Half Metals and Their Properties

LLNL-BOOK-424888
Half metallic materials and their

properties
C.Y. Fong, J.E. Pask, L.H. Yang
March 2, 2010
Half metallic materials and their properties

Disclaimer
This document was prepared as an account of work sponsored by an agency of the United States
government. Neither the United States government nor Lawrence Livermore National Security, LLC,
nor any of their employees makes any warranty, expressed or implied, or assumes any legal liability or
responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or
process disclosed, or represents that its use would not infringe privately owned rights. Reference herein
to any specific commercial product, process, or service by trade name, trademark, manufacturer, or
otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the
United States government or Lawrence Livermore National Security, LLC. The views and opinions of
authors expressed herein do not necessarily state or reflect those of the United States government or
Lawrence Livermore National Security, LLC, and shall not be used for advertising or product
endorsement purposes.
April 2, 2012 4:12 World Scientific Book - 9in x 6in HalfMetals
Publishers’ page
Publishers’ page
Publishers’ page
Publishers’ page
To our families
vi Half metallic materials and their properties

Preface
Since the first theoretical prediction of a half metallic material—one in

which the electrons of one spin participate in conduction while those of the
other do not—was published in 1983, much effort has been exerted to under-
stand, predict, and grow new half metallic materials. A main impetus for
this effort has been the great potential for half metals in spintronics—a new
generation of electronics in which the electron spin, as well as its charge, is
exploited to achieve substantial reductions in size and/or improvements in
performance. In a half metal, the electrons of one spin channel are metal-
lic while those of the other are insulating, the spin polarization at Fermi
level is complete (100%), and the magnetic moment is an integer. Partially
spin-polarized spintronic materials such as Co/Fe layered structures, have
already led to breakthroughs in information storage and processing tech-
nologies, such as magnetoresistance based read heads, magnetic random
access memory (MRAM), and spin-current switches. Because half metals
exhibit very large (ideally, infinite) magnetoresistances, it is not surprising
that the design and realization of half metal devices have been among the
hottest topics in condensed matter physics and materials science in recent
years.
Tremendous theoretical, experimental, and technological progress in the
understanding of half metallic materials has been achieved in the past few
years, and the scientific literature has grown correspondingly. It is oppor-
tune, therefore, to summarize the main concepts, results, and advances in
order to spark and facilitate new research and provide a solid and coherent
foundation for new researchers in this exciting field. These are the goals
we set out to achieve in the present monograph, at a level appropriate for
advanced undergraduate or graduate students in physics, chemistry, and
materials science.
vii
viii Half metallic materials and their properties
In the introductory chapter, we give an overview of the main features,

similarities, and differences among the three main classes of half metals
which have emerged to date. We discuss key features from both a the-
oretical and technological point of view. We emphasize the three basic
interactions underlying half metallic properties in all such materials in the
hope that new half metallic materials may be realized by tuning each of
the interactions independently. For device applications, it is necessary to
understand both electronic and magnetic properties. Substantial progress
has been achieved both experimentally and theoretically in the endeavor to
understand and predict half metallic materials properties and we provide in
Chapter 2 a discussion of the key experimental and theoretical techniques,
as appropriate for each class of material, which have been employed to date
in order to achieve that understanding and predictive capability.
The three main classes of half metallic materials which have emerged to
date are the Heusler alloys, transition metal oxides, and pnictides, chalco-
genides, and carbides with the zincblende structure; each with its own
rapidly evolving literature. We discuss each class in detail in Chapters 3,
4, and 5, respectively. In each chapter, we discuss the crystal structure,
experimental characterization, roles of the basic quantum mechanical in-
teractions, and resulting electronic, magnetic, and transport properties.
Quantum structures with tunable properties can be constructed from
half metals in various combinations with normal metals, semiconductors,
and insulators. In Chapter 5, we discuss some of the more common con-
figurations such as superlattices, quantum dots, digital ferromagnetic het-
erostructures, and quantum wires. For superlattices, we include a discus-
sion of active research areas such as transport properties. There is as yet
relatively little known about quantum dots in the context of half met-
als. However, capping and competition of ferromagnetic and antiferromag-
netic phases have emerged as key issues. Digital ferromagnetic heterostruc-
tures constitute an ideal configuration to investigate magnetic coupling,
and mechanisms to enhance coupling between transition metal elements, in
particular. We include a discussion of theoretical and experimental work
in this rapidly evolving area. Quantum wires have emerged as another
promising configuration for device applications. In these, the half metal-
lic properties have been shown to be robust with respect to the spin-orbit
interaction and lattice vibrations. We include also a discussion of the sim-
ilarities and differences of orbital hybridizations in zincblende structures
and quantum wires.
Acknowledgments
We are grateful for the support of the National Science Foundation with
Grants ESC-0225007 and ECS-0725902. Work at Lawrence Livermore Na-
tional Laboratory was performed under the auspices of the U.S. Department
of Energy under Contract DE-AC52-07NA27344. We thank Professor Kai
Liu for a critical reading of portions relating to experiments, Dr. R. Du-
mas for providing a diagram of the setup of the radio frequency sputtering
scheme, Dr. R. Rudd for a number of helpful comments throughout the
writing, and Dr. Michael Shaughnessy for a critical reading of Chapter 5.
We also thank the many research groups who have graciously allowed us to
include their results.
ix
x Half metallic materials and their properties

Contents
Preface vii
Acknowledgments ix
1. Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Classes of half metals . . . . . . . . . . . . . . . . . . . . 4
1.3 Half metallic devices . . . . . . . . . . . . . . . . . . . . . 10
2. Methods of studying half metals 13

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 Molecular beam epitaxy (MBE) . . . . . . . . . . . . . . . 14
2.2.1 Schematic setup . . . . . . . . . . . . . . . . . . . 14
2.2.2 Issues concerning growth . . . . . . . . . . . . . . 15
2.3 Characterization of samples . . . . . . . . . . . . . . . . . 17
2.3.1 Reflection high energy electron diffraction (RHEED) 17
2.3.2 X-ray reflectometry (XRR) . . . . . . . . . . . . . 19
2.3.3 Scanning tunneling microscopy (STM) . . . . . . 21
2.3.4 Auger electron spectroscopy (AES) . . . . . . . . 23
2.4 Methods of determining physical properties . . . . . . . . 24
2.4.1 Magnetic properties . . . . . . . . . . . . . . . . . 24
2.4.2 Transport properties . . . . . . . . . . . . . . . . 31
2.4.3 Half metallic properties . . . . . . . . . . . . . . . 37
2.5 Theoretical methods . . . . . . . . . . . . . . . . . . . . . 52
2.5.1 Density functional theory (DFT) . . . . . . . . . . 53
2.5.2 Kohn-Sham equations . . . . . . . . . . . . . . . . 54
2.5.3 Methods of calculating electronic properties . . . 58
xi
xii Half metallic materials and their properties
2.5.4 Methods of calculating Curie temperature, TC . . 65
3. Huesler alloys 71
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.2 Half-Heusler and full-Heusler alloys . . . . . . . . . . . . . 72
3.3 Methods of growing Heusler alloys . . . . . . . . . . . . . 75
3.3.1 Bulk Heusler alloys . . . . . . . . . . . . . . . . . 75
3.3.2 Thin films . . . . . . . . . . . . . . . . . . . . . . 76
3.4 Characterization of Heusler alloys . . . . . . . . . . . . . . 81
3.4.1 Bulk Heusler alloys . . . . . . . . . . . . . . . . . 81
3.4.2 Thin films . . . . . . . . . . . . . . . . . . . . . . 82
3.5 Physical properties of bulk Heusler alloys . . . . . . . . . 83
3.5.1 Magnetic moments and the Slater-Pauling rule . . 83
3.5.2 Insulating gap in half metallic Heusler alloys . . . 85
3.5.3 Polarization at EF . . . . . . . . . . . . . . . . . . 89
3.5.4 Magnetic moments . . . . . . . . . . . . . . . . . 90
3.5.5 Curie temperature . . . . . . . . . . . . . . . . . . 92
3.5.6 Other magnetic properties . . . . . . . . . . . . . 92
3.5.7 Disorder in Heusler alloys . . . . . . . . . . . . . . 95
3.6 Physical properties of Heusler alloys in thin-film form . . 100
3.6.1 NiMnSb . . . . . . . . . . . . . . . . . . . . . . . 100
3.6.2 Co2 MnSi . . . . . . . . . . . . . . . . . . . . . . . 104
3.6.3 Co2 FeSi . . . . . . . . . . . . . . . . . . . . . . . . 117
4. Half metallic oxides 131

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 131
4.2 CrO2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
4.2.1 Structure . . . . . . . . . . . . . . . . . . . . . . . 132
4.2.2 Growth . . . . . . . . . . . . . . . . . . . . . . . . 132
4.2.3 Characterization . . . . . . . . . . . . . . . . . . . 135
4.2.4 Transport properties . . . . . . . . . . . . . . . . 136
4.2.5 Half metallic properties . . . . . . . . . . . . . . . 140
4.2.6 Electronic properties . . . . . . . . . . . . . . . . 141
4.2.7 Magnetic properties . . . . . . . . . . . . . . . . . 143
4.2.8 Theoretical studies of electronic and magnetic
properties . . . . . . . . . . . . . . . . . . . . . . 145
4.3 Fe3 O4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
4.3.1 Structure . . . . . . . . . . . . . . . . . . . . . . . 148
Contents xiii
4.3.2 Growth . . . . . . . . . . . . . . . . . . . . . . . . 150

4.3.3 Characterization . . . . . . . . . . . . . . . . . . . 153
4.3.4 Physical properties . . . . . . . . . . . . . . . . . 155
4.3.5 One-electron theory . . . . . . . . . . . . . . . . . 164
4.4 La1−x (Sr, Ca, Ba)x MnO3 . . . . . . . . . . . . . . . . . . 172
4.4.1 Structure . . . . . . . . . . . . . . . . . . . . . . . 173
4.4.2 Growth and characterization . . . . . . . . . . . . 174
4.4.3 Physical properties . . . . . . . . . . . . . . . . . 175
4.5 Magnetic interactions in the oxides . . . . . . . . . . . . . 181
4.5.1 Superexchange . . . . . . . . . . . . . . . . . . . . 182
4.5.2 Double exchange . . . . . . . . . . . . . . . . . . . 183
4.5.3 Magnetism in CrO2 and Fe3 O4 . . . . . . . . . . . 185
5. Half metals with simple structures 191

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 191
5.2 Half metals with zincblende structure . . . . . . . . . . . 192
5.2.1 Experiment . . . . . . . . . . . . . . . . . . . . . . 192
5.2.2 Theory . . . . . . . . . . . . . . . . . . . . . . . . 202
5.3 Half metallic superlattices . . . . . . . . . . . . . . . . . . 212
5.4 Quantum dots . . . . . . . . . . . . . . . . . . . . . . . . 218
5.4.1 Experiment . . . . . . . . . . . . . . . . . . . . . . 218
5.4.2 Theory . . . . . . . . . . . . . . . . . . . . . . . . 222
5.5 Digital ferromagnetic heterostructures . . . . . . . . . . . 224
5.5.1 Experiment . . . . . . . . . . . . . . . . . . . . . . 224
5.5.2 Theory . . . . . . . . . . . . . . . . . . . . . . . . 232
5.6 One dimensional half metals . . . . . . . . . . . . . . . . . 238
Appendix A Appendix 243

A.1 Anisotropic magnetoresistance . . . . . . . . . . . . . . . 243
Bibliography 245
Index 269
Chapter 1
Introduction
1.1 Background
The primary ingredient of computing hardware is the integrated circuit

(IC). The number of transistors that can be placed on an IC is character-
ized by Moore’s law (Moore, 1965) which describes a long-term trend in
the history of computing hardware. According to Moore’s law, the packing
density of transistors on a chip doubles approximately every two years. In
the last decade, advances in magnetic hard drive technology have increased
the capacity of a typical hard drive from tens of gigabytes (GB) to terabytes
(TB); while the bit density of magnetic heads has increased 60–100% annu-
ally to 10.7 gigabits/in, which suggests the law still holds. However, there
is a limit to such packing due to the finite interatomic distance of a few Å in
typical materials,1 suggesting that Moore’s law must eventually be violated.
Furthermore, another important issue is the volatility of the information
stored. For example, in dynamic random access memory (DRAM), data is
stored in the form of charge on capacitors. Therefore, whenever the power
is switched off the information is lost. The approaching limit in packing
density and volatility of information stored must be overcome to produce
smaller, faster devices.
Innovative technologies have been developed after the discovery of giant
magnetoresistance (GMR) (Baibich et al., 1988; Wolf et al., 2001) to utilize
the electron spin degree of freedom for information manipulation, storage,
and transmission. Metal-based devices in the form of heterostructures have
already found applications in ultra-high density magnetic recording (Ross,
2001), as GMR sensors (Baibich et al., 1988; Parkin et al., 1991; Dieny et al.,
1992), in prototype magnetic random access memory (MRAM) (Daughton,
1 http://www.almaden.ibm.com/st/
1
2 Half metallic materials and their properties
1992; Savtchenko et al., 2003; Bussmann et al., 1999), and in spin-current

switches (Albert et al., 2000; Grollier et al., 2001). These devices utilize
ordinary ferromagnetic materials, such as Fe, Co, and Ni, that have a spin
imbalance, or finite electronic spin polarization P at Fermi energy EF .
Information can be encoded into the finite resistance change resulting from
spin-dependent electric transport through these magnetic heterostructures.
The advantages of these devices over conventional non-magnetic devices
are their increased integration densities, nonvolatility, faster speed in data
processing, and lower power consumption.
To develop further these new technologies exploiting electron spin, it is
advantageous to leverage established technologies. One emerging field com-
bines semiconductors and magnetism. The term “spintronics” was coined
by S. Wolf in 1996 (Wolf et al., 2001). It is characterized by the transport
of spin, or charge and spin, in semiconductor based materials. Therefore,
a crucial initial step in developing next-generation spin-based technolo-
gies is the development of spintronic materials. These materials can be
composed of ferromagnetic metals, magnetic semiconductors with or with-
out doping, and/or so called “half metals” (HM). Such materials stand
to enable new technologies with dramatically increased speeds, decreased
sizes, and nonvolatile storage. Figure 1.1 illustrates the development of
these next-generation technologies. Materials issues are crucial concern-
ing design, growth, and contact to electrodes to efficiently transfer spin.
Associated devices may be expected to include spin field-effect-transistors,
memory devices, and components for quantum computing.
Fig. 1.1 Developing next-generation spin-based technologies and applications
Normally, a nonferromagnetic material, such as silicon (Si) for example,

has spin degeneracy in its energy states—i.e., energies are independent of
spin and occupation of spin-up (↑) and spin-down (↓) states is equal, lead-
Introduction 3
ing to zero net spin polarization. In a ferromagnetic metal, such as iron

(Fe), on the other hand, this degeneracy is broken and more states of one
spin channel (the “majority” channel) are occupied than the other (“minor-
ity” channel), leading to nonzero net spin polarization and ferromagnetic
properties. Since coordinate axes can be chosen as desired, it is typical to
denote, without loss of generality, the majority-spin channel as ↑ and the
minority-spin channel as ↓ and we shall adopt this convention here. The
spin polarization PN is defined by
N↑ − N↓
PN = , (1.1)
N↑ + N↓
where N↑ and N↓ are the number of ↑ spin and ↓ spin states at EF , respec-
tively. Under this definition, PN measures the spin imbalance of mobile
electrons. A typical value for PN in ferromagnetic metals is 40 ∼ 50% at
room temperature (RT). It has also been defined, alternatively, as the net
fractional spin polarization near EF . In this case, we denote it by P to
distinguish from PN , with
d↑ − d↓
P = , (1.2)
d↑ + d↓
where d↑ and d↓ are the density of states (DOS) of ↑ spin and ↓ spin channels
at EF , respectively. P can therefore be directly determined from the DOS.
de Groot et al. (1983) predicted a remarkable property in the half
Heusler alloy, NiMnSb. In this compound, the ↑ spin states exhibit metal-
lic properties while the ↓ spin states are insulating. At EF , N↓ and d↓
vanish identically. Consequently, all states at EF are ↑ spin and NiMnSb
has P = 100%. These properties define half metallicity and materials pos-
sessing these properties are called half metals. The earliest Heusler alloys
were grown by Heusler (1903). Materials such as NiMnSb are called half-
Heusler alloys. Many of the so called full-Heusler alloys, e.g., Co2 MnSi,
are also predicted to be HM theoretically and some have been confirmed
experimentally. The difference between half- and full-Heusler alloys lies in
the occupation of sites within the unit cell, as we discuss further below.
There are also predictions of HMs in other structures, such as oxides
(Schwarz, 1986) in the rutile structure. Among them, only CrO2 has been
verified experimentally as a HM at low temperature. Other oxides as well as
chromium arsenide (CrAs) (Akinaga et al., 2000a) and manganese carbide
(MnC) (Pask et al., 2003) in the zincblende (ZB) structure were predicted
theoretically to be HMs. In the latter compound, the usual roles played by
the majority- and minority-spin channels are reversed.
In all of the above cases, the transport properties are determined solely
by electrons in states in the vicinity of EF with a single spin polarization.
In some applications, both charge and spin transport can be envisioned.
Transport utilizing the spin degree of freedom provides completely new
prospects for information storage and transmission. It is, all but certainly,
only a matter of time to incorporate this type of transport, with all its
possibilities, into semiconductor technologies. However, there are concerns
about the disappearance of half metallicity at room temperature (RT) in
these materials. Those materials experimentally demonstrated to reach
over 95% spin polarization, such as CrO2 , cannot sustain their spin po-
larization above RT (Dowben and Skomski, 2004). Structural transitions,
collective excitations, e.g., spin waves and phonons, correlations associated
with onsite Coulomb interactions, and spin polarons (Katsnelson et al.,
2008) which show non-Fermi liquid behavior and are formed in the gap
near EF , can cause loss of half metallicity. In this monograph, we focus
manily on the design, growth, and basic understanding of the electronic and
magnetic properties of half metallic materials determined experimentally at
low temperature and predicted theoretically at or near T = 0 K. We will,
however, briefly comment on issues at higher temperature as appropriate.
1.2 Classes of half metals
Up to now, three main classes of ferromagnetic HMs have been found

with distinct crystal structures. They are the Heusler alloys (de Groot
et al., 1983; Galanakis, 2002b), such as Co2 CrAl; the oxides, such as CrO2
(Schwarz, 1986); and those having the ZB structure (e.g.,CrAs) (Akinaga
et al., 2000a; Pask et al., 2003) or diamond structure (Qian et al., 2006a).
The unit cell of a typical full-Heusler alloy is shown in Fig. 1.2 and that of
CrO2 is shown in Fig. 1.3. In Fig. 1.2, the Co atoms are shown as filled and
open circles, the Cr atoms are denoted by open triangles, and the Al atoms,
by open squares. A common feature of all these HMs is that they contain
at least one TM atom, such as Co, Mn, or Fe, in the unit cell. Their d-
states play a key role in the half metallicity. The interaction of the d-states
with the states of other atoms—e.g., the d-states of other TM atoms, the
p-states of oxygen atoms, pnictides, or Group IV elements—gives rise to a
number of distinguishing properties.
To get a basic understanding of the half metallic properties and differ-
ences among the three classes, we give an overview in terms of the density
Introduction 5
Fig. 1.2 L21 crystal structure of full-Heusler alloy (X2 YZ), such as Co2 CrAl. The
X=Co atoms are denoted by filled and open circles, the Y=Cr atoms by open triangles,
and the Z=Al atoms by open squares. The outermost cube edge has length a. If the
sites with open circles (X(2)) are unoccupied, the structure corresponds to a half-Heusler
alloy (XYZ) and is denoted by C1b (Galanakis et al., 2002a).
Fig. 1.3 Crystal structure of CrO2
of states (DOS) based on one-electron properties. The DOS for each class
shows distinct features as a consequence of the atoms involved and distinct
crystal structure. A typical DOS exhibits s-like states from the non-TM
atoms, such as oxygens, pnictides, and Group IV elements, in the low en-
ergy region of the valence manifold. The d-states of the TM are split into
triply and doubly degenerate multiplets in a cubic or tetrahedral environ-
ment. In the cubic case, the triply degenerate states are labeled t2g and
have lower energy than the doubly degenerate states labeled eg because the
lobes of the dxy , dyz , and dzx states point toward neighboring atoms while
the lobes of the eg states point toward second-nearest neighbors.2 In the
2 The t2g and eg labels are not the only ones used to denote triply and doubly degenerate
states, especially in condensed matter physics. For example, the labels Γ25′ and Γ12′ are
also used in cubic cases (having inversion symmetry). In the ZB structure, Γ15 is used
to label the three-fold degenerate states. In recent years, however, the t2g /eg notation
tetrahedral environment, the order of the two sets is reversed. Depending

on the strength of the d-d interaction between neighboring TM elements
relative to the d-p interaction between TM and neighboring non-TM ele-
ments, the highest occupied states can be either d or d-p hybrid in nature.
The ferromagnetic exchange interaction determines the relative energies of
the majority and minority states and the value of EF . Considering the d-d
and d-p interactions, the half metallicity in all three classes of compounds
can be understood in terms of the crystal field, hybridization, and exchange
interaction.
In Fig. 1.4, we show the DOS near EF of a half-Heusler alloy. Typically,
a Heusler alloy is metallic in the majority-spin channel. In this case, the d-d
interaction between the X and Y atoms (Galanakis et al., 2002a) determines
the states near EF . The Fermi level intersects the eg states in the majority-
spin channel. There is an overlap between the eg and t2g states in this
channel due to the d-d interaction. In the minority-spin channel, the gap
is formed between the hybridizing d (X-atom) and p (Z atom) states (p-t2g
hybrid) and antibonding d states (e∗g ) from the TM (Y-atom). A common
feature of the Heusler alloys is that EF is located just below the unoccupied
↓ spin states—i.e., bottom of the conduction band in the minority-spin
channel. This feature has important consequences, as we describe later.
Fig. 1.4 Schematic DOS of a half-Heusler alloy. ∆↓ is the energy gap for the ↓ spin
states.
Coey and Venkatesan (2002) classified the DOS of half metallic oxides
into three types. They are IA , IB , and IIB and are shown in Figs. 1.5, 1.6,
and 1.7. The label A indicates that the ↑ spin channel is conducting. The
label B indicates that the ↓ spin channel is conducting. In all three types,
the lowest energy s-states are below the energy range shown. The next
has become more prevalent (even when not strictly applicable to the symmetry at hand)
and we shall use this convention here.
Introduction 7
group of states are oxygen 2p in nature. They are bonding states. Both
↑ spin and ↓ spin channels are equally occupied. They experience a small
effect from ferromagnetic exchange.
Fig. 1.5 Type IA DOS, ∆↓ is the insu- Fig. 1.6 Type IB DOS.
lating gap for the ↓ spin channel.
For the type IA HMs (Fig. 1.5), the next structure in the DOS is de-
rived from the d-states of the majority-spin channel. They are partially
occupied and the Fermi energy EF passes through the t2g states. The eg
states overlap with the t2g states and are located at higher energy. This
is a consequence of the octahedral crystal field. The next higher energy
structure is the unoccupied d-states of the minority-spin channel. Just as
for the occupied majority-spin channel, the eg states are at higher energy
and overlap with the t2g states. The insulating (semiconducting) gap (∆↓ )
is in the minority-spin channel between the t2g states and bonding oxygen
p-states. An example of this type of HM is CrO2 . As shown later, the
difference between this schematic DOS and the calculated DOS is that the
latter shows some d-p hybridization in the occupied states, with dominant
oxygen p.
Fig. 1.7 Type IIB DOS. Fig. 1.8 Type IIIA DOS.
For the type IB HMs (Fig. 1.6), the next structure above the bonding
oxygen p-states is also derived from the d-states of the majority-spin chan-
nel. In this case, however, they are fully occupied. The next higher energy
structure is the overlapping t2g and eg states of the minority-spin channel
with EF passing through t2g states. The insulating gap is formed between
fully occupied majority-spin d states and anti-bonding oxygen s states. An
example of this type of HM is Sr2 FeMnO6 .
For type IIB HMs (Fig. 1.7), the d manifolds of majority- and minority-
spin states do not overlap. The lower-energy majority-spin states are fully
occupied. The mechanism of conduction differs from that of type IA and
IB HMs. The minority t2g electrons form polarons. Fe3 O4 is an example
of this type of compound. The half metallic properties of this type of HM,
like others, can be affected by impurities. For example, if the sublattice
occupied by TM elements is substituted by other elements, the sample can
become a Mott insulator.
Based on the mobility µ and the effective mass m∗ of the mobile carri-
ers, rather than integer magnetic moment/unit-cell, Coey and Venkatesan
(2002) considered two additional types of HM (types IIIA and IVA ). As
shown in the DOS (Fig. 1.8) for a type IIIA HM, both spin channels in-
tersect EF . However, the majority-spin electrons are localized while the
minority-spin electrons are delocalized. There is thus a large difference of
µ and m∗ for carriers of different spin. The electrons in the majority-spin
channel are essentially immobile; and so conduction is confined mainly to
the minority-spin channel.
The main features of types I and II HMs can be qualitatively understood
in terms of two key facts: (i) oxygen atoms have large electronegativity and
(ii) TM elements have tightly bound d-states. Consequently, the oxygen
atom essentially ionizes the electrons of neighboring cations to fill its 2p
states. The d states of the cation split into three-fold t2g and two-fold
eg states in the octahedral field of the surrounding oxygen atoms. The
exchange interaction shifts the energies of the minority-spin states up rel-
ative to the majority-spin states. The occupied d states remain localized
at the cation site with more or less atomic-like features. The p states of
the oxygen atoms show less hybridization with the d states than do the p
states of the pnictide or Group IV elements in the Heusler alloys due to the
large electronegativity of the oxygen atoms. There is no d-d hybridization
because the TM elements are surrounded by the oxygen atoms.
For a HM with the ZB structure, we first give a qualitative discussion of
the bonding and then comment on key features of the DOS. The anion can
Introduction 9
Fig. 1.9 MnAs: As atoms (filled circles) are at the corners and face centers. A Mn
atom (open circle) is surrounded by four As atoms and is located at (1/4,1/4,1/4) along
the body diagonal. A schematic diagram of the d-p hybridization is shown at right.
be a valence IV, V, or VI element. Its electronegativity is in general weaker

than that of an O atom. Due to the tetrahedral environment, the anion
s and p states form sp3 type orbitals, which point toward the neighboring
cations. The cation, a TM element, has its five-fold degenerate d-orbitals
split into t2g and eg type states in the tetrahedral environment. The t2g
states are higher in energy than the eg states and are comprised of dxy ,
dyz , and dzx states. Linear combinations of these form directional orbitals
pointing toward neighboring anions. These orbitals can interact with the
sp3 type orbitals of the anions to form bonding and antibonding states.
Figure 1.9 shows the structure and d-p hybridization in MnAs. At left, the
As atoms are indicated by filled circles and the Mn atoms by open circles.
One Mn atom is located at (1/4,1/4,1/4)a along the cubic body diagonal,
where a is the length of the cube edge. The primitive cell is defined by one
Mn atom and one As atom. At the right of Fig. 1.9, the overlap of a Mn
d-orbital and As sp3 orbital is depicted. This overlap gives rise to bonding
and antibonding states. The bonding states have covalent character, i.e.,
charge sharing with neighbors. These are the d-p hybrid states. The energy
of the bonding p-t2g hybrid states is lower than the eg states. The ordering
of energies is shown schematically in Fig. 1.10. At the left and right ends,
the energy levels of the d-states of a TM element and the s- and p-states of
a chalcogenide, pnictide, or carbide are shown. Moving toward the center,
the effects of the crystal field are illustrated. For the TM atom, the five-fold
degenerate d-states split into triply degenerate t2g and doubly degenerate
eg states. The non-TM element forms sp3 type orbitals. At the middle
of the figure are the bonding (p-t2g ) and antibonding (t∗2g ) states resulting
from the d-p hybridization.
Fig. 1.10 Schematic diagram of the Fig. 1.11 Schematic DOS of a HM

crystal-field splitting of 3d orbitals for with ZB structure.
one spin under cubic symmetry and d-p
hybridization in the ZB structure. Su-
perscripts indicate degeneracy.
The ordering of these states is shown in the schematic DOS of Fig. 1.11.
Only states in the vicinity of EF are shown. The eg states are comprised of
dz2 and dx2 −y2 states. They point toward to second neighbors rather than
nearest neighboring cations and form the non-bonding states or bands in
the crystal. They can overlap in energy with the d-p bonding states (p-t2g )
or can be separated from them to form a gap.
As shown in the ↓ spin channel, a gap exists between the bonding p-
t2g states and non-bonding eg states. EF passes through this gap. For
the majority-spin states, an overlap between the bonding p-t2g and non-
bonding eg states is typical, as shown in the figure. The bonding states
have significant anion-p character. To accommodate the total number of
valence electrons in the unit cell, the lowest-energy antibonding states in
the majority-spin channel are occupied due to the exchange splitting of
majority- and minority-spin states. These antibonding states are d-p hy-
brid states with predominantly transition-element d character. This partial
occupation in the majority-spin channel gives rise to the half metallicity.
Among the three classes of HMs, the d-states are dominant near EF in
the Heusler alloys and the oxides. Hybridization is the strongest in the ZB
HMs. The states at EF for these are d-p hybrid in character.
1.3 Half metallic devices
Since HMs offer the possibility of unprecedented magnetoresistances (infi-

nite in principle), there is huge potential for device applications, improving
Introduction 11
substantially essentially all GMR based devices in current use and offering
a host of new possibilities. Hence, this has been an active field of research.
There have been many theoretical proposals, based on extensive quantum
mechanical calculations, as we detail in subsequent chapters. Here, we note
some recent progress on practical device realization.
Recent efforts in spintronic device fabrication have focused mainly in
the areas of spin valves and spin transistors. The latter was proposed by
Datta and Das (1990) and is similar to standard FETs. However, the re-
quirements for effective FETs, such as large polarized current, capability
of amplification, and simple material structure, are extremely demanding,
and there has not yet been success fabricating practical devices in this
form. There has been more progress in fabricating HM based spin-valves.
In these devices, spin is injected from HMs to semiconductors. It has been
identified by Wang and Vaedeny (2009) that four conditions should be met
for an effective spintronic device. Among the four, the most important is
efficient spin injection from the metallic materials normally serving as elec-
trodes into semiconductors. To maintain the signal, electrons in the semi-
conductor should have long lifetimes for conduction and spin relaxation.
In crystalline semiconductors, the existence of the spin-orbit interaction
can change the spin moment of the carriers. Wang and Vaedeny (2009)
used organic semiconductors instead of conventional semiconductors. They
made a device with La2/3 Sr1/3 MnO3 (LSMO) and Co serving as electrodes.
The organic semiconductor CVB (4,4’–bis–(ethyl–3–carbazovinylene)–1,1’–
biphenyl) was used to reduce the effect of the spin-orbit interaction. The
measured spin polarized carrier density was 60–20% in the temperature
range of 60–120 K. Kodama et al. (2009) grew the layered structure
Cr/Ag/Cr/Co2MnSi/Cu/Co2 MnSi/Co0.75 Fe0.25 /Ir0.22 Mn0.78 /Ru on MgO
(001) substrate. Cu was included between Heusler layers to facilitate the
transport of spin polarized carriers from one Heusler alloy to the other,
which would be impeded by Cr. The Ag layer between the Cr layers served
to enhance the perpendicular current. The ratio of magnetoresistance was
found to be 8.6 % at room temperature and 30.7% at T = 6 K. Hence, more
work is needed to get magnetoresistance and robustness to desired levels,
and much work continues.

Chapter 2
Methods of studying half metals
2.1 Introduction
To realize spintronic devices based on half metals (HMs), the growth of

these materials must be addressed. Growth methods for the three classes
of HMs are not universal; some are specific to a particular class or classes.
In this chapter, we discuss a method which has been used to grow Heusler
alloys and HMs with the ZB structure. Having discussed growth, we next
consider methods used to characterize the quality of the samples grown—
a crucial step for all subsequent understanding and application. We then
consider techniques used to measure the properties of the resulting mate-
rials, including electronic properties, magnetic properties, half metallicity,
transport properties, and TC . Most of the characterization methods and
measurement techniques are commonly applied to all classes of HMs. Since
de Groot et al. (1983) first predicted half metallic properties in NiMnSb,
theoretical modeling has been shown to be extraordinarily effective in un-
derstanding and predicting new HMs. This requires methods with substan-
tial predictive power for atomic arrangements quite unlike any known or
considered heretofore. Therefore, first-principles methods based on density
functional theory (DFT) are among the most popular for this purpose and
we include a discussion of the basic concepts and practical implementations
of these.
In this chapter, we first discuss the key experimental method used for
growing the Heusler alloys and HMs with ZB structure in thin-film form.
We then discuss the methods of characterization and techniques for mea-
surement commonly employed. Finally, we conclude with a brief discussion
of theoretical methods most commonly employed, since many excellent re-
views are available (see, e.g., Martin (2004)).
13
2.2 Molecular beam epitaxy (MBE)
Because HM crystal structures are not always the stable bulk ground state
structures, and because desired magnetoresistance (MR) properties are of-
ten achieved by multilayer structures, the ability to grow HM structures
in thin-film form is of crucial importance. The molecular beam epitaxy
(MBE) method is an ideal way to grow quality thin films. It was originally
designed to grow layered semiconductors and semiconductor heterostruc-
tures (Chang and Ploog, 1985).
Its application to grow NiMnSb was carried out by Van Roy et al. (2000)
and Turban et al. (2002). Van Roy et al. (2000) used the Riber 32P chamber
while a commercial chamber was used by Turban et al. (2002). Ambrose
et al. (2000) also used a commercial chamber to grow Co2 MnGe. With the
MBE method, Akinaga et al. (2000a) were the first group to successfully
grow CrAs thin film with ZB structure and Zhao et al. (2001) grew ZB
CrSb with a single monolayer. Hereinafter, we will refer to these types of
samples as HMs with the ZB structure.
2.2.1 Schematic setup

A schematic diagram for the setup of the MBE method is shown in Fig. 2.1.
The growth chamber is shown as a circle. The residual pressure in the
chamber is commonly maintained at the level of 10−9 mbar.
Fig. 2.1 A schematic diagram of the setup for the MBE method. The substrate is just
above the holder. Sample is shown in black horizontal rectangle.
The beam sources are indicated by open rectangles connected to the

chamber. Depending on desired number of constituent atoms, more than
two beam sources are possible. The sample holder is shown in black. The
substrate is on top of the sample holder; and the sample, on top of the
Methods of studying half metals 15
substrate. In general, characterizations of the sample are made in situ,

measurements are carried out without moving the sample out the chamber.
In that case, the chamber will have to accommodate such equipment.
The source cells are, of course, specifically dependent on what is to
be grown. For the growth of NiMnSb, a nonvalved cracker cell for Sb
is sometimes used. The cracker zone was kept at 930 ◦ C. Ni and Mn
atoms with 1:1 ratio are furnished (Ambrose et al., 2000). Alternatively,
Turban et al. (2002) prepared 99.99% pure Ni and Mn and 99.999% pure
Sb Knudsen cells. They also used a quartz-crystal microbalance installed
at the sample position to monitor the flux of the atomic beams. To grow
Co2 MnGe, atomic fluxes are supplied by individual Knudsen cell sources
monitored by a quadrupole mass analyzer. For the growth of thin-film
forms of CrAs having ZB structure, GaAs (001) serves as the substrate
(Akinaga et al., 2000a). Below it, a GaAs buffer layer of the order of 20
nm thick was grown first at 580 ◦ C. This helps to obtain flat surface of the
substrate, to reduce the stress of the substrate. The substrate should be
annealed at 600 ◦ C for about 10 min in the chamber in order to remove
oxidation layers or other contaminations. During the growth, the substrate
is under rotation and is irradiated by Cr and As beams at once. The rate
of growth was 0.017 nm/s. After the growth, the sample was capped by
low temperature GaAs to prevent oxidation when the sample is exposed to
air. The thickness of the capping layer was 5 nm.
2.2.2 Issues concerning growth

MBE has been applied to the growth of Heusler alloys and HMs with ZB
structure. The quality of a thin-film sample depends on the selection of the
substrate and the control of the substrate temperature. These are general
issues for all thin-film growths but they are sample specific. The following
discussions on these two issues will focus on Heusler alloys and HMs with
ZB structure.
2.2.2.1 Substrate for Heusler alloys

For Heusler alloys, both of growths by Van Roy et al. (2000) and Ambrose
et al. (2000) used GaAs(001) as the substrate based on the compatibility of
the crystal structures between the growth sample and substrate and based
on the matching of lattice constants between them. On the other hand,
Turban et al. (2002) used single crystal MgO with (001) surface. MgO has
a relatively smaller lattice constant which is 4.21 Å. There should be a large
lattice constant mismatch. This kind of substrates, in general, can be used
to grow polycrystalline samples in film forms (Bauer, 2010). Therefore,
as described above the primary consideration for the choice of a substrate
is determined by matching lattice constants between the sample and the
substrate. Alternative choices to grow single crystal thin films can be made
based on experience by adding additional buffer layers.
2.2.2.2 Substrate for HMs with ZB structure

For growths of HMs with ZB structure, it is more critical to take the match-
ing of lattice constants between the thin film and substrate into considera-
tion because the ZB structure for these compounds is not the ground state
structure. If there is additional stress due to lattice constant mismatch,
the metastable ZB HM will cause devices made of these materials to have
limited life times. Another possibility is that the half metallicity can be lost
or obtained under stress. As we discussed earlier, Akinaga et al. (2000a)
used GaAs(001) as the substrate for CrAs growth, while Zhao et al. (2001)
grew CrSb on GaSb.
2.2.2.3 Temperature
The temperature of the substrate is an important factor to determine the
quality of the films. For growing NiMnSb on GaAs(001), the best films
are obtained when the temperature of the substrate Tsub is kept at 300 ◦ C
(Van Roy et al., 2000). Turban et al. (2002) demonstrate clearly that Tsub
at about 350 ◦ C is the best to grow NiMnSb by measuring the intensity
of the reflection high energy electron diffraction (RHEED) to be described
later. On the other hand, the substrate temperature is maintained at 175
◦
C for Co2 MnGe growths. To grow a better CrAs, the temperature of the
substrate cannot be maintained at 580 ◦ C required to have a flat surface
of the substrate. It is necessary to bring down the temperature to 200 ◦ C
and 300 ◦ C range under As pressure.
During or after a sample is grown, it is necessary to characterize its
quality and to measure its physical properties. In the following two sec-
tions, we will discuss methods of characterization, in particular samples in
thin-film forms and of determining various physical properties relevant to
spintronic applications. These methods are generally applied to all HMs.
The results, however, will be sample dependent and they will be discussed
specifically for each sample.
2.3 Characterization of samples
Two popular methods to characterize the quality of HMs grown in thin-film

forms are the reflection high energy electron diffraction (RHEED) (Harris
et al., 1981) and low energy electron diffraction (LEED) (Heinz, 1995).
The thickness of a film can be determined by either high-resolution X-ray
diffraction (HRXRD) (Mukhamedzhanov et al., 2000) or X-ray reflectom-
etry (XRR) (Toney and Brennan, 1989). To correlate whether one period
of oscillation corresponds to the growth of one atomic layer or of one unit
cell depends on the sample. Methods to distinguish such correlations are
XRR and scanning tunneling microscopy (STM) (Binnig et al., 1983).
Although RHEED is a powerful method, it is noted that the results of
the method cannot distinguish between chemically ordered and disordered
structures; in particular, the degree of intermixing between atoms of the
sample and substrate. In this respect, Auger electron spectroscopy (AES)
(Wang et al., 2005a) is an appropriate method and should be carried out.
Between RHEED and LEED methods, RHEED method provides more in-
formation. Since it is the more powerful method between the two, we shall
give a detailed discussion. This will be followed by methods of XRR and
STM. The former can serve two functions as described above and the latter
is now commonly used. Since intermixing near the layer boundary is one
of the important issues relevant to the quality of the film, we shall discuss
AES afterward.
2.3.1 Reflection high energy electron diffraction (RHEED)

2.3.1.1 Basic information
A beam of electrons is used as the probe in this method. The energy of
the electron is in the range of 15.0 keV. The beam is incident at a glancing
angle of about 2.0◦ on the sample. The reflected electron beam provides
the interesting information.
2.3.1.2 What is measured

There are two quantities to be measured: the interference pattern of the
reflected beam that is called the RHEED pattern and the intensity at the
center labeled as (00) of the pattern.
RHEED pattern The pattern is in striped form and provides the struc-
tural information of the sample. The primary stripes are the constructive
interference pattern relevant to the fundamental structure of the sample
surface. Therefore, one of the features of RHEED is the capability to char-
acterize a surface structure. Before the growth, it is important to check the
pattern of the substrate. When the primary stripes coincide with those of
the substrate, lattice-constant matched growth can be inferred. If there are
additional weak streaks, it indicates the surface is reconstructed. A typical
RHEED pattern for a surface structure is shown in Fig. 2.2. The pattern
Fig. 2.2 A typical RHEED pattern of a surface with streaks (Wang et al., 2005a). The
primary ones are shown in dark regions. The streaks are shown in light stripes.
shown in Fig. 2.2 is for a clean GaAs with electron beam focused along the
[11̄0] crystal axis. The stripes with elongated dark regions are the primary
lines. Spacing between these primary lines can fill with other weak streaks
which are related to the reciprocal lattice vector in the crystallographic
direction of the surface for the sample with a reconstruction. This is one
way to extract lattice-constant information of the sample. In this figure,
there is clearly a reconstruction with the periodicity of six times the lattice
constant—each primary line is accompanied by five weak streaks.
Oscillatory behavior of spot intensity at the center (00) This os-

cillatory behavior reveals the growth mode is in the layer-by-layer form or
the Frank-Van der Merwe growth mode (see Bauer (2010)). The absence
of this behavior indicates the growth is not in the layer-by-layer mode. It
is not observed at the beginning of the growth process since there are no
complete layers. The general feature of this intensity is an initial drop.
Then, the oscillatory behavior appears. The reason for the initial drop can
be attributed to the intermixing between the growing layer and substrate or
poor layer-by-layer growth. After that, the layer-by-layer growth prevails.
An example of intensity as a function of time for growing Co2 MnSi is given

in Fig. 2.3.
Fig. 2.3 The oscillatory behavior of RHEED intensity at (00) varying with time (Wang
et al., 2005a).
2.3.2 X-ray reflectometry (XRR)

Many methods, such as atomic force microscopy (AFM) and X-ray photo-
electron spectroscopy are limited by their sensitivities to a few nm. X-ray
reflectometry—which was developed for monitoring the quality of transis-
tors, in particular the interface of Si/SiO2 and the layer thickness of SiO2 —
is still a viable tool for determining the thickness of a film on a nm scale,
with an advantage of nondestructivity (Wang et al., 2005a).

The source of X-ray is typically Cu K lines with wavelength at 0.154 nm
for semiconductor applications. Synchrotron radiations with wavelength of
the order of 1.0 nm have also been used.
2.3.2.2 Schematic setup

A schematic diagram for X-ray reflectometry is shown in Fig. 2.4. X-ray
source is indicated by the light grey rectangle. X-ray is reflected from a
graded multilayer mirror called Gobel mirror (tilted black rectangle) and
is passed to the first slit, S1 , before arriving at the sample (the thin tilted
rectangle). The reflected X-rays pass two slits, S2 and S3 (both also shown
as broken strips) to block the stray X-rays, then reach the detector (slightly
titled vertical rectangle).
Fig. 2.4 A schematic setup for X-ray reflectometry.

Two quantities are measured: the specular reflected X-ray and diffuse X-
ray. They are functions of angle 2θ—the angle between the incident and
reflected X-rays. Typical results of these two X-rays are shown in Figs. 2.5
and 2.6. In Fig. 2.5, the specular X-ray signal is measured in units of
counting rate (cps) and is shown as a function of θ. At small θ, the intensity
is constant. Then, it follows a region of having a negative slope. The angle θ
at which the slope changes from zero to negative can be used to determine
the density of the substrate. The slope is affected by the roughness of
the substrate, the interface roughness and thickness of the film causing the
reduction of intensity. The modulations (wavy form) are due to interference
of reflections between the surface of the sample and the interface of the
sample and substrate. In Fig. 2.6, a typical diffuse signal is plotted by
Bahr et al. (1993). This set of curves can only be measured when there is
surface or interface roughness. The two side peaks come from total external
reflection (sample-air) with incident and exit angles at the critical angle of
total reflection. They are called Yoneda wings (Yoneda, 1963). The center
peak originates from specular reflection. The curve without the center
peak is determined by the surface layer. A recent development of X-ray
reflectometry is reported by Sacchi et al. (2007). They used pinholes to
control the coherence of the X-ray and constructed a holographic image of
the surface.
Fig. 2.5 Measured intensity of specu- Fig. 2.6 Intensity of diffuse X-ray re-
lar X-ray reflection as a function of θ flection as a function of θ (Bahr et al.,
(Toney and Brennan, 1989). 1993). The Yoneda peaks are marked
by arrows.
2.3.3 Scanning tunneling microscopy (STM)

The scanning tunneling microscope (STM) was developed by Binnig et al.
(1983). It is an extremely useful tool for surface science. It involves a
metallic tip and the sample for which the surface structure is to be deter-
mined. Over the years, the technique has been much refined. A schematic
diagram is shown in Fig. 2.7. The tip scans across a section of the surface.
The height h is the separation between the tip and top layer of the surface.
It is controlled by a feedback mechanism to either maintain at an absolute
height—to form the STM image providing the morphology of the surface—
or at a fixed value with respect to the top layer of the surface by keeping
the measured current constant.
Fig. 2.7 A schematic diagram of STM. h is the separation between the tip and surface.
An ammeter (A) is shown as a large circle. The voltage “V” is maintained between the
tip and surface.

The tunneling current, the reading in the ammeter, as a function of h at
any point on the surface is measured. The theory of the tunnelling current
was worked out by Tersoff and Hamann (Tersoff and Hamann, 1985). Two
quantities are obtained from measurements of the tunnelling current: if h
is kept constant, the current in the ammeter is measured. On the other
hand, if a constant current is maintained, then h is measured. Then from
the amplitude of the current, a STM image can be deduced. This latter
scheme was adopted by Wang et al. (2005a). The STM image for Co2 MnSi
obtained by Wang et al. (2005a) is shown in Fig. 2.8. By guiding the tip
to move along a line, h as a function of position along the line determines
the profile of the surface morphology and the thickness of the layer growth.
From the height of the profile, it is possible to determine the layer thickness.
As shown in Fig. 2.8, the height of a/2 (2.85 Å) is obtained, where a is the
cubic edge. It reflects the height of islands formed above the substrate
during the growth. This distance is half of the cubic edge for the full
Heusler alloy.
Fig. 2.8 STM line profile along a line parallel to the surface. a/2 = 2.85 Å, where a is
the lattice constant (Wang et al., 2005a).
This STM height determination inferred by the RHEED period of os-

cillations reflects the growth of two atomic layers instead of four for a full
unit cell. This half of the unit cell growth is required to satisfy the chemical
composition and electrical neutrality.
2.3.4 Auger electron spectroscopy (AES)

2.3.4.1 Basic processes
A photon with energy higher than 1.0 keV excites a core state of an atom in
a solid. The hole created in the core region is annihilated by a valence elec-
tron. The energy released from the recombination excites another valence
electron. Pictorially, the processes shown in Fig. 2.9. The left figure shows a
core hole (open circle) being created by the incident X-ray. The right figure
describes the hole being annihilated by a valence electron (dark circle) and
another valence electron (gray circle) being excited above the vacuum level.
The processes involve core states which can exhibit the characteristics of
each atom.
Fig. 2.9 Processes in AES. Left figure shows a core hole (open circle) being created by
an incident X-ray. Right figure describes the hole being annihilated by a valence electron
(dark circle) and another valence electron (gray circle) being excited above the vacuum
level.
2.3.4.2 Simplified experimental setup

In most cases, it is desirable to determine the extent of the intermixing of
atoms in the substrate in situ. Therefore, the setup is more or less identical
to the growth process except that the source of the X-ray is installed in the
chamber to have the beam incident on the sample and a spherical mirror to
detect the secondary electrons is positioned above the sample. The source
of X-rays depends on the atoms to be detected. If GaAs(100) is used as a
substrate, then Mg (1253.6 eV) or Al (1486.6 eV) K-line is used to detect
the Ga line (1070 eV) and As line (1228 eV).

The peak-to-peak intensity of the atoms is plotted against the thickness
of the sample. A typical example of Co2 MnSi on GaAs(001) is shown in
Fig. 2.10. By measuring the energy of the emitted secondary electron, the
gray circle shown in Fig. 2.9, the chemical species can be determined.
Fig. 2.10 Peak-to-peak intensity as a function of thickness in Co2 MnSi at 450 K (Wang
et al., 2005a).
2.3.4.4 Remarks
Because of the hole created in the core, AES can involve significant many-
body interactions due to possible polarization effects induced by the hole.
For simply determining the presence of certain species of atoms, however,
it is possible to ignore such complications.
2.4 Methods of determining physical properties
In this section, we focus on methods of determining magnetic, transport,

and half metallic properties. These methods are generally applied to all
HMs.
2.4.1 Magnetic properties

With spintronic applications in mind, the important magnetic proper-
ties for a half metallic sample are the saturation magnetization, magnetic
anisotropy, spin polarization at EF —polarization P, and the Curie temper-
ature (TC ). All of these properties have been studied both experimentally
and theoretically. The experimental aspects will be discussed first.
Among all of these properties, the saturation magnetization is the most

fundamental for ferromagnetic properties. It is the saturation value of a
hysteresis loop B as a function of H, where B is the magnetic induction
and H is the applied magnetic field. For half metallic compounds, it is
important that this quantity is observed at or above room temperature
(RT) in order to fabricate practical spintronic devices. Its value also can be
compared to the theoretically predicted magnetic moment/unit-cell. The
methods used to measure a hysteresis loop include:
• Superconducting quantum interference device (SQUID) based mag-

netometer.
• Technique utilizing the magneto-optical Kerr effect (MOKE).
2.4.1.1 SQUID based magnetometer

The hysteresis loops of both bulk and thin-film samples can be measured
by this method. Since SQUID—a sensitive device to measure the magnetic
field—is commercially available, we shall not elaborate on the method.
Basically, a dc magnetic field H is applied along the easy axis of the
sample. For thin-film samples, this axis, in general, lies in the plane of the
film. The magnetization is measured as a function of the strength of the
dc magnetic field.
Schematic setup In Fig. 2.11, a schematic setup of the measurement is

shown. The large white area is the substrate, the sample is denoted by a
small gray parallelogram. A dc magnetic field H is applied to the sample,
indicated by an arrow. The SQUID is shown as a loop. It measures the B
field from the sample.
Fig. 2.11 A schematic setup for the SQUID magnetometer. A dc magnetic field H is
applied to the sample. The SQUID is shown as a loop. It measures B field from the
sample
What is measured The hysteresis loop is in general measured. Often,

the quantity B is converted into magnetic moment. In Fig. 2.12, magnetic
moment of Co2 FeSi as a function of H is shown (Wurmehl et al., 2005).
Fig. 2.12 Magnetic moment of Co2 FeSi as a function of H at different temperatures

(Wurmehl et al., 2005).
2.4.1.2 Magneto-optical Kerr effect (MOKE)

Basic idea The Kerr effect is the linear polarization of an electric field
of incident light reflected from a magnetized sample is changed to ellipti-
cal polarization—the Kerr rotation. It was first applied to study surface
magnetism by measuring the hysteresis loop of a magnetic thin film in the
monolayer range (Bader, 1991). Later, it has been used to search for mag-
netic ordering, to identify major magnetic anisotropy, to correlate TC to
the layer thickness, and even characterize the critical exponent in a 2D
phase transition.
Schematic setup The setup is relatively simple, a schematic diagram is

shown in Fig. 2.13. It can be incorporated into the growth vacuum chamber
by adding an inlet for the incident light and an outlet for the reflected light
if an external dc magnetic field H can be provided to the sample. If H
is parallel to the film, it is called the longitudinal Kerr effect. The light
source is a polarized light. The detection should include a polarizer.
What is measured The main quantities to be measured are the intensity

of the reflected light and Kerr rotation which is proportional to the magnetic
moment of the sample. By measuring the intensity of the reflected light,
a hysteresis loop can be determined, and the coercivity can be extracted.
The Kerr rotation itself does not provide the absolute value of the magnetic
Fig. 2.13 Schematic setup of MOKE.
moment of the sample. The temperature dependence of the Kerr rotation

is useful to determine M(T)/M(0)—a relative quantity.
There are two ways to exhibit a hysteresis loop. One is to measure the
Kerr rotation as a function of H. The shape of the loop and remanence
are also functions of the layer thickness. The other one is to measure the
intensity through the polarizer. With the polarization direction fixed in the
polarizer, the intensity of the light passing through the polarizer changes as
the strength of the dc magnetic field H varies. The two hysteresis loops are
shown in Fig. 2.14. The sub-loop shown in light gray in Fig. 2.14(a) relate
Fig. 2.14 (a) Hysteresis loop shown in Kerr rotation as a function of H. Results from
different thickness are shown in black and light gray lines. Black loop is for thicker and
light gray loop is for thinner samples. (b) Hysteresis loop shown in terms of the Kerr
intensity as a function of H (Bader, 1991).
to the thickness of the sample. Let us take Co2 MnSi grown on GaAs(001) as
an example. After growing three bilayers, the sample starts to exhibit an in-
plane uniaxial magnetic anisotropy having the easy axis pointing along the
[11̄0] direction. The experiment of measuring the hysteresis loop was carried
out to probe the magnetization along the hard axis—the [110] direction.
Between 4 and 20 bilayers, there are two easy axes in the (110) plane.
Hysteresis loops are developed which are shown in gray in Fig. 2.14(a) as a
function of positive and negative external magnetic fields. The remanence
field is zero. As the film grows thicker—over 35 bilayers, the sub-loops
disappear. The following empirical relation is found: let Hs1 and Hs2 be
the fields in the middle of sub-loops at positive and negative H, then
Hs = (Hs1 − Hs2 )/2. (2.1)
The empirical relation is Hs ≈ 1/d, where d is the thickness of the film.
When H is applied along the easy axis, a simple hysteresis loop is obtained.
In Fig. 2.14(b), the hysteresis loop is in terms of the Kerr intensity as a
function of H.
The Kerr rotation under the remanence field along the easy axis can
be measured as a function of temperature. If the Kerr rotation is assumed
to be proportional to the magnetization, M, then the measured results fit
well with the Bloch formula indicating the excitations are spin waves. The
Bloch formula is
M (T ) = M (0)(1 − bT 3/2 ). (2.2)
Because MOKE cannot measure the absolute magnetization, it can only
probe the magnetization in the surface region with penetration depth of the
order of 10 nm. For Heusler alloys, this is a real disadvantage, because a
complex layer stacking can happen to diminish magnetic properties at the
surface. Alternative methods to measure the hysteresis loop are the alter-
native gradient magnetometer (AGM) and vibrating sample magnetometer
(VSM). Since these two methods are specific for Heusler alloys, the related
discussions will be given in Chapter 3.
2.4.1.3 X-ray magnetic circular dichroism (XMCD)

Samples like Co2 MnSi can have its moment contributed by multiple atomic
species, such as Co and Mn. In order to determine individual contributions,
XMCD can be used. The basic principle of the method has been discussed
by Stöhr (1999).
Basic principles Within one-electron model, an X-ray with either right-

or left-hand circular polarization excites a core electron to a state above the
vacuum level. Core states of a transition metal (TM) element being excited
normally have orbital angular momentum of l = 1, the 2p core states. With
the spin-orbit coupling, core states are identified as L3 (l + s) and L2 (l − s)
lines, where s is the spin angular momentum. These lines are called L-
edges. The XMCD signal is the normalized difference in absorption between
the right- and left-hand polarized X-rays. It contains the information of
exchange splitting and the spin-orbit coupling of the initial and final states.
By controlling the energy of the X-ray, it is possible to have XMCD probe
a particular element. After absorbing the X-ray, the core electron changes
its angular momentum due to the circularly polarized light. It is excited to
either an s- or a d-like state above the vacuum level. Spin-polarized bands
respond differently for different polarizations of the X-ray. Measurements of
photoelectrons with respect to the spin polarization of the sample provide
information about the magnetic moment of each TM element in the sample.
The processes are depicted in Fig. 2.15.
Fig. 2.15 Processes in XMCD. L2 and L3 are the core energy levels. The thick arrow
indicates the X-ray with the left and right polarization shown as circular arrows. Vertical
double arrows indicate possible transitions of core electrons. The spin-polarized density
of states (DOS) are shown in shaded areas (Stöhr, 1999).
Schematic setup It is not easy to get a commercial X-ray source for

frequencies in the desired range of a specific TM element. For Mn, energies
of L3 and L2 lines are 638.5 eV and 651.4 eV, respectively. Tunable X-ray
sources are available at Stanford Synchrotron Radiation Laboratory in the
U.S., the Deutsches Elektronen-Synchrotron in Germany and First Dragon
Beamline of NSRRC in Taiwan. With the light source, a schematic diagram
of a typical setup is illustrated in Fig. 2.16.
Bias rings are used to eliminate the photoelectron possibly from the
gold mesh and stray electrons from the sample. The photocurrent detector
measures the photoelectrons from the sample. The detector is made of
materials which are transparent to the X-ray.
Fig. 2.16 Simple schematic setup of XMCD. X-ray is incident from the left. The dashed
line is a Au mesh. The middle line at the right of the dashed line is the Al window.
Horizontal lines are bias rings. The gray rectangle is the sample. The black box is the
magnet with its south pole close to the sample and is separated from the sample by
a quartz window. Emitted electrons are collected by the photocurrent detector (gray
vertical line) (Stöhr, 1999).
Fig. 2.17 XMCD at L2,3 absorption edge of Mn for 45Å thick Co2 MnSi films at RT
(Wang et al., 2005a).
What is measured The main quantities to be measured are the so-

called white lines. They are the total photocurrents from L2 and L3 lines.
By reversing the dc magnetic field, two sets of white lines are measured.
The important magnetic circular dichroism (MCD) curves are obtained by
subtracting two sets of white lines. A typical set of data for the difference
of photocurrents associated with L2 and L3 lines (Wang et al., 2005a) is
shown in Fig. 2.17.
2.4.2 Transport properties

2.4.2.1 Magnetic tunnel junctions (MTJs)
Basic principles These junctions allow electrons to tunnel from one fer-
romagnetic film through an insulating region to another ferromagnetic film
serving as a detector of spin polarization. The spin polarizations in such
electrodes are defined by the DOS at EF (Eq. (1.1)) and are denoted by P1
and P2 , respectively. These junctions can be used to measure the resistance
change if the magnetization of the detector is reversed by some controlled
magnetic field.
Schematic setup The schematic setup is shown in Fig. 2.18. The two
ferromagnetic thin films are indicated by gray and black regions and are in
contact with an ammeter denoted by an open polygon and a battery. The
mid region between the gray and black regions is the insulating region.
Fig. 2.18 Schematic diagram of magnetic tunnel junctions.
What is measured Spin polarization of the sample is determined by

measuring the tunnel magnetoresistance (TMR). The TMR can be ex-
pressed in terms of spin polarizations:
2P1 P2
T MR = , (2.3)
1 − P1 P2
where P1 and P2 are spin polarizations at EF in the electrodes, and have
been defined in Eq. (1.1). Note that it is also possible to measure directly
the magnetoresistance, MR, by reversing the magnetization in the detecting
electrode. The magnetoresistance is expressed as:
R↑↓ − R↑↑
MR = , (2.4)
R↑↑
where R↑↑ and R↑↓ are the resistance in parallel and antiparallel configura-
tions in the two electrodes. In practice, measurements of TMR and MR are
complicated by the presence of insulating layers. Cautions in interpreting
the data are called for due to scatterings such as phonons.
These two kinds of results are given in Fig. 2.19. Figure 2.19(a) shows
TMR as a function of bias voltage and Fig. 2.19(b) shows MR as a function
of external field. The results also depend on the temperature and magne-
tization of electrodes. Figure 2.19(a) shows the RT case. The asymmetry
shown in Fig. 2.19(a) originates from the tunneling process dependent on
the direction of the bias voltage.
Fig. 2.19 (a)A typical plot of TMR as a function of bias voltage (Sakuraba et al., 2006),
and (b) a typical plot of MR as a function of external field (Gong et al., 1997)
.
To see how the polarization can be determined, we follow the method

suggested by Pesavento et al. (2004) and Bezryadin et al. (1998). Here we
use the conductance instead of the resistance. For the parallel magnetiza-
tion configuration of two electrodes, we expect that
G↑↑ = n1 n2 + (1 − n1 )(1 − n2 ) (2.5)
G↑↓ = n1 (1 − n2 ) + (1 − n1 )n2 , (2.6)
where G↑↑ and G↑↓ are conductances of the parallel and antiparallel configu-
rations, respectively. n1 is the fractional number of majority-spin electrons
in the DOS at EF in electrode 1, and n2 is the corresponding fractional
number of electrons in electrode 2. The number of minority-spin electrons
is 1 − ni , i = 1, 2. G↑↑ is the sum of the conductance from electrons with
both electrodes having parallel orientations of majority- and minority-spin
channels. G↑↓ is for opposite spin orientations in the two electrodes. G↑↑
and G↑↓ are measured to determine n1 and n2 . The polarizations in elec-

trodes are:
P1 = 2n1 − 1 (2.7)
P2 = 2n2 − 1. (2.8)
2.4.2.2 Resistivity
General discussion For bulk materials, samples are often in rod form
with axis along the growth direction. For samples in thin-film form, the re-
sistivity ρ is commonly measured in the plane of the sample. The resistivity
is measured as a function of temperature. The temperature dependence is
assumed to be:
ρ = ρo + cT n , (2.9)
where ρo is the residual resistivity—the low temperature limit. The expo-

nent n is determined by fitting ln(ρ − ρo ) as a function of ln(T ). Three
distinct temperature ranges were identified for the linear fits to Eq. (2.9)
(Ritchie et al., 2003). This information is used to find whether there is a
structural or magnetic transition. At TC , the resistivity shows a change in
slope. In most cases, the slope changes from a mild slope, n ≈ 2, to a steep
one. The mild slope in the ferromagnetic phase has been attributed to the
absence of magnetic fluctuations; the sample behaves like an ordinary fer-
romagnetic material. n = 3 indicates the contribution of phonons to the
resistivity. A magnetic phase transition can be detected by the observed
power-law shift.
Comments on setup For both bulk and thin-film samples, the standard
way of measuring resistivity is to attach four electrodes to the sample.
This is called the “four-probe” arrangement. In Fig. 2.20, a typical four-
probe arrangement is shown. One of the advantages of the four-probe
arrangement is that it reduces the effects of the contact resistance between
the sample and electrode (Pesavento et al., 2004; Bezryadin et al., 1998).
What is measured The simplest measurement is ρ as a function of T.

Typical data is shown in Fig. 2.21(a). In order to get the exponent n, the
data are analyzed in terms of a plot of ln(ρ − ρo ) as a function of ln(T ).
Figure 2.21(b) shows a typical plot of ln(ρ − ρo ) as a function of ln(T ).
Fig. 2.20 A typical four-probe arrangement, (a) the side view and (b) the top view
(Pesavento et al., 2004).
Fig. 2.21 (a)A typical plot of Resistivity (ρ) as a function of temperature. (b) A typical
plot of ln(ρ − ρo ) as a function of ln(T ).
2.4.2.3 Hall conductivity

Basic principles The Hall effect is a result of the application of mutu-
ally perpendicular electric and magnetic fields on a conductor (Fig. 2.22).
Moving charges experience a force, called the Lorentz force, when a mag-
netic field (Hz ) is present that is perpendicular to their motion (Ix ) and
their paths are curved so that charges accumulate on the upper side of
the material. An electric field Ey is then developed along the y-direction.
When Ey = −ωc τ Ex , where ωc is the cyclotron frequency determined by
Hz and τ is the lifetime of an electron when the moving electrons are no
longer deflected, a voltage VH = Ey y is developed. This is the Hall effect,
where VH is called the Hall voltage. The quantity of interest is called the
“Hall coefficient” which is defined as:
RH = VH z/(Ix Hz ), (2.10)
or
RH = Ey /(jx Hz ), (2.11)
where VH = Ey y, Ix = jx Ayz , and Ayz is the cross sectional area perpen-

dicular to the current. The current density J in normal metals or heavily
doped semiconductors is given by
J = σH H × E, (2.12)
where σH is called the Hall conductivity.
When Hall carried out his experiment on Fe, he found a coefficient about
ten times larger than that of Ag. This effect in ferromagnetic materials is
now known as the “anomalous Hall effect” (AHE). Husmann and Singh
(2006) expressed the current as:
J′ = σ ′ E × M, (2.13)
′ ′
where J is the anomalous Hall current density, and σ is the corresponding
conductivity. The total current is the sum of these two currents J and
J′ . Physical origins of AHE have been classified into extrinsic and intrin-
sic mechanisms. The extrinsic mechanism includes spin-orbit interaction,
impurities, and phonons. The intrinsic mechanism has been attributed to
either the band structure (Yao et al., 2004) or a topological spin background
(Ye et al., 1999).
Fig. 2.22 The relative orientations of the magnetic field, current, and electric field. The
sample is shown in gray. Current Ix flows in the x direction. The magnetic field Hz is
in the z direction (into the paper shown by the cross). The electric field Ey is in the y
direction.
Schematic setup A schematic setup for a Hall measurement is shown

in Fig. 2.23 (Moos et al., 1995). The Hall voltage VH is probed by the
potential difference between electrodes at the two sides of the material. The
two electrodes in the lower side of the material can improve the accuracy
of VH measurement. Initially, when H=0, the resistor R is adjusted so
that VH is zero. The electrodes used to probe VH need to have small cross
sections in order to reduce stray current.
Fig. 2.23 Schematic diagram for Hall measurement. The sample is shown as the long
gray bar. The current is measured by the ammeter A. H is the strength of a magnetic
field pointing into the paper (shown as a cross). VH is the Hall voltage.
Fig. 2.24 (a) Longitudinal resistivity ratio (∆ρxx /ρxx ) as a function of H (Husmann
and Singh, 2006); (b) Transverse resistivity (ρxy ) as a function of H (Lee et al., 2004);
and (c) Magnetization M, as a function of H (Lee et al., 2004).
What is measured Two quantities, the longitudinal σxx and the trans-
verse σxy Hall conductivities as a function of H at different temperatures
are commonly measured. Husmann and Singh (2006) measured ∆ρxx /ρxx
as an alternative. The anomalous Hall coefficient RAH is then extracted
with the information from either longitudinal resistivity ratio ∆ρxx /ρxx as
a function of H (Fig. 2.24(a)) or transverse resistivity ρxy as a function of
H (Fig. 2.24(b)) measurements at different temperatures. For pure ferro-

magnetic samples, RAH is typically positive and monotonically decreasing
with increasing temperature (Lee et al., 2004). With disorder, RAH can be
negative.
2.4.3 Half metallic properties

There are many predicted and grown half metallic compounds which also
exhibit ferromagnetic (FM) properties. Therefore, one would expect that
spin-polarized photoemission (SPP), FM-superconductor tunnel junction
experiments, and the Andreev reflection method (Andreev, 1964) can be
used to detect half metallicity. The objective is to determine half metallic
properties of the samples and the spin polarization P at EF . However,
the first two experiments are not easily to be carried out. The energy
resolution of the SPP method is needed within 10−3 eV, because only spin
polarization at EF is probed. The FM-superconductor tunnel junction
requires a strictly uniform oxide layer formed on top of the superconductor
such that the sample can be grown on top of the oxide layer for sensitive
measurements of spin polarization. A more appealing approach to probe
half metallicity is the Andreev reflection method. This method has been
applied to CrO2 by Soulen et al. (1998) and Ji et al. (2001).
The SPP method is useful to determine the spin-polarized band struc-
ture and the FM-superconductor tunnel junction was the first one being
introduced for measuring the spin polarization. While a more powerful one
is the spin-polarized angular resolved photoemission method which we shall
discuss first.
2.4.3.1 Spin-polarized angular resolved photoemission

spectroscopy (ARPES)
Basic principles A simple description of the geometry in an ARPES
experiment is shown in Fig. 2.25(a). Light (the wavy line) with energy
~ω is incident on a solid. An electron (gray arrow) is then emitted. The
energy of the emitted electrons satisfies the following conditions depending
on whether the electron originates at EF or from a valence state below EF .
Ekin = ~ω − Φ (2.14)
Ekin = ~ω − Φ − Eb , (2.15)
where Ekin is the maximum energy of the emitted electron from a state
at EF , ~ω is the energy of the incident light, Φ is the work function, and
Eb is the binding energy with respect to EF . The momentum relations,

between the electron inside and electron outside the sample, determining
the angular resolved spectra are
kout
k
= kink
±G (2.16)
r
out 2mEkin out 2
k⊥ = − kk , (2.17)
~2
where kout
k is the momentum of the emitted electron parallel to the sur-
face of the sample, kin k is the parallel component of the momentum of the
electron inside the sample, G is a reciprocal lattice vector parallel to the
out
surface, and k⊥ is the magnitude of the momentum of the emitted electron
out
perpendicular to the surface. Both kout k and k⊥ can be specified by the
polar and azimuthal angles (θ, φ) shown in Fig. 2.25(a).
Fig. 2.25 Geometric configurations for (a) regular energy and angular resolved photoe-
mission, (b) longitudinal magnetization, and (c) transverse magnetization. θ and φ in
(a) are the polar and azimuthal angles.
An electron excited by the incident light can involve in many-body in-

teractions as it travels toward the surface. The effects of these interactions
can be discerned by comparing measured spectra to band structure results
(Braun, 1996).
Remarks To probe electronic states in a ferromagnetic sample using pho-

toemission process, one must distinguish the two distinct polarizations as-
sociated with spins of electrons. It was not until 1979, with the work of
Celotta et al. (1979), that significant advances were made experimentally.
Earlier, only the geometry of the longitudinal magnetization (Fig. 2.25(b))
was considered. In this geometry, a large external magnetic field is required
to align domains of the sample in such a way that the magnetization is per-
pendicular to the surface. Consequently, the resultant magnetic field can
deflect the emitted electrons— making it impossible to carry out angular

dependent photoemission measurements.
Celotta et al. (1979) suggested the use of a long, thin-plate form of
sample, with the magnetization parallel to the surface of the sample with-
out applying external field. In the transverse magnetization experiment
(Fig. 2.25(c)), the sample is permanently magnetized. The stray field out-
side the sample is in general too small to have a significant effect on the
trajectory of the emitted electron. One may also employ a particular elec-
tron extraction lens system to ensure that emitted electrons will not be
affected by the presence of the stray field.
The emitted electrons can have two spin polarizations with respect to a
chosen axis. To detect net polarizations, one carries out double scattering
measurements. The first scattering selects a particular spin polarization.
The next step utilizes the scattering process proposed by Mott to determine
the cross section of each spin polarization. It is called Mott scattering.
Mott scattering In its original form, the Mott scattering characterizes

the Coulomb scattering cross section modified by exchange effects when two
equivalent electrons scatter (Bohm, 1951). To understand spin-polarized
photoemission, Mott scattering is applied to an electron having its spin
momentum coupled to its orbital angular momentum when the electron
is scattered by the Coulomb interaction from a target composed of heavy
element, such as gold (Au). The effect of this spin-orbit (S-O) interaction
causes a beam of electrons with a particular spin orientation passing at
the left and right of the target to show asymmetry. We expect that a new
generation of light source for photoemission experiments will be available
in the near future, it is therefore worth to provide the background of the
S-O interaction. We start with the Dirac equation in 2 × 2 matrix form
(Kessler, 1985).
E − eΦ − cα · (p − eA/c) − βmc2 ψ = 0,

(2.18)
where E is the energy of the electron, Φ is the external potential, and α
and β are 2 × 2 matrices

0 σi 10
αi = ;β = (2.19)
−σi 0 01
where i = 1, 2, 3 or x, y, z, and

01 0 −i 1 0
σ1 = , σ2 = , σ3 = , (2.20)
10 i 0 0 −1
p is the momentum of the electron, A is the vector potential for an exter-

nally applied magnetic and electric fields, Φ is the Coulomb potential from
the nuclei and m is the mass of the electron. By expressing Eq. (2.18) in
a 2 × 2 matrix equation and the wave function into large and small com-
ponents with respect to the rest mass, the resultant equation of the large
component wave function is the 2 × 2 Schrödinger equation:

1 eA 2 e~ e~ e~
(p − ) + eΦ − σ·B+i E·p− σ · EC × p ψ = Es ψ,
2m c 2mc 4m2 c2 4m2 c2
(2.21)
where the first two terms are the usual Schrödinger equation for an electron
in the presence of an external field. The third term is the Zeeman term from
e~
the spin moment µ = 2mc σ of the electron in the external magnetic field
B. The fourth term is from an external electric field. The last term can
be reduced to the spin-orbit (S-O) interaction. In the electron reference
frame, where the electron with velocity v sees the nucleus of the heavy
atom moving at v. The magnetic field BC due to motion of the nucleus
and the Coulomb interaction from the nucleus (the current associated with
the nucleus) is
EC EC × p
BC = −v × = , (2.22)
c mc
and EC is the electric field from the Coulomb potential, V.
1 dV r
EC = − . (2.23)
e dr r
Now, the magnetic moment of the electron interacts with BC . With s = ~2 σ,
Eqs. (2.22), and (2.23), the last term in Eq. (2.21) can be expressed as:

e 1 dV r 1 1 dV
−µ · BC = s· − ×p = (l · s), (2.24)
2m2 c2 e dr r 2m2 c2 r dr
where l = r × p.
The effects of this term can be seen as follows: let an electron pass
to the right of the target (an Au atom). The orbital angular moment is
oriented along the z-axis in the frame of the moving electron. When the
electron occupies the ↑ spin state, with orientation aligned along the z-
axis, it experiences an additional positive potential from Eq. (2.24) with
respect to the Coulomb potential. This resultant potential V↑ is shown as
a solid curve in Fig. 2.26. The Coulomb potential is shown as a dashed
curve. Similarly, the ↓ spin electron experiences an additional negative
potential and the resultant potential V↓ is shown as a dash-dot curve in
Fig. 2.26. Different potentials for the ↑ and ↓ spin states result in different
cross-sections. If an electron passes to the left side of the target, the orbital
angular momentum reverses direction, so that the extra potential due to the
last term acting on the ↑ and ↓ spin states is also reversed. The phenomenon
is depicted schematically in Fig. 2.27.
Fig. 2.26 Different potentials experienced by the ↑ and ↓ spin states in Mott scattering.
Fig. 2.27 An electron passing to the left (top) (dark grey arrows for the spin and
velocity) and right (bottom) (light grey arrows) of an Au atom experiences different S-O
interactions due to the reversal of the effective B field. v is the velocity of the electron
and S is the spin of the electron.
Consequently, when a beam of electrons with a definite spin polarization

scatter from a target of heavy atoms, there is an asymmetry recorded in
the detectors located at the left and right of the target. This is the basic
idea behind the Mott detector.
As discussed above, the setups can involve delicate lens systems. One
of the more detailed setups is designed by Raue et al. (1984) and enables
one to determine the energy and angular resolution of spin-polarized pho-
toelectrons. A schematic diagram of the setup is shown in Fig. 2.28(a). In
which a vacuum chamber contains a ferromagnetic sample, manipulator,
mirror with focusing system, and other electron lens systems. At the end
of the chamber, a 90◦ spherical deflector deflects photoelectrons into the
Mott detector or Mott spin polarimeter by passing through another set of
lens systems. The spherical deflector serves also as an energy analyzer. At
its exit plane, an assembly of electron multiplier arrays are installed for
multichannel detections.
Fig. 2.28 (a) Schematic of a photoemission experiment. (b) A retarding-potential type

Mott detector.
The Mott detector is based on the principle described in Mott scattering

section. The detector is either a conventional high energy type used by Get-
zlaff et al. (1998) or a retarding-field type used by Uhrig et al. (1989). The
former exhibits a drawback of poor discrimination against inelastic scatter-
ing incurred by electrons. The retarding-field type Mott detector consists of
an energy analyzer and electron deflectors. It has the capability to discrim-
inate the energy of electrons and monitor the polarization of an electron
even with spatial asymmetry caused by the S-O interaction. However, a
slight asymmetric alignment of the instrument can smear the polarization
measurements. A way to eliminate this smear is to include measurements
with reversed spin polarizations of electrons. A retarding-field type Mott
detector is shown in Fig. 2.28(b). The outer two hemispheres provide the
acceleration to incoming photoelectrons. The electrons scatter from the
gold film with the S-O interaction. The scattered electrons enter the left
detector (DL ) and the right detector (DR ). Both detectors are equipped
with electron multiplies to increase the sensitivity of measurements.
The operational details are given below: The asymmetry is defined as
the ratio of the left current IL and right current IR with respect to the
target
IL
A= . (2.25)
IR
Fig. 2.29 Sherman function as a function of θ, the dashed curve is for the electronic
energy at 50 keV and the solid curve is for 120 keV (Gellrich and Kessler, 1991).
The polarization in the Mott detector is given as

A−1
PMott = Sef f P = , (2.26)
A+1
where Sef f is call the Sherman function (Fig. 2.29) and
N↑ − N↓
PN = . (2.27)
N↑ + N↓
The Sherman function is related to the scattering amplitude and phase
shifts, as can be seen as follows: Let us first assume that the wave function
of the ↑ spin channel after the scattering is given asymptotically (r → ∞)
by
f1 (θ, φ)eikr
ψ ↑ → eikz + . (2.28)
r
Because of the S-O interaction, there is a possibility of having a spin-flip
transition,
g1 (θ, φ)eikr
ψ ↑→↓ → . (2.29)
r
Similarly for the ↓ spin channel with quantization axis along the z-direction,
we have
f2 (θ, φ)eikr
ψ ↓ → eikz + (2.30)
r
with the spin flip transition,
g2 (θ, φ)eikr
ψ ↓→↑ → (2.31)
r
The amplitudes fi and gi , i = 1, 2, can be expanded into partial waves
with the corresponding phase shifts ηl and η−l−1 . The scattered waves fi
and gi can be written as
1 P∞ 2iηl
− 1) + l(e2iη−l−1 − 1) Pl (cos θ)

f1 (θ, φ) = 2ik l=0 (l + 1)(e (2.32)
= f (θ) = f2 (θ, φ),
1
P∞ 2iηl
+ e2iη−l−1 Pl1 (cos θ)eiφ

g1 (θ, φ) = 2ik l=0 −e (2.33)
= g(θ)eiφ ,
and
g2 (θ, φ) = −g(θ)e−2iφ , (2.34)
where Pl (cosθ) is the Legendre polynomial and Pl1 (cosθ) is the associated
Legendre polynomial. For an incident wave with an arbitrary spin orienta-
tion, the spinor is:

1 ikz 0 ikz
ψ=A e +B e . (2.35)
0 1
ikr ikr ikr

Af − Bge−2iφ eikr

a e f1 e g2 e
=A +B = . (2.36)
b r g1 r f2 r Bf + Ageiφ r
The cross-section is
|a|2 +|b|2 2 2 ∗ iφ ∗ −iφ
σ(θ, φ) = |A|2 +|B|2
= (|f | + |g| ) + (f g ∗ − f ∗ g) −AB |A|
e +A Be
2
+|B|2
2 2 ∗ iφ ∗ −iφ (2.37)
= (|f | + |g| )(1 + S(θ)) −AB |A|
e +A Be
2
+|B|2
.
S(θ) is defined as
(f g ∗ − f ∗ g)
S(θ) = 2 2 . (2.38)
|f | + |g|
It does not depend on ϕ and it can be determined by spin unpolarized

photoemission experiments.
Since the Mott detector is sensitive to misalignment of the incident beam
and electrons backscattered from walls, the experiments and interpretations
of the results must be carried out with great care.
What is measured The angle-resolved energy distribution curve (EDC)

(I0 = I↑ +I↓ ) and spin-polarized EDC (P = (I↑ −I↓ )/(I↑ +I↓ )) are measured
simultaneously (Raue et al., 1984). From these two measurements, the two
spin-resolved contributions I↑ and I↓ can be calculated as shown in the
lower part of Fig. 2.30. Experimental results on HMs, CrO2 in particular,
will be discussed in Chapter 4.
In the following, we present two schemes for probing spin polarized
electrons.
Fig. 2.30 Left panels: EDC and spin polarization for a normal emission from Ni(110);
Right panels: calculated spin-resolved EDCs (Raue et al., 1984).
2.4.3.2 Ferromagnet-superconductor tunneling

Basic principles Meservey and Tedrow (1994) were the first to carry
out the experiment on a Al–Al2 O3 –Ni junction, where Al is in a thin-film
form and exhibiting superconducting properties. It demonstrates the spin
polarized electron tunneling effects. The density of states of Al and Ni
are schematically drawn in Fig. 2.31(a). Without the external magnetic
field, the density of states of Al are denoted by solid curves. Because of
the superconducting gap, 2∆, is small (≈ kB T), Ni density of states are
approximated by constants, with the ↑ spin states marked by the dotted
line and ↓ spin states marked by the dashed line, respectively. The middle
region is an insulating barrier. When an external magnetic field H is applied
parallel to the film, the Al density of states are Zeeman splitted (shown as
the dashed (↓ spin state) and dotted (↑ spin state) curves, respectively).
For a finite H and a normal metal used as an electrode, the tunneling
current is expected to be symmetric with respect to spin polarization as
the voltage reverses. If the non-superconducting electrode is a ferromagnet,
then an asymmetry is expected for the ↑ spin and ↓ spin peaks. An expected
tunneling conductance is shown in Fig. 2.31(b). The asymmetry is indicated
by a, b, c, d peaks at H = 22.6 and 33.7 kOe.
What is measured The tunneling conductance, dI/dV , as a function of

the bias voltage, V, is measured. Figure 2.31(b) shows the typical results.
In practice, effects of the interface and S-O interaction for heavy elements
should be dealt with.
Fig. 2.31 (a) The Al (superconducting) density of states denoted by solid curves are
shown in the upper panel. ∆ is half of the gap. The middle horizontal region is an
insulating barrier. The bottom panel illustrates the Ni density of states. When an
external magnetic field H is applied paralleled to the Al thin film, the superconductor
exhibits Zeeman splitting. Dashed curves are for the ↓ spin states while dotted curves
are for the ↑ spin states. (b) Ratio of conductance as a function of voltage for different
values of H (Meservey and Tedrow, 1994).
2.4.3.3 Andreev reflection

Basic principles The basic principle of the Andreev reflection method
is illustrated in Fig. 2.32. It consists of the density of states involved in
the tunneling experiments of normal metal/superconductor (Fig. 2.32(a))
and HM/superconductor (Fig. 2.32(b)). It is one of the methods providing
unambiguous determination of a sample to be a half metal.
For the normal metal/superconductor junction (Fig. 2.32(a)), both
majority- (↑) and minority- (↓) spin states are occupied at EF . If a
majority-spin electron is incident (under a bias voltage, V, less than ∆)
to the superconductor region, a majority-spin hole (open circle) will con-
tribute to the reflected current. At the same time, a minority-spin electron
with proper momentum forms a Cooper pair with the incident electron and
leaves behind a minority-spin hole in the metal region (shown as an open
square in Fig. 2.32(a)). The extra current caused by this hole is the An-
dreev reflection. The conductance is contributed by two electrons passing
through the junction. If the normal metal is replaced by a ferromagnetic
metal, similar process applies even though there are more majority-spin
electrons than minority-spin electrons. On the other hand, if a HM is
considered, the tunneling carriers from the HM side have only one spin
orientation. It is impossible for the incident carrier entering the supercon-
ductor to pair with minority-spin electron due to the presence of the gap
(shown as an open circle in the HM region). As a result, a reflection by
the hole generated by the electron in the insulating channel can not appear
in the HM. The conductance is therefore zero. To apply this method to

ferromagnetic half metals, the junction is in a point contact form so the
uniform oxide layer is not needed.
Fig. 2.32 (a) Energy as a function of DOS of tunneling between a normal metal and
a superconductor. (b) Energy as a function of DOS of tunneling between a HM and a
superconductor. Arrows mark the spin polarizations of states in the metal.
Schematic diagram of setup Niobium (Nb) metal is commonly used

as the superconductor in Andreev reflection method. It is in the form of a
tip. A schematic diagram of the setup is shown in Fig. 2.33.
Fig. 2.33 Schematic of Andreev reflection method for detecting half metallic properties.
The gray rectangle is the sample. The Nb metal is in the tapered form serving as a point
contact. The circle “A” is the ammeter. “V” represents the battery.
What is measured The tunneling current, from the reading of the am-
meter, is the crucial quantity to be measured. To contrast characteristics
relevant to processes of the normal-metal-superconductor tunneling and
HM-superconductor tunneling, the corresponding tunneling currents as a
function of voltage are shown in Fig. 2.34. Note that in Fig. 2.34(a), the
current is zero only when the bias voltage is zero. However, there is a range
of 2∆ of the voltage for the tunneling current to be zero in Fig. 2.34(b).
Fig. 2.34 (a) The tunneling I-V curve between a normal metal and a superconductor;
(b) the tunneling I-V curve between a HM and a superconductor.
A phenomenological approach to the tunneling of spin polarized electron

has been given by Blonder, Tinkham, and Klapwijk (Blonder et al., 1982)
(BTK).
Analysis BTK introduced a model repulsive potential Vint δ(x) at the

interface to explain the metal-superconductor tunneling. One of the im-
portant parameters used in the model is the normalized barrier strength Z
(= Vint /(~υF )), where υF is the Fermi velocity, to characterize the inter-
face and effects of a barrier. BTK considered the current passing through
a plane in the normal metal (NM) region since it is contributed by single
particles instead of Cooper pairs. Furthermore, contributions to the cur-
rent are parametrized by four probabilities A, B, C, and D depending on
(E, ∆, Z) and the tunneling between NM and NM or between NM and SC,
where E is the energy of the incident electron and 2∆ is the gap of the
superconductor. The simplified notations are A(E), B(E), C(E), and D(E).
They characterize probabilities for the four processes.
To see the four processes described by Blonder et al. (1982), a simplified
band diagram is shown in Fig. 2.35 for a normal or ferromagnetic metal
(NM) and a SC in contact. At equilibrium, quasiparticle states labeled 0
and 5 at left and 1 to 4 at right are occupied with the same probability
f0 (E). There is a hole state at left labeled by 6. The lines for momenta in
the two regions are taken as positive when they are on the right-hand-side of
energy axes and negative when they are on the left-hand-side. Consider now
an electron labeled by “0” with momentum kin and E > ∆ incident from the
NM side into the SC. By matching the slope and value of the wavefunction
Fig. 2.35 The band diagram of a normal (NM) or ferromagnetic metal and a supercon-
ductor (SC) in point contact (Blonder et al., 1982). 2∆ is the gap of the SC. I (small
filled circle) is the incident electron. T− and T+ are the possible transmitted electronic
states. The hole state labeled by A is for the Andreev reflection.
at the interface, the incident electron can be subjected to transmission and

reflection processes labeled by A, B, C, and D. For example, C(E) is the
normal transmission probability through the interface without change of
sign of momentum q + → k + . D(E) is the transmission probability for
q + = −k − crossing the Fermi surface. A(E) characterizes the Andreev
reflection probability for a hole on the other side of the Fermi surface. A
hole appearing in the NM region is required for the incident electron to form
a Cooper pair in the SC. B(E) is the probability for an ordinary reflection.
The four parameters are obtained by solving Bogoliubov equations. These
four parameters satisfy the sum rule for the conservation of one electron:
A(E) + B(E) + C(E) + D(E) = 1. (2.39)
The tunneling current I is first expressed in terms of the difference of
the right going f→ (E) and left going f← (E) distribution functions,
Z∞
I = 2eAα Nσ (EF )υF σ [f→ (E) − f← (E)]dE, (2.40)
−∞
where Aα is the effective cross-section of the contact, Nσ (EF ) is the DOS

at EF for the σ spin state, υF σ is the Fermi velocity for the σ spin state,
and E is the energy of the incident electron. For complicated cases, the
two distribution functions must be solved self-consistently (Blonder et al.,
1982). Consider a situation:
f→ (E) = f0 (E − eV ) (2.41)
and
f← (E) = A(E)(1−f→ (−E))+B(E)f→ (E)+(C(E)+D(E))f0 (E), (2.42)
with
f→ (−E) = f0 (E + eV ), (2.43)
where f0 is the Fermi-Dirac distribution function and V is the poten-
tial energy due to an externally applied voltage. Substituting Eqs (2.41)
and (2.42) into Eq (2.40), the current from the normal metal to the super-
conductor is,
Z∞
IN M→SC = 2eAα Nσ (EF )υF σ {f0 (E−eV )−f0 (E)}{1+A(E)−B(E)}dE,
−∞
(2.44)
From this expression, the conductance can be obtained by taking the
derivative with respect to V. For an NM-NM contact, A(E) is zero be-
cause there is no Andreev reflection and D(E) is also zero due to no cross
Fermi surface scattering. With Eq. (2.42) and the sum rule Eq (2.39), the
conductance is
dIN M−N M
= 2eAα Nσ (EF )υF σ C(E) = Gn . (2.45)
dV
Gn is also the conductance for the NM-SC contact with V ≫ ∆. The
integrations in Eq. (2.44) were carried out by Blonder et al. (1982) for Z =
0.0, 0.5, 1.5, and 5.0, respectively. The results are shown in Fig. 2.36. For
Z=0 case, the normalized conductance is 2 for V < ∆. This is due to the
presence of a Cooper pair and Gn should correspond to one electron. For
|V | ≫ ∆, we have
G(V )
= 1. (2.46)
Gn
Fig. 2.36 Numerical G(V ) = dI/dV normalized to Gn as a function of V for Z = 0.0,

0.5, 1.5, and 5.0, respectively (Blonder et al., 1982).
To see how to extract the polarization information, we consider the

definition of the spin polarization. It is known for the tunneling junction
that the spin polarization PT can be determined by
N↑ (EF ) |T↑ |2 − N↓ (EF ) |T↓ |2
PT = 2 2, (2.47)
N↑ (EF ) |T↑ | + N↓ (EF ) |T↓ |
where Tσ is the tunneling matrix element for the σ spin channel.
For the point contact, it is not the tunneling matrix elements which
govern the process but the speed of electrons. The spin polarization for a
point contact Pc is expressed as
N↑ (EF )υF ↑ − N↓ (EF )υF ↑
Pc = . (2.48)
N↑ (EF )υF ↑ + N↓ (EF )υF ↑
Because the product Nσ (EF )υF σ is the current Iσ for σ channel, Pc can be
rewritten in terms of currents:
I↑ − I↓
Pc = . (2.49)
I↑ + I↓
This expression is based on the assumption of ballistic transport (i.e. no
scattering) at the point contact. In general, Pc should relate to the param-
eter Z introduced by Blonder et al. (1982) to characterize the scattering at
the contact. Strijkers et al. (2001) modified the current expression into the
sum of fully polarized current (Ipol ) and fully unpolarized current (Iunpol )
for the NM and SC contact,
I = (1 − Pc )Iunpol + Pc Ipol , (2.50)
where (1 − Pc )Iunpol is the unpolarized part of the current and can include
contribution from the Andreev reflection , Pc Ipol is the polarized part of
the current and does not have any contribution from the Andreev reflection
. The conductance can now be related to Pc by:
dI (V, T ; Pc , Z) dIunpol (V, T ; Z) dIpol (V, T ; Z)
G(V ) = = (1 − Pc ) + Pc .
dV dV dV
(2.51)
Pc is obtained by fitting Eq. (2.51) to the normalized conductance from
experiments with different Z values (Fig. 2.36). Figure 2.36 shows that the
conductance relevant to the Andreev reflection decreases as Z increases.
2.4.3.4 Curie temperature, TC

General comments In order to ensure that a device made of a HM can
be operated at RT, the determination of TC of a HM is essential. There
are several methods to determine TC of a HM. A commonly used one is to

measure hysteresis loops at different temperatures. An alternative method
is to measure the saturation magnetization as a function of temperature.
Measurements of hysteresis loops have been given in Section 2.4.1. Here,
we shall briefly describe measurements of the saturation magnetization.
Saturation magnetization SQUID is commonly used to measure the

saturation magnetization. It is a very sensitive way to determine the mag-
netization. The data are normalized to the magnetization at zero temper-
ature. A typical set of data is shown in Fig. 2.37. TC is then obtained by
extrapolation.
Fig. 2.37 Saturation magnetization as a function of temperature. TC is obtained by

extrapolation.
2.5 Theoretical methods
The electron density of states (DOS) and magnetic moment of half metal-
lic compounds are the primary interest. In addition, energetics for sta-
ble structures and the effects of dopings on half metallic properties at T
= 0 K are of interest. Theoretical methods based on density functional
theory (DFT) (Hohenberg and Kohn, 1964) in the local density approxi-
mation (LDA) (Kohn and Sham, 1965) or generalized gradient approxima-
tion (GGA) (Langreth and Perdew, 1980; Langreth and Mehl, 1983) for
exchange-correlation are the most often used. We shall give a brief review
of DFT and comment on popular methods of calculating the total energy
of HMs and corresponding electronic band structures.
2.5.1 Density functional theory (DFT)

The basic idea of DFT is that in a many-electron system the ground state is
assumed to be nondegenerate and the charge density ρ(r) is the fundamental
quantity. If ρ is known, other physical properties of the many-electron
system can be determined. There are many reviews of DFT in the physics
and chemistry literatures. We shall not dwell on details of the theory but
rather will just provide an overview of the physical meaning and important
consequences. The theory is based on two theorems by Hohenberg and
Kohn (1964).
2.5.1.1 Hohenberg-Kohn theorem I

There is a one-to-one correspondence between the ground state charge den-
sity ρ(r) of a many-electron system and external potential Vext . As a con-
sequence of this theorem, there is a unique ground state charge density for
any given external potential acting on a many-electron system.
2.5.1.2 Hohenberg-Kohn theorem II

The ground state charge density of a many-electron system minimizes the
total energy of the system. Mathematically, the total energy functional F of
a many-electron system at its ground state composes of the kinetic energy T ,
external potential energy Vext [ρ], and interaction energy of electrons. This
interaction energy can be divided into the Coulomb interaction between
any pair of electrons and exchange-correlation energy Exc . These energies
are functionals of the charge density, which is a function of the electronic
position r. Let the total energy functional of the system be F ,
1 ρ(r)ρ(r′ ) 3 3 ′
Z ZZ Z
F = T [ρ]d3 r + Vext [ρ] + d rd r + Exc [ρ]d3 r, (2.52)
2 |r − r′ |
where Vext [ρ] is expressed in terms of a single particle potential υext (r),
Z
Vext [ρ] = ρ(r)vext (r)d3 r. (2.53)
The free energy functional F of the system is minimized with respect

to ρ(r) with the constraint,
Z
N = ρ(r)d3 r. (2.54)
N is the total number of electrons in the system. By introducing a Lagrange

multiplier λ in Eq. (2.52), an one-particle equation is obtained through the
minimization of F with respect to ρ(r),

ρ(r′ ) 3 ′
Z
t[ρ(r)] + vext (r) + d r + Vext [ρ(r)] − λ = 0, (2.55)
|r − r′ |
where t[ρ(r)] = δT /δρ and vxc [ρ(r)] is δExc [ρ(r)]/δρ. Now, a few crucial
questions remain:
• What is the form of t[ρ(r)]?

• What is the form of εxc [ρ(r)]?
• How does one carry out practical calculations?
2.5.2 Kohn-Sham equations

Kohn and Sham (1965) introduced a scheme leading to the self-consistent
method for calculating ground state electronic properties of a many-body
system. There are several versions from the original formulation. We refer
to the one by Grossu and Paravocicino (2000). The first step is to modify
the total energy functional by specifying the density and add and subtract
a non-interacting kinetic energy functional:
X
ρ(r) = ψi∗ (r)ψi (r), (2.56)
i=occ
where ψi are fictitious, single-particle wave functions of the system,

X −~2 ∇2

T0 [ρ(r)] = ψi
ψi , (2.57)
i 2m
and define a new exchange-correlation energy functional as:
′
Exc [ρ(r)] = Exc [ρ(r)] + T [ρ(r)] − T0 [ρ(r)] (2.58)
The total energy functional is now modified,

1 ρ(r)ρ(r′ ) 3 3 ′
Z ZZ Z
F = T0 [ρ]d3 r + Vext [ρ] + d rd r + ′
Exc [ρ]d3 r (2.59)
2 |r − r′ |
The single-particle Schrödinger eqation is now modified to:
2 2
ρ(r′ ) 3 ′

−~ ∇
Z
′
+ vext (r) + d r + vxc [ρ(r)] − λ ψi (r) = 0. (2.60)
2m |r − r′ |
The self-consistent scheme is to solve Eqs. (2.56) and (2.60) self-
consistently. We are still left with vxc [ρ((r)] unspecified. Kohn and Sham
(1965) proposed the local density approximation (LDA) for the exchange-
correlation energy functional.
2.5.2.1 Local density approximation (LDA)

′
In DFT, the exchange-correlation energy functional Exc [ρ] in Eq. (2.58) is:
Z
′ ′
Exc [ρ] = vxc [ρ(r)]ρ(r)d3 r, (2.61)
with
g(r, r′ )ρ(r′ ) 3 ′
Z
′
vxc [ρ(r)] = d r, (2.62)
|r − r′ |
where g(r, r′ ) is the pair correlation function. It is a non-local function
specifying the probability for a particle at r′ when there is a particle at r.
′
In LDA, vxc is approximated as follows: Let
Z
Exc [ρ] = εxc [ρ(r)]ρ(r)d3 r,
LDA
(2.63)
where εxc [ρ(r)] is the exchange-correlation energy per particle. εxc [ρ(r)] is
′
determined locally by the density at r. Thus vxc [ρ(r)] can be expressed as:
LDA

′ ∂Exc [ρ] ∂εxc [ρ]
vxc [ρ(r)] = = εxc [ρ(r)] + ρ . (2.64)
∂ρ ∂ρ ρ=ρ(r)
Eqs. (2.63) and (2.64) are the essence of LDA. εxc [ρ(r)] is the sum of the
exchange energy per particle, εx [ρ(r)], and correlation energy per particle,
εc [ρ(r)]. In terms of rs —the average spacing between particles—Perdew
and Zunger (1981) fit these energies to Monte Carlo results of Ceperley
and Alder (1980) to obtain
0.4582
εx [ρ(r)] = − , (2.65)
rs
and
− 1+1.05290.1423
√
rs +0.3334rs rs ≥ 1
εc [ρ(r)] =
−0.048 + 0.311 ln(rs ) − 0.0116rs + 0.002rs ln(rs ) rs < 1.
(2.66)
2.5.2.2 Spin-polarized Kohn-Sham equations

Half metals are a class of ferromagnetic materials. To study their magnetic
properties, it is necessary to include the spin degree of freedom. In principle,
we should review DFT in spin-polarized form. The primary difference from
the formalism given in Section 2.5.2 is to generalize the energy density
functional in terms of 2 × 2 matrix form which was first given by von Barth
and Hedin (1972). Readers who are interested in this subject can read the
review by Zeller (2006). Here we only discuss the essentials of spin-polarized
Kohn-Sham equations (von Barth and Hedin, 1972) which are relevant to
half metallic properties. Let α denote the ↑ spin and β denote the ↓ spin
states. We follow the formalism in LDA by defining
X
∗
ραβ (r) = ψα,i (r)ψβ,i (r). (2.67)
i=occ
For diagonal terms,i.e. α = β, the two densities are denoted by either ρα (r)
or ρβ (r) and the Schrödinger equation is now in a coupled form,
X 2
−~ 2 X Z ργγ (r)
3 ′ ′
δµν ∇ + vext (r) + d r + vxc [ρ(r)]µν − λ ψν = 0,
ν 2m γ |r − r′ |
(2.68)
where µ, ν, and γ are either α or β. With the local spin density approxi-
mation (LSDA), the exchange-correlation functional ELSDA xc is
Z
LSDA
Exc [ραβ (r)] = εxc [ρα (r), ρβ (r)] (ρα (r) + ρβ (r)) d3 r (2.69)
LDA
′ ∂Exc [ρ]
vxc [ρ(r)]αβ = . (2.70)
∂ραβ
Note the matrix forms in Eqs. (2.69) and (2.70). Additional information
about the magnetic moment at a point inside a unit cell and the magnetic
moment per unit cell can be obtained,
m(r) = ρα (r) − ρβ (r), (2.71)
and
Z
M= m(r)d3 r. (2.72)
Although simple, the LDA (or LSDA) results in a realistic description

of the atomic structure, elastic, and vibrational properties for a wide range
of systems. However, LDA is generally not accurate enough to describe the
energetics of chemical reactions (heats of reaction and activation energy
barriers), leading to an overestimate of the binding energies of molecules
and solids in particular. To improve this, generalized gradient approxima-
tion (GGA) for the exchange-correlation energy functional is often used.
It has several versions. The commonly used is the one by Perdew et al.
(1992).
2.5.2.3 Generalized gradient approximation (GGA)

Basic idea To improve the calculated total energy of a physical system,
Langreth and Mehl (1983) , Becke (1988), and Perdew and Yue (1986)
developed the GGA exchange-correlation. The essential idea is to generalize
the LSDA exchange-correlation energy functional, ELSDA xc [ραβ (r)] given in
Eq. (2.69) to
Z
Exc [ρα (r), ρβ (r)] = f [ρα (r), ρβ (r), ∇ρα (r), ∇ρβ (r)] d3 r,
GGA
(2.73)
where f is a functional of spin-polarized densities and their derivatives.

There are several forms in the literature. The PBE (Perdew et al., 1996,
GGA
1997) exchange-correlation functional is a popular one which divides Exc
GGA GGA
into the correlation part Ec and exchange part Ex .
Expressions for EcGGA and ExGGA For facilitating the numerical meth-
ods, PBE proposed analytic expressions for both energy functionals satis-
fying energetically significant conditions and having constant parameters.
We start with EcGGA ,
Z
GGA
Ec [ρα (r), ρβ (r)] = ρ(r) {εc [ρα (r), ρβ (r)] + C (rs , ρα−β (r), t)} d3 r,
(2.74)
where εc is the one used in Eq. (2.66) and
ρα−β (r) = (ρα (r) − ρβ (r))/(ρα (r) + ρβ (r)). (2.75)
p
t = |∇ρ(r)| /(2φkT F ρ) with kT F = (4kF )/(πaB ) as the Thomas-Fermi
screening wave number, where kF is the Fermi momentum and aB is the
Bohr radius. Let φ = (1/2) (1 + ρ)2/3 + (1 − ρ)2/3 , then

e2 3 β(1 + At2 )

2
C (rs , ρα−β (r), t) = γφ ln 1 + t , (2.76)
aB γ(1 + At2 + A2 t4 )
where A = (β/γ){exp(−εc /(e2 γφ3 /aB ) − 1}−1 , γ = (1 − ln 2)/π 2 and β =
0.066725. EcGGA is given as
Z
ExGGA [ρα (r), ρβ (r)] = ρ(r) {εx [ρ(r)]Fx (s)} d3 r, (2.77)
η |∇ρ|
where εx [ρ(r)] = −3e2 kF /(4π), Fx (s) = 1 + η − 1+0.21951s2 , s = 2k ρ , and
F
η = 0.804.
With the basic concept of DFT, we now discuss the essence of several
popular methods for calculating the electronic properties of various mate-
rials including HMs based on DFT.
2.5.3 Methods of calculating electronic properties

For the methods to be discussed, the electron exchange-correlation interac-
tions are treated with either within LSDA or GGA. The differences among
these methods are in the treatments of ionic potentials, whether including
core states or not, and the choice of basis functions.
2.5.3.1 Linearized augmented plane wave (LAPW) method

An important feature of this method is the inclusion of all electrons, both
core and valence, in the calculation. Therefore, it is sometimes called an
“all-electron” method. In principle, it is up-to-now the most accurate
method for calculating electronic properties. Because of the inclusion of
all electrons, it is generally restricted to materials with a relatively small
number (< 50) of atoms per unit cell. The APW method on which it is
based was originally proposed by Slater.
Basic idea In order to make the method tractable, Slater (1951) em-
ployed an approximation that the potential has spherical symmetry near
an atom in a solid and is constant in the interstitial region: the so-called
muffin-tin (MT) approximation. A schematic diagram of the MT potential
is given in Fig. 2.38. The lattice is indicated by mesh. At each intersection,
there is a lattice point and an atom. The MT potential around each atom
is shown by a filled circle known as the MT sphere. Its radius is called the
MT radius, Rm . The arrows mark the interstitial region.
Fig. 2.38 Schematic diagram of MT potential. Intersections of lattice lines represent

lattice points of the crystal. The spherical potentials about each atom are shown as filled
circles. The interstitial region is indicated by arrows.
With the MT potential, the wave function inside the MT sphere can be
expanded in spherical harmonics, Ylm (θ, φ), where θ and φ are the stan-
dard angular variables. In the interstitial region the wave function can be
expanded in plane waves eiG·r . An augmented plane wave (APW) is then
defined as
P
l,m Alm Rl (r, E)Ylm (θ, φ) r ≤ Rm
ψAP W (r) = (2.78)
iG·r
e interstitial,
where Rl (r, E) is the solution of the radial equation with angular momen-
tum l—the quantum number—at energy E and is a function of both r and
E. Alm is determined by matching the expansion inside the MT sphere and
plane wave at Rm . The set of all such APW’s is then used as the basis for
the wave functions in the solid.
Secular equation The variational principle is applied to the expansion

of the wave function in terms of APWs. There are two points to note:
• The derivatives of the basis functions at r = Rm are discontinues.

• Because of the energy dependence of the Rl (r, E) in the APW, the
matrix elements of the secular equation are energy dependent.
Therefore, many existing powerful diagonalization algorithms cannot be

applied. A linearized formulation of the APW method was proposed by
Andersen (1975), by expanding the wave function inside the MT on a com-
plete set of states at a fixed energy El , and by Koelling and Arbman (1975),
by making use of wave functions and their energy derivatives at different El .
This method is now called the linearized augmented plane wave (LAPW)
method.
Remarks Since the basis functions of APW’s are used, the matrix ele-
ments of the Coulomb and exchange-correlation potentials can be readily
calculated. Over the years, a number of software implementations of the
method have been developed. One popular and versatile implementation is
the WIEN package1 . For a detailed discussion of the LAPW method, the
reader is referred to the excellent book of Singh and Nordström (2006).
2.5.3.2 Korringa-Kohn-Rostoker (KKR) method

Basic idea The KKR method treats ionic potentials inside a solid as
scattering centers. In a solid having one atom/unit-cell, the lattice vectors
Rn are defined with respect to a chosen origin. Here, n is in fact a set of
1 http://www.wien2k.at
three integers to specify the location of the n-th unit cell. At the origin of
the lattice, a Wigner-Seitz (WS) cell is defined. The MT sphere Sm and the
corresponding MT potentials are inside the WS cell. The Green’s function
is the response at a position r due to a unit source located at a point r′
and is scattered by the MT potentials at other lattice sites.
In order to facilitate the calculations, non-overlapping MT potentials
Σn Vm (r − Rn ) are used, where the summation runs over all lattice points
and Vm (r − Rn ) is the MT potential at Rn . The original KKR formulation
applied to solids is based on the work of Dupree and Beeby (1967). It has
been applied to half metallic compounds by Podloucky et al. (1980) and
Akai et al. (1985) in Germany and Japan, respectively.
Green’s function The Schrödinger equation to be solved is

2
−~ 2 X
∇ + Vm (r − Rn ) ψ(r) = Eψ(r). (2.79)
2m n
The Green’s function associated with Eq. (2.79) satisfies the following
equation:
{H0 + E} G(r + Rn , r′ + Rn′ ; E) = −δ(r − r′ )δnn′ (r), (2.80)
where Ho is (−~2 /2m) ∇2 . By constructing a functional Λ and setting
∂Λ/∂Ψ∗(r) = 0, an integral equation is obtained, where
ψ ∗ (r) { n Vm (r − Rn )} ψ(r)d3 r
R P
Λ= RR
ψ ∗ (r′ ) { n Vm (r − Rn )} G(r, r′ ; E) { n′ Vm (r′ − Rn′ )} ψ(r′ )d3 rd3 r′
P P
+
(2.81)
and
Z nX o
ψ(r) = G(r, r′ ; E) ′
Vm (r′ − Rn′ ) ψ(r′ )d3 r′ . (2.82)
n
′
By changing the variable r and applying the Bloch theorem,
ψk (r + Rn ) = eik·Rn ψk (r), (2.83)
where k is a vector inside the first Brillouin zone (BZ), Green’s function is
now k-dependent and is given by:
X
Gk (r − r′ ; E) = e−ik·Rn Gk (r − r′ − Rn ; E). (2.84)
n
In Eq.(2.84), we assume that space is homogeneous: Green’s function

depends on relative displacements of the source and point of observation.
The boundary condition satisfied by the wave function ψ(r) is obtained by
multiplying Eq. (2.79) by G and Eq. (2.80) by ψ(r), then subtracting the
two resulting equations, and finally setting the surface integral term to be
zero:
∂ψ(r′ ) ′
Z
′ ∂Gk (r − r ; E)
Gk (r − r′ ; E) − ψ(r ) dS = 0. (2.85)
S ∂r′ ∂r′
To apply Eq. (2.85), an expression for G is necessary. Eq. (2.80) can be
easily solved for a single site,
1
G(r − r′ ; E) = . (2.86)
E − H0
By using the eigenfunctions of Ho —the plane waves—the spectral rep-
resentation of Green’s function with summation over BZ can be obtained.
After carrying out the sum, the real space representation of Green’s func-
tion is given as:
′
eiκ|r−r |
G(r − r′ ; E) = − , (2.87)
4π |r − r′ |
√ p
where κ is E if E is positive, and is i |E| if E is negative. Equation (2.87)
can be expanded in spherical harmonics, the so-called Neumann expansion,
′
eiκ|r−r | ′ ′ ∗ ′ ′ ′
P
− 4π|r−r ′| = κ l,m jl (κr)Ylm (θ, φ) {nl (κr ) − ijl (κr )} Ylm (θ , φ ) r < r ,
(2.88)
where (r, θ, φ) is the usual spherical coordinates. jl is the spherical Bessel
function of the first kind and nl is the spherical Bessel function of the second
kind. For r > r′ , one exchanges r and r′ at the right hand side of Eq (2.88).
Structure factor Since the KKR method is based on the idea that the
electron wave function at a point r is a consequence of interferences of
scattered waves from the potential centers, the important quantity is the
so-called structure factor which specifies phase relations of the potential
centers. For one atom per unit cell, Eq. (2.84) provides the basic form of
the structure factor. By substituting Eq. (2.87) into Eq. (2.84), one gets
cos(κ |r − r′ |) X
Gk (r−r′ ; E) = − + Γlm (k, E)jl (κ |r − r′ |)Ylm (θr−r′ , φr−r′ ),
4π |r − r′ | l,m
(2.89)
where θr−r′ is the polar angle and ϕr−r′ is the azimuthal angle of the vector
r − r′ , respectively. Γlm (k,E) is the structure factor and is given by
X δl0 δm0
Γlm (k, E) = κ e−ik·Rn {nl (κRn ) − ijl (κRn )} − iκ √ . (2.90)
n 4π
For the case of many atoms per unit cell, Eq. (2.90) can be easily modi-
fied to accommodate the situation. The most recent improvement including
the coherent potential method is given by Akai and Dederichs (1993).
Eigenvalue problem After obtaining Green’s function, one should be

able to find the eigenvalues by using a trial wave function and Eq. (2.84).
A possible trial wave function is
X
ψ(k, r) = clm (k)Rl (E, r)Ylm (θ, φ), (2.91)
l,m
where Rl is the regular solution inside the MT sphere. Outside the MT

sphere, the potential is expressed in terms of phase shifts which are func-
tions of energy E. From Eq. (2.84), a set of linear coupled equations for
clm (k) can be set up. The corresponding secular equation gives eigenvalues
at each k-point.
2.5.3.3 Pseudopotential method

This is one of the popular methods for calculating electronic properties in
condensed matter. There are many well-documented algorithms available
either in the market, such as VASP 2 or free, e.g. ABINIT3 , QuantumE-
spresso4 , and CP2K5 . In this section, we only discuss the basic idea of
pseudopotentials and comments of recent developments on this subject.
Basic idea The simplest way to express a valence electron in a mate-

rial is to use a planewave |ki = √1Ω eik·r . The amplitude of a planewave
is a constant and Ω is the unit-cell volume. In reality, the wave function
of a valence state must be orthogonal to core states. Herring (1940) in-
troduced the orthogonalized plane wave, OPW, to take into account this
orthogonalization constrain.
X
|OP W i = |ki − |ci hc| ki , (2.92)
c
where |ci is the core wave function in Bloch form. For treating electronic
states in a real solid with periodicity, the description of a simple plane wave
is not sufficient. A Bloch form of an |OP W i should be more appropriate.
Then, the orthogonal processes become a tremendous burden. Phillips and
Kleinman (1959) reformulated the orthogonalization process into a form of
potential—the pseudopotential. The form of a pseudopotential is
X
Vps (r) = V (r) + (E − Ec ) |ci hc| , (2.93)
c
where V (r) is the crystal potential. E is the eigenvalue for a valence state.
Ec is the energy of a core state. Some features of pseudopotentials are:
2 http://cms.mpi.univie.ac.at/vasp
3 http://www.abinit.org
4 http://www.pwscf.org
5 http://cp2k.berlios.de
• The second term is repulsive because E > Ec . |Vps (r)| is therefore

weaker than |V (r)| in the core region.
• Vps (r) is energy dependent.
• Vps (r) is nonlocal because of the presence of projection operators.
Key developments The semi-nonlocal form–only spherical harmonics

Ylm are used in the projection operators–was used by Lee and Falicov
(1968) to calculate the band structure of potassium, and by Fong and
Cohen (1970) to calculate the band structure of copper. Because of the
feature of the semi-nonlocal operators, not only there is a large number of
off-diagonal matrix elements needed to be stored but also is very time con-
suming to calculate these matrix elements. Kleinman and Bylander (1982)
proposed an approximate scheme of treating nonlocal operators efficiently
to reduce the computational efforts. In the late 1970’s and beginning of
1980’s, implementation of first-principles norm-conserving pseudopotential
in plane-wave based DFT was developed (Hamann et al., 1979). Since the
pseudopotential in the core region is not unique (Austin et al., 1962), and
it is appealing to use plane-wave basis for its flexibility and efficiency, Van-
derbilt (1990) proposed ultrasoft pseudopotentials for elements with tightly
bounded, e.g. O 2p and Ni 3d, orbitals. This development enables one to
treat TMs efficiently with pseudopotentials and plane-wave basis. Recent
developments by Blöchl (1994) and Kresse and Joubert (1999) of the pro-
jector augmented wave (PAW) method have shown the PAW method can
be as accurate as the all-electron approach if implemented in plane-wave
codes. These potentials are now available in widely used VASP and ABINIT
implementations.
2.5.3.4 LDA+U
There are some concerns about electron-correlation effects on electronic
properties of Heusler alloys and oxides. Here we briefly discuss the essence
of electron-correlation treated in these two classes of HMs.
The most important electron-correlation effect due to localized d-
electrons in TM elements, such as Co, Cr, Fe, and Mn involved in Heusler
alloys and oxides, is the Coulomb repulsion experienced by localized elec-
trons. The LDA+U (including LSDA+U) scheme is designed to model lo-
calized states when on-site Coulomb interactions become important. Since
most modern DFT implementations such as VASP and WIEN have in-
cluded the LDA+U scheme, we explain the basic ideas and comment on
cautions one should take when using the scheme.
Basic idea The concept of the U term characterizing the Coulomb inter-
action at an atomic site in a solid was introduced by Hubbard (1963). The
starting point is a many-electron Hamiltonian,
X ~2 X X
e2
H= − ∇2i + Vext (ri ) + , (2.94)
i 2m i j6=i |ri − rj |
where i and j are dummy variables to enumerate the electrons. The

2
Coulomb interaction |rie−rj | can be replaced by screened electron-electron
interaction Ve−e . By using a set of localized functions, say, the Wannier
functions, the Hamiltonian is second quantized. Let the set of Wannier
functions be {w (ri )}, then the matrix element of the single-particle kinetic
~2
energy and the external potential, hw (ri )|− 2m ∇2i +Vext (ri ) |w (ri )i, defines
~2
the on-site energy εi . The matrix element, hw(ri )|− 2m ∇2i |w(rj )i, describes
the hopping integral tij of an electron hopping from site i to site j. The ma-
trix element of the two-particle operator Ve−e (|ri − rj |) gives the strength
of Coulomb repulsion hw(ri )w(rj )| Ve−e (|ri − rj |) |w(ri )w(rj )i. This term
is approximated as U ni↑ nj↓ , where niσ = a†iσ aiσ and a†iσ and aiσ are the
creation and annihilation operators at the i-th site with spin σ. If i and
j are equal, it defines the on-site U term. The so-called Hubbard model
Hamiltonian HH is,
X
εi a†iσ aiσ + †
X X X
HH = ′
t iσ,jσ ′ a iσ a jσ ′ + H.c. + U ni↑ ni↓ ,
i,σ i,σ j6=i,σ i
(2.95)
where H.c. is the Hermitian adjoint of a†iσ ajσ′ .
Eq. (2.95) is a many-body expression. To implement U in DFT formal-
ism, the general procedure is to add the U term to either LDA or GGA
total energy. The total energy of a physical system can then be written as
(Hubbard, 1963):
h i h i h i
n′ l′ m′ nlm n′ l′ m′ nlm n′ l′ m′
E T ot ρσ , ρσ′ ; onlm
iσ , oiσ ′ = E LDA
[ρ σ , ρ σ ′ ]+E
U
oiσ , oiσ ′ −E dc
oiσ , oiσ ′ ,
(2.96)
where E LDA is the total energy of LDA or GGA functional, E U is the
total energy derived from the U term, and E dc is the double counting total
energy to correct the overlap between E LDA and E U energies. onlm iσ is the
nlm,n′ l′ m′
trace of the local density matrix Ôiσ,jσ ′ , where i is the site index, nlm
is the product of the principal, angular momentum, and magnetic quantum
numbers.
Implementation of Hubbard U To implement the last two terms of

Eq. (2.96), it is important to realize that:
• E U should be calculated within the atomic region where atomic
characters of electronic states can be identified. Assuming it is
possible to identify atomic orbitals as |inl; mσi within an atomic
sphere then only |mσi is of interest. Thus the notation of |inl; mσi
can be simplified as |mi.
′′ ′′′
 
′′ ′ ′′′ m,m′ m ,m
 hm, m | V e−e |m , m i Ô σ,σ Ô−σ,−σ

U
h
nlm n′ l′ m′
i 1X 
′′ ′ ′′′ ′′ ′′′ ′

E oiσ , oiσ′ = + [hm, m | Ve−e |m , m i − hm, m | Ve−e |m , m i] .
2 {m},σ   Ôm,m′ Ôm′′ ,m′′′ 

σ,σ σ,σ
(2.97)
m,m′ m,m
If the density matrix is diagonal then Ôσ,σ = Ôσ,σ δm,m′ .
• If there is no orbital polarization, Eq. (2.96) should be reduced
to E LSDA and E dc can be expressed in terms of U and J—the
exchange integral (Liechtenstein et al., 1995).
h i 1 1
n′ l′ m′
E dc onlm
iσ , oiσ ′ = U {o(o − 1)}− J {o↑ (o↑ − 1) + o↓ (o↓ − 1)} ,
2 2
(2.98)
m,m′
where o = o↑ + o↓ and oσ are trace of Ôσ,σ . In practice, an
effective single-particle potential is deduced from above equations
and added to V LSDA (,
m′′ ,m′′′
)
′′ ′ ′′′
′ hm, m | Ve−e |m , m i Ô
Vσm,m = m′′ ,m′′′ −σ,−σ
P
m′′ ,m′′′
+ [hm, m′′ | Ve−e |m′ , m′′′ i − hm, m′′ | Ve−e |m′′′ , m′ i] Ôσ,σ
−U (o − 21 ) + J(oσ − 21 ).
(2.99)
• The matrix elements of Ve−e are expressed in terms of Slater inte-
gerals F k (nl, nl′ ) with 0 ≤ k ≤ 2l.
hm, m′′ | Ve−e |m′ , m′′′ i = k ak (m, m′ , m′′ , m′′′ )F k ,
P
4π Pk
ak (m, m′ , m′′ , m′′′ ) = 2k+1 ′ ′′
q=−k hlm|Ykq |lm i hlm | Ykq |lm i ,
∗ ′′′
(2.100)
where Ykq is a spherical harmonics.
Typically, U value is approximately 20 eV in a free TM element and is
on the order of a few eV in a solid.
2.5.4 Methods of calculating Curie temperature, TC

To calculate TC , there are a few methods based on DFT. The foundation of
these methods is the “magnetic force theorem” proved first by Mackintosh
and Andersen (1980) and Heine (1980). It is then extended by Oswald et al.
(1985) for systems with magnetic impurities. The theorem is frequently
used to compute exchange interaction parameters and adiabatic spin-wave
spectra of ferromagnets. The interest of this approach is the energy differ-
ence of different spin orientations at the atoms can be determined by the
change of the one-particle energy. It allows the calculation of the change
of the total energy to be carried out non-self-consistently. Thus, the theo-
rem provides a way to reduce the computational effort. Liechtenstein et al.
(1987) used the theorem for treating spin wave spectra of ferromagnetic
systems within mean field theory (MFT). Bruno (2003) and his collabora-
tors extended the applications to varieties of ferromagnetic metallic systems
and improved the MFT scheme proposed by Liechtenstein et al. (1987).
2.5.4.1 Determination of the dominant excitation

Pajda et al. (2001) based on the physical ground to determine the dominant
excitation in ferromagnetic materials. In transition metals, such as Fe, Co,
and Ni — the so-called itinerant ferromagnets, there are possibly two kinds
of magnetic excitations: the Stoner excitation and the spin-wave excitation.
The former involves interband transitions from the occupied ↑ spin band to
an unoccupied ↑ spin band. They contribute to the longitudinal fluctuation
of the magnetization. The latter is a collective excitation of spin moments
of atoms. These collective excitations are called spin waves or magnons and
contribute to the transverse fluctuation of the magnetization. The lower
energies and corresponding large DOS of the magnon excitations cause them
to be the dominant excitations at low temperature. Therefore, Pajda et al.
(2001) considered only magnon excitations to determine TC . This type of
excitations is also adopted by others (Bouzerar et al., 2006). For HMs, it
is possible to use the same approach to determine TC because the metallic
channel contributes to the ferromagnetic property.
2.5.4.2 Basic idea

The basic idea is to first map the interested spin system onto an effective
Heisenberg Hamiltonian (Pajda et al., 2001),
X
Hef f = − Jij Si · Sj , (2.101)
i<j
where Jij is the pairwise exchange interaction energy, Si is the spin moment
at site-i. If Jij is obtained, then the magnon energy ~ω(q)), where q is the
momentum of the magnon, can be calculated,

4µB X
J0j 1 − eiq·Roj ,

~ω(q) = (2.102)
m j
where m is the magnetic moment per atom R0j = R0 − Rj is the lattice
vector at the j-lattice point from some chosen reference R0 .
To determine TC within the mean field approximation (MFA) (Bouzerar
et al., 2006), one then lets the energy of the magnetic system expressed in
DX E DX E
E = gµB hi · Si = − hi mi = − hh · mi , (2.103)
i i
where hi = gµ2B j6=i Jij Sj —the mean magnetic field at site i produced
P
by magnetic moments from other sites—g is the so called g factor—the

gyromagnetic ratio. For electron without spin orbit coupling g = 2. The
magnetic moment at site-i mi is −gµB Si . Or, the energy is expressed in
term of an effective mean field h = λm, λ is the mean field parameter, and
P
m = i mi . The thermal average of Sj is expressed as:
h i
−i E
T r Sj e kB T
hSj i = h i , (2.104)
−i E
T r e kB T
where kB is the Boltzmann constant. By solving Eqs. (2.103) and (2.104)
self-consistently, TCMF A is given by
λ
2
kB TCMF A = m , (2.105)
3

2
where m is calculated at T = 0K and λ is assumed to be independent
of temperature.
A more accurate way is to use Green’s function method. Callen (1963)
used temperature dependent Green’s function and decoupled the hierarchy
of the equation of motion by a physical intuitively appealing scheme. The
magnetization of general spin is simply expressed as the z-component of the
averaged spin operator for each site hS z i, and relates to the Bose-Einstein
(B-E) distribution function Φ characterizing spin wave excitations.
(S − Φ)(1 + Φ)2S+1 + (1 + S + Φ)Φ2S+1
hS z i = , (2.106)
(1 + Φ)2S+1 − Φ2S+1
where
1 X 1
Φ= ~ω(q)
, (2.107)
N q
e kB T − 1
where N is the number of q points. Bouzerar et al. (2006) used the ran-
dom phase approximation (RPA) to decouple Green’s function and derived
similar expression for the magnetization. By expanding the Bose-Einstein

distribution, TC is given by:
1 6µB 1 X 1
RP A
= lim . (2.108)
kB TC M z→0 N q z − ~ω(q)
1
The term N1 q z−~ω(q)
P
is in fact the magnon Green’s function.
2.5.4.3 Comments on practical calculations

In practice, calculations of TC involve several demanding steps. The first
is the determination of Jij , then, is the summation over q, Eq. (2.108).
Determination of Jij There are two ways to determine Jij and the
excitation energy. One is the real space scheme and the other is the q-
space scheme or the scheme of frozen magnon.
Real-space approach In Eq. (2.101), for each i one calculates Jij by consid-
ering the j-th shell of atoms. First-principles methods can be used
to determine Jij by calculating the change in energy when spin
moments at the i-th and j-th sites are rotated by equal and oppo-
site amounts with respect to the ground state (all spins aligned),
and subtracting the changes in energy for the isolated rotations at
the two sites, respectively. The tight-binding linear combination
of muffin-tin orbital (TB-LMTO) method has been used by Pajda
et al. (2001) with j up to 172-th shell for fcc structure and 195-th
shell for bcc lattice. In order to get converged ω(q), about 200
terms of Jij are needed.
Momentum-space approach In this approach, one specifies the spin mo-
ments on the neighboring atoms to have a spiral configuration and
carries out first-principles calculations to get the ~ω as a function of
q relations for the configuration. Where Jij is obtained by the in-
verse Fourier transform–an important contribution by Sandratskii
(1998) so large supercell is not necessary. The main relevant physics
is illustrated in Fig. 2.39 using a one-dimensional model.
In Fig. 2.39, each arrow makes a polar angle θ with the z-axis and
an azimuth angle ϕ with respect to the x-axis. The lattice vector
is Rn where n represents the n-th lattice site. The moment at the
n-th lattice site for a spiral structure with momentum q is given
by
Mn = |Mn | {cos(q · Rn + ϕn ) sin θn x̂ + sin(q · Rn + ϕn ) sin θn ŷ + cos θn ẑ} ,
(2.109)
Fig. 2.39 One dimensional model showing the relative phase of the spin spiral structure
(Sandratskii, 1998).
where |Mn | is the magnitude of the spin moment at the n-th lat-
tice site. It is easy to see that when there is a lattice translation
Rn , correspondingly there is a spin moment rotation by an an-
gle of −q · Rn which leaves the invariance of the spiral structure.
By analogy to notations used in the space group, one can define
({q · Rn |e| Rn }) as a generalized translation of the spiral structure,
where e indicates that it does not have any real space coordinate
transformation, such as a rotation or a reflection and commutes
with the Hamiltonian of the spiral structure. Because of e, it has
no effect on the periodic part of the Bloch wave function—just a
unit operator. The effect of generalized translation on a spinor
wave function with electron momentum k is
Pk ({q · Rn |e| Rn }) ψk = U (αS ) e−ik·Rn ψk , (2.110)
where Pk is a unitary operator associated with the generalized
translation operator. The active point of view is adopted here
for the coordinate transformation on the wave!function.
i
e− 2 q·Rn 0
U (αS ) = i , (2.111)
0 e 2 q·Rn
Eq. (2.110) is the generalized Bloch theorem. For each k-point, the
eigenvalue εn (k) of the nonmagnetic case is changed to εn k − σq 2
, where σ is “+” or “-”, the spin projection along the z-axis. E(q)
can be obtained by summing all occupied state energies and sub-
tracting the energy of the reference state. From E(q) one can get
J(q), then Jij is obtained by inverse Fourier transform.
Comparing the two schemes, the k-space scheme is simpler owing
to this generalized Bloch theorem. However, the calculations are
still demanding in order to obtain accurate ω(q).
Summation of q To calculate TC , there is still the q-space sum to be

carried out independent of MFA and RPA. It has to be done with extreme
care about the q = 0 point. Large number of q-points should be used in
order to have converged result.
Chapter 3
Huesler alloys
3.1 Introduction
Heusler alloys are ternary intermetallic compounds. It was Heusler who

first synthesized Cu2 MnAl in 1903 (Heusler, 1903). Excellent reviews have
been given by Westerholt et al. (2005) and Hirohata et al. (2006). In 1971,
Webster (1971) reported the chemical and magnetic structure of Co2 MnX
alloys, with X = Si, Ge, and Sn. In 1983, de Groot et al. (1983) predicted
that NiMnSb, now classified as a “half”-Heusler (HH) alloy, should exhibit
half metallic properties. Since then, more than 1,000 Heusler alloys have
been synthesized. A typical chemical formula, X2 YZ, is adopted for the
class of “full”-Heusler (FH) alloys that are predicted to have half metallic
properties, where X is a transition metal (TM) element, e.g., Co, Fe, Ni,
or Pt; Y is another TM element, such as Cr, Mn, or Ti; and Z is an atom
belonging to Group III, IV, or V, e.g., Al, Ge, Si, or Sb.
The recent development of spintronics has generated much interest in
Heusler alloys. For spintronic applications, two criteria are essential:
(1) Effective injection of spin-polarized carriers into semiconductors.

(2) Sufficiently high TC for devices to operate at room temperature (RT).
The Heusler alloys are expected to meet both of these criteria because their
structures are compatible with many semiconductors in both elemental and
compound forms. The TC ’s of Heusler alloys are well above RT. However,
only NiMnSb (Hanssen et al., 1990; Hordequin et al., 1996) has been found
to exhibit half metallic properties within 1/100 of an electron accuracy per
formula-unit experimentally. A number of these alloys have been predicted
to be half metallic theoretically (Wurmehl et al., 2005). These are excellent
candidates for spintronic materials.
71
In this chapter, we focus on NiMnSb, Co2 FeSi, and a number of other

Heusler alloys predicted by theory to exhibit half metallic properties. There
are two categories (Galanakis, 2002b) of Heusler alloys: the half-Heusler
alloys, such as NiMnSb and PtMnSb, and the full-Heusler alloys, such as
Co2 CrAl, Co2 FeAl, Co2 FeGe, Co2 MnSi, and Co2 FeSi. The structure of the
first category C1b (Fig. 3.1) differs from the second category L21 in the
occupation of TM sites in the unit cell. The C1b structure corresponds to
the L21 structure with the (X(2)) sublattice unoccupied (Galanakis et al.,
2002a; Mancoff et al., 1999).
Fig. 3.1 L21 crystal structure with chemical formula, X2 YZ. X(1) sites are denoted by
filled circles, X(2) by open circles, Y by open triangles, and Z by open squares. This
structure corresponds to a FH alloy. The outermost cube edge has length a. If those
sites with open circles (X(2)) are unoccupied, the structure corresponds to a HH alloy
and is denoted by C1b (Galanakis et al., 2002a).
In Section 3.2, we list the half metallic HH and FH alloys determined

by experiment or theory. The growth of these Heusler alloys will be dis-
cussed in Section 3.3, including both bulk and thin-film forms. Structural
properties of FH and HH alloys will be discussed in Section 3.4.1. Bulk and
thin-film properties will be described in Section 3.5. Finally, in Section 3.6
we discuss one specific HH alloy, NiMnSb, followed by the much studied
FH alloy, Co2 MnSi, and FH alloy with highest TC , Co2 FeSi.
3.2 Half-Heusler and full-Heusler alloys
A great many Heusler alloys have been studied. However, no half metal
(HM) has been confirmed experimentally except NiMnSb (Hanssen et al.,
1990; Hordequin et al., 1996). They have been mostly predicted by first-
principles calculations. Tables 3.1 and 3.2 lists half-Heusler (HH) and
full-Heusler (FH) alloys, respectively, known to exhibit half metallic prop-
Huesler alloys 73
Table 3.1 The magnetic moment/unit-cell and the measured Curie tempera-
ture (TC ) of half-Heusler alloys.
Compound Magnetic moment TC Reference
(µB /unit-cell) (K)
Exp. Theory Exp.
CoMnSb 4.0 490 Webster and Ziebeck (1988)
2.949 Galanakis (2005)
4.2 478 Otto et al. (1989)
FeMnSb 1.930 Galanakis (2005)
NiMnSb 4.0 Hanssen et al. (1990)
4.01±0.02 Hordequin et al. (1996)
4.02 Ritchie et al. (2003)
4.0 de Groot et al. (1983)
3.6 Clowes et al. (2004)
3.85 730 van Engen et al. (1983)
3.9±0.2 Turban et al. (2002)
4.0 Halilov and Kulatov (1991)
4.0 Block et al. (2004)
4.2 728 Otto et al. (1989)
PdMnSb 3.95 500 van Engen et al. (1983)
PtMnSb 3.97 582 van Engen et al. (1983)
4.14 572 Webster and Ziebeck (1988)
3.96 572 Otto et al. (1989)
PtMnSn 3.42 330 van Engen et al. (1983)
3.5 330 Otto et al. (1989)
erties experimentally or predicted theoretically to have integer magnetic

moment/unit-cell and the density of states (DOS) showing a metallic be-
haviour in one spin channel and a insulating behavior in the other. We also
give the available measured and predicted magnetic moment/unit-cell, M,
for each compound. M should, in principle, be an integer for a half metallic
compound, especially determined by theory.
Table 3.3 lists a few other full Heusler alloys besides Co-related com-
pounds predicted to show half metallic properties.
Table 3.2 The magnetic moment/unit-cell and the measured Curie temperature (TC ) of ful-
l-Heusler alloys.
Compound Magnetic moment TC Reference
(µB /unit-cell) (K)
Exp. Theory Exp.
Co2 CrAl 4.811 Antonov et al. (2005)
0.53–0.86 310–330 Hirohata et al. (2005)
3.0 Block et al. (2004)
Co2 FeAl 5.29 (4K) Elmers et al. (2004)
4.9 Kelekar and Clemens (2004)
4.6–4.8 Hirohata et al. (2005)
Co2 FeSi 5.492 (RT),5.907(10.2K) >980 Niculescu et al. (1977)
5.492 (RT),5.97(5.0K) 6.000 1100 Wurmehl et al. (2005)
5.91 Hashimoto et al. (2005)
Co2 MnGe 5.11 Ambrose et al. (2000)
4.84 905 Brown et al. (2000)
5.00 Fuji et al. (1990)
4.93 Miyamoto et al. (2004)
5.0 Picozzi et al. (2002)
5.11±0.05 905±3 Webster (1971)
Co2 MnSi 4.96 985 Brown et al. (2000)
5.0 Fuji et al. (1990)
4.7(375◦ C) Kämmerer et al. (2004)
5.0 Kandpal et al. (2006)
4.78 Ritchie et al. (2003)
4.95±0.25 Singh et al. (2004a)
5.1 Singh et al. (2006)
5.0 Wang et al. (2005a)
5.07±0.05 985±5 Webster (1971)
5.07 Westerholt et al. (2005)
Co2 MnSn 4.78 829 Brown et al. (2000)
5.03 Fuji et al. (1990)
5.08±0.05 829±4 Webster (1971)
5.10 Zhang et al. (2005)
Co2 TiAl 0.70 148±2 Souza et al. (1987)
0.54–0.64 Ishida et al. (1982)
Co2 TiGa 0.82 128 Furutani et al. (2009)
0.80 130±2 Souza et al. (1987)
Co2 TiGe 1.94 2.0 380(5) Barth et al. (2010)
1.78 386(4) Carbonari et al. (1993)
Co2 TiSi 1.96 2.0 380(5) Barth et al. (2010)
1.10 375(4) Carbonari et al. (1993)
Co2 VAl 1.84 310(4) Carbonari et al. (1993)
Co2 VSn 1.20 105 Carbonari et al. (1993)
Huesler alloys 75
Table 3.3 Theoretically predicted half metallic other than Co-related

Heusler alloys.
Compound Magnetic moment (µB /unit-cell) Reference
Fe2 MnAl 1.98 Galanakis (2005)
Fe2 MnSi 2.935 Wu et al. (2005)
Ni2 MnSb 3.70 Rusz et al. (2006)
3.882 Wu et al. (2005)
3.3 Methods of growing Heusler alloys
The methods of growing Heusler alloys can be generally classified into two
categories: the growth of bulk samples and the growth of thin-film sam-
ples. Owing to the anticipation that spintronic device applications require
samples to be in thin-film forms especially, the MR properties are going to
be utilized, why is it interested in growing bulk alloys if most of the appli-
cations require samples having thin-film forms? One answer lies in the fact
that several thin-film growth methods, such as the radio frequency mag-
netron sputtering method and pulsed laser deposition method, use bulk
crystals or polycrystals (Giapintzakis et al., 2002; Caminat et al., 2004;
Shen et al., 2004) as sources. They add more incentives to grow bulk sam-
ples even polycrystals. Therefore, we decide to discuss methods of growing
both bulk samples and samples in thin-film forms.
3.3.1 Bulk Heusler alloys

There are two popular methods to grow Heusler alloys, namely:
• The arc-melting method for growing polycrystalline ingots.

• The tri-arc Czochralski method.
3.3.1.1 Arc-melting method

A typical arc-melting growth scheme (Wurmehl et al., 2005) starts with a
proper or a stoichiometric mixing of pure (> 99.99%) constituent elements.
The mixture was heated under argon atmosphere in a copper hearth cooled
by water. Turning and rotating the hearth can achieve the homogeneity of
the ingots. After that, the processes depend on the usages of samples, or
they are pulverized and sieved (Raphael et al., 2002).
Co2 FeSi samples have been grown by Wurmehl et al. (2005) using this
method under an argon atmosphere at 10−4 mbar. To avoid oxygen con-
tamination, these authors evaporated Ti inside the vacuum chamber before
melting the constituent materials and carrying out additional purification

of the process gas. The polycrystalline ingots were annealed in a sealed
near-vacuum quartz tube at 1300 K for 21 days. The samples were of the
L21 structure. The compositions were checked by X-ray photoemission in-
cluding electron spectroscopy of chemical analysis (ESCA) to insure the
quality of crystals after the Ar+ ion beam bombardment to remove na-
tive oxides at the surfaces. Additional checking of the structure of samples
was carried out by X-ray diffraction using CuKα or MoKα spectra. They
also carried out X-ray absorption fine structure (EXAFS) measurements to
probe the short range order of the structures.
3.3.1.2 Tri-arc Czochralski method

The tri-arc Czochralski method (Raphael et al., 2002) is, in a sense, an
extension of the arc melting method. For single crystal growth, the ingots
are melted again by three directed arcs on a water cooled rotating copper
hearth. This three-arc scheme is the essence of the tri-arc Czochralski
method. Another way is to use seed crystallites. The small seed crystals
can be grown by the tri-arc method using some starting seed. For the
growths of Co2 MnSi and Co2 MnGe, polycrystalline Fe has been used as
starting seeds. The boule obtained this way consists of a number of grains.
The crystals grown from these multigrained boules can have perfect lattice
match to the desired semiconductors, such as GaAs or InAs crystal.
3.3.2 Thin films

Besides the commonly used growth method for HMs discussed in Chapter 2,
there are two other specific methods to grow Heusler alloys in thin-film form,
namely the radio frequency magnetron sputtering method (Schneider et al.,
2007) and the pulsed laser deposition (PLD) method (Shen et al., 2004).
In this section, we first revisit the MBE method in the context of Heusler
alloys, then discuss the radio frequency and PLD methods.
Since there are three choices of methods to grow Heusler alloys in thin-
film form, the first step is to determine which growth method is to be used.
The intended use of the grown sample and properties to be examined are
the main factor in this determination. The next considerations will be:
• The selection of substrates.

• The temperature at the growth, in particular the temperature of the
substrates.
Huesler alloys 77
We shall select a few examples for each of the three methods to illustrate
how the substrates are chosen and the effects of temperature.
3.3.2.1 MBE method

In general, the MBE method is well suited to grow low defect samples and
hybrid structures in thin-film form. Achieving efficient spin injections from
a magnetic sample to a semiconductor has motivated the synthesis of hybrid
structures. In the following, we discuss growths of HH alloys and a hybrid
structure.
Seed or buffer layers The issue of substrate is the first to be addressed.

A proper choice of substrate is critical to the quality of the sample. If there
is a lattice constant matched substrate, the choice can be straightforward.
However, under certain circumstances a large lattice constant mismatch
between the sample and selected substrate may be preferred. An example
was discussed in Chapter 2 in the choice of MgO as a substrate for the
growth of Heusler alloys. The large difference in lattice constants between
the sample and substrate inevitably affects the quality of the sample. A
seed or buffer layer can minimize the effect as demonstrated by Turban
et al. (2002). They grew HH alloys using a (001) V seed layer of the order
of 0.5 − 5.0 nm thick on (001) MgO at 400 K.
Hybrid structure with matching lattice constants Hashimoto et al.

(2005) realized that the lattice constant of Co2 FeSi (5.658 Å) differs from
that of GaAs (5.653 Å) by just 0.08%. This could lead to a better inter-
face between the two crystals so that the efficiency of spin injection from
magnetic material to semiconductor may be improved. To test that, they
chose the MBE method to grow a single crystal of Co2 FeSi on GaAs (001)
as a hybrid structure.
They first prepared 100 nm-thick GaAs templates in the III-V growth
chamber. As-terminated c(4x4) reconstructed GaAs(001) surfaces were
grown by cooling the sample to 420 ◦ C under As4 pressure, so that for-
mation of defects at the top of the surface could be prevented. The GaAs
sample was then transferred to an ultra-high-vacuum (UHV) chamber with-
out any As atoms by a degas process at 580 ◦ C. The chamber was kept at
pressure 6.6x10−10 mbar. In growing Co2 FeSi, they started by growing lay-
ered binary alloy Co0.66 Fe0.34 having a body-center cubic (bcc) structure.
By comparing the lattice constant of the alloy and taking into account the
tetragonal distortion of layers, the composition was determined. They then
added Si to obtain Co2 FeSi with fluxes of Fe and Co kept at the optimized
values. By varying the growth temperature between 100 ◦ C and 400 ◦ C,
they searched the optimal growth condition. This range of temperature
is considered low. Therefore, these authors adopted a low growth rate at
0.1nm/min in order not to degrade the quality of the crystal. These con-
ditions are essential to the growth of high quality crystal and interface.
The temperature of the Si cell (TSi ) varies between 1280 ◦ C and 1335 ◦ C
for the best stoichiometry. Characterization processes and magnetization
measurements confirm the quality of the single crystal and interface. Such
hybrid structures have been discussed in some detail by Hirohata et al.
(2006).
3.3.2.2 Radio frequency magnetron sputtering method

This method was applied in 1997 to grow thin-film forms of NiMnSb
by Caballero et al. (1997). More recently, it has been applied to grow
Co2 FeSi by Schneider et al. (2006) on MgO(100) and Al2 O3 (112̄0) sub-
strates, Co2 MnSi on GaAs(001) (Kohn et al., 2007) and to grow Co2 FeSi
on SiO2 and MgO(001) substrates (Inomata et al., 2006).
Typical setup A typical setup is shown in Fig. 3.2. The chamber is

denoted by the rectangular box and is under Ar pressure. The valve at the
left corner controls the flow of the Ar gas. The pressure gauge is installed at
the right corner. The base pressure in the deposition chamber is maintained
under 10−8 mbar. The pressure is maintained by the vacuum pump, located
at the bottom center of the chamber.
The growth region is composed of a substrate holder attached to the
chamber. The substrate is cleaned by ultrasound with chemicals such as
alcohol. It is usually heated to 500 ◦ C or 800 ◦ C then cooled to the growth
temperature.
Two guns are placed near the middle. The targets, which are inside the
guns, are cut from polycrystalline ingots obtained typically by the methods
described in Chapter 2. They are controlled by the radio frequency power
supply to sputter atoms onto the substrate. The typical radio frequency
power is 15–100 W.
Substrate The choice of a substrate is largely empirical. The success-

ful ones for NiMnSb are Corning Glass (Caballero et al., 1997) and for
Co2 FeSi are MgO(100) and Al2 O3 (112̄0) surfaces (Schneider et al., 2006).
The matching of lattice constants between the sample and substrate is
Huesler alloys 79
Fig. 3.2 Schematic setup of radio frequency sputtering method (Courtesy of Randy
Dumas).
not a major concern. For example, the lattice constant mismatch between
Co2 FeSi and MgO is 5.6%.
Temperature The temperatures of substrates in the growth of NiMnSb

are between 200 and 500 ◦ C. The two substrates for the growth of full
Heusler alloy are maintained at 700 ◦ C to obtain the best quality films
with thickness between 60 and 80 nm. In this case, any film grown under
lower temperature conditions exhibits poorer quality—i.e. the appearance
of disorder.
Schneider et al. (2006) obtained the best results of thin-film form of
Co2 FeSi by using stoichiometric composition targets in UHV with a base
pressure below 2x10−7 mbar and kept initially the base pressure of the
chamber at 10−8 mbar. Then the actual growth condition is under 10−2
mbar of Ar pressure and at 700 ◦ C for the substrate with the deposition
rate of 5Å/s. Finally, they covered the top of the grown samples with 4
nm of Al at 350 ◦ C to prevent oxidation. The structure of the films was
determined with X-ray diffraction of the Cu-Kα line and was analyzed with
X-ray four-circle diffractometer. For the case of MgO as the substrate, they
observed the (200) and (400) reflections without any impurity phases. The
rocking curve has a width of 0.3◦ , indicating a good out-of-plane growth.
When Al2 O3 was used as a substrate, the rocking curve shows only 0.1◦
width.
3.3.2.3 Pulsed laser deposition (PLD)

This method is well suited to grow quality metallic thin-layered and mul-
tilayered structures (Shen et al., 2004). It has recently been used by Giap-
intzakis et al. (2002) to grow the HH alloy, NiMnSb, and by Wang et al.
(2005a) to grow the FH alloy, Co2 MnSi. The targets for growing NiMnSb
were pressed and sintered polycrystalline pellets. Another possibility is to
obtain the pellets by radio frequency melting in Ar gas environment. For
Co2 MnSi, stoichiometric polycrystalline pellets were used.
The KrF excimer laser with the wavelength at 248 nm and pulse width
of 34 ns is used as the laser source for both growths. The laser beams are
incident on the rotating target at 45◦ . The energy of pulse is approximately
300 mJ and can have a maximum of 600 mJ. The repetition rate is 10 Hz.
Schematic diagram of setup The setup of the PLD method is shown

schematically in Fig. 3.3. The chamber is shown as a large circle. The
pressure in the chamber is maintained at around 10−9 mbar. The laser
beam is provided by an excimer laser and is focused on the target by a
lens. The target is composed of a bulk form of the sample. The evaporated
atoms are deposited on the substrate. The sample holder is indicated by
the dashed line. The thin film of the desired compound is shown as a stripe.
Wang et al. (2005a) alternatively used a multi-chamber system combined
with MBE under ultra high vacuum (UHV ∼ 10−13 mbar) conditions.
Fig. 3.3 The schematic diagram of PLD setup.
Substrate The substrate for the growth of NiMnSb using PLD is poly-
crystalline InAs or Si(111) surface. For the growth of Co2 MnSi films,
GaAs(001) is used as an effective substrate. The substrates are cleaned.
One process for cleaning the GaAs(001) substrate is as follows: The sub-
strate is initially obtained from a commercial source. The first step is then
to degas in UHV at temperature up to 580 ◦ C. After that, the substrate is
sputtered by a 0.6 keV Ar+ ion beam with current density 4 µÅ/cm2 at 600
◦
C for 30 min, at an incident angle of 45◦ . The quality of the GaAs surface
is checked by AES and LEED to ensure absence of surface impurities and
Huesler alloys 81
desired reconstruction.
Temperature The temperature of the substrate for growth of NiMnSb

on a semiconductor is an issue. In the MBE and radio frequency magnetron
sputtering growth, better samples could be obtained with the temperature
of the substrate between 350 and 400 ◦ C. However, these materials are not
suitable for making multilayer devices. The reason is that interdiffusion of
atoms causes granular formations in the sample. A better growth temper-
ature is approximately 200 ◦ C. Both Giapintzakis et al. (2002) and Wang
et al. (2005a) chose this temperature to successfully grow NiMnSb and
Co2 MnSi. In addition, with the choice of this lower growth temperature
they avoided the post-annealing process.
3.4 Characterization of Heusler alloys
3.4.1 Bulk Heusler alloys

The crystal structure of bulk samples is determined by the Laue method.
Practical rotation method is used for powder samples. Since these methods
are described in elementary solid state physics text books, such as “In-
troduction to Solid State Physics” by Kittel (2004), we shall not provide
further discussions.
At low temperature, the structure of a FH alloy is called L21 . The
associated space group is F m3m. The primitive cell of the L21 structure
consists of four inter-penetrating fcc cubes. An L21 crystal structure is
shown in Fig. 3.1.
Table 3.4 The origins of the

interpenetrating fcc cubes of
the L21 structure. a is the out-
ermost cube edge
Element Origin (a)
X(1) (0.0, 0.0, 0.0)
X(2) (1/2, 1/2, 1/2)
Y (1/4, 1/4, 1/4)
Z (3/4, 3/4, 3/4)
Another related structure is the B2 structure. In which, X(2) atoms are

the same as X(1) atoms. Table 3.5 lists lattice constants determined either
by experiments or calculations of some half metallic Heusler alloys.
Table 3.5 Lattice constants of three half-Heusler alloys and Co-based

full-Heusler alloys having half metallic properties.
Alloy Lattice constant (Å) Reference
NiMnSb 5.904 Van Roy et al. (2000)
PtMnSb 6.210 Matsubara et al. (1999)
PdMnSb 6.260 Matsubara et al. (1999)
Co2 CrAl 5.735 Yoshimura et al. (1985)
Co2 FeAl 5.730 Buschow and van Engen (1981)
Co2 FeSi 5.640 Wurmehl et al. (2005)
Co2 MnAl 5.756 Webster (1971)
5.749 Buschow and van Engen (1981)
Co2 MnGa 5.770 Webster (1971)
Co2 MnGe 5.743 Webster (1971)
5.75 Cheng et al. (2001)
Co2 MnSi 5.654 Webster (1971)
5.66 Cheng et al. (2001)
Co2 MnSn 6.000 Webster (1971)
Co2 TiAl 5.85 Ziebeck and Webster (1974)
Co2 TiGa 5.85 Ziebeck and Webster (1974)
Co2 TiSi 5.849 Barth et al. (2010)
Co2 TiSn 6.07 Ziebeck and Webster (1974)
Co2 VAl 5.772 Buschow et al. (1983)
Co2 VGa 5.779 Buschow et al. (1983)
3.4.2 Thin films

We use Co2 MnSi as an example. The AES and low energy electron diffrac-
tion (LEED) methods are the most widely used to characterize the samples.
3.4.2.1 Auger electron spectroscopy (AES)

As discussed in Chapter 2, the AES can probe the degree of intermixing of
atoms in the substrate and film sample. One examines the evolutions of the
1070 eV Ga line, 1228 eV As line, and 780 eV Co line. The intermixing can
be investigated by fitting the intensities of Ga and As lines to the following
relation (Wang et al., 2005a):

t
I Ga,As = I0Ga,As exp − , (3.1)
cos φ · λGa,As
where I Ga,As is the intensity of either Ga or As in the presence of Co2 MnSi,
and IoGa,As is the intensity of either element without Co2 MnSi. t is the
thickness of Co2 MnSi. φ is the mean opening angle of the spherical mirror
analyzer. λGa is 15 Å for the 1070 eV Ga Auger electrons and λAs is 17
Å for the 1228 eV As electrons. Data show no intermixing for Ga atoms,
Huesler alloys 83
whereas there is a significant segregation of As to the top of Co2 MnSi. The

segregations were gradually buried into the films when the thickness of the
film increases.
3.4.2.2 Low energy electron diffraction (LEED)

The LEED patterns of thin-film samples were compared to those of GaAs.
The results confirm good quality of the epitaxial growth. However, it is
important to note that the patterns are impossible to distinguish the or-
dered (1x1) structure from the disordered ones where same atomic species
occupy several sites.
In Chapter 2, we discussed characterizations of layered structures using
RHEED and STM. They have been applied to the Heusler alloys as well.
3.5 Physical properties of bulk Heusler alloys
The Heusler alloys are magnetic materials and many of them were pre-
dicted to show half metallic properties—they are the consequence of d-d
or d-p interactions. One property that is of central importance is the spin
polarization, P, at EF in the metallic channel. We therefore will focus on
magnetic moments of Heusler alloys, insulating gaps, and spin polarizations
of half metallic Heusler alloys. Disorder in a sample should also be con-
sidered because of the complication of atomic arrangements in a unit cell.
We divide the discussions of physical properties of Heusler alloys into three
main subjects:
• Systematics of magnetic moments in half metallic Heusler alloys.

• Individual Heusler alloys.
• Spin polarization at EF and other magnetic properties.
3.5.1 Magnetic moments and the Slater-Pauling rule

The magnetic moment is of central importance in spintronics. Tables 3.1
and 3.2 list the magnetic moments of HM Heusler alloys. The question
arises: Is there a systematic way to estimate these magnetic moments?
Most Heusler alloys contain 3d TM elements. Slater and Pauling devel-
oped an empirical rule correlating the magnetic moment per atom m of
3d elements and their alloys to the average number of valence electrons
per atom, now known as the Slater-Pauling rule. We shall discuss how a
Slater-Pauling type rule applies to magnetic moments in Heusler alloys.

Let the average number of the valence electrons per atom be nv . The
rule divides m (magnetic moment per atom) into two regions with Fe (nv =
8) on the border. In addition, since it is often favorable for the majority d-
states to be fully occupied (nd↑ = 5), consistent with Hund’s rule, one can
define the magnetic valence nM = 2nd↑ − nv . Malozemoff et al. (1984) and
Kübler (1984) used nM rather than nv in their analyses of ferromagnetism in
metallic systems. A plot of m vs. nM is called the generalized Slater-Pauling
curve. Figure 3.4 shows a plot of m vs. nM (nv ) for a series of Heusler
alloys and associated TM elements obtained by Wurmehl et al. (2005).
For nv ≤ 8, m increases with nv , consistent with m being contributed by
local moments. For nv > 8, m decreases with nv , consistent with m being
contributed by mobile carriers. This is the region of itinerant magnetism.
Fig. 3.4 The Slater-Pauling curve for 3d TM elements and associated Heusler alloys,
filled circles are for Co2 -based alloys and filled grey circles are for other Heusler com-
pounds(Wurmehl et al., 2005). nv (nM ) is the average number of (magnetic) valence
electrons per atom.
If one defines n↑ (n↓ ) as the number of majority-spin (minority-spin)

valence electrons per atom, then the magnetic moment per atom (in µB ) is
m = n↑ − n↓ = (nv − n↓ ) − n↓ = nv − 2n↓ . (3.2)
Fe and its binary alloys typically have approximately three valence elec-
trons per atom in the minority channel (n↓ ≈ 3). Due to the gap in the
minority channel, the half metallic Heusler alloys have exactly three elec-
trons per atom in the minority channel (n↓ = 3). Hence, from m = nv −2n↓ ,
we have
mHM = nv − 6. (3.3)
For ordered HH alloys, there are three atoms/formula-unit and the
moment/formula-unit predicted by the above rule becomes
MHH = Nv − 18, (3.4)
Huesler alloys 85
where Nv is the total number of valence electrons per formula unit. Ex-
tending this argument to the FH alloys, there are 4 atoms/formula-unit for
a total 12 occupied bands in the ↓ spin channel, so that
MF H = Nv − 24. (3.5)
For example, half metallic Co2 FeSi has Nv = 30. The above rule then gives
MF H = 6µB , consistent with theory (Table 3.2).
One should bear in mind that the above rules (Eqs.(3.4) and (3.5))
determine only the magnetic moment/formula-unit for a known HM sample.
They do not predict whether a given sample is HM or not. Rather, the
values of n↓ are determined from a band structure for each alloy as suggested
by Kübler (1984).
3.5.2 Insulating gap in half metallic Heusler alloys

In Heusler alloys, all the evidence shows that the majority (↑) spin states
exhibit metallic properties while the minority (↓) spin channel shows insu-
lating behavior. The insulating gap is one of the essential characteristics to
have half metallic properties. As we discussed in Chapter 1, the gap of the
insulating minority spin channel in HH alloys is generally formed between
the bonding p-t2g states and antibonding e∗g states. However, as we shall
see, this picture is not completely consistent with conclusions from some
band structure calculations which can depend on the electronic structure
methods and atomic arrangements in the unit cell.
3.5.2.1 Half-Heusler alloys

In the mid 1980’s, Heusler alloys have already attracted much attention.
van der Heidet et al. (1985) analyzed the ellipsometric near infrared data
and determined the insulating gaps of NiMnSb and PtMnSb as 0.7 and 0.9
eV, respectively. Later, infrared and optical absorption measurements were
carried out by Kirillova et al. (1995) to determine the insulating gaps of
NiMnSb, PdMnSb, and PtMnSb. They only reported a value of 0.4 eV for
NiMnSb. The difficulty in determining such a gap is that there is a Drude
contribution in metallic majority-spin states.
There are many calculations of NiMnSb and other HH alloys. Calcula-
tions of bulk properties have been carried out by a number of groups using
different theoretical approaches. Aside from de Groot et al. (1983) who
pioneered half metallic calculations on NiMnSb, Galanakis et al. (2002a)]
used KKR method. The LMTO method was used by Kulatov and Mazin
(1990), Youn and Min (1995), and Halilov and Kulatov (1991). The spin-
orbit effect has been considered by Youn and Min (1995). Antonov et al.
(1997) extended the results of de Groot et al. (1983) to study the magneto-
optic Kerr effect in NiMnSb, PdMnSb, PtMnSb, and other ferromagnetic
ternary compounds. These results are listed in Table 3.3. Wang et al.
(1994) carried out calculations on NiMnSb and PtMnSb with the atomic-
sphere-approximation tight-binding LMTO method.
Most of these calculations of HH alloys show a general feature: the gap
is indirect between the occupied states at the Γ point and unoccupied state
at the X point. The valence states are predominantly the hybridizing p-
states of the Sb and d-states of the Ni (Pd, Pt) atom. The conduction state
at X is derived from antibonding d-states of the Mn atom. Therefore, p-d
hybridization plays a crucial role in determining half metallic properties
of HH alloys. The spin-orbit interaction does not diminish half metallic
properties in these alloys except for PtMnSb.
Fig. 3.5 PDOS of ↓-spin states at Γ-point of NiMnSb (Galanakis et al., 2002a).
Galanakis et al. (2002a) investigated electronic properties of a series of

HH alloys using KKR method. They did not discuss band structures for
both spin states for NiMnSb because they found them to be similar to those
presented by de Groot et al. (1983). However, they noted significantly qual-
itative differences between the two sets of calculations, in particular, the
roles played by the Ni and Mn atoms. The most critical issue in the calcula-
tions of de Groot et al. (1983) is that there is no Ni-Mn d-d interaction even
though Ni and Mn are nearest neighbors (nn). Furthermore, because Mn
and Sb are second nearest neighbors, de Groot et al. (1983) concluded that
it is unlikely that d- and p-states associated with these two atoms should
hybridize. In contrast, calculations performed by Galanakis et al. (2002a)
concluded that Mn d-states do mix with Ni d-states, forming states near
Huesler alloys 87
Table 3.6 Summary of Hubbard U values

Element U (eV) Reference
Co 2.5–5.0 (GGA) Kandpal et al. (2006)
4.8 (GGA) Wurmehl et al. (2005)
1.92 (LDA) Kandpal et al. (2006)
Fe 2.4–4.8 (GGA) Kandpal et al. (2006)
4.5 (GGA) Wurmehl et al. (2005)
1.80 (LDA) Kandpal et al. (2006)
EF . The hybridization is purely from Sb p-states and Ni d-states around

3.5 eV below EF because the two atoms are nn. Thus they concluded that
Sb not only stabilizes the C1b structure, it also contributes states control-
ling the half metallic properties. As shown in the PDOS (Fig. 3.5) of the
↓-spin states, the bonding states at the Γ point are contributed from both
Sb p- and Ni d-states.
With regard to the Slater-Pauling rule, Galanakis et al. (2002a) dis-
cussed the 18-electron rule for a binary sample to be a semiconductor: 8
electrons from the Sb and 10 electrons from the TM atom. For a HM,
it should be replaced by a 9-electron rule for the insulating channel: 4
electrons from the Sb and 5 electrons from the TM element.
3.5.2.2 Full-Heusler alloys

There are no experimental values for the band gap of FH alloys. A few
groups have calculated electronic band structures of several FH alloys us-
ing FP-LAPW (Kandpal et al., 2006) and KKR (Wurmehl et al., 2005)
methods.
The M value (5.29 µB ) calculated by FP-LAPW method with GGA
exchange-correlation for Co2 FeSi does not account for the anticipated mag-
netic moment/formula-unit (6.0 µB ) based on the Slater-Pauling rule. and
the position of the gap which can be below EF . Only when the Hubbard
U term—the on-site Coulomb repulsion—is introduced for the electron-
electron correlations, do the results show HM properties. Kandpal et al.
(2006) determined U for Co to be between 2.5 and 5.0 eV and between 2.4
and 4.8 eV for Fe. Wurmehl et al. (2005) found U values of 4.8 and 4.5 eV
for Co and Fe, respectively. A summary of U values for different elements
is listed in Table 3.6.
There is a general agreement about the characteristic of insulating gaps
in FH alloys. The calculated gaps for the insulating channel are also in-
direct, as in NiMnSb, with the top of valence bands at the Γ point and
bottom of conduction bands at the X point. The conduction states are
Table 3.7 The insulating gaps for Heusler alloys.

Alloy Gap (eV) Reference
Exp. Theory
NiMnSb 0.4 Kirillova et al. (1995)
0.7 van der Heidet et al. (1985)
0.6 Galanakis et al. (2000)
0.61 Youn and Min (1995)
0.6 Antonov et al. (1997)
0.44 Kulatov and Mazin (1990)
0.43 Wang et al. (1994)
PdMnSb 0.4 Antonov et al. (1997)
PtMnSb 0.65 Kirillova et al. (1995)
0.9 van der Heidet et al. (1985)
0.8 Galanakis et al. (2002a)
0.91 Youn and Min (1995)
0.2 Antonov et al. (1997)
0.3 Wang et al. (1994)
Co2 MnSi 0.5 Fuji et al. (1990)
Co2 FeSi 0.8 Wurmehl et al. (2005)
1.0 Kulatov and Mazin (1990)
sensitive to the U values and are derived from the antibonding states of
one of TM elements. Whether it is from the Co or Mn atom is not yet
clear. Similarly, the nature of the top of valence states is yet unclear.
In Table 3.7, we summarize reported values of the insulating gap in a
few Heusler alloys. For FH alloys, there are only theoretical values.
The nature of band-edge states of the minority-spin channel in Co2 MnSi
was examined by Galanakis et al. (2002b). They noted a stronger d-d
hybridization between the two Co atoms than between Co and Mn atoms.
Figure 3.6(a) shows the resulting states due to Co-Co interaction. The five
bonding states (lower group of five states in the middle of Fig. 3.6(b)) then
hybridize with the five states of the Mn atom, as shown in Fig. 3.6(b).
They form bonding and antibonding states. The five antibonding states of
the Co-Co interaction (top group of five states at the left of Fig. 3.6(b))
are located between the bonding and antibonding states resulting from the
hybridization of the Co and Mn atoms, and form nonbonding states between
Co and Mn hybridized states. These nonbonding states split into t1u and
Huesler alloys 89
eu type states (states below and above EF in the middle of Fig. 3.6(b)).
EF falls in the gap between the t1u and eu states. Therefore, the gap states
are primarily from d-states of the Co atoms.
Fig. 3.6 Nature of gap states in the minority-spin channel: (a) the resulting states due
to Co-Co interaction, and (b) states around EF after Co-Co states interact with states
from the Mn atom (Galanakis et al., 2002b).
3.5.3 Polarization at EF
The polarization at EF , P, is a crucial quantity if any Heusler alloy is go-
ing to be used for spintronic applications. Table 3.8 lists the measured
and calculated values of P for bulk HH and Co-based FH alloys. A the-
ory of spin-polarized positron-annihilation on NiMnSb was put forward by
Hanssen and Mijnarends (1986).
Experiments were performed later by Hanssen et al. (1990) with the
positron momentum px in three different directions: [100], [110], and [111].
The [100] direction was analyzed at T=27 K and the other two directions
at 8 K. These experiments measured a P value of 100%, confirming half
metallic properties.
The point contact Andreev reflection method was recently applied to
Co2 Crx Fe1−x Si (Karthik et al., 2007). The measured P is 0.64 ± 0.01 at x
= 0.02. As compared to the value (0.57 ± 0.01) at x = 0, the presence of
Cr increases the P value. The authors suggest this increase is due primarily
to the doping effect that improves the L21 ordering structure as concluded
from their XRD and Mössbauer spectra.
Two groups have carried out first-principles calculations of these alloys.
Both groups used the KKR method with muffin-tin potentials. Fuji et al.
(1990) used the local spin density (LSD) approximation of von Barth-Hedin
(von Barth and Hedin, 1972) parametrized by Janak et al. (1975) for treat-
Table 3.8 P values for several Heusler alloys

Alloy P (%) Temperature (K) Reference
Exp. Theory
CoMnSb 99.0 Galanakis et al. (2000)
FeMnSb 99.3 Galanakis et al. (2000)
NiMnSb 44.0 4.2 Clowes et al. (2004)
45.0 4.2 Ritchie et al. (2003)
99.3 Galanakis et al. (2000)
PdMnSb 40.0 Galanakis et al. (2000)
PtMnSb 66.5 Galanakis et al. (2000)
Co2 MnSi 56.0 4.2 Ritchie et al. (2003)
55.0 Singh et al. (2006)
100.0 Fuji et al. (1990)
Co2 FeSi 57.0 4.2 Karthik et al. (2007)
Co2 Cr0.02 Fe0.98 Si 64.0 4.2 Karthik et al. (2007)
ing electron-electron correlation and calculated the electronic structure of

crystalline Co2 MnX, where X is Al, Ga, Si, Ge, or Sn. Galanakis et al.
(2000) performed calculations on Sb-based Heusler alloys. In Table 3.8, we
include the temperature at which P is measured. Most measurements were
carried out using Andreev reflection from the free surface of bulk samples.
In general, the spin-orbit interaction is small in these alloys. There
are large discrepancies between experimental and theoretical results for the
P values (Table 3.8). Many of the experimental authors attributed this
to the presence of nonmagnetic atoms (Clowes et al., 2004; Ritchie et al.,
2003) and the surface interrupting the tetrahedral environment around Mn
and nonmagnetic atoms. de Wijs and de Groot (2001) calculated ideal
surfaces of NiMnSb and showed that they are not HMs due to symmetry
breaking at the surface. However, when the tetrahedral environment formed
between (111) surface of NiMnSb and CdS is restored, the half metallicity
is recovered.
3.5.4 Magnetic moments

Magnetic moments of Ni- and Co-based HH alloys were calculated by
Galanakis et al. (2002b). Table 3.9 lists total magnetic and local moments
for some of pnictides and chalcogenides. These calculated results are based
on experimental lattice constant for each compound. CoVSb was also stud-
ied theoretically by Tobola et al. (1998). They concluded that it is a HM
with a spin moment of 0.965 µB . Its experimental lattice constant is 5.801
Å. For all of the compounds studied in this paper, the interesting feature
Huesler alloys 91
Table 3.9 Calculated magnetic moments of TM elements for

some HH alloys, XYZ. X is either Ni or Co. Y denotes V,
Cr, Mn, or Fe. Z is In, Sn, Sb, or Te. “Void” is for a site
not occupied by any atom. These values were obtained by
Galanakis et al. (2002b) except CoVSb which was calculated
by Tobola et al. (1998).
Compound Ni Mn X Void Total
(µB ) (µB ) (µB ) (µB ) (µB )
NiMnIn 0.192 3.602 -0.094 0.003 3.704
NiMnSn 0.047 3.361 -0.148 -0.004 3.256
NiMnSb 0.264 3.705 -0.060 0.052 3.960
NiMnTe 0.467 3.996 0.101 0.091 4.656
NiCrSb 0.059 2.971 -0.113 0.059 2.976
NiFeSb 0.404 2.985 -0.010 0.030 3.429
NiVSb 0.139 1.769 -0.040 0.073 1.941
CoVSb -0.126 1.074 -0.021 0.038 0.965
CoCrSb -0.324 2.335 -0.077 0.032 1.967
CoMnSb -0.132 3.176 -0.098 0.011 2.956
CoFeSb 0.927 2.919 -0.039 0.012 3.819
Table 3.10 Comparison of computed magnetic moments of NiMnSb and CoMnSb

under compression. “Void” corresponds to the missing atom in the L21 structure.
Compound Lattice constant Ni or Co Mn Sb Void Total
(Å) (µB ) (µB ) (µB ) (µB ) (µB )
NiMnSb–aI[exp] 5.93 0.20 3.85 -0.09 0.04 4.00
NiMnSb–aII 5.68 0.32 3.68 -0.05 0.05 4.00
NiMnSb–aIII 5.62 0.33 3.64 -0.04 0.05 3.97
CoMnSb–aI[exp] 5.87 -0.32 3.41 -0.11 0.02 3.00
CoMnSb–aII 5.22 0.45 2.57 -0.06 0.04 3.00
CoMnSb–aIII 5.17 0.48 2.52 -0.05 0.04 2.99
is the magnetic moment vs. valance of the low-valence TM elements. For

Cr replacing Mn and V replacing Cr, the magnetic moment reduces by 1.0
µB —indicating the existence of local moments. For Fe replacing Mn, due
to the fact that majority d-states are now filled, an extra electron fills the
minority-spin states and thus half metallic properties are lost. Galanakis
et al. (2002b) also remarked on replacing Sb by Te or Sn, which destroys
the half metallicity. For Te, the additional electron puts EF above the gap
of the minority-spin channel while for Sn, EF falls below the gap.
Table 3.10 lists the effect of lattice parameter on magnetic moments of
a few HH alloys calculated by Şaşoǧlu et al. (2005b), where aI[exp] corre-
sponds to the experimental bulk lattice constant. For both NiMnSb and
CoMnSb, EF lies in the lower part of the insulating gap. The use of the lat-
tice constant aII shifts EF to the upper edge of the insulating gap. Further
Table 3.11 The Curie temperature (TC ) for

Heusler alloys.
Alloy TC (K) Reference
NiMNSb 728 Otto et al. (1989)
728 Ritchie et al. (2003)
Co2 MnGa 694 Brown et al. (2000)
Co2 MnGe 905 Brown et al. (2000)
Co2 MnSi 985 Brown et al. (2000)
985 Ritchie et al. (2003)
Co2 MnSn 829 Brown et al. (2000)
Co2 FeGa > 1100 Brown et al. (2000)
Co2 FeSi 1100 Wurmehl et al. (2005)
Co2 TiSn 359 Brown et al. (2000)
370 Souza et al. (1987)
compression of 1%, to aIII , places EF slightly above the gap. The effect of
lattice parameter on the local magnetic moment is determined within the
atomic sphere. For example, for NiMnSb, the lattice contraction increases
the hybridization between Ni and Mn atoms and increases the magnetic
moment of the Ni atom while decreasing the moment of the Mn atom. In
CoMnSb, which has a larger insulating gap, the moments on the Co and
Mn atoms are antiparallel. As a result, the transition of EF to the upper
edge of the insulating gap requires a large lattice contraction of ∼ 11%.
The magnetic moment at each atom is sensitive to the lattice parameter
while the total moment is relatively insensitive.
3.5.5 Curie temperature

The Curie temperature (TC ) of a Heusler alloy is of central interest for
room-temperature spintronic device applications. Most of the measured
TC values have been obtained by inverting the magnetic susceptibility χ.
A summary of TC values for Heusler alloys is given in Table 3.11. It is most
encouraging to note that so many Heusler alloys have TC well above RT.
3.5.6 Other magnetic properties

X-ray magnetic circular dichroism (XMCD) spectra of NiMnSb were stud-
ied experimentally by Kimura et al. (1997) and Yablonskikh et al. (2000,
2001) and theoretically by Galanakis et al. (2000) (Fig. 3.7). The dif-
ference between the two experiments is that absorption spectra (Kimura
et al., 1997) and emission spectra (Yablonskikh et al., 2000, 2001) were
measured, respectively. The measured quantity in X-ray absorption spec-
Huesler alloys 93
tra is the difference of intensities with the photon helicity (spin) parallel
(I+ ) and antiparallel (I− ) to the magnetization. The L2,3 excitations from
Mn 2p core states were measured.
Physically, when a core electron in the 2p level is excited, the absorption
process is governed by selection rules of the transition to 3d states of the Mn
atom. If the sample is half metallic, the final states of the absorption process
are majority-spin states at EF . The insulating minority-spin channel is not
expected to have any states to contribute. As the photon energy increases
to reach conduction states of the minority-spin channel, it is then possible
to obtain a finite intensity I. The difference I+ − I− can therefore reveal
half metallic properties. The XMCD Mn-L2,3 edge spectra of NiMnSb are
shown in Fig. 3.7. The negative part centred around 634.0 eV (L3 line) is
stronger than the positive contribution centred at 645.0 eV (L2 line).
Fig. 3.7 Experimental (Kimura et al., 1997) and theoretical (Galanakis et al., 2000)
absorption and XMCD spectra of Mn L2,3 lines in NiMnSb.
The XMCD absorption spectrum of Mn-based HH alloys can be calcu-

lated similarly to any optical transitions except with the additional contri-
bution from the spin polarization.
2π X X

ε2,± (ω) = j± mj± e± ·p |nki hnk| p·e± j± mj± δ ~ω − Enk + Ej± ,
~ m
j± n,k
(3.6)
where j± mj± is an initial core state having energy Ej± , |nki is a final state
with energy Enk , p is the electron momentum, and e± are polarizations of
photons. The δ-function enforces conservation of energy. The calculated
peak positions agree well with experimental results. The two structures are
at 634 eV and 644.5 eV, respectively, (Galanakis et al., 2000) which agree
well with experimental results. The linewidth of the lower energy peak does
not agree due primarily to the fact that theoretical results were from ideal
crystals.
The emission process involves an electron near EF dropping into a core-
hole state. One measures the emitted X-ray. This process is more com-
plicated than absorption because the presence of a equilibrium hole in the
core region can affect energies of valence states. The corresponding struc-
tures in the measured spectra are at 640.5 eV and 652 eV, respectively
(Yablonskikh et al., 2000, 2001).
Another element-specific, surface-sensitive magnetic technique—the
spin resolved appearance potential spectroscopy (SRAPS)—has been ap-
plied to probe the surface magnetization of the NiMnSb(001) surface (Kolev
et al., 2005). The basic idea is related to core-core-valence Auger transi-
tions (Hörmandinger et al., 1988). In Fig. 3.8, the basic processes are
illustrated. The direct process has core-core transition between states 3
and 1 and valence-valence transition between states 4 and 2. The exchange
process is characterized by valence-core transition from state 4 to state 1
and core-valence transition from state 3 to state 2.
Fig. 3.8 Schematic energy level diagram for core-valence Auger transitions. Numbers
1 and 3 are core states; 2 and 4 are valence states (Hörmandinger et al., 1988).
Kolev et al. (2005) formulated a simple expression for the intensity of

SRAPS:
Z E+Ec X
Iσ (E) ∼ nκµ (E ′ )nγδ (E + Ec − E ′ )Wκµ,γδ
σ
(E ′ , E + Ec − E ′ ),
0 κµ,γδ
(3.7)
where the first Greek letter labels the relativistic spin-orbit quantum num-
ber of the state and the second Greek subscript denotes the magnetic quan-
tum number. n is the LDOS at the atom of interest and can be obtained
from first-principles calculations. W is the effective spin dependent cross
section. The measured results show that the SRAPS amplitude is depen-
Huesler alloys 95
dent on the location at the surface for the X-ray to create holes. The
measured surface SRAPS spectra are compared to bulk SRAPS spectra
in Fig. 3.9. The key feature in Fig. 3.9 is the ratio of amplitudes for the
majority- and minority-spin channels. The experimental result is approxi-
mately three times smaller than the theory based on the bulk model. This
indicates the reduction of spin polarization at the surface. This result agrees
with those of other experiments.
Fig. 3.9 Comparison of theoretical (bulk) and experimental (surface) APS spectra for
the ↓- and ↑-spin states (Kolev et al., 2005).
The XMCD emission spectra of L2,3 in Co2 MnSb were measured by

Yablonskikh et al. (2000, 2001). The two structures relating to the
majority- and minority-spin channels are at 638 eV and 648 eV, respec-
tively. The energies of these structures depend on the photon energy to
excite a hole in the core region. The higher energy peak does not appear
until the exciting photon has an energy of 652 eV.
3.5.7 Disorder in Heusler alloys

Because of complex crystal structures for HH and FH alloys, disorder in
the samples is inevitable. We first comment briefly on probing the disorder
in these alloys and discuss the nature of disorder in HH and FH alloys.
3.5.7.1 Experimental probes

There are two methods to probe local disorder in Heusler alloys: Mössbauer
spectroscopy including the determination of the hyperfine field (Khoi et al.,
1978), and spin echo nuclear magnetic resonance. Neutron and X-ray scat-
tering are methods for determining the degree of disorder.
Mössbauer spectroscopy This method makes use of the recoil speed of

isotopes while the nuclei emit or absorb γ-ray in order to probe the presence
of disorder from linewidths of the radiation. Alloys composed of Fe, Mn,
and Sn can be readily probed. By including the interaction between the
nuclear spin Iz and effective magnetic field He at the site, the hyperfine
field Hhf =−γIz He can be determined, where γ is the gyromagnetic ratio.
For example, 119 Sn Mössbauer spectroscopic measurements on Co2 MnSn
with Fe or Cr substituting Mn at RT have been carried out by Zhang
et al. (2005). Mössbauer spectra for Co2 Mn1−x Fex Sn is shown in Fig. 3.10.
There is a well-defined peak centered at about 2 mm/s for x = 0. When x
≥ 0.2, there is a doublet structure appeared on the peak, suggesting a small
percentage of Sn forming a non-Heusler phase. Wurmehl et al. (2006a) used
57
Co to probe the disorder in a powdered sample of Co2 FeSi at 85 K. They
obtained a sextet pattern with an isomer shift of 0.23 mm/s. This pattern
is typical for magnetically ordered systems. They also observed an 57 Fe line
with a width of 0.15 mm/s, indicating a well ordered sample. This width
is comparable to the 0.136 mm/s width for α-Fe at 4.2 K. More recently,
Karthik et al. (2007) reported Mössbauer spectra of Co2 Crx Fe1−x Si. The
results show a sextet pattern as well. For x ≥ 0.2, there is a doublet
structure.
Fig. 3.10 Mössbauer spectra of Co2 Mn1−x Fex Sn (Zhang et al., 2005).
Spin echo nuclear magnetic resonance It is theoretically possible to

calculate the hyperfine field at a nucleus if the chemical environment is
known. Consequently, direct measurements of nuclear magnetic proper-
ties are effective to determine the disorder in the samples. The nuclear
magnetic resonance (NMR) frequencies are of the order of MHz. The
linewidths of NMR peaks reveal the degree of disorder in the samples.
Huesler alloys 97
For off-stoichiometric and substitutional alloys, two quantities are mea-

sured: the hyperfine field on impurity atoms, and positions of NMR satellite
lines originating from host atoms with nearest neighbor and second nearest
neighbor impurity atoms, in particular for low concentrations. The total
local effective magnetic fieldX at the i-th atom isX
Hi = hoi µo + h<nn>i µ<nn> + h<sn>i µ<sn> + ....., (3.8)
where o means at the i-th site, < nn > and < sn > denote the first and
second nearest neighbors, respectively. hji is a parameter. Let Ai be the
hyperfine coupling constant of the i-th atom, we can calculate hji /Ai to
compare contributions at the i-th atom due to different neighbors.
The spin echo experiments provide more information about the disorder
in a sample. Wojcik et al. (2002) found three lines at 217, 263, and 199
MHz in NiMnSb. These lines were identified from 123 Sb, 55 Mn, and 121 Sb,
respectively. The authors also suggested that there was twinning in their
samples.
Wurmehl et al. (2007) applied spin echo technique to Co2 Mn1−x Fex Si.
By defining the probability for the local environment surrounding an Fe
atom to have n Mn atoms and 12 − n Fe atoms as
N!
P (n, x) = (1 − x)N −n xn , (3.9)
(N − n)!n!
P (n, x) can be directly compared to the intensity of spin echo measure-
ments. In this particular experiment, 55 Mn spin echo intensity as a func-
tion of frequency in Co2 Mn0.5 Fe0.5 Si (Fig. 3.11(a)) and relative area of the
resonance line vs. number of Fe atoms in the third coordination shell of
55
Mn (Fig. 3.11(b)) were obtained.
Other methods Neutron diffraction experiments estimate less than 10%

atomic disorder in NiMnSb (Helmholdt et al., 1984). Kautzky et al. (1997)
carried out X-ray diffraction (XRD) measurements on two single phase
(111) oriented thin films of PtMnSb and found also 10% disorder in each
film. The possible disorders in two different films are given in Table 3.12.
These results were obtained by measuring the (111), (333), and (444) peaks,
then fitting their integrated intensities with structure factors characterizing
different types of disorder. Note that the vacant site (X(2)) in C1b sublattice
can easily cause disorder.
3.5.7.2 Theoretical investigations of disorder

Half-Heusler alloys Structural defects in NiMnSb were investigated the-
oretically by Orgassa et al. (1999, 2000) using KKR method with the co-
Fig. 3.11 (a) 55 Mn spin echo intensity as a function of frequency in Co2 Mn0.5 Fe0.5 Si.
The distribution of Fe atoms in the third coordination shell of the 55 Mn is also given; (b)
Relative area of resonance line vs. number of Fe atoms in the third coordination shell of
55 Mn. Shown are the results for the Co Mn
2 0.5 Fe0.5 Si compound with ideal stoichiometry
x = 0.5 (squares) and optimized stoichiometry x = 0.517 (triangles) (Wurmehl et al.,
2007).
Table 3.12 Disorder determined from XRD measurements on two

films of PtMnSb. Each value in parentheses is the fractional occu-
pancy of sites on a particular C1b sublattice which are occupied by
that atom type.
Film thickness (Å) Site (see Fig. 1.2) Fractional occupancy
1090 X(1) Pt(0.72) Pt(0.77)
X(2) Pt(0.15) Pt(0.10)
Mn(0.15) Sb(0.10)
Y Mn(0.79) Mn(0.94)
Z Sb(1.00) Sb(0.90)
543 X(1) Pt(0.80) Pt(0.83)
X(2) Pt(0.11) Pt(0.07)
Mn(0.11) Sb(0.07)
Y Mn(0.89) Mn(1.00)
Z Sb(1.00) Sb(0.92)
herent potential approximation at the experimental lattice constant, 5.927

Å. They considered three defect models. Recall that there are four inter-
penetrating fcc cubes in the crystal structure. The cube related to X(2)
site in NiMnSb is not occupied. In Table 3.13, three models labelled as
types 1, 2, and 3 are listed. In all cases the composition is kept stoichio-
metric. The question is whether the disorder will destroy the half metallic
properties. These authors calculated P values for the three defect models
with 10 and 15% of disorder for each structure. The results are summa-
rized in Table 3.14 which shows that in general the value of P decreases
monotonically with increasing disorder. The worst case scenario is type-3
disorder. Physically, when the unoccupied X(2) site is occupied by a TM
Huesler alloys 99
Table 3.13 Three defect models studied by Orgassa et al. (1999, 2000).
Defect structure Type Site occupation
X(1) X(2) Y Z
X(1) ↔ Y 1 Ni1−x Mnx Nix Mn1−x Sb
X(1)Y ↔ X(2) 2 Ni1−x Nix Mnx Mn1−x Sb
YZ ↔ X(2) 3 Ni Mnx Sbx Mn1−x Sb1−x
Table 3.14 The effects of disorder on P in

NbNiSb studied by Orgassa et al. (1999).
Defect type Disorder (%) P (%)
1 5 52
10 29
2 5 67
10 31
3 5 24
10 10
element, the d-d interaction is changed in such a way as to diminish the

half metallicity.
Full-Heusler alloys Picozzi et al. (2004) investigated the effects of dis-

order in a 32-atom supercell due to the exchange of Mn and Co atoms and
antisites in Co2 MnSi and Co2 MnGe. The Mn antisite is defined as a Co site
which is occupied by a Mn atom, while the Co antisite is a Mn site occupied
by a Co atom. The method of calculations was the FP-LAPW method with
GGA exchange-correlation. The muffin-tin radius for all atoms was 1.11 Å.
The formation energies were computed as follows:
∆E = Edef − Eideal + nMn µ0,Mn + nCo µ0,Co + nX µ0,X , (3.10)
where the first term is the total energy with disorder, the second term is
the total energy without disorder, ni is the number of atoms transferred to
and from a chemical reservoir of the i-th element, µ0,i is the corresponding
chemical potential, and X denotes Si or Ge. The chemical potentials of Mn,
Co and X were determined from fcc antiferromagnetic Mn, hcp ferromag-
netic Co, and diamond structure Si and Ge. The formation energies and
the total magnetic moments in the supercell are summarized in Table 3.15.
In the ideal (perfectly ordered) case, the total magnetic moment is 40 µB .
The Mn antisite has the smallest formation energy. They suggested that
it can be easily formed during growth, especially in the tri-arc Czochralski
method for Co2 MnSi at 1523 K. Experiments done by Raphael et al. (2002)
show there are 0.36x1022 cm−3 Mn antisites.
Table 3.15 Formation energy ∆E and total magnetic moment M in

a unit cell for the four types of disorder calculated by Picozzi et al.
(2004).
Disorder ∆E (eV) M (µB )
Co2 MnSi Co2 MnGe Co2 MnSi Co2 MnGe
Co antisite 0.80 0.84 38.01 38.37
Mn antisite 0.33 0.33 38.00 38.00
Co-Mn swap 1.13 1.17 36.00 36.00
Mn-Si swap 1.38 40.00
3.6 Physical properties of Heusler alloys in thin-film form
In this section, we shall discuss the physical properties of a few Heusler

alloys in thin-film forms: First, we shall discuss the most stable HH alloy,
NiMnSb, then some FH alloys, including the most studied Co2 MnSi.
3.6.1 NiMnSb
3.6.1.1 Stability of structure and half metallicity
Based on the theoretical studies of bulk properties of HH alloys, the
NiMnSb-type structure is the most stable among all HH alloys, with re-
spect to an interchange of atoms (Larson et al., 2000). Orgassa et al. (1999,
2000) showed that the half metallicity in NiMnSb is not destroyed by a few
% disorder. Galanakis et al. (2002b) attributed the gap in the non-metallic
channel to the d-d interaction between Ni and Mn atoms, which differs from
the explanation in terms of the d-p interaction between Mn and Sb atoms
proposed by de Groot et al. (1983). However, Galanakis et al. (2002b)
argued that the presence of Sb atoms is crucial to stabilize the structure
because NiMn does not favor any open structure.
3.6.1.2 Spin polarization

The NiMnSb films were grown on GaAs(001) by Van Roy et al. (2000)
using the MBE method and on MgO(001) and Si(001) by Schlomka et al.
(2000), using argon-ion sputtering onto water-cooled targets. The thickness
of the films in the first experiment was 260–350 nm. The single-crystal
samples were grown in the [001] orientation. In the second experiment, the
thickness ranged from 1.0 nm to 80 nm. The surfaces showed roughness.
As the temperature was increased from 150◦C to 250◦ C, the roughness was
increased by a factor of 5. These experiments found that the NiMnSb films
were not HM.
Huesler alloys 101
On the other hand, Zhu et al. (2001) grew polycrystalline samples on Si

substrates using the e-beam evaporation method of Kabani et al. (1990) and
carried out spin-resolved photoemission measurements. They obtained 40%
spin polarization which is smaller than an earlier result of 50% obtained by
Bona et al. (1985).
To explain the much smaller measured polarization as compared to the
theoretically predicted value, two suggestions were made. The first is that
the gap in the minority-spin channel is less than the calculated value of 0.5
eV (Kang et al., 1995). The second suggestion is that surface segregation
occurs (Park et al., 1998a).
Ristoiu et al. (2000b) and Komesu et al. (2000) examined the
effects of surface composition on the polarization. They grew
MgO(110)/Mo(100)/NiMnSb(100) thin films by sputtering and capped the
samples with a 1000 Å Sb layer. Inverse photoemission experiments were
carried out to measure the polarization with capping layers removed from
the samples. The P value was found to be 67 ± 9% (Ristoiu et al.,
2000b). The polarization decreased as the thickness of the Sb layer in-
creased (Komesu et al., 2000).
Fig. 3.12 A schematic diagram of the DOS of a HM with the majority-spin channel
showing metallic behavior. The Fermi energy of the Heusler alloy is indicated as EF
and is located right below the conduction band edge by an spin-flip gap δ. ∆ is the
fundamental gap of the insulating channel.
Beside the issue of determining experimentally the polarization P at

Fermi level, there is a transition in NiMnSb from a HM to a normal fer-
romagnet at about 80 K (Hordequin et al., 2000; Borca et al., 2001), the
temperature is referred to as T∗ . One possible reason to explain the tran-
sition is that EF is located very close to the bottom of conduction bands
in the minority-spin channel (Fig. 3.12). As the temperature increases,
electrons at EF in the majority-spin channel can be thermally excited to
the conduction band edge with their spin orientations flipped from the ↑
state to the ↓ spin state. As the conduction band of the ↓ spin channel is
occupied, the half metallicity is lost. The temperature for this to happen in
NiMnSb is 88 K (i.e., δ ≤ 88 K). Consequently, it is unlikely that NiMnSb
can be used for fabricating the RT spintronic devices.
3.6.1.3 Surface and interface effects

Using angular resolved X-ray photoemission spectroscopy, Ristoiu et al.
(2000b) suggested that it is possible to have MnSb or Mn-rich surface layers
on top of NiMnSb thin-film samples depending on how the surfaces were
prepared. Ristoiu et al. (2000a) found that Mn segregations (Mn-rich) and
vacancies in Ni layers, instead of MnSb, terminate the surface of the films by
preparing the samples with sufficient annealing for stoichiometric surfaces.
The effect of Mn-rich surface structure is the cause of the reduction of spin
polarization.
Theoretical study by Jenkins (2004) was performed to examine the
effects of the hexagonal phase of MnSb formed at the surface of
NiMnSb. They used planewave pseudopotential method to examine MnSb-
terminated (001) (1x1) NiMnSb surface. The supercell consists of a slab
having five MnSb/Ni bilayers. A slight buckling of 0.06 Å in the top MnSb
layer was found by relaxing the surface—the Sb atoms relax outward rela-
tive to their bulk positions, while the Mn atoms relax inwards. The second
layer buckles by 0.09 Å in the reverse sense to the top layer. The third layer
has 0.03 Å buckling. The calculated magnetic moment of the Mn atom at
the surface is 4.21 µB , increased by 9.0% as compared to the bulk value.
The symmetry breaking due to the presence of surface introduces surface
states, making bands of the minority-spin channel cross EF . Therefore, in
the surface region, the sample is metallic. The occupied surface states at
k=0 of the surface BZ originated from Mn dx2 −y2 orbitals. The surface
states above EF are derived from dxz orbitals of the Mn atoms mixed with
px states of the Sb atoms. The surface band structures along Γ-K and Γ-M
with different terminations at the surface are shown in Fig. 3.13. In addi-
tion, Wojcik et al. (2002) observed MnSb in epitaxially grown NiMnSb on
a GaAs(001) surface. For Sb-poor samples, the MnSb inclusions disappear.
de Wijs and de Groot (2001) designed a NiMnSb/CdS supercell with

different interface structures and used planewave pseudopotential method
to examine the interface structure on the spin polarization. They found
Huesler alloys 103
Fig. 3.13 Surface band structure of the NiMnSb(111) surface (Jenkins, 2004). The
upper surface terminated by Mn is denoted as Aα , by Ni as Aβ , and by Sb as Aγ . The
bottom surface terminated by Sb is denoted by Bα .
that (100) interface diminished the spin polarization. On the other hand,
the (111) interface with interface composed of the Sb and S atoms as shown
in Fig. 3.14 restores half metallicity. It was also found that (100) and (111)
surfaces of NiMnSb do not exhibit half metallic properties.
Fig. 3.14 Structure of the half metallic NiMnSb(111)/CdS(111) interface (de Wijs and
de Groot, 2001).
The magnetic properties of half metallic NiMnSb and CoMnSb surfaces

have been calculated by Galanakis (2002b). The results are summarized
in Table 3.16, along with the corresponding bulk values. In general, the
magnetic moment at the surface is increased mainly due to the presence of
TM elements with increased local moments at the surface.
Magnetization and Kerr rotation measurements were carried out on thin
films of NiMnSb by Kabani et al. (1990). The results of the hysteresis loop
for a film sample of 220 nm measured at 4.2 K show significant anisotropy
(Fig. 3.15). The origin of the anisotropy is not identified. The magnetic
moment of 3.9±0.2 µB per formula-unit was determined. These authors
Table 3.16 Spin moments in µB for NiMnSb and CoMnSb

compounds in the case of: (i) bulk compounds; (ii) the Mn
and Sb atoms in the surface; and (iii) the Ni- or Co-terminated
surfaces. The “total” moment denotes the sum of moments in
the surface and subsurface layers that include Sb and vacancy
(Galanakis, 2002b).
Interface NiMnSb CoMnSb
Ni Mn Total Co Mn Total
Bulk 0.26 3.70 3.96 -0.13 3.18 2.96
(001)MnSb 0.22 4.02 4.19 -0.06 3.83 3.65
(001)Ni 0.46 3.84 4.30
(001)Co 1.19 3.31 4.43
also measured the Kerr rotation at RT on the same film using the MOKE
method. A laser source with a wavelength of 632.8 nm was incident on the
film surface at 45◦ . The magnetic field was swept between ±3.5 kOe. At
the limiting value of the magnetic field and in the polar configuration, the
Kerr rotation is 0.06◦ as compared to the bulk value of 0.10◦ ; It seems that
the film has magnetic properties close to those of the bulk.
Fig. 3.15 The hysteresis loop of a 220 nm NiMnSb film at 4.2 K (Kabani et al., 1990).
3.6.2 Co2 MnSi

Among all Heusler alloys, bulk Co2 MnSi has several appealing properties.
It is predicted to be a HM with a large insulating gap of about 0.4 eV in the
↓ spin channel (Fuji et al., 1990). It possesses the second highest TC (985 K)
among all Heusler alloys (Brown et al., 2000). For spintronic applications,
samples in thin-film form are preferred. Therefore, it has attracted much
study in thin-film forms. GaAs has been used as the substrate. Amorphous
Al-O and MgO have been used as the barrier in tunnel junctions. The high
polarization deduced from the magnetic tunnel junction made of Co2 MnSi
Huesler alloys 105
shows great potential for spintronic devices.
3.6.2.1 Growth in thin-film form

Co2 MnSi in thin-film form is among the most studied full Heusler alloys.
It has been grown on a GaAs(001) surface by PLD method (Wang et al.,
2005a), inductively coupled plasma-assisted magnetron sputtering method
and dc magnetron sputtering (MS) method (Kohn et al., 2007). We shall
discuss in detail the growth by PLD and dc MS methods because of some
unique requirements during the growth. Toward device applications, we
shall discuss magnetic tunnel junctions (MTJs), in particular Co2 MnSi/Al-
O/Co2 MnSi, which were fabricated using the inductively coupled plasma-
assisted MS method by Sakuraba et al. (2006).
In 2005, Wang et al. reported the use of polycrystalline Co2 MnSi pellet
targets to grow thin-film forms of Co2 MnSi by PLD on GaAs(001) sub-
strates. The lattice constant of GaAs (5.65 Å) matches that of Co2 MnSi.
These substrates were commercially available and were degassed in UHV
up to 580◦C. Finally, the surfaces of substrates were sputtered for 30 min.
at 600◦C by an Ar+ ion beam with energy at 0.6 keV. The beam was inci-
dent at 45◦ with current density 4.0 µA/cm2 . A KrF excimer laser of 248
mm wavelength was used with 34 ns pulse width, 300 mJ energy, and 10 Hz
repetition rate. The pressure inside the chamber was kept under 5.0x10−11
mbar. Previous experience showed difficulties in growing quality Heusler
alloys on semiconductor surfaces at high deposition temperatures (> 450
K). The present growth was carried out with substrate temperature at 450
K. Quality thin films of around 60 Å were obtained. Monitoring RHEED
oscillations provided the evidence of the film quality. Regular oscillations
started at the third layer and continued up to 10 layers. The lack of os-
cillations at the first two layers was attributed to the poor layer-by-layer
growth and the possibility of interface mixing. The characterization of the
quality of the films by AES has been given in Section 3.3.
Most recent growth details were given by Kohn et al. (2007). The GaAs
(001) substrates were positioned below the targets on a Ta strip heater.
The substrates were chemically cleaned and annealed at 595◦ C for 10 min
to remove oxides and to have a 4x2 reconstruction. The base pressure in
the deposition chamber was 2.67x10−9 mbar. The temperature was then
lowered to 380◦ C and the system was pumped for one and half hours to
remove the As atoms completely—avoiding the formation of Mn2 As com-
pound. The rate of deposition was 0.10 nm/s under an Ar pressure of 32
mbar. Energy dispersive X-ray analysis (EDX) in a scanning electron mi-

croscope was carried out and determined stoichiometric samples within 1.5
at.%. X-ray diffraction patterns showed the films having single-phase (001)
orientation with a lattice constant of 5.63±0.01Å. Film thicknesses of 15 to
260 nm were obtained. There was intermixing between Mn and As atoms
in the films up to 15 nm. For thicker films, the L21 structure is the main
polycrystalline phase.
3.6.2.2 Magnetic properties

X-ray absorption and magnetic circular dichroism The spin po-
larization, P, at EF is the quantity of interest and is determined by the
magnetic moments of TM elements. The X-ray absorption spectra (XAS)
and X-ray magnetic circular dichroism (XMCD) methods, combined with
sum rules, are widely used for determining magnetic moments. Typically,
the L2,3 absorption edges of the Mn and Co atoms are used. The sum rule
relates the integrated intensity difference of the two lines to the orbital and
spin moments of the atoms, corrected for the incomplete degree of circular
polarization and angle of incidence.
The XAS spectra for Co-L2,3 of a Co2 MnSi film of 17 Å thickness at RT
obtained by Wang et al. (2005a) are shown in Fig. 3.16. The main features
are as follows:
• For the Co atom, XAS shows lines at 776.5 eV and 791.5 eV corre-
sponding to 2p3/2 → 3d (first peak, L3 ) and 2p1/2 → 3d (second peak,
L2 ) transitions, respectively. The edges of L2,3 lines for the Mn atom
are at 638.5 eV and 651.4 eV.
• The line widths of the Mn atom are narrower than those of the Co
atom.
• The high energy line of Mn exhibits a doublet structure at 649.4 eV
and 650.8 eV.
• There is some variance with respect to the thickness of the films. For
example, the doublet structure of the Mn L2 line is smeared in a thicker
(45 Å) film.
• In contrast to the L3 line of the Co atom in other Heusler alloys such
as Co2 TiSn, Co2 ZrSn, and Co2 NbSn (Yamasaki et al., 2002), showing
multiple structures, this one does not exhibit any such structures.
The narrowness of the Mn lines can be attributed to the number of Mn

and Co atoms in the sample. The Mn atoms suffer less local field effect due
Huesler alloys 107
to their environments.
The doublet structure in the higher energy line shows the atomic char-
acteristic of the Mn atom due to the fact that the Mn-Mn distance is larger
than the separation of the Co atoms and is attributed by Wang et al. (2005a)
to the interplay of (1) the exchange and Coulomb interactions between core
holes and unpaired valence electrons, and (2) the hybridization between 3d
orbitals and surrounding states. In a thicker film, the smearing out of the
structure according to Wang et al. (2005a) is due to the reduction of orbital
contributions as the thickness of the film increases. They also attributed
the absence of multiple structures appearing in Sn-based Heusler alloys to
more metallic Co d-states in Co2 MnSi.
The XMCD spectra are the normalized difference of the left- and right-
handed polarized X-rays. They probe the exchange splitting and spin-orbit
coupling of both initial core and final valence states.
Fig. 3.16 XAS Co-L2,3 lines (first and second peaks) of the Mn (a) and Co (b) atoms
from a Co2 MnSi film of 17 Å thick (Wang et al., 2005a).
Fig. 3.17 XMCD Co-L2,3 lines (first and second peaks) of the Mn (a) and Co (b) atoms
from a Co2 MnSi film of 45 Å thick (Wang et al., 2005a).
A summary of the features in Fig. 3.17 is as follows:
• The structures in XMCD are in general 0.5 eV lower than the corre-
sponding peaks of XAS.
• For the Co atom, the L3 line is approximately a factor of two stronger
than the L2 line.
• For the Mn atom, the L3 line is more than a factor of three stronger than
the L2 line and about a factor of two stronger than the corresponding
Co line.
• Total intensities deduced from the sum rule correspond to 2.24 unoc-
cupied states for the Co atom and 4.52 for the Mn atom.
For the Mn atom, the large negative dichroism within the L3 region
(lower panel in Fig. 3.17) is attributed to the large separation between the
Mn atoms and smaller coordination numbers, resulting in weak Mn-Mn
magnetic coupling. The Mn atom, consequently, exhibits a large magnetic
moment. With 45 Å thick film, the sum rules provide information about the
spin moments: 1.04 µB per Co atom, comparing well with the predicted
value of 1.06 µB . However, one must bear in mind that the sum rules
underestimate both the effective spin moment and effective orbital moment.
Magneto-optical Kerr effect (MOKE) Wang et al. (2005a) carried

out MOKE measurements on film samples of Co2 MnSi at 70 K. The be-
havior of the Kerr rotation in the remanence along the [11̄0] easy axis as a
function of the thickness of the films is shown in Fig. 3.18. The results are
summarized as follows:
• With thickness less than two bilayers, there is no magnetization. This

could be due to magnetically inert layers or small islands preventing
magnetic ordering.
• With thickness greater than two bilayers and an applied field of 1 kG,
there is a nonzero Kerr signal. Beyond four bilayers, the behavior is
close to linear.
• In-plane uniaxial magnetic anisotropy with the easy axis oriented along
[11̄0] is detected when the thickness increases beyond three bilayers.
This is the critical layer thickness for the onset of ferromagnetic long
range order.
The anisotropy in the plane of the film has been examined by measuring
the Kerr rotation in a 60 Å film under a magnetic field with different angles
between the easy [11̄0] and hard [110] axes. The geometry and results are
Huesler alloys 109
Fig. 3.18 Thickness dependence of longitudinal Kerr rotation measured at 70 K for

Co2 MnSi films. The open dots are the results for the films thinner than four bilayers (the
superparamagnetic phase) in an applied field of 1 kG. The full dots show the saturated
Kerr intensity of the ferromagnetic phase (above four bilayers), which is equivalent to
the intensity in remanence in this case(Wang et al., 2005a).
Fig. 3.19 Geometric diagram of the Fig. 3.20 Kerr rotation as a function
magnetization and applied magnetic of magnetic field applied in the plane of
field (Wang et al., 2005a). a 60 Å film (Wang et al., 2005a).
shown in Fig. 3.19 and Fig. 3.20. Experiments were carried out at 80 K.
The important results shown in Fig. 3.20 are:
• The appearance of sub-loops begins with magnetic field angle α greater

than 60◦ with respect to the easy axis.
• For 65◦ < α < 75◦ , there are three loops, including a central loop.
• For α = 90◦ , the three loops coalesce into one loop.
The appearance of sub-loops is explained by the applied magnetic field

having enough strength to magnetize the sample along the hard axis. The
origin of the anisotropy is unclear.
The temperature dependence of Kerr rotations in different film thick-
nesses suggests that the magnetization can be well described by the usual
Bloch formula, in particular for T < 0.6TC ,

M(T) = M(0)(1 − bT3/2 ), (3.11)
where M (0) and b depend on the thickness of the films. In Fig. 3.21, b is
shown to be proportional to the inverse of the film thickness (Wang et al.,
2005b).
Physically, this formula describes the spin-wave excitations of a ferro-
magnetic sample at finite temperature. The linear dependence shown in
Fig. 3.21 suggests that the film thickness decreases with the magnetization
contributed by the surface and interface.
Fig. 3.21 Average spin-wave parameter b as a function of the inverse film thickness d
of the Co2 MnSi films grown on GaAs(001). The solid line is a linear fit (solid line) to
the experimental values (Wang et al., 2005b).
Spin-resolved photoemission spectroscopy (SRPES) The SRPES

spectra discussed below were obtained by Wang et al. (2005a) using the
UE56/2-PGM2 beamline at BESSY in Berlin. The film samples were mag-
netized in-plane along the [11̄0] direction and have thicknesses ranging from
17 Å to 45 Å. The resolution of experiment is of the order of 0.5 eV. The
results of spin spectra and polarization measurements are given in Fig. 3.22.
The important features include:
• The overall intensity of the majority-spin spectrum is higher than that
of the minority spin.
• There is a structure at the binding energy of 0.9 eV which is conjectured
to arise from a mixture of metallic Co and Mn components.
• The spin polarization at Fermi level P = (I↑ − I↓ )/(I↑ + I↓ ) were
calculated. The values at RT were 10% and 18% for the 17 Å and 45
Å samples, respectively. By correlating the temperature dependence of
P to that of the magnetization, the extrapolated P at T = 0 is found
Huesler alloys 111
to be 12%, independent of film thickness. This is quite low compared

to the theoretical polarization of 100%.
Fig. 3.22 SRPES of two Co2 MnSi thin films measured at RT. (a) and (c) are spin
resolved spectra. (b) and (d) are spin polarization (Wang et al., 2005b).
Point-contact Andreev reflection (PCAR) Singh et al. (2004b) car-

ried out PCAR measurements on the Co2 MnSi thin films. The samples
were grown by the dc magnetron co-sputtering method on an array of a-
plane sapphire substrates. The base pressure was 2.67x10−9 mbar. The
deposition rate was 0.10 nm/s. The substrate temperature ranged from
545 K to 715 K. An optimized stoichiometric film was grown at 715 K with
400 nm thick and the L21 structure. However, it was in polycrystalline
form with strong (110) texture. The P value measured with PCAR method
was 56% at 4.2 K.
3.6.2.3 Transport properties

Measurements of resistivity ρ as a function of T reveal the operative scat-
tering mechanisms. For example, if ρ is proportional to T2 , the scattering
may be electron-magnon or electron-electron in nature. A common method
for measuring resistivity is the four-point DC method. For the stoichiomet-
ric films of 400 nm thick Co2 MnSi grown at various Tsub , the temperature
of the a-plane sapphire substrate, Singh et al. (2004b) measured ρ at 4.2 K
as a function of Tsub . As shown in Fig. 3.23, the results show two distinct
linear behaviors in different regions of Tsub divided at 570 K. The decrease
of ρ is due to the increase of grain sizes.
It is more interesting to examine ρ as a function of T for a thin-film
sample. Singh et al. (2004b) carried out these types of measurements on a
sample grown at Tsub of 715 K. For 295 K > T > 100 K, ρ decreases lin-
Fig. 3.23 ρ as a function of Tsub of a Co2 MnSi film (Singh et al., 2004b).
early with T, indicating electron-phonon scattering. For T < 100 K, ρ (T)

fits well with T2 +T9/2 . These authors attributed the second contribution
to the electron-two-magnon scattering. They suggested that the T2 depen-
dence originated from either the electron-one-magnon or electron-electron
scattering.
3.6.2.4 Magnetic tunnel junctions (MTJs)

The quantity of interest in MTJs is the tunnel magnetoresistance (TMR).
The basic principle was discussed in Chapter 2. The junctions have nu-
merous device applications. They are essential components of sensors and
memory devices.
Since fabricating MTJs is technologically important and an early theory
predicted that disorder, such as the antisites, can destroy half metallic
properties of Co2 MnSi (Ishida et al., 1998). it is, therefore, important to
control precisely the microstructure of the sample. In the following, we
discuss methods of growing films and junctions.
Kämmerer et al. (2003) grew Co2 MnSi films onto a vanadium(V)-buffer
layer which can assist in (110) texture formation at RT using combined
dc and radio frequency magnetron sputtering. They also found that it
is effective to use V as the seed layer. However, the samples show low
saturation magnetization. These authors (Kämmerer et al., 2004) therefore
suggested the following strategy for successful growth of Co2 MnSi films for
MTJs.
• Prepare a magnetically optimized layer of Co2 MnSi through the evo-
lution magnetic moment as a function of annealing temperature on a
V-buffer layer.
• Grow an oxide layer, such as AlOx , as a barrier by oxidizing an Al
Huesler alloys 113
layer.
• Grow the upper magnetic electrode.
• Grow the upper current lead.
In practice, details of each step are important. During the growth, the
base pressure was 1x10−7 mbar. In order to have a large TMR, it is crucial
to have clean interfaces between Co2 MnSi and AlOx and between AlOx and
Co70 Fe30 —the upper magnetic electrode. This requires the whole growth
process to take place in a vacuum chamber without any interruption.
The steps for growing MTJs having SiO2 (substrate)/ V (42 nm)/
Co2 MnSi(100 nm)/ AlOx (1.8 nm)/ Co70 Fe30 (5.1 nm)/ Mn83 /Ir17 (10 nm)
are:
• Use dc magnetron sputtering to grow the V buffer, Co2 MnSi, and Al
layers.
• Oxidize the Al layer in pure oxygen plasma for 150 s.
• Anneal for 40 min at temperatures between 400 and 500 ◦ C.
The last step is an important one. It can result in ferromagnetically
textured Co2 MnSi with minimal disorder and homogenize the AlOx barrier.
It is tested by monitoring the magnetic moment/formula-unit, M, of the
ferromagnetic layer. After 1 hour at temperature above 250◦C, the value
of M is more than 4.0 µB . At 375◦ C, M reaches 4.7 µB . The annealing
helps the (110) texture formation in the Co2 MnSi layer as shown in the
X-ray diffraction pattern (Fig. 3.24). The existence of (110) texture was
evidenced by the gray (220) line (Kämmerer et al., 2004).
This line indicates that there is a periodic structure with periodicity
twice or four times longer than the distance between adjacent (110) planes.
When V is used, these two lines accompanied by the V(110) line and a
shoulder of the V(220) line are clearly seen (Fig 3.24).
• Oxidize the top AlOx layer for 50 seconds to clean the surface.
• Sputter Co70 Fe30 by dc magnetron sputtering to form the upper mag-
netic electrode.
• Use radio frequency magnetron sputtering to put an antiferromagnetic
Mn83 /Ir17 layer on top.
• Apply dc magnetron sputtering again to form a Cu/Ta/Au multilayer
lead.
• Impose an exchange bias between Mn83 /Ir17 and upper ferromagnetic
Co70 Ir30 electrode by annealing the multilayer structure for 1 hour at
275◦C under an external magnetic field with a strength of 100 mT.
• Pattern the multilayer material using optical lithography and ion beam
etching to form a 200 µm quadratic (four different layers) MTJs.
Fig. 3.24 X-ray diffraction pattern of Fig. 3.25 ∆R/R(Hext ) of a MTJ as a

a 100 nm Co2 MnSi film on a 42 nm function of external magnetic field at
V buffer layer taken after an anneal- RT(Kämmerer et al., 2004).
ing process of 1 hour at temperature
above 250◦ C, the (hkl)-peaks belong-
ing to Co2 MnSi are indicated in gray
(Kämmerer et al., 2004).
The next issue is the determination of the TMR. Two quantities were
measured by Kämmerer et al. (2004) to determine the TMR of MTJs: The
resistance under a magnetic field and the change of resistance ∆R which is
defined as the difference of the resistances when the magnetization between
the two electrodes are parallel and antiparallel. The ratio ∆R/R(Hext )
of these two quantities is related to the TMR, where R is in general the
resistance measured when the two electrodes having parallel magnetization.
The ratio plotted against the external field of a is shown in Fig. 3.25.
Figure 3.26 explains schematically the tunnel processes (Sakuraba et al.,
2006). Note that in a case to be discussed, both the sample and the upper
electrode are made of the same materials. The upper panels are for the
parallel configuration of magnetizations in the sample (at right) and upper
electrode (at left). A bias potential e∆V is applied such that a positive
bias causes an electron to tunnel from the right to the left. The rectangular
area indicates the tunneling barrier. The density of states showing the half
metallicity of the sample and upper electrode are plotted on both sides
of the barrier. The gap in the minority-spin channel is denoted by EG .
δCB is the energy separation between the bottom of conduction bands and
EF , and δV B is the energy separation between EF and the top of valence
bands. When e∆V = 0, EF is aligned on both sides. An electron in the
Huesler alloys 115
majority-spin channel can tunnel from the sample to the upper electrode.
When e∆V > EG , EF in the upper electrode is shifted down by e∆V. A
new tunneling channel between minority-spin states is open as the result.
Fig. 3.26 Schematic of electron tunneling process at finite bias voltage in the
Co2 MnSi/Al–O/Co2 MnSi system for parallel (upper panels) and antiparallel (lower pan-
els) configurations(Sakuraba et al., 2006).
The lower panels are for the antiparallel configuration of magnetizations

in the two electrodes. Schematically, it is represented by a reversal of
spin channels in the sample. At e∆V = 0, an electron in the minority
(↓) spin channel of the sample cannot tunnel to the left because of the
insulating gap in the upper electrode, neither can a valence electron in a
majority- (↑) spin state due to the Pauli principle—the state in the upper
electrode is filled. With finite bias, electrons in both channels can eventually
tunnel. The change of current at small bias voltage gives ∆R when a parallel
configuration is switched to an antiparallel configuration.
A simple analysis given by Kämmerer et al. (2004) expresses TMR as a
function of the difference between θsample and θupper , where θsample is the
angle between magnetization of the sample (e.g., Co2 MnSi) and external
magnetic field, and θupper is the angel between upper electrode, Co70 Ir30
(Kämmerer et al., 2004) or Co2 MnSi (Kämmerer et al., 2003) and external
field.
ATMR (T)
TMR(Hext ) = [1 − cos(θsample − θupper )], (3.12)
2
Msample cos(θsample ) Mupper cos(θupper)
M(Hext ) = + , (3.13)
[tsample /(tsample + tupper )] [tsample /(tsample + tupper )]
where AT MR is the measured amplitude of the ratio ∆R/R(Hext ). It is
called the TMR-effect amplitude and is a function of T. The measured
∆R/R(Hext ) as a function of external magnetic field for MTJs is shown
in Fig. 3.25. Using Eqs. (3.12) and (3.13), the corresponding TMR as
a function of T is shown in Fig. 3.27. The TMR-effect amplitude at low
temperature is 86%. More recently, Sakuraba et al. (2006) fabricated a
Co2 MnSi/AlO/Co2 MnSi MTJs and determined the TMR as a function of
T. At RT, the value of the TMR approaches 67%, while at 2K the value is
570%.
Fig. 3.27 TMR-effect amplitude as a function of T in

Co2 MnSi/AlOx /Co70 Fe30 Mn83 Ir17 (open circles), Mn83 Ir17 Co70 Fe30 /AlOx /Co70 Fe30
(filled triangles), and Mn83 Ir17 Co70 Fe30 /AlOx /Ni80 Fe20 (filled squares) (Kämmerer
et al., 2004).
We now discuss the relationship between TMR-effect and spin polar-

ization P using Jullière’s formula. The relation between TMR and polar-
izations of the Co2 MnSi (sample) and Co70 Ir30 (electrode) at 10K can be
found through:
2Psample (10K)Pelectrode (10K)
T M R(10K) = . (3.14)
1 − Psample (10K)Pelectrode (10K)
One can deduce Psample at 10 K from this relation when the measured
TMR-effect is saturated. Using P ∼ 50% for Co70 Ir30 determined by
Thomas (2003) and TMR-effect amplitude 86%, Eq. (3.14) gives P ∼ 61%
for the Co2 MnSi sample (Kämmerer et al., 2004).
Huesler alloys 117
For MTJs fabricated by Sakuraba et al. (2006), its P value can also
be determined from Jullière’s formula. Using P = 89% for the Co2 MnSi
electrode and TMR-effect amplitude of 570%, the P value of the Co2 MnSi
sample is 83%. In addition, Schmalhorst et al. (2004) grew MTJs similar to
that of Kämmerer et al. (2004) except they replaced Co70 Fe30 by an anti-
ferromagnetic layer of Mn83 Ir37 . The P value for this sample is determined
to be 86% at 10 K.
By comparing the values of P determined from different magnetic mea-
surements, it is found that single crystalline Co2 MnSi films grown on
GaAs(001) substrates show the smallest P values. The possible reasons
for this discrepancy are: (i) there is another nonstoichiometric phase dis-
tributed inside the film, (ii) an nonmagnetic metallic phase in the sur-
face/interface regions could add an equal contribution to both spin com-
ponents and thus reduce the polarization. Wang et al. (2005a) suggested
that the possibility of the second reason is unlikely according to their Kerr
effect as a function of thickness measurements which indicate that the sur-
face/interface regions are magnetic.
To explain the 61% value of P which is substantially less than the pre-
dicted 100%, Schmalhorst et al. (2004) attributed the reduced value to spin
scattering of tunneling electrons from the paramagnetic Mn ions and atomic
disorder in the interface region. There is also the possibility of intermixing
of Co and Mn, and Co atoms occupying Mn sites. As shown by Picozzi
et al. (2004), with Mn sites occupied by the Co atoms, half metallicity per-
sists in Co2 MnSi. Neutron diffraction experiments (Raphael et al., 2002)
suggest that there are approximately 10–15% Co-Mn antisites exit in the
sample, with Mn sites occupied by the Co atoms.
Based on the above discussions, it should be clear that the TMR of
MTJs with the Co2 MnSi as the lower electrode (sample) depend critically
on the choices of the seed layer and materials for the upper electrode.
3.6.3 Co2 FeSi

This Heusler alloy has the highest TC (1100 K) and largest magnetic
moment/formula-unit (5.97±0.05 µB ) at 5 K known to date (Wurmehl
et al., 2006a). Since it has more attractive properties than the Co2 MnSi, a
number of groups have focused on thin-film forms of Co2 FeSi. We shall dis-
cuss the growth of these forms, their characterizations, magnetic properties,
and transport properties. Finally, magnetic tunnel junctions composed of
Co2 FeSi will be discussed.
3.6.3.1 Growth in thin-film form

The radio frequency sputtering method is one of the methods used to grow
thin-film forms of Co2 FeSi on either MgO(100) (Schneider et al., 2006) or
Al2 O3 (112̄0) (Schneider et al., 2006) substrates. Another one is the MBE
method (Hashimoto et al., 2005) using GaAs(001) as the substrate. In the
following, we discuss the radio frequency method. The target preparation
was discussed in Chapter 2. The base pressure in the chamber was 5.0x10−8
mbar. The rate of deposition was 5 Å/s. The best samples were obtained
with an Ar pressure of 1.0x10−2 mbar and Tsub of 700◦C. After the films
were grown, a 4 nm Al layer was deposited on the top at 350◦ C to prevent
the oxidation.
3.6.3.2 Characterizations
Bragg reflection of X-rays has been used to characterize the films (Schneider
et al., 2006). In Fig. 3.28, the Bragg lines of a Co2 FeSi film grown on
MgO(100) substrate are shown. The (200) and (400) lines are sharp with
widths of 0.3◦ . The results indicate that the growth in the out-of-plane
direction is good. For the films grown on Al2 O3 (112̄0) substrate, there are
three different epitaxial domains. Only Co2 FeSi films grown on MgO are
fully epitaxial (defect-free layers), as revealed by LEED pattern (Fig. 3.29).
The spots clearly show the four-fold symmetry.
Fig. 3.28 X-ray Bragg reflection lines Fig. 3.29 The LEED pattern identify-
from a Co2 FeSi film ing epitaxial growth of Co2 FeSi grown
grown on MgO(100) (Schneider et al., on MgO (Schneider et al., 2006). The
2006). primary electron beam energy is 318 eV.
Huesler alloys 119

XAS and XMCD experiments Magnetization was measured by
XMCD using the UE56/1–SGM beamline at the BESSY-II Synchrotron
light source (Schneider et al., 2006). All measurements were under an ex-
ternal magnetic field of 16 kG to saturate the magnetization. The field was
perpendicular (⊥) to the film surface. The measurements were carried out
at RT. The polarization of the X-ray was kept constant and the external
field was reversed in direction. Total electron yield (TEY) measurements
were carried out simultaneously.
The samples were 68 nm layers of Co2 FeSi grown on different substrates.
Co L2,3 and Fe L2,3 lines were measured. For the XAS experiments, the
total electron yield was measured in addition to the XAS. The TEY signal
originates from the upper interface region within the electron escape depth.
In order to collect all the escaping electrons, a conducting tube with a
positive bias voltage (100 V) is used to shield the film under investigation.
The photon flux of XAS transmitted through the sample is converted into a
signal detected by a GaAs photodiode (PD)—the photodiode current IP D .
A transparent cap layer, that is used to prevent the detection of any X-
ray fluorescence from the sample, protects the photodiode. The setup is
schematically shown in Fig. 3.30 (Kallmayer et al., 2007) . IT EY measures
the intensity of TEY. The XMCD spectra were measured by flipping the
magnetic field while keeping the polarization of the X-ray fixed.
Fig. 3.30 Schematic diagram of TEY measurements (Kallmayer et al., 2007). PD is a

photodiode.
Co2 FeSi (110) film on Al2 O3 (112̄0) substrate Two Co L2,3 lines deter-
mined by TEY are recorded in both XAS and XMCD experiments.
The results are shown in Fig. 3.31. Assuming that the luminiscence
signal of the substrate I± lum is proportional to the intensity of the

transmitted X-ray and Iref is the reference intensity from the bare
substrate.h µ± are the aborption coefficients and can be calculated
+
i
1 I (hυ)

by − d ln lum Iref (hυ) , where d is the thickness of the film.
+ −
IT EY (hυ) = Ilum (hυ) + Ilum (hυ). Two peaks at 781 eV and 797.2
eV are observed in the XAS spectra. A smaller structure located at
3.5 eV above the 781 eV line is also observed, which was noted as an
indicator of the quality of the sample. It is noted that only the L21
structure shows this peak structure and it is absent in known disordered
and selectively oxidized Heusler alloys. The corresponding structures
in the XMCD spectra are shifted down by about 0.5 eV as in the case
of Co2 MnSi.
Fig. 3.31 XAS and XMCD spectra Fig. 3.32 XAS and XMCD spectra
of Co lines from a 68 nm Co2 FeSi of Fe lines from a 68 nm Co2 FeSi
film grown on a Al2 O3 (112̄0) substrate film grown on a Al2 O3 (112̄0) substrate
(Schneider et al., 2006). (Schneider et al., 2006).
For the Fe lines on the similar film sample, the two peaks in the XAS
spectra are located at 711 and 724 eV, respectively (Fig. 3.32). Only
a very weak shoulder at 715 eV is seen. The shifts of energies in the
XMCD spectra are relatively small compared to the Co case. Using
the sum rule with assumed number of d-holes for each element, nd (Fe)
= 3.4 and nd (Co) = 2.5 as reported for the pure elements (Wurmehl
et al., 2006b), magnetic moments of the Co and Fe atoms were deter-
mined. They are summarized in Table 3.17. Where TMD indicates
that the value was deduced from transmission data. The ratios of the
orbital and spin moments for Co and Fe are 0.12 and 0.04, respectively,
from the TEY results. The ratio for the Co atom (0.13) agrees well
with that in crystalline Co (Carra et al., 1993). However, for Fe, the
Huesler alloys 121
Table 3.17 Element specific magnetic mo-

ments (µB ) for Co and Fe in Co2 FeSi films
grown on a Al2 O3 (112̄0) substrate.
Moment type Co Fe Co2 FeSi
Spin (TEY) 1.13 2.47 4.73
Spin (TMD) 1.25 2.43 4.93
Orbit (TEY) 0.14 0.10 0.38
Orbit (TMD) 0.11 0.07 0.29
Sum (TEY) 1.27 2.57 5.11
Sum (TMD) 1.36 2.50 5.22
SQUID 4.8
Table 3.18 Element specific magnetic mo-

ments (µB ) for Co and Fe in Co2 FeSi films
grown on MgO(100) substrate.
Moment type Co Fe Co2 FeSi
Spin (TEY) 1.07 2.46 4.60
Spin (TMD) 1.28 2.46 5.02
Orbit (TEY) 0.04 0.05 0.13
Orbit (TMD) 0.13 0.12 0.38
Sum (TEY) 1.11 2.51 4.73
Sum (TMD) 1.41 2.58 5.40
SQUID 4.8
present result compared to the crystalline value reported by Carra et al.

(1993) is approximately 33 times smaller. One possible explanation is
that moments of the neighboring Co atoms can orient in the opposite
direction relative to that of the Fe atom.
The sum of the spin and orbital momenta for the two measurements
are larger than the result obtained by SQUID. The SQUID measure-
ment was carried out at 300 K. The discrepancy was attributed to the
assumed number of holes used in determining the magnetic moments
from the X-ray data.
Co2 FeSi(100) film on MgO(100) substrate The results of XAS and XMCD
measurements of Co and Fe lines from a 68 nm Co2 FeSi layer grown on
MgO(100) substrate are similar to those shown in Figs. 3.31 and 3.32.
The element specific magnetic moments for the Co and Fe atoms de-
duced from the measurements and sum rule are given in Table 3.18.
The orbital moments of the Co and Fe atoms deduced from the TEY
spectra are three and two times smaller than those found on the Al2 O3
(112̄0) substrate. For the transmission spectra, the results for both
substrates agree reasonably well.
The experimental results of Schneider et al. (2006) were compared to
the XAS spectra of pure Co and Fe by Chen et al. (1995) shown in

Fig. 3.33. The peak positions shown in Figs. 3.31 and 3.32 agree well
with those shown in Fig. 3.33.
Fig. 3.33 XAS spectra of Co and Fe (Chen et al., 1995).
All these results suggest that localized properties, such as the element
specific spin moment, are not affected by whether the sample is in thin-
film form or bulk. Theoretical predictions of spin moments of the Co
and Fe atoms based on LDA by Galanakis1 are 2.3 µB and 3.5 µB , re-
spectively. The spin moment calculated with LDA+U (Kandpal et al.,
2006) of Co is 1.53µB and that for Fe is 3.25 µB . The magnetic mo-
ment is 6.0 µB /unit-cell. The last value was confirmed experimentally
(Wurmehl et al., 2005). The deduced element specific spin moments
given in Tables 3.17 and 3.18 are smaller than the theoretical values,
in particular, the LDA results.
Magneto-optical Kerr effect (MOKE) As discussed previously in the

case of Co2 MnSi, MOKE provides measurements of the anisotropy of a film.
It is a first order effect and is divided into:
• Polar MOKE (PMOKE), which is proportional to the out-of-plane mag-

netization.
• Longitudinal MOKE (LMOKE) ML , which is related to the in-plane
magnetization parallel to the plane of incidence light.
• Transverse MOKE (TMOKE) MT , which provides information about
the in-plane magnetization perpendicular to the plane of incidence light.
The relative orientations of the magnetization, external magnetic fields

Hi , and crystal axes of the film are shown in Fig. 3.34 (Hamrle et al., 2007).
1 Unpublished
Huesler alloys 123
Fig. 3.34 (a) Relative orientations of the incident s-polarized light (S), magnetization
in the film (ML and MT ), and the external magnetic field H8 . (b) Relative orientations
of crystal axes, magnetization of the film, and the external magnetic field in different
orientations Hi . The electric field is perpendicular to the beam velocity (Hamrle et al.,
2007).
The MOKE loops in most Heusler alloy films are symmetric with re-
spect to angle α (Fig. 3.34). As experiments have progressed, evidences
have emerged that MOKE loops can be asymmetric. Postava et al. (2002)
attributed the asymmetry to the quadratic behavior of MOKE, QMOKE,
ML MT , and ML2 − MT2 .
Physically, the observation of the Kerr effect depends on both the spin-
orbit and exchange interactions. Unlike PMOKE and LMOKE, where the
magnetization M is parallel to the wave vector of the light k, QMOKE
requires M to be perpendicular to the plane of incidence.
Therefore, the first order contribution of the spin-orbit interaction does
not play a role. The second order contribution of the spin-orbit interaction
is smaller so QMOKE is a smaller effect. To show that QMOKE depends
on the quadratic contributions of ML and MT , we outline the macroscopic
derivations (Foner, 1956) as follows: The response function of a magnetic
medium to the light can be described by the dielectric tensor,
(0)
εij =εij +Kijk Mk +Gijkl Mk Ml +..., (3.15)
(0)
where εij is the linear dielectric tensor, Kijk and Gijk are the linear and
nonlinear magneto-optical tensors, respectively. Repeated indices are used
as the summation convention. The symmetries required by the Onsager
relation give the following conditions:
(0) (0)
εij (M) =εij (−M) = ε0 δij ,
Kijk = −Kjik , Kiik = 0, i 6= j 6= k, (3.16)
Gijkl = Gjikl = Gjilk Gijlk .
In a cubic crystal, the number of components of K and G can be further

reduced to
Kijk = K,
Giiii = G11 , Giijj = G12 , (3.17)
G1212 = G1313 = G2323 = G44 ,
where δij is the Kronecker delta function. With in-plane magnetization,
the off-diagonal εij are given as follows:
εxy = εyx = 2G44 ML MT ,
εxz = −εzx = KML , (3.18)
εyz = −εzy = −KMT ,
for α 6= 0,
εxy = εyx = [2G44 + Gmap (1 − cos(4α))] ML MT − Gmap sin(4α) ML2 − MT2 ,

4
εxz = −εzx = KML ,
εyz = −εzy = −KMT ,
(3.19)
where Gmap = G11 −G12 −2G44 and is called the magneto-optical anisotropy
parameter. Using the well established relations between Kerr rotations for
the s polarization (the electric field of the incident light is parallel to the
velocity of the incident light) and the dielectric tensor εij , the Kerr s effect
Θs can be expressed as:
K2

Gmap K
Θs = As G44 + (1 − cos(4α)) + ML MT −As Gmap sin(4α)(ML2 −MT2 )−Bs ,
2 ε0 KL
(3.20)
where As and Bs are the weighting optical factors and are an even and odd
functions of angle ϕ (Fig.3.34). A similar equation is for the Kerr p effect
Θp with sign changes on A and B. Both these constants depend on the
incident angle ϕ.
The QMOKE measurements on the Co2 FeSi film at α = –22.5◦ were
carried out by Hamrle et al. (2007). The sample was a 21 nm Co2 FeSi
film on the MgO (100) substrate. A red laser of wavelength 670 nm with
s polarization was used in the experiments. The laser spot diameter was
approximately 300 µm. To determine the saturation signal of QMOKE,
the Kerr angle was measured after the external field Hi was applied in the
eight different directions (Fig. 3.34) with strengths sufficient to saturate
the magnetization. These measurements can give rotations associated with
ML MT and ML2 − MT2 terms. In Fig. 3.35, QMOKE rotations with respect
to the external field H8 k y and saturation are shown. The filled circle at
Huesler alloys 125
H = 0 is the signal of QMOKE at saturation. The linear part arises from

the stray field of the magnet on the optical components of the experimen-
tal setup. As the field increases from saturation, the signal of QMOKE
decreases to zero as indicated by the arrows, one for the field in positive
and the other in negative directions. At that point, the average values of
ML MT , < ML MT >, and ML2 − MT2 , < ML2 − MT2 >, over the laser spot
are equal to zero simultaneously. This indicates the presence of magnetic
domains during the reversal process. Otherwise the two quantities can not
be equal to zero simultaneously. Furthermore, the height of the QMOKE
signal peaks reaches 30 mdeg indicating that there is a large second order
or even higher order spin-orbit coupling in the sample.
Fig. 3.35 QMOKE Kerr rotation loops in a 21 nm Co2 FeSi film with ϕ = 0.5◦ and α
= –22.5◦ . The external magnetic field is in H8 k y direction. The filled circle at H = 0
is the QMOKE signal in saturation as determined by the 8-directional method for α =
–22.5◦ (Hamrle et al., 2007).
Spin-resolved photoemission spectroscopy (SRPES) Schneider

et al. (2006) have also carried out the experiments of SRPES. The ba-
sic principle of SRPES has been discussed in Chapter 2. In this section, we
discuss some practical aspects. The film samples were placed in an ultra-
high vacuum (UHV) with base pressure to be below 10 × 10−10 mbar. The
Al capping layer on top of the film was removed by sputtering with Ar+
ions of energy 500 eV. The surface was then subjected to repeated cycles of
sputtering and prolonged annealing at 570 K. The films were magnetized
by an external in-plane magnetic field. The experiments were carried out
at 300 K.
The incident light was an s-polarized 4-th harmonic narrow band with
energy centered at 5.9 eV in the form of a pulse generated by Ti:sapphire
oscillator. The light was incident on the sample at 45◦ angle with respect
to the surface normal.

The photoemitted electrons were analyzed in a normal emission geome-
try by a commercial cylindrical sector analyzer equipped with an additional
spin detector based on spin-polarized LEED. The energy resolution and ac-
ceptance angle of the analyzer are 150 meV and ±13◦ , respectively.
The SRPES spectra for films grown on Al2 O3 and MgO substrates are
shown in Fig. 3.36. For comparison, shown in Fig. 3.37 is the bulk DOS
of Co2 FeSi from the LDA+U band structure calculations (Wurmehl et al.,
2005). The SRPES spectra show relatively smooth variation near EF and
broad asymmetric peaks centered at 1.3 eV in both spin channels.
The theory shows a wide peak in the ↓ spin channel near –1.0 eV.
There is some agreement between SPRES spectra and the calculated DOS
for minority-spin states in the energy range between –1.5 eV to EF (EF is
set to be zero).
Fig. 3.36 SRPES spectra of (a) Co2 FeSi/Al2 O3 and (b) Co2 FeSi/MgO (Schneider et al.,
2006).
Fig. 3.37 Calculated DOS of bulk Co2 FeSi (Wurmehl et al., 2005).
Experimentally, the majority-spin states have a slightly higher intensity

Huesler alloys 127
over the whole energy range of measurements for both films. Therefore, the
spin polarization is positive. The maximum spin polarization is 4% for the
Co2 FeSi/Al2 O3 film and 12% for the Co2 FeSi/MgO film. P (EF ) is 0 for
the former and 6% for the latter. The reduction of the spin polarization is
not confined to the top surface layer. It extends 4-6 nm into the sample
for the Co2 FeSi/MgO film. Compared to XAS and XMCD, which measure
the integrated spectra, SRPES probes states near EF in a very thin region
of the surface. These spectra should therefore be more sensitive to disorder
in the surface region.
Magnetization Schneider et al. (2006) measured the saturation mag-

netization for Co2 FeSi films grown on MgO and Al2 O3 substrates. The
extrapolated value to T = 0 K is 5.0 µB /formula-unit for those films on
Al2 O3 . There is a uniaxial anisotropy in films grown on Al2 O3 with the
easy axis along the [110] direction. Films deposited on the MgO substrate
show no significant anisotropy.
There is a T3/2 dependence in M (T) for Co2 FeSi grown on MgO. The
result is shown in Fig. 3.38. This dependence indicates spin wave excitations
in these films.
Fig. 3.38 M(T) as a function of temperature for a Co2 FeSi film grown on the MgO
substrate (Schneider et al., 2006).
Curie temperature As we have mentioned previously, Co2 FeSi has the

highest TC among all Heusler alloys. Wurmehl et al. (2005) measured
the high temperature magnetization of Co2 FeSi using the vibrating sample
magnetometer (VSM) method (Fig. 3.39). The measurements were carried
out under a magnetic field of 1 kG. The quantity determined is the inverse
of the susceptibility 1/χ(T) for T > TC . This is based on the Curie-Weiss
law:
C
χ(T ) = . (3.21)
T − TC
The value of TC is 1100±20 K (Wurmehl et al., 2005).
Fig. 3.39 A schematic setup of VSM. The sample coil senses the signal proportional to
χ. The cup transmits the sound wave to the sample (Foner, 1956).
3.6.3.4 Transport properties

Temperature dependence of resistivity The van der Pauw method,
four-probe scheme, was used by Schneider et al. (2006) to measure the
resistivity ρ. The temperature range was 0 < T < 300 K. Results for a
typical film of Co2 FeSi are shown in Fig. 3.40. The residual resistivity ρo
defined at T = 4.0 K is 40 µΩcm. The corresponding ratio between the
resistance at 300 K and 4.0 K is 1.5. These values fall in the range of other
Heusler alloys in thin-film form. The key feature of ρ(T) is the exponent.
The resistivity curve shown in Fig. 3.40 can be fit to:
• ρ(T) ∼ T7/2 , T < 70 K.
• (ρ − ρo ) ∼ T1.65 , T > 70 K.
These results do not indicate that the dominant scattering mechanism
is the one-magnon process. Other processes, such as incoherent scattering,
and the scattering between s- and d-states, are possible causes of exponents
deviating from 2.0. It is also possible to get better agreement with a one-
magnon process if the shape of the Fermi surface is taken into account.
Magnetoresistance This quantity measures the current-voltage re-

sponse of a sample under an external magnetic field H. The resistance
can have two components:
Huesler alloys 129
Fig. 3.40 Resistivity as a function of temperature in a typical Co2 FeSi film (Schneider
et al., 2006).
• ρk , the resistivity parallel to the magnetic field,

• ρ⊥ , the resistivity perpendicular to the magnetic field.
Fig. 3.41 Magnetoresistivity of a thin-film Co2 FeSi at 4 K (Schneider et al., 2006).
The measured results by Schneider et al. (2006) are shown in Fig. 3.41.
ρ⊥ (the top curve) is larger than ρk (the bottom curve). This is attributed
to the spin-orbit interaction. At H = 0, the spontaneous anisotropy can be
determined from the expression,

ρk − ρ⊥
1
. (3.22)
3 ρk + 2ρ⊥
The ratio is found to be –0.8%, which is considered small. At large field

(H > 15 kG), both ρ’s show linear behavior with different slopes which
are functions of temperature. Similar behaviors have been observed in
conventional ferromagnets. The resistivity has been attributed to spin-flip
scattering processes.
3.6.3.5 Tunnel magnetoresistance (TMR)

The TMR of Co2 FeSi films grown on MgO has been measured with the
film as one of the electrodes (Niculescu et al., 1977). At RT, TMR is 41%.
The value increases to 60% at 5K when the film grown at 473 K was post
annealed at 573 K. The results show that TMR is highly sensitive to the
sample preparation.
Chapter 4
Half metallic oxides
4.1 Introduction
Among the various oxides, a few are now known to show half metallic
properties. They are:
• Chromium dioxide (CrO2 )

• Magnetite (Fe3 O4 )
• Manganites, La2/3 Sr1/3 MnO3 (LSMO) and La1−x Bax MnO.
The last of these emerged from the study of high TC superconductors. All
three are appealing for spintronic applications because they are predicted
to have 100% spin polarization at EF . Experimentally, CrO2 exhibits P up
to 97% at 4 K (Ji et al., 2001). These materials are expected to show large
tunnel magnetoresistance (TMR) when used to form electrodes sandwiching
a thin insulating region, and so are good candidates for spintronic devices
such as spin valves and sensors.
The above three oxides are of comparable complexity to the Heusler
alloys with respect to growth and structure. In this chapter, we shall discuss
each one separately, considering the growth, characterization, and magnetic
and electronic properties for each in detail.
4.2 CrO2
Experimental studies of CrO2 go back to the late 1960’s (Stoffel, 1969).

Schwarz (1986) was the first to predict CrO2 to be a half metallic ferro-
magnet. This study was motivated by the potential for using this material
to make read-write magneto-optical memory devices.
131
4.2.1 Structure
CrO2 has the simplest structure among the three oxides discussed above.
It crystallizes in the rutile (TiO2 ) structure (Fig. 1.3).
The space group is D14 4h ( P42 /mnm), which is a nonsymmorphic group.
There are two formula units per unit cell. The Cr atoms form a body-
centered tetragonal lattice, as shown in Fig. 1.3. Each of the Cr atom is
surrounded by an octahedron of O atoms. The orientations of the two oc-
tahedra differ by a 90◦ rotation about the z-axis (parallel to the shorter
lattice vector, c). The positions of the Cr and O atoms are given in Ta-
ble 4.1. Denoting the longer lattice vector by a in the x-y plane, the c/a
ratio is 0.65958 (Schwarz, 1986). The lattice constant, a, is determined by
Thamer et al. (1957) to be 4.42 Å. Later, Porta et al. (1972) confirmed
these results.
Table 4.1 Positions of Cr and
O atoms in rutile structure ex-
pressed in reduced coordinates,
where δ is 0.3053.
Atom Coordinates (x,y,z)
Cr (0,0,0)
Cr (0.5, 0.5, 0.5)
O (δ, δ, 0)
O (-δ, -δ, 0)
O (0.5+δ, 0.5-δ, 0.5)
O (0.5-δ, 0.5+δ, 0.5)
4.2.2 Growth
Cr forms many competing oxides, namely Cr3 O4 , Cr2 O5 , CrO3 , and Cr2 O3 .
Crystalline CrO2 is the only oxide which is ferromagnetic at room tem-
perature (RT). But it is metastable at atmospheric pressure and is easily
decomposed to Cr2 O3 with heat. Therefore, it can be difficult to grow.
The successful growth techniques can be classified into two categories: the
“high-pressure” and “low-pressure” techniques. Thermal decomposition
(Ranno et al., 1997; Hwang and Cheong, 1997) is the high-pressure tech-
nique. Chemical vapor deposition (CVD) (Li et al., 1999; Ivanov et al.,
2001; Kämper et al., 1987; Gupta et al., 2000; Anguelouch et al., 2001) is
the low-pressure technique. More recently, pulsed laser deposition (PLD)
has been used to grow CrO2 (Shima et al., 2002). PLD is a low-pressure
technique and has been discussed in Chapter 2. In the following, we shall
Half metallic oxides 133
focus on the thermal decomposition and CVD techniques.
4.2.2.1 Thermal decomposition

This technique was successfully used by Ranno et al. (1997) to grow powder
and film samples of CrO2 . For the powder growth, CrO3 was decomposed
in a closed vessel with different oxygen pressures under controlled temper-
ature. The reaction can be described by the chemical formulas:
1
CrO3 → CrO2 + O2 (4.1)
2
and
3
2CrO3 → Cr2 O3 + O2 . (4.2)
2
One must bear in mind that the end product can be Cr2 O3 . Ranno et al.
(1997) used the thermopiezic analysis (TPA) to record the decomposition
temperature and oxygen release. An instructive result of the TPA shows
the range of pressure and temperature for obtaining CrO2 from a powdered
sample of CrO3 as given in Fig. 4.1. CrO3 decomposes at 250 ◦ C at ambient
pressure. Between 300 and 400 ◦ C is the region forming the mixture of CrO2
and Cr2 O5 with a trace of Cr2 O3 . CrO2 powder decomposes at 450 ◦ C.
To grow thin-film forms of CrO2 , a single crystal of TiO2 or Al2 O3 having
an area of a few mm2 and thickness of 0.5 mm serves as the substrate
and powdered CrO3 as the precursor in a brass reactor with Al seal. The
reactor was put into a furnace and was heated at 0.5 ◦ C/min from RT.
Before bringing back to RT at a rapid rate, the reactor was held at 425 ◦ C
for about an hour.
Fig. 4.1 The oxygen pressure as a function of temperature recorded by TPA for an 18
mg CrO3 sample under heating at 0.5 ◦ C/min (Ranno et al., 1997).
4.2.2.2 Chemical vapor deposition (CVD)

The methods used to grow CrO2 using CVD differ from one another mainly
in precursor. Most have used the decomposition of CrO3 in a closed reactor
onto heated substrates of RuO2 , TiO2 , or Al2 O3 . In a typical case, the
reaction tube has a two-zone furnace (Li et al., 1999). Powdered CrO3 was
loaded in the first zone and kept at 260 ◦ C. Oxygen carriers flow from the
first to the second zone at the rate of 0.5 cc/min at atmospheric pressure.
The substrate was in the second zone and was heated to 390–400 ◦ C.
The grown films were put into an ultra high vacuum (UHV) chamber
and the surfaces were cleaned by soft sputter cycles of 30 seconds with a
defocused 500 eV Ne ion beam under almost grazing angle of incidence.
This method was used because the surface of CrO2 can easily transform
into Cr2 O3 at 200 ◦ C. The samples are in polycrystalline form.
Gupta et al. (2000) grew finer-grain polycrystalline CrO2 films on a seed
layer of TiO2 . The samples grown on Al2 O3 are highly oriented due to the
six-fold symmetry of the substrate with multiple grains. A year earlier,
Gupta et al. (1999) reported the growth of CrO2 thin films by CVD with
selective-area and lateral overgrowth scheme. The substrate was a single
crystal TiO2 . The selected-area and lateral overgrowth was accomplished
using a prepatterned SiO2 mask.
Anguelouch et al. (2001) used a liquid precursor CrO2 Cl2 . It was placed
in a bubbler and kept at 0 ◦ C. Their reactor was a quartz tube inside a
furnace. The tube was kept at 400 ◦ C. The substrate was TiO2 oriented
in the (100) plane. It was placed on a tilted glass susceptor after the use
of organic solvents and a dilute hydrofluoric acid to clean the tube. With
oxygen gas flowing at 40 cc/min, these authors grew thin-film forms of
CrO2 at a deposition rate of 72 Å/min.
Ivanov et al. (2001) provided some insight into the growth of CrO2 from
CrO. They realized that in the CVD technique the oxygen gas serves as the
mechanical carrier of the CrO3 and is not crucial to the growth. They tried
to use MBE with a CrO3 beam but failed to get any CrO2 . The work of
Norby et al. (1991) identified that one of the intermediates is Cr8 O21 . This
led Ivanov et al. (2001) to prepare Cr8 O21 as the sole precursor by heating
powdered CrO3 to 250 ◦ C for a period of eight hours. With this precursors,
they finally grew high quality epitaxial CrO2 layers on Al2 O3 (0001) and
TiO2 (110) substrates. They also pointed out that further heating of CrO3
to 330 ◦ C for twelve hours produced Cr2 O5 . This sample cannot be used
as a precursor to grow CrO2 . Therefore, one must be extremely careful to
control the temperature in growing CrO2 .
4.2.3 Characterization
There are three techniques to characterize the quality of CrO2 in thin-film
form: RHEED (Li et al., 1999), X-ray diffraction (Gupta et al., 1999), and
AFM (Anguelouch et al., 2001).
4.2.3.1 Reflection high energy electron diffraction (RHEED)

Li et al. (1999) aligned the electron beam along orthogonal zone axes and
determined the RHEED pattern ex situ for TiO2 and Al2 O3 substrates,
respectively. The results are shown schematically in Figs. 4.2(a) and 4.2(b).
The patterns are sharp and streaky on TiO2 , suggesting the surface is well-
ordered and flat. The pattern shown in Fig. 4.2(c) is formed by small spots
even though it is still sharp. The pattern in Fig. 4.2(d) shows hexagonal
spots. These patterns suggest the films were grown in three dimension. The
patterns for different beam directions in the growth on a TiO2 substrate
show that the film axes are aligned with the axes of the substrate. However,
the beam directions indicate there are multiple domains with the b-axis of
CrO2 aligned along the equivalent directions of < 112̄0 > on Al2 O3 .
Fig. 4.2 The RHEED patters obtained by Li et al. (1999). (a) On TiO2 (100), beam
< 001 >, (b) on TiO2 (100), beam < 110 >, (c) on Al2 O3 (0001), beam < 11̄00 >, (d) on
Al2 O3 < 0001 >, beam < 112̄0 >.
4.2.3.2 X-ray diffraction

By using the selective-area and lateral growth techniques, Gupta et al.
(1999) obtained epitaxial growth of CrO2 . Three surfaces of the substrate
are used, (100), (110), and (001). The normal X-ray intensity as a func-
tion of 2θ of those three cases is given in Fig. 4.3. The first peak in each
case is associated with TiO2 and the second peak with CrO2 , because the
lattice constants of the sample and substrate are different. The lines for
TiO2 are very narrow for all three directions, and are reproducible, indicat-
ing that substrate surfaces were all properly cleaned. The lattice constant
mismatches are large in the three orientations. For (100) and (010) orien-
tations, it is 3.79%, and for (001) it is 1.48%. Therefore, there is a tensile
strain in the in-plane direction and a compressive strain in the perpendic-
ular direction. The peaks in case (c) show the largest shift to the higher
angle. The largest full width at half maximum for the rocking curve1 of the
second peak in case (b) is 0.4◦ , which is considered to be narrow.
Fig. 4.3 X-ray intensity as a function of 2θ (◦ ) with the substrate surfaces: (a) (200),
(b) (220), and (c) (002) (Gupta et al., 1999).
4.2.3.3 Atomic force microscopy (AFM)

The AFM image of 1050 Å CrO2 sample obtained by Anguelouch et al.
(2001) is shown in Fig. 4.4. The distinguishable lines correspond to atomic
planes of CrO2 . The RMS roughness in an area of 1.0 µm2 is 4.6 Å. The
height of each step is either 4.4 Å or 8.8 Å. These steps correspond to one
or two planes height. The films are atomically smooth.
4.2.4 Transport properties

Since applications of magnetoresistance (MR) are common for many oxides
besides CrO2 and a technique— the point contact method —has been de-
1 The rocking curve provides information about compositional and thickness variations.
In addition, lattice constant difference between a substrate and a sample can be deter-
mined within 10 parts per million from the peak separation and the ratio of integrated
intensities of the two peaks of the rocking curve.
Fig. 4.4 AFM image of a 1050 Å CrO2 sample. The distinguishable lines correspond to
the atomic planes of CrO2 . The inset shows a sample cross section of the region marked
A with 4.4 Å steps (Anguelouch et al., 2001).
veloped for powdered forms of oxides, we shall discuss first this technique
and the associated powder magnetoresistance (PMR). We are particularly
interested in the half metallic properties of CrO2 . The experimental de-
termination of these properties will be discussed. Finally, the electronic
properties of this compound will be discussed based on photoemission ex-
periments and theoretical grounds.
4.2.4.1 Point contacts and powder magnetoresistance

Many oxides are grown in powdered form. One of the effective experimental
techniques for determining the PMR of the oxides was developed by Coey
et al. (2002) is the point contact method:
• Definition: A point contact is a small region of atomic scale continuity

between two crystallites (Fig. 4.5). It is possible to have a grain bound-
ary at the contact. This last property differs from a nanoconstriction
in which there is crystallographic coherence and no grain boundary.
• Both the point contact and nanoconstriction can have domain walls
which can be pinned at the contact.
• The tunneling process to get magnetoresistance is through the point
contacts.
Let δ be the wall width and λ be the mean free path (MFP). If λ >
δ, then the mechanism for the transport is ballistic. The transmission
coefficient Tb for an oxide HM under low bias and zero temperature (Coey
et al., 2002) is
θ12
Tb = cos2 ( ), (4.3)
2
where θ12 is the angle of misalignment of the magnetization in the domains

on each side of the contact. For λ ≪ δ , the transport is diffusive and the
transmission coefficient Td (Coey et al., 2002) is
1 θ12 ν−1
Td = [cos( )] , (4.4)
2 2ν
where ν = δ/λ. The magnetoresistance is
M R ∼ (1 − Tb/d ). (4.5)
Note that θ12 = π gives the largest MR.

R(0) − R(H) G(H) − G(0)
MR = = , (4.6)
R(0) + R(H) G(H) + G(0)
where G is the conductance. If the samples are pressed powder and mag-
netic moments in the two domains are not completely aligned along the
external magnetic field, then
m2
PMR = , (4.7)
(1 + m2 )
where m2 =< cosθ12 >2 .
Fig. 4.5 Schematic of a point contact (Coey et al., 2002). The shaded triangular areas
are two magnetic domains. Their magnetizations are oriented relative to each other by
θ12 (not indicated). The dashed lines with arrows indicate a path of ballistic transport
while the solid lines show the diffusive transport from the left to the right.
Setup The setup for point contact magnetoresistance measurement is

shown in Fig. 4.6. The separation between the sample (M1 ) and tip (M2 )
is controlled by screws. A modified version uses a piezo system to control
the separation. Magnetic field is applied by a pair of Helmholtz coils (1
and 2).
Fig. 4.6 Basic schematic setup of point contact magnetoresistance measurement (Garcı́a
et al., 1999). M1 and M2 form the point contact. The battery connected to M2 applies
the bias voltage.
Measured results The PMRs of CrO2 for a pressed powder and for a
powder diluted with 75 wt% insulating Cr2 O3 particles with similar shape
were measured as a function of temperature. The results are shown in
Fig. 4.7. At 5 K, PMR is approximately 50% for the diluted powder com-
pact (Coey et al., 2002). The value of MR for undiluted CrO2 compact
(black squares) is approximately 30%. The corresponding polarizations are
62% and 82%, respectively. However it is only 0.1% at RT (Coey et al.,
2002).
Fig. 4.7 PMR results for CrO2 (black Fig. 4.8 Magnetization (solid curve)
squares) and 25% powder compact of and MR (dashed curve) at 5 K as func-
CrO2 (open circles) (Coey et al., 2002). tion of fields up to 60 kGauss (kG) for
a CrO2 powder compact (Coey et al.,
2002).
4.2.4.2 Magnetization and magnetoresistance vs. applied external

field
The response of powder CrO2 under a dc magnetic field up to 60 kG in
terms of magnetization and MR was measured by Coey et al. (2002). The
magnetization was measured by a SQUID magnetometer. These results
at 5 K are shown in Fig. 4.8. The MR exhibits a butterfly pattern. The
magnetization shows a hysteresis loop with coercivity 0.99 kG. At RT, it
is 0.59 kG. The mass magnetization at saturation Ms is approximately 115
emu/g.
4.2.5 Half metallic properties

While measuring the magnetoresistance, it is possible to deduce the polar-
ization P. For large P, it is possible that the sample is a HM. To confirm
this, however, it is necessary to carry out Andreev reflection or other more
direct measurement techniques.
4.2.5.1 Point-contact Andreev reflection (PCAR)

The basic principle of the Andreev reflection was discussed in Chapter 2.
To study the reflection between a HM and a superconductor (SC), Soulen
et al. (1998) developed the point contact scheme which was discussed in
section 4.2.4.1.
Anguelouch et al. (2001) used the point-contact Andreev reflection
(PCAR) to analyze CrO2 . Lead (Pb) is used as SC and the sample CrO2
is in film form. The advantages of using PCAR over the use of tunneling
junctions and nanolithography are two-fold:
• The measurements are not fabrication dependent.

• The measured results are independent of the characteristics of the in-
terface between the sample and superconductor.
Tip and Probes The tips were obtained from wires with 0.03 inch in
diameter (Ji et al., 2001), treated first by mechanical polishing and then
electrochemical etching. Since the conductance as a function of voltage was
to be determined, a conventional four-probe scheme (e.g., Fig. 2.20) was
used.
Setup The setup is shown in Fig. 2.33. The tip was controlled by a
differential screw mechanism for microscopic movements toward the sample
surface. To keep the measurements at low temperature, the whole setup

was encased in a vacuum jacket. The jacket was immersed in a liquid helium
bath. The temperature at the measurements was 1.85 K.
What is measured The contact resistances or conductances dI/dV were

measured by standard ac lock-in techniques.
Anguelouch et al. (2001) analyzed their data and extracted the polar-
ization by using the modified model of Strijkers et al. (2001), originally
proposed by Blonder et al. (1982)—the BTK model for understanding the
conductance as a function of voltage measured from a point contact of
nonmagnetic-metal (NM)/superconductor (SC).
Ji et al. (2001) reported their experimental results for G(V )/Gn for the
case Pc = 96% at T = 1.85 K (Fig. 4.9) which agree well with the calculated
results using the same parameters. At very low temperature, CrO2 shows
close to 100% polarization.
Fig. 4.9 Measured normalized conductance of Pb/CrO2 (open circles) as a function of

V for Z=0, Pc = 0.96, ∆=1.12 meV, R=4.0Ω, and T=1.85 K (Ji et al., 2001). The solid
line is a fit to the BTK model using the same parameters.
4.2.6 Electronic properties

In this section, we are concerned mainly with photoemission spectra. We
shall start with the basic principles of conventional angular resolved pho-
toemission spectroscopy (ARPES), then we shall discuss the spin-polarized
photoemission, the measurements and setups, and finally the results for
CrO2 .
4.2.6.1 Spin-polarized angular resolved photoemission

spectroscopy (ARPES)
The quantity measured by a photoemission measurement is the energy dis-
tribution curves (EDCs): The intensity of the emitted electrons at certain
angles (θ and ϕ) and certain photon energy as a function of binding energy
of electrons with respect to EF (see 2.4.3.1 for more details). A spin- and
energy-resolved photoemission study of polycrystalline CrO2 films was car-
ried out by Kämper et al. in 1987. The light source was a He gas discharge
lamp with photon energy hν = 21.2 eV. The films were magnetized in the
film plane using a coil with pulse field of about 500 Oe. The emitted elec-
trons were detected normal to the film plane by means of a 180◦ spherical
energy analyzer with energy and angle resolution of 100 meV and ±3◦ , re-
spectively. In Fig. 4.10(a), the photoemission EDC of polycrystalline CrO2
films measured at 300 K for three different sputter cleaning times (tsp ) are
given (Kämper et al., 1987). The sputter cleaning process is to change the
surface condition.
The most crucial information provided by the measurements is the ap-
pearance of the shoulder around 2 eV for tsp = 120 s. It persists up to tsp =
335 s. Figure 4.10(b) shows the corresponding spin-polarization spectrum
for tsp = 120 s. Although the near 100% spin polarization was observed
near 2 eV below EF , they did not observe a large spin polarization near
EF .
Fig. 4.10 (a) Photoemission EDC of polycrystalline CrO2 films measured at 300 K for
different times of sputter cleaning. (b) Spin-polarized spectrum of polycrystalline CrO2
films at 300 K after tsp = 120 s (Kämper et al., 1987).
Figure 4.11 shows measured spin-polarized EDC at a particular k-point

of CrO2 with (100) surface by Dedkov (2004) for different tsp ’s. The exper-
iments were carried out at RT. The important features of the results are
summarized as follows:
• After tsp = 210 s, a shoulder developed at energy 2.3 eV below EF
for the total EDC. This feature agrees well with the earlier results of
Kämper et al. (1987). A similar feature is exhibited in the ↑ spin chan-
nel. The ↓ spin states do not show any such structure. The polarization
at EF is 95%.
• After tsp = 750 s, the shoulder of the total EDC is shifted to 2.0 eV
and its strength is increased.
• After tsp = 750 s and additional annealing for 12 hours at 150 ◦ C, the
shoulder feature persists. The polarization at EF is 85%.
Fig. 4.11 Photoemission EDC of CrO2 (100) surface for different sputter cleaning times
(Dedkov, 2004). The energy scale is referenced to EF .
In summary, the experimental results are interpreted as follows:

• The –2.3 eV shoulder is primarily from the 3d-states.
• The difference of the total EDC at tsp = 210 s and tsp = 750 s is due
to the sputtering process which can either increase disorder near the
surface of the sample or reduce d-p hybridization between the Cr and
O atoms.
• P ∼ 85% in for tsp = 750 s is due to the annealing process which
recovered the quality of the surface and removed contaminants.
4.2.7 Magnetic properties

4.2.7.1 Curie temperature
CrO2 is a ferromagnetic material. Most of the studies of the magnetic prop-
erties have used samples in thin-film form. Kuneš et al. (2002) measured
temperature dependence of the in-plane and out-of-plane magnetizations in
CrO2 film grown by CVD on Al2 O3 (0001) substrates. The easy axis is in
the plane of the film. Both in-plane and out-of-plane magnetizations van-
ish at 386 K, as shown in Fig. 4.12(a), which led these authors to conclude
that TC is 385 K. An earlier experiment (Kouvel and Rodbell, 1967) on
compressed powdered stoichiometric CrO2 reported a TC of 386.5 K. These
two results agree well with each other.
Fig. 4.12 (a) M(T) as a function of T for thin-film form of CrO2 on Al2 O3 (0001)
substrates (Kuneš et al., 2002). (b) Hysteresis loops of CrO2 (100) films epitaxially
grown on a TiO2 (100) substrate with at T = 10 K (dots) and 300 K (open squares)
(Spinu et al., 2000).
4.2.7.2 Saturation magnetic moment

The saturation magnetization for a thin film of CrO2 at T = 10 K was
found to be 600 emu/cm3 , and it is 450 emu/cm3 at T = 300 K (Spinu
et al., 2000).
4.2.7.3 Hysteresis loops

The hysteresis loops of CrO2 (100) films epitaxially grown on a TiO2 (100)
substrate were measured by a SQUID magnetometer with magnetic field
applied along the easy axis of the films (Spinu et al., 2000). The results are
shown in Fig. 4.12(b).
As shown later in the band structure, because EF does not intersect the
d-bands, the magnetism cannot be of Slater-Stoner-Wohlfarth type. The
localized moments must be aligned by some kind of inter-atomic exchange
interaction. One possibility is the superexchange through the oxygen atom.
4.2.8 Theoretical studies of electronic and magnetic prop-

erties
4.2.8.1 Electronic properties
Band structure and density of states The band structure of CrO2
has been calculated by many authors using first-principles methods
based on DFT. Schwarz (1986) was the first to use the spin-polarized
augmented spherical wave method of Williams et al. (1979). Later,
LSDA/Linear-Combination-of-Gaussian-Orbitals (LCGO) (Brener et al.,
2000), GGA/LAPW (Mazin et al., 1999), and LSDA+U/LMTO (Korotin
et al., 1998) methods were applied. As an example, the ↑ (majority)
and ↓ (minority) spin band structures calculated with the GGA exchange-
correlation by Brener et al. (2000) are shown in Fig. 4.13.
Fig. 4.13 Band structure of CrO2 (Brener et al., 2000). (a) ↑ (majority) spin states;
(b) ↓ (minority) spin states.
All these calculations show that CrO2 is a HM with a magnetic moment

of 2.0 µB /formula-unit. Other properties, such as the insulating gap in the
minority-spin channel and DOS of the majority-spin channel at EF , are
listed in Table 4.2. Shown in Fig. 4.14 are the density of states of the two
channels based on the GGA exchange-correlation functional (Brener et al.,
2000).
Table 4.2 Comparison of theoretical density of states.

DOS (states/eV-cell) at EF Gap (eV) Reference
Majority-spin states Minority-spin states
∼0.5 (a minimum) 1.5 Schwarz (1986)
2.33 (close to a minimum) 1.34 Brener et al. (2000)
0.95 (a minimum) 1.3 Mazin et al. (1999)
∼0.3 (a minimum) 1.5 Korotin et al. (1998)
Fig. 4.14 Spin-projected densities of states of CrO2 (Brener et al., 2000).
Qualitatively, the three results given in Table 4.2 show a minimum in the
DOS at EF of the majority-spin channel. The presence of a minimum agrees
reasonably well with the photoemission spectra. However quantitatively,
the amplitude of the density of states differs by nearly a factor of eight.
Whereas the agreement of the gap values for minority-spin states is much
better.
All the theoretical calculations agree that the d-states of the Cr split
into a triply (t2g ) and a doubly (eg ) degenerate sets of states under the crys-
tal field of the O atoms. The bonding properties of CrO2 show both ionic
character and d-p hybridization of more covalent character. Korotin et al.
(1998) show that it is the eg states originating from the Cr atom which con-
tribute to the hybridization. The crystal field and hybridization combined
with the exchange interaction gives rise to the half metallic properties.
Fermi surface The Fermi surface of CrO2 crystal is rather complicated.

The results of Brener et al. (2000) are shown in Fig. 4.15 in different sec-
tions of the Brillouin zone (BZ). There is a roughly cubical electron surface
around the Γ point with hole surfaces located between the electron surface
and kz zone boundaries. The points of hole surfaces closest to the Γ point
touch the electron surface. These results differ from Mazin et al. (1999) in
two aspects:
• The hole surface determined by Mazin et al. (1999) touches the A-Z
plane and forms hammerheads
• Mazin et al. (1999) find also needle shaped surfaces.
Hubbard U The Hubbard-U value obtained by Korotin et al. (1998)

is 3.0 eV. The effect of U is to shift the unoccupied minority-spin states
Fig. 4.15 The BZ and Fermi surfaces in the sections of ΓZAM and ΓZRX (Brener
et al., 2000). The grey region is the electron surface while the dashed regions are the
hole surfaces.
with respect to the LSDA results and to make the dip in the DOS at EF
of majority-spin channel more pronounced. Given these effects, it is still
controversial whether it is necessary and appropriate to determine the band
structure of CrO2 with LSDA + U.

Magnetic moment The value of the magnetic moment was first ex-
plained by Schwarz (1986) using an ionic model—two oxygen atoms take
a total of 4 electrons from the Cr atoms. This leaves two electrons at the
Cr site. Hund’s first rule requires the spin moments of the two electrons to
be parallel. The spin angular momentum of an electron along a given axis
is ~/2. With the g-factor of 2, the magnetic moment in the cell is then 2.0
µB . Brener et al. (2000) object to the ionic model based on the calculations
by Schwarz (1986)—the total number of electrons within the Wigner-Seitz
sphere of the Cr atom being 4.2 rather than 2.0. The alternative explana-
tion proposed by Brener et al. (2000) is that the ↑ and ↓ d-p hybridized
states are fully occupied so that they do not contribute to the magnetic
properties of the crystal. The remaining two electrons at the Cr site give
the 2 µB /unit-cell.
Mechanism of ferromagnetism The band structure obtained by Ko-

rotin et al. (1998) shows that the dyz and dzx orbitals of the Cr atom
hybridize with p-states of the O atoms. These hybridized states are mobile
and can move into the region where the dxy orbital is localized. Hund’s first
rule requires the three electron spins to be aligned, resulting in the ferro-
magnetic coupling between the localized states. The mechanism is similar
to Zener’s double exchange mechanism (Zener, 1951). More discussion will

be given in Section 4.5.2.
4.3 Fe3 O4
Fe3 O4 is sometimes called magnetite. It is a well-known magnetic material

with a TC of 858 K (van Dijken et al., 2004). It has a more complicated
crystalline structure than CrO2 . The Fe ions have two different valences,
Fe2+ and Fe3+ . In addition, it exhibits a Verwey transition (Verwey, 1939):
A sudden decrease of conductivity at TV = 120 K, which is attributed to
an order-disorder transition of Fe2+ and Fe3+ on the B sublattice and the
accompanying distortion of the lattice, as we shall discuss further below.
4.3.1 Structure
The crystal structure above TV has been determined by X-ray diffraction
and its magnetic structure has been probed by neutron scattering (Shull
et al., 1951). It is the cubic spinel structure. The primitive lattice is fcc
with lattice constant of 8.396 Å. Associated with each fcc lattice point,
there are two smaller cubes as shown in Fig. 4.16. The space group of the
crystal is Fd3̄m. Each primitive fcc cell has the following configurations of
atoms:
• Fe2+ sites: The unit cell contains two Fe2+ ions, one of them is at the
origin of the unit cell—at the lower left corner (black polygon), and
the other one is at the center of one of the small cubes (open polygon).
They are at the tetrahedral (A) sites.
• Fe3+ sites: The other Fe ions are 3+ and are located at four corners of
the smaller cube (gray polygons) with a vacancy at its center. These
Fe ions are at the octahedral (B) sites.
• O sites: The O atoms occupy four corners of each small cube (dia-
monds).
In Table 4.3, we list the atom positions in reduced coordinates of the
lattice constant a. The magnetic structure of Fe3 O4 was first proposed
by Nèel (1948). All Fe atoms at B sites are ferromagnetically aligned
with each other and antiferromagnetically aligned with those at A sites.
This structure was experimentally confirmed by Shull et al. (1951).
Fe3+ ions on the A sublattice have magnetic moments of 5 µB while
Fe2+ ions have magnetic moments of 4 µB and 5 µB .
Fig. 4.16 Part of the spinel structure associated with one lattice point (lower left corner)
of the conventional fcc structure. The tetrahedral (A) sites (black and open polygons) ,
the octahedral (B) sites (gray polygons) and the oxygen atoms (diamonds) in Fe3 O4 .
Table 4.3 Atomic positions in reduced coordinates and ef-

fective charges of the Fe ions in Fe3 O4 .
Element Atomic position Effective charge (e)
Fe (0.000, 0.000, 0.000) 2+
Fe (0.250, 0.250, 0.250) 2+
Fe (0.125, 0.125, 0.625) 3+
Fe (0.375, 0.375, 0.625) 3+
Fe (0.125, 0.375, 0.875) 3+
Fe (0.375, 0.125, 0.875) 3+
O (0.375, 0.125, 0.125)
O (0.125, 0.375, 0.125)
O (0.125, 0.125, 0.375)
O (0.375, 0.375, 0.375)
O (0.375, 0.125, 0.625)
O (0.125, 0.375, 0.625 )
O (0.125, 0.125, 0.875)
O (0.375, 0.375, 0.875)
Below TV , Iizumi et al. (1982) determined the structure using X-ray

diffraction method on a single crystal of synthetic Fe3 O4 . The unit cell
determined at 10 K is monoclinic
√ √with three lattice constants along a-
, b-, and c-directions being 2a, 2a, and 2a, respectively, where a is
8.394 Å and is for the cubic phase above TV . The corresponding space
groups going from the cubic to monoclinic structure are Fd3̄m → Cc. In
2002, Wright et al. (2002) used high resolution neutron and synchrotron X-
ray powder-diffraction measurements to refine the structural information of
Fe3 O4 below TV . The refined structure information of the monoclinic unit
cell has the space group P2/c instead of Cc with lattice parameters a =
5.944 Å, b = 5.924 Å, c = 16.775 Å, and angle β = 90.2363◦. The unit cell
is shown in Fig. 4.17. The positions of the atoms at 90 K were constrained

by the operators of the orthorhombic space group Pmca and are given in
Table 4.4 with the origin at an inversion center located between two A sites.
Fig. 4.17 The unit cell (outlined in thick lines) of Fe3 O4 at temperature below TV
(Wright et al., 2002).
Table 4.4 Atomic positions of of Fe3 O4 with

Pmca symmetry at T = 90 K (Wright et al.,
2002).
Atom x (1/a) y (1/b) z (1/c)
Fe (A1 ) 0.25 0.0034 0.0637
Fe (A2 ) 0.25 0.5061 0.1887
Fe (B1a ) 0.00 0.5000 0.0000
Fe (B1b ) 0.50 0.5000 0.0000
Fe (B2a ) 0.00 0.0096 0.2500
Fe (B2b ) 0.50 0.0096 0.2500
Fe (B3 ) 0.25 0.2659 0.3801
Fe (B4 ) 0.25 0.7520 0.3766
O (1) 0.25 0.2637 -0.0023
O (2) 0.25 0.7461 -0.0029
O (3) 0.25 0.2447 0.2542
O (4) 0.24 0.7738 0.2525
O (5a) -0.0091 0.0095 0.1277
O (5b) 0.4909 0.0095 0.3723
O (6a) -0.0081 0.5046 0.1246
O (6b) 0.4919 0.5046 0.3754
4.3.2 Growth
Tiny single crystals and thin films of Fe3 O4 have been grown.
4.3.2.1 Single crystal growth

A recent report of growing 5–10 g single crystal samples was given by Mc-
Queeney et al. (2006). These authors used a radio frequency induction
melting technique to grow the samples from the powdered Fe3 O4 . Initially,
the powdered materials were in a crucible lined with a solid of the same
composition to minimize the possible contamination of the melt. After slow
cooling, the crystals were extracted from the fractured boule. The single
crystals were re-annealed under appropriate CO2 /CO atmospheres. The
chamber for the re-annealing is specially designed Shepherd and Sandberg
(1984). A schematic diagram of the annealing chamber is shown in Fig 4.18.
The chamber is made of Alumina and is partly enclosed inside a furnace.
Fig. 4.18 A schematic diagram of Fig. 4.19 A schematic of the apparatus

the annealing chamber (Shepherd and for growing thin-film Fe3 O4 (Reisinger
Sandberg, 1984). et al., 2003).
4.3.2.2 Thin-film growth

Many researchers have grown thin-film Fe3 O4 using different methods. We
comment on the PLD method and epitaxial growth.
Pulsed laser deposition Recently, Bollero et al. (2005) used PLD to

grow 26 to 320 nm thick Fe3 O4 films on MgAl2 O4 (100) substrates. The
lattice constant of the substrate is 8.083 Å. There is a –4% lattice mis-
match between the substrate and samples. The films show antiphase do-
main boundaries. A Lambda-Physik 248 nm KrF excimer laser was used.
Its repetition rate was 10 Hz and the fluence was 0.65 J/cm2 . The sub-
strate was kept at 500 ◦ C during the growth in a background of oxygen at
pressure 5.0 × 10−6 mbar.
In 2003, Reisinger et al. (2003) reported the PLD growth of Fe3 O4
on Si(111) substrates. Instead of using other oxides, such as MgO, as

substrates, this experiment used Si as the substrate. They had in mind to
integrate the magnetic properties of Fe3 O4 into Si technologies.
The growths involved several stages, namely the preparation of Si sub-
strates, pulsed laser depositions of buffer layers made of TiN and MgO to
have a better lattice constant match with the grown samples, and finally
the deposition of Fe3 O4 . A schematic diagram is shown in Fig. 4.19.
It is advisable to have in situ characterization of the substrates and
samples by RHEED during the thin-film growth. To grow the oxides, an
issue related to the in situ measurement needs to be addressed. The samples
have to be grown in high oxygen partial pressure. The oxygen gases interact
strongly with the high energy electron beams. To prevent the scattering of
the electron beam in the oxygen gas from causing misinterpretation of the
results, the authors built a RHEED system operated under pressure (10.0
Pa). In addition, it has been shown that it is more advantageous to heat
the substrate to high temperature in epitaxial growth of thin films. Ohashi
et al. (1999) introduced a heating scheme using an infrared (IR) diode laser.
With this scheme, the substrate temperature up to 1300 ◦ C was provided
by a laser beam with 95 µm wavelength and a maximum power of 100
W. The beam was focused on the backside of the substrate, in the region
painted by silver. Temperature of the substrate was measured by a high
resolution pyrometer operating at 1.5 µm rather than a commonly used
thermocouple.
The Si substrates were a commercial Si (001) single crystal wafer. Ace-
ton and isopropanol were used in an ultrasonic bath to pre-clean the sub-
strates. A possible oxide layer can be formed during the cleaning procedures
but can be removed by an in situ cleaning scheme using the IR laser.
A pulsed KrF excimer laser with wave length 248 nm and energy den-
sity 2–5 J/cm2 per shot was used to deposit TiN, MgO, and Fe3 O4 on
Si(001) surfaces. The repetition rate of the laser was between 2 and 10 Hz
depending on the materials to be deposited.
After the growth of the buffer layers, Fe3 O4 was epitaxially grown on
top of MgO. Other than the repetition rate being changed to 2 Hz, the
other conditions remained the same. The RHEED pattern indicated that
the growth was layer by layer with smooth structure.
Laser ablation Huang et al. (2002) used an ultra-violet (UV) laser to

grow epitaxially 150 monolayers (ML) of Fe3 O4 on MgO. MgO has a rock-
salt structure with lattice constant 4.21 Å, about half that of the conven-
tional cell of Fe3 O4 above TV . Before the growth, MgO was cleaved to MgO
(100) and annealed at 650 ◦ C for 1–2 hours under an oxygen pressure of
1.3×10−7 mbar. The base pressure in the growth chamber was maintained
at 1.3 ×10−10 mbar. The Fe atoms were emitted from a water-cooled ef-
fusion cell. They formed thin films of Fe3 O4 with the oxygen atoms under
pressure in the order of 6.7×10−7 mbar.
Molecular beam epitaxy (MBE) MBE growth of Fe3 O4 films ranging

from 3 to 100 nm in thickness were also reported by Eerenstein et al. (2002).
The background pressure of the growth chamber was 10−10 mbar. The
oxygen pressure was 10−6 mbar during the growth.
van der Zaag et al. (2000) used the MBE method to grow Fe3 O4 /CoO
bilayers on α-Al2 O3 (0001) and SrTiO3 (100) substrates. They used a Balz-
ers UMS630 multichamber system. The deposit rate was between 0.02–0.05
nm/s at a substrate temperature of 523 K under oxygen pressure of 3 ×10−3
mbar.
A multilayer growth of Fe3 O4 was reported by Cai et al. (1998) using a
platinum (Pt) (111) substrate. The top of the Pt(111) surface was cleaned
using bombardment of Ar+ ions with energy 1.0 keV, then annealed to T
= 1350 K under oxygen pressure of 10−7 mbar. A well-ordered surface
was obtained. After the substrate surface was prepared, a submonolayer
of Fe was grown on the surface by vapor deposition at RT and was heated
for 2–3 min around 1000 K under oxygen pressure of 1.0×10−6 mbar. At
this point, a stoichiometric bilayer of FeO(111) was obtained. Then, Fe
deposition and oxidation cycles were repeated to form Fe3 O4 . The islands
of Fe3 O4 were grown on top of the FeO bilayer. The well-known Stranski-
Krastanov (Stranski and Krastanow, 1939) growth mode —forming islands
—dominates in the growth process. Finally, the islands of Fe3 O4 coalesce
into thick layers.
4.3.3 Characterization
Most layered forms of Fe3 O4 have been characterized by RHEED, TEM,
and STM methods. Since we have discussed the growth of Fe3 O4 by PLD,
we shall discuss the characterizations based on the same work.
The characterizations discussed by Reisinger et al. (2003) involve several
stages:
4.3.3.1 Si substrate
The Si substrate was heated to 600 ◦ C, then treated at Tsub = 900 ◦ C and
1150 ◦ C to remove amorphous Si oxides on Si(100), and eventually cooled
back down to 600 ◦ C for the growth of the buffer layers. RHEED patterns
were taken at the [110] azimuth. The sequence of the patterns is shown in
Fig. 4.20.
Figure 4.20(a) shows the RHEED pattern before the heat treatment.
It shows a homogeneous background and a few spots. The background is
due to the presence of amorphous Si oxides. The spots are the constructive
interference pattern of Si(001) surfaces. As the temperature increases to
about 900 ◦ C, it exhibits repeating long and short lines (Fig. 4.20(b)),
indicating a 2×1 reconstruction, with the lines somewhat streaky. The
homogeneous background is no longer visible. As the temperature increases
to 1150 ◦ C, the pattern characterizing the 2×1 reconstruction disappears.
This suggests that the dimmer bonds forming the 2×1 reconstruction have
dissolved. When the Si substrates are cooled down to 1000 ◦ C, sharp Laue
spots are clearly visible, indicating that the oxides have been removed. The
five spots indicate that a clean 2×1 surface has been obtained.
Fig. 4.20 The RHEED patterns of Si substrates during heat treatments (Reisinger
et al., 2003). (a) Tsub = 600 ◦ C, (b) after heating to 890 ◦ C, (c) 1150 ◦ C (the maximum
temperature), and (d) 1000 ◦ C .
4.3.3.2 TiN buffer layer

The RHEED patterns of the TiN films depend on the number of laser pulses
at constant deposition temperature. After 70 pulses at Tsub = 600 ◦ C, the
film thickness is 0.4 nm. The pattern shown in the left panel of Fig. 4.21
exhibits varieties of spots due to the growth of superstructures. When the
film is thicker than 0.8 nm, the pattern changes to a few strong elongated
spots, indicating the growth of islands (right panel).
Fig. 4.21 The RHEED pattern of TiN on a Si (100) substrate at Tsub = 600 ◦C
(Reisinger et al., 2003). (a) 70 pulses, and (b) 1000 pulses.
4.3.3.3 Fe3 O4 film

Finally, the growth of Fe3 O4 on the buffer layer was characterized by
RHEED after 5000 pulses. The thickness of the magnetite was 40 nm.
As shown in Fig. 4.22, a well defined pattern with stripes is observed. The
spacing between the stripes is half of that shown in the case of TiN. This
is consistent with the fact that the lattice constant of Fe3 O4 is twice that
of TiN.
Fig. 4.22 The RHEED pattern of Fe3 O4 on a TiN buffer layer after 5000 pulses
(Reisinger et al., 2003).
4.3.4 Physical properties

Fe3 O4 has two attractive features:
• It has a ferrimagnetic transition temperature of 860 K (Smit and Wijn,

1959).
• Crystallites of the magnetite can be grown. It was predicted to show
half metallic property by Yanase and Siratori (1984) and Zhang and
Satpathy (1991).
In this section, we discuss first the most relevant property to this

monograph—the half metallicity of Fe3 O4 . What is the experimental ev-
idence? Then, we shall look into some surface issues. Finally, we shall
discuss theoretical studies of the electronic properties of Fe3 O4 and the
mechanism for its magnetic properties.
4.3.4.1 Half metallic properties

Evidence for half-metallic ferromagnetic properties of Fe3 O4 is furnished by
an experiment of the spin-polarized photoemission measurements (Dedkov
et al., 2002). The experiment was carried out using the photon energy of
21.2 eV. The samples are epitaxially grown Fe3 O4 (111) films. The temper-
ature at the measurements was RT and the pressure in the chamber was
1.0×10−10 mbar. The energy resolution was 100 meV and the angular res-
olution was ±3◦ . An external pulse of magnetic field of 500 Oe is applied
along the easy axis, in-plane < 11̄0 >, of the films.
The spectra were obtained in normal emission by a 180◦ hemispher-
ical energy analyzer connected to a 100 KeV Mott detector. Two sets
of data are collected. One is the energy distribution curve (EDC) N (E)
and the other is the polarization P(E). With the spin-polarized case, the
EDC is proportional to I(↑) + I(↓). The polarization is proportional to
I(↑) − I(↓). In general, the structures in the EDC are not prominently
exhibited. Therefore, sometimes additional measurements of the second
order derivative d2 N (E)/dE 2 need to be carried out. The polarization P is
–(80±5)% for 300 layers of oxidized Fe after annealing at 250 ◦ C (Fig. 4.23).
This led Dedkov et al. (2002) to conclude that Fe3 O4 is a HM.

The magnetotransport properties of Fe3 O4 have been probed by several re-
searchers. We separate the discussions into MR, junction MR (JMR), more
recent MR measurements using nanocontacts and spin valve structures, and
finally, neutron measurements to understand the underlying mechanism of
the ferromagnetism.
Fig. 4.23 Spin polarization as function of binding energy for a 300-layer oxidized Fe
after annealing at 250 ◦ C (Dedkov et al., 2002). The P value is –(80±5)%. The thin
solid line through the data points was fitted by a three-point averaging fast Fourier
transformation procedure.
Magnetoresistance For polycrystalline thin-film samples of Fe3 O4 ,

Feng et al. (1975) reported, as early as 1975, the observation of negative
MR for temperatures between 105 and 250 K. At 130 K and 23 kG, the
MR reaches 17%. The sign of the MR has physical significance. If it is
positive, the majority-spin polarization at EF of the sample is parallel to
the magnetization of the bulk or some reference systems, such as a trilayer.
If it is negative, then the minority-spin polarization at EF is parallel to the
magnetization of the bulk. Gridin et al. (1996) examined single crystals of
Fe3 O4 and obtained a sharp peak near TV with a maximum of 17% under a
magnetic field of 77 kG. About the same time, Gong et al. (1997) measured
the MR of Fe3 O4 samples grown epitaxially on MgO and observed a similar
peak around TV (∼119 K). In addition, they found that the MR increases
monotonically as the temperature decreases below 105 K. For a film of 6600
Å thick, the MR is 32% at 60 K and field of 40 kG. The results are shown
in Fig. 4.24. The dashed lines are the fit using Mott’s formula, Eq (4.8),
based on the mechanism of variable range hopping.
1
B
ρ = Ae( T ) 4 , (4.8)
where A and B are parameters. B can be interpreted as the activation
energy and is 1.61×108 K by fitting the zero field case. Based on these
fits, the authors attributed the origin of the MR increasing with decreasing
temperature to a magnetic field dependence of the activation energy for
hopping conduction below TV .
Besides the effects of the Verwey transition on the MR, Ziese and Blythe
(2000) found also anisotropic MR (AMR) 2 in thin-film forms of Fe3 O4
2 The background of AMR is discussed in Appendix A.1.
Fig. 4.24 Measured MR at magnetic field of 40 kG (down triangles), 20 kG (triangles),

10 kG (squares), and 5 kG (dots) in the temperature range 60–140 K for a Fe3 O4 film.
Fits to Eq (4.8) are indicated by dashed lines (Gong et al., 1997).
grown on MgO. The thicknesses of the films were 200, 50, and 15 nm,
respectively, with 10% uncertainty. The current was along the [100] axis
and the magnetic fields were applied along the [100] and [010] directions.
For the thickest films, additional measurements were made with the current
flowing in the [110] direction and magnetic fields oriented along the [110]
and [11̄0] axes. The four-point geometry with silver-paste contacts was
used to measure the resistivity. The MR was determined with respect to
the resistivity at the coercive field at which there is no magnetization in
the sample. A typical result of the MR showing the anisotropy is given in
Fig. 4.25. It is clear that the parallel and perpendicular MRs are different.
However, clear also is the anisotropy of each. Ziese and Blythe (2000)
attributed the anisotropy of the MR to the anisotropy of the samples.
Fig. 4.25 The magnetoresistance ratio Fig. 4.26 The magnetization as a func-
ρ/ρ0 of the Fe3 O4 single crystal and 200 tion of H of an Fe3 O4 /CoCr2 O4 /LSMO
nm thick film. The solid lines are mea- trilayer (Hu and Suzuki, 2002).
surements in longitudinal geometry and
the dashed lines in transverse geometry
(Ziese and Blythe, 2000).
Junction (tunneling) magnetoresistance The MR can also be mea-

sured by either magnetic junction (Hu and Suzuki, 2002) or nanocontact
method (Versluijs and Coey, 2001). Occasionally, the results obtained from
the former method are called the junction magnetoresistance (JMR) . Hu
and Suzuki (2002) measured the JMR of a trilayer junction made of Fe3 O4 /
CoCr2 O4 / La0.7 Sr0.3 MnO3 (LSMO). CoCr2 O4 serves as a weak paramag-
netic barrier (for T > 95 K) and has 0.8% lattice mismatch with Fe3 O4 .
The trilayers can be grown on (110) SrTiO3 by PLD. LSMO grown on
1900 Å SrTiO3 exhibits ferromagnetic half metallic properties with a TC
of 360 K (Park et al., 1998a). Between 0.4 eV (below EF ) and EF , the
spin-polarized photoemission spectra measured at 40 K show ∼ 100% po-
larization. Therefore, it is used as a spin analyzer for Fe3 O4 .
The JMR measurements were carried out with the current flowing in the
direction perpendicular to the plane and the external magnetic field parallel
to the plane of the film, along the easy [001] axis. To ensure that there is no
exchange coupling between the magnetized layers, the magnetization as a
function of magnetic field (H) was measured. The result of an unpatterned
trilayer at 80 K is shown in Fig. 4.26 and indicates that coercive fields of
the magnetic layers, LSMO and Fe3 O4 , are clearly separated at 280 Oe and
1.0 kOe, respectively. Figure 4.27 shows JMR as a function of H at 80 K
for a 20 µm × 20 µm cross-section. Using the resistance measured at 4.0
kOe (R4kOe ) as a reference, JMR is defined as
R4kOe − RH
JM R = . (4.9)
R4kOe
The results show negative JMR with the maximum JMR at –20%.
Fig. 4.27 The magnetoresistance de- Fig. 4.28 I-V

fined in Eq. (4.9) as a function of H for curves of Fe3 O4 nanocontacts (Versluijs
an Fe3 O4 /CoCr2 O4 /LSMO trilayer (T and Coey, 2001). The dashed lines are
= 80 K) (Hu and Suzuki, 2002). the fitted results.
Magnetoresistance determined by nanocontacts Transport mea-

surements using nanocontacts were developed after the STM method. One
notable feature is the possible observation of quantized conductance. Ver-
sluijs and Coey (2001) applied this method to Fe3 O4 . They used two
magnetite crystals with volumes less than 0.1 mm3 . These crystals were
brought together into mechanical contact. They were then pulled apart by
a piezo stack. The experiments were done at ambient conditions. A set of
Helmholtz coils was used to apply a magnetic field of 70 G in the parallel
and perpendicular directions of the current. There was a bias voltage up to
±0.3 V applied between the two crystals. An I-V converter with a low noise
amplifier (OP37) was used to measure the current. After the amplification,
the current was fed into a PC ADC card for analysis.
The results do not show quantized conductance. As the crystals were
slowly pulled apart, the resistance was continuously increased. The I-V
curves at two magnetic fields, B= 0 and B = 70 G, are shown in Fig. 4.29
with the curves fitted with I = GV + cV 3 , where G and c are fitting
parameters. The MR was defined similarly to the JMR (Eq. (4.9)) with
reference R at B = 0. The MR as a function of conductance is shown in
Fig. 4.29. The maximum MR is 85% for resistive contacts and the MR
decreases as the conductance increases. The authors ruled out that the
behavior of the MR is caused by magnetostrictive effects. They attributed it
to the formation of a domain wall at the nanocontact based on the nonlinear
results of the I-V measurements.
Magnetoresistance of spin valves involving Fe3 O4 Spin valves are

made of two magnetic layers separated by a normal metal. The negative
MR shown by other experiments suggests that Fe3 O4 in spin valve config-
urations could yield giant MR (GMR). This could be a significant techno-
logical advance. van Dijken et al. (2004) fabricated two spin valves made
of Fe3 O4 /Au/Fe3 O4 and Fe3 O4 /Au/Fe/Au, respectively, by dc-magnetron
sputtering on MgO(001) substrates under a base pressure of 10−7 mbar.
The samples had 10×10 mm2 areas. The current was measured in a van
der Pauw configuration having contacts at the four corners of the surface
of the sample. The magnetic field was applied in-plane with parallel and
perpendicular directions of the current. The magnetization curves were
determined using a SQUID magnetometer.
In Fig. 4.30, the longitudinal and transverse MR of 30nm–Fe3 O4 /5nm–
Au/10nm–Fe3 O4 on MgO(001) spin valve, the first spin valve, as a function
of applied magnetic field are shown in Fig. 4.30(b). For comparison, data for
a 50 mn Fe3 O4 film on MgO(001) are shown in Fig. 4.30(a). Figure 4.30(c) is

for 30nm–Fe3 O4 /5nm–Au/10nm–Fe/2nm–Au spin valve on MgO(001), the
second spin valve. The measurements were carried out at 300 K. The GMR
and AMR are much reduced in both spin valve cases, indicating current
passing through the Au region. In fact, no GMR is observed in the first
spin valve due to the reversal of the magnetization in each Fe3 O4 layer (van
Dijken et al., 2004). The AMR of the second spin valve is more pronounced.
Because all the contacts were made on the surface of the top Fe3 O4 layer, it
is reasonable to conclude that the AMR originates in this Fe3 O4 region and
possibly from the structural and stoichiometric differences in the regions
between Fe3 O4 and Au and between Fe3 O4 and MgO substrate. The second
spin valve shows quite different MR behavior as compared to the other two
cases, showing an isotropic GMR at low external magnetic field. The AMR
of the second spin valve is shown by the decrease of the longitudinal MR
but increase of the transverse MR as the magnetic field is cycled between –1
to 1 kG. van Dijken et al. (2004) suggested that this is due to the reversal
of magnetization in the Fe layer. They also attributed the isotropic GMR
in the second spin valve to the independent magnetizations in the Fe3 O4
and Fe layers.
Fig. 4.29 MR as a function of conductance at bias voltage of 0.3 V (filled circles) and
zero bias (open squares) for Fe3 O4 nanocontacts (Versluijs and Coey, 2001).
4.3.4.3 Charge ordering

The Verwey transition has a charge-ordered ground state at TV ∼ 120
K. The crystal exhibits ordered mixed valence of the Fe ions at the B
sites. According to Nazarenko et al. (2006) there are six non-equivalent
iron atoms. Two (Fe3+ ) are at the tetrahedral (A) sites and four (Fe2+ ,
Fe3+ ) ions occupy the octahedral (B) site. The charge ordering happens at
Fig. 4.30 MR as a function of magnetic field for Fe3 O4 grown on MgO(001) (a); and
for the two spin valves (b) and (c) (van Dijken et al., 2004).
the B sites. Two of them, Fe (B1 ) and Fe (B2 ), are at the centers of oxygen
octahedra. Fe (B3 ) and Fe (B4 ) are off center.
Resonance X-ray diffraction was used because the technique yields both
the site selective diffraction and the local absorption spectroscopy regarding
atomic species. Reflections in the range of tens of electron volts can be
recorded around the absorption edge of an element or elements where there
are evidence showing strong energy and angular dependencies (Nazarenko
et al., 2006). Physically, it involves virtual photon absorption-emission by
an electron initially occupying a core state and making a resonant transition
to some intermediate state near EF . This technique is particularly effective
for probing charge, orbital or spin orderings under distortions. For example,
different resonance frequencies can be correlated to the charge difference
at atoms suffering dissimilar distortions. By measuring the shift of the
absorption edges, the valences of the atoms were determined to be 5.38,
5.62, 5.40, 5.60 e for Fe (B1 ) to Fe (B4 ), respectively. On the other hand,
Wright et al. (2001) deduced from their neutron and X-ray measurements
the valences of the four Fe ions to be 5.6, 5.4, 5.4, and 5.6 e, respectively.
Therefore, there are some discrepancies for Fe (B1 ) and Fe (B2 ) between
these two measurements.
4.3.4.4 Surface properties

Because of the intense interest in magnetic thin films and multilayers, Kim
et al. (2000b) investigated the surface properties of Fe3 O4 (111) by grow-
ing the sample on Fe(110) in situ. They first used X-ray absorption spec-
troscopy (XAS), MCD, and LEED—allowing them to determine the ionic
state and element-specific magnetic information. This combination of ex-
perimental methods identified unambiguously the magnetic overlayer to be
ferrimagnetic Fe3 O4 when the layer thickness exceeds 600 layers. Further-
more, they found that the polarization in Fe3 O4 layers is in the opposite
direction of that in the Fe substrate.
These authors also studied bilayer forms of Fe3 O4 (111)/Fe(110) by
LEED, spin-polarized photoemission spectroscopy (SPPS), and magnetic
linear dichroism (MLD) (Kim et al., 2000a). The photon energy was 120
eV. The spin and angular resolved photoelectrons were analyzed using a
commercial VSW 50-mm spherical analyzer coupled with a low-energy dif-
fuse scattering spin polarimeter. The resolution of the electronic energy
was 0.1 eV and the angular resolution was around 2◦ . The measurement
chamber was maintained at the base pressure of 1.3×10−11 mbar. The sam-
ple was annealed at 250 ◦ C and was magnetized along the in-plane [001]
direction, which is the easy axis of the thick Fe film.
Figure 4.31 shows the experimental spin-polarized photoemission spec-
tra. In Fig. 4.31(a), the integrated spectra of the ↑ and ↓ spins are given
as a function of binding energy at different thickness. Figure 4.31(b) shows
the spin-resolved spectra. The pure Fe spectra are shown at the bottom of
the figures. The binding energy is measured with respect to EF .
For the pure Fe case, the width of d-states is approximately 3.0 eV. The
weak peak at about 8.0 eV below EF is attributed to s-states of Fe. As the
Fe3 O4 grows to 100 layers, a strong peak at 6.0 eV appears. It originates
from p-states of the O atoms. When the thickness increases to 600 layers,
several fine structures are present. By examining the spin-resolved spectra
in Fig. 4.31(b), it is clear that the fine structures are due to magnetic
ordering. It is also clear that the samples are not magnetic when the films
are less than 300 layers in thickness.
The MLD spectra reveal the magnetic characteristics of the samples.
In particular, the difference between spectra with opposite magnetizations
can provide interesting information. In Fig. 4.32, the MLD spectra and
differences between magnetizations of four films of different thickness are
plotted with respect to the binding energy. In the MLD spectra, the pure
Fig. 4.31 (a) Intensity of photoemission spectra of opposite magnetization (dots and
solid lines) as a function of binding energy for different layer thicknesses (Kim et al.,
2000a). (b) The corresponding spin-resolved curves. The spectra for the pure Fe are
shown at the bottom of both panels. The binding energy is measured with respect to
EF .
Fe sample shows a narrow peak at around 53.5 eV. As the thickness in-
creases, the peak becomes broader because in Fe3 O4 the Fe ions have two
different charges, Fe2+ and Fe3+ . The difference spectra indicate antipar-
allel coupling between the oxide layers and the Fe substrate. This can be
seen in Fig. 4.32(b). For the Fe, the difference spectrum shows a positive
slope at 53.0 eV. The 900-layer case (top curve) shows negative slope at
around 54.0 eV. The authors confirmed this by growing more Fe on top
of Fe3 O4 . As the Fe film thickness increases (1.2 ML), the onset of MLD
changes its sign: i.e., the Fe metal and thick oxide overlayer have opposite
magnetization. Kim et al. (2000a) attributed this to an antiparallel mag-
netic coupling in the Fe3 O4 (111)/Fe(110) bilayer. This essential feature
can be seen in a schematic diagram (Fig. 4.33). As Compared to SPPS for
the bulk-terminated surface of Fe3 O4 , there are two important features:
• The intensity of the photoemission spectra at EF are higher in the films

grown on Fe(110).
• The film samples show lower spin polarization (16%) than the bulk-
terminated surfaces (40–60%).
There is yet no convincing explanation of the intensity difference. One

possible reason for the smaller polarization in the film samples is the lack
of perfect stoichiometry.
4.3.5 One-electron theory

Zhang and Satpathy (1991) reported the first calculations of the spin-
Fig. 4.32 (a)The MLD spectra and (b) the corresponding differences of MLD signals
for four films of different thickness as a function of binding energy (Kim et al., 2000a).
Fig. 4.33 Schematic diagram showing the antiparallel magnetization of the

Fe3 O4 (111)/Fe(110) structure. The arrows indicate the local direction of magnetiza-
tion in the Fe planes. In Fe3 O4 the magnetization is supported by the Fe2+
octh ions (solid
arrows), while the Fe3+ magnetic moments cancel (dashed arrows) (Kim et al., 2000a).
polarized band structure of Fe3 O4 in 1991. They used the linear muffin-
tin orbital method within the atomic sphere approximation (LMTO-ASA)
(Andersen, 1975) and von Barth and Hedin exchange-correlation functional
(von Barth and Hedin, 1972) . The LMTO-ASA method is extensively dis-
cussed by Skriver (1983).
The Fe3 O4 structure is shown in Fig. 4.16. A lattice constant of 8.397
Å and 18 empty spheres were used. In Fig. 4.16, the Fe atoms are located
at the A and B sites. In terms of crystallographic terminology (Shepherd
and Sandberg, 1984), they are at 8a and 16d sites, respectively. The frozen
core approximation was applied to 3p-states of Fe and 1s-states of O. We
summarize the sites and muffin-tin sphere radii used in Table 4.5. Muffin-
tin orbitals of s, p, d on the Fe and O atoms and s and p type on the
empty spheres were used as basis functions. The total charge density was
constructed using a total of 56 k-points in the irreducible part of BZ.
There are three important contributions given by the LMTO-ASA cal-

culations:
• Fe3 O4 is predicted as a HM by the spin-polarized band structure and
DOS.
• The relation between the non-spin-polarized band structure and the
Stoner model.
• The Verwey transition was studied by a three-band model Hamiltonian
with all the parameters determined from first-principles calculations.
We shall discuss further the first two contributions, since we are interested
mainly in the half metallic and magnetic properties.
Table 4.5 Atomic site index and muffin-tin sphere radii of Fe3 O4 . The
site index is in crystallographic terminologies.
Atom Site index Muffin-tin sphere radius(Å)
Fe at A site 8a 0.95
Fe at B site 16d 1.15
O 32e (x = 0.379) 1.20
Empty sphere 1 16c 1.12
Empty sphere 2 8b 0.80
Empty sphere 3 48f (x = 0.25) 0.84
4.3.5.1 Spin-polarized band structure and DOS

The spin-polarized band structure and corresponding DOS (Fig. 4.34) in
Fe3 O4 were calculated by Zhang and Satpathy (1991). Key features in these
results are:
• The half metallic behavior is characterized by the intersection of EF
with states in the minority-spin channel only. The states at EF are
identified as the t2g states of the Fe atoms at B sites. In fact, the
trigonal crystal field at the B site splits the t2g states into doubly and
singly degenerate states. eg states originating from dx2 −y2 and dz2
(d3z2 −r2 ) states form those states 2.0 eV above EF . The Fe atoms at
A sites do not contribute significantly to those states near EF . They
are occupied states with binding energy of about 2.0 eV with respect
to EF . The gap in the majority-spin states is approximately 1.8 eV.
• The p-bands associated with the O atoms are separated by a gap of
about 1.0 eV from d-states of Fe atoms at A sites in the metallic chan-
nel; while the corresponding p- and d-bands in the majority-spin chan-
nel are separated by a small gap.
• The calculated magnetic moment M is 4.0 µB /Fe3 O4 . For each Fe atom,

the calculated M is 3.5 µB . This latter value is considered an estimate
because it depends on the chosen muffin tin radii. The moments on the
sublattices A and B are aligned antiferromagnetically.
Fig. 4.34 The spin-projected densities of states (DOS) and numbers of states for mag-
netites (NOS, dot lines) of the majority- and minority-spin channels in Fe3 O4 (Zhang
and Satpathy, 1991). The net magnetic moment is 52 − 44 = 8µB per unit cell which
consists of two formula units.
Comparisons to experiment The integer M and finite DOS at EF in

the ↓ spin states show that Fe3 O4 is a HM. The metallic behavior of this HM
arises from the minority-spin channel. This is consistent with the negative
MR observed in the experiments.
Polarized neutron scattering experiments have determined M at the A
site to be 3.82 µB . The calculated M at the A site discussed above agrees
reasonably well with the experiments.
One point not discussed by Zhang and Satpathy (1991) is the extent of
d-p hybridization between the Fe atoms at A and B sites with the O atoms.
Based on the DOS shown in Fig. 4.34, Fe atoms at B sites hybridize with
O atoms more strongly than those at A sites.
4.3.5.2 Non-spin-polarized band structure and Stoner model

The Stoner model (Stoner, 1939) explains successfully and intuitively the
ferromagnetism in Fe. The essence of the model characterizing the mag-
netism in a ferromagnetic metal is a consequence of the Pauli principle. This
results in two competing energies. The first is the increase in kinetic energy
as electrons are forced to occupy higher energy states and the second is the
reduction of the Coulomb repulsion. To see how the non-spin-polarized
band structure can be used in the Stoner theory3 , we let n = n↑ + n↓ be

the total number of d-electrons/atom and m = n↑ − n↓ be the magnetiza-
tion/atom in the crystal. The net energy of the two competing energies is
given by
1 m m′ m′2
Z
E(m) = dm′ − I n n
↑ ↓
, (4.10)
2 0 N (n, m′ ) 4
where In↑ n↓ is the Stoner parameter. Physically, it is a measure of the
strength of the intra-atomic Coulomb interaction between electrons con-
tributing to the magnetization and is part of the Hubbard U (Kim, 1999).
But in the present context, it is defined under the condition of a finite
magnetization. Here,
m
N (n, m) = (4.11)
ε↑ − ε↓
is the DOS/atom-spin averaged over the spin-flip gap ε↑ − ε↓ . Where ε↑
and ε↓ are two energies around EF for the ↑ spin occupancy n↑ = 12 (n + m)
and ↓ spin occupancy n↓ = 21 (n − m) in units of states/atom-spin. The
solution having net magnetization m0 can be obtained by minimizing the
energy E(m) with respect to m. The condition to have a ferromagnetic
state is
In↑ n↓ N (n, m0 ) = 1. (4.12)
This is the Stoner criterion. In addition, the stable ferromagnetic state
requires the second order derivative of E(m) to be negative. Otherwise,
the ferromagnetic state is metastable. Eq. (4.11) can also be derived from
considering the chemical potentials for the ↑ and ↓ spin subsystems. The
chemical potential for either spin channel is defined by the energy required
to add or remove an electron from the corresponding spin band. These two
chemical potentials are expressed as
µσ = εσ + In↑ n↓ n−σ . (4.13)
These chemical potentials suggest that the exchange interaction favors the
addition of a spin state to the majority-spin band. Eq. (4.11) is obtained
by equating the two chemical potentials.
The Stoner model can be generalized for Fe3 O4 , where there are two
sublattices (A and B) for the Fe atoms. With the two sublattices, it is pos-
sible that the number of d-electrons can change due to the magnetization.
Thus, the two chemical potentials at site A can be expressed as:
InA↑ n↓ A
µA
↑/↓ = ε (n
↑/↓ 0
A
, m A
) + (n A
− n A
0 )U A
+ (n ∓ mA ), (4.14)
2
3 Marcus and Moruzzi (1988) have generalized the Stoner model to reproduce spin-
polarized band structure calculations.
where nA A
0 (n ) is the total number of d-electrons at site A before (after)
the system exhibits magnetization. U A is the on-site Coulomb repulsion
energy at site A. A similar set of equations can be written for site B. Since
the two sites are not distinguished in Eq. (4.12), the present theory cannot
predict the relative orientation of the magnetizations at the two sites. The
Stoner conditions for the A and B sites are:
InA↑ n↓ N A (nA , mA ) = 1
(4.15)
InB↑ n↓ N B (nB , mB ) = 1.
By equating the two chemical potentials, µ↑ and µ↓ , at each site, one has:
εA A A A A A A B B B B B
↑ + ε↓ + 2(n − n0 )U + In↑ n↓ n = ε↑ + ε↓ + 2(n − n0 )U + In↑ n↓ n
B B
(4.16)
There is also the condition of conservation of total electron number with
and without the magnetization.
nA + 2nB = nA B
0 + 2n0 . (4.17)
The factor of 2 is due to the fact that there are twice as many Fe atoms
at B sites. The n’s and m’s are obtained by solving Eqs. (4.13) –(4.15)
self-consistently. Instead of carrying out self-consistent calculations, one
can estimate the magnetizations at the A and B sites. In Eq. (4.14), the
quantity U is the largest among all the terms. To satisfy Eq. (4.17), nA =
nA B B
0 and n = n0 are obtained. The Stoner parameters can be calculated by
using the scheme of Poulsen et al. (1976). They turn out to be between 60–
80 meV. The magnetizations at the two sites mA and mB are determined by
using the calculated N (n, m) for the two sites from the non-spin-polarized
band structure. mA and mB are approximately 3.5 µB .
4.3.5.3 Local spin density approximation with and without U

Anisimov et al. (1996, 1997) used the LMTO method to calculate the band
structure of Fe3 O4 with LSDA and LSDA + U, respectively. More re-
cently, Jeng et al. (2004) carried out similar calculations using the VASP
implementation of PAW method.
The U value used by Anisimov et al. (1996, 1997) was 4.51 eV. The par-
tial DOS for the two approximations, LSDA and LSDA+U, are compared
in Figs. 4.35(a) and (b). In both cases, the O p-states are well separated in
the region of –8 to –4 eV with respect to EF . Thus, the results indicate the
p-d hybridization between the O and Fe atoms is not strong. In the ↑-spin
channel, the five d-bands of the Fe atom at the B site are occupied with the
eg states above the t2g states. The d-states of the Fe at the A site form the
conduction bands with the eg state just above EF . For the ↓-spin states,
the role of the A and B sites is reversed. Furthermore, the t2g states at the
B site form the bottom of the conduction band. Since the calculations were
intended to study charge ordering, the B sites were divided into B1 and B2
sites for Fe3+ and Fe2+ , respectively. At the B1 site, nB1 = n0 + δn and
B2 has nB2 = n0 − δn. The LSDA+U DOS shows that the t2g states of
the B2 site in the ↓-spin channel experience a d-d interaction so that they
split into bonding and antibonding states. All the other d-states including
states at B1 site are unoccupied. Furthermore, the t2g states of Fe at the
B1 site have energy below the antibonding states.
Fig. 4.35 (a) PDOS with LSDA, and (b) PDOS with LSDA + U (Anisimov et al.,
1996, 1997). A, tetrahedral coordinated Fe ions; B, octahedral Fe ions (B1 corresponds
to Fe3+ and B2 to Fe2+ ions. The Fermi level (EF ) is set at 0.
The LSDA predicts that Fe3 O4 should be a HM with conducting states

from the ↓-spin channel. On the other hand, in the LSDA+U calculations,
the DOS shows that EF touches the top of the valence band for both spin
channels. The effect of U is to push up the ↑-spin eg states at the B site
(Fig. 4.35(a)) and lower the ↓-spin t2g states at the B2 site. The top edge
of the t2g states almost aligns with the eg states.
Jeng et al. (2004) calculated the band structures of both crystal struc-
tures above and below TV . For the low temperature case, they included U
in both LSDA and GGA exchange-correlation. The U and J values used
by these authors are given in Table 4.6.
The magnitudes of U are comparable to that used byAnisimov et al.
(1996) and the results of these calculations are:
• With the refined low temperature structure of Wright et al. (2001), the
calculated results show not only the charge ordering consistent with
the results of Wright et al. (2001) but also the orbital ordering.
Table 4.6 The U and J values used by

Jeng et al. (2004).
Approximation U (eV) J (eV)
GGA 5.0 0.89
4.5 0.5
4.5 0.7
4.5 1.1
LSDA 4.5 0.89
5.0 0.89
5.5 0.89
• The charge states on the Fe atoms are insensitive to the choice of U

and J. By integrating the charge density in a sphere of radius 1.0 Å
around each Fe atom, the charge states on the Fe atoms are listed in
Table 4.7. The calculated charges agree well with the measured values.
Table 4.7 The integrated charge in a sphere of radius 1.0Å

and the calculated M for the four Fe atoms.
Atom Charge (e) M (µB )
Exp. (Wright et al., 2001) Thoery Theory
Fe(B1 ) 5.6 5.57 3.45
Fe(B2 ) 5.4 5.41 3.90
Fe(B3 ) 5.4 5.44 3.81
Fe(B4 ) 5.6 5.58 3.39
• The total energy in LSDA+U for the monoclinic (low temperature)

structure is lower than the cubic structure by 0.35 eV/formula-unit
(f.u.).
• Both GGA without U applied to the distorted lattice and LSDA+U
applied to the undistorted cubic structure give metallic band structures
in the minority-spin channel and a gap in the majority-spin states.
• The values of M listed in Table 4.7 are correlated to the charge states
of the atoms. For Fe (B1 ) and Fe (B4 ) there is extra charge at the two
sites causing a reduction of the net spin moment at these sites.
• The charge ordering is explained by the lowering of one of the t2g
states of the Fe atoms at the B1 and B4 sites due to the distortions
associated with the Verwey transition. A schematic level diagram after
the distortion and corresponding filled orbitals are shown in Fig. 4.36.
At the B1 sites, the orbitals are dyz and dzx , respectively while the
one at the B4 site is dxy . They are now occupied and are depicted in
Fig. 4.37. The shift of the orbitals is called orbital ordering and it is
the driver of the charge ordering.
Fig. 4.36 Schematic energy level dia- Fig. 4.37 Orbital ordering in B-size
gram for the Fe ↓-spin d-orbitals un- sublattice, dyz and dzx are the orbitals
der distortion at B sites in Fe3 O4 (Jeng associated with the B1 atoms and dxy
et al., 2004). orbitals are with the B4 atoms (Jeng
et al., 2004).
The disagreement between the observed charge ordering (Wright et al.,

2001) and Anderson’s pattern (Anderson, 1956)—minimal electrostatic
repulsion—is explained. Anderson’s pattern requires each tetrahedron
with a corner sharing B type Fe to have two Fe2+ (B1 and B4 ) and
two Fe3+ (B2 and B3 ) ions. This follows from Anderson’s use of point
charges. In reality, the charges are not point charges. Therefore, the
screening effect is different.
• To obtain the charge ordered state in agreement with the experimental
results of Wright et al. (2001), it does not require an intersite Coulomb
interaction as introduced by Anisimov et al. (1996). It is also unneces-
sary to use different U’s for Fe2+ and Fe3+ (Antonov et al., 2001).
4.4 La1−x (Sr, Ca, Ba)x MnO3
La1−x Srx MnO3 (LSMO) is considered as an alloy of LaMnO3 with a La

atom substituted by a Sr atom. LaMnO3 is a perovskite-type oxide. It
shows colossal MR and exhibits half metallic properties for x = 0.3 (Park
et al., 1998b). At x = 0.3, it has been shown that its spin polarization is
95% at 4 K (Bowen et al., 2003) and ∼90% at 100 K (Bowen et al., 2005).
Experimentally, as early as 1997, Wei et al. (1997) provided evidence that
La1−x Cax MnO3 (LCMO) is a HM. La2/3 Ba1/3 MnO3 (LBMO) shows giant
negative MR (von Helmolt et al., 1993). These alloys are potential materials
for spintronic applications.
4.4.1 Structure
The crystal structure of perovskite is shown in Fig. 4.38. The Mn atom is
surrounded by six O atoms forming an octahedron. This octahedraon is the
backbone of the ferromagnetic oxides. LSMO exhibits a structural phase
transition at TS = 280 K with zero external magnetic field. TS is a func-
tion of the external magnetic field and decreases to 220 K at magnetic filed
70 kG. At low temperature, LSMO has the orthorhombic structure with
space group Pnma. The high temperature phase has the rhombohedral
structure and belongs to the R3c space group. All these lattice parameters
and angles depend on the value of x. For example at x = 0.17, the lattice
constants are a = 5.547 Å, b = 7.790 Å, and c = 5.502 Å. The structure
consists of six formula units—six perovskite structures. Three of the oc-
tahedra stack up vertically and two of them align in the a-b plane. For
the rhombohedral phase, the lattice constant is 5.475 Å and the angle α
is 60.997◦. Paiva-Santos et al. (2002) refined earlier crystal parameters of
the La0.65 Sr0.35 MnO3 rhombohedral structure determined by Alonso et al.
(1997) using the Rietveld method (Rietveld, 1969). The method is now im-
plemented in a software package (Young et al., 1995). The lattice constant
a is 5.5032 Å and c is 13.3675 Å. The value of a agrees reasonably well with
the one obtained by Asamitsu et al. (1996). The interatomic distances of
La or Sr to the O and Mn atoms are summarized in Table 4.8.
Fig. 4.38 The unit cell of the perovskite structure. The Mn atoms (open circle at the
center) occupy the A sites and the La atoms (black circles at corners of the cube) occupy
the B sites. Oxygen atoms are shown as gray circles at face centers of the cube.
In fact, in La1−x Srx MnO3 the structural transition between the or-
thorhombic (O) and rhombohedral (R) phases depends on the temperature
and Sr concentration x. The phase diagram expressed in terms of temper-
ature and x is shown in Fig. 4.39, obtained from the neutron scattering
Table 4.8 Refined inter-atomic distances between TM elements and

an O atom, and the distance to the Mn atom in La0.65 Sr0.35 MnO3 .
TM element Distance (Å)
La or Sr above and below the O atom 2.746(4)
La or Sr nearly in the same plane of the O atom 2.524(4)
Mn 1.954(4)
experiments (Kawano et al., 1996b). There are three sets of data points.
The circles and squares mark the onsets of antiferromagnetic and ferro-
magnetic to paramagnetic phase transitions, respectively. The filled and
open diamonds indicate the R → O transition temperature. The triangles
indicate the R → O structural phase transitions for x = 0.125.
The neutron scattering results show also that there are two O phases,
O′ and O∗ . The difference between these two phases was examined based
on the powder pattern using Rietvelt analysis. The O′ phase has the MnO6
octahedra√distorted due to the Jahn-Teller effect and its lattice constants
∗
√ 2 < c < a. The O phase exhibits the pseudocubic structure
satisfy b/
with b/ 2 ∼ c ∼ a and without any evidence of the Jahn-Teller effect.
Fig. 4.39 Magnetic (a) and structural (b) phase diagrams of La1−x Srx MnO3 with x ≤
0.17. CAF is canted antiferromagnetic, FM is ferromagnetic, and PM is paramagnetic.
The circles and squares indicate the onset of FM and AF components, respectively.
The diamonds indicate the R → O transition temperature. The triangles denote the
structural phase transitions for x = 0.125 (Kawano et al., 1996b).
4.4.2 Growth and characterization

Depending on the properties to be probed, the growth involves bulk sam-
ples, polycrystalline forms, and thin films. Bulk La1−x Srx MnO3 , x = 0.35,
samples were grown by firing stoichiometric mixtures of Mn2 O3 , La2 O3 , and
SrO3 at 1300 ◦ C for 16 hrs with intermediate grinding (Paiva-Santos et al.,
2002). X-ray diffraction was used to characterize the crystals and probe the
possible refinement. Polycrystalline samples of LCMO were grown in two
ways. One was to first grow powered forms using similar mixtures as de-
scribed above. The mixture was calcined three times in the air at 1050 ◦ C
for 24 hrs with intermittent grinding. There is a slight difference in detailed
settings for the temperature and time (Asamitsu et al., 1996; Kawano et al.,
1996a). The resulted powders were pressed into rods under a hydrostatic
pressure of 1.4 ton/cm2 then fired at 1100 ◦ C in air for 38 hrs. Both groups
used X-ray powder diffraction to characterize the samples. Another way to
grow LCMO is to grow single crystals using the floating zone method. The
initial samples were prepared in rod form and were loaded into a flow zone
furnace. The feeding speed is 5–10 mm/hr. The single crystal samples were
then powered. To form a tunneling junction, the thin films of LCMO were
grown by PLD on LaAlO3 (100) substrates (Wei et al., 1997). The growth
temperature was 700 ◦ C and in an environment of 13.3 mbar of oxygen.
The samples were then annealed at 900 ◦ C in 1 atm oxygen for several
hours. XPS and STM were used to check the quality of the films. The XPS
results showed very low contamination indicating the surfaces of the films
were clean. The STM was operated at 1 nA constant-current mode with Pt
tip which was biased at about –2.0 eV. The images were taken at RT and
77 K in either high vacuum or ultra-pure He gas. They show atomically
smooth “rice-paddy” type of terraces with the step height of one unit cell.
4.4.3 Physical properties

Among the physical properties, half metallicity is the prime interest with
respect to the spintronic applications. In addition, magnetic and transport
properties have attracted much attention. A recent review was given by
Salamon and Jaime (2001).
4.4.3.1 Half metallicity

Both LCMO and La0.5 Ba0.5 MnO3 were predicted to be half metallic by ab
initio band structure calculations (Hamada and Terakuar, 1995; Satpathy
and Vukajlović, 1996). The first experimental evidence on LSMO was given
by Okimoto et al. (1995) using optical conductivity measurements for x =
0.175 and by Hwang et al. (1996) measuring the MR for x = 1/3. Wei et al.
(1997) carried out TJ measurements for LCMO with x = 0.3. The measured
tunneling conductances for LCMO with x = 1/3 at T = 77 K (solid-dot
curve) and 300 K (open circles) are shown in Fig. 4.40 for cases below and
above TC . The normalized conductances in Fig. 4.40 are proportional to
the DOS. In practice, it normalizes the inherent dependence of the STM
transmission probability on the voltage.
Fig. 4.40 Normalized conductance as a function of bias voltage of LCMO with x=1/3,
at T = 77 K (solid-dot curve) and T = 300 K (open circles) (Wei et al., 1997).
The important features are the two peaks at ±1.75 eV appear in the low
temperature results. These two peaks agree with most of the DOS derived
from the spin-polarized band structures of LSMO. The DOS results exhibit
the half metallic properties for LSMO (Singh and Pickett, 1998; Ma et al.,
2006). The –1.75 eV peak corresponds to the peak in the occupied states of
the majority-spin channel and the +1.75 eV peak corresponds to the peak
in the unoccupied minority-spin states above EF . The difference is the
strength of the exchange splitting. The half-metallic feature is explicitly
shown by the presence of the spin splitting peaks in the low temperature
results. At 300 K, there are no such peaks consistent with the paramagnetic
properties.
4.4.3.2 Magnetic and transport properties

Curie temperature is the primary interest regarding the magnetic properties
of the oxide alloys involving La. The magnetization and hysteresis loops
will also be discussed. As for the transport properties, we shall discuss
primarily the various MRs.
Magnetic properties TC in these oxide alloys varies with respect to

the composition x. Determination of TC as a function of x for LSMO was
carried by Jonker and Santen (1950). Table 4.9 lists TC for several of these
oxides for a few x compositions. Shown in Fig. 4.41 are the magnetization
M(T) as a function of temperature and hysteresis loop of LCMO at x = 0.3

(Park et al., 1998b). The sample was epitaxially grown in film form on a
SrTiO3 (001) substrate by PLD. M(T) was measured by SQUID and TC is
extrapolated to be 360 K. The hysteresis loop was determined by MOKE.
It shows that the sample has a single domain as indicated by nearly 100%
remanent M and very low coercive field Hco . For bulk samples, Hco is
approximately 30–50 Oe and M is saturated at 5 kG (McCormack et al.,
1994).
Table 4.9 Summary of TC for the alloying oxides.

Alloy x TC (K) Reference
LSMO 0.17 264 Asamitsu et al. (1996)
0.175 283 Okimoto et al. (1995)
0.3 360 Park et al. (1998b)
1/3 370 Bowen et al. (2003)
0.3 – 0.5 ∼380 Asamitsu et al. (1996)
LCMO (film) 0.3 260 Wei et al. (1997)
LaBa (LB) MO 0.33 343 von Helmolt et al. (1993)
Fig. 4.41 (a) M(T)/M(T = 0 K) as a function of temperature and (b) a hysteresis loop
at T = 300 K for LCMO with x = 0.3 (Park et al., 1998b).
Transport properties Experiments providing evidence for LBMO with

x = 1/3 exhibiting giant negative MR were carried out by von Helmolt
et al. (1993). The samples were films in thickness of 150±10 nm grown
by PLD on (100)- and (110)-oriented SrTiO3 (STO) substrates. The unit
cell is rhombohedral. MR is defined as M R = [R(0) − R(50kG)]/R(0). At
RT, MR is more than 60% larger than that of Cu/Co multilayer structures.
In metallic multilayers, MR is attributed to the spin-dependent scattering
at the interface between the magnetic and non-magnetic layers. The MR
in LBMO is independent of the relative orientation of H, current I, and
axis of the samples. The high MR and formation of partially localized

states at the magnetic impurities were the basic reasons for von Helmolt
et al. (1993) to suggest that the carriers contributing to the transport are
magnetic polarons.
Tunnel magnetoresistance (TMR) and inverse TMR measurements were
carried out by Lu et al. (1996) and Teresa et al. (1999), respectively. The
TMR ratio is ∆R/R = (R↑↓ − R↑↑ )/R↑↓ , where R↑↓ and R↑↑ are resistances
for the two sides of the junction in antiparallel and parallel configurations.
The trilayer sample of LSMO/STO/LSMO was grown by PLD. The two
LSMO layers are approximately 500 Å thick and the barrier is 30–60 Å.
The MR value of 83% was obtained. From the result, the value of P de-
duced from Jullière formula was 54%, which is in good agreement with the
value predicted by Singh and Pickett (1998) (53%) for La2/3 Ba1/3 MnO3 .
The inverse TMR measurement was carried out with trilayers of LSMO
(x=0.3)/STO/Co. The thicknesses of the three layers were 35/2.5/30 nm,
also grown by PLD. In this layer configuration, the LSMO is used as a
spin analyzer for the polarization in Co. Experiments were carried out at
T = 5 K and –50% inverse TMR was measured. The “inverse” is defined
as the negative TMR, having small resistance in the antiparallel configura-
tion. The carriers originate from the negatively polarized electrons in the
Co layer. It is well known that the DOS of Co at EF for the majority-
spin channel is smaller than the corresponding DOS of the minority-spin
states. In Fig. 4.42, the data for the MR exhibit the inverse TMR. A qual-
itative explanation of the MR results is given in Fig. 4.43. The DOS in the
minority-spin channel of Co at EF is larger than that of LSMO. At bias
voltage V=0V (Fig. 4.43(a)), the current flows from the left (majority) to
the right (minority). From the TMR ratio, R↑↑ is larger. When V is in the
range of 0.7 and –2 V (Fig. 4.43(b)), EF in the Co region is lowered and
the current flows from the left to the right due to relative feature of the
DOS for LSMO and Co is not changed. Current reverses its direction when
V is greater than 1.15 V because R↑↓ is larger due to the DOS of the Co
majority-spin states is now larger (Fig. 4.43(b)).
More recently, the spin-polarized TMR in LSMO (350 Å)/STO (001)
(28 Å)/LSMO (100 Å) with x = 0.3 was determined by Bowen et al. (2003).
These trilayers differ from the earlier samples in three aspects:
• Higher quality interfaces of LSMO/STO.

• Smaller sizes of junctions.
• Pinning the spin of the top layer of LSMO. The pinning was accom-
Fig. 4.42 MR as a function of bias voltage of two LSMO/STO/Co samples at T = 5 K

and 30 K (Teresa et al., 1999).
Fig. 4.43 Schematic diagram of the tunneling processes corresponding to MR in

Fig. 4.42. The vertical axis in each panel is the energy, the horizontal axis is the DOS,
and EF is indicated by dashed lines. The thick arrows indicate the direction of current
flows (Teresa et al., 1999).
plished as follows: The samples were first inserted into an rf sputtering

system and were covered by a 150 Å Co layer on top. Then, they were
etched by an oxygen-rich plasma to form a CoO layer of 25 Å in thick-
ness. Finally, the samples were capped by a 150 Å Au layer . The TMR
is 1800% at T = 4.2 K under a bias voltage of 1 mV. The deduced P
is 95%. The TMR vanishes at T = 280 K. The improved TMR was
attributed to the sample preparation.
One-electron properties In this section, we discuss one-electron theory,

in particular the band structures. Effects of strains, the Jahn-Teller dis-
tortion, and rotations of spin moments on some Mn atoms are the primary
issues. Most of the calculations were done using LSDA with the LAPW
method as by Singh and Pickett (1998). The FLEUR code4 having the
spin polarization and other modern features is a popular implementation,
as used by Ma et al. (2006).
Singh and Pickett (1998) have studied effects of the rotation of spin
moments and the Jahn-Teller distortion in Mn-O6 octahedra. Similarly, Ma
et al. (2006) also used a 20-atom model with Pnma symmetry and a 5-atom
perovskite unit cell. The model is shown in Fig. 4.44. The lattice constants
and MT radii used by Ma et al. (2006) are summarized in Table 4.10.
Fig. 4.44 A 20-atom model of LSMO (Ma et al., 2006). Sr is shown as a large gray
circle, filled black circles are La, small gray circles are Mn, and open circles are O.
Table 4.10 Summary of lattice constants and MT radii of LSMO with x =

1/3 used in LAPW calculations (Ma et al., 2006).
Lattice constant (Å) MT radius (Å) U (eV) J (eV)
a c La Sr Mn O
3.90 3a 1.32 1.32 1.06 0.794 2 and 3 0.7
The conclusions of their study are summarized as follows:
• Mn(1) has 3.11 µB moment and Mn(2) has 3.26 µB .

• Using GGA exchange-correlation, the total magnetic moment/unit-cell
is 10.89 µB . With GGA+U, the moment is 11.0 µB . The half metal-
licity of LSMO with x = 1/3 is therefore sensitive to the choice of U
4 http://www.flapw.de
value used in the calculations.

• The effects on the VBM (valence band maximum) and CBM (conduc-
tion band minimum) due to the in-plane compression were examined
with and without U. For U = 0, the VBM is higher and the CBM is
lower with respect to EF by a few tenths of an eV. With U = 2.0 eV,
the VBM is essentially unchanged while the CBM shifts down by about
the same amount as for U = 0. Consequently, the effects of the strain
are not drastic.
• The occupied states near EF are the eg states. The in-plane strain lifts
the degeneracy of x2 − y 2 and 3z 2 − r2 states.
• The origin of the magnetism was attributed to the mixed valence of the
Mn ions, Mn3+ and Mn4+ , and an electron hopping between them.
In treating the LSMO alloy within the DFT calculations , Singh and
Pickett (1998) used virtual crystal approximation for La2/3 Ba1/3 MnO3 .
The size of the unit cell is that of LCMO with x = 1/3 exhibiting colossal
MR (CMR). Relaxations were performed in ferromagnetic and antiferro-
magnetic configurations. Their results are summarized as follows:
• The spin moment on the Mn atom is 3.40 µB .

• The DOS does not show any half metallic property.
• The FM phase has a lower energy than the AFM phase by 0.014 eV/Mn.
• The rotations of the octahedra affect the eg states which form the gap
near EF .
• The lattice distortions in FM alloys do not couple to electrons as strong
as in the spin-disordered paramagnetic state.
4.5 Magnetic interactions in the oxides
The understanding the physical origins of the magnetic properties in the

oxides remains a challenging task. We shall first consider the basic mag-
netic interactions discussed in the literature. Then, we discuss some recent
results. The key interactions are:
(1) Superexchange interaction.

(2) Double exchange interaction.
4.5.1 Superexchange
4.5.1.1 Model approach
First proposed by Kramers (1934), it was Anderson (1950) who simplified
Kramers’ general formalism and provided a simple picture. The effective
interaction based on Kramers’ formulation between two Mn atoms with a
common neighboring O atom is called superexchange. The essential idea of
the interaction is shown in Fig. 4.45. The O2− ion is located between two
Mn2+ ions. There are no free carriers in Anderson’s view of superexchange.
Fig. 4.45 Superexchange interaction. The dashed arrows indicate that the electrons
can hop.
The O2− ion has a filled p-shell so that it has a net charge of –2e. The
two electrons occupy one of the p-orbitals with spins pointing in opposite
directions. The Mn2+ ions have five d-electrons and align their local spins
according to Hund’s first rule. O p-electrons can hop to the neighboring
unoccupied Mn d-states while conserving spin. This delocalization reduces
the energy of the O2− ion. Because the spins of p-electrons are oppositely
oriented, this hopping can only occur if the neighboring d electrons are
antiferromagnetically aligned. The effective interaction was reformulated
in terms of spin operators and the orthogonality of the wave functions.
In reality, the wave functions at the neighboring atoms are not orthogo-
nal. This led Yamashita and Kondo (1958) to reconsider the superexchange
interaction. These authors pointed out that if the ionic model shown in
Fig. 4.45 is considered, then it is necessary to expand the total energy
to 4-th order in the overlap integral S in order to have the total energy
dependent on the spin arrangements. They argued that the ionic states
should include some mixing of electronic configurations for excited states
and investigated the possible contributions from 4 different configurations:
• The Slater Model (Slater, 1953). This model consists of [Mn2+ ; O2−
(2p)6 ](the ionic model) + [Mn2+ ; O2− (2p↑)2 (3s↑)(2p↓)3 ] +[Mn2+ ;
O2− (2p↑)3 (3s↓) (2p↓)2 ].
• The Kramer-Anderson model (Anderson, 1950). It has [Mn2+ ; O2− ] +
[Mn+ ;O− ].
• The Goodenough model (Goodenough, 1955). Two configurations are
[Mn2+ ; O2− ] + [Mn+ —O—Mn+ ].
• The band model (Yamashita and Kondo, 1958). The model also has
two configurations, [Mn2+ ; O2− ] + [Mn3+ —O2− —Mn+ ].
The Kramer-Anderson model can have both ferromagnetic and antifer-

romagnetic states. All the other models give antiferromagnetic states due
to symmetry or the Pauli principle. Furthermore, both the Goodenough
model and band model indicate that the Mn ions can have mixed valence.
The important point is that it is the overlap matrix elements which con-
tribute to superexchange.
4.5.1.2 First-principles approach

Oguchi et al. (1984) approached superexchange in TM oxides and sulfides
such as MnO, NiO, and MnS, from the band structure point of view. They
discuss the “itinerant versus localized picture of superexchange.” The use of
the word “itinerant” conflicts with Anderson’s localized picture of superex-
change. In electronic band structures, localized electrons have flat bands.
The authors pointed out that the superexchange is associated with second
neighbor antiferromagnetic interactions characterized by the exchange con-
stant J2 originating from the eg -p-eg coupling. It is known that the eg states
in Mn form narrow bands. In this sense, it is consistent with the localized
picture of Anderson. The approach has no empirical parameters and the
interaction strength for a crystal is inversely proportional to the exchange
splitting instead of the U term in the model approach. Qualitatively, this
explains the increasing trend of the Néel temperatures from MnO to NiO.
4.5.2 Double exchange

Zener (1951) proposed a mechanism to explain the ferromagnetic proper-
ties of oxides, now known as the double exchange. The double exchange
interaction is like that of superexchange in that it causes alignment of neigh-
boring magnetic ions to minimize the kinetic energy of hopping electrons.

However, in double exchange the alignment is ferromagnetic rather than
antiferromagnetic as in superexchange. It typically occurs in systems with
mixed valence magnetic ions, such as Fe3 O4 . This is because such systems
can have neighboring magnetic ions of the same species but different va-
lences: for example, the Fe2+ and Fe3+ ions at the octahedral B sites in
Fe3 O4 . In such a case, hopping of the unpaired d-electron of the Fe2+ (3d6 )
to the Fe3+ (3d5 ) can occur only if the remaining d-electrons are aligned
ferromagnetically. Thus kinetic energy is reduced by this configuration.
The interaction can be understood in terms of two degenerate configura-
tions, Fe2+ —O2− —Fe3+ and Fe3+ —O2− —Fe2+ . They are connected by
double exchange matrix elements characterizing an electron hopping from
Fe2+ to O2− and from O2− to Fe3+ . It is important to realize that the
transfer matrix element in this case is nonzero only when the local spins at
Fe sites are aligned according to Hund’s first rule. There is a resonance of
local spins in neighboring Fe atoms if the spins are parallel. The splitting
due to double exchange matrix elements is of the order of kB T, where kB
is the Boltzmann constant.
Fig. 4.46 Double exchange interaction. S1 and S2 are the spin moments of the Mn ions
contributed by the localized d-electrons. s is the spin of the itinerant electron.
Anderson and Hasegawa (1955) revisited Zener’s theory (Zener, 1951)

using a model for treating classically the localized spins at the Mn ions and
quantum mechanically the spin of a mobile electron characterized by the
double exchange matrix elements. A simple picture is shown in Fig. 4.46.
The Mn ions have their d-electron spins aligned according to Hund’s first
rule. The spin moments for the left and right Mn ions are S1 and S2 , re-
spectively. The right Mn ion has a mobile electron with spin moment s.
Three cases listed in Table 4.11 are considered. The important quantities
are the intra-atomic exchange integral J between mobile electron and local
d-orbitals, and the magnitude of the transfer matrix element b characteriz-
ing the effective coupling between case 1 and case 2. The authors showed
that by coupling states from two Mn ions with b, the splitting of the Zener
levels is proportional to cos( 2θ ), where θ is the angle between classical spins
S1 and S2 at the two Mn ion sites and the effective transfer integral is
Θ (S0 + 21 )
tef f = b cos( )=b , (4.18)
2 2s + 1
where S0 is the sum of the spin moments of the two magnetic ions, and
s is the spin of the mobile electron. With b > J, they showed that the
ferromagnetic phase can occur through the mobile electron. It is possible
to conclude that double exchange favoring ferromagnetism should involve:
• Localized spin at each TM ion site.

• Hund’s first rule coupling the localized spin and spin of the mobile
electron.
Consequently, the motion of the mobile electron between the two sites
is correlated. Most of the investigations of the magnetic properties of the
metal oxides have been based on either superexchange or double exchange
or both, depending on the particular oxide of interest.
Table 4.11 Three cases consid-

ered in double exchange. d1
and d2 are the localized d-states
contributing to S1 and S2 . d is
the state of the mobile electron
and p2 and p are the doubly and
singly occupied states of the O
atom.
Case Mn1 O Mn2
1 d1 p2 d2 , d
2 d1 , d p2 d2
3 d1 , d p d2 , d
4.5.3 Magnetism in CrO2 and Fe3 O4

There has been much work investigating the magnetic interactions in CrO2
and Fe3 O4 . Because much has been learned about these two important
cases, we consider each in turn below.
4.5.3.1 CrO2
For CrO2 , there is general consensus that it is the double exchange mech-
anism that is responsible for its magnetic properties. Schlottmann (2003)
combined single-particle results with collective excitations to show that
double exchange in CrO2 depends critically on the distortion of the oc-
tahedron surrounding the Cr atom and therefore differs from the double
exchange mechanism in the manganites.
The physical picture is as follows: The Cr ion in the octahedron formed
by the O atoms has valence 4+. Only two d-electrons remain in valence.
Due to the cubic environment, the five d-states split into t2g (triply de-
generate) and eg (doubly degenerate) states, with the t2g states at lower
energy. Therefore, the t2g states are partially occupied and eg states are
unfilled. The t2g states are composed of dxy , dyz , and dzx orbitals. The
octahedron undergoes a Jahn-Teller distortion. This tetrahedral distortion
lifts the degeneracy of the t2g states with the dxy state having the lowest
energy. The other two orbitals form √12 (dyz +dzx ) and √12 (dyz –dzx ) states.
One of the two electrons occupies the dxy state and the other electron has
a 50% probability of occupying each of the combined orbitals. The level
scheme is shown in Fig. 4.47.
Fig. 4.47 Level diagram of the Cr4+ ion under cubic crystal field and tetragonal dis-
tortion. One of the two electrons occupying the dxy state is localized while the other
electron sharing √1 (dyz ± dzx ) states is itinerant.
2
To simplify the model for analysis, the following approximations are

made:
• Direct hopping between the Cr atoms rather than mediation by the O

p-states.
• The localized state at each site does not participate in hopping.
• Hopping between the same orbital angular momentum states.
Letting the hopping integral between the two sites, 1 and 2, be t, the
hopping Hamiltonian is
X †
Ht = −t c1ξσ c2ξσ + c†2ξσ c1ξσ + c†1ζσ c2ζσ + c†2ζσ c1ζσ , (4.19)
σ
where ζ and ξ label orbital angular momentum states and σ labels the spin.
The other term required is the on-site Coulomb energy U to model the
full Hamiltonian. To apply the model Hamiltonian to CrO2 , we note that
there are two Cr ion sites, with two electrons each. One of them occupies
the dxy localized state and the other is itinerant. From Hund’s first rule
at each site, the orbital wave function is odd under permutation of the
two electrons and the spin state is triplet. Combining the two sites, the
resultant spin momenta, S = 2, 1, 0, can be constructed and serve as the
states acted on by Ht . The S = 2 case consists of three states having an
orbital triplet and one state having an orbital singlet. The triplet is even
under permutation of two mobile electrons while the singlet is odd.
ζζ
ψeven (S = 2) = c†1ζ↑ c†2ζ↑ |1 ↑ 2 ↑i
(4.20)
ζξ
ψeven (S = 2) = √12 (c†1ζ↑ c†2ξ↑ + c†1ξ↑ c†2ζ↑ ) |1 ↑ 2 ↑i
ξξ
ψodd (S = 2) = c†1ξ↑ c†2ξ↑ |1 ↑ 2 ↑i
ζξ (4.21)
ψodd (S = 2) = √12 (c†1ζ↑ c†2ξ↑ − c†1ξ↑ c†2ζ↑ ) |1 ↑ 2 ↑i ,
where c†1ξ↑ is the creation operator of an electron at site 1 occupying state

ξ—one of the linear combinations of dyz and dzx with ↑ spin. The even
states are unchanged under Ht . The energy of the odd state is lowered by
4t2 /U. This is the largest shift among all values of S. Therefore, ferromag-
netic ordering is favored, consistent with the double exchange mechanism.
4.5.3.2 Fe3 O4
The Fe ions in Fe3 O4 have two different valences, 2+ at the B2 and B3
sites and 3+ at the B1 and B4 sites. In addition, the Verwey transition
complicates investigations of the origin of the ferromagnetism in this com-
pound. At present, the origins of the ferromagnetism in Fe3 O4 and the
La oxide alloys are still somewhat controversial. In this section, we dis-
cuss a treatment considering the double exchange mechanism but taking
the antiferromagnetic ordering between A and B sites into account. The
approach differs from that of the preceding section, considering instead the
free energy and minimizing it with respect to the magnetization at the B

site (Loos and Novák, 2002).
The double exchange mechanism proposed by Loos and Novák (2002)
was based on the quantum mechanical formulation of Kubo and Ohata
(1972). Its original application was to La oxide alloys. Grave et al. (1993)
applied a modified formulation (Kubo and Ohata, 1972) to Fe3 O4 . In the
original formulation for La oxides alloys, the picture is that the Mn3+ has
three electrons occupying the t2g manifold and one electron in the higher
energy eg states. The eg states on the Mn4+ are unoccupied. The electron
in the eg states of Mn3+ is considered as itinerant. It can hop from Mn3+ to
Mn4+ and vice versa. The assumptions are that the intra-atomic exchange
integral J is larger than the hopping integral tij , J ≫ |t|, the charge effect
due to alloying is averaged out, and the alloys can be treated within the
virtual crystal approximation. The Hamiltonian characterizing the double
exchange is
X
† †
X
H = −J ′
S i σss′ c cis′
is + ′
t ij c is c js′ , (4.22)
i,s,s i,j,s,s
where i and j are site indices, S is the total spin of the localized states, σss′
is the Pauli matrix, and the c’s are the creation and annihilation operators
for the itinerant electron. Recall that in the spinel structure of Fe3 O4 shown
in Fig. 4.16, there are tetrahedral (A) sites and octahedral (B) sites. The B
sites contribute to the local spins and itinerant electrons (Loos and Novák,
2002). The mean field approach was adopted to find an effective field acting
on an atom at a B site due to its neighbors.
Above the Verwey temperature, there are NB sites with each B site
having a local spin moment S of (5/2)~ and NB /2 itinerant electrons. Let
S ′ be the effective local spin moment resulting from the spin moment of
the itinerant electron coupled to the local spin through Hund’s first rule.
Grave et al. (1993) obtained a value of 2.9 for Fe3 O4 through their hy-
perfine measurements based on the double exchange formulation of Kubo
and Ohata (1972). This differs quite significantly from the ideal estimation
of 2.0. Loos and Novák (2002) addressed this discrepancy by including su-
perexchange between site A and site B. This is done by using effective fields
which are proportional to the z-components of the equilibrium magnetiza-
tions, mA,eq =< SA,z >eq and mB,eq =< SB,z >eq , acting respectively on
site B and site A ions. mA , therefore, can be expressed as
mA = SBS (λAB mB,eq + λAA mA ) , (4.23)
where BS is the Brillouin function. The λ’s characterize the superexchange
interactions, with
λij = βB SJij zj (i), (4.24)
where Jij is the strength of the superexchange interaction between sites i
and j, zj (i) is coordination number of the i-th atom with j nearest neigh-
bors, and βB is kB1T . Let λ, the mean field at T, be βSgµB Hef
z
f acting at
a B site; the magnetization at the same site is
z
mB (λ) = hSB (λ)i = (1 − x)SBS (λ) + xS ′ BS ′ (S ′ λ/S), (4.25)
where g is the Landé g-factor and x is the fraction of itinerant electrons
relative to the total at site B. S is the total local spin and S ′ is the resultant
spin at site B including the local spin S and the spin of the itinerant electron
coupled by Hund’s rule (Loos and Novák, 2002). The effective Hamiltonian
due to the superexchange and the free energy per B site fB can then be
given:
m2

1
hHSE i = − mB λBA mA,eq + λBB B (4.26)
βB S 2
fB = xµ + Ω + hHSE i − T S (S), (4.27)

where µ is the chemical potential for the itinerant electron. It is expressed
in terms of the DOS, g(ε), of the free carriers,
g(ε)
Z
µ = dε β (ε−µ) , (4.28)
e B +1
grand canonical potential
Z
Ω= dεg(ε) ln(eβB (ε−µ) + 1), (4.29)
and entropy,

(S)
X
λm/S
X λmB λm/S
S = kB (1 − x) ln( e ) + x ln(e ) − .
|m|≤ 52 |m|≤2 S
(4.30)
mB can then be determined by minimizing fB with respect to mB . The
theory has some shortcomings. If S ′ is set equal to 2, then TC is 1590
K, almost a factor of 2 larger than experiment. Loos and Novák (2002)
attributed this shortcoming to the mean field theory.

Chapter 5
Half metals with simple structures
5.1 Introduction
Both Heusler alloys and transition metal (TM) oxides are appealing for
spintronic applications. However, due to the possible large Coulomb corre-
lation at the TM sites, defects, phase transitions (Borca et al., 2001), and
surface and interface effects in thin-film samples discussed in Chapter 3, at
present, neither Heusler alloys nor TM oxides have been shown experimen-
tally to exhibit half metallic properties at room temperature (RT). Thus at
the beginning of this century, researchers began to search for half metals
(HM) with simpler structures, such as the zincblende (ZB) structure. Ex-
citing results were obtained by Akinaga et al. (2000a). They first predicted
CrAs in the ZB structure to be a HM by first-principles calculations then
grew it in thin-film form. In this new HM, the unit cell consists of just two
atoms, as in GaAs. It is especially appealing in light of the following:
• The simple structure can eliminate the disorder problem.

• It can be integrated readily into well developed semiconductor tech-
nologies because its structure conforms to semiconductors in common
use today.
• If single crystals or thin films can be grown, transport properties can
be improved; in contrast to Heusler alloys which are more susceptible
to defects.
As for most of the Heusler alloys, there is as yet no experimental evi-

dence of half metallicity for ZB compounds such as CrAs. The results of
Akinaga et al. (2000a) have nevertheless stimulated much research to de-
sign new HMs with simple structures. In this chapter, we shall first discuss
the efforts undertaken by various researchers to grow (Zhao et al., 2001;
191
Etgens et al., 2004) and predict (Galanakis, 2002a; Pask et al., 2003; Xie
et al., 2003; Xu et al., 2002; Liu, 2003; Galanakis and Mavropoulos, 2003;
Continenza et al., 2001; Zhao et al., 2002; Şaşoǧlu et al., 2005a) new HMs
in the ZB structure. Since the growth of these compounds is the major
hurdle, due to the fact that the ZB structure is not the ground state struc-
ture (e.g., many have the hexagonal NiAs structure as the ground state,
in which they are not half metallic), we give the details of the growth
techniques adopted by various researchers to provide a foundation for fu-
ture developments. The interactions giving rise to the half metallicity in
these compounds will be discussed. We shall then describe the efforts to
grow (Mizuguchi et al., 2002; Akinaga and Mizuguchi, 2004) and design
quantum structures, including superlattices (Fong et al., 2004; Fong and
Qian, 2004), quantum dots (Qian et al., 2004a; Shirai, 2004), and digital
ferromagnetic heterostructures (DFH) (Sanvito and Hill, 2001; Qian et al.,
2006a; Wu et al., 2007; Zhu et al., 2008), and discuss their physical prop-
erties. Half metallic Si-based DFH’s will be elaborated in particular, due
to the wide availability of mature Si-related technologies. Finally, we shall
discuss attempts to design even simpler structures. In this respect, carbon
nanowires doped with TM elements are a promising direction (Dag et al.,
2005; Durgun et al., 2006).
5.2 Half metals with zincblende structure
Two TM pnictides with ZB structure have been grown in thin-film form.

All other TM pnictides, a carbide, and chalcogenides have been predicted
by calculations.
5.2.1 Experiment
We shall focus on the experimental efforts on the growth, characterization,
and magnetic properties of HMs with ZB structure.
5.2.1.1 Growth
There are two forms for the growth: thin film and multilayers.
Thin films of CrAs In 2000, Akinaga et al. (2000a) used MBE to grow
CrAs on a GaAs(100) substrate. A molybdenum holder with indium solder
was used to hold the substrate. A 20 nm GaAs buffer layer was first grown
Half metals with simple structures 193
after the chamber was degassed. The CrAs was then grown on top at 0.02
nm/s. The temperature of the growth was 200 ◦ C. Finally, a 4.0 nm layer
of gold was used to cap the CrAs for preventing oxidations. The resulting
thickness of the CrAs layer with the ZB structure is 2 nm.
The results reported by Akinaga et al. (2000a) were not without con-
troversy. Etgens et al. (2004) investigated the structure of CrAs epilayers
grown on GaAs(001) and also used the MBE method to grow their sam-
ples. A 100 nm undoped GaAs buffer layer was first grown on GaAs(001)
substrate. The surface was annealed under As to improve the flatness of
the surface. CrAs was then grown at 200 ◦ C with a rate of 8 Å/min under
As rich conditions. At 295 K, CrAs has the orthorhombic MnP structure
with lattice constants, a = 5.637 Å, b = 3.445 Å, and c = 6.197 Å. Above
1100 K, it is grown in the hexagonal NiAs structure.
Thin films of CrSb Zhao et al. (2001) reported the growth of CrSb on
GaSb. They also used the MBE method to grow CrSb films on a GaAs(001)
substrate with three different buffer layers, GaAs, AlGaSb, and GaSb, re-
spectively. The layer configurations are shown in Fig. 5.1. The solid sources
of Ga, As, Sb, Al and Cr were provided. Before depositing the buffer layer,
the oxide layer on top of the substrate was cleaned at 580 ◦ C. The buffer
layers have different thickness as shown in Fig. 5.1. For example, case A has
500 nm thick GaAs and was grown at 560 ◦ C. During the growth on this
buffer layer, the As and Ga fluxes were terminated simultaneously at the
end of the growth. The Sb fluxes were immediately released to prevent the
sublimation of Group V elements from the surface. At this time, the tem-
perature of the substrate was lowered to grow CrSb. When the substrate
temperature reached 400 ◦ C, the flux of Sb atoms was stopped to avoid the
build-up of Sb on the surface. As the temperature of the substrate reached
250 ◦ C, both Cr and Sb fluxes were then turned on. Typically, the intensity
of the Cr flux was 2×1013cm−2 s−1 . The beam equivalent pressure ratio of
Sb/Cr was maintained at about 10. By monitoring RHEED patterns that
show ZB features in the sample, these authors specified the thickness in
terms of depositing time. Typically, the growth time was 60 s. The results
are summarized in Table 5.1.
Multilayer growth Recognizing that the ZB structure is not the ground

state structure for CrAs and CrSb, the growth of multilayer films has been
pursued by Mizuguchi et al. (2002). They used a MBE system to grow
a CrAs/GaAs multilayer with period z on semi-insulating GaAs(001) sub-
Fig. 5.1 Layer configurations of growing CrSb. The time (seconds) of deposition for
the GaSb layers is also shown (Zhao et al., 2001).
Table 5.1 Summary of information for growing CrSb with the ZB structure on three different buffer
layers.
Sample Buffer Growth time Duration of Substrate temp. Cr flux
layer (s) ZB RHEED pattern (s) (◦ C) (1013 cm−2 s−1 )
A GaAs 60 40 250 2
B AlGaSb 64 60 250 3
C GaSb 64 60 250 2
strates. A sample of the multilayer having a period z consists of x layers of

CrAs and y layers of GaAs is shown in Fig. 5.2.
Fig. 5.2 Schematic diagram showing the layer configurations, x, y, and z of CrAs/GaAs
multilayers. LT means low temperature (220 ◦ C) with respect to the buffer layer tem-
perature (580 ◦ C) (Mizuguchi et al., 2002).
We now discuss the growth of the multilayers. In the ultra-high vacuum

chamber, thermal cleaning was carried out by annealing the substrate at
600 ◦ C for 10 minutes to remove the surface oxidation layer on the GaAs
substrate. A GaAs buffer layer of 20 nm thickness was grown at 580 ◦ C.
Then CrAs and GaAs layers were grown alternately. They were controlled
by opening shutters of Knudsen cells for each pair of elements simultane-

ously at 220 ◦ C. The vapor pressure ratio of As /Cr was set at 100 to 1000.
The period of the growth z was set at 10 or 100. The thickness of the
last GaAs layers was set at 5 nm to prevent the oxidation of the multilayer
structures. The substrate temperature (bottom GaAs layer in Fig. 5.2) was
maintained at 200 ◦ C.
5.2.1.2 Characterization using X-rays

Various X-ray techniques have been used to characterize the layer growth.
The most common one is RHEED. In addition, X-ray absorption spec-
troscopy (XAS), X-ray photoelectron spectroscopy (XPS), and ex-situ high
resolution transmission electron microscopy (HRTEM) have also been used.
RHEED for CrAs epilayers The RHEED patterns of ZB CrAs were

used to characterize the quality of films by Akinaga and Mizuguchi (2004).
In Fig. 5.3, the RHEED patterns for a CrAs surface on GaAs along the
[110] and [100] directions of the substrate are shown. The streaky features
and symmetry of the intensities are the signature of the ZB structure. By
looking at the temporal change of the patterns from streaky to spotty, these
authors identify a critical thickness of 3 nm for the film to exhibit the ZB
structure. Further confirmation of the ZB structure for a film of 2 nm was
obtained by HRTEM.
Fig. 5.3 The RHEED patterns of CrAs films with the electron beam incident along
the [110] (a and c) and [100] (b and d) directions of a GaAs substrate (Akinaga and
Mizuguchi, 2004).
Etgens et al. (2004) also used in situ RHEED patterns to characterize

a CrAs film. The patterns started with a well-ordered GaAs streaky form
but faded into a strong diffuse background at 1 Å deposition. After 2 Å

deposition, the faded streaky patterns changed to elongated spots. The
spotty patterns persisted with increasing growth thickness. These authors
tried to anneal the films at 300 ◦ C under As pressure. The streak patterns
became more elongated in the diffuse background. Inter-diffusion between
layers did not appear as confirmed by core-level photoemission experiments.
GIXD for CrAs epilayers The Grazing Incidence X-ray Diffraction

(GIXD) technique is effective to investigate the crystallography of thin films
and epilayers (Robinson, 1991) because the X-ray probes the substrate at
total reflection condition to enhance the signal from epilayers with respect
to the signal from the bulk. Etgens et al. (2004) used this scheme to exam-
ine three CrAs film samples. The thicknesses of the samples were 12, 25,
and 45 Å, respectively. These samples are labeled as X12, X25, and X45.
Before carrying out the GIXD experiments, the surface of each sample was
cleaned and verified by Auger spectroscopy indicating that there were no C
and O contaminations. The X-ray measurements were indexed with cubic
GaAs(001) surfaces expressed in terms of the bulk GaAs, they are:
1 1 1
aSGaAs = [11̄0]GaAs, bSGaAs = [110]GaAs, cSGaAs = [001]GaAs . (5.1)
2 2 2
The calculated CrAs lattice constant of 5.8 Å (Galanakis, 2002a) was used.
This value gives a 2.6% mismatch with that of GaAs. Therefore, they
searched the strained and relaxed CrAs layered samples using the GIXD
method. The results displayed diffraction peaks which can be associated
with an orthorhombic structure for all thicknesses. The X-ray spots are
shown in Fig. 5.4. The ones associated with GaAs are denoted by * and
with CrAs are indicated by ⋄. For the thicknesses less than or equal to 25 Å,
the patterns are similar. They differ from the thickest case. The thinnest
case has [100]CrAs parallel to [110]GaAs (the a-axis) and [010]CrAs parallel
to [11̄0]GaAs (the b-axis). The c-axis of CrAs is parallel to the growth
direction. Along the a- and b-axes, there are 6% and 13% mismatch with
respect to the corresponding lattice constants of the substrate. With this
thin layer, only one epitaxy was observed. The result suggests that aSGaAs
and bSGaAs in Eq. (5.1) are not equivalent.
For the thickest case X45, there are two different domains. The ori-
entations of the epilayers were determined and are denoted as (a,c) and
(a∗ ,c∗ ) in Fig. 5.4. The orientations of the axes of the surface cells have
been identified to be [100]CrAs k [100]GaAs and [001]CrAs k [010]GaAs for
(a,c) and [001]CrAs k [100]GaAs and [100]CrAs k [010]GaAs for (a∗ ,c∗ ).
Fig. 5.4 The GIXD patterns for the three samples with different thicknesses (Etgens
et al., 2004). The axes are in reciprocal lattice units.
The b-axis of CrAs bCrAs is in the growth direction. From the spots, the
length of a differs from |aCrAs | by 0.23%. On the other hand, the one in
the direction of c has 9.5% mismatch. These basis vectors correspond to
the fully relaxed epilayer. Comparing the different GIXD patterns of the
thin to thick epilayers, the single phase can be due to a metastable phase
formed at the beginning of the growth. Consequently, the growth results
can depend on the film thickness. Since characteristics of the ZB structure
were not observed in these experiments, Etgens et al. (2004) attributed the
failure of finding the ZB structure to the lattice mismatching. However,
there is the possibility that this growth did not control the orientation of
CrAs and GaAs, so that CrAs is not in the ZB structure.
RHEED and HRTEM for CrSb epilayers The RHEED and high-
resolution cross sectional transmission electron microscopy (HRTEM) were
used to characterize the substrates and film forms of CrSb (Zhao et al.,
2001). The HRTEM spectra exhibit features of a film in real space.
The RHEED picture shows a streaky (1×3) pattern during the CrSb
growth on a Sb-terminated GaSb surface at the substrate temperature
ranging from 250 ◦ C to 400 ◦ C. This initial pattern became weaker and
disappeared after 40 seconds. For the growth on other substrates, such as
Al0.84 Ga0.16 Sb and GaAs, the streaky characteristic of the ZB GaAs buffer
layer disappeared after depositing CrSb for 60 seconds.
The results of HRTEM for a CrSb layer grown on GaAs for 30 s at the
substrate temperature of 250 ◦ C and with the Cr flux of 2×1013 cm−2 s−1
were obtained by Zhao et al. (2001). The film sample was capped by GaAs.
Both the limited and wider areas show 1 ML of CrSb without any defects,
such as dislocations, at the interfaces. The results provide an evidence of
a more or less ideal 1 ML growth. This characterization should serve as a

standard for a δ-layer growth.
RHEED, XAS, and TEM for CrAs/GaAs multilayers The

RHEED, XAS (Mizuguchi et al., 2002), and cross-sectional transmission
electron microscopy (TEM) (Akinaga and Mizuguchi, 2004) have been used
as the primary methods to characterize the CrAs/GaAs multilayers.
The surface conditions were monitored by RHEED during all growth
processes by Mizuguchi et al. (2002) and Akinaga and Mizuguchi (2004).
The RHEED patterns for different samples with and without capping of
GaAs layers are summarized in Table 5.2. In the table, A, B, C, and D
represent different CrAs/GaAs multilayer configurations. E and F are the
cases of substituting CrAs layers by alloys of (Ga,Cr)As. E is for 50% of Cr
while F is for 1% of Cr. Their multilayer configurations are the same as for
the multilayer A. The streaky pattern of RHEED is a characteristic of the
ZB CrAs layer. In the multilayer A, the streaks related to CrAs are weaker
than those of the GaAs buffer layer. The streaky pattern persists up to
the tenth CrAs layer in the multilayer A. These results suggest that two
layers of CrAs grown on two layers of GaAs can be formed in multilayers.
The spotty patterns in the multilayer B and C indicate that the growth
mode is not epitaxial, where the samples have more than two layers of
CrAs (x > 2) with less than two layers of GaAs (y < 2). Therefore, it
is concluded that the set of (x, y) = (2, 2) is the optimum condition to
realize the epitaxial ZBCrAs/GaAs multilayer. The thick multilayer D
shows initially epitaxial growth. However, a change of the patterns from
streaky to spotty happens gradually. At z ∼ 50, the patterns appear to have
extra spots, indicating that some unknown phase is present. For multilayer
D, the RHEED eventually shows a perfect spotty pattern. It is interpreted
as the CrAs layers losing the ZB structure. Akinaga and Mizuguchi (2004)
increased the substrate temperature to 300 ◦ C to grow superlattices having
two layers of CrAs and two layers of GaAs up to 100 periods without any
spotty pattern appeared in the RHEED.
XAS measurements were performed to probe the chemical bonding of
the alloys using the multilayer A as the reference (Mizuguchi et al., 2002).
The photon energy varies continuously from 570 to 595 eV with the incident
angle set at 60◦ from the normal. The fluorescence yield from the Cr L line
edge was measured using a Si detector with the detection angle normal to
the sample surface. The results are shown in Fig. 5.5. The spectra were
normalized by the peak height. The integral backgrounds were subtracted
Table 5.2 The CrAs/GaAs multilayer samples and characteristics of RHEED

analysis. x represents the number of CrAs multilayers, y is the number of
GaAs multilayers, and z denotes the number of periods of these sublayers in
the growth direction (Mizuguchi et al., 2002).
Multilayer sample x (ML) y (ML) z (period) RHEED pattern
A 2 2 10 Streaky
B 2 1 10 spotty
C 3 3 10 spotty
D 2 2 100 spotty
E spotty
F Streaky
in each case. The measured spectra originated from the Cr L3 and L2

peaks of multilayer A. The spectra for the multilayer E are very similar.
For the multilayer F, there is at least a 0.5 eV down shift with respect to
the multilayer A. The possible reasons are as follows:
• In the multilayer E, CrAs can be in a segregated state while in multi-

layer F, the Cr atoms are in the dilute magnetic semiconductor state.
• In the multilayer E, CrAs precipitates.
• There is a possibility for Cr to diffuse into the GaAs region in the
multilayer E.
Fig. 5.5 XAS results of multilayer CrAs/GaAs samples of A, E, and F listed in Table 5.2.
Unfilled and filled ▽’s denote peak positions (Mizuguchi et al., 2002).
The TEM image for a multilayer CrAs/GaAs with x, y, and z of 4,

4, 10, respectively, was obtained by Akinaga and Mizuguchi (2004), which
clearly shows the multilayer region. The sample was grown on a 20 nm
GaAs buffer layer. At the interface, there are possible defects. But there
is no Cr inter-diffusion into GaAs buffer and cap layers as studied by the
secondary ion mass spectroscopy (SIMS).

In addition to investigating the sample growths and characterizations, they
also studied the magnetic properties. Among them, two groups carried out
the measurements using the SQUID magnetometer.
CrAs epilayers Akinaga et al. (2000a) measured the hysteresis loop for
an epitaxial film of CrAs grown on GaAs(001) at RT. The magnetic field
was applied parallel to the film plane. The saturation magnetization is 560
emu/cm3 . This is equivalent to 3.0 µB per formula-unit and agrees with
theoretical predictions (Akinaga et al., 2000a; Galanakis, 2002a; Pask et al.,
2003). For different samples under identical growth conditions, there is ap-
proximately 10% spread in values of the saturation magnetization. The
coercivity is small. Based on these RT measurements, the Curie tempera-
ture (TC ) is estimated to be greater than 400 K.
Fig. 5.6 The hysteresis loop of an epi- Fig. 5.7 Magnetization as a function
taxial CrAs grown on GaAs. The mea- of thickness of a CrAs film. The inset
surement was done at 300 K (Akinaga shows the hysteresis loops for the two
et al., 2000a). CrAs films with different thickness (Et-
gens et al., 2004).
The magnetization as a function of field shown in Fig. 5.6 is a char-

acteristic of CrAs with the ZB structure. One of the possible structures
for CrAs is the MnP structure. It is known to show a helimagnetic para-
magnetic transition at 265 K (Suzuki and Ido, 1993). Contamination of
the Mn atoms was excluded by SIMS because the concentration of the Mn
atoms is 4 × 1017 atoms/cm3 . Under this concentration, it is difficult to
form ferromagnetic MnAs. The ground state structure of CrAs is the NiAs
structure. It is antiferromagnetic with a Néel temperature of 710 K. It is
unlikely that the NiAs structure contributes to the hysteresis loop at RT.
Etgens et al. (2004) also reported results of experiments on CrAs but
did not discuss the methods employed. The main result is the saturation
magnetization as a function of thickness of the films shown in Fig. 5.7.
When the layers are ∼ 4.0 nm, the saturation magnetization of different
samples with the same thickness exhibits more than a factor of 3 fluctua-
tion with significant uncertainties. At 20 Å (2 nm), the saturation value is
1000 emu/cm3 , which gives ∼ 3µB per Cr atom for an orthorhombic CrAs,
assuming that only Cr atoms contribute to the magnetic moment. This
number agrees with that obtained by Akinaga et al. (2000a). As the layer
thickness increases, the magnetization reaches a value below 250 emu/cm3 .
The hysteresis loops for the two films with different thicknesses are shown
in the inset. The coercive field and remanent magnetization are also com-
parable to those given by Akinaga et al. (2000a).
CrSb epilayers Zhao et al. (2001) used SQUID to measure the hysteresis
loops of CrSb thin-film samples grown on three different buffer layers. They
also reported the remanent magnetization as a function of temperature.
The hysteresis loops at 300 K for the CrSb thin-film samples are shown
in Fig. 5.8. The contribution from the diamagnetic GaAs was subtracted.
Even the layer thickness was only 1 ML, the films exhibit ferromagnetic
properties at RT. The saturation magnetizations at 5 K are between 3
and 5 µB /formula-unit. The value of 3.0 µB /formula-unit agrees with the
theoretical calculations by Pask et al. (2003).
Fig. 5.8 Hysteresis loops of CrSb thin- Fig. 5.9 Temperature dependence of
film samples grown on three different the remanent magnetization of the three
buffer layers–GaAs (triangles), AlGaSb CrSb thin-film samples (Zhao et al.,
(circles), and GaSb (squares) (Zhao 2001).
et al., 2001).
The temperature dependence of the remanent magnetization of the three

samples measured at B = 0 is shown in Fig. 5.9. The results of the films
grown on GaAs and GaSb are very close while the one grown on AlGaSb
alloy is higher. At 400 K, all the samples still exhibit finite remanent
magnetization indicating that the TC is higher than 400 K.
CrAs/GaAs multilayers The SQUID measurements were used to de-

termine the magnetic hysteresis loops of CrAs/GaAs multilayers. An ex-
ample of the hysteresis loop for [(CrAs)2 /(GaAs)2 ]100 capped by 10 nm
GaAs grown on a 20 nm GaAs buffer layer is shown in Fig. 5.10. The loop
is very narrow. The remanent magnetization is similar to the one of the
CrAa epilayer. The saturation magnetization is 400 emu/cm3 and is equiv-
alent to 2.0 µB /formula-unit which is smaller than the theoretical value of
3.0 µB /formula-unit for bulk CrAs (Pask et al., 2003).
Fig. 5.10 The hysteresis loop of a multilayer structure: [(CrAs)2 /(GaAs)2 ]100 (Akinaga
and Mizuguchi, 2004).
5.2.2 Theory
Since the pioneering work of Akinaga and Mizuguchi (2004) on CrAs, much
theoretical work has been devoted to the prediction of new HMs with ZB
structure. Among them, Fong et al. (2004) and Fong and Qian (2004)
designed superlattices with HM properties. Later, Qian et al. (2004a) re-
ported an integer magnetic moment in a quantum dot composed of MnAs
(Qian et al., 2006b). Dag et al. (2005) and Durgun et al. (2006) designed
quantum wires showing half metallic properties. Sanvito and Hill (2001)
examined the half metallic properties in δ-layer-doped Mn in GaAs. Such
structures are now known as digital ferromagnetic heterostructure (DFH).
In this section, we shall review these efforts by starting with compounds.
We shall discuss the basic interactions causing the half metallicity in these
compounds. Then we shall discuss the associated quantum structures.
Despite the fact that the ZB structure is not the energetic ground state
structure for the TM compounds, many half metallic pnictides, a carbide,
and several chalcogenides have been examined theoretically. Table 5.3 lists
some of these theoretical efforts, including the bulk lattice constants, lattice
constants at which the compound exhibits half metallicity, and correspond-
ing magnetic moments.
The agreement among the theories is very good for all properties except
the magnetic moment; some show non-integer values. The discrepancies
are mainly caused by different methods. There was some doubt whether
the pseudopotential method would be suitable to treat these compounds.
As illustrated in Table 5.3, however the results obtained by the plane-
wave pseudopotential (PWPP) method agree very well with those of the
all-electron linearized augmented plane wave (LAPW) method.
Table 5.3 Theoretica predictions of HMs with ZB structure.

Compound Lattice constants (Å) Magentic moment Reference
Bulk HM (µB /Formula-unit)
CrAs 5.66 5.66 3 Akinaga et al. (2000a) [LAPW]
5.659 5.53–5.65 3 Galanakis (2002a) [KKR]
5.67 5.67 3 Pask et al. (2003) [LAPW]
5.67 5.50–5.87 3 Pask et al. (2003) [PWPP]
5.65 3 Galanakis (2002a) [KKR]
CrP 5.35 5.48 3 Pask et al. (2003) [LAPW]
5.42 5.47–5.60 3 Pask et al. (2003) [PWPP]
CrSb 6.15 6.15 3 Pask et al. (2003) [LAPW]
6.14 5.89–6.39 3 Pask et al. (2003) [PWPP]
6.139 5.69–7.17 3 Liu (2003) [LAPW]
CrSe 5.62–5.82 4 Şaşoǧlu et al. (2005a) [KKR]
5.833 5.833 4 Xie et al. (2003) [LAPW]
MnAs 5.72 5.7 4 Pask et al. (2003) [LAPW]
5.74 5.77–5.97 4 Pask et al. (2003) [PWPP]
5.643 3.75 Continenza et al. (2001) [LAPW]
5.70 5.70 3.5 Sanvito and Hill (2001) [Tight-binding]
5.73–5.87 4 Şaşoǧlu et al. (2005a) [KKR]
MnBi 6.399 5.399 4 Xu et al. (2002) [LAPW]
MnGe 5.61–5.65 3 Şaşoǧlu et al. (2005a) [KKR]
MnP 5.308 2.73 Continenza et al. (2001) [LAPW]
MnSb 6.166 3.77 Continenza et al. (2001) [LAPW]
6.19 6.19 4 Pask et al. (2003) [LAPW]
6.21 5.96–6.46 4 Pask et al. (2003) [PWPP]
6.128 4 Xu et al. (2002) [LAPW]
MnSi 5.52–5.65 3 Şaşoǧlu et al. (2005a) [KKR]
MnC 4.20–4.23 1 Şaşoǧlu et al. (2005a) [KKR]
4.36 4.26 1 Pask et al. (2003) [LAPW]
4.39 3.60–4.28 1 Pask et al. (2003) [PWPP]
VAS 5.54 1.939 Galanakis and Mavropoulos (2003) [KKR]
Table 5.4 Compound, range of lattice constants a within ∼4% of the corre-
sponding equilibrium values over which half metallicity is predicted and the
suggested substrates for growth. Substrate lattice constants (Wyckoff, 1963)
are given in parentheses.
Compound Lattice constants (Å)
LAPW PWPP Substrate
CrP > 5.48 > 5.47 AlP (5.53)
CrAs > 5.51 > 5.50 GaAs (5.64), AlAs (5.62)
CrSb > 5.87 AlSb (6.13), GaSb (6.12), InAs (6.04)
MnAs > 5.75 > 5.77 InP (5.87)
MnSb > 6.06 > 5.96 AlSb (6.13), GaSb (6.12), InSb (6.48)
MnC > 4.26 > 4.20 Diamond (3.57), cubic SiC (4.35)
The energetic ground state structures of these compunds have also been
considered. Pask et al. (2003) compared NiAs and ZB structures using both
pseudopotential and LAPW methods with GGA exchange-correlation. The
NiAs structure is confirmed to be the ground state for the pnictides. Re-
alizing that the ZB structure is a metastable structure and thin-film forms
can be grown, a list of possible substrates (Table 5.4) for each compound
was suggested Galanakis and Mavropoulos (2003) also suggested GaAs as
a substrate to grow CrAs.
In the following, we focus first on a typical pnictide, MnAs, where
the underlying interactions associated with the half metallicity will be dis-
cussed. We then discuss a TM carbide, MnC, which shows some interesting
and opposite properties from the pnictides and chalcogenides.
5.2.2.1 MnAs
Several groups (Pask et al., 2003; Continenza et al., 2001; Şaşoǧlu et al.,
2005a) predicted MnAs in the ZB structure to exhibit half metallic prop-
erties. The ZB structure of MnAs is shown in Fig. 1.9, where the filled
grey circles are As atoms, and they are located at the corners and centers
of the cubic faces. Open circles are the Mn atoms. It is important to note
that the Mn and As atoms can form chains, for example, along the [110]
direction.
With the structure in mind, we now discuss the half metallic and elec-
tronic properties:
Density of states The density of states (DOS) is a convenient physical

quantity to exhibit half metallic properties of a compound. Another quan-
tity is the magnetic moment/unit-cell (or formula-unit). In Fig 5.11, the
DOS of MnAs at the lattice constant of 5.77 Å is shown (Pask et al., 2003).
The upper panel shows the DOS of the majority- (↑) spin channel and the
lower panel is for the minority (↓) spin states. EF is set to be zero on the
energy scale. The features of the DOS are:
• At EF , a finite DOS, 0.77 states/eV-cell, appears in the ↑ channel.
• For the ↓ spin states, EF falls within a gap of 1.70 eV.
• P value at EF is 100%.
The position of EF in the insulating channel is lattice constant depen-
dent. When the lattice constant increases, EF drops toward the top of the
valence band. As a result, the half metallicity disappears when the lattice
constant reaches a critical value.
Fig. 5.11 The density of states of MnAs (Pask et al., 2003).
Magnetic moment/unit-cell The calculated magnetic moment for

MnAs is 4.0 µB /unit-cell. Because the total number of electrons in a unit
cell is an integer, the insulating channel requires integer number of elec-
trons to fill the valence bands, and the g-factor of an electron is 2, the
magnetic moment should then be an integer. This is a necessary condition
for a compound to be a HM.
Electronic band structure The spin-polarized band structure of MnAs

from the pseudopotential method (Pask et al., 2003) is shown in Fig. 5.12
along a few symmetry directions of the BZ for the fcc structure. As a
reference, the BZ is shown in Fig. 5.12. The lowest bands in both spin
channels are formed by the s-like state of the As atom. The next set of
bands at the Γ-point (labeled as Γ15 ) are triply degenerate t2g states. They
are the bonding states as the consequence of d-p hybridization. They have
predominantly the character of the As p-like states. The next higher en-
ergy bands are denoted as Γ12 or eg . They are doubly degenerate and are
called the non-bonding d-states of the Mn atom. The lobes of the charge
distribution of these two states point toward the second neighbors of the
Mn atoms. In this cubic environment, it is possible to identify that they
are originated from the dz2 and dx2 −y2 states. In the ↑ spin channel, these
doubly degenerate states are occupied, while those in the ↓ spin channel
lie above EF . These bands separate from the lower energy Γ15 (t2g ) bands
by a gap. Therefore, the minority channel exhibits insulating behavior. In
the majority-spin (↑) states, we label another Γ15 (t∗2g ) above Γ12 . These
are antibonding states resulted from d-p hybridization. The broad fea-
ture of these bands make them partially occupied along the Γ–X direction
contributing to the metallic behavior of the majority-spin states. These
features show the half metallic properties from the band structure point of
view.
Fig. 5.12 Band structure of MnAs along symmetry directions of the BZ (Pask et al.,
2003). The BZ of the fcc lattice is shown on the right.
Charge densities To examine the bonding properties in this ZB half

metallic compound, the charge densities of the occupied t2g and eg states
associated with the majority-spin channel are shown in Fig. 5.13. Both
sections contain the zig-zag chain formed by the Mn and As atoms along
the [110] direction.
The left panel shows the charge distribution of the t2g states. The
bond charge is clearly exhibited between the two atoms. In the present
case, the high concentration at the Mn atom manifests the tightly bound
character of the d-state. If we translate this unit cell in both x and y
directions, the four lobes of the d-state can be identified. It is a linear
combination of the dxz and dyz states. The right panel is for the eg states.
We can easily identify the lobes around the Mn atom associated with the
dz2 state. These lobes point toward the second neighbor Mn atoms. There
is no distribution pointing toward the nearest neighbor As atoms. The

non-bonding characteristic of these states is clearly demonstrated.
Fig. 5.13 The charge densities of the t2g (left) and eg (right) states for the majority-
spin states. The horizontal axis is in the [110] direction and the vertical axis is the z-axis
of a cube (Pask et al., 2003).
5.2.2.2 MnC
Half metallic phases Among all the compounds studied by Pask et al.
(2003), MnC shows interesting half metallic properties. The majority- and
minority-spin channels reverse their roles for the metallic and insulating
behaviors as compared to the pnictides and chalcogenides. The calculated
magnetic moment is only 1.0 µB /unit-cell which will be explained in sec-
tion 5.2.2.3. Qian et al. (2004b) found that MnC in fact has two different
ranges of lattice constants for which the half metallic properties exist. The
lattice constant dependent total energy and magnetic moment are shown
in Fig. 5.14. From the total energy curve (the lower panel), the optimized
lattice constant is found to be 4.39 Å. The interesting behavior of the mag-
netic moment/unit-cell is also shown. For lattice constant greater than 5.0
Å, the magnetic moment is 3.0 µB /unit-cell and can be accounted for by
the ionic model. For lattice constant between 3.28 and 4.23 Å, there is
another region with integer magnetic moment. The value of the magnetic
moment is 1.0 µB /unit-cell.
Band structure at smaller lattice constants The band structures for

lattice constants at 5.0 and 4.20 Å were calculated (Qian et al., 2004b). The
band structure at a=5.0 Å is similar to the one of MnAs and is not given.
The band structure at a=4.20 Å is shown in Fig. 5.15(a). The lowest s-band
is not shown. The ordering of the bonding, nonbonding, and antibonding
Fig. 5.14 Magnetic moment/unit-cell and total energy as a function of lattice constant
in MnC (Qian et al., 2004b).
states indicated by the energy levels at Γ(Fig. 5.15(b)) are the same as in
MnAs.
Fig. 5.15 (a) The band structure of MnC at a = 4.20 Å, (b) the energy levels at the
Γ-point as a function of lattice constant (Qian et al., 2004b).
Periodic local moment There is another interesting property of MnC

at lattice constants between 3.60 and 4.28 Å. Under such a condition, the
magnetic moment of MnC is still determined by the majority-spin states.
However, the majority-spin states exhibit insulating properties. The cor-
responding charge density is localized in the bonding region and around
the Mn atom. Therefore, the local magnetic moment can show a periodic
distribution. At low temperature, these moments can have spin wave exci-
tations. An electron in the minority-spin channel occupying a state at EF
can scatter with the spin waves at low temperature. It is anticipated that
the resistivity should show a T5 characteristic (Qian et al., 2004b).
Because of these interesting properties, it is worthwhile to explore the
possibility of growing this compound by choosing the proper substrates.
BP, BN, and SiC have been suggested (Fong and Qian, 2004; Fong et al.,
2008). Based on the results predicted by Fong and Qian (2004), only a
monolayer MnC could be grown on a SiC(100) substrate.
Underlying interactions for half metallicity in ZB structure There

are three types of interactions involved in the half metallicity in TM com-
pounds with ZB structure. The first is due to the environment of the
atoms in the unit cell—the crystal field. The second is due to the interac-
tion between nearest neighbors—the d-p hybridization. And the third is a
magnetic interaction—the exchange interaction.
Crystal field As shown in Fig. 1.9, each of the two atoms in a unit cell is
surrounded by four neighbors forming a tetrahedron. Neighboring ions
of the TM atom exert Coulomb fields on the d-states of the atom. The
effect is to split the five-fold degeneracy of the d-states into triply de-
generate t2g and doubly degenerate eg states. Therefore, a gap between
the t2g and eg states is formed.
d-p hybridization The triply degenerate t2g states of the TM element are
composed of dxy , dyz , and dzx states. Their linear combinations form
orbitals pointing toward the nearest neighbors of the metal element.
The sp3 orbitals of the non-metal atoms point their lobes toward their
nn. The orientations of a d-orbital and sp3 orbital are shown schemat-
ically in Fig. 1.9. The d- and p-states overlap and hybridize. This
hybridization forms the bonding and antibonding gap characterized by
the two Γ15 states at k = 0 in the band structure. These two interac-
tions can be summarized in terms of energy levels as shown in Fig. 1.10.
The Coulomb fields due to the neighboring TM elements cause the s-
and the p-states of the As atom to form sp3 directional orbitals. Under
this circumstance, the p-states remain degenerate.
Exchange interaction Up to this point, the bands in a compound are spin
degenerate. Now, the exchange interaction lifts the degeneracy of the
↑ and ↓ states and shifts the ↓ state energies upwards.
With the proper combinations of these three interactions, EF intersects

one or more of the bands of the majority-spin channel and falls in a gap
of the minority-spin states. The compound is then a HM. A reversal of
roles played by the majority- and minority-spin states is also possible due
to variations of the three interactions, as manifested in MnC.
The strengths of the three interactions can be estimated from the band
structure. Take MnAs as an example. The energy difference between Γ12
and the lowest energy Γ15 states can be used to specify the strength of the
crystal field. The difference of energies for the bonding and antibonding
Γ15 states can be used to estimate the magnitude of the hybridization. A
more refined approach is to determine first the average energies for the
states derived from the two Γ15 states and then take the difference. The
energy difference of the lowest Γ15 states for the majority- and minority-
spin channels is related to the strength of the exchange interaction.
To understand the smaller lattice constant of MnC in terms of the
strength of the three interactions, we compare the band structures shown in
Fig. 5.12 and Fig. 5.15(a). The major differences shown in Fig. 5.15(a) are:
(i) EF intersects the minority-spin states, (ii) the gap is formed within the
majority-spin states. The lattice dependent energy levels at the Γ-point
are shown in Fig. 5.15(b). For the majority-spin channel, the bonding-
nonbonding and bonding-antibonding gaps increase as the lattice constant
decreases. These results indicate that the strength of the crystal field and
p-d hybridization increase under compression. The minority-spin states
show similar behavior, except that the nonbonding (Γ12 ) state is lowered
with compression. This may be due to the fact that the Γ12 states are
not occupied. The increase of dispersion of the bonding and antibonding
states with respect to the case of larger lattice constant cause EF to shift.
The strength of the exchange interaction, gauged by the differences of the
lower Γ15 states of both ↑ and ↓ states, is lessened as the lattice constant
decreases.
5.2.2.3 Qualitative explanation of magnetic moments

The magnetic moment of MnAs is 4.0 µB /unit-cell which can be qualita-
tively understood in terms of the so called “ionic model” (Schwarz, 1986)
based on:
Atomic configurations The atomic configurations of the valence states

of the Mn and As atoms are (3d)5 (4s)2 and (4s)2 (4p)3 , respectively.
Charge transfer Since As is a group-V element, it has a large ionicity.

Three electrons from the Mn atom are transferred to the As atom to form
bonding and antibonding states through d-p hybridization. These three
electrons pair their spins with p states from the As atom and therefore do
not contribute to the magnetic moment of the crystal. Four d-electrons
remain at the Mn site.
Hund’s first rule Since d-states can accommodate five electrons in each
spin channel, the four electrons can align their spins as required by Hund’s
first rule. This is essentially the manifestation of Pauli’s exclusion principle
and the minimization of the Coulomb interaction considering the position
and spin as two dynamical variables to specify the states of the electrons.
Since the four d-electrons have the same spin states, they cannot occupy the
same spatial position. They are kept apart from one another. Consequently,
the effect is to reduce the Coulomb repulsion. The compound can therefore
have lower energy. The total spin moment of the compound is 2~. With
the g-factor of 2 for each electron, the resultant magnetic moment is 4.0
µB /unit-cell. The contribution from the orbital motion to the magnetic
moment is negligible due to electrons occupying 3d states.
It should be noted that for a HM the magnetic moment/unit-cell should
be an integer. The reasons are as follows:
• The total number of electrons is an integer in a compound.
• One of the spin channels, say the ↓ channel, is insulating. Therefore, it
is necessary to have an integer number of electrons, N↓ , to fill up the
top of the valence band.
• The number of electrons in the conducting channel, N↑ , must then be
an integer.
The magnetic moment/unit-cell is:
M = N↑ − N↓ . (5.2)
Therefore, for any theoretical prediction of a compound to be a HM, it is
necessary to have an integer value of the magnetic moment/unit-cell.
The above explanation applies to MnC having a large lattice constant.
Can we understand why the spin moment changes from 3.0 to 1.0 µB /unit-
cell in MnC as its lattice constant decreases? Let us start with the case
having a larger lattice constant, 5.0 Å. From the ionic model, the three elec-
trons remaining at the Mn atom align their spins. Two of them occupy the
Γ12 -related states and one occupies the low energy tail of the antibonding
state (see Fig. 5.15 for reference). At this lattice constant, there is ample
space for the electrons to spatially avoid each other consistent with Pauli’s
principle. As the lattice constant decreases, the charges of the d-states
around the Mn atom are pushed closer to the C atom as manifested by
the increase of the bonding (lower Γ15 states) and antibonding (upper Γ15
states) gap and the crystal field effect (Fig. 5.15(b)). As the lattice constant
decreases further, we take the value of 4.20 Å as an example, the space for
the three d-electrons to align their spins becomes so limited that one of the
three majority-spin electrons flips its spin. This effect is demonstrated in
Fig. 5.15(b) by the lowering in energy of the Γ12 states in the minority-spin
channel to accommodate this spin-flipped electron. To be more explicit,
the magnetic charge (spin density) distribution, defined as the difference
between the ↑ spin and ↓ spin charge densities, of the Γ12 related states
in the minority-spin channel is plotted in Fig. 5.16. The section of the
charge distribution is the (110) plane including the zig-zag chain of Mn and
C atoms. The dot-dash contour denotes the zero magnetic charge density.
The solid and the dotted contours exhibit the positive and negative values
of the density.
A possible method of detecting volume-dependent half metallic phases
is to carry out the measurement of the saturation magnetic moment of the
sample. If the saturation magnetization is over 500 emu/cm3 , then the
sample has 3 µB /unit-cell. If the measured result is around 180 emu/cm3 ,
the corresponding saturation magnetization is 1 µB /unit-cell.
Fig. 5.16 Magnetic charge densities of MnC at 4.20 Å in the (110) plane including the
zig-zag chain of the Mn and C atoms.
5.3 Half metallic superlattices
In addition to CrAs/GaAs multilayered structures grown by Akinaga and

Mizuguchi (2004), there are two types of superlattices studied theoretically.
One is the superlattice composed of half metallic TM pnictides having the
ZB structure (Fong et al., 2004). The other one is a combination of one
half metallic pnictide with a semiconductor (Fong et al., 2008). We shall
discuss them separately.
5.3.0.4 CrAs/MnAs superlattices

The first theoretical report on the (CrAs)n /(MnAs))n superlattices was by
Fong et al. (2004) with n = 1 and 2. Since the qualitative features for these
two cases are very similar, the following discussions will be focused on n =
1. In Fig. 5.17, the supercell for (CrAs)1 /(MnAs)
√ 1 is√shown. They three√
axes
√ of the supercell are defined as: a1 = (a/ 2, a/ 2, 0), a2 = (-a/ 2,
a/ 2, 0), and a3 =(0, 0, 1), where a is the cubic edge of the conventional
cubic cell. The Cr atoms are shown by the gray circles, the Mn atoms are
denoted by the filled circles and the As atoms are depicted by the open
circles. The value of a in Fig. 5.17 is optimized to be 5.70 Å and is in
between the lattice constant of CrAs (5.66 Å) and MnAs (5.77 Å).
Fig. 5.17 The supercell model of (CrAs)1 /(MnAs)1 .
Total and partial density of states The total and partial DOS (PDOS)
of the (CrAs)1 /(MnAs)1 superlattice are shown in Fig. 5.18. The As occu-
pied s-states form bands below 10 eV from EF which is set to be zero. For
the occupied d-states, the Mn d-manifold is located about 2 eV lower than
those of the Cr d-states. This is expected because Mn has valence 7 while
Cr has 6. The eg states of both atoms are higher than the center of gravity
of the t2g states. As shown in Fig. 1.10, under the crystal field the t2g states
have higher energy than the eg states. The t2g states in the superlattice
are the hybridized bonding states. Another feature shown in the PDOS is
the eg states of both metallic elements in the minority-spin channel located
above EF . The valence states of this channel are predominantly the As p-
and Mn t2g -hybridized states.
Charge densities Total charge densities in a section containing the zig-

zag chain of the two TM elements and As atom are shown in Fig. 5.19
Fig. 5.18 Total and partial DOS of the majority- and minority-spin channels. TDOS
are in units of states/eV-unit-cell.
for the two spin channels. The contours reflect the features of the DOS
and PDOS. In the majority-spin channel, the bond charges are explicitly
indicated by contours located between the Mn and As atoms. For the
minority-spin states, contours around the two TM elements exhibit the
four-lobe characteristic of the d-states.
Fig. 5.19 Total valence charge densities of (CrAs)1 /(MnAs)1 of majority-spin states
(upper panel) and minority-spin channel (lower panel) in a section containing the zig-
zag chain of the two TM elements and As atom.
Magnetic moment and magnetization One of the key quantities for

spintronic applications is the magnetic moment. Based on the assumption
that the exchange splitting is sufficient to push the minority eg states above
EF , with minority anion-s and p-t2g hybrid states fully occupied, Pask et al.
(2003) suggested the following relation for the magnetic moment/unit-cell
(M) of a HM in the ZB structure:
M = Nmajority − Nminority = Ztotal − 2Nminority , (5.3)

where Ztotal is the total number of valence electrons, and Nmajority and
Nminority are the number of the majority- and minority- spin states, re-
spectively. For (CrAs)1 /(MnAs)1 , Ztotal = 23 and Nminority = 8 give M =
7 µB . Similarly, Eq. (5.3) gives a moment of 14 µB for (CrAs)2 /(MnAs)2 .
These values are confirmed by numerical calculations. They are the sum of
M for each constituent compound making up the superlattice. It is impor-
tant to note that the increased magnetic moment/unit-cell for larger cells
implied by Eq. (5.3) does not necessary mean an increase of the saturation
magnetization. The latter is defined as the magnetic moment density. Both
superlattices have the same saturation magnetization of 672.8 emu/cm3 .
Comparing the saturation magnetizations of CrAs (572.4 emu/cm3 , 3.0
µB )—which agrees well with the measured 560 emu/cm3 —and MnAs (763.2
emu/cm3 , 4.0 µB ), the calculated value for the superlattices is approxi-
mately the average of the two constituents. Therefore, it is not viable to
increase the saturation magnetization by growing superlattices. These re-
sults are summarized in Table 5.5 along with DOS at EF for the majority-
spin channel and energy gap (Eg ) of the minority-spin states.
Table 5.5 Total DOS at EF of the majority-spin channel, Eg of the minor-

ity-spin states, the magnetic moment/unit-cell for (CrAs)1 /(MnAs)1 , and
(CrAs)2 /(MnAs)2 . Superscript a is referred to Pask et al. (2003).
Sample DOS at EF (states/eV-cell) Eg (eV) M (µB )
(CrAs)1 (MnAs)1 1.94 1.65 7.0
(CrAs)2 (MnAs)2 3.47 1.62 14.0
CrAsa 0.85 1.85 3.0
MnAsa 0.77 1.70 4.0
Substrates for growth To determine possible substrates for growing

CrAs/MnAs superlattices, Fong et al. (2004) calculated the half metallicity
for a range of lattice constants. Between 5.60 and 6.03 Å, the superlattices
retain their half metallic properties. This range spans the experimental
lattice constants of AlAs (5.62 Å), GaAs (5.65 Å), and InP (5.81 Å). These
semiconductors can serve as substrates imposing the least strain for growing
CrAs/MnAs superlattices to exhibit half metallic properties.
5.3.0.5 Superlattice showing spin-polarized ballistic transport

Unit cell In all theoretical efforts, the objective has been whether the
sample exhibits half metallic properties. Qian et al. (2004b) designed a
half metallic superlattice which can exhibit ballistic transport properties.
The supercell consists of one layer of GaAs, one layer of MnAs, two layers
of CrAs, and two capping layers of GaAs. The supercell model is shown in
Fig. 5.20. The unit cell is outlined by the solid and dashed lines. The a-axis
is along
√ the [110] direction of the conventional cell in the ZB structure. |a|
= 2ao and |b| = 3a o , where ao =5.722 Å is the optimized lattice constant
for GaAs.
Fig. 5.20 A superlattice model exhibits the spin-polarized ballistic transport properties.
a is along the [110] direction of the conventional cubic cell (Qian et al., 2004b).
Total and partial density of states The total density of states (TDOS)
and the PDOS are shown in Fig. 5.21. The qualitative features are similar to
those of CrAs/MnAs superlattice discussed in section 5.3.0.4. The majority-
spin states exhibit metallic behavior (the top panel of Fig. 5.21). They are
primarily contributed by the d-states of the Cr atom (the third panel). The
occupied d-manifold of Cr is higher in energy than that of the Mn atom (the
second and the third panels). The eg states of the Cr atom are occupied.
The minority-spin channel exhibits a gap at EF . The gap value is 0.95 eV
and is nearly half of the values of CrAs (1.88 eV) and MnAs (1.74 eV). The
occupied d-states are the bonding t2g states and those eg states are above
EF .
Charge densities In Fig. 5.22, the charge densities are plotted in a sec-
tion consisting of a zig-zag chain of atoms. There are three panels. The
top panel shows the total valence charge density of ↑ spin states. In the
region between the Mn and Cr atoms, the distribution is similar to the
CrAs/MnAs superlattice. The As atoms are located between the labeled
atoms as their nn to form a chain. In the GaAs regions, bond charges
are formed between the Ga atoms and their nn. The middle panel shows
the charge distribution of ↓ spin states. Similar to the CrAs/MnAs case,
Fig. 5.21 Total DOS and PDOS of a superlattice showing the ballistic transport (Qian
et al., 2004b).
the four lobes associated with d-states of Cr and Mn atoms are illustrated.
They are mainly the t2g type of bonding states.
Fig. 5.22 Total charge densities of ↑ spin states (top), ↓ spin states (middle), and the ↑
spin channel (bottom) in an energy ranging between EF and EF + 0.3 eV. The section
contains the zig-zag chain of the TM elements, Ga, and As atoms (Qian et al., 2004b).
Ballistic conductance In order to see the conducting channel of this su-

perlattice along the direction perpendicular to the layers, Qian et al. (2005)
plotted the charge density of the states in the majority-spin channel and
within 0.3 eV above EF . This charge distribution exhibits the coherent
extended feature allowing an electron injected into the superlattice at the
left side of GaAs region by an electric field whose strength does not disturb
the charge distribution in the superlattice. Consequently, the electron trav-
els through the superlattice without suffering any scattering. In addition,
they also calculated Fermi velocity in the travelling direction. This velocity
relates to the ballistic conductance Gσ by the following relation:
Ae2 1 X
Z
Gσ = hΨkn,σ |Pz | Ψkn,σ i δ(εkn,σ − EF )d3 k, (5.4)
2m 8π 3 n
where σ labels the spin channel. A is a finite cross section, e is the electronic
charge, Pz is the z-component of momentum operator. Ψkn,σ is the wave
function for band n and at a k point inside the BZ with energy εkn,σ . The
results of the conductance for the two spin channels are shown in Fig. 5.23.
The upper panel is for the ↑ channel and shows a finite conductance at
EF . The conductance of the ↓ states is shown in the lower panel. There
is zero conductance at EF . All these results illustrate that the superlattice
shown in Fig. 5.20 can have a spin-polarized ballistic transport. This kind
of superlattices can be used to fabricate spin filters.
Fig. 5.23 Ballistic conductance along the c-axis for the superlattice is shown as a func-
tion of energy (Qian et al., 2005). EF is set to be zero.
5.4 Quantum dots
Advances in nanoscience are based on the studies of superlattices, quantum

dots, and other quantum structures with length on the order of nanome-
ters. Having the possibility of growing HMs in thin-film form, it is natural
to explore the possibilities for spintronic applications using quantum dots.
There have been several attempts (Ono et al., 2002; Okabayashi et al.,
2004) at growing quantum dots using compounds predicted to exhibit half
metallic properties. Theoretical designs of half metallic quantum dots have
also been initiated by Qian et al. (2006b). We shall discuss first the exper-
imental efforts then the theoretical results.
5.4.1 Experiment
Growths of quantum dots composed of TM pnictide, such as MnAs, have
been reported by Ono et al. (2002) and Okabayashi et al. (2004), respec-
tively. We discuss the growths first, then characterizations, and finally, the
experimental determination of the electronic properties.
5.4.1.1 MnAs quantum dots

Growth To grow MnAs quantum dots with the MBE technique, Ono
et al. (2002) used a scheme developed by Akinaga et al. (2000b) for growing
MnSb granular films on the sulfur-passivated GaAs substrate. The reason
for passivating GaAs by a group-VI element is to lower the surface energy so
that self-assembled growth can take place. Ono et al. (2002) passivated the
n+ GaAs(001) substrate with the S atoms by first dipping the substrate
into a solution of (NH4 )2 Sx for one hour then rising the substrate with
pure water. The substrate was then heated to 200 ◦ C. The source beams
were finally switched on with the flux ratio of As/Mn set to be 4–5. To
ensure that the MnAs quantum dots are in the ZB structure, these authors
also grew the NiAs-type bulk MnAs and Ga1−x Mnx As films on GaAs(001)
substrates for comparisons. For the Ga1−x Mnx As films, they grew first
a 15 nm GaAs buffer layer on a GaAs substrate after the removal of an
oxide layer at the surface by heating the substrate to 580 ◦ C. Then the
temperature of the buffer layer was cooled to 200 ◦ C. Another GaAs buffer
layer of 10 nm was grown on top. Finally, the films were grown. For the
alloy and bulk films, the ratio of As/Mn is also set at 4–5.
Characterization of samples After rinsing away (NH4 )2 Sx and setting

the substrate temperature at 200 ◦ C, the RHEED pattern changed from a
halo to a 1 × 1 streaky form. During the growth of MnAs quantum dots,
the pattern became spotty.
To probe whether the MnAs quantum dots are in the ZB structure,
high resolution cross-sectional TEM and selected area electron diffraction
(SAED) were used. The results are shown in Fig. 5.24. On the left, the
TEM images of two cross-sections of the interfaces between the MnAs and
GaAs substrate are exhibited. The details of the images are very simi-
lar. On the right, the SAED patterns are shown. The patterns for the
two sections are also very similar. These results indicate that both the
quantum dots and substrate have the same structure. The lattice constant
mismatch between the MnAs and substrate can be also detected by both
TEM and SAED. A 0.7% mismatch was estimated in the plane parallel to
the interface.
Fig. 5.24 The TEM image and SAED patterns of nanoscale MnAs and GaAs substrates
(Ono et al., 2002).
Size distribution The size distribution of MnAs quantum dots can be

determined by high resolution scanning electron microscopy (SEM). The
results obtained by Ono et al. (2002) are shown in Fig. 5.25. Most of the
dots have diameters near 16.3 nm.
Fig. 5.25 Size distribution of MnAs quantum dots determined by SEM (Ono et al.,
2002).
Electronic properties To probe the Mn 3d partial density of states of

the nanoscale MnAs dots, Okabayashi et al. (2004) used the 3p–3d resonant
photoemission spectroscopy. The photon energies were between 46 and 55
eV and the light was not polarized. These authors compared spectra of
different density of quantum dots to those of an in situ prepared alloy in
ZB-like structure and a NiAs-like MnAs film.
The comparisons of the on-resonance (hν=50 eV) photoemission spectra
are shown in Fig. 5.26. When the photon energy is at and above 50 eV,
a peak at 4.0 eV below EF appears in all spectra. A shoulder follows in
the low energy side (around –7 ∼ –9 eV) of the peak only for the alloy and
quantum dots but not for the film. It was therefore concluded that this is
another evidence showing that the quantum dots have the ZB structure.
Fig. 5.26 (a) The on-resonance (hν=50 eV) photoemission spectra of Ga1−x Mnx As
alloy, medium density quantum dots, high density quantum dots, and a NiAs-like MnAs
film. (b) The difference spectra between the on-resonance (hν=50 eV) and off-resonance
(48 eV) spectra can be obtained as a measure of the Mn 3d partial density of states
(Okabayashi et al., 2004).
Magnetic properties Magnetic properties of MnAs quantum dots were

measured by Ono et al. (2002) using SQUID. From Fig. 5.27(a), the fer-
romagnetic ordering in the MnAs quantum dots can be clearly seen in the
hysteresis loop measurement at 50 K. The TC for the quantum dots is
estimated to be 280 K based on the temperature dependence of the mag-
netization measurement (Fig. 5.27(b)).
Fig. 5.27 (a) Hysteresis loop, and (b) the temperature dependence of the magnetization
in MnAs quantum dots (Ono et al., 2002).
NiAs-type bulk MnAs and Ga1−x Mnx As films on GaAs(001)

The bulk MnAs in the NiAs structure is hexagonal. The TEM image and
SAED pattern are distinct from GaAs but similar to MnAs quantum dots
grown on a GaAs(001) substrate without any passivation of the S atoms.
For the alloys, the low temperature growth on the 10 nm buffer layer shows
a 1 × 2 surface reconstruction in the RHEED pattern. The results differ
from those for the quantum dots.
5.4.2 Theory
5.4.2.1 MnAs quantum dot
Qian et al. (2006b) built a model of MnAs quantum dot in the ZB struc-
ture. The model, a cluster, is composed of a total of 41 atoms. They
calculated the electronic properties and magnetic moment using ab initio
pseudopotential method and DFT with GGA exchange-correlation.
Cluster model The cluster was built by choosing a center atom in the
cubic crystalline MnAs and by retaining up to the third shells of the
atoms. In terms of number atoms in each shell, the cluster is specified
as MnAs4 Mn12 As24 . It has a radius of 6.1 Å. To saturate the dangling
bonds at the edge of the cluster, 60 hydrogen atoms were used. Since these
H atoms are attached to the outer shell As atoms, their effective charge is
chosen at 0.75 e, where e is the electron charge. The cluster is put in a
cubic supercell having a lattice constant of 22.96 Å, four times the size of
of the bulk MnAs with the ZB structure. The nearest neighbor distance
between atoms in neighboring clusters is 4.06 Å.
Effects of relaxation The cluster was relaxed in both ferromagnetic

and antiferromagnetic configurations with respect to the shells of the Mn
atoms. After relaxation with the ferromagnetic configuration, the aver-
age bond length of MnAs4 at the center is 2.491 Å, which is larger than
the unrelaxed value of 2.486 Å. When the spin of the center Mn atom is
flipped (antiferromagnetic configuration), the corresponding average bond
length is reduced to 2.470 Å. This contraction is caused by the so-called
exchange striction (Solovyev and Terakura, 2003). The total energy of the
antiferromagnetic configuration is lower by 125 meV with respect to the
ferromagnetic configuration. Based on this result, the antiferromagnetic
phase is energetically more stable than the ferromagnetic phase.
Electronic properties and magnetic moment The total and partial

DOS for the spin channels were calculated and are shown in Fig. 5.28 for
the ferromagnetic phase (FM) and antiferromagnetic phase (AFM), respec-
tively. Note that the AFM is defined as the spin orientations of the center
Mn and the rest Mn atoms are opposite. The partial DOS are the projected
DOS on one atom. As usual, EF is set to be zero. The major difference
between FM and AFM is that in FM the minority-spin channel exhibits a
gap of 1.83 eV as compared to 0.97 eV in AFM. The calculated magnetic
moments for FM and AFM are 52 µB /cluster and 42 µB /cluster, respec-
tively. With the integer values of magnetic moments and gaps appearing in
the minority-spin channel, the cluster in either the ferromagnetic or antifer-
romagnetic phase can be a HM based on the criterion of having an integer
magnetic moment of the sample.
Fig. 5.28 TDOS and PDOS for the ferromagnetic phase (a) and antiferromagnetic phase
(b) of MnAs quantum dot (Qian et al., 2006b) .
The major contribution of d-states in both ferromagnetic and antiferro-

magnetic phases is in the region between –1.0 and –4.5 eV. In the majority-
spin channel, they contribute dominantly in the region between –3.2 and
–4.0 eV (Fig. 5.28(a) and (b)). By examining the PDOS, there appears a
difference between the two magnetic phases. In Fig. 5.28(a), these states
are originated from the t2g states in the majority-spin channel of the Mn
atom at the center of the cluster and the second shell Mn atoms. These d-
states hybridize with p-states of their neighboring As atoms. The p-states
are concentrated at –4.0 eV. In Fig. 5.28(b), on the other hand, the t2g
states are primarily from the minority-spin states of the second shell Mn
atoms. These d-states hybridize with p-states of the As atoms.
Comparing to the spin-unpolarized photoemission data, the theoretical
results can account for the measured peak shown in Fig. 5.26. From the
PDOS in Fig. 5.28, the origin of the peak can be identified from the hy-
bridized bonding p-states of the As atoms. Near the region of the measured
peak, d-states of the Mn atom contribute as well. It will be interesting to
carry out spin-polarized photoemission measurements to distinguish the fer-

romagnetic and antiferromagnetic phases and other measurements, such as
PCAR and positron annihilation, to show half metallic properties in these
MnAs quantum dots. If the ferromagnetic phase is the stable one and the
HM can be exhibited then these quantum dots can potentially be one of
the future materials for spintronic devices.
5.5 Digital ferromagnetic heterostructures
Digital ferromagnetic heterostructures (DFH) is defined as an ideal δ-

doping semiconductor with TM elements. It was first attempted by
Kawakami et al. (2000) to grow such heterostructure in the form of doping
GaAs(100) with Mn. It is now called (Ga,Mn)As-DFH. A theoretical study
of this DFH was carried out by Sanvito and Hill (2001). The possible en-
hancement of ferromagnetic coupling in Mn/GaAs-DFH has been examined
by Wang and Qian (2006). (Ga,Mn)Sb-DFH has been grown by Chen et al.
(2002). Recently, Qian et al. (2006a) designed a DFH involving a δ-layer
doping Si with Mn atoms. It is called Si-based DFH. The interstitial sites
for Mn in Si was investigated by Wu et al. (2007).
5.5.1 Experiment
In the following, the growth and characterization of (Ga,Mn)As- and
(Ga,Mn)Sb-DFH will be discussed first, then focus will be on the asso-
ciated physical properties.
5.5.1.1 Growth and characterization of (Ga,Mn)As-DFH

The MBE method was employed by Kawakami et al. (2000) to grow
(Ga,Mn)As-DFH. It is important to use this method to control the positions
of Mn atoms in the growth direction. The growth chamber was an ultra-
high vacuum Varian/EPI Gen-II system. By starting with a GaAs(100)
substrate, a 250 nm GaAs buffer layer was grown on top at 580 ◦ C. There
was a 2 × 4 reconstruction detected by RHEED at this stage. The substrate
is then cooled to 300 ◦ C in an As2 flux. Then the RHEED pattern changed
to a c(4 × 4) reconstruction. The actual growth temperature was between
240 and 280 ◦ C. Another 100 nm GaAs buffer layer was grown to allow
the substrate temperature to equilibrate with the Ga shutter open. The
RHEED pattern showed a 1 × 1 reconstruction. The DFH is then grown
by alternately opening the Ga and Mn shutters while holding a constant

As2 flux. A valve on the cracking source maintained at 700 ◦ C controlled
the flux. The ratio of As/Ga was about 25. During the deposition of Mn
atoms, the RHEED pattern changed to a 1 × 2 reconstruction, then the
1 × 1 pattern was recovered when Ga was deposited. The RHEED oscil-
lations were monitored to determine the deposition rates, about 0.5 ML/s
for GaAs and 0.06 ML/s for MnAs. The DFH is capped by a GaAs layer
having about 40 nm thickness. During the growth, measurements of the
optical absorption spectrum of the GaAs substrate were used to monitor
its temperature. The temperature of the substrate was also measured by a
thermocouple. Its reading was 30 to 40 ◦ C lower than the actual temper-
ature due to the radiative heating of the Ga cell. The final samples have
about 100 repetitions of 10 ML of GaAs and 0.5 ML of Mn. It is labeled
as (10/0.5)100 .
The (10/0.5)100 sample was subjected to X-ray diffraction analysis. The
plot of X-ray counts as a function of 2θ is shown in Fig. 5.29 for the sample
grown at 280 ◦ C. The DFH(0) peak is at 2θ = 66◦ indicating a 0.4% ex-
pansion of the lattice constant along the growth direction. The two small
peaks labeled as ±1 reflect the periodicity of the DFH. Transmission elec-
tron microscopy (TEM) measures the image of the cross-section at the
interface. The high resolution TEM images show that Mn is distributed
over a thickness of 3–5 ML for all growth temperatures.
Luo et al. (2002) used MBE to grow (Ga,Mn)As-DFH with a temper-
ature variation in growing buffer layers. The growth started with a GaAs
buffer layer of 200 Å on GaAs(100) substrate at 580 ◦ C. The substrate tem-
perature was then brought down to 275 ◦ C and a layer of low-temperature
GaAs was grown to a thickness of a few hundred Å. Finally, (Ga,Mn)As-
DFH was prepared first by MBE to grow GaAs on buffer layers and then
atomic layer epitaxy (ALE) for Mn layers. They used RHEED to monitor
the growth, in particular, for the MnAs precipitates. Typically, samples of
9 layers of GaAs with 0.2–0.5 ML Mn were obtained.
5.5.1.2 Growth and characterization of (Ga,Mn)Sb-DFH

Chen et al. (2002) used MBE to grow (Ga,Mn)Sb-DFH on a GaAs(100)
substrate. To avoid the large lattice constant mismatch (7.5%) between
the GaSb and GaAs, a 500 nm GaSb buffer layer was grown on a GaAs
substrate. RHEED was used to monitor the growth. The samples have
about 50 periods of 0.5 ML of Mn atoms with various GaSb thickness.
Fig. 5.29 X-ray diffraction result of Fig. 5.30 The TEM image of a
(10/0.5)100 DFH grown at 280 ◦ C (12/0.5)50 -DFH, the Mn atoms are con-
(Kawakami et al., 2000). tained in dark lines (Chen et al., 2002).
TEM image of a (12/0.5)50 DFH is shown in Fig. 5.30. The light and dark
lines are clearly identifiable. The dark lines are those containing the Mn
atoms.
5.5.1.3 Physical properties of (Ga,Mn)As-DFH

The properties probed by experimental means are the magnetic proper-
ties and TC on (Ga,Mn)As-DFH. For (Ga,Mn)Sb-DFH, the magnetic and
transport properties have been measured.
Fig. 5.31 (a) The SQUID hysteresis loop for a (10/0.5)100 (Ga,Mn)As-DFH, and (b)
the remanent magnetization as a function of temperature (Kawakami et al., 2000).
Magnetic properties of (Ga,Mn)As-DFH The magnetic properties

were measured by Kawakami et al. (2000) using a SQUID magnetometry.
In Fig. 5.31, the hysteresis loop of a (10/0.5)100 DFH grown at 280 ◦ C is
compared to the one for a dilute alloy Ga0.949 Mn0.051 As. The experiment
was carried out at 5 K. The magnetic field is along the [100] in-plane axis—
the easy axis. The results are presented with the contribution of a linear
background from GaAs being subtracted. Solid lines are for the DFH, while
the dashed lines are for the alloy. There is a 100% remanence showing the
ferromagnetic ordering along an easy magnetization axis. The temperature
dependence of the remanent magnetization is shown in Fig. 5.31(b). The
magnetization M was initially achieved by applying an external magnetic
field of 1000 Oe at 5 K. Then, measurements of M(T) were carried out
at zero field as temperature increases. At about and beyond 50 K, the
magnetic moment of the DFH is zero. The alloy magnetization persists up
to about 70 K.
SQUID magnetometry was used by Luo et al. (2002) to measure the
magnetizations of the samples. There are some differences between the
samples grown by Luo et al. (2002) and those synthesized by Kawakami
et al. (2000). Instead of using M, Luo et al. (2002) defined the term called
“effective spin density” (ESD) from the saturation magnetization expressed
in terms of the total spin aligned by the field per unit-area. The experiments
were carried out at 5 K with magnetic fields up to 55 kGauss (kG). ESD
plotted as a function of Mn fraction is shown in Fig. 5.32, where ESD
exhibits a maximum value of 40 in units of 1013 /cm2 at Mn fraction 0.25.
At Mn fraction 0.45, ESD is only about 10 × 1013 /cm2 .
Fig. 5.32 Effective spin density (ESD) of a (Ga,Mn)As-DFH as a function of Mn con-

centration in the layer (Luo et al., 2002).
Magnetic properties of (Ga,Mn)Sb-DFH For (Ga,Mn)Sb-DFH,

Chen et al. (2002) also used a SQUID magnetometer to measure its magne-
tization. All the samples show hysteresis loops up to 400 K—the limiting
temperature of their SQUID magnetometer. The results of the magnetic
moment at different temperatures are given in Fig. 5.33. These authors con-
cluded that the ferromagnetic properties exhibited at RT are contributed
by DFH not by any MnSb precipitates. According to Abe et al. (2000), the
room temperature ferromagnetism has been observed in GaMnSb if there
are MnSb precipitates. But such samples show temperature independent
hysteresis loops with the coercive field at RT to be the same value as at 5
K. The results shown in Fig. 5.33 do not display such features in GaMnSb
with precipitates.
Fig. 5.33 (a) Hysteresis loops of (Ga,Mn)Sb measured at different temperatures. (b)
Magnetization as a function of temperature (Chen et al., 2002).
The temperature dependences of magnetization given in Figs. 5.31(b)

and 5.33(b) are quite different. It is difficult to compare shapes of the
magnetic moment between (Ga,Mn)As- and (Ga,Mn)Sb-DFH’s. On the
other hand, the magnetizations M(T) shown in both figures expose the
difference between these two types of DFH’s. M(T) in (Ga,Mn)As-DFH
is not zero at T > 50 K, while M(T) is zero for (Ga,Mn)Sb-DFH. The
temperature dependent M(T) are significantly different.
Curie temperature of GaAs- and GaSb-based DFH For

(Ga,Mn)As-DFH, Kawakami et al. (2000) reported measurements of TC
vs. GaAs thickness (Fig. 5.34). All the samples were grown at 280 ◦ C.
There is a factor of two difference between values of TC for samples having
0.5 ML of MnAs coverage (50 K) and the one with 0.25 ML of MnAs cov-
erage (22 K) on 10 to 15 layers of GaAs. TC of the 0.5 ML case decreases
from 50 K as the GaAs layer thickness increases. It reaches a minimum
(35 K) at about 50 layers of GaAs. Then, the value attained a maximum
(40 K) at 100 layers. At 200 layers, TC reduces to 33 K. These authors
remarked that TC is sensitive to the growth temperature.
The (Ga,Mn)As-DFH samples obtained by Luo et al. (2002) reach the
maximum TC of 37 K at 0.25 ML of Mn. This value is higher than the one
measured by Kawakami et al. (2000). However, with samples at 0.40 ML of
Fig. 5.34 Curie temperature (TC ) as a function of GaAs layer thickness for (10/0.5)100
and (10/0.25)100 DFH’s (Kawakami et al., 2000).
Mn, the value determined by Kawakami et al. (2000) is higher by about 10

K. Whether the discrepancy can be attributed to the thicker GaAs buffer
layer used by Luo et al. (2002) remains to be answered.
Curie temperature of DFH’s based on GaSb were estimated from the
measurements of hysteresis loops by Chen et al. (2002). Their estimated
value is higher than 400 K. The result seems to be consistent with the
measurements on MnAs and MnSb in the NiAs structure. The former has
a TC at 310 K and the latter is at 580 K. We summarize the values of TC
in these known III-V compound based DFH’s in Table 5.6.
Table 5.6 TC ’s of (Ga,Mn)As- and (Ga,Mn)Sb-DFH’s. The thickness of the substrate or buffer
layers is represented in monolayers (ML).
DFH Thickness (ML) Growth temp. (◦ C) TC (K) Reference
(Ga,Mn)As
(10/0.5)100 10–15 280 50 Kawakami et al. (2000)
(10/0.5)100 100 280 40 Kawakami et al. (2000)
(10/0.5)100 200 280 33 Kawakami et al. (2000)
(10/0.25)100 100 280 22 Kawakami et al. (2000)
(10/0.5)100 260 19 Kawakami et al. (2000)
(10/0.5)100 240 5 Kawakami et al. (2000)
(9/0.25) > 100 275 33 Luo et al. (2002)
(9/0.4) > 100 275 17 Luo et al. (2002)
(Ga,Mn)Sb
(various/0.5)50 > 400 Chen et al. (2002)
Transport properties of (Ga,Mn)As- and (Ga,Mn)Sb-DFH Two

types of measurements were carried out by Luo et al. (2002) on (Ga,Mn)As-
DFH and by Chen et al. (2002) on (Ga,Mn)Sb-DFH: magnetoresistance
(MR) and anomalous Hall effect (AHE). The measurements were carried
out using the van der Pauw configurations in a cryostat having a 170
kG (kGauss) superconducting magnet and Hall bar configuration. All the
(Ga,Mn)As-DFH samples do not exhibit any metallic features. The sheet
resistance of a sample, (9/0.5), is fitted well with
1
ln R = ln Ro − (T/To ) 2 , (5.5)
where To is 61 K and Ro is 850 Ω. This exponential dependence of R with
respect to T is known for materials such as an n-type δ-doped GaAs where
the conduction is contributed by a variable range of hopping. On the other
hand, all the (Ga,Mn)Sb-DFH’s are metallic based on the MR results. This
is in distinct contrast to the (Ga,Mn)As-DFH case.
Anomalous Hall effect (AHE) is able to probe the information about
the interaction between mobile carriers and local spin moment of magnetic
elements at low external magnetic field. The expression of the transverse
(perpendicular to the external applied electric field) resistivity ρxy contains
a term induced by the magnetization M.
ρxy = Ro B + 4πRa M, (5.6)
where Ro is the normal Hall coefficient, B is the strength of the external
magnetic field, and Ra is anomalous Hall coefficient. The first term is the
normal Hall effect. The second term defines the anomalous Hall effect.
In Fig. 5.35, the sheet and Hall resistances of (Ga,Mn)As-DFH and the
sheet and anomalous Hall resistance for (Ga,Mn)Sb-DFH are shown. For
(Ga,Mn)As-DFH, the sheet resistance decreases with temperature. The
shapes do not change significantly as the temperature varies from 12 to 52
K. They are peaked at B = 0. With the four-terminal van der Pauw tech-
nique, the Hall resistance Rxy is equivalent to the Hall resistivity ρxy (Met-
alidis and Bruno, 2006). The magnitude of Rxy (RHall ) given in Fig. 5.35
for a (Ga,Mn)As-DFH decreases with temperature. Rxy changes from neg-
ative to positive value in a very narrow (∼ 20 kG) region when the positive
external magnetic field reverses its sign. The question of the physical ori-
gins causing the temperature behavior, the double structures in the sheet
resistances, and their small Rxy values compared to the large ones (one
order of magnitude larger) is not answered.
Since the above-mentioned question is not answered, we just point out
the difference of the magnetotransport properties of the two (Ga,Mn)-
pnictide-DFH’s. For (Ga,Mn)Sb-DFH, both the sheet and Hall resistances
behave quite differently from the case of (Ga,Mn)As-DFH. Let us first com-
pare the sheet resistances shown in Figs. 5.35(a) and 5.36(a). At low tem-
perature (T < 40 K), the shapes are somewhat similar. However, the
Fig. 5.35 Sheet resistance of a (Ga,Mn)As-DFH at different temperatures as a function

of external magnetic field (a), and Hall resistance of a (Ga,Mn)As-DFH at different
temperatures as a function of external magnetic field (b) (Luo et al., 2002).
magnitudes differ by more than a factor of 10 at about 20 K. In addi-

tion, those of (Ga,Mn)Sb-DFH’s do not exhibit the double structure. The
shapes of Rsheet shown in Fig. 5.36 are also changed drastically at higher
temperature (T ≥ 40 K). Instead of decreasing, the sheet resistances in-
crease as the magnetic field increases positively and decreases negatively.
Two qualitative features of Rxy in both DFH’s are in agreement: (i) The
decrease in magnitudes of the resistance as the temperature increases, (ii)
the sharp reverse of Rxy (change from positive to negative) vs. external
magnetic field at low temperature. At high temperature (T = 400 K), Rxy
of (Ga,Mn)Sb-DFH shows a smooth transition from positive to negative as
the magnetic field changes from negative to positive. No such behavior was
detected for temperature range between 21 and 32 K, just below the TC of
(Ga,Mn)As-DFH.
Fig. 5.36 (a) Sheet resistance of a (Ga,Mn)Sb-DFH at different temperatures as a

function of external magnetic field, and (b) the Hall resistivity at T = 4 and 400 K as a
function of external magnetic field (Chen et al., 2002).
5.5.2 Theory
5.5.2.1 GaAs-based DFH
Structure Immediately after publication of the experimental work by
Kawakami et al. (2000), Sanvito and Hill (2001) reported their theoretical
study of an (NGaAs /1)∞ (Ga,Mn)As-DFH, where NGaAs is the number of
GaAs layers and ∞ stands for a superlattice. They used SIESTA algorithm
with the local spin density approximation (LSDA) (Sánchez-Portal et al.,
1997). The experimentally determined lattice constant of GaAs, 5.65 Å,
was used. These authors also examined antisite defects.
Electronic and magnetic properties The band structures near EF for

the majority- and minority-spin channels of a (15/1)∞ (Ga,Mn)As-DFH are
shown in Fig. 5.37(a) and (b), respectively. The directions of k are in the
plane perpendicular to the GaAs layers. X1 is at the edge of the square BZ
and X2 is located at the corner, the diagonal, of the square. The ↑ spin
states show metallic behavior while the ↓ spin channel exhibits a gap of 0.65
eV. The DFH is a two-dimensional HM. The band dispersions along the two
directions show anisotropy. There is no identification of the origins of those
states. The widths of those bands intercepted by EF become narrower as
the separation between Mn layers increases.
Fig. 5.37 Band structures near EF of the (a) majority- and (b) minority-spin channels
for a (15/1)∞ (Ga,Mn)As-DFH (Sanvito and Hill, 2001).
By comparing the ferromagnetic and antiferromagnetic configurations

of the Mn atoms, the authors found that the ferromagnetic configuration
is more favorable. The coupling strength is stronger in a DFH than in
a random Ga1−x Mnx As alloy. The energy difference ∆F A between ferro-
and antif-erromagnetic configurations is 515 meV in the (15/1)∞ structure
while ∆F A in the random alloy is only 160 meV. For DFH’s, this energy
difference depends weakly on the GaAs thickness for small N. When N =

4, 6, and 8, the values of ∆F A are 531, 533, and 515 meV, respectively.
Defects Sanvito and Hill (2001) investigated the effects of As antisites,

that is to have As atoms occupy Ga sites. In the supercell model of (11/1)∞
they replaced half of the Ga atoms by the As atoms in the layer close to
the MnAs layer. The ferromagnetic phase is still favored with ∆F A to be
70 meV. They cited the Zener model in which the coupling between the
hole spin and local spin moment at the Mn atom causes the long-range
ferromagnetic behavior.
Transport properties The ballistic transport properties of (Ga,Mn)As-

DFH’s were investigated by Sanvito and Hill (2001). They calculated the
current in the Mn plane termed as CIP and the one perpendicular to the
Mn plane, CPP. The Landauer-Büttiker formula (Büttiker et al., 1985) was
used to calculate the conductance of each of the two spin channels.
2e X
T r tσ (k)t†σ )k) ,

Γσ = (5.7)
h k
where T r is the notation of trace, tσ is the transmission matrix for spin

σ. The transmission matrix is the scattering matrix with matrix elements
between the incoming and outgoing waves.
The results of the calculated spin-polarized CIP conductances of a
(15/1)∞ DFH are shown in Fig. 5.38. The ↑ spin channel of the CIP
conductance is finite in the region around EF . From the partial CIP con-
ductance, the contributions are dominated by p-states of the As atoms.
The t2g states of the Mn atoms contribute less. The eg states of the Mn
atom hardly make a significant contribution. These results are consistent
with the crystal field and hybridization effects. The contributions from As
p-states and Mn t2g states are the consequence of hybridization. With the
tetrahedral symmetry, the eg states are the nonbonding states. They do not
hybridize strongly with p-states of the neighboring As atoms (Fig. 1.10).
The ↓ spin states do not contribute to any conductance near EF . The DFH
is a 2-D HM. Because of the thick GaAs (15) layers, it is not expected to
have any CPP conductance.
5.5.2.2 Si-based DFH

Structure From what we have discussed so far, it is evident that enor-
mous efforts have been devoted to the study of HMs. However, at this
moment, we still do not have any realization of spintronic devices using
Fig. 5.38 Total and partial spin-polarized CIP and CPP conductances of a (15/1)∞
DFH (Sanvito and Hill, 2001).
HMs. The major obstacle is the growth of Heusler alloys, oxides, and TM
pnictides. Then, there is the T∗ (∼ 88 K) at which the half metallicity
disappears in a Heusler alloy due to spin-flip transitions and the fact that
P vanishes precipitously as the temperature approaches RT.
Qian et al. (2006a) proposed a DFH by doping a δ-layer of Mn atoms in
Si (Si-based DFH) motivated by the fact that Si technologies are the most
mature among the semiconductors so the growth problems can possibly
be eliminated. They used a supercell model of a (31/1)∞ DFH shown in
Fig. 5.39. The planewave pseudopotential method, and the GGA exchange-
correlation within DFT, were used to calculate the density of states of Si-
based DFH for the investigation of whether there is a T∗ in this DFH.
Fig. 5.39 A supercell model of a (31/1)∞ DFH (Qian et al., 2006a).
Effect of lattice relaxation The presence of 1.0 ML of Mn in 31 layers

of Si necessitates examining various properties affected by the lattice relax-
ation. The effects on the magnetic moment, DOS at EF in the conducting
channel, ∆F A of the Mn atoms, and energy gaps are summarized in Ta-

ble 5.7. The magnetic moment is 3.0 µB with finite DOS at EF (N(EF )) for
the majority-spin channel. EF M and EAF M are the total energies for the
ferromagnetic and antiferromagnetic configurations, respectively. Eg is the
energy gap in the minority-spin states. Relaxation of the atoms inside the
unit cell maintains the magnetic moment/Mn atom and improves N(EF )
and Eg . EF M is reduced but the ferromagnetic phase is still favorable.
Table 5.7 Comparison of the unrelaxed and relaxed (31/1)∞ -DFH. m is the
magnetic moment per Mn atom, N(EF ) is the DOS of the conducting channel
at EF , ∆F A = EF M - EAF M , and Eg is the energy gap.
Structure m (µB ) N(EF ) (states/eV-cell) ∆F A (meV) Eg (eV)
Unrelaxed 3.0 1.06 –523.91 0.20
Relaxed 3.0 1.25 –442.38 0.25
Density of states The spin-polarized DOS can manifest whether a sam-

ple is a HM. The calculated results are shown in Fig. 5.40. The top panel
shows the total density of states. The second one is for a Mn atom. The
third and fourth panels are for the Si atoms located at the nearest and
second layers of the Mn layer. At EF , TDOS shows that the majority-spin
channel is metallic while the minority states exhibits insulating behavior.
The gap is 0.25 eV. Combining with the integer spin moment per unit cell
(3 µB /unit-cell), the DFH is a half metal.
Fig. 5.40 TDOS (states/eV-unit-cell) and PDOS (states/eV) for a (31/1)∞ DFH.
PDOS are for the nearest and second nearest neighbor Si atoms (Qian et al., 2006a).
The PDOS enable one to identify states around EF . With the energy
window between EF –1.0 and EF +2.0 eV, the majority-spin states are pri-
marily the hybridized p-states of the nearest neighbor Si atom (Si(I)) to
the Mn atom and the t2g states of the Mn atom (the second and the third
panels). As will be shown later, these are part of the antibonding states.
The strongest region of the hybridization is at 2.6 eV below EF . The eg
states of the Si(I) atom are centered around –2.5 eV with a width of about
1.0 eV. In the range between EF –1.0 and EF +2.0 eV, the second neighbor
Si(II) atom has hardly any significant contributions. Between EF –2.0 and
EF –4.0 eV, those p-states have a uniform distribution.
In the minority-spin channel, those states just below the valence band
maximum are the p(Si(I))-d(Mn(t2g )) hybridized bonding states. EF is
located near the top of the valence bands. The eg states of Mn atom are
not occupied. The conduction band minimum is contributed by the low
energy end of the eg states. Neither Si(I) nor Si(II) have contributions to
the states at the conduction band minimum.
Band structures and Fermi surfaces The band structures near EF of

the ↑ and ↓ spin channels are plotted in Figs. 5.41(a) and (b), respectively.
Γ–Z is in the direction perpendicular to the layers. The top of the valence
band along this direction is relatively flat as compared to lower energy
bands. There are gaps at the Γ- and Z-points. Along this direction, there
is no conduction. Point R is at the corner of the two dimensional BZ.
The X point is along the [100] direction. The occupied bands closest to
EF at these two points are derived from the antibonding states at the Γ
point. Therefore, the Γ point is expected to be surrounded by a hole Fermi
surface while the R point should have electron pockets. The Fermi surfaces
thus can have both electron and hole surfaces in the two dimensional BZ
and are shown in Fig. 5.42. In Fig. 5.42(a) the contour surrounding the Γ
point is the hole surface. The curves at the R point are the boundaries of
electron surfaces. The hole charge distribution for states near EF is shown
in Fig. 5.42(b).
T∗ in Si-based DFH According to recent experiments on dilutely doped

Mn in Si, Mnx Si1−x alloys, TC can be above 400 K (Zhang et al., 2004;
Bolduc et al., 2005; Bandaru et al., 2006). It is especially encouraging
because Bandaru et al. (2006) verified that the high TC and ferromagnetism
are intrinsic properties, not due to clustering of Mn atoms. A relevant
issue, the question of whether this DFH can have TC at or above RT,
needs to be addressed in order to determine whether this DFH will be a
potential spintronic material. In Fig. 5.43, we make use of the one shown
in section 3.6.1.2 (Fig. 3.12). EF of the Heusler alloy is denoted by “EF of
HA” and is located right below the conduction band edge (CBMin) in the
Fig. 5.41 (a) The band structure of the majority-spin states and (b) the band structure
of the minority-spin channel (Qian et al., 2006a).
Fig. 5.42 (a) The Fermi surface in the X-Y section of the two dimensional BZ, and (b)
the hole charge distribution for the states near EF (Qian et al., 2006a).
↓ spin channel.
We now focus on the Si-based DFH. Its EF is explicitly labeled and
is located near the valence band maximum (VBM) of the ↓ spin states.
The gap of the insulating minority-spin states is denoted by ∆↓ . The
important gap is labeled as δ which is measured from EF to CBMin in the
minority-spin channel. Its value is approximately 0.25 eV (Table 5.7) which
is about 10 times larger than RT. Therefore, the probability is drastically
reduced for an electron in the majority-spin channel to make a spin-flip
transition from EF to CBMin of the minority-spin channel. However, one
can ask the following question: what about the possibility for an electron
from VBM to make a spin-flip transition to a state at EF ? As we see
from the PDOS (Fig. 5.40), the valence band is dominated by p-states of
Si(I). Consequently, one expects the corresponding charge density to be
concentrated near the Si atom. The overlap between the Si(I) p-states
and Mn d-states will be small. As shown in Fig. 5.40, the hole states
Fig. 5.43 Schematic diagram of the DOS of a HM with the majority-spin channel show-
ing metallic behavior. The Fermi energy of the Heusler alloy is now indicated as “EF
of HA” and is right below CBMin. The Fermi energy of the Si-based DFH is labeled
as EF and is located close to the valence band maximum (VBM) of the DFH. δ is the
energy difference between CBMin and EF . ∆↓ is the fundamental gap of the insulating
channel.
near EF have their charge distributions concentrated on the Mn atoms.

The transition matrix element reduces the probability of this kind of spin-
flip transition. The conclusion is that there will not be a T∗ in the Si-
based DFH. Therefore, It can be a promising material for making spintronic
devices.
5.6 One dimensional half metals
Even though DFH’s behave as two dimensional systems, their basic struc-
tures are still three dimensional. To explore the possibility of finding half
metallicity in truly low dimensional systems, Dag et al. (2005) investigated
one dimensional systems by doping TM elements in carbon wires, Cn (TM).
They used a pseudopotential method within DFT with spin-polarized GGA
exchange-correlation.
Model The wires are modeled by a tetragonal supercell with a = b = 10

Å and c depends on the length of the carbon wires. Sometimes, c is set to
be twice the length of the wires if Peierls instability (Peierls, 1955) is the
concern. While they covered different n′ s and many TM elements, such as
Cr, Fe, Mn and Ti, they reported mainly on Cn Cr with n ranging from 2
to 7. These results are summarized in Table 5.8.
Half metallic properties and magnetic moments Except the CCr

compound, the other compounds with n ≥ 2 are HMs. These systems favor
Table 5.8 Summary of calculated total energy difference

∆ET between the spin-polarized and spin-unpolarized cases,
optimized value of c, magnetic moment/unit-cell (M), and
the sample type: semiconductor (S) or a half metal (HM).
For a HM, the energy gap is also given.
Compound ∆ET c M Type Gap (eV)
(eV) (Å) (µB ) ↑ ↓
CCr –1.8 3.7 2.0 S 0.7 1.0
C2 Cr –2.8 5.2 4.0 HM 3.3
C3 Cr –3.0 6.5 4.0 HM 0.4
C4 Cr –3.0 7.9 4.0 HM 2.9
C5 Cr –2.5 9.0 4.0 HM 0.6
C6 Cr –3.1 10.3 4.0 HM 2.4
C7 Cr –2.5 11.6 4.0 HM 0.5
spin polarization and have lower energies than the corresponding unpolar-
ized cases. The values of the magnetic moment/unit-cell (4.0 µB ) are larger
than the one for MnC with larger lattice constant. A simple explanation is
that each of the two neighboring C atoms transfers one electron from the
Cr atom to doubly occupy its p-orbital. It is interesting to note that the
half metallic behavior depends on n. The metallic feature alternates be-
tween the majority- and minority-spin states as n increases from 2 to 7. For
even n, the majority-spin channel is metallic. In addition to the electronic
properties, these authors also investigated the stability of the wires and the
effects of spin-orbit interaction. Based on their findings, they suggested a
way to grow these HMs.
Spin-polarized band structure and DOS The spin-polarized band

structures and density of states of C3 Cr and C4 Cr are shown in Fig. 5.44,
because of their different half metallic properties as compared to other
compounds. The band structures are plotted along the direction of the
wire, z-direction (Γ–Z). The solid lines are for the majority (↑)-spin states
and dashed lines are for the minority (↓)-spin channel. EF is set to be zero.
vσ and mσ are the highest valence and lowest conduction bands for spin-σ
channel, respectively. The two mσ bands are intercepted by EF having
different spin orientations. cσ is the next conduction band with energy
larger than the mσ state. If the TM element has symmetric environment,
n=4, a band in the majority-spin channel is intercepted by EF . On the
other hand in the asymmetric cases, n=3, the band intercepted by EF
belongs to the minority-spin states. Looking closer from the bond lengths,
double bonds are formed between all atoms for n=3, where the bond lengths
in Å are distributed as C-1.28-C-1.28-C-1.95-Cr. For n=4, triple and single
bonds form alternately between C atoms and a longer bond between the
Cr atom and its neighbors. The distributions are C-1.25-C-1.33-C-1.25-C-
2.1-Cr. The different bond lengths between the Cr and C atoms result in
different band structures.
From the PDOS, the mσ bands are formed by Cr-3d states hybridized
with 2p-state of the neighboring C atoms. The vσ band is composed of s-
states and d-states of the Cr atom. The unoccupied cσ band is contributed
by the hybridized antibonding C p-states and Cr d-states.
Fig. 5.44 The spin-polarized band structures and density of states of (a) C3 Cr and (b)
C4 Cr (Dag et al., 2005).
Effect of spin-orbit interaction To accurately investigate the spin-

orbit (S-O) interaction, the total energy difference was calculated using
WIEN2K. It is found that the effect is small, –7.9 meV with the S-O in-
teraction relative to the non-S-O case. The TDOS with and without S-O
interaction for the case C3 Cr are shown in Fig. 5.45. EF is set at zero. The
shift of the states in the majority-spin states due to the S-O interaction
does not destroy the half metallicity.
Stability These authors carried out an extensive search of various struc-

tures. Local minima of the total energy were probed by optimizing the
structure starting from a transversely displaced chain of atoms at various
lattice constants. The linear chain structure has been found to be stable
and is more favorable than a zig-zag structure. They also calculated the
phonon frequencies Ω. They are all positive. For n=3, ΩT O (k=0) = 89,
92, 411 cm−1 and ΩLO (k=0) = 421, 1272, 1680 cm−1 . The corresponding
frequencies for n= 4 are 13, 71, 353, 492 cm−1 and 489, 1074, 1944, 2102
cm−1 . However, for n=9 some of the frequencies become negative. To as-
sure the stability of the chains with small n (< 9), they also carried out
Fig. 5.45 TDOS of the two spin channels with (solid lines) and without (shaded areas)
S-O interaction for C3 Cr (Dag et al., 2005).
ab initio molecular dynamics simulations with T set between 750 and 1000
K using Nosé thermostat. The atoms were displaced in random directions.
The results confirm the stability of the chains at small n.
There are still the possibilities of the Peierls instability and strain issues.
These authors expanded the models by a factor 2 and calculated the band
structures. The splitting of bands at EF should manifest the effect of the
Peierls instability. There is no splitting of metallic bands at EF in Cn Cr
chain structures. Axial strains were applied to examine whether strain can
destroy half metallicity. The half metallicity is robust for n = 4 under ǫz
= ±0.05. However, for n = 5, its half metallic properties remain only at
ǫz ≤0.05. With ǫz = -0.05, C5 Cr is transformed to a ferromagnetic metal.
At ǫz = 0.10, C3 Cr is a semiconductor while it becomes a ferromagnetic
metal having the magnetic moment of 3.1 µB /unit-cell when ǫz = -0.10.
Fig. 5.46 (a) Energetics of doping path for C5 Cr, and (b) a possible path of growing
C7 Cr wire (Dag et al., 2005).
Growth In Fig. 5.46(a), the doping path gives the energetics of a process
of doping Cr in C5 . The steps are indicated as A, B, C, and D. From A to
D, there is a barrier of 1.86 eV. This is considered to be a large barrier to
overcome. Therefore, the doping process is not a viable method to grow a
half metallic wire. The growth path in Fig. 5.46(b) suggests a possible way
to grow a C7 Cr wire without a large barrier. The process A shows to first
grow a C7 chain, then a Cr atom is deposited at one end of the C7 . The
processes of growing C7 Cr wire by adding more C atoms to one end of Cr
atom are indicated as B and C. From the interaction energy with distance
d there is no significant barrier.
Appendix A
Appendix
A.1 Anisotropic magnetoresistance
In a ferromagnetic metal, the resistivity (the resistance multiplied by cross-

section and divided by height for a cylindrical sample) can change when the
magnetization (M ) of the metal changes under (i) change of temperature
and/or (ii) change of direction of external field. Anisotropic magnetore-
sistance (AMR) arises under condition (ii). Experimentally, one measures
the resistivities, ρk and ρ⊥ vs. applied magnetic field Hext , where ρk is
measured parallel and ρ⊥ is measured perpendicular to the applied field.
AMR is defined by ∆ρ = ρk − ρ⊥ when both change slopes under Hext .
In Fig. A.1, the resistivities at A and B are used to define the AMR. The
normalized AMR is defined as ∆ρ/ρav , where ρav = (1/3)ρk + (2/3)ρ⊥ .
The advantage of using the normalized AMR is that it is dimensionless.
Thus, there is no need to know the dimension of the sample.
Fig. A.1 The resistivity of a ferromagnetic metal vs. Hext . The resistivities at A and
B are used to define ∆ρ.
243

Bibliography
Abe, E., Matsukura, F., Yasuda, H., Ohno, Y. and Ohno, H. (2000). Molecular
beam epitaxy of III–V diluted magnetic semiconductor (Ga,Mn)Sb, Physica
E: Low-dimensional Systems and Nanostructures 7, pp. 981–985.
Akai, H., Akai, M. and Kanamori, J. (1985). Electronic structure of impurities
in ferromagnetic iron. II. 3d and 4d impurities, J. Phys. Soc. Jpn. 54, pp.
4257–4264.
Akai, H. and Dederichs, P. H. (1993). Local moment disorder in ferromagnetic
alloys, Phys. Rev. B 47, pp. 8739–8747.
Akinaga, H., Manago, T. and Shirai, M. (2000a). Material design of half-metallic
zinc-blende CrAs and the synthesis by molecular-beam epitaxy, Jpn. J.
Appl. Phys. Lett. 39, p. L1118.
Akinaga, H. and Mizuguchi, M. (2004). Zinc-blende CrAs/GaAs multilayers
grown by molecular-beam epitaxy, J. Phys.: Condens. Mat. 16, pp. S5549–
S5553.
Akinaga, H., Mizuguchi, M., Ono, K. and Oshima, M. (2000b). Room-
temperature thousandfold magnetoresistance change in MnSb granular
films: Magnetoresistive switch effect, Appl. Phys. Lett. 76, pp. 357–359.
Albert, F. J., Katine, J. A., Buhrman, R. A. and Ralph, D. C. (2000). Spin-
polarized current switching of a Co thin film nanomagnet, Appl. Phys. Lett.
77, pp. 3809–3811.
Alonso, J. A., Martinez-Lope, M. J., Casais, M. T., Macmanus-Driscoll, J. L.,
de Silva, P. S. I. P. N., Cohen, L. F. and Fernandez-Diaz, M. T. (1997).
Non-stoichiometry, structural defects and properties of LaMnO3+δ with
high δ values (0.11 ≤ δ ≤ 0.29), J. Mater. Chem. 7, pp. 2139–2144.
Ambrose, T., Krebs, J. and Prinz, G. (2000). Epitaxial growth and magnetic
properties of single-crystal Co2 MnGe Heusler alloy films on GaAs(001),
Appl. Phys. Lett. 76, pp. 3280–3282.
Andersen, O. K. (1975). Linear methods in band theory, Phys. Rev. B 12, pp.
3060–3083.
Anderson, P. W. (1950). Antiferromagnetism. theory of superexchange interac-
tion, Phys. Rev. 79, pp. 350–356.
Anderson, P. W. (1956). Ordering and antiferromagnetism in ferrites, Phys. Rev.
245
102, pp. 1008–1013.

Anderson, P. W. and Hasegawa, H. (1955). Considerations on double exchange,
Phys. Rev. 100, pp. 675–681.
Andreev, A. (1964). Thermal conductivity of the intermediate state of supercon-
ductors, Sov. Phys. - JETP 19, pp. 1228–1231.
Anguelouch, A., Gupta, A., Xiao, G., Abraham, D., Ji, Y., Ingvarsson, S. and
Chien, C. L. (2001). Near-complete spin polarization in atomically-smooth
chromium-dioxide epitaxial films prepared using a CVD liquid precursor,
Phys. Rev. B 64, p. 180408(R).
Anisimov, V. I., Aryasetiawan, F. and Lichtenstein, A. I. (1997). First-principles
calculations of the electronic structure and spectra of strongly correlated
systems: the LDA + U method, J. Phys.: Condens. Mat. 9, p. 767.
Anisimov, V. I., Elfimov, I. S., Hamada, N. and Terakura, K. (1996). Charge-
ordered insulating state of Fe3 O4 from first-principles electronic structure
calculations, Phys. Rev. B 54, pp. 4387–4390.
Antonov, V., Dürr, H., Kucherenko, Y., Bekenov, L. and Yaresko, A. (2005).
Theoretical study of the electronic and magnetic structures of the Heusler
alloys Co2 Cr1−x Fex Al, Phys. Rev. B 72, p. 054441.
Antonov, V., Oppeneer, P., Yaresko, A., Perlov, A. and Kraft, T. (1997). Com-
putationally based explanation of the peculiar magneto-optical properties
of PtMnSb and related ternary compounds, Phys. Rev. B 56, p. 13012.
Antonov, V. N., Harmon, B. N., Antropov, V. P., Perlov, A. Y. and Yaresko, A. N.
(2001). Electronic structure and magneto-optical Kerr effect of Fe3 O4 and
Mg2+ - or Al3+ -substituted Fe3 O4 , Phys. Rev. B 64, p. 134410.
Asamitsu, A., Moritomo, Y., Kumai, R., Tomioka, Y. and Tokura, Y. (1996).
Magnetostructural phase transitions in La1−x Srx MnO3 with controlled car-
rier density, Phys. Rev. B 54, pp. 1716–1723.
Austin, B., Heine, V. and Sham, L. (1962). General theory of pseudopotentials,
Phys. Rev. 127, p. 276.
Bader, S. D. (1991). SMOKE, J. Magn. Magn. Mater. 100, pp. 440–454.
Bahr, D., Press, W., Jebasinski, R. and Mantl, S. (1993). X-ray reflectivity and
diffuse-scattering study of CoSi2 layers in Si produced by ion-beam synthe-
sis, Phys. Rev. B 47, p. 4385.
Baibich, M. N., Broto, J. M., Fert, A., Dau, F. N. V., Petroff, F., Etienne, P.,
Creuzet, G., Friederich, A. and Chazelas, J. (1988). Giant magnetoresis-
tance of (001)Fe/(001)Cr magnetic superlattices, Phys. Rev. Lett. 61, p.
2472.
Bandaru, P. R., Park, J., Lee, J. S., Tang, Y. J., Chen, L.-H., Jin, S., Song, S. A.
and O’Brien, J. R. (2006). Enhanced room temperature ferromagnetism in
Co- and Mn-ion-implanted silicon, Appl. Phys. Lett. 89, p. 112502.
Barth, J., Fecher, G. H., Balke, B., Ouardi, S., Graf, T., Felser, C., Shkabko,
A., Weidenkaff, A., Klaer, P., Elmers, H. J., Yoshikawa, H., Ueda, S. and
Kobayashi, K. (2010). Itinerant half-metallic ferromagnets Co2 TiZ (Z=Si,
Ge, Sn) : Ab initio calculations and measurement of the electronic structure
and transport properties, Phys. Rev. B 81, p. 064404.
Bauer, E. (2010). Phänomenologische theorie der kristallabscheidung an
Bibliography 247
oberflächen. I, Zeitschrift für Kristallographie 110, pp. 372–394.

Becke, A. D. (1988). Density-functional exchange-energy approximation with cor-
rect asymptotic behavior, Phys. Rev. A 38, pp. 3098–3100.
Beeby, J. L. (1967). The density of electrons in a perfect or imperfect lattice,
Proc. R. Soc. Lond. A 302, pp. 113–136.
Bezryadin, A., Verschueren, A., Tans, S. and Dekker, C. (1998). Multiprobe trans-
port experiments on individual single-wall carbon nanotubes, Phys. Rev.
Lett. 80, p. 4036.
Binnig, G., Rohrer, H., Gerber, C. and Weibel, E. (1983). 7 x 7 reconstruction
on Si(111) resolved in real space, Phys. Rev. Lett. 50, p. 120.
Blöchl, P. E. (1994). Projector augmented-wave method, Phys. Rev. B 50, pp.
17953–17979, doi:10.1103/PhysRevB.50.17953.
Block, T., Carey, M., Gurney, B. and Jepsen, O. (2004). Band-structure calcu-
lations of the half-metallic ferromagnetism and structural stability of full-
and half-Heusler phases, Phys. Rev. B 70, p. 205114.
Blonder, G. E., Tinkham, M. and Klapwijk, T. M. (1982). Transition from metal-
lic to tunneling regimes in superconducting microconstrictions: Excess cur-
rent, charge imbalance, and supercurrent conversion, Phys. Rev. B 25, pp.
4515–4532.
Bohm, D. (1951). Quantum Theory (Prentice Hall Inc., New Jessery).
Bolduc, M., Awo-Affouda, C., Stollenwerk, A., Huang, M. B., Ramos, F. G.,
Agnello, G. and LaBella, V. P. (2005). Above room temperature ferromag-
netism in Mn-ion implanted Si, Phys. Rev. B 71, p. 033302.
Bollero, A., Ziese, M., Höhne, R., Semmelhack, H. C., Köhler, U., Setzer, A. and
Esquinazi, P. (2005). Influence of thickness on microstructural and magnetic
properties in Fe3 O4 thin films produced by PLD, J. Magn. Magn. Mater.
285, pp. 279–289.
Bona, G. L., Meier, F., Taborelli, M., Bucher, E. and Schmidt, P. H. (1985).
Spin polarized photoemission from NiMnSb, Solid State Communications
56, pp. 391–394.
Borca, C. N., Komesu, T., Jeong, H.-K., Dowben, P. A., Ristoiu, D., Horde-
quin, C., Nozières, J. P., Pierre, J., Stadler, S. and Idzerda, Y. U. (2001).
Evidence for temperature dependent moments ordering in ferromagnetic
NiMnSb(100), Phys. Rev. B 64, p. 052409.
Bouzerar, R., Bouzerar, G. and Ziman, T. (2006). Why RKKY exchange in-
tegrals are inappropriate to describe ferromagnetism in diluted magnetic
semiconductors, Phys. Rev. B 73, p. 024411.
Bowen, M., Barthélémy, A., Bibes, M., Jacquet, E., Contour, J.-P., Fert, A.,
Ciccacci, F., Duò, L. and Bertacco, R. (2005). Spin-polarized tunneling
spectroscopy in tunnel junctions with half-metallic electrodes, Phys. Rev.
Lett. 95, p. 137203.
Bowen, M., Bibes, M., Barthelemy, A., Contour, J.-P., Anane, A., Lemaitre, Y.
and Fert, A. (2003). Nearly total spin polarization in La2/3 Sr1/3 MnO3 from
tunneling experiments, Appl. Phys. Lett. 82, pp. 233–235.
Braun, J. (1996). The theory of angle-resolved ultraviolet photoemission and its
applications to ordered materials, Reports on Progress in Physics 59.
Brener, N. E., Tyler, J. M., Callaway, J., Bagayoko, D. and Zhao, G. L. (2000).
Electronic structure and Fermi surface of CrO2 , Phys. Rev. B 61, pp.
16582–16588.
Brown, P. J., Neumann, K. U., Webster, P. J. and Ziebeck, K. R. A. (2000). The
magnetization distributions in some Heusler alloys proposed as half-metallic
ferromagnets, J. Phys.: Condens. Mat. 12, p. 1827.
Bruno, P. (2003). Exchange interaction parameters and adiabatic spin-wave spec-
tra of ferromagnets: A“renormalized magnetic force theore”, Phys. Rev.
Lett. 90, p. 087205.
Buschow, K. H. J. and van Engen, P. G. (1981). Magnetic and magneto-optical
properties of Heusler alloys based on aluminium and gallium, J. Magn.
Magn. Mater. 25, pp. 90–96.
Buschow, K. H. J., van Engen, P. G. and Jongebreur, R. (1983). Magneto-optical
properties of metallic ferromagnetic materials, J. Magn. Magn. Mater. 38,
pp. 1–22.
Bussmann, K., Prinz, G. A., Cheng, S.-F. and Wang, D. (1999). Switching of ver-
tical giant magnetoresistance devices by current through the device, Appl.
Phys. Lett. 75, pp. 2476–2478.
Büttiker, M., Imry, Y., Landauer, R. and Pinhas, S. (1985). Generalized many-
channel conductance formula with application to small rings, Phys. Rev. B
31, pp. 6207–6215.
Caballero, J., Park, Y., Cabbibo, A., Childress, J., Petroff, F. and Morel, R.
(1997). Deposition of high-quality NiMnSb magnetic thin films at moderate
temperatures, J. Appl. Phys. 81, pp. 2740–2744.
Cai, Y. Q., Ritter, M., Weiss, W. and Bradshaw, A. M. (1998). Valence-band
structure of epitaxially grown Fe3 O4 (111) films, Phys. Rev. B 58, pp. 5043–
5051.
Callen, H. (1963). Green function theory of ferromagnetism, Phys. Rev. 130, p.
890.
Caminat, P., Valerio, E., Autric, M., Grigorescu, C. and Monnereau, O. (2004).
Double beam pulse laser deposition of NiMnSb thin films at ambient tem-
perature, Thin Solid Films 453-454, pp. 269–272.
Carbonari, A. W., Pendl, W., Attili, R. N. and Saxena, R. N. (1993). Magnetic
hyperfine fields in the heusler alloys Co2 YZ (Y=Sc, Ti, Hf, V, Nb; Z=Al,
Ga, Si, Ge, Sn), Hyperfine Interactions 80, pp. 971–976.
Carra, P., Thole, B. T., Altarelli, M. and Wang, X. (1993). X-ray circular dichro-
ism and local magnetic fields, Phys. Rev. Lett. 70, pp. 694–697.
Celotta, R., Pierce, D., Wang, G., Bader, S. and Felcher, G. (1979). Surface
magnetization of ferromagnetic ni(110): A polarized low-energy electron
diffraction experiment, Phys. Rev. Lett. 43, p. 728.
Ceperley, D. and Alder, B. (1980). Ground state of the electron gas by a stochastic
method, Phys. Rev. Lett. 45, p. 566.
Chang, L. and Ploog, K. (1985). Molecular beam epitaxy and heterostructures,
NATO Adv. Sci. Inst. Ser. E87.
Chen, C. T., Idzerda, Y. U., Lin, H. J., Smith, N. V., Meigs, G., Chaban, E.,
H, G., Ho, Pellegrin, E. and Sette, F. (1995). Experimental Confirmation
Bibliography 249
of the X-Ray Magnetic Circular Dichroism Sum Rules for Iron and Cobalt,
Physical Review Letters 75, pp. 152–155.
Chen, X., Na, M., Cheon, M., Wang, S., Luo, H., McCombe, B. D., Liu, X.,
Sasaki, Y., Wojtowicz, T., Furdyna, J. K., Potashnik, S. J. and Schiffer,
P. (2002). Above-room-temperature ferromagnetism in GaSb/Mn digital
alloys, Appl. Phys. Lett. 81, pp. 511–513.
Cheng, S., Nadgomy, B., Bussmann, K., Carpenter, E., Das, B., Trotter, G.,
Raphael, M. and Harris, V. (2001). Growth and magnetic properties of
single crystal Co2 MnX (X=Si,Ge) Heusler alloys, IEEE Transactions on
Magnetics 37, pp. 2176–2178, doi:10.1109/20.951116.
Clowes, S., Miyoshi, Y., Bugoslavsky, Y., Branford, W., Grigorescu, C., Manea,
S., Monnereau, O. and Cohen, L. (2004). Spin polarization of the transport
current at the free surface of bulk NiMnSb, Phys. Rev. B 69, p. 214425.
Coey, J. M. D. and Venkatesan, M. (2002). Half-metallic ferromagnetism: Exam-
ple of CrO2 (invited), J. Appl. Phys. 91, pp. 8345–8350.
Coey, J. M. D., Versluijs, J. J. and Venkatesan, M. (2002). Half-metallic oxide
point contacts, J. Phys. D: Appl. Phys. 35, pp. 2457–2466.
Continenza, A., Picozzi, S., Geng, W. T. and Freeman, A. J. (2001). Coordination
and chemical effects on the structural, electronic, and magnetic properties
in Mn pnictides, Phys. Rev. B 64, p. 085204.
Şaşoǧlu, E., Galanakis, I., Sandratskii, L. M. and Bruno, P. (2005a). Stability
of ferromagnetism in the half-metallic pnictides and similar compounds: a
first-principles study, J. Phys.: Condens. Mat. 17, p. 3915.
Şaşoǧlu, E., Sandratskii, L. M., Bruno, P. and Galanakis, I. (2005b). Exchange
interactions and temperature dependence of magnetization in half-metallic
Heusler alloys, Phys. Rev. B 72, p. 184415.
Dag, S., Tongay, S., Yildirim, T., Durgun, E., Senger, R., Fong, C. Y. and Ciraci,
S. (2005). Half-metallic properties of atomic chains of carbon-transition
metal compounds, Phys. Rev. B 72, p. 155444.
Datta, S. and Das, B. (1990). Electronic analog of the electrooptic modulator,
Appl. Phys. Lett. 56, pp. 665–667.
Daughton, J. (1992). Magnetoresistive memory technology, Thin Solid Films 216,
pp. 162–168.
de Groot, R. A., Mueller, F., Engen, P. and Buschow, K. (1983). New class of
materials: Half-metallic ferromagnets, Phys. Rev. Lett. 50, pp. 2024–2027.
de Wijs, G. A. and de Groot, R. A. (2001). Towards 100% spin-polarized charge-
injection: The half-metallic NiMnSb/CdS interface, Phys. Rev. B 64, p.
020402.
Dedkov, Y. (2004). Spin-Resolved Photoelectron Spectroscopy of Oxidic Half-
Metallic Ferromagnets and Oxide/Ferromagnet Interfaces, Ph.D. thesis.
Dedkov, Y. S., Rüdiger, U. and Güntherodt, G. (2002). Evidence for the half-
metallic ferromagnetic state of Fe3 O4 by spin-resolved photoelectron spec-
troscopy, Phys. Rev. B 65, p. 064417.
Dieny, B., Humbert, P., Speriosu, V. S., Metin, S., Gurney, B. A., Baugart,
P. and Lefakis, H. (1992). Giant magnetoresistance of magnetically soft
sandwiches: Dependence on temperature and on layer thicknesses, Phys.
Rev. B 45, pp. 806–813.

Dowben, P. A. and Skomski, R. (2004). Are half-metallic ferromagnets half met-
als? (invited), J. Appl. Phys. 95, pp. 7453–7458.
Dupree, T. H. (????). Electron scattering in a crystal lattice, Ann. Phys.-New
York 15, pp. 63–78.
Durgun, E., Senger, R. T., Mehrez, H., Dag, S. and Ciraci, S. (2006). Nanospin-
tronic properties of carbon-cobalt atomic chains, EPL (Europhysics Letters)
73, p. 642.
Eerenstein, W., Palstra, T. T. M., Hibma, T. and Celotto, S. (2002). Origin of the
increased resistivity in epitaxial Fe3 O4 films, Phys. Rev. B 66, p. 201101.
Elmers, H., Wurmehl, S., Fecher, G., Jakob, G., Felser, C. and Schönhense,
G. (2004). Field dependence of orbital magnetic moments in the Heusler
compounds Co2 FeAl and Co2 Cr0.6 Fe0.4 Al, Appl. Phys. A–Mater. 79, pp.
557–563.
Etgens, V. H., de Camargo, P. C., Eddrief, M., Mattana, R., George, J. M. and
Garreau, Y. (2004). Structure of ferromagnetic CrAs epilayers grown on
GaAs(001), Phys. Rev. Lett. 92, p. 167205.
Feng, J. S. Y., Pashley, R. D. and Nicolet, M. A. (1975). Magnetoelectric proper-
ties of magnetite thin films, J. Phys. C: Solid State Phys. 8, pp. 1010–1023.
Foner, S. (1956). Vibrating sample magnetometer, Rev. Sci. Instrum. 27, pp.
548–548.
Fong, C., Qian, M., Pask, J. E., Yang, L. and Dag, S. (2004). Electronic and
magnetic properties of zinc blende half-metal superlattices, Appl. Phys.
Lett. 84, p. 239.
Fong, C. Y. and Cohen, M. L. (1970). Energy band structure of copper by the
empirical pseudopotential method, Phys. Rev. Lett. 24, p. 306.
Fong, C. Y., Qian, M., Liu, K., Yang, L. H. and Pask, J. (2008). Design of
spintronic materials with simple structures, J. Nanosci. Nanotechnol. 8, p.
3652.
Fong, C. Y. and Qian, M. C. (2004). New spintronic superlattices composed of
half-metallic compounds with zinc-blende structure, J. Phys.: Condens.
Mat. 16, pp. S5669–S5676.
Fuji, S., Sugimurat, S., Ishidat, S. and Asano, S. (1990). Hyperfine fields and
electronic structures of the Heusler alloys Co2 MnX (X=Al, Ga, Si, Ge,
Sn), J. Phys.: Condens. Mat. 2, p. 8583.
Furutani, Y., Nishihara, H., Kanomata, T., Kobayashi, K., Kainuma, R., Ishida,
K., Koyama, K., Watanabe, K. and Goto, T. (2009). Field-induced-moment
nuclear coupling for 59 Co in a Heusler alloyCo2 TiGa, J. Phys.: Conf. Ser.
150, p. 042037.
Galanakis, I. (2002a). Surface half-metallicity of CrAs in the zinc-blende struc-
ture, Phys. Rev. B 66, p. 012406.
Galanakis, I. (2002b). Surface properties of the half- and full-Heusler alloys, J.
Phys.: Condens. Mat. 14, pp. 6329–6340.
Galanakis, I. (2005). Orbital magnetism in the half-metallic Heusler alloys, Phys.
Rev. B 71, p. 012413.
Galanakis, I., Dederichs, P. and Papanikolaou, N. (2002a). Origin and properties
Bibliography 251
of the gap in the half-ferromagnetic Heusler alloys, Phys. Rev. B 66, p.

134428.
Galanakis, I., Dederichs, P. and Papanikolaou, N. (2002b). Slater-Pauling behav-
ior and origin of the half-metallicity of the full-Heusler alloys, Phys. Rev.
B 66, p. 174429.
Galanakis, I. and Mavropoulos, P. (2003). Zinc-blende compounds of transition
elements with N, P, As, Sb, S, Se, and Te as half-metallic systems, Phys.
Rev. B 67, p. 104417.
Galanakis, I., Ostanin, S., Alouani, M., Dreyssé’, H. and Wills, J. (2000). Ab
initio ground state and L2,3 x-ray magnetic circular dichroism of Mn-based
Heusler alloys, Phys. Rev. B 61, p. 4093.
Garcı́a, N., noz, M. M. and Zhao, Y.-W. (1999). Magnetoresistance in excess of
200 in ballistic Ni nanocontacts at room temperature and 100 Oe, Phys.
Rev. Lett. 82, pp. 2923–2926.
Gellrich, A. and Kessler, J. (1991). Precision measurement of the Sherman asym-
metry function for electron scattering from gold, Physical Review A 43, pp.
204–216.
Getzlaff, M., Heidemann, B., Bansmann, J., Westphal, C. and Schonhense, G.
(1998). A variable-angle electron spin polarization detection system, Rev.
Sci. Instrum. 69, pp. 3913–3923.
Giapintzakis, J., Grigorescu, C., Klini, A., Manousaki, A., Zorba, V., An-
droulakis, J., Viskadourakis, Z. and Fotakis, C. (2002). Low-temperature
growth of NiMnSb thin films by pulsed-laser deposition, Appl. Phys. Lett.
80, pp. 2716–2718.
Gong, G. Q., Gupta, A., Xiao, G., Qian, W. and Dravid, V. P. (1997). Mag-
netoresistance and magnetic properties of epitaxial magnetite thin films,
Phys. Rev. B 56, pp. 5096–5099.
Goodenough, J. B. (1955). Theory of the role of covalence in the perovskite-type
manganites [La,M(II)]MnO3 , Phys. Rev. 100, pp. 564–573.
Grave, E. D., Persoons, R. M., Vandenberghe, R. E. and de Bakker, P. M. A.
(1993). Mössbauer study of the high-temperature phase of Co-substituted
magnetites, Cox Fe3−x O4 . I. x ≤ 0.04, Phys. Rev. B 47, pp. 5881–5893.
Gridin, V. V., Hearne, G. R. and Honig, J. M. (1996). Magnetoresistance ex-
tremum at the first-order verwey transition in magnetite (Fe3 O4 ), Phys.
Rev. B 53, pp. 15518–15521.
Grollier, J., Cros, V., Hamzic, A., George, J. M., Jaffres, H., Fert, A., Faini,
G., Youssef, J. B. and Legall, H. (2001). Spin-polarized current induced
switching in Co/Cu/Co pillars, Appl. Phys. Lett. 78, pp. 3663–3665.
Grossu, G. and Paravocicino, G. P. (2000). Solid State Physics (Academis Press,
New York).
Gupta, A., Li, X. W., Guha, S. and Xiao, G. (1999). Selective-area and lateral
overgrowth of chromium dioxide (CrO2 ) films by chemical vapor deposition,
Appl. Phys. Lett. 75, pp. 2996–2998.
Gupta, A., Li, X. W. and Xiao, G. (2000). Magnetic and transport properties
of epitaxial and polycrystalline chromium dioxide thin films (invited), J.
Appl. Phys. 87, pp. 6073–6078.
Halilov, S. V. and Kulatov, E. T. (1991). Electron and magneto-optical proper-

ties of half-metallic ferromagnets and uranium monochalcogenide, J. Phys.:
Condens. Mat. 3, p. 6363.
Hamada, H. S. N. and Terakuar, K. (1995). Spectroscopy of Mott Insulators and
Correlated Metal, p. 95.
Hamann, D., Schlüter, M. and Chiang, C. (1979). Norm-conserving pseudopoten-
tials, Phys. Rev. Lett. 43, p. 1494.
Hamrle, J., Blomeier, S., Gaier, O., Hillebrands, B., Schneider, H., Jakob, G.,
Postava, K. and Felser, C. (2007). Huge quadratic magneto-optical Kerr
effect and magnetization reversal in the Co2 FeSi Heusler compound, J.
Phys. D: Appl. Phys. 40, pp. 1563–1569.
Hanssen, K. and Mijnarends, P. (1986). Positron-annihilation study of the half-
metallic ferromagnet NiMnSb: Theory, Phys. Rev. B 34, p. 5009.
Hanssen, K., Mijnarends, P., Rabou, L. and Buschow, K. (1990). Positron-
annihilation study of the half-metallic ferromagnet NiMnSb: Experiment,
Phys. Rev. B 42, p. 1533.
Harris, J. J., Joyce, B. A. and Dobson, P. (1981). Oscillations in the surface
structure of Sn-doped GaAs during growth by MBE, Surf. Sci. 103, pp.
L90–L96.
Hashimoto, M., Herfort, J., Schonherr, H.-P. and Ploog, K. (2005). Epitaxial
Heusler alloy Co2 FeSi/GaAs(001) hybrid structures, Appl. Phys. Lett. 87,
pp. 102506–3.
Heine, V. (1980). Electronic structure from the point of view of the local atomic
environment, pp. 1–123.
Heinz, K. (1995). LEED and DLEED as modern tools for quantitative surface
structure determination, Rep. Prog. Phys. 58, pp. 637–704.
Helmholdt, R. B., de Groot, R. A., Mueller, F. M., van Engen, P. G. and Buschow,
K. H. J. (1984). Magnetic and crystallographic properties of several C1b
type Heusler compounds, Journal of Magnetism and Magnetic Materials
43, pp. 249–255.
Herring, C. (1940). A new method for calculating wave functions in crystals, Phys.
Rev. 57, p. 1169.
Heusler, F. (1903). über magnetische manganlegierungen, Verh. Dtsch. Phys. Ges.
5, p. 219.
Hirohata, A., Kikuchi, M., Tezuka, N., Inomata, K., Claydon, J., Xu, Y. and
van der Laan, G. (2006). Heusler alloy/semiconductor hybrid structures,
Current Opinion in Solid State and Materials Science 10, pp. 93–107.
Hirohata, A., Kurebayashi, H., Okamura, S., Kikuchi, M., Masaki, T., Nozaki,
T., Tezuka, N. and Inomata, K. (2005). Structural and magnetic proper-
ties of epitaxial L21 –structured Co2 (Cr,Fe)Al films grown on GaAs(001)
substrates, J. Appl. Phys. 97, pp. 103714–8.
Hohenberg, P. and Kohn, W. (1964). Inhomogeneous electron gas, Phys. Rev.
136, pp. B864–B871.
Hordequin, C., Pierre, J. and Currat, R. (1996). Magnetic excitations in the
half-metallic NiMnSb ferromagnet: From Heisenberg-type to itinerant be-
haviour, Journal of Magnetism and Magnetic Materials 162, pp. 75–84.
Bibliography 253
Hordequin, C., Ristoiu, D., Ranno, L. and Pierre, J. (2000). On the cross-over
from half-metal to normal ferromagnet in NiMnSb, Euro. Phys. J. B 16,
pp. 287–293.
Hörmandinger, G., Weinberger, P., Marksteiner, P. and Redinger, J. (1988). The-
oretical calculations of core-core-valence Auger spectra: Applications to the
L3 M2,3 V transitions of Ti in nonstoichiometric Ti-C, Ti-N, and Ti-O, Phys.
Rev. B 38, pp. 1040–1046.
Hu, G. and Suzuki, Y. (2002). Negative spin polarization of Fe3 O4 in
magnetite/manganite-based junctions, Phys. Rev. Lett. 89, p. 276601.
Huang, D. J., Chang, C. F., Chen, J., Tjeng, L. H., Rata, A. D., Wu, W. P.,
Chung, S. C., Lin, H. J., Hibma, T. and Chen, C. T. (2002). Spin-resolved
photoemission studies of epitaxial Fe3 O4 (100) thin films, J. Magn. Magn.
Mater. 239, pp. 261–265.
Hubbard, J. (1963). Elecron correlations in narrow energy bands, Proc. R. Soc.
Lond. A 276, pp. 238—257, doi:10.1098/rspa.1963.0204.
Husmann, A. and Singh, L. (2006). Temperature dependence of the anomalous
Hall conductivity in the Heusler alloy Co2 CrAl, Phys. Rev. B 73, p. 172417.
Hwang, H. Y. and Cheong, S. W. (1997). Enhanced intergrain tunneling magne-
toresistance in half-metallic CrO2 films, Science 278, pp. 1607–1609.
Hwang, H. Y., Cheong, S.-W., Ong, N. P. and Batlogg, B. (1996). Spin-polarized
intergrain tunneling in La2/3 Sr1/3 MnO3 , Phys. Rev. Lett. 77, pp. 2041–
2044.
Iizumi, M., Koetzle, T. F., Shirane, G., Chikazumi, S., Matsui, M. and Todo, S.
(1982). Structure of magnetite Fe3 O4 below the verwey transition temper-
ature, Acta Crystall. B 38, pp. 2121–2133.
Inomata, K., Okamura, S., Miyazaki, A., Kikuchi, M., Tezuka, N., Wojcik, M.,
Jedryka, E. and et al, K. I. (2006). Structural and magnetic properties and
tunnel magnetoresistance for Co2 (Cr,Fe)Al and Co2 FeSi full-Heusler alloys,
J. Phys. D: Appl. Phys. 39, p. 816.
Ishida, S., Akazawa, S., Kubo, Y. and Ishida, J. (1982). Band theory of Co2 MnSn,
Co2 TiSn and Co2 TiAl, J. Phys. F: Met. Phys. 12, p. 1111.
Ishida, S., Masaki, T., Fujii, S. and Asano, S. (1998). Theoretical search for half-
metalliic films of Co2 MnZ (Z = Si, Ge), Physica B: Condensed Matter 245,
pp. 1–8.
Ivanov, P. G., Watts, S. M. and Lind, D. M. (2001). Epitaxial growth of CrO2
thin films by chemical-vapor deposition from a Cr8 O21 precursor, J. Appl.
Phys. 89, pp. 1035–1040.
Janak, J., Moruzzi, V. and Williams, A. (1975). Ground-state thermomechanical
properties of some cubic elements in the local-density formalism, Phys. Rev.
B 12, p. 1257.
Jeng, H.-T., Guo, G. Y. and Huang, D. J. (2004). Charge-orbital ordering and
Verwey transition in magnetite, Phys. Rev. Lett. 93, p. 156403.
Jenkins, S. J. (2004). Ternary half-metallics and related binary compounds: Sto-
ichiometry, surface states, and spin, Phys. Rev. B 70, p. 245401.
Ji, Y., Strijkers, G., Yang, F., Chien, C., Byers, J., Anguelouch, A., Xiao, G. and
Gupta, A. (2001). Determination of the spin polarization of half-metallic
CrO2 by point contact Andreev reflection, Phys. Rev. Lett. 86, p. 5585.
Jonker, G. H. and Santen, J. H. V. (1950). Ferromagnetic compounds of man-
ganese with perovskite structure, Physica 16, pp. 337–349.
Kabani, R., Terada, M., Roshko, A. and Moodera, J. S. (1990). Magnetic prop-
erties of NiMnSb films, J. Appl. Phys. 67, pp. 4898–4900.
Kallmayer, M., Schneider, H., Jakob, G., Elmers, H. J., Balke, B. and Cramm, S.
(2007). Interface magnetization of ultrathin epitaxial Co2 FeSi(110)/Al2 O3
films, J. Phys. D: Appl. Phys. 40, pp. 1552–1557.
Kämmerer, S., Heitmann, S., Meyners, D., Sudfeld, D., Thomas, A., Hütten,
A. and Reiss, G. (2003). Room-temperature preparation and magnetic
behavior of Co2 MnSi thin films, J. Appl. Phys. 93, pp. 7945–7947, doi:
10.1063/1.1556249.
Kämmerer, S., Thomas, A., Hütten, A. and Reiss, G. (2004). Co2 MnSi Heusler
alloy as magnetic electrodes in magnetic tunnel junctions, Appl. Phys. Lett.
85, pp. 79–81.
Kämper, K. P., Schmitt, W., Güntherodt, G., Gambino, R. J. and Ruf, R. (1987).
CrO2 –a new half-metallic ferromagnet? Phys. Rev. Lett. 59, pp. 2788–
2791.
Kandpal, H., Fecher, G., Felser, C. and Schöuml;nhense, G. (2006). Correlation
in the transition-metal-based Heusler compounds Co2 MnSi and Co2 FeSi,
Phys. Rev. B 73, p. 094422.
Kang, J.-S., Park, J.-G., Olson, C. G., Younq, S. J. and Min, B. I. (1995). Va-
lence band and Sb 4d core level photoemission of the XMnSb-type Heusler
compounds (X=Pt,Pd,Ni), J. Phys.: Condens. Mat. 7, pp. 3789–3800.
Karthik, S., Rajanikanth, A., Nakatani, T., Gercsi, Z., Takahashi, Y., Fu-
rubayashi, T., Inomata, K. and Hono, K. (2007). Effect of Cr substitution
for Fe on the spin polarization of Co2 Crx Fe1−x Si Heusler alloys, J. Appl.
Phys. 102, pp. 043903–6.
Katsnelson, M. I., Irkhin, V. Y., Chioncel, L., Lichtenstein, A. I. and deGroot,
R. A. (2008). Half-metallic ferromagnets: From band structure to many-
body effects, Rev. Mod. Phys. 80, pp. 315–378.
Kautzky, M. C., Mancoff, F. B., Bobo, J. F., Johnson, P. R., White, R. L. and
Clemens, B. M. (1997). Investigation of possible giant magnetoresistance
limiting mechanisms in epitaxial PtMnSb thin films, J. Appl. Phys. 81, pp.
4026–4028.
Kawakami, R. K., Johnston-Halperin, E., Chen, L. F., Hanson, M., Guebels, N.,
Speck, J. S., Gossard, A. C. and Awschalom, D. D. (2000). (Ga,Mn)As as a
digital ferromagnetic heterostructure, Appl. Phys. Lett. 77, pp. 2379–2381.
Kawano, H., Kajimoto, R., Kubota, M. and Yoshizawa, H. (1996a). Canted an-
tiferromagnetism in an insulating lightly doped La1−x Srx MnO3 with x ≤
0.17, Phys. Rev. B 53, pp. 2202–2205.
Kawano, H., Kajimoto, R., Kubota, M. and Yoshizawa, H. (1996b).
Ferromagnetism-induced reentrant structural transition and phase diagram
of the lightly doped insulator La1−x Srx MnO3 with x ≤ 0.17, Phys. Rev. B
53, pp. R14709–R14712.
Kelekar, R. and Clemens, B. (2004). Epitaxial growth of the Heusler alloy
Bibliography 255
Co2 Cr1−x Fex Al, J. Appl. Phys. 96, p. 540.

Kessler, J. (1985). Polarized Electrons (Springer Netherlands).
Khoi, L. D., Veillet, P. and Campbell, I. A. (1978). Hyperfine fields and magnetic
interactions in Heusler alloys, Journal of Physics F: Metal Physics 8, p.
1811.
Kim, D. J. (1999). New Perspectives in Magentisim of Metals (Plenum Publica-
tion (New York)).
Kim, H.-J., Park, J.-H. and Vescovo, E. (2000a). Fe3 O4 (111)/Fe(110) magnetic
bilayer: Electronic and magnetic properties at the surface and interface,
Phys. Rev. B 61, pp. 15288–15293.
Kim, H.-J., Park, J.-H. and Vescovo, E. (2000b). Oxidation of the Fe(110) surface:
An Fe3 O4 (111)/Fe(110) bilayer, Phys. Rev. B 61, pp. 15284–15287.
Kimura, A., Suga, S., Shishidou, T., Imada, S., Muro, T., Park, S., Miyahara,
T., Kaneko, T. and Kanomata, T. (1997). Magnetic circular dichroism in
the soft-x-ray absorption spectra of Mn-based magnetic intermetallic com-
pounds, Phys. Rev. B 56, p. 6021.
Kirillova, M., Makhnev, A., Shreder, E., Dyakina, V. and Gorina, N. (1995).
Interband optical absorption and plasma effects in half-metallic XMnY fer-
romagnets, physica status solidi (b) 187, pp. 231–240.
Kittel, C. (2004). Introduction to solid state physics, 8th edition, Wiley & Sons,
New York, NY .
Kleinman, L. and Bylander, D. (1982). Efficacious form for model pseudopoten-
tials, Phys. Rev. Lett. 48, p. 1425.
Kodama, K., furubayashi, T., Sukegawa, H., Nanktani, T. M., Inomata, K. and
Hono, K. (2009). Current-perpendicular-to-plane giant magnetoresistance
of a spin valve using Co2 MnSi Heulster alloy electrodes, J. Appl. Phys. 105,
p. 07E905.
Koelling, D. D. and Arbman, G. O. (1975). Use of energy derivative of the radial
solution in an augmented plane wave method: application to copper, J.
Phys. F: Met. Phys. 5, p. 2041.
Kohn, A., Lazarov, V., Singh, L., Barber, Z. and Petford-Long, A. (2007). The
structure of sputter-deposited Co2 MnSi thin films deposited on GaAs(001),
J. Appl. Phys. 101, pp. 023915–11.
Kohn, W. and Sham, L. (1965). Self-consistent equations including exchange and
correlation effects, Phys. Rev. 140, p. A1133.
Kolev, H., Rangelov, G., Braun, J. and Donath, M. (2005). Reduced surface
magnetization of NiMnSb(001), Phys. Rev. B 72, p. 104415.
Komesu, T., Borca, C. N., Jeong, H.-K., Dowben, P. A., Ristoiu, D., Nozières,
J. P., Stadler, S. and Idzerda, Y. U. (2000). The polarization of Sb overlayers
on NiMnSb(100), Phys. Lett. A 273, pp. 245–251.
Korotin, M. A., Anisimov, V. I., Khomskii, D. I. and Sawatzky, G. A. (1998).
CrO2 : A self-doped double exchange ferromagnet, Phys. Rev. Lett. 80, pp.
4305–4308.
Kouvel, J. S. and Rodbell, D. S. (1967). Magnetic critical-point behavior of CrO2 ,
J. Appl. Phys. 38, pp. 979–981.
Kramers, H. A. (1934). L’interaction entre les atomes magnètogènes dans un
cristal paramagnètique, Physica 1, pp. 182–192.

Kresse, G. and Joubert, D. (1999). From ultrasoft pseudopotentials to the pro-
jector augmented-wave method, Phys. Rev. B 59, pp. 1758–1775.
Kübler, J. (1984). First principle theory of metallic magnetism, Physica B+C
127, pp. 257–263.
Kubo, K. and Ohata, N. (1972). A quantum theory of double exchange. I, J.
Phys. Soc. Jpn. 33, pp. 21–32.
Kulatov, E. and Mazin, I. I. (1990). Extended stoner factor calculations for the
half-metallic ferromagnets NiMnSb and CrO2 , J. Phys.: Condens. Mat. 2,
pp. 343–350.
Kuneš, J., Novák, P., Oppeneer, P. M., König, C., Fraune, M., Rüdiger, U.,
Güntherodt, G. and Ambrosch-Draxl, C. (2002). Electronic structure of
CrO2 as deduced from its magneto-optical Kerr spectra, Phys. Rev. B 65,
p. 165105.
Langreth, D. C. and Mehl, M. J. (1983). Beyond the local-density approximation
in calculations of ground-state electronic properties, Phys. Rev. B 28, pp.
1809–1834.
Langreth, D. C. and Perdew, J. P. (1980). Theory of nonuniform electronic sys-
tems. I. analysis of the gradient approximation and a generalization that
works, Phys. Rev. B 21, pp. 5469–5493.
Larson, P., Mahanti, S. D. and Kanatzidis, M. G. (2000). Structural stability of
Ni-containing half-Heusler compounds, Phys. Rev. B 62, pp. 12754–12762.
Lee, M. J. G. and Falicov, L. M. (1968). The de Haas-van Alphen effect and the
Fermi surface of potassium, Proc. R. Soc. Lond. A 304, pp. 319–334.
Lee, W.-L., Watauchi, S., Miller, V. L., Cava, R. J. and Ong, N. P. (2004). Dissipa-
tionless anomalous Hall current in the ferromagnetic spinel CuCr2 Se4−x Brx ,
Science 303, pp. 1647–1649.
Li, X. W., Gupta, A., McGuire, T. R., Duncombe, P. R. and Xiao, G. (1999).
Magnetoresistance and Hall effect of chromium dioxide epitaxial thin films,
J. Appl. Phys. 85, pp. 5585–5587.
Liechtenstein, A., Anisimov, V. and Zaanen, J. (1995). Density-functional the-
ory and strong interactions: Orbital ordering in Mott-Hubbard insulators,
Phys. Rev. B 52, p. R5467.
Liechtenstein, A. I., Katsnelson, M. I., Antropov, V. P. and Gubanov, V. A.
(1987). Local spin density functional approach to the theory of exchange
interactions in ferromagnetic metals and alloys, J. Magn. Magn. Mater. 67,
pp. 65–74.
Liu, B.-G. (2003). Robust half-metallic ferromagnetism in zinc-blende CrSb, Phys.
Rev. B 67, p. 172411.
Loos, J. and Novák, P. (2002). Double exchange and superexchange in a ferri-
magnetic half-metal, Phys. Rev. B 66, p. 132403.
Lu, Y., Li, X. W., Gong, G. Q., Xiao, G., Gupta, A., Lecoeur, P., Sun, J. Z.,
Wang, Y. Y. and Dravid, V. P. (1996). Large magnetotunneling effect at
low magnetic fields in micrometer-scale epitaxial La0.67 Sr0.33 MnO3 tunnel
junctions, Phys. Rev. B 54, pp. R8357–R8360.
Luo, H., McCombe, B. D., Na, M. H., Mooney, K., Lehmann, F., Chen, X., Cheon,
Bibliography 257
M., Wang, S. M., Sasaki, Y., Liu, X. and Furdyna, J. K. (2002). Transport
and magnetic properties of ferromagnetic GaAs/Mn digital alloys, Physica
E: Low-dimensional Systems and Nanostructures 12, pp. 366–369.
Ma, C., Yang, Z. and Picozzi, S. (2006). Ab initio electronic and magnetic struc-
ture in La0.66 Sr0.33 MnO3 : Strain and correlation effects, J. Phys.: Con-
dens. Mat. 18, pp. 7717–7728.
Mackintosh, A. R. and Andersen, O. K. (1980). The electronic structure of tran-
sition metals, chap. 5.
Malozemoff, A., Williams, A. and Moruzzi, V. (1984). “Band-gap theory“ of
strong ferromagnetism: Application to concentrated crystalline and amor-
phous Fe- and Co-metalloid alloys, Phys. Rev. B 29, p. 1620.
Mancoff, F., Boboo, J., Richter, O., Bessho, K., Johnson, P., Sinclair, R., Nix,
W., White, R. and Clemens, B. (1999). Growth and characterization of
epitaxial NiMnSb/PtMnSb C1b Heusler alloy superlattices, J. Materi. Res.
14, p. 1560.
Marcus, P. M. and Moruzzi, V. L. (1988). Stoner model of ferromagnetism and
total-energy band theory, Phys. Rev. B 38, pp. 6949–6953.
Martin, R. M. (2004). Electronic Structure: Basic Theory and Practical Methods
(Cambridge University Press, Cambridge).
Matsubara, K., Anno, H., Kaneko, H. and Imai, Y. (1999). Electrical proper-
ties of half-metallic PtMnSb-based Heusler alloys, Eighteenth International
Conference on Thermoelectrics (1999). , pp. 60–63.
Mazin, I. I., Singh, D. J. and Ambrosch-Draxl, C. (1999). Transport, optical, and
electronic properties of the half-metal CrO2 , Phys. Rev. B 59, pp. 411–418.
McCormack, M., Jin, S., Tiefel, T. H., Fleming, R. M., Phillips, J. M. and
Ramesh, R. (1994). Very large magnetoresistance in perovskite–like La–
Ca–Mn–O thin films, Appl. Phys. Lett. 64, pp. 3045–3047.
McQueeney, R. J., Yethiraj, M., Montfrooij, W., Gardner, J. S., Metcalf, P.
and Honig, J. M. (2006). Investigation of the presence of charge order in
magnetite by measurement of the spin wave spectrum, Phys. Rev. B 73, p.
174409.
Meservey, R. and Tedrow, P. M. (1994). Spin-polarized electron tunneling, Physics
Reports 238, pp. 173–243.
Metalidis, G. and Bruno, P. (2006). Inelastic scattering effects and the Hall resis-
tance in a four-probe ring, Phys. Rev. B 73, p. 113308.
Miyamoto, K., Kimura, A., Iori, K., Sakamoto, K., Xie, T., Moko, T., Qiao, S.,
Taniguchi, M. and Tsuchiya, K. (2004). Element-resolved magnetic mo-
ments of Heusler-type ferromagnetic ternary alloy Co2 MnGe, J. Phys.:
Condens. Mat. 16, p. S5797.
Mizuguchi, M., Akinaga, H., Manago, T., Ono, K., Oshima, M., Shirai, M., Yuri,
M., Lin, H. J., Hsieh, H. H. and Chen, C. T. (2002). Epitaxial growth of
zinc-blende CrAs/GaAs multilayer, J. Appl. Phys. 91, pp. 7917–7919.
Moore, G. E. (1965). Cramming more components onto integrated circuits, Elec-
tronics 38, p. 114.
Moos, R., Menesklou, W. and Härdtl, K. H. (1995). Hall mobility of undoped n-
type conducting strontium titanate single crystals between 19 K and 1373
K, Appl. Phys. A–Mater. 61, pp. 389–395.

Mukhamedzhanov, E. K., Bocchi, C., Franchi, S., Baraldi, A., Magnanini, R.
and Nasi, L. (2000). High-resolution x-ray diffraction, x-ray standing-wave,
and transmission electron microscopy study of Sb-based single-quantum-
well structures, J. Appl. Phys. 87, pp. 4234–4239, doi:10.1063/1.373058.
Nazarenko, E., Lorenzo, J. E., Joly, Y., Hodeau, J. L., Mannix, D. and Marin, C.
(2006). Resonant x-ray diffraction studies on the charge ordering in mag-
netite, Phys. Rev. Lett. 97, p. 056403.
Nèel, L. (1948). Magnetic properties of ferrites: Ferrimagnetism and antiferro-
magnetism, Ann. Phys.-Paris 3, p. 137.
Niculescu, V., Burch, T., Raj, K. and Budnick, J. (1977). Properties of Heusler-
type materials Fe2 TSi and FeCo2 Si, J. Magn. Magn. Mater. 5, pp. 60–66.
Norby, P., Christensen, A. N., FjellvøAg, H. and Nielsen, M. (1991). The crys-
tal structure of Cr8 O21 determined from powder diffraction data: Ther-
mal transformation and magnetic properties of a chromium-chromate-
tetrachromate, J. Solid State Chem. 94, pp. 281–293.
Oguchi, T., Terakura, K. and Williams, A. R. (1984). Transition-metal monox-
ides: Itinerant versus localized picture of superexchange, J. Appl. Phys. 55,
pp. 2318–2320.
Ohashi, S., Lippmaa, M., Nakagawa, N., Nagasawa, H., Koinuma, H. and
Kawasaki, M. (1999). Compact laser molecular beam epitaxy system us-
ing laser heating of substrate for oxide film growth, Rev. Sci. Instrum. 70,
pp. 178–183.
Okabayashi, J., Mizuguchi, M., Ono, K., Oshima, M., Fujimori, A., Kuramochi,
H. and Akinaga, H. (2004). Density-dependent electronic structure of zinc-
blende-type MnAs dots on GaAs(001) studied by in situ photoemission
spectroscopy, Phys. Rev. B 70, p. 233305.
Okimoto, Y., Katsufuji, T., Ishikawa, T., Urushibara, A., Arima, T. and Tokura,
Y. (1995). Anomalous variation of optical spectra with spin polarization
in double-exchange ferromagnet: La1−x Srx MnO3 , Phys. Rev. Lett. 75, pp.
109–112.
Ono, K., Okabayashi, J., Mizuguchi, M., Oshima, M., Fujimori, A. and Akinaga,
H. (2002). Fabrication, magnetic properties, and electronic structures of
nanoscale zinc-blende MnAs dots (invited), J. Appl. Phys. 91, pp. 8088–
8092.
Orgassa, D., Fujiwara, H., Schulthess, T. C. and Butler, W. H. (1999). First-
principles calculation of the effect of atomic disorder on the electronic
structure of the half-metallic ferromagnet NiMnSb, Phys. Rev. B 60, pp.
13237–13240.
Orgassa, D., Fujiwara, H., Schulthess, T. C. and Butler, W. H. (2000). Disor-
der dependence of the magnetic moment of the half-metallic ferromagnet
NiMnSb from first principles, J. Appl. Phys. 87, pp. 5870–5871.
Oswald, A., Zeller, R., Braspenning, P. J. and Dederichs, P. H. (1985). Interaction
of magnetic impurities in Cu and Ag, J. Phys. F: Met. Phys. 15, pp. 193–
212.
Otto, M. J., van Woerden, R. A. M., van der Valk, P. J., Wijngaard, J., van
Bibliography 259
Bruggen, C. F., Haas, C. and Buschow, K. H. J. (1989). Half-metallic ferro-

magnets. i. structure and magnetic properties of NiMnSb and related inter-
metallic compounds, Journal of Physics: Condensed Matter 1, p. 2341.
Paiva-Santos, C. O., Marques, R. F. C., Jafelicci, J. M. and Varanda, L. C.
(2002). X-ray powder data and bond valence of La0.65 Sr0.35 MnO3 after
Rietveld refinement, Powder Diffraction 17, pp. 149–152.
Pajda, M., Kudrnovský, J., Turek, I., Drchal, V. and Bruno, P. (2001). Ab
initio calculations of exchange interactions, spin-wave stiffness constants,
and Curie temperatures of Fe, Co, and Ni, Phys. Rev. B 64, p. 174402.
Park, J. H., Vescovo, E., Kim, H. J., Kwon, C., Ramesh, R. and Venkatesan, T.
(1998a). Direct evidence for a half-metallic ferromagnet, Nature 392, pp.
794–796.
Park, J.-H., Vescovo, E., Kim, H.-J., Kwon, C., Ramesh, R. and Venkatesan, T.
(1998b). Magnetic properties at surface boundary of a half-metallic ferro-
magnet La0.7 Sr0.3 MnO3 , Phys. Rev. Lett. 81, pp. 1953–1956.
Parkin, S. S. P., Bhadra, R. and Roche, K. P. (1991). Oscillatory magnetic ex-
change coupling through thin copper layers, Phys. Rev. Lett. 66, pp. 2152–
2155.
Pask, J. E., Yang, L. H., Fong, C. Y., Pickett, W. E. and Dag, S. (2003). Six
low-strain zinc-blende half metals: An ab initio investigation, Phys. Rev. B
67, p. 224420.
Peierls, R. E. (1955). Quantum Theory of Solids (Oxford).
Perdew, J. and Yue, W. (1986). Accurate and simple density functional for
the electronic exchange energy: Generalized gradient approximation, Phys.
Rev. B 33, p. 8800.
Perdew, J. and Zunger, A. (1981). Self-interaction correction to density-functional
approximations for many-electron systems, Phys. Rev. B 23, p. 5048.
Perdew, J. P., Burke, K. and Ernzerhof, M. (1996). Generalized gradient approx-
imation made simple, Phys. Rev. Lett. 77, pp. 3865–3868.
Perdew, J. P., Burke, K. and Ernzerhof, M. (1997). Errata: Generalized gradient
approximation made simple [Phys. Rev. Lett. 77, 3865 (1996)], Phys. Rev.
Lett. 78, p. 1396.
Perdew, J. P., Chevary, J. A., Vosko, S. H., Jackson, K. A., Pederson, M. R.,
Singh, D. J. and Fiolhais, C. (1992). Atoms, molecules, solids, and surfaces:
Applications of the generalized gradient approximation for exchange and
correlation, Phys. Rev. B 46, pp. 6671–6687.
Pesavento, P. V., Chesterfield, R. J., Newman, C. R. and Frisbie, C. D. (2004).
Gated four-probe measurements on pentacene thin-film transistors: Con-
tact resistance as a function of gate voltage and temperature, J. Appl. Phys.
96, pp. 7312–7324.
Phillips, J. and Kleinman, L. (1959). New method for calculating wave functions
in crystals and molecules, Phys. Rev. 116, p. 287.
Picozzi, S., Continenza, A. and Freeman, A. J. (2002). Co2 MnX (X=Si, Ge, Sn)
Heusler compounds: An ab initio study of their structural, electronic, and
magnetic properties at zero and elevated pressure, Phys. Rev. B 66, p.
094421.
Picozzi, S., Continenza, A. and Freeman, A. J. (2004). Role of structural defects

on the half-metallic character of Co2 MnGe and Co2 MnSi Heusler alloys,
Phys. Rev. B 69, p. 094423.
Podloucky, R., Zeller, R. and Dederichs, P. H. (1980). Electronic structure of
magnetic impurities calculated from first principles, Phys. Rev. B 22, pp.
5777–5790.
Porta, P., Marezio, M., Remeika, J. P. and Dernier, P. D. (1972). Chromium diox-
ide: High pressure synthesis and bond lengths, Materials Research Bulletin
7, pp. 157–161.
Postava, K., Hrabovsky, D., Pistora, J., Fert, A. R., Visnovsky, S. and Yamaguchi,
T. (2002). Anisotropy of quadratic magneto-optic effects in reflection, J.
Appl. Phys. 91, pp. 7293–7295.
Poulsen, U. K., Kollar, J. and Andersen, O. K. (1976). Magnetic and cohesive
properties from canonical bands (for transition metals), J. Phys. F: Met.
Phys. 6, p. L241.
Qian, M., Fong, C. Y., Pickett, W. and Wang, H. (2004a). An ab initio inves-
tigation on the zinc-blende MnAs nanocrystallite, J. Appl. Phys. 95, pp.
7459–7461.
Qian, M., Fong, C. Y. and Yang, L. H. (2004b). Coexistence of localized magnetic
moment and opposite-spin itinerant electrons in MnC, Phys. Rev. B 70, p.
052404.
Qian, M. C., Fong, C. Y., Liu, K., Pickett, W. E., Pask, J. E. and Yang, L. H.
(2006a). Half-metallic digital ferromagnetic heterostructure composed of a
δ-doped layer of Mn in Si, Phys. Rev. Lett. 96, p. 027211.
Qian, M. C., Fong, C. Y. and Pickett, W. E. (2006b). Enhancement of ferro-
magnetic coupling in Mn/GaAs digital ferromagnetic heterostructure by
free-hole injection, J. Appl. Phys. 99, p. 08D517.
Qian, M. C., Fong, C. Y., Pickett, W. E., Pask, J. E., Yang, L. H. and Dag,
S. (2005). Spin-polarized ballistic transport in a thin superlattice of zinc
blende half-metallic compounds, Phys. Rev. B 71, p. 012414.
Ranno, L., Barry, A. and Coey, J. M. D. (1997). Production and magnetotrans-
port properties of CrO2 films, J. Appl. Phys. 81, pp. 5774–5776.
Raphael, M., Ravel, B., Huang, Q., Willard, M., Cheng, S., Das, B., Stroud,
R., Bussmann, K., Claassen, J. and Harris, V. (2002). Presence of antisite
disorder and its characterization in the predicted half-metal Co2 MnSi, Phys.
Rev. B 66, p. 104429.
Raue, R., Hopster, H. and Kisker, E. (1984). High-resolution spectrometer for
spin-polarized electron spectroscopies of ferromagnetic materials, Rev. Sci.
Instrum. 55, pp. 383–388.
Reisinger, D., Schonecke, M., Brenninger, T., Opel, M., Erb, A., Alff, L. and
Gross, R. (2003). Epitaxy of Fe3 O4 on Si(001) by pulsed laser deposition
using a TiN/MgO buffer layer, J. Appl. Phys. 94, pp. 1857–1863.
Rietveld, H. M. (1969). A profile refinement method for nuclear and magnetic
structures, J. Appl. Cryst. 2, pp. 65–71.
Ristoiu, D., Nozieres, J. P., Borca, C. N., Borca, B. and Dowben, P. A. (2000a).
Manganese surface segregation in NiMnSb, Appl. Phys. Lett. 76, pp. 2349–
Bibliography 261
2351.
Ristoiu, D., Nozir̀es, J. P., Borca, C. N., Komesu, T., k. Jeong, H. and Dowben,
P. A. (2000b). The surface composition and spin polarization of NiMnSb
epitaxial thin films, EPL (Europhysics Letters) 49, p. 624.
Ritchie, L., Xiao, G., Ji, Y., Chen, T., Chien, C., Zhang, M., Chen, J., Liu, Z.,
Wu, G. and Zhang, X. (2003). Magnetic, structural, and transport prop-
erties of the Heusler alloys Co2 MnSi and NiMnSb, Phys. Rev. B 68, p.
104430.
Robinson, I. K. (1991). Handbook of synchrotron radiation, Vol. 3.
Ross, C. (2001). Patterned magnetic recording media, Ann. Rev. Mater. Res. 31,
pp. 203–235.
Rusz, J., Bergqvist, L., Kudrnovskỳ, J. and Turek, I. (2006). Exchange interac-
tions and Curie temperatures in Ni2−x MnSb alloys: First-principles study,
Phys. Rev. B 73, p. 214412.
Sacchi, M., Spezzani, C., Carpentiero, A., Prasciolu, M., Delaunay, R., Luning,
J. and Polack, F. (2007). Experimental setup for lensless imaging via soft
x-ray resonant scattering, Rev. Sci. Instrum. 78, pp. 043702–6.
Sakuraba, Y., Hattori, M., Oogane, M., Ando, Y., Kato, H., Sakuma, A.,
Miyazaki, T. and Kubota, H. (2006). Giant tunneling magnetoresistance
in Co2 MnSi/Al–O/Co2 MnSi magnetic tunnel junctions, Appl. Phys. Lett.
88, pp. 192508–3.
Salamon, M. B. and Jaime, M. (2001). The physics of manganites: Structure and
transport, Rev. Mod. Phys. 73, pp. 583–628.
Sánchez-Portal, D., Ordejón, P., Artacho, E. and Soler, J. M. (1997). Density-
functional method for very large systems with LCAO basis sets, Interna-
tional Journal of Quantum Chemistry 65, pp. 453–461.
Sandratskii, L. M. (1998). Noncollinear magnetism in itinerant-electron systems:
theory and applications, Advances in Physics 47, pp. 91– 160.
Sanvito, S. and Hill, N. (2001). Ab initio transport theory for digital ferromagnetic
heterostructures, Phys. Rev. Lett. 87, p. 267202.
Satpathy, S. and Vukajlović, Z. S. P. F. R. (1996). Electronic structure of the
perovskite oxides: La1−x Cax MnO3 , Phys. Rev. Lett. 76, pp. 960–963.
Savtchenko, L., Engel, B., Rizzo, N., Deherrera, M. and Janesky, J. (2003).
Method of writing to scalable magnetoresistance random access memory
element, Us patent.
Schlomka, J.-P., Tolan, M. and Press, W. (2000). In situ growth study of NiMnSb
films on MgO(001) and Si(001), Appl. Phys. Lett. 76, pp. 2005–2007.
Schlottmann, P. (2003). Double-exchange mechanism for CrO2 , Phys. Rev. B 67,
p. 174419.
Schmalhorst, J., Kämmerer, S., Sacher, M., Reiss, G., Hütten, A. and Scholl,
A. (2004). Interface structure and magnetism of magnetic tunnel junctions
with a Co2 MnSi electrode, Phys. Rev. B 70.
Schneider, H., Herbort, C., Jakob, G., Adrian, H., Wurmehl, S. and Felser, C.
(2007). Structural, magnetic and transport properties of Co2 FeSi Heusler
films, Journal of Physics D: Applied Physics 40, p. 1548.
Schneider, H., Jakob, G., Kallmayer, M., Elmers, H., Cinchetti, M., Balke, B.,
Wurmehl, S., Felser, C., Aeschlimann, M. and Adrian, H. (2006). Epitax-

ial film growth and magnetic properties of Co2 FeSi, Phys. Rev. B 74, p.
174426.
Schwarz, K. (1986). CrO2 predicted as a half-metallic ferromagnet, J. Phys. F:
Met. Phys. 16, pp. L211–L215.
Shen, J., Gai, Z. and Kirschner, J. (2004). Growth and magnetism of metallic thin
films and multilayers by pulsed-laser deposition, Surface Science Reports
52, pp. 163–218, doi:doi:DOI:10.1016/j.surfrep.2003.10.001.
Shepherd, J. P. and Sandberg, C. J. (1984). Measurement and control of oxygen
fugacity during annealing, Rev. Sci. Instrum. 55, pp. 1696–1699.
Shima, M., Tepper, T. and Ross, C. A. (2002). Magnetic properties of chromium
oxide and iron oxide films produced by pulsed laser deposition, J. Appl.
Phys. 91, pp. 7920–7922.
Shirai, M. (2004). The computational design of zinc-blende half-metals and their
nanostructures, J. Phys.: Condens. Mat. 16, p. S5525.
Shull, C. G., Wollan, E. O. and Koehler, W. C. (1951). Neutron scattering and
polarization by ferromagnetic materials, Phys. Rev. 84, pp. 912–921.
Singh, D. J. and Nordström, L. (2006). Pseudopotentials, and the LAPW method.
Singh, D. J. and Pickett, W. E. (1998). Pseudogaps, Jahn-Teller distortions, and
magnetic order in manganite perovskites, Phys. Rev. B 57, pp. 88–91.
Singh, L., Barber, Z., Kohn, A., Petford-Long, A., Miyoshi, Y., Bugoslavsky,
Y. and Cohen, L. (2006). Interface effects in highly oriented films of the
Heusler alloy Co2 MnSi on GaAs(001), J. Appl. Phys. 99, pp. 013904–4.
Singh, L., Barber, Z., Miyoshi, Y., Branford, W. and Cohen, L. (2004a). Struc-
tural and transport studies of stoichiometric and off-stoichiometric thin
films of the full Heusler alloy Co2 MnSi, J. Appl. Phys. 95, pp. 7231–7233.
Singh, L. J., Barber, Z. H., Miyoshi, Y., Bugoslavsky, Y., Branford, W. R. and
Cohen, L. F. (2004b). Structural, magnetic, and transport properties of thin
films of the Heusler alloy Co2 MnSi, Appl. Phys. Lett. 84, pp. 2367–2369.
Skriver, H. L. (1983). The LMTO method (Spinger-Verlag, Berlin).
Slater, J. C. (1951). Magnetic effects and the Hartree-Fock equation, Phys. Rev.
82, pp. 538–541.
Slater, J. C. (1953). Quarterly progress reports, Tech. rep.
Smit, J. and Wijn, H. P. J. (1959). Ferrites (John Wiley & Sons, New York).
Solovyev, I. V. and Terakura, K. (2003). Orbital degeneracy and magnetism of
perovskite manganese oxides, chap. 6, pp. 253–295.
Soulen, R., Byers, J., Osofsky, M., Nadgorny, B., Ambrose, T., Cheng, S., Brous-
sard, P., Tanaka, C., Nowak, J., Moodera, J., Barry, A. and Coey, J. (1998).
Measuring the spin polarization of a metal with a superconducting point
contact, Science 282, pp. 85–88.
Souza, S. D., Saxena, R., Shreiner, W. and Zawislak, F. (1987). Magnetic hy-
perfine fields in Heusler alloys Co2 YZ (Y=Ti,Zr; Z=Al,Ga,Sn), Hyperfine
Interactions 34, pp. 431–434.
Spinu, L., Srikanth, H., Gupta, A., Li, X. W. and Xiao, G. (2000). Probing
magnetic anisotropy effects in epitaxial CrO2 thin films, Phys. Rev. B 62,
pp. 8931–8934.
Bibliography 263
Stoffel, A. M. (1969). Magnetic and magneto-optic properties of FeRh and CrO2 ,

J. Appl. Phys. 40, pp. 1238–1239.
Stöhr, J. (1999). Exploring the microscopic origin of magnetic anisotropies with
x-ray magnetic circular dichroism (XMCD) spectroscopy, J. Magn. Magn.
Mater. 200, pp. 470–497.
Stoner, E. C. (1939). Collective electron ferromagnetism. ii. energy and specific
heat, Proc. R. Soc. Lond. A 169, pp. 339–371.
Stranski, I. N. and Krastanow, L. V. (1939). Abhandlungen der Mathematisch-
Naturwissenschaftlichen Klasse, Akademie der Wissenschaften und der Lit-
eratur in Mainz 146, p. 797.
Strijkers, G. J., Ji, Y., Yang, F. Y., Chien, C. L. and Byers, J. M. (2001). Andreev
reflections at metal/superconductor point contacts: Measurement and anal-
ysis, Phys. Rev. B 63, p. 104510.
Suzuki, T. and Ido, H. (1993). Magnetic-nonmagnetic transition in cras and the
related compounds, J. Appl. Phys. 73, pp. 5686–5688.
Teresa, J. M. D., Barthémy, A., Fert, A., Contour, J. P., Lyonnet, R., Mon-
taigne, F., Seneor, P. and Vaurès, A. (1999). Inverse tunnel magnetoresis-
tance inCo/SrTiO3 /La0.7 Sr0.3 MnO3 : New ideas on spin-polarized tunnel-
ing, Phys. Rev. Lett. 82, pp. 4288–4291.
Tersoff, J. and Hamann, D. (1985). Theory of the scanning tunneling microscope,
Phys. Rev. B 31, p. 805.
Thamer, B. J., Douglass, R. M. and Staritzky, E. (1957). The thermal decom-
position of aqueous chromic acid and some properties of the resulting solid
phases1, J. Am. Chem. Soc. 79, pp. 547–550.
Thomas, A. (2003). Ph.D. thesis, University of Bielefeld.
Tobola, J., Pierre, J., Kaprzyk, S., Skolozdra, R. V. and Kouacou, M. A. (1998).
Crossover from semiconductor to magnetic metal in semi-Heusler phases as
a function of valence electron concentration, J. Phys.: Condens. Mat. 10,
p. 1013.
Toney, M. and Brennan, S. (1989). Measurements of carbon thin films using x-ray
reflectivity, J. Appl. Phys. 66, pp. 1861–1863.
Turban, P., Andrieu, S., Kierren, B., Snoeck, E., Teodorescu, C. and Traverse, A.
(2002). Growth and characterization of single crystalline NiMnSb thin films
and epitaxial NiMnSb/MgO/NiMnSb(001) trilayers, Phys. Rev. B 65, p.
134417.
Uhrig, M., Beck, A., Goeke, J., Eschen, F., Sohn, M., Hanne, G. F., Jost, K. and
Kessler, J. (1989). Calibration of a Mott detector using circularly polarized
impact radiation from helium, Rev. Sci. Instrum. 60, pp. 872–878.
van der Heidet, P. A. M., Baeldet, W., deGroott, R. A., de Vrooment, A. R., van
Engent, P. G. and Buschow, K. H. J. (1985). Optical properties of some
half-metallic ferromagnets, J. Phys. F: Met. Phys. 15, p. L75.
van der Zaag, P. J., Bloemen, P. J. H., Gaines, J. M., Wolf, R. M., van der Heijden,
P. A. A., van de Veerdonk, R. J. M. and de Jonge, W. J. M. (2000). On
the construction of an Fe3 O4 -based all-oxide spin valve, J. Magn. Magn.
Mater. 211, pp. 301–308.
van Dijken, S., Fain, X., Watts, S. M., Nakajima, K. and Coey, J. M. D. (2004).
Magnetoresistance of Fe3 O4 /Au/Fe3 O4 and Fe3 O4 /Au/Fe spin-valve struc-

tures, J. Magn. Magn. Mater. 280, pp. 322–326.
van Engen, P., Buschow, K., Jongebreur, R. and Erman, M. (1983). PtMnSb, a
material with very high magneto-optical Kerr effect, Appl. Phys. Lett. 42,
pp. 202–204.
Van Roy, W., Boeck, J. D., Brijs, B. and Borghs, G. (2000). Epitaxial NiMnSb
films on GaAs(001), Appl. Phys. Lett. 77, pp. 4190–4192.
Vanderbilt, D. (1990). Soft self-consistent pseudopotentials in a generalized eigen-
value formalism, Phys. Rev. B 41, pp. 7892–7895.
Versluijs, J. J. and Coey, J. M. D. (2001). Magnetotransport properties of Fe3 O4
nanocontacts, J. Magn. Magn. Mater. 226-230, pp. 688–689.
Verwey, E. J. W. (1939). Electronic conduction of magnetite (Fe3 O4 ) and its
transition point at low temperatures, Nature 144, pp. 327–328.
von Barth, U. and Hedin, L. (1972). A local exchange-correlation potential for
the spin polarized case. I, J. Phys. C: Solid Stat. Phys. 5, p. 1629.
von Helmolt, R., Wecker, J., Holzapfel, B., Schultz, L. and Samwer, K. (1993).
Giant negative magnetoresistance in perovskitelike La2/3 Ba1/3 MnOx ferro-
magnetic films, Phys. Rev. Lett. 71, pp. 2331–2333.
Wang, F. and Vaedeny, Z. V. (2009). Organic spin valves: the first organic spin-
tronics devices, J. Materi. Chem. 19, pp. 1685–1690.
Wang, H.-Y. and Qian, M. C. (2006). Electronic and magnetic properties of
Mn/Ge digital ferromagnetic heterostructures: An ab initio investigation,
J. Appl. Phys. 99, p. 08D705.
Wang, W. H., Przybylski, M., Kuch, W., Chelaru, L., Wang, J., Lu, Y., Barthel,
J., Meyerheim, H. and Kirschner, J. (2005a). Magnetic properties and
spin polarization of Co2 MnSi Heusler alloy thin films epitaxially grown
on GaAs(001), Phys. Rev. B 71, p. 144416.
Wang, W. H., Przybylski, M., Kuch, W., Chelaru, L. I., Wang, J., Lu, Y. F.,
Barthel, J. and Kirschner, J. (2005b). Spin polarization of single-crystalline
Co2 MnSi films grown by PLD on GaAs(0 0 1), J. Magn. Magn. Mater. 286,
pp. 336–339.
Wang, X., Antropov, V. P. and NHarmon, B. (1994). First principles study of
magneto-optical properties of half metallic Heusler alloys: NiMnSb and
PtMnSb, IEEE T. Magn. 30, pp. 4458–4460.
Webster, P. J. (1971). Magnetic and chemical order in Heusler alloys containing
cobalt and manganese, Journal of Physics and Chemistry of Solids 32, pp.
1221–1231.
Webster, P. J. and Ziebeck, K. R. A. (1988). Alloys and compounds of d-elements
with main group elements, Landolt-Börnstein, New Series, Group III, ed
H. R. J. Wijn (Berlin: Springer) 19c, pp. 75–184.
Wei, J. Y. T., Yeh, N.-C. and Vasquez, R. P. (1997). Tunneling evidence of
half-metallic ferromagnetism in La0.7 Ca0.3 MnO3 , Phys. Rev. Lett. 79, pp.
5150–5153.
Westerholt, K., Bergmann, A., Grabis, J., Nefedov, A. and Zabel, H. (2005).
Half-metallic alloys — fundamentals and applications, p. 67.
Williams, A., Kübler, J. and Gelatt, C. (1979). Cohesive properties of metallic
Bibliography 265
compounds: Augmented-spherical-wave calculations, Phys. Rev. B 19, p.

6094.
Wojcik, M., Van Roy, W., Jedryka, E., Nadolski, S., Borghs, G. and De Boeck,
J. (2002). NMR evidence for MnSb environments within epitaxial NiMnSb
films grown on GaAs(0 0 1), Journal of Magnetism and Magnetic Materials
240, pp. 414–416.
Wolf, S. A., Awschalom, D. D., Buhrman, R. A., Daughton, J. M., Molnár, S.,
Roukes, M., Chtchelkanova, A. and Treger, D. (2001). Spintronics: A spin-
based electronics vision for the future, Science 294, pp. 1488–1495.
Wright, J. P., Attfield, J. P. and Radaelli, P. G. (2001). Long range charge or-
dering in magnetite below the Verwey transition, Phys. Rev. Lett. 87, p.
266401.
Wright, J. P., Attfield, J. P. and Radaelli, P. G. (2002). Charge ordered struc-
ture of magnetite Fe3 O4 below the Verwey transition, Phys. Rev. B 66, p.
214422.
Wu, H., Kratzer, P. and Scheffler, M. (2005). First-principles study of thin mag-
netic transition-metal silicide films on Si(001), Phys. Rev. B 72, p. 144425.
Wu, H., Kratzer, P. and Scheffler, M. (2007). Density-functional theory study of
half-metallic heterostructures: Interstitial Mn in Si, Phys. Rev. Lett. 98, p.
117202.
Wurmehl, S., Fecher, G., Kandpal, H., Ksenofontov, V., Felser, C. and Lin, H.-
J. (2006a). Investigation of Co2 FeSi: The Heusler compound with highest
Curie temperature and magnetic moment, Appl. Phys. Lett. 88, pp. 032503–
3.
Wurmehl, S., Fecher, G., Kandpal, H., Ksenofontov, V., Felser, C., Lin, H.-J. and
Morais, J. (2005). Geometric, electronic, and magnetic structure of Co2 FeSi
: Curie temperature and magnetic moment measurements and calculations,
Phys. Rev. B 72, p. 184434.
Wurmehl, S., Fecher, G. H., Kroth, K., Kronast, F., Dürr, H. A., Takeda,
Y., Saitoh, Y., Kobayashi, K., Lin, H.-J., Schönhense, G. and Felser, C.
(2006b). Electronic structure and spectroscopy of the quaternary Heusler
alloy Co2 Cr1x Fex Al, J. Phys. D: Appl. Phys. 39, pp. 803–815.
Wurmehl, S., Kohlhepp, J. T., Swagten, H. J. M., Koopmans, B., Wojcik, M.,
Balke, B., Blum, C. G. F., Ksenofontov, V., Fecher, G. H. and Felser, C.
(2007). Probing the random distribution of half-metallic Co2 Mn1−x Fex Si
Heusler alloys, Appl. Phys. Lett. 91, pp. 052506–3.
Wyckoff, R. W. G. (1963). Crystal Structures, 2nd edition, Vol. 1, p. 110.
Xie, W.-H., Xu, Y.-Q., Liu, B.-G. and Pettifor, D. (2003). Half-metallic ferromag-
netism and structural stability of zincblende phases of the transition-metal
chalcogenides, Phys. Rev. Lett. 91, p. 037204.
Xu, Y.-Q., Liu, B.-G. and Pettifor, D. G. (2002). Half-metallic ferromagnetism
of MnBi in the zinc-blende structure, Phys. Rev. B 66, p. 184435.
Yablonskikh, M., Grebennikov, V., Yarmoshenko, Y., Kurmaev, E., Butorin,
S., Duda, L., Sthe, C., Käämbre, T., Magnuson, M. and Nordgren, J.
(2000). Magnetic circular dichroism in X-ray fluorescence of Heusler alloys
at threshold excitation, Solid State Communications 117, pp. 79–82.
Yablonskikh, M., Yarmoshenko, Y., Grebennikov, V., Kurmaev, E., Butorin, S.,
Duda, L.-C., Nordgren, J., Plogmann, S. and Neumann, M. (2001). Origin
of magnetic circular dichroism in soft x-ray fluorescence of Heusler alloys
at threshold excitation, Phys. Rev. B 63, p. 235117.
Yamasaki, A., Imada, S., Suga, S., Kanomata, T. and Ishida, S. (2002). Magnetic
circular dichroism in the soft x-ray absorption spectra of Co-based Heusler
alloys, Surface Review and Letters 9, pp. 955–960.
Yamashita, J. and Kondo, J. (1958). Superexchange interaction, Phys. Rev. 109,
pp. 730–741.
Yanase, A. and Siratori, K. (1984). Band structure in the high temperature phase
of Fe3 O4 , J. Phys. Soc. Jpn. 53, pp. 312–317.
Yao, Y., Kleinman, L., MacDonald, A., Sinova, J., Jungwirth, T., sheng Wang,
D., Wang, E. and Niu, Q. (2004). First principles calculation of anomalous
Hall conductivity in ferromagnetic bcc Fe, Phys. Rev. Lett. 92, p. 037204.
Ye, J., Kim, Y., Millis, A., Shraiman, B., Majumdar, P. and Tescaronanovicacute,
Z. (1999). Berry pphase theory of the anomalous hall effect: Application to
colossal magnetoresistance manganites, Phys. Rev. Lett. 83, p. 3737.
Yoneda, Y. (1963). Anomalous surface reflection of x rays, Phys. Rev. 131, p.
2010.
Yoshimura, K., Miyazaki, A., Vijayaraghavan, R. and Nakamura, Y. (1985). Hy-
perfine field of the Co2 YZ Heusler alloy (Y = V, Cr, Mn and Fe; Z = Al
and Ga), Journal of Magnetism and Magnetic Materials 53, pp. 189–198.
Youn, S. and Min, B. (1995). Effects of the spin-orbit interaction in Heusler com-
pounds: Electronic structures and Fermi surfaces of NiMnSb and PtMnSb,
Phys. Rev. B 51, p. 10436.
Young, R. A., Sakthivel, A., Moss, T. S. and Paiva-Santos, C. O. (1995). DBWS –
9411 – an upgrade of the DBWS *.* programs for Rietveld refinement with
PC and mainframe computers, J. Appl. Cryst. 28, pp. 366–367.
Zeller, R. (2006). Computational nanoscience: do it yourself! p. 419.
Zener, C. (1951). Interaction between the d–shells in the transition metals. II.
ferromagnetic compounds of manganese with perovskite structure, Phys.
Rev. 82, pp. 403–405.
Zhang, F. M., Liu, X. C., Gao, J., Wu, X. S., Du, Y. W., Zhu, H., Xiao, J. Q. and
Chen, P. (2004). Investigation on the magnetic and electrical properties of
crystalline Mn0.05 Si0.95 films, Appl. Phys. Lett. 85, pp. 786–788.
Zhang, W., Jiko, N., Mibu, K. and Yoshimura, K. (2005). Effect of substitution
of Mn with Fe or Cr in Heusler alloy of Co2 MnSn, J. Phys.: Condens. Mat.
17, p. 6653.
Zhang, Z. and Satpathy, S. (1991). Electron states, magnetism, and the Verwey
transition in magnetite, Phys. Rev. B 44, pp. 13319–13331.
Zhao, J. H., Matsukura, F., Takamura, K., Abe, E., Chiba, D. and Ohno, H.
(2001). Room-temperature ferromagnetism in zincblende CrSb grown by
molecular-beam epitaxy, Appl. Phys. Lett. 79, pp. 2776–2778.
Zhao, Y.-J., Geng, W. T., Freeman, A. J. and Delley, B. (2002). Structural,
electronic, and magnetic properties of α– and β–MnAs: LDA and GGA
investigations, Phys. Rev. B 65, p. 113202.
Bibliography 267
Zhu, W., Sinkovic, B., Vescovo, E., Tanaka, C. and Moodera, J. S. (2001). Spin-
resolved density of states at the surface of NiMnSb, Phys. Rev. B 64, p.
060403.
Zhu, W., Zhang, Z. and Kaxiras, E. (2008). Dopant-assisted concentration en-
hancement of substitutional Mn in Si and Ge, Phys. Rev. Lett. 100, p.
027205.
Ziebeck, K. R. A. and Webster, P. J. (1974). A neutron diffraction and magneti-
zation study of Heusler alloys containing Co and Zr, Hf, V or Nb, Journal
of Physics and Chemistry of Solids 35, pp. 1–7.
Ziese, M. and Blythe, H. J. (2000). Magnetoresistance of magnetite, J. Phys.:
Condens. Mat. 12, p. 13.

Index
Alternative gradient magnetometer Coercivity, 26, 140, 200

(AGM), 28 Coulomb interaction, 4, 39, 40, 53,
Andreev reflection, 37, 46, 47, 49–51, 63, 64, 107, 168, 172, 211
89, 90, 140 Coulomb scattering, 39
Angular resolved photoemission CrAs, 191–199, 215
spectroscopy (ARPES), 141 CrO2 , 131, 140
Anisotropic magnetoresistance CrSb, 193
(AMR), 157, 243
Crystal field, 6, 7, 9, 146, 166, 186,
Anomalous Hall effect (AHE), 35,
209–211, 213, 233
229, 230
Curie temperature (TC ), 24, 51, 65,
Atomic force microscopy (AFM), 19,
73, 74, 92, 127, 143, 176, 200, 228,
135–137
229
Auger electron spectroscopy (AES),
17, 23, 24, 81, 82, 105
dc magnetron sputtering (MS)
Auger transitions, 94
method, 105
Augmented plane wave (APW), 59
Density functional theory (DFT), 13,
Ballistic conductance, 217, 218 52, 53, 55, 57, 63–65, 145, 181, 222,
Boltzmann constant, 67, 184 234, 238
Bose-Einstein (B-E) distribution Density of states (DOS), 3, 5, 29, 45,
function, 67 46, 52, 73, 114, 145, 146, 204, 205,
Brillouin zone (BZ), 60, 61, 102, 146, 213, 216, 220, 221, 234, 235, 239
147, 165, 205, 206, 218, 232, 236, Diffuse X-ray reflection, 20, 21
237 Digital ferromagnetic
heterostructures (DFH), 192, 202,
C1b structure, 87, 97, 98 224–230
Carbon nanowire, 192 Double exchange, 148, 181, 183–188
Chemical vapor deposition (CVD), Dynamic random access memory
132–134, 144 (DRAM), 1
Co2 MnSi, 11, 19, 22, 24, 27, 28, 30,
80, 88, 90, 92, 99, 100, 104–117, Energy dispersive X-ray analysis
120, 122 (EDX), 106
269
Energy distribution curve (EDC), 44, Junction magnetoresistance (JMR),

142, 143, 156 159
Exchange-correlation energy, 53–57
Kerr rotation, 26, 108
Fe3 O4 , 131, 148–153, 155–163, Knudsen cells, 195
165–169, 184, 187, 188 Kohn-Sham equations, 54
Fermi surface, 146 Korringa-Kohn-Rostoker (KKR)
Ferromagnet-superconductor method, 59, 61, 85, 87, 89, 97, 203
tunneling, 45
First-principles norm-conserving La2/3 Sr1/3 MnO3 (LSMO), 131
pseudopotential, 63
Laser ablation, 152
FM-superconductor tunnel junction,
Linearized augmented plane wave
37
(LAPW) method, 58, 59
Frank-Van der Merwe growth mode,
Local density approximation (LDA),
18
52, 55
Full-Heusler alloy, 71
Local spin density approximation
GaAs, 198 (LSDA), 56
Generalized gradient approximation Low energy electron diffraction
(GGA), 52 (LEED), 17, 81–83, 118, 126, 163
Giant magnetoresistance (GMR), 1,
160 Magnetic force theorem, 65
Gobel mirror, 19 Magnetic tunnel junctions (MTJs),
Grazing incidence X-ray diffraction 31, 105, 112, 114, 116, 117
(GIXD), 196 Magneto-optical anisotropy
parameter, 124
Half-Heusler alloy, 71 Magneto-optical Kerr effect (MOKE),
Hall conductivity, 35 26, 108, 122, 177
Hall effect, 34 Magnetoresistance (MR), 14, 31
Hall measurement, 36 Many-electron system, 53
Heisenberg Hamiltonian, 66 Mean field approximation (MFA), 67
Heusler alloy, 3–6, 8, 71 Mean field theory (MFT), 66
High energy electron diffraction Mean free path (MFP), 137
(RHEED), 17
MnAs, 215
High-resolution cross sectional
MOKE method, 104
transmission electron microscopy
Molecular beam epitaxy (MBE), 14
(HRTEM), 197, 198
Hohenberg-Kohn theorem, 53 Mott detector, 42
Hund’s rule, 184, 185, 187–189, 210 Mott scattering, 39
Hysteresis loops, 25, 27, 28, 52, 144, Muffin-tin (MT) approximation, 58
176, 200–202, 227–229
NiMnSb, 14
Integrated circuit(IC), 1
One dimensional half metals, 238
Jahn-Teller distortion, 180, 186 Orthogonalized plane wave (OPW),
Jahn-Teller effect, 174 62
Bibliography 271
Pairwise exchange interaction energy, Spin-polarized photoemission

66 spectroscopy (SPPS), 163
Plasma-assisted magnetron Spin-resolved photoemission
sputtering method, 105 spectroscopy (SRPES), 125
Point-contact Andreev reflection Spin-wave excitation, 66
(PCAR), 111, 140, 224 Spinel structure, 148, 188
Powder magnetoresistance (PMR), Spiral configuration, 68
137, 139 Stoner excitation, 66
Projector augmented wave (PAW), 63 Stoner model, 166, 167, 169
Pseudopotential method, 62 Stranski-Krastanov, 153
Pulsed laser deposition (PLD), 76, Structure factor, 61
79, 80, 105, 133, 151, 175, 177
Superconducting quantum
interference device (SQUID), 25, 52
Quantum dot, 218, 222
Superexchange, 144, 181–183, 188,
189
Radio frequency magnetron
sputtering method, 75, 76, 78
Thermopiezic analysis (TPA), 133
Random phase approximation (RPA),
67 Total electron yield (TEY), 119, 120
Reflection high energy electron Transmission electron microscopy
diffraction (RHEED), 16, 17, 83, (TEM), 198
105, 135, 152, 153, 155, 193, 195, Tri-arc Czochralski method, 75, 76,
197, 198, 219, 222, 224, 225 99
Resonance X-ray diffraction, 162 Tunnel magnetoresistance (TMR),
31, 112–115, 117, 130, 131, 178
Saturation magnetic moment, 144
Saturation magnetization, 25, 140 Verwey transition, 148, 157, 161, 166,
Scanning tunneling microscopy 171, 187
(STM), 17, 21 Vibrating sample magnetometer
Sherman function, 43 (VSM), 28
Slater-Pauling rule, 83, 87
Specular X-ray reflection, 20 Wigner-Seitz (WS) cell, 60
Spin echo nuclear magnetic
resonance, 96 X-ray absorption spectra (XAS), 106,
Spin filter, 218 107, 119–122, 127, 163
Spin polarization, 100 X-ray absorption spectroscopy
Spin resolved appearance potential (XAS), 195, 198
spectroscopy (SRAPS), 94 X-ray diffraction (HRXRD), 17
Spin valve, 131, 156, 160 X-ray magnetic circular dichroism
Spin-orbit (S-O) interaction, 39, 40 (XMCD), 28–30, 92, 93, 95,
Spin-polarized angular resolved 106–108, 119–121, 127
photoemission spectroscopy X-ray photoelectron spectroscopy, 19
(ARPES), 37, 142 X-ray reflectometry (XRR), 17, 19
Spin-polarized photoemission (SPP),
37 Yoneda wings, 20
Zincblende (ZB) structure, 3–5, 8, 10,

13–16, 191, 192

Half Metals and Their Properties

Uploaded by

Copyright:

Available Formats

Half Metals and Their Properties

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Half Metals and Their Properties

Uploaded by

Copyright:

Available Formats

LLNL-BOOK-424888

Half metallic materials and their

C.Y. Fong, J.E. Pask, L.H. Yang

Half metallic materials and their properties

vi Half metallic materials and their properties

Since the first theoretical prediction of a half metallic material—one in

viii Half metallic materials and their properties

In the introductory chapter, we give an overview of the main features,

x Half metallic materials and their properties

2. Methods of studying half metals 13

xii Half metallic materials and their properties

2.5.4 Methods of calculating Curie temperature, TC . . 65

4. Half metallic oxides 131

4.3.2 Growth . . . . . . . . . . . . . . . . . . . . . . . . 150

5. Half metals with simple structures 191

Appendix A Appendix 243

The primary ingredient of computing hardware is the integrated circuit

2 Half metallic materials and their properties

1992; Savtchenko et al., 2003; Bussmann et al., 1999), and in spin-current

Fig. 1.1 Developing next-generation spin-based technologies and applications

Normally, a nonferromagnetic material, such as silicon (Si) for example,

ing to zero net spin polarization. In a ferromagnetic metal, such as iron

4 Half metallic materials and their properties

1.2 Classes of half metals

Up to now, three main classes of ferromagnetic HMs have been found

Fig. 1.3 Crystal structure of CrO2

6 Half metallic materials and their properties

tetrahedral environment, the order of the two sets is reversed. Depending

8 Half metallic materials and their properties

be a valence IV, V, or VI element. Its electronegativity is in general weaker

10 Half metallic materials and their properties

Fig. 1.10 Schematic diagram of the Fig. 1.11 Schematic DOS of a HM

1.3 Half metallic devices

Since HMs offer the possibility of unprecedented magnetoresistances (infi-

12 Half metallic materials and their properties

Methods of studying half metals

To realize spintronic devices based on half metals (HMs), the growth of

14 Half metallic materials and their properties

2.2 Molecular beam epitaxy (MBE)

2.2.1 Schematic setup

The beam sources are indicated by open rectangles connected to the

Methods of studying half metals 15

substrate. In general, characterizations of the sample are made in situ,

2.2.2 Issues concerning growth

2.2.2.1 Substrate for Heusler alloys

16 Half metallic materials and their properties

2.2.2.2 Substrate for HMs with ZB structure

Methods of studying half metals 17

2.3 Characterization of samples

Two popular methods to characterize the quality of HMs grown in thin-film

2.3.1 Reflection high energy electron diffraction (RHEED)

2.3.1.2 What is measured

18 Half metallic materials and their properties

Oscillatory behavior of spot intensity at the center (00) This os-

Methods of studying half metals 19

An example of intensity as a function of time for growing Co2 MnSi is given

2.3.2 X-ray reflectometry (XRR)

2.3.2.1 Basic information

2.3.2.2 Schematic setup