
view looking down [110] on the right shows the lack of octahedral tilting about the a- or
b-axes (right). The higher temperature I4/m structure has an identical tilting pattern.
120
A.3 Electrical Conductivity Measurements
with decreasing temperature. This data was found not to follow a linear relationship with
T-1 as one would find with activated transport in a conventional semiconductor, but rather
to be linear with T-1/4 consistent with variable range hopping transport, whereby electrons
hop between localized states. These linear relationships can be fitted via equation A.1
below to extract the A coefficient. There is a deviation from the linear fitting of the
model which occurs near the structural phase transition, leading to two separate linear
fits, above and below this deviation. The fitted A factors found above and below this
!/! )
σ=σ0𝑒 (!!/! (A.1)
121
Figure A.5 Electrical transport behavior of Sr2CoOsO6 as shown by log resistivity vs
temperature.
122
A.4 Density Functional Theory Calculations
Additional details about the DFT calculations and the method used to fit the exchange
configuration, where the Co and Os spins order independently with the Co (Os) moments
aligned parallel or antiparallel along a particular direction in the ac- (ab-) plane. Based
and Os spins are hypothesized to be non-interacting or only weakly interacting with each
other, it would be a reasonable argument to make their moments collinear, i.e. make them
align parallel or antiparallel while lying on the same plane. In order to test this
spin−orbit coupling effects) with ten different magnetic configurations: (a) ferromagnetic
(FM), where the spins on all the Co and Os ions are aligned parallel to each other, (b)
ferrimagnetic (FiM) (AF1), where the spins on Co ions are aligned parallel to each other
and antiparallel to the spins on Os ions (consequently the spins on Os ions are aligned
parallel to those on other Os ions), (c) C-type antiferromagnetic, with chains of parallel
aligned Co and Os ions along the c-axis and the neighboring chains being aligned
antiparallel to each other, (d) A-type antiferromagnetic (AF2), where the spins on Co and
Os ions lying on any equatorial plane (ab-plane) are parallel to each other while the
ordering between any two neighboring ab-planes is anti-parallel, (e) the collinear version
in Figure A.11.The calculations were performed with the following three combinations of
(UCo, UOs): (0, 0), (4.1, 2.1) and (4.1, 4.1) eV. For all the three combinations, the energy
of Exp(col) was found to be the lowest making it the ground state. This also lends support
to our conjecture that the Co and Os spins might be non- or only weakly interacting. The
relative energies of the other four configurations with respect to the ground state
configurations calculated using UCo = 4.1 eV and UOs = 2.1 eV are shown in Table A.3.
Figure A.7 shows the spin- and atom-resolved densities of states (DOS) for the ground
state magnetic configuration for calculations (a) with GGA, i.e excluding strong
correlation effects and (b) with UCo and UOs = 4.1 and 2.1 eV, respectively. Although the
two calculations predict the same ground state magnetic configuration energetically, from
Figure A.7 it can been seen that inclusion of strong-correlation effects on both Co and Os
experimentally. Since the three combinations of UCo and UOs give similar results when
comparing the energies of different magnetic configurations and as UCo = 4.1 eV and UOs
= 2.1 eV also reproduce the experimentally observed insulating nature, we use this
particular combination of UCo and UOs for the remainder of the work.
124
Figure A.6 16 formula-unit cell of Sr2CoOsO6 showing (a) the experimentally observed
magnetic configuration which is non-collinear and (b) the collinear version of the
experimentally observed magnetic configuration where all the moments have been
aligned along the [001] direction. Only Os (red) and Co (blue) atoms are shown for
clarity. The arrows show the direction of the magnetic moment on every Co and Os ion.
Figure A.7 Spin- and atom- resolved density of states with collinear arrangement of
independently ordered Co and Os sublattices obtained using (a) GGA without Hubbard-U
and (b) with UCo = 4.1 eV and UOs = 2.1 eV.
125
Table A.3 Relative energy of different magnetic configurations examined in this work,
along with the spin and orbital-moments of Co atoms and OsO6 units obtained for each
configuration.
126
Figure A.8 Spin- and orbital-resolved density of states of (a) Co and (b) Os in the 16
formula-unit cell of Sr2CoOsO6 with collinear arrangement of independently ordered Co
and Os sublattices obtained using UCo = 4.1 eV and UOs = 2.1 eV.
Figure A.8 shows the orbital- and spin-resolved DOS for the Co and Os atoms for the
ground state configuration. With this particular combination of U, we find the magnetic
moments on the Co and Os ions to be 2.72 and 1.43 µB/ion, respectively. In addition, we
also find the O ions to have a magnetic moment of 0.07 µB/ion, with their sign being
equal to that of the Os ion to which it is connected to, which implies significant charge
sharing between the extended 5d states of the Os ions and the 2p states of the O ions
within the OsO6 octahedra, considering that the six O ions together contribute a moment
of 0.42 µB/OsO6 unit. On adding the magnetic moments of an Os ion and its neighboring
127
O atoms, we get a magnetic moment of 1.82 µB/ion for every OsO6 octahedron. These
values are in good agreement with the experimentally (from neutron powder diffraction)
determined values. Since these calculations were performed without including spin-orbit
Heavy ions like Os, which has an atomic number Z = 76, can have strong spin-orbit
interactions. This is because the magnitude of spin-orbit coupling (SOC) depends on the
the 16 f.u. cells by including SOC as implemented within VASP, wherein fully-
relativistic calculations are performed for the core electrons and valence electrons are
treated in a scalar relativistic approximation. Since SOC couples the electron spin to the
different calculations on the 16 f.u. cells by including SOC effects: (a) A collinear
calculation with the Exp(col) configuration as discussed above with all moments aligned
along the c-axis, and (b) a non-collinear calculation with the moments on each ion
oriented along the experimentally observed directions, i.e. the Co and Os moments lying
along a direction in the ac- and ab-plane, respectively. From Table A.3, we can see that
the structure with non-collinear arrangement of moments has the lowest energy, followed
by the Exp(col) configuration with and without SOC, respectively. Figure 4.4 in the
Chapter 4 shows the total and atom-resolved density of states for the non-collinear
128
configuration. The spin- and orbital-moment of the Co atoms and OsO6 units for the two
As expected, the OsO6 units have a larger orbital moment, 0.72 µB, than Co with an
orbital moment of 0.25 µB/ion, due to the higher Z of Os,. The total magnetic moment
(spin+orbital) on the Os ions can be represented by the vector ±(0.97 xˆ − 0.82 yˆ − 0.1zˆ )
µB/ion, where x̂ , ŷ and ẑ are unit-vectors along the three Cartesian axes. Thus, the Os
moments lie predominantly on the ab-plane, similar to the experimental results where the
moments can be described with the vector ±(1.28 xˆ − 1.28 yˆ + 0 zˆ ) µB/ion. On the other
hand, we find two different sets of Co ions with one set having their moments as
±(2.01xˆ − 0.04 yˆ − 1.82 zˆ) µB/ion and the other having ±(2.27 xˆ − 0.45 yˆ − 1.41zˆ) µB/ion.
While the former set of ions have their moments predominantly along the ac plane,
similar to the experimental moments that can be described with the vector
±(2.05 xˆ + 0 yˆ − 2.05 zˆ) µB/ion, the latter set of Co ions have a small component of
magnetization along the b axis. The total moment on the two different sets of Co ions
however is the same. Although the absolute value of the moments is known to change
strongly with the U values, it is important to note that the arrangement of the moments
for the non-collinear calculation match well with the experiments, with the exception that
our calculations suggest two different sets of Co ions, one with its moments lying on the
ac plane, while the other set of ions have a small component along the c axis, whereas
experimentally, only one set of Co ions was observed. Thus overall, the DFT calculations
129
Co and Os ions with their ordering pattern similar to that of Exp(col) configuration, to be
the most stable configuration and are in good agreement with the experiments.
The total DOS of the three lowest energy configurations, i.e., Exp(col), Exp(col) + SOC
and the non-collinear configuration Exp(non_col) are shown together for comparison in
Figure A.9. There are some changes in the peak positions near the Fermi energy,
especially between the non-collinear and the collinear calculations. Their overall features
are however very similar. Also the ordering between of the moments within the Co and
Os sublattices is similar for the two configurations. These combined with the fact that
even while excluding SOC effects, Exp(col) is found to be most stable amongst other
130
Figure A.9 The total DOS for the 16 f.u. cell with (a) Exp(non_col), (b) Exp(col)+SOC
and (c) Exp(col) magnetic configurations.
In order to shed light on the reason behind this highly unusual magnetic coupling in
between the Co and Os ions, with the interaction pathways shown in Figure A.10 [3]. In
addition to FM, FiM (AF1), A-type AFM (AF2), C-type, and Exp(col), we have used
four additional magnetic configurations labeled as AF4-AF7 , all of which were done
using 16 f.u. supercells and are shown schematically in Figure A.11 and listed in Table
131
A.5. We have used a Heisenberg model to express the total energy of each magnetic
configuration, within a 2 f.u. cell, as a sum of a non-magnetic part with energy E0 and
H = E0 − ∑ J ij Si ⋅ S j , (A.2)
i< j
where Jij are the exchange constants for interaction between spin Si and Sj at sites i and j,
respectively. If Jij is positive (negative) it favors parallel (antiparallel) coupling of the two
spins. For simplicity, we have expressed Jij in terms of J1−J7 for different pairs of Co−Co,
Co−Os, and Os−Os interactions, as shown in Figure A.10. As discussed earlier, the
magnetic moments of Co and OsO6 units for the Exp(col) calculation are 2.72 and 1.82
µB/ion, respectively, and remain within ±0.2 µB for all the other configurations, roughly
corresponding to S = 3/2 for Co and S = 1 for Os. These moments are indicative of a
valence state of Co2+ in its high-spin and Os6+ in its low-spin configuration and also agree
132
Figure A.10 Schematic showing the different exchange interactions between Co and Os
ions.
The Hamiltonian above has eight unknown parameters: E0 and J1 – J7. Hence, using the
eight magnetic configurations discussed above, we have solved the set of Eqs. A.3−A.10
2 2
EFM = (−4 J 4 − 6 J 2 − 8 J 5 ) SCo + (−8 J 6 − 6 J 3 − 4 J 7 ) SOs − 12 J1SCo ⋅ SOs (A.3)
2 2
EAF1 = ( −4 J 4 − 6 J 2 − 8 J 5 ) SCo + ( −8 J 6 − 6 J 3 − 4 J 7 ) SOs + 12 J1SCo ⋅ SOs (A.4)
2
EAF2 = ( −4 J 4 − 6 J 2 + 8 J 5 ) SCo 2
+ (−6 J 3 + 4 J 7 ) SOs − 4 J1SCo ⋅ SOs (A.5)
2
EC-type = (−4 J 4 − 6 J 2 + 8J 5 )SCo 2
+ (−6 J 3 + 4 J 7 ) SOs + 4 J1SCo ⋅ SOs (A.6)
2
EAF4 = ( −2 J 4 − 4 J 2 − 4 J 5 ) SCo 2
+ (−8 J 6 − 6 J 3 − 4 J 7 ) SOs − 9 J1SCo ⋅ SOs (A.7)
133
2
EAF5 = (−4 J 4 − 6 J 2 − 8 J 5 ) SCo 2
− 6 J 3 SOs − 6 J1SCo ⋅ SOs (A.8)
2
EAF6 = ( −2 J 4 − 2 J 2 ) SCo 2
+ (−2 J 6 − 2 J 3 ) SOs − 4 J 1S Co ⋅ S Os (A.9)
2
EAF7 = ( −2 J 4 − 4 J 2 + 4 J 5 ) SCo 2
+ (−4 J 3 + 2 J 7 ) SOs + 2 J1SCo ⋅ SOs (A.10)
In order to compare to the relative strengths of the spin interactions between the Co2+
(S=3/2) and Os6+ (S=1), we use their effective spin exchange constants J ijeff = Si S j J ij ,
Table A.4 Spin exchange parameters J ijeff = Si S j J ij for Sr2CoOsO6 (in meV).
134
Table A.5 Comparison of energy per f.u. of the different magnetic configurations
135
Figure A.11 Schematic showing the spin arrangements in two neighboring ab planes in
Sr2CoOsO6.
136
References
3. Tian,C.; Wibowo, A. C.; Zur Loye, H.-C.; Whangbo, M-H. On the Magnetic
Insulating States, Spin Frustration, and Dominant Spin Exchange of the Ordered
Double-Perovskites Sr2CuOsO6 and Sr2NiOsO6: Density Functional Analysis
Inorg. Chem. 2011, 50, 4142-4148
137
Appendix B. Supplementary Information from Chapter 5
Sr2CrReO6 (001) epitaxial films are deposited on three kinds of substrates or buffer
(SCNO) buffer layer on LSAT. The Sr2CrReO6 (SCRO) films are grown using an off-
axis ultrahigh vacuum (UHV) sputtering technique we recently developed for epitaxial
and Sr2CrReO6), Yttrium ion garnet (Y3Fe5O12 or YIG), Heusler compounds (e.g.
Co2FeAl0.5Si0.5). The SCNO buffer layer on LSAT was deposited to a thickness of ~200
nm to allow the SCNO buffer layer to relax to near its bulk lattice parameters away from
700 °C using a DC power source at a constant current of 55 mA. For optimal deposition,
we used a total O2/Ar sputtering gas pressure of 12.5 mTorr with an oxygen partial
pressure of 26.25 µTorr. The Sr2CrNbO6 buffer layer was deposited at Ts = 600 °C using
mTorr; the oxygen partial pressure was 17.5 µTorr. The deposition rates of Sr2CrReO6
138
B.2 Diamagnetic Background Subtraction in Squid Magnetometry
Measurements
First, we show in Supplemental Material Figure B.1 the raw magnetometry data taken at
T = 20 K for the Sr2CrReO6 film on LSAT. It is readily seen that the film gives almost no
distinguishable signal above that of the diamagnetic LSAT substrate for the hard axis
(field applied perpendicular to the film) hysteresis loop. Additionally, the easy axis loop
does not show a linear high field region. This implies that the film on LSAT is not nearly
ferrimagnetic response for the SCRO film (after diamagnetic substrate background
subtraction) exhibiting a high field magnetization close to the expected value (of which
we have prior knowledge). As a result of the manual fitting of the LSAT substrate
background and low out-of-plane magnetic signal from the SCRO/LSAT film, we are
unable to extrapolate the anisotropy field Hu for SCRO/LSAT and simply estimate it to
139
Figure B.1 Raw magnetic data for the 90 nm SCRO/LSAT film taken with a Quantum
Design SQUID for both in-plane and out-of-plane geometries at T = 20 K. The hard axis
loop (field applied perpendicular to the film) shows almost no signal differentiating the
SCRO film from the background of the LSAT substrate, so a reliable anisotropy field
analysis could not be performed. We estimate Hu to be 10's of tesla.
Material Figure B.2 the raw magnetic data for the 90-nm Sr2CrReO6 film on SrTiO3.
Magnetometry measurements are taken in both in-plane (red) and out-of-plane (blue)
geometries. Raw data includes the magnetic signal from the film as well as the
diamagnetic background from the SrTiO3 substrate. The diamagnetic background of the
substrate is not expected to have orientation dependence. We fit the substrate background
using the high field regions of the easy-axis measurement (in-plane) because at high
fields the SCRO film is almost saturated and thus the high field region data gives the best
estimate of the diamagnetic background slope. Supplemental Material Figure B.2 shows
the substrate background (green) slope matches the SCRO/STO high field region slope.
140
After the substrate background is subtracted from the raw data, we normalize the raw
signal (in µemu) to the sample volume and present the data in units of µB/f.u., where f.u.
Figure B.2 Raw magnetic data for 90-nm SCRO/STO film taken with a Quantum Design
SQUID magnetometer for both in-plane and out-of-plane geometries at T = 20 K. The
(linear) diamagnetic background of the substrate is shown.
Application of this analysis to the SCRO/LSAT film [panel (a) of Fig. 5.3 in the Chapter
5] results in negative magnetizations at positive fields, which suggests the film (in the
easy-axis in-plane measurement) is not nearly saturated at fields of 7 T and our fitting
procedure is not applicable. We manually fit the LSAT substrate background to acquire a
ferrimagnetic response exhibiting a high field magnetization close to the expected value.
As a result of the manual fitting of the LSAT substrate background and low out-of-plane
141
magnetic signal from the SCRO/LSAT film, we are unable to extrapolate the anisotropy
field HU for SCRO/LSAT and simply estimate it to be at least 10’s of tesla (T).
For the SCRO/SCNO/LSAT, we use the same analysis procedure to fit the substrate
background, which includes contributions from both the SCNO double perovskite buffer
layer and LSAT substrate. The easy axis in this case is along the axis normal to the film
surface and so we fit the substrate background using the high field regions of the out-of-
See Supplemental Material Table B.1 for DFT-calculated rotation angles of the octahedra
Table B.1 DFT-calculated rotation angles of the Cr and Re oxygen octahedra as well as
equatorial (deq) and axial (dax) lengths of the Cr–O and Re–O bonds for the Sr2CrReO6
films grown on LSAT, SrTiO3 and Sr2CrNbO6/LSAT.
142
B.4 Atomically Resolved Spin and Orbital Magnetic Moments for
SCRO Films on LSAT, STO, and SCNO/LSAT
See Supplemental Material Table B.2 for calculated spin and orbital moments for the Cr,
Re, and O atoms in SCRO for each strain state (and both in-plane and out-of-plane).
Table B.2 First principles calculated in-plane (average of [100] and [010] and out-of-
plane ([001]) spin and orbital magnetic moments for Cr, Re, and O atoms in SCRO films
of different strain states.
See Supplemental Material Figure A.3 for the calculated DOS of Re t2g (a) xy and (b) yz
and xz suborbitals near the Fermi energy in SCRO for tetragonal disortions c/a = 1.025
143
Figure B.3 Calculated DOS of Re t2g (a) xy and (b) yz and xz suborbitals near the Fermi
energy in SCRO for tetragonal distortions c/a = 1.025 and c/a = 0.99.
144
Appendix C. Supplementary Information for Chapter 7
A comparison of the calculated geometry optimized lattice parameters compared with the
Table C.1 An HSE hybrid functional with spin orbit coupling was used for the
computations.
The diffuse reflectance plots have been transformed into pseudo absorbance using the
145
Figure C.1 Kubleka-Munk plots for the silver-bismuth (blue) and lead compounds (red),
with chloride compounds on the left and bromide compounds on the right.
Plots of the X-ray powder diffraction patterns of the Cs2AgBiCl6 and the Cs2AgBiBr6
samples after exposure to air in the dark for 2 weeks, exposure to both air and light for 2
146
Figure C.2 XRPD overlays of Cs2AgBiCl6 (upper) and Cs2AgBiBr6 (lower) showing
the light stability of the compounds after a month of exposure to sunlight. The change in
the powder diffractogram for the bromine compound indicates that structural changes are
accompanying the visual darkening of the pigment.
147
References
1. Piskunov, S., Heifets, E., Eglitis, R. . & Borstel, G. Comput. Mater. Sci. 29, 165–
178 (2004).
2. Van Benthem, K., Elsa sser, C. & French, R. H. J. Appl. Phys. 90, 6156 (2001).
6. Lufaso, M. W. & Woodward, P. M. Acta Crystallogr. Sect. B Struct. Sci. 60, 10–
20 (2004).
7. Popov, G., Greenblatt, M. & Croft, M. Phys. Rev. B 67, 024406 (2003).
8. Kim, B. G., Hor, Y. S. & Cheong, S. W. Appl. Phys. Lett. 79, (2001).
10. Mishra, R., Soliz, J. R., Woodward, P. M. & Windl, W. Chem. Mater. 24, 2757–
2763 (2012).
13. Woodward, P. M., Karen, P., Evans, J. S. . & Vogt, T. Functional Materials.
(Unpublished).
148
14. Mishra, R., Restrepo, O. D., Woodward, P. M. & Windl, W. Chem. Mater. 22,
6092–6102 (2010).
15. Popov, G., Greenblatt, M. & Croft, M. Phys. Rev. B 67, 024406 (2003).
16. Vasala, S. & Karppinen, M. Prog. Solid State Chem. 43 1-36 (2015).
3. Sholl, D.S. & Steckel, J.A. Density Functional Theory: A Practical Introduction
(John Wiley and Sons, Hoboken, NJ, 2009).
4. Himmetoglu, B., Floris, A., de Gironcoli, Stefano & Cococcioni, M. Int. J. Quant.
Chem. 114, 14-49 (2014).
3. Lufaso, M.W.; BarnesP. W.;Woodward, P.M. Acta Cryst. B 2006, 62, 397-410.
5. Rogado, N. S.; Li, J.; Sleight, A. W.; Subramanian, M.A. Adv. Mater. 2005, 17
(18), 2225-2227.
6. Sleight, A. W.; Longo, J.; Ward, R. Inorg. Chem. 1962, 1 (2), 245–250.
149
7. Macquart, R; Kim, S. J.; Gemmill, W. R.; Stalick, J. K.; Lee, Y.; Vogt, T.; Zur
Loye, H. C . Inorg. Chem. 2005, 44, 9676-9683.
8. Lufaso, M. W.; Gemmill, W. R.; Mugavero, S. J.; Kim, S. J.; Lee, Y.; Vogt, T.;
zur Loye, H. C. J. Solid State Chem. 2008, 181 (3), 623-627.
9. Huq A.; Hodges J. P.; Gourdon, O.; Heroux, L. Zeitschrift für Kristallographie
Proceedings 2011, 1, 127-135.
12. Larson, A. C.; Von Dreele, R. B. Los Alamos National Laboratory Report LAUR
86-748 2000.
13. Toby, B. H. EXPGUI, a graphical user interface for GSAS. J. Appl. Cryst. 1991,
34, 210-213.
14. Kobayashi, K.; Nagao, T.; Ito, M. Acta Crys. A 1991, 67, 473-480.
20. Hobbs, D.; Kresse, G.; Hafner, J. Phys. Rev. B 1991, 62, 11556–11570.
21. Dudarev, S. L.; Botton, G. A.; Savrasov, S. Y.; Humphreys, C. J.; Sutton, A. J.
Phys. Rev. B 1991, 57, 1505–1509.
22. Anisimov, V. I.; Zaanen, J.; Andersen, O. K. Phys. Rev. B 1991, 44, 943-954.
23. Viola, M. C; Martinez-Lope, M. J.; Alonso, J. A.; De Paoli, J. M.; Pagola, S.;
Pedregosa, J. C.; Fernandez-Diaz, M. T.; Carbonio, R. E. Chem. Mater. 2003, 15
(8), 1655-1663.
24. Ivanov, S. A.; Eriksson, S. G.; Tellgren, R.; Rundlöf, H.; Tseggai, M. Mater. Res.
Bull. 2005, 40, 840-849.
150
25. Lopez, C. A.; Saleta, M. E.; Curiale, J.; Sanchez, R. D. Mater. Res. Bull. 2012,
47, 1158-1163.
27. Choy, J. H.; Kim, D. K.; Kim., J. Y. Solid State Ionics. 1998, 108, 159–163.
28. Chmaissem, O.; Kruk, R.; Dabrowski, B.; Brown, D. E.; Xiong, X.; Kolesnik, S.;
Jorgensen, J. D.; Kimball C. W. Phys. Rev. B 2000, 62, 14197-14206.
31. Currie, R. C.; Vente, J. F.; Frikkee, E.; IJdo, D. J. W. J. Solid State Chem. 1995,
116 (1), 199-204.
32. Tian,C.; Wibowo, A. C.; Zur Loye, H.-C.; Whangbo, M-H. Inorg. Chem. 2011,
50, 4142-4148.
33. Wang, J.; Meng, J.; Wu, Z. Chem. Phys. Lett. 2011, 501, 324-329.
1. M. Seifert, V. Neu, and L. Schultz, Appl. Phys. Lett. 94, 022501 (2009).
(2012).
151
7. F. Y. Yang, C. L. Chien, E. F. Ferrari, X. W. Li, G. Xiao, and A. Gupta, Appl.
Phys. Lett. 77, 286 (2000).
19. X. Chen, D. Parker, K. P. Ong, M. H. Du, and D. J. Singh, Appl. Phys. Lett. 102,
102403 (2013).
152
21. C. H. Du, R. Adur, H. L. Wang, A. J. Hauser, F. Y. Yang, and P. C. Hammel,
Phys. Rev. Lett. 110, 147204 (2013).
22. H. L. Wang, C. H. Du, Y. Pu, R. Adur, P. C. Hammel, and F. Y. Yang, Phys. Rev.
B 88, 100406(R) (2013).
30. V. I. Anisimov, J. Zaanen, and O. K. Andersen, Phys Rev B 44, 943–954 (1991).
31. H.-T. Jeng and G. Y. Guo, Phys. Rev. B 67, 094438 (2003).
35. I. Yang, S. Y. Savrasov, and G. Kotliar, Phys. Rev. Lett. 87, 216405 (2001).
153
37. S. Gold, E. Goering, C. König, U. Rüdiger, G. Güntherodt, and G. Schütz, Phys.
Rev. B 71, 220404 (2005).
38. H.-T. Jeng and G. Y. Guo, Phys. Rev. B 65, 094429 (2002).
1. D. Serrate, J.M. De Teresa, and M.R. Ibarra, J. Phys. Condens. Matter 19, 023201
(2007).
3. A.J. Hauser, J.R. Soliz, M. Dixit, R.E.A. Williams, M.A. Susner, B. Peters, L.M.
Mier, T.L. Gustafson, M.D. Sumption, H.L. Fraser, P.M. Woodward, and F.Y.
Yang, Phys. Rev. B 85, 161201 (2012).
4. H. Asano, N. Kozuka, A. Tsuzuki, and M. Matsui, Appl. Phys. Lett. 85, 263
(2004).
9. H. Das, P. Sanyal, T. Saha-Dasgupta, and D.D. Sarma, Phys. Rev. B 83, 104418
(2011).
13. A.J. Hauser, J.M. Lucy, H.L. Wang, J.R. Soliz, A. Holcomb, P. Morris, P.M.
Woodward, and F.Y. Yang, Appl. Phys. Lett. 102, 032403 (2013).
14. B.T. Thole, P. Carra, F. Sette, and G. van der Laan, Phys. Rev. Lett. 68, 1943
(1992).
15. C. Chen, Y. Idzerda, H. Lin, and N. Smith, Phys. Rev. Lett. 75, 152 (1995).
20. V.I. Anisimov, J. Zaanen, and O.K. Andersen, Phys. Rev. B 44, (1991).
21. H.-T. Jeng and G. Guo, Phys. Rev. B 67, 094438 (2003).
25. P.M. Woodward, Acta Crystallogr. Sect. B Struct. Sci. 53, 32 (1997).
26. P.M. Woodward, Acta Crystallogr. Sect. B Struct. Sci. 53, 44 (1997).
27. D. Huang, H.-T. Jeng, C. Chang, G. Guo, J. Chen, W. Wu, S. Chung, S. Shyu, C.
Wu, H.-J. Lin, and C. Chen, Phys. Rev. B 66, 174440 (2002).
29. R. Mishra, O.D. Restrepo, P.M. Woodward, and W. Windl, Chem. Mater. 22,
6092 (2010).
155
30. R. Leapman, L. Grunes, and P. Fejes, Phys. Rev. B 26, 614 (1982).
32. T.L. Daulton and B.J. Little, Ultramicroscopy 106, 561 (2006).
34. J.P. Clancy, N. Chen, C.Y. Kim, W.F. Chen, K.W. Plumb, B.C. Jeon, T.W. Noh,
and Y.-J. Kim, Phys. Rev. B 86, 195131 (2012).
35. J. van Bokhoven A., H. Sambe, D. Koningsberger C., and D. Ramaker E., J. Phys.
IV Fr. 7, 835 (1997).
156
7. Fthenakis, V. M.; Kim, H. C.; Alsema, E. Emissions from photovoltaic life
cycles. Environ. Sci. Technol. 2008, 42, 2168–2174.
8. Held, M.; Ilg, R. Update of environmental idicators and energy payback time of
CdTe PV systems in Europe. Prog. Photovoltaics Res. Appl. 2011, 19, 614–626.
9. Stoumpos, C. C.; Malliakas, C. D.; Kanatzidis, M. G. Semiconducting tin and
lead perovskites with organic cations: Phase transitions, high mobilities, and near-
infrared photoluminescent properties. Inorg. Chem. 2013, 52, 9019–9038.
10. Noel, N. K.; Stranks, S. D.; Abate, A.; Wehrenfennig, C.; Guarnera, S.;
Haghighirad, A.-A.; Sadhanala, A.; Eperon, G. E.; Pathak, S. K.; Johnston, M. B.;
Petrozza, A.; Herz, L. M.; Snaith, H. J. Lead-free organic–inorganic tin halide
perovskites for photovoltaic applications. Energy Environ. Sci. 2014, 7, 3061.
11. Hao, F.; Stoumpos, C. C.; Cao, D. H.; Chang, R. P. H.; Kanatzidis, M. G. Lead-
free solid-state organic–inorganic halide perovskite solar cells. Nat. Photonics
2014, 8, 489–494.
12. Stoumpos, C. C.; Malliakas, C. D.; Peters, J. A.; Liu, Z.; Sebastian, M.; Im, J.;
Chasapis, T. C.; Wibowo, A. C.; Chung, D. Y.; Freeman, A. J.; Wessels, B. W.;
Kanatzidis, M. G. Crystal growth of the perovskite semiconductor CsPbBr3: A
new material for high-energy radiation detection. Cryst. Growth Des. 2013, 13,
2722–2727.
13. Myagkota, S.; Voloshinovskii, A.; Stefanskii, I.; Mikhailik, M.; Pashuk, I.
Reflection and emission properties of lead-based perovskite-like crystals. Radiat.
Meas. 1998, 29, 273–277.
14. Kulbak, M.; Cahen, D.; Hodes, G. Planar How important is the organic part of
lead halide perovskite solar cells? Efficient CsPbBr3 solar cells. J. Phys. Chem.
Lett. 2015, 6, 2452–2456.
15. Kulbak, M.; Gupta, S.; Kedem, N.; Levine, I.; Bendikov, T.; Hodes, G.; Cahen,
D. Cesium enhances long-term stability of lead bromide perovskite-based solar
cells J. Phys. Chem. Lett. 2016, 7, 167−172.
16. Topas Academic, General Profile and Structural Analysis for Powder Diffraction
Data; Bruker AXS: Karlsruhe, Germany 2004.
17. Kresse, G.; Hafner, J. Ab initio Molecular-Dynamics for Liquid- Metals. Phys.
Rev. B 1993, 47, 558–561.
18. Kresse, G.; Hafner, J. Ab Initio Molecular-Dynamics Simulation of the Liquid-
Metal Amorphous-Semiconductor Transition in Germanium. Phys. Rev. B 1994,
49, 14251–14269.
157
19. Blöchl, P.E. Projector augmented-wave method. Phys. Rev. B. 1994. 50, 17953-
17979.
20. Perdew, J. P.; Burke, K.; Ernzerhof, M. Generalized gradient approximation made
simple. Phys. Rev. Lett. 1996, 77, 3865–3868.
21. Heyd, J.; Scuseria, G. E.; Ernzerhof, M. Hybrid Functionals Based on a Screened
Coulomb Potential. J. Chem. Phys. 2003, 118, 8207–8215.
22. Heyd, J.; Scuseria, G. E.; Ernzerhof, M. Hybrid Functionals Based on a Screened
Coulomb Potential. J. Chem. Phys. 2006, 124, 219906.
23. Paier, J.; Marsman, M.; Hummer, K.; Kresse, G.; Gerber, I. C.; Angyan, J. G.
Screened Hybrid Density Functionals Applied to Solids. J. Chem. Phys. 2006,
125, 249901.
24. Giorgi, G.; Fujisawa, J. I.; Segawa, H.; Yamashita, K. Small photocarrier
effective masses featuring ambipolar transport in methylammonium lead iodide
perovskite: A density functional analysis. J. Phys. Chem. Lett. 2013, 4, 4213–
4216.
25. Melissen, S. T. A. G.; Labat, F.; Sautet, P.; Le Bahers, T. Electronic properties of
PbX3CH3NH3 (X = Cl, Br, I) compounds for photovoltaic and photocatalytic
applications. Phys. Chem. Chem. Phys. 2015, 17, 2199-2209.
26. Yin, W.-J.; Yang, J.-H.; Kang, J.; Yan, Y.; Wei, S.-H. Halide perovskite materials
for solar cells: a theoretical review J. Mater. Chem. A 2015, 3, 8926-8942.
27. Park, B. W.; Philippe, B.; Zhang, X.; Rensmo, H.; Boshchloo, G.; Johansson, E.
M. J. Bismuth Based Hybrid Perovskites A3Bi2I9 (A: Methylammonium or
Cesium) for Solar Cell Application, Adv. Mater. 2015, 27, 6806–6813.
28. Lehner, A. J.; Fabini, D. H.; Evans, H. A.; Hebert, C. A.; Smock, S. R.; Hu, J.;
Wang, H.; Zwanziger, J. W.; Chabinyc, M. L.; Seshadri, R. Crystal and Electronic
Structures of Complex Bismuth Iodides A3Bi2I9 (A = K, Rb, Cs) Related to
Perovskite: Aiding the Rational Design of Photovoltaics, Chem. Mater. 2015, 27,
7137−7148.
29. Saparov, B.; Hong, F.; Sun, J.-P.; Duan, H.-S.; Meng, W.; Cameron, S.; Hill, I.
G.; Yan, Y., Mitzi, D. B. Thin-Film Preparation and Characterization of Cs3Sb2I9:
A Lead-Free Layered Perovskite Semiconductor, Chem. Mater. 2015, 27,
5622−5632.
30. Morss, L. R.; Siegall, M.; Stenger, L.; Edelstein, N. Preparation of cubic chloro
complex compounds of trivalent metals: Cs2NaMCl6 Inorg. Chem.,1970, 9, 1771-
1775.
158
References from Appendix A
3. Tian,C.; Wibowo, A. C.; Zur Loye, H.-C.; Whangbo, M-H. On the Magnetic
Insulating States, Spin Frustration, and Dominant Spin Exchange of the Ordered
Double-Perovskites Sr2CuOsO6 and Sr2NiOsO6: Density Functional Analysis
Inorg. Chem. 2011, 50, 4142-4148
159